Вы находитесь на странице: 1из 1131

Handbook of Polymer

Reaction Engineering

Edited by
Th. Meyer, J. Keurentjes

Handbook of Polymer Reaction Engineering. Edited by T. Meyer, J. Keurentjes

Copyright 8 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
ISBN: 3-527-31014-2
Further Titles of Interest

E. S. Wilks (Ed.)

Industrial Polymers Handbook

Products, Processes, Applications

ISBN 3-527-30260-3

H.-G. Elias (Ed.)

Vols. 1–4

ISBN 3-527-31172-6, 3-527-31173-4, 3-527-31174-2,

M. F. Kemmere, Th. Meyer (Eds.)

Supercritical Carbon Dioxide

in Polymer Reaction Engineering

ISBN 3-527-31092-4

M. Xanthos (Ed.)

Functional Fillers for Plastics

ISBN 3-527-31054-1

R. C. Advincula, W. J. Brittain, K. C. Caster, J. Ru̇

u̇he (Eds.)

Polymer Brushes
ISBN 3-527-31033-9

H.-G. Elias (Ed.)

An Introduction to Plastics
Second, Completely Revised Edition
ISBN 3-527-29602-6
Handbook of Polymer Reaction Engineering

Edited by
Thierry Meyer, Jos Keurentjes
Editors 9 All books published by Wiley-VCH are
carefully produced. Nevertheless, authors,
Dr. Thierry Meyer editors, and publisher do not warrant the
Swiss Federal Institute of Technology information contained in these books,
Institute of Process Science including this book, to be free of errors.
EPFL, ISP-GPM Readers are advised to keep in mind that
1015 Lausanne statements, data, illustrations, procedural
Switzerland details or other items may inadvertently be
Prof. Jos T. F. Keurentjes
Process Development Group Library of Congress Card No.: Applied for
Eindhoven University of Technology
P.O. Box 513 British Library Cataloging-in-Publication
5600 MB Eindhoven Data: A catalogue record for this book is
The Netherlands available from the British Library

Bibliographic information published by

Die Deutsche Bibliothek
Die Deutsche Bibliothek lists this publica-
tion in the Deutsche Nationalbibliografie;
detailed bibliographic data is available in
the Internet at hhttp://dnb.ddb.dei.

8 2005 WILEY-VCH Verlag GmbH & Co.

KGaA, Weinheim
All rights reserved (including those of
translation into other languages). No part of
this book may be reproduced in any form –
nor transmitted or translated into 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.

Printed in the Federal Republic of Germany

Printed on acid-free paper

Composition Asco Typesetters, Hong Kong

Printing betz-druck gmbh, Darmstadt
Bookbinding Litges & Dopf Buchbinderei
GmbH, Heppenheim

ISBN-13 978-3-527-31014-2
ISBN-10 3-527-31014-2


A principal difference between science and engineering is intent. Science is used

to bring understanding and order to a specific object of study – to build a body of
knowledge with truth and observable laws. Engineering is more applied and prac-
tical, focused on using and exploiting scientific understanding and scientific prin-
ciples to make products to benefit mankind. A polymer reaction engineer seeks the
applied and practical as the title implies, but the path to success is most often
through polymer science. This truth is steeped in history – there are many exam-
ples of polymeric products commercialized without adequate understanding of the
chemistry and physics of the underlying polymerization. Polymer reaction engi-
neers, faced with detriments in process safety, product quality or product cost, be-
come the driving force behind many polymer science developments. As such, poly-
mer reaction engineering is more a collaboration of polymer science and reaction
engineering. A collaboration where polymer reaction engineers develop a firm un-
derstanding of the many aspects of polymer chemistry and physics to successfully
apply chemical engineering principles to new product developments. Only through
the integration of science and engineering are such products realized.
This handbook is a testimony to this melding of polymer science and chemical
engineering that defines polymer reaction engineering. Thierry Meyer and Jos
Keurentjes have compiled a strong list of contributors with an even balance from
academia and industry. The text offers a comprehensive view of polymer reaction
engineering. The text starts with an overview describing the important integration
of science and engineering in polymer reaction engineering and ends with recent
and potential breakthrough developments in polymer processing. The middle
chapters are divided into three sections. The first section is devoted to the science
and chemistry of the major types of polymerization. Included are step and chain
growth polymerizations with chapters devoted specifically to several different chain
growth methods. The central section of the middle chapters is dedicated to poly-
mer reaction engineering tools and methods. The very important topics of safety
and process control are detailed and help frame the conditions through which suc-
cessful scale-ups are achieved. The last section of the middle chapters is focused on
the physics and physical nature of formed polymers including their physical and
mechanical structure. In these chapters, many of the processes that modify poly-
VI Foreword

mers through man-made and natural change are discussed. The details of polymer
end use are also presented.
This tome represents the first published handbook on polymer reaction engi-
neering and should be well received in academia and industry. Polymer reaction
engineering became recognized as a separate engineering discipline in the 1970’s.
It is well past due that such a handbook be published. The broad scope and depth
of coverage should make it an important reference for years to come.

Michael C. Grady, ScD

Senior Engineering Associate
Philadelphia, Pennsylvania


Volume 1

Foreword V

Preface XXIX

List of Contributors XXXI

1 Polymer Reaction Engineering, an Integrated Approach 1

Th. Meyer and J. T. F. Keurentjes
1.1 Polymer Materials 1
1.2 A Short History of Polymer Reaction Engineering 4
1.3 The Position of Polymer Reaction Engineering 5
1.4 Toward Integrated Polymer Reaction Engineering 7
1.5 The Disciplines in Polymer Reaction Engineering 9
1.5.1 Polymerization Mechanisms 11
1.5.2 Fundamental and Engineering Sciences 12
1.6 The Future: Product-inspired Polymer Reaction Engineering 14
1.7 Concluding Remarks 15
References 15

2 Polymer Thermodynamics 17
Theodoor W. de Loos
2.1 Introduction 17
2.2 Thermodynamics and Phase Behavior of Polymer Solutions 18
2.2.1 Thermodynamic Principles of Phase Equilibria 18
2.2.2 Fugacity and Activity 18
2.2.3 Equilibrium Conditions 20
2.2.4 Low-pressure Vapor–Liquid Equilibria 21
2.2.5 Flory–Huggins Theory and Liquid–Liquid Equilibria 21
2.2.6 High-pressure Liquid–Liquid and Vapor–Liquid Equilibria 25
2.3 Activity Coefficient Models 29
2.3.1 Flory–Huggins Theory 30
VIII Contents

2.3.2 Hansen Solubility Parameters 32

2.3.3 Correlation of Solvent Activities 34
2.3.4 Group Contribution Models 35
2.4 Equation of State Models 39
2.4.1 The Sanchez–Lacombe Lattice Fluid Theory 40
2.4.2 Statistical Associating-fluid Theory 44 SAFT and PC-SAFT Hard Chain Term 44 SAFT Dispersion Term 45 The PC-SAFT Dispersion Term 46 SAFT and PC-SAFT Applications 47 Extension to Copolymers 48
2.5 Conclusions 50
Notation 52
References 54

3 Polycondensation 57
Mário Rui P. F. N. Costa and Rolf Bachmann
3.1 Basic Concepts 57
3.1.1 A Brief History 57
3.1.2 Molecular Weight Growth and Carothers’ Equation 59
3.1.3 The Principle of Equal Reactivity and the Prediction of the Evolution of
Functional Group Concentrations 62
3.1.4 Effect of Reaction Media on Equilibrium and Rate Parameters 62
3.1.5 Polycondensation Reactions with Substitution Effects 64
3.1.6 Exchange Reactions 66
3.1.7 Ring-forming Reactions 67
3.1.8 Modeling of Polymerization Schemes 68
3.2 Mass Transfer Issues in Polycondensations 69
3.2.1 Removal of Volatile By-products 69
3.2.2 Solid-state Polycondensation 80
3.2.3 Interfacial Polycondensation 82
3.3 Polycondensation Processes in Detail 85
3.3.1 Polyesters 85 Structure and Production Processes 85 Acid-catalyzed Esterification and Alcoholysis 86 Catalysis by Metallic Compounds 87 Side Reactions in Aromatic Polyester Production 89 Side Reactions in the Formation of Unsaturated Polyesters 90 Modeling of Processes of Aromatic Polyester Production 91 Modeling of Processes for Unsaturated Polyester Production 92
3.3.2 Polycarbonates 93 General Introduction 93 Interfacial Polycondensation 94 Melt Transesterification 96
3.3.3 Polyamides 98
Contents IX Introduction 98 Kinetic Modeling 98 Nonoxidative Thermal Degradation Reactions 100 Process Modeling 101
3.3.4 Polymerizations with Formaldehyde: Amino Resins (Urea and
Melamine) and Phenolics 102 Formaldehyde Solutions in Water 102 Amino Resins 102 Phenolic Resins 107
3.3.5 Epoxy Resins 108
3.3.6 Polyurethanes and Polyureas 109
3.4 Modeling of Complex Polycondensation Reactions 113
3.4.1 Overview 113
3.4.2 Description of Reactions in Polycondensations of Several Monomers
with Substitution Effects 113
3.4.3 Equilibrium Polycondensations with Several Monomers 117
3.4.4 Kinetic Modeling of Irreversible Polycondensations 129
3.4.5 Kinetic Modeling of Linear Reversible Polycondensations 133
Notation 136
References 144

4 Free-radical Polymerization: Homogeneous 153

Robin A. Hutchinson
4.1 FRP Properties and Applications 153
4.2 Chain Initiation 154
4.3 Polymerization Mechanisms and Kinetics 156
4.3.1 Homopolymerization 157 Basic Mechanisms 157 Kinetic Coefficients 161 Additional Mechanisms 169
4.3.2 Copolymerization 179 Basic Mechanisms 179 Kinetic Coefficients 183 Additional Mechanisms 187
4.3.3 Diffusion-controlled Reactions 190
4.4 Polymer Reaction Engineering 193
4.4.1 Types of Industrial Reactors 195 Batch Processes 195 Semi-batch Processes 196 Continuous Processes 196
4.4.2 Mathematical Modeling of FRP Kinetics 197 Method of Moments 198 Modeling of Distributions 201
4.4.3 Reactor Modeling 203 Batch Polymerization 204
X Contents Continuous Polymerization 204 Complex Flowsheets 205 Computational Fluid Dynamics (CFD) 205 Model-based Control 206
4.5 Summary 206
Notation 206
References 209

5 Free-radical Polymerization: Suspension 213

B. W. Brooks
5.1 Key Features of Suspension Polymerization 213
5.1.1 Basic Ideas 213
5.1.2 Essential Process Components 214
5.1.3 Polymerization Kinetics 214
5.1.4 Fluid–Fluid Dispersions and Reactor Type 215
5.1.5 Uses of Products from Suspension Polymerization 216
5.2 Stability and Size Control of Drops 216
5.2.1 Stabilizer Types 217
5.2.2 Drop Breakage Mechanisms 218
5.2.3 Drop Coalescence 222
5.2.4 Drop Size Distributions 223
5.2.5 Drop Mixing 224
5.3 Events at High Monomer Conversion 228
5.3.1 Breakage of Highly Viscous Drops 229
5.3.2 Polymerization Kinetics in Viscous Drops 229
5.3.3 Consequences of Polymer Precipitation Inside Drops 230
5.4 Reaction Engineering for Suspension Polymerization 234
5.4.1 Dispersion Maintenance and Reactor Choice 234
5.4.2 Agitation and Heat Transfer in Suspensions 235
5.4.3 Scaleup Limitations with Suspension Polymerization 237
5.4.4 Reactor Safety with Suspension Processes 238
5.4.5 Component Addition during Polymerization 238
5.5 ‘‘Inverse’’ Suspension Polymerization 239
5.5.1 Initiator Types 239
5.5.2 Drop Mixing with Redox Initiators 240
5.6 Future Developments 240
5.6.1 Developing Startup Procedures for Batch and Semi-batch Reactors 240
5.6.2 Maintaining Turbulence Uniformity in Batch Reactors 242
5.6.3 Developing Viable Continuous-flow Processes 242
5.6.4 Quantitative Allowance for the Effects of Changes in the Properties of
the Continuous Phase 242
5.6.5 Further Study of the Role of Secondary Suspending Agents 243
5.6.6 Further Characterization of Stabilizers from Inorganic Powders 243
Notation 243
References 244
Contents XI

6 Emulsion Polymerization 249

José C. de la Cal, José R. Leiza, Jose M. Asua, Alessandro Buttè, Guiseppe Storti,
and Massimo Morbidelli
6.1 Introduction 249
6.2 Features of Emulsion Polymerization 250
6.2.1 Description of the Process 250
6.2.2 Radical Compartmentalization 254
6.2.3 Advantages of Emulsion Polymerization 254
6.3 Alternative Polymerization Techniques 256
6.4 Kinetics of Emulsion Polymerization 258
6.4.1 Monomer Partitioning 259
6.4.2 Average Number of Radicals per Particle 260
6.4.3 Number of Polymer Particles 264 Heterogeneous Nucleation 264 Homogeneous Nucleation 266 Simultaneous Heterogeneous and Homogeneous Nucleation 267 Coagulative Nucleation 267
6.5 Molecular Weights 267
6.5.1 Linear Polymers 268 Zero–One System (Smith–Ewart Cases 1 and 2) 268 Pseudo Bulk System (Smith–Ewart Case 3) 270
6.5.2 Nonlinear Polymers: Branching, Crosslinking, and Gel Formation
6.6 Particle Morphology 273
6.7 Living Polymerization in Emulsion 275
6.7.1 Chemistry of LRP 275 Nitroxide-mediated Polymerization (NMP) 277 Atom-transfer Radical Polymerization (ATRP) 277 Degenerative Transfer (DT) 278 Reversible Addition–Fragmentation Transfer (RAFT) Polymerization
6.7.2 Polymerization of LRP in Homogeneous Systems 280
6.7.3 Kinetics of LRP in Heterogeneous Systems 282
6.7.4 Application of LRP in Heterogeneous Systems 284 Ab-initio Emulsion Polymerization 284 Miniemulsion Polymerization 285
6.8 Emulsion Polymerization Reactors 286
6.8.1 Reactor Types and Processes 286 Stirred-tank Reactors 286 Tubular Reactors 287
6.8.2 Reactor Equipment 288 Mixing 289 Heat Transfer 290
6.9 Reaction Engineering 290
6.9.1 Mass Balances 291
XII Contents

6.9.2 Heat Balance 292

6.9.3 Polymer Particle Population Balance (Particle Size Distribution)
6.9.4 Scaleup 295
6.10 On-line Monitoring in Emulsion Polymerization Reactors 296
6.10.1 On-line Sensor Selection 297 Latex Gas Chromatography 298 Head-space Gas Chromatography 298 Densimetry 298 Ultrasound 299 Spectroscopic Techniques 299 Reaction Calorimetry 302
6.11 Control of Emulsion Polymerization Reactors 305
Notation 312
References 317

7 Ionic Polymerization 323

Klaus-Dieter Hungenberg
7.1 Introduction 323
7.2 Anionic Polymerization 325
7.2.1 Anionic Polymerization of Hydrocarbon Monomers – Living
Polymerization 326 Association Behavior/Kinetics 326 Molecular Weight Distribution of Living Polymers 331 Side Reactions 336 Copolymerization 338 Tailor-made Polymers by Living Polymerization – Optimization 341 Industrial Aspects – Production of Living Polymers 343
7.2.2 Anionic Polymerization of Vinyl Monomers Containing Heteroatoms
7.2.3 Anionic Polymerization of Monomers Containing Hetero Double Bonds
7.2.4 Anionic Polymerization via Ring Opening 346
7.3 Cationic Polymerization 351
7.3.1 Cationic Polymerization of Vinyl Monomers 351
7.3.2 Cationic Ring-opening Polymerization 353
7.4 Conclusion 356
Notation 357
References 359

8 Coordination Polymerization 365

João B. P. Soares and Leonardo C. Simon
8.1 Polyolefin Properties and Applications 365
8.1.1 Introduction 365
8.1.2 Types of Polyolefins and Their Properties 366
Contents XIII

8.1.3 The Importance of Proper Microstructural Determination and Control

in Polyolefins 369
8.2 Catalysts for Olefin Polymerization 372
8.2.1 Ziegler–Natta, Phillips, and Vanadium Catalysts 378
8.2.2 Metallocene Catalysts 379
8.2.3 Late Transition Metal Catalysts 381
8.3 Polymerization Kinetics and Mechanism with Coordination Catalysts
8.3.1 Comparison between Coordination and Free-radical Polymerization
8.3.2 Polymerization Kinetics with Single-site Catalysts 383 Homopolymerization 383 Copolymerization 388
8.3.3 Polymerization Kinetics with Multiple-site Catalysts 392
8.3.4 Long-chain Branch Formation 395
8.4 Single Particle Models – Mass- and Heat-transfer Resistances 399
8.5 Macroscopic Reactor Modeling – Population Balances and the Method
of Moments 408
8.5.1 Homopolymerization 408
8.5.2 Copolymerization 413
8.6 Types of Industrial Reactors 416
8.6.1 Gas-phase Reactors 420
8.6.2 Slurry Reactors 422
8.6.3 Solution Reactors 423
8.6.4 Multizone Reactors 425
Notation 425
References 428

9 Mathematical Methods 431

P. D. Iedema and N. H. Kolhapure
9.1 Introduction 431
9.2 Discrete Galerkin h–p Finite Element Method 432
9.3 Method of Moments 435
9.3.1 Introduction 435
9.3.2 Linear Polymerization 435
9.3.3 Nonlinear Polymerization 438
9.4 Comparison of Galerkin-FEM and Method of Moments 440
9.5 Classes Approach 444
9.5.1 Introduction 444
9.5.2 Computing the CLD of Poly(vinyl acetate) for a Maximum of One TDB
per Chain 444
9.5.3 Multiradicals in Radical Polymerization 446
9.6 Pseudo-distribution Approach 449
9.6.1 Introduction 449
9.6.2 CLD/DBD for Mixed-metallocene Polymerization of Ethylene 451
XIV Contents Formulation of Pseudo-distribution Problem 451 Construction of the Full 2D Distribution 456
9.6.3 CLD/Number of Terminal Double Bonds (TDB) Distribution for
Poly(vinyl acetate) – More than one TDB per Chain 458 General Case 458 TDB Pseudo-distribution Approach for a Maximum of one TDB per
Chain 466 TDB Pseudo-distribution Approach for More than one TDB per Chain
9.6.4 Radical Polymerization of Ethylene to Low-density Polyethylene (LDPE)
469 Introduction 469
9.6.5 Radical Copolymerization 473 Introduction 473 Balance Equations 474
9.7 Probability Generating Functions 480
9.7.1 Introduction 480
9.7.2 Probability Generating Functions in a Transformation Method 480
9.7.3 Probability Generating Functions and Cascade Theory 481
9.8 Monte Carlo Simulations 485
9.8.1 Introduction 485
9.8.2 Weight-fraction Sampling of Primary Polymers: Batch Reactor,
Transfer to Polymer 486
9.8.3 Example 490
9.8.4 CSTR with Transfer to Polymer 491
9.8.5 Comparison of Galerkin-FEM Classes Model and CSTR with Transfer
to Polymer 492
9.8.6 Batch Reactor, Terminal Double Bond Incorporation 493
9.8.7 CSTR, Terminal Double Bond Incorporation 497
9.8.8 Incorporation of Recombination Termination 498
9.8.9 Incorporation of Random Scission, Linear Chains, Batch Reactor 498
9.8.10 Combined Scission/Branching 501
9.8.11 Scission in a CSTR 501
9.9 Prediction of Branched Architectures by Conditional Monte Carlo
Sampling 502
9.9.1 Introduction 502
9.9.2 Branched Architectures from Radical Polymerization in a CSTR 503
9.9.3 Branched Architectures from Polymerization of Olefins with Single
and Mixed Branch-forming Metallocene Catalysts in a CSTR 505 Introduction 505 Single-catalyst System 505 Synthesis of Topology 505 Mixed-catalyst System 508
9.9.4 Mathematical Methods for Characterization of Branched Architectures
510 Graph Theoretical Connectivity Matrices 510
Contents XV Characterization of Architectures by Radius of Gyration 511 Characterization of Architectures by Seniorities and Priorities 512
9.10 Computational Fluid Dynamics for Polymerization Reactors 517
9.10.1 Introduction 517 Modeling Challenges 517
9.10.2 Development and Optimization of Modern Polymerization Reactors
518 Benefits of CFD 519 Limitations of CFD 519
9.10.3 Integration of CFD with Polymerization Kinetics 520 Classification and Complexity of CFD Models 521 Treatment of Polymerization Kinetics 522 Illustration of Homogeneous Reactor Model Formulation 522
9.10.4 Target Applications 523 Illustrative Case Studies 523
9.10.5 Concluding Remarks 528
Acknowledgments 530
References 530

10 Scaleup of Polymerization Processes 533

E. Bruce Nauman
10.1 Historic and Economic Perspective 533
10.2 The Limits of Scale 533
10.3 Scaleup Goals 534
10.4 General Approaches 535
10.5 Scaleup Factors 537
10.6 Stirred-tank Reactors 537
10.7 Design Considerations for Stirred Tanks 541
10.8 Multiphase Stirred Tanks 542
10.9 Stirred Tanks in Series 542
10.10 Tubular Reactors 543
10.11 Static Mixers 545
10.12 Design Considerations for Tubular Reactors 546
10.13 Extruder and Extruder-like Reactors 549
10.14 Casting Systems 549
10.15 Concluding Remarks 550
Notation 550
References 551

Volume 2

11 Safety of Polymerization Processes 553

Francis Stoessel
11.1 Introduction 553
11.2 Principles of Chemical Reactor Safety Applied to Polymerization 554
11.2.1 Cooling Failure Scenario 554
XVI Contents

11.2.2 Criticality Classes Applied to Polymerization Reactors 557 Description of the Criticality Classes 558
11.2.3 Heat Balance of Reactors 559 Heat Production 559 Heat Exchange 560 Heat Accumulation 561 Convective Heat Transport due to Feed 561 Stirrer 561 Heat Losses 562 Simplified Expression of the Heat Balance 562
11.2.4 Dynamic Control of Reactors 562
11.2.5 Thermal Stability of Polymerization Reaction Masses 563
11.3 Specific Safety Aspects of Polymerization Reactions 564
11.3.1 Kinetic Aspects 564
11.3.2 Thermochemical Aspects 565
11.3.3 Factors Leading to Changing Heat Release Rates 568
11.4 Cooling of Polymerization Reactors 570
11.4.1 Indirect Cooling: Heat Exchange Across the Reactor Wall 570
11.4.2 Hot Cooling: Cooling by Evaporation 574
11.4.3 Importance of the Viscosity 578
11.5 Chemical Engineering for the Safety of Polymerization Processes
11.5.1 Batch Processes 579
11.5.2 Semi-batch Processes 580 Initiation 581 Feed 582 Final Stage 583 Practical Aspects 583
11.5.3 Continuous Processes 584 Concentration Stability 584 Particle Number Stability 584
11.5.4 Design Measures for Safety 585 Process Design 586 Reactor Design 586 Control of Feed 587 Emergency Cooling 587 Inhibition 588 Quenching 588 Dumping 588 Controlled Depressurization 588 Pressure Relief 588 Time Factor 589
11.6 Conclusion 589
References 590
Notation 591
Contents XVII

12 Measurement and Control of Polymerization Reactors 595

John R. Richards and John P. Congalidis
12.1 Introduction 595
12.1.1 Definitions 595
12.1.2 Measurement Error 597
12.2 Measurement Techniques 598
12.2.1 Temperature 599 Resistance Thermometers 599 Thermocouples 600 Expansion Thermometers 601 Radiation Pyrometers 601
12.2.2 Pressure Measurement 602
12.2.3 Weight 604
12.2.4 Liquid Level 605
12.2.5 Flow 608
12.2.6 Densitometry, Dilatometery, and Gravimetry 617
12.2.7 Viscosity 619
12.2.8 Composition 620
12.2.9 Surface Tension 622
12.2.10 Molecular Weight Distribution (MWD) 622
12.2.11 Particle Size Distribution (PSD) 623
12.3 Sensor Signal Processing 625
12.3.1 Sensors and Transmitters 625
12.3.2 Converters 626
12.3.3 Indicators 626
12.3.4 Filtering Techniques 627
12.4 Regulatory Control Engineering 627
12.4.1 General 627
12.4.2 Process Dynamics 630 First-order System 631 Second-order System 632 High-order and Dead Time Systems 636 First-order Plus Dead Time System 636 Integrating System 638 Integrator plus Dead Time System 639
12.4.3 Controllers 639 Proportional Control 640 Integral Control 641 Derivative Control 641 PI, PD, and PID Control 642 Digital Controllers 642 Controller Tuning 644 On–Off Controllers 646 Self-operated Regulators 647
12.4.4 Valve Position Controllers 650
XVIII Contents

12.4.5 Single-loop Controllers 650

12.4.6 Digital Control Systems 650
12.4.7 Actuators 652
12.5 Advanced Control Engineering 656
12.5.1 Feedforward Control 659 Steady-state Model Feedforward Control 660 Ratio Control 660
12.5.2 Cascade Control 661
12.5.3 Feedforward–Feedback Control 663
12.5.4 State Estimation Techniques 666
12.5.5 Model Predictive Control 668
12.5.6 Batch and Semi-batch Control 669 Operation and Variability 669 Statistical Process Control 671
12.5.7 Future Trends 671
Notation 672
References 675

13 Polymer Properties through Structure 679

Uday Shankar Agarwal
13.1 Thermal Properties of Polymers 679
13.1.1 Crystalline and Amorphous Polymers 680
13.1.2 Influence of Polymer Structure on Crystallizability of Polymers 682
13.1.3 The Glass Transition Temperature 683
13.1.4 Influence of Polymer Structure on Tg of Polymers 684
13.1.5 The Crystallization Temperature and the Melting Point 686
13.1.6 Tuning Polymer Crystallization for Properties 686
13.1.7 Morphology of Crystalline Polymers 688
13.1.8 Tailoring Polymer Properties through Modification, Additives, and
Reinforcement 690 New Morphologies through Block Copolymers 691 Polymeric Nanocomposites 692
13.2 Polymer Conformation and Related Properties 692
13.2.1 The Chain Conformation 692
13.2.2 Solubility of Polymers 694
13.2.3 Dilute Solution Zero-shear Viscosity 695 Polymers as Dumbbells 696 Polymers as Chains of Beads and Springs 697
13.2.4 Viscosity of Concentrated Solutions and Melts 698
13.2.5 Nonlinear Polymers 699
13.2.6 Rigid Rod-like Polymers 701
13.3 Polymer Rheology 702
13.3.1 The Viscous Response: Shear Thinning 702
13.3.2 Normal Stresses during Shear Flow 703
13.3.3 Extensional Thickening 705
Contents XIX

13.3.4 The Elastic Response 706 Ideal Elastic Response 706 Rubberlike Elasticity 706
13.3.5 The Viscoelastic Response 707 Linear Viscoelasticity in Dynamic Oscillatory Flow 709
13.3.6 Influence of Polymer Branching Architecture in Bulk Polymers 711
13.3.7 Polymers as Rheology Modifiers 712
13.3.8 Rheological Control with Block Copolymers 714
13.3.9 Polymer-like Structures through Noncovalent Associations 715
13.4 Summary 715
Notation 716
References 718

14 Polymer Mechanical Properties 721

Christopher J. G. Plummer
14.1 Introduction 721
14.1.1 Long-chain Molecules 721
14.1.2 Simple Statistical Descriptions of Long-chain Molecules 722
14.2 Elasticity 724
14.2.1 Deformation of an Elastic Solid 724
14.2.2 Thermodynamics of Rubber Elasticity 725
14.2.3 Statistical Mechanical Approach to Rubber Elasticity 727
14.3 Viscoelasticity 729
14.3.1 Linear Viscoelasticity 729
14.3.2 Time–Temperature Superposition 734
14.3.3 Molecular Models for Polymer Dynamics 736
14.3.4 Nonlinear Viscoelasticity 740
14.4 Yield and Fracture 741
14.4.1 Yield in Polymers 741
14.4.2 Models for Yield 744
14.4.3 Semicrystalline Polymers 746
14.4.4 Crazing and Fracture 748
14.5 Conclusion 752
References 755

15 Polymer Degradation and Stabilization 757

Tuan Quoc Nguyen
15.1 Introduction 757
15.2 General Features of Polymer Degradation 759
15.2.1 Degradative Reactions 759 Initiation 760 Propagation 760 Chain Branching 761 Termination 762
15.2.2 Some Nonradical Degradation Mechanisms 763
XX Contents

15.2.3 Physical Factors 763 Glass Transition Temperature 764 Polymer Morphology 766
15.3 Degradation Detection Methods 767
15.3.1 Mechanical Tests 768
15.3.2 Gel Permeation Chromatography 771
15.3.3 Fourier Transform Infrared Spectroscopy 773
15.3.4 Magnetic Resonance Spectroscopy 775 Nuclear Magnetic Resonance (NMR) 775 Electron Spin Resonance (ESR) 776
15.3.5 Oxygen Uptake 776
15.3.6 Chemiluminescence 778
15.4 Thermal Degradation 778
15.4.1 Thermal Stability 779
15.4.2 Polymer Structure and Thermal Stability 779
15.4.3 Computer Simulation 780
15.4.4 Thermal Oxidative Degradation of Polypropylene 782 Initiation 782 Propagation 784 Chain Branching 785 Termination 786 Secondary Reactions 786 Formation of Volatile Compounds 788
15.4.5 Homogeneous versus Heterogeneous Kinetics 789
15.4.6 Applications of Thermal Degradation 790 Analytical Pyrolysis 790 Introduction of New Chemical Functionalities 791 Chemical Modification of Polymer Structure 791 Metal Injection Molding (MIM) 792 Recycling 792
15.5 Photodegradation 793
15.5.1 Absorption of UV Radiation by Polymers 793
15.5.2 The Solar Spectrum 796
15.5.3 Photo-oxidation Profile 796
15.5.4 Influence of Wavelength: the Activation and Action Spectrum 799
15.5.5 Photodegradation Mechanisms 802 Photoinitiation 802 The Norrish Photoprocesses 803 Photo-Fries Rearrangement 803
15.6 Radiolytic Degradation 805
15.6.1 Interaction of High-energy Radiation with Matter 805
15.6.2 Radiation Chemistry 807
15.6.3 Radiolysis Stabilization 810
15.6.4 Applications 811 Radiation Sterilization 812
Contents XXI Controlled Degradation and Crosslinking 812

15.7 Mechanochemical Degradation 813
15.7.1 Initiation by Mechanical Stresses 813 Effect of Tensile Stress on Chemical Reactivity 813 Breaking Strength of a Covalent Bond 814 Rate of Stress-activated Chain Scission 815
15.7.2 Extrusion Degradation 816
15.7.3 Applications 817
15.8 Control and Prevention of Aging of Plastic Materials 818
15.8.1 Antioxidants 818 Radical Antioxidants 818 Hindered Amine Stabilizers (HAS) 819 Peroxide Decomposers 821
15.8.2 Photostabilizers 822
15.8.3 PVC Heat Stabilizers 823
15.8.4 Other Classes of Stabilizers 824 Metal Deactivators 824 Antiozonants 824
15.9 Lifetime Prediction 824
15.10 Conclusions 826
Notation 827
References 830

16 Thermosets 833
Rolf A. T. M. van Benthem, Lars J. Evers, Jo Mattheij, Ad Hofland, Leendert J.
Molhoek, Ad J. de Koning, Johan F. G. A. Jansen, and Martin van Duin
16.1 Introduction 833
16.1.1 Thermoset Materials 833
16.1.2 Networks 834
16.1.3 Advantages 835
16.1.4 Curing Resins 835
16.1.5 Functionality 835
16.1.6 Formulation 836
16.1.7 Production 837
16.1.8 General Areas of Application 837
16.2 Phenolic Resins 838
16.2.1 Introduction 838
16.2.2 Chemistry 838 Resols 840 Novolacs 840 Epoxy-novolacs 841 Discoloration 841
16.2.3 Production 842
16.2.4 Properties and Applications 842
16.3 Amino Resins 843
XXII Contents

16.3.1 Introduction 843

16.3.2 Chemistry 843 Polymerization Chemistry 845
16.3.3 Production 848
16.3.4 Properties and Applications 849
16.4 Epoxy Resins 849
16.4.1 Introduction 849
16.4.2 Chemistry 850 Cure 851
16.4.3 Production 853 Standard Liquid 853 Taffy Process 854 Advancement Process 854
16.4.4 Properties and Applications 855
16.5 Alkyd Resins 855
16.5.1 Introduction 855
16.5.2 Chemistry 856 The Alkyd Constant 858 Autoxidative Drying 858
16.5.3 Production 859
16.5.4 Properties and Applications 861 Short Oil Alkyds 861 Long Oil Alkyds 861 Medium Oil Alkyds 861
16.5.5 Alkyd Emulsions 861 The Inversion Process 862
16.6 Saturated Polyester Resins 862
16.6.1 Introduction 862
16.6.2 Chemistry 863
16.6.3 Production 865 Monitoring the Reaction 865
16.6.4 Properties and Applications 866
16.6.5 Powder Coatings 866 Application 867 Crosslinking 868 Advantages 869
16.7 Unsaturated Polyester Resins and Composites 869
16.7.1 Introduction 869
16.7.2 Chemistry 869 Crosslinking 871 Styrene Emission 871 Vinyl Ester Resins 873
16.7.3 Production 874
16.7.4 Reinforcement 875
16.7.5 Fillers 878
Contents XXIII

16.7.6 Processing 879 Hand Lay-up and Spray-up 882 Continuous Lamination 882 Filament Winding 882 Centrifugal Casting 882 Pultrusion 883 Cold-press Molding 883 Resin Infusion 883 Resin-transfer Molding 883 Hot-press Molding 883 Casting, Encapsulation, and Coating (Non-reinforced Applications)
16.7.7 Design Considerations: Mechanical Properties of Composites 886
16.8 Acrylate Resins and UV Curing 889
16.8.1 Introduction 889
16.8.2 Chemistry 890
16.8.3 Production 891 Epoxy Acrylates 891 Polyester Acrylates 891 Urethane Acrylates 892 Inside-out 893 Outside-in 894 Comparing Inside-out with Outside-in 894 Stabilization 894 Dilution 895 Safety 895
16.8.4 Properties 895
16.8.5 Introduction to UV Curing 896 General Introduction to UV-initiated Radical Polymerization 896 Photoinitiators for Radical Polymerization 897 Resin 897 Reactive Diluent 898 Curing Process 899 Cationic Curing 900 Base-mediated Curing 901
16.9 Rubber 901
16.9.1 Introduction 901 Types of Rubber 902
16.9.2 Polymerization 903
16.9.3 Crosslinking 904 Sulfur Vulcanization 904 Peroxide Curing 905 Processing 906
16.9.4 Properties and Applications 907 Advantages and Disadvantages 907
XXIV Contents Thermoplastic Vulcanizates 907

Notation 908
References 909

17 Fibers 911
J. A. Juijn
17.1 Introduction 911
17.1.1 A Fiber World 911
17.1.2 Scope of this Chapter 912
17.2 Fiber Terminology 912
17.2.1 Definitions: Fibers, Filaments, Spinning 912
17.2.2 Synthetic Yarns 914
17.2.3 Titer: Tex and Denier 914
17.2.4 Tenacity and Modulus: g denierC1, N texC1 , or GPa 915
17.2.5 Yarn Appearance 916
17.2.6 Textile, Carpet, and Industrial Yarns 917
17.2.7 Physical Structure 918
17.3 Fiber Polymers: Choice of Spinning Process 920
17.3.1 Polymer Requirements 920
17.3.2 Selection of Spinning Process 920
17.3.3 Spinnability 922
17.4 Melt Spinning 923
17.4.1 Extrusion 923
17.4.2 Polymer Lines and Spin-box 924
17.4.3 Spinning Pumps 925
17.4.4 Spinning Assembly 926 Filtration 926 Spinning Plate 926
17.4.5 Quenching 928
17.4.6 Finish 929
17.4.7 Spinning Speed 931
17.4.8 Winding 931
17.4.9 Drawing 931
17.4.10 Relaxation and Stabilization 934
17.4.11 Process Integration 934
17.4.12 Rheology 934 Shear Viscosity 934 Elasticity 936 Elongational Viscosity 936
17.4.13 Process Calculations 936 Mass Flow 937 Volume Flow 937 Extrusion Speed and Elongation in the Spin-line 937 Pressure Drop over the Spinning Holes 938
Contents XXV

17.4.14 Polyester (Poly(ethylene terephthalate), PET) 938 PET Polymer 938 Spinning of PET 939 PET Staple Fiber 939 PET Textile Filament Yarns 940 PET Industrial Yarns 940
17.4.15 Polyamide (PA6 and PA66) 941 PA Polymer 941 PA Spinning 941 PA Staple Fiber 942 PA Textile Filament Yarns 942 PA Industrial Yarns 942
17.4.16 Polypropylene (PP) 943 PP Polymer 943 PP Spinning 943 PP Staple Fiber 943 PP Split Fiber 943 PP Filament Yarns 944
17.5 Solution Spinning 944
17.5.1 Preparation of Spinning Dope 944
17.5.2 Dry Spinning 944 Cellulose Acetate 945 Acrylics 946 Poly(vinyl alcohol) 946
17.5.3 Wet Spinning 946 Viscose Rayon 948 Acrylics 951 Poly(vinyl alcohol) 952
17.6 Comparison of Melt and Solution Spinning 953
17.7 High-modulus, High-strength Fibers 956
17.7.1 Air-gap Spinning 956 Aramids 956 Other Liquid-crystalline Polymers 960
17.7.2 Gel Spinning 961 Theory 961 Gel Spinning of Polyethylene 962 Other Gel-spun or Superdrawn Fibers 964
17.7.3 Carbon Fiber 965 Carbon Fiber from PAN 965 Carbon Fiber from Pitch 966 Applications of Carbon Fibers 966
17.7.4 Other Advanced Fibers 966
Notation 967
Acknowledgments 969
References 969
XXVI Contents

18 Removal of Monomers and VOCs from Polymers 971

Marı́a J. Barandiaran and José M. Asua
18.1 Introduction 971
18.2 Polymer Melts and Solutions 972
18.2.1 Devolatilization 973 Fundamentals 973 Implementation of Devolatilization 975 Equipment 975
18.3 Polyolefins 979
18.4 Waterborne Dispersions 979
18.4.1 Post-polymerization 980 Modeling Post-polymerization 981
18.4.2 Devolatilization 981 Modeling 982 Rate-limiting Steps 985 Devolatilization under Equilibrium Conditions 986 Equipment 986
18.4.3 Combined Processes 988
18.4.4 Alternative Processes 989
18.5 Summary 989
Notation 990
References 991

19 Nano- and Microstructuring of Polymers 995

Christiane de Witz, Carlos Sánchez, Cees Bastiaansen, and Dirk J. Broer
19.1 Introduction 995
19.1.1 Patterning Techniques 996
19.1.2 Photoembossing 998
19.2 Materials and their Photoresponsive Behavior 999
19.3 Single-exposure Photoembossing 1001
19.4 Dual-exposure Photoembossing 1007
19.5 Complex Surface Structures from Interfering UV Laser Beams 1007
19.6 Surface Structure Development under Fluids 1010
19.7 Conclusion 1012
Acknowledgments 1012
Notation 1013
References 1013

20 Chemical Analysis for Polymer Engineers 1015

Peter Schoenmakers and Petra Aarnoutse
20.1 Introduction 1015
20.2 Process Analysis 1017
20.2.1 Near-infrared Spectroscopy 1017
20.2.2 In-situ Raman Spectroscopy 1018
20.2.3 At-line Conversion Measurements 1020
Contents XXVII

20.3 Polymer Analysis 1022

20.3.1 Basic Laboratory Measurements 1022 Conversion 1022
20.3.2 Detailed Molecular Analysis 1023 FTIR Spectroscopy 1023 NMR Spectroscopy 1024 Mass Spectrometry 1025
20.3.3 Polymer Distributions 1030 Molecular Weight Distributions 1030 Functionality-type Distributions 1034 Chemical Composition Distributions (CCDs) 1037 Degree of Branching Distributions 1040 Complex Polymers (Multiple Distributions) 1041
Notation 1044
References 1045

21 Recent Developments in Polymer Processes 1047

Maartje Kemmere
21.1 Introduction 1047
21.2 Polymer Processes in Supercritical Carbon Dioxide 1048
21.2.1 Interactions of Carbon Dioxide with Polymers and Monomers 1050 Solubility in Carbon Dioxide 1051 Sorption and Swelling of Polymers 1052 Phase Behavior of Monomer–Polymer–Carbon Dioxide Systems 1054
21.2.2 Polymerization Processes in Supercritical Carbon Dioxide 1055
21.2.3 Polymer Processing in Supercritical Carbon Dioxide 1058 Extraction 1060 Impregnation and Dyeing 1061
21.3 Ultrasound-induced Radical Polymerization 1062
21.3.1 Ultrasound and Cavitation in Liquids 1063
21.3.2 Radical Formation by Cavitation 1065
21.3.3 Cavitation-induced Polymerization 1067 Bulk Polymerization 1067 Precipitation Polymerization 1069 Emulsion Polymerization 1070
21.3.4 Cavitation-induced Polymer Scission 1072
21.3.5 Synthesis of Block Copolymers 1073
21.4 Concluding Remarks and Outlook for the Future 1074
Acknowledgments 1076
Notation 1076
References 1077

Index 1083


Freshly started as chairman and secretary of the Working Party on Polymer Reac-
tion Engineering it never crossed our mind to edit a book on this subject. This
changed when Wiley-VCH asked if the working party would be able to provide a
translation of the Handbuch der Technischen Polymerchemie, written in 1993 by
Adolf Echte. We decided to do so, but not exactly. Very rapidly we were convinced
that we needed a completely new book, covering the field of polymer reaction engi-
neering in a modern, broad and multidisciplinary approach. Many of the working
party members directly agreed to participate, others needed somewhat stronger
persuasion techniques, and for some chapters we hired authors from other in-
stitutions. In June 2003 we had completed the list of contributors, coming from
Europe, Canada and the USA. Now, roughly one year later, the new handbook is
The quality an edited book like this very much depends on the quality of the in-
dividual contributions. It has been a great pleasure for us to see that all authors
have taken their writing jobs very seriously. With these contributions, we are sure
that this book represents the state of the art in polymer reaction engineering. It
is intended to attract equally readers that are new in the field as well as readers
that may be considered expert in some of the topics but want to broaden their
knowledge. We are convinced that the multidisciplinary and synergetic approach
presented in this book may act as an eye-opener for research and development ac-
tivities going on in strongly related areas. We hope the reader will take advantage
of this approach, where the references given in the various chapters may be a start-
ing point for further reading.
Reading books, you often read the preface as well. We have seen numerous ex-
amples from which the frustration is quite obvious. Of course things may not al-
ways work out the way you plan, that has also been the case for this book. Maybe
we were just lucky, but we have greatly enjoyed doing this. Editing this book has
also been a starting point for the editors to become friends, including Swiss cheese
fondue and Dutch ‘‘Friese nagelkaastaart’’ in a friendly home setting. From that
perspective also Francine and Maartje have had their part both of the workload
but also of the fun of all this.
Finally, we would like to thank Karin Sora and Renate Doetzer from Wiley-VCH
XXX Preface

for their help with the editing process. They really know to find the balance be-
tween waiting and pushing in order not to diverge too far from the schedule.

Lausanne & Eindhoven, fall 2004

Thierry Meyer & Jos Keurentjes

List of Contributors

P. Aarnoutse Prof. M. J. Barandiaran

Polymer-Analysis Group The University of the Basque Country
Department of Chemical Engineering Institute for Polymer Materials
Faculty of Science, University of Manuel Lardizabal, 3
Amsterdam 20018 Donostia-San Sebastián
Nieuwe Achtergracht 166 Spain
1018 WV Amsterdam
The Netherlands Dr. C. W. M. Bastiaansen
Eindhoven University of Technology
Dr. U. S. Agarwal Den Dolech 2
Polymer Technology Group 5600 MB Eindhoven
Department of Chemical Engineering The Netherlands
and Chemistry
Eindhoven University of Technology Prof. D. J. Broer
P.O. Box 513 Philips Research Laboratories
5600 MB Eindhoven Prof. Holstlaan 4
The Netherlands 5656 AA Eindhoven
The Netherlands
Prof. J. M. Asua
The University of the Basque Country
Institute for Polymer Materials Eindhoven University of Technology
(POLYMAT) Den Dolech 2
Paseo Manuel Lardizabal 3 5600 MB Eindhoven
20018 Donostia-San Sebastián The Netherlands
Dr. R. Bachmann Dutch Polymer Institute (DPI)
Bayer AG P.O. Box 902
ZT-TE-SVT 5600 AX Eindhoven
51368 Leverkusen The Netherlands
XXXII List of Contributors

Polymer Technology Prof. J. C. de la Cal

Department of Chemical Engineering The University of the Basque Country
and Chemistry Institute for Polymer Materials
Eindhoven University of Technology (POLYMAT)
P.O. Box 513 Paseo Manuel Lardizabal, 3
5600 MB Eindhoven 20018 Donostia-San Sebastián
The Netherlands Spain

Prof. B. W. Brooks Dr. A. J. de Koning

Loughborough University DSM Research
Department of Chemical Oude Postbaan 1
Engineering 6167 RG Geleen
Loughborough The Netherlands
Leicestershire, LE11 3TU
United Kingdom Dr. T. W. de Loos
Delft University of Technology
Dr. A. Buttè Faculty of Applied Sciences
Swiss Federal Institute of Department Chemical Technology
Technology Julianalaan 136
Zurich, ETHZ 2628 BL Delft
Institut für Chemie- und The Netherlands
Gruppe Morbidelli C. de Witz
ETH Hoenggerberg/HCI F135 Philips Research Laboratories
8093 Zurich Prof. Holstlaan 4
Switzerland 5656 AA Eindhoven
The Netherlands
Dr. J. P. Congalidis
E.I. du Pont de Nemours and Dr. L. J. Evers
Company DSM Melamine
DuPont Central Research and Oude Postbaan 1
Development 6167 RG Geleen
Experimental Station The Netherlands
Wilmington, DE 19880
USA Dr. A. Hofland
DSM Coating Resins
Dr. M. R. P. F. N. Costa Ceintuurbaan 5
Faculty of Engineering 8022 AW Zwolle
University of Porto The Netherlands
Rua Roberto Frias, s/n
4200-465 Porto Dr. K.-D. Hungenberg
Portugal BASF AG
Polymer Technology, B1
67056 Ludwigshafen
List of Contributors XXXIII

Prof. R. A. Hutchinson N. H. Kolhapure

Department of Chemical Engineering DuPont Engineering Research and
Queen’s University Technology
Dupuis Hall, 19 Division St. 1007 N. Market St.
Kingston, Ontario K7M 2G9 Wilmington, DE 19898-0001
Canada USA

Dr. P. D. Iedema Prof. J. R. Leiza

Department of Chemical Engineering The University of the Basque Country
University of Amsterdam Institute for Polymer Materials
Nieuwe Achtergracht 166 (POLYMAT)
1018 WV Amsterdam Paseo Manuel Lardizabal 3
The Netherlands 20018 Donostia-San Sebastián
Dr. J. F. G. A. Jansen
DSM Research J. Mattheij
Oude Postbaan 1 DSM Melamine
6167 RG Geleen Oude Postbaan 1
The Netherlands 6167 RG Geleen
The Netherlands
Dr. J. A. Juijn
Research Institute Dr. T. Meyer
Department QRI Swiss Federal Institute of Technology
P.O. Box 9600 Institute of Process Science
6800 TC Arnheim EPFL, ISP-GPM
The Netherlands 1015 Lausanne
Dr. M. F. Kemmere
Process Development Group L. J. Molhoek
Department of Chemical Engineering DSM Coating Resins
and Chemistry Ceintuurbaan 5
Eindhoven University of Technology 8022 AW Zwolle
P.O. Box 513 The Netherlands
5600 MB Eindhoven
The Netherlands Prof. M. Morbidelli
Swiss Federal Institute of Technology
Prof. J. T. F. Keurentjes Zurich, ETHZ
Process Development Group Institut für Chemie- und
Eindhoven University of Technology Bioingenieurwissenschaften
P.O. Box 513 Gruppe Morbidelli
5600 MB Eindhoven ETH Hoenggerberg/HCI F135
The Netherlands 8093 Zurich
XXXIV List of Contributors

Prof. E. B. Nauman Prof. P. J. Schoenmakers

The Isermann Department of Chemical Polymer-Analysis Group
and Biological Engineering Department of Chemical Engineering
Rensselaer Polytechnic Institute (ITS)
Troy, NY 12180 Faculty of Science, University of
USA Amsterdam
Nieuwe Achtergracht 166
Dr. Q. T. Nguyen 1018 WV Amsterdam
Laboratory of Polymers (LP) The Netherlands
Ecole Polytechnique Fédérale
de Lausanne Prof. L. C. Simon
1015 Lausanne Department of Chemical Engineering
Switzerland University of Waterloo
200 University Avenue West
J. C. Plummer Waterloo, Ontario N2L 3G1
Laboratory of Composite and Polymer Canada
Technology (LTC)
Ecole Polytechnique Fédérale Prof. J. B. P. Soares
de Lausanne Department of Chemical Engineering
1015 Lausanne University of Waterloo
Switzerland 200 University Avenue West
Waterloo, Ontario N2L 3G1
Dr. J. R. Richards Canada
E. I. du pont de Nemours and Company
DuPont Engineering and Research Prof. F. Stoessel
Technology Swiss Institute for the Promotion of
Experimental Station Safety and Security
Wilmington, DE 19880 Chemical Process Safety Consulting
USA Klybeckstrasse 141
Dr. C. Sánchez 4002 Basel
Eindhoven University of Technology Switzerland
Den Dolech 2
5600 MB Eindhoven Prof. G. Storti
The Netherlands Swiss Federal Institute of Technology
Zurich, ETHZ
Institut für Chemie- und
Dutch Polymer Institute (DPI) Bioingenieurwissenschaften
P.O. Box 902 Gruppe Morbidelli
5600 AX Eindhoven ETH Hoenggerberg/HCI F125
The Netherlands 8093 Zurich
List of Contributors XXXV

Prof. R. A. T .M. van Benthem Dr. M. van Duin

Coating Technology DSM Research
Department of Chemical Engineering Oude Postbaan 1
and Chemistry 6167 RG Geleen
Eindhoven University of Technology The Netherlands
P.O. Box 513
5600 MB Eindhoven
The Netherlands

Polymer Reaction Engineering, an Integrated

Th. Meyer and J. T. F. Keurentjes

Polymer Materials

Synthetic polymers can be denoted as the materials of the 20th century. Since
World War II the production volume of polymers has increased by a factor of 50
to a current value of more than 120 million tonnes annually (Figure 1.1). The con-
sumption per capita has also increased over the years to a worldwide average of ap-
proximately 20 kg per annum in the year 2000. In terms of volumetric output, the
production of polymers exceeds that of iron and steel. The enormous growth of
synthetic polymers is due tot the fact that they are lightweight materials, act as in-
sulators for electricity and heat, cover a wide range of properties from soft packag-
ing materials to fibers stronger than steel, and allow for relatively easy processing.

Annual Production 120
106 to/an
kg/hab 80

World Population , 20
109 people
0 0
1940 1950 1960 1970 1980 1990 2000 2010
Fig. 1.1. Polymer production and the evolution of the population since 1940 [1].

Handbook of Polymer Reaction Engineering. Edited by T. Meyer, J. Keurentjes

Copyright 8 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
ISBN: 3-527-31014-2
2 1 Polymer Reaction Engineering, an Integrated Approach

Tab. 1.1. Applications and 2002 Western European markets for the major thermoplastics [1].

Thermoplastic Market Applications

[10 3 tonnes]

LDPE 7935 pallet and agricultural film, bags, toys, coatings,

containers, pipes
PP 7803 film, battery cases, microwave-proof containers, crates,
automotive parts, electrical components
PVC 5792 window frames, pipes, flooring, wallpaper, bottles, cling
film, toys, guttering, cable insulation, credit cards,
medical products
HDPE 5269 containers, toys, housewares, industrial wrappings and
films, pipes
PET 3234 bottles, textile fibers, film food packaging
PS/EPS 3279 electrical appliances, thermal insulation, tape cassettes,
cups and plates, toys
PA 1399 film for food packaging (oil, cheese, ‘‘boil-in-bag’’), high-
temperature engineering applications, textile fibers
ABS/SAN 788 general appliance moldings
PMMA 317 transparent all-weather sheet, electrical insulators,
bathroom units, automotive parts

Moreover, parts with complex shapes can be made at low cost and at high speed by
shaping polymers or monomers in the liquid state.
The polymer market can be divided into thermoplastics and thermosets. The ma-
jor thermoplastics include high-density polyethylene (HDPE), low-density polyeth-
ylene (LDPE), polyethylene terephthalate (PET), polypropylene (PP), polystyrene
(PS and EPS), poly(vinyl chloride) (PVC), polyamide (PA), poly(methyl methacry-
late) (PMMA) and styrene copolymers (ABS, SAN). The most important applica-
tions of thermoplastics are summarized in Table 1.1. The total Western European
demand for thermoplastics was 37.4 million tonnes in 2002, a growth of about 9%
as compared to 2001 [1]. Thermoplastics are used not only in the manufacture of
many typical plastics applications such as packaging and automotive parts, but also
in non-plastic applications such as textile fibers and coatings. These non-plastic ap-
plications account for about 14% of all thermoplastics consumed.
The major thermosets include epoxy resins, phenolics, and polyurethanes (PU),
for which the major applications are summarized in Table 1.2. It has to be noted,

Tab. 1.2. Applications and 2002 Western European markets for the major thermosets [1].

Thermoset Market [10 3 tonnes] Applications

PU 3089 coatings, finishes, cushions, mattresses, vehicle seats

Phenolics 912 general appliance moldings, adhesives, appliances,
automotive parts, electrical components
Epoxy resins 420 adhesives, automotive components, E&E components,
sports equipment, boats
1.1 Polymer Materials 3


Large industry
Electrical and 2.5%

Building and Packaging

construction 38.1%
Fig. 1.2. Plastic consumption in 2002 by industry sectors in Western Europe [1].

however, that about one-third of the market for thermosets is for relatively small-
scale specialty products. The total Western European market for thermosets was
10.4 million tonnes in 2002, about 1% below the 2001 value.
The major application areas of polymers can be defined as follows (Figure 1.2).

Automotive industry Motorists want high-performing cars combined with reliabil-

ity, safety, comfort, competitive pricing, fuel efficiency, and, increasingly, reassur-
ance about the impact on the environment. Lightweight polymeric materials are
increasingly used in this sector (Daimler Benz’s Smart is a nice example), also con-
tributing to a 10% reduction in passenger fuel consumption across Europe.

Building and construction Polymeric materials are used in the building and con-
struction sector, for example for insulation, piping, and window frames. In 2002
this sector accounted for 17.6% of the total polymer consumption.

Electrical and electronic industry Many applications in this field arise from newly
designed polymeric materials, for example for polymeric solar cells and holo-
graphic films. It is interesting to note that, while the number of applications in
this field is increasing, the weight of the polymers used per unit is decreasing.

Packaging The packaging sector remains the largest consumer of synthetic poly-
mers, approximately 38% of the total market. This is mainly due to the fact that
these materials are lightweight, flexible, and easy to process, and are therefore
increasingly being substituted for other materials. Although polymer packaging
ranks first in terms of units sold, it is only third if judged on weight.

Agriculture As agricultural applications account for about 2.5% of the total of syn-
thetic polymers consumed in Europe, they play only a marginal role. Irrigation and
4 1 Polymer Reaction Engineering, an Integrated Approach

drainage systems provide effective solutions to crop growing, and polymeric films
and greenhouses can increase horticultural production substantially. The use of so-
called ‘‘super absorbers’’ for increased irrigation efficiency in arid areas can be con-
sidered an important emerging market.

A Short History of Polymer Reaction Engineering

In Table 1.3 a comprehensive overview of the major developments in the polymer

industry is given. In the 19th century, polymers produced by Nature, such as cellu-
lose, Hevea brasiliensis latex (natural rubber), and starch, were processed to manu-
facture useful products. This practice was often based on experimental discoveries.
As an example, in 1839 Goodyear discovered by mistake the sulfur vulcanization of
natural rubber, making it possible for Ford to develop the automotive market. In
those times no polymers were produced synthetically.
Early in the 20th century (1920), the first empirical description of macromole-
cules was developed by Staudinger [2]. At the same time, new methods were devel-
oped to determine the specific characteristics of these materials. In the 1930s many
research groups (for examples see refs. 3–7) developed models for the chain length
distribution in batch reactors resulting from different polymer chemistries, a meth-
odology that was further developed in the 1940s leading to more complex and com-
prehensive models, some of which are still being used today.

Tab. 1.3. The history of polymers in brief.

19th century natural polymer and derivatives (vulcanized rubber, celluloid)

1920 concept of macromolecules postulated by Staudinger
1930–1940 first systematic synthesis of polymers
synthesis of polyamides (nylon) by Carothers at DuPont
discovery of polyethylene at ICI (Fawcett and Gibson)
1940–1950 synthetic rubbers and synthetic fibers
1950–1960 stereospecific polymerizations by Ziegler and Natta, the birth of
discovery of polymer single crystals (Keller, Fischer, Till)
development of polycarbonate
1960–1970 discovery of PPO at GE by Hay and commercialization of PPO/PS
blends (Noryl2 )
1970–1980 liquid-crystalline polymers
1980–1990 superstrong fibers (Aramid2 , polyethylene)
functional polymers (conductive, light-emitting)
1990–2000 metallocene-based catalysts; novel polyolefins hybrid systems
(polymer/ceramic, polymer/metals)
2000– Nature-inspired catalysts
synthesis of polymers by bacteria and plants
1.3 The Position of Polymer Reaction Engineering 5

Around 1940, partly inspired by World War II, a more systematic search for new
synthetic polymer materials as a replacement for scarce natural materials led to the
development of nylon (DuPont) and polyethylene (ICI) [8, 9]. This was followed by
the development of synthetic rubbers and synthetic fibers. In the same period,
Denbigh [10] was one of the first to introduce chemical reaction engineering con-
cepts into polymer science by considering polymerization reactions at both the
chemical and at the process levels. Processes were classified as homocontinuous
and heterocontinuous, depending on the mixing level. This pioneering approach
also acted as a catalyst for the further development of polymer reaction engineer-
ing (PRE).
The development of catalysts based on transition metals by Ziegler and Natta
[11] allowed the development of stereospecific propylene polymerization processes
and ethylene polymerization in the 1950s. Several process schemes were developed
at that time, of which some are still in use. The major problem in process develop-
ment has been to deal with the heat of polymerization, an issue that was solved, for
example, by using an inert solvent as a heat sink or by flashing of monomer fol-
lowed by condensation outside the reactor. In the same period, polycarbonate and
(somewhat later) poly(propylene oxide) (PPO) were developed. The main character-
istic of the polymers developed so far was that they were bulk materials, to be pro-
duced in extremely large quantities.
In the 1970s, a paradigm shift occurred when polymers with more specific prop-
erties started to be produced. This included various liquid crystalline polymers
leading, for example, to the production of superstrong fibers such as Aramid2 /
Kevlar 2 [12]. The development of functional polymers for the conduction of light
and electricity and optical switches also started then [13]. In the near future this
will probably lead to highly effective and flexible polymer solar cells [14].
In the 1990s, metallocene catalysts were developed for polyolefin production that
surpassed the Ziegler–Natta catalysts in terms of selectivity and reactivity [15, 16].
Additionally, various hybrid materials were combining properties of both the poly-
mer (lightweight, flexible) and a solid material, which could be metal (conductive)
or ceramic (insulating), leading to materials with specific properties applicable, for
example, as protective coatings [17].
Current developments include the mimicking of nature (enzymes) for the syn-
thesis of quite complex polymers like natural silk. Also, bacteria and plants are be-
ing modified to produce polymers of interest [18]. However, this can be expected
to require polymer reaction engineering developments that are as yet difficult to

The Position of Polymer Reaction Engineering

Traditional chemical reaction engineering has its basis in the application of scien-
tific principles (from disciplines such as chemistry, physics, biology, and mathe-
matics) and engineering knowledge (transfer of heat, mass, and momentum) to
6 1 Polymer Reaction Engineering, an Integrated Approach

Energy Integrated heat recovery Less energy

demanding processes

Raw material Coal, Oil Natural gas Renewable feedstock

Integrated and
Safety Plant operation inherent safety

Market Capacity Quality control Flexible production

Pollution Recycling
Water Air Green solvents
control Clean processes

Process Optimization Intensification

Scientific From empiricism to strategy Multidisciplinarity


1980 1990 2000 2010

Fig. 1.3. Changing priorities in industrial chemical engineering research.

the solution of problems of practical, industrial, and societal importance. Since the
1970s, a changing focus in chemical reaction engineering can be observed, which
is summarized in Figure 1.3.
To deal with more stringent requirements in terms of energy consumption re-
quires a shift from heat loss minimization toward novel intensified process con-
cepts that intrinsically require less energy. Safety should now be considered as an
intrinsic plant property rather than a responsive action, and the plant needs to be
flexible to be able to respond quickly to changes in the market. Last but not least,
new concepts will be required to provide a basis for sustainable future develop-
ments, that is, the use of renewable resources and processes based on ‘‘green’’
solvents. As a result of this changing focus, a shift toward a multidisciplinary
approach can be observed.
For PRE this implies the combination of several disciplines such as polymer
chemistry, thermodynamics, characterization, modeling, safety, mechanics, phys-
ics, and process technology. PRE problems are often of a multi-scale and multi-
functional nature to achieve a multi-objective goal. One particular feature of PRE
is that the scope ranges from the micro scale on a molecular level up to the macro
scale of complete industrial systems. PRE plays a crucial role in the transfer of in-
formation across the boundaries of different scale regions and to provide a compre-
hensive and coherent basis for the description of these processes [19].
As depicted in Figure 1.4, there is a direct link between time and size scale, from
which it is obvious that the micro and macro scales are not related to the same
time scale [20]. As an example, molecular dynamics calculations are addressing a
time scale in the order of femto- to nanoseconds, whereas process system integra-
tion evolves on the scale of years. Engineers have traditionally been working at the
meso scale, which is represented by the middle portion of Figure 1.4, using phe-
1.4 Toward Integrated Polymer Reaction Engineering 7

Time scale

Years System integration

Environmental, Global
Engineering design
Days Process models
Molecular level
Elementary reactions
Molecular modeling
Nanoseconds Chemical equilibrium

Atomic level
Picoseconds Fundamental
Quantum techniques
Quantum chemistry

1Å 10Å 100Å 1µm 1mm 1m 1km

Size scale
Fig. 1.4. Activities in PRE with their corresponding time and size scales.

nomenological and continuum models. Today these limits are pushed in two direc-
tions, both toward a more fundamental understanding and at the same time to-
ward a more global scale. In the past, the ‘‘micro-region’’ has traditionally been
the domain of physicists and chemists, whereas the ‘‘macro-region’’ has been the
field, rather, of process or plant engineers. Today, it becomes obvious that only us-
ing a multidisciplinary, parallel, and synergetic approach can lead to successful de-
velopments. Polymer reaction engineering will play an essential role as the core
and the coordinator of this complex process.

Toward Integrated Polymer Reaction Engineering

As will be obvious from the foregoing discussion, PRE is composed of many disci-
plines all linked together. These disciplines can be either mature or emergent, but
they have a common gateway (see Figure 1.5). Although there is not necessarily a
direct connection between them, there exists a common core in which the different
disciplines make their own specific contribution to a general objective.
The frontiers in PRE are determined by what we know, understand, and are able
to quantify, and these frontiers are moving with growing knowledge, competences,
and experience. Efforts to push these limits will induce innovative developments
leading to emerging technologies and products, and will also strengthen the multi-
disciplinary approach. In general terms, PRE can be defined as the science that
8 1 Polymer Reaction Engineering, an Integrated Approach

Materials sciences

Novel processes

Inherent safety Materials


Recycling, Disposal
Modeling and
Polymer Reaction simulation
Process integration (PRE)
optimization New products

Polymer chemistry Post-reaction

Reaction kinetics processes

Nano-, Micro-
Novel processes Bio-

Measurement and control

Fig. 1.5. The expanding sphere of polymer reaction engineering.

brings molecules to an end-use product. We can either consider it like a black box
(Figure 1.6) or we can try to define the interconnected disciplines that compose
this black box (Figure 1.7). Provided the required product properties can be met,
we expect that sustainability is the common denominator for all the disciplines in-
volved in this process.
The process of transforming raw materials into valuable end-use products is not
a one-way procedure but rather an iterative process in which we try to optimize all
the parameters involved. The selection of the proper chemistry and technology
should include an evaluation of environmental, safety, and economic parameters.
Moreover, questions regarding the possible use of renewable resources and mini-
mizing the energy requirement will have to be answered. Defining PRE in this
manner appears to be very close to the procedure of life cycle analysis (LCA) [21].

Polymer Reaction
Raw materials End use product

Fig. 1.6. PRE as a black box process.

1.5 The Disciplines in Polymer Reaction Engineering 9


Materials sciences

Novel processes

Raw materials Inherent safety Materials

Application Products
Environment Profit
Needs Recycling, Disposal
Modeling and
simulation Satisfaction
Integrated PRE
Process integration Knowledge
optimization New products

Polymer chemistry Post-reaction

Reaction kinetics processes

Nano-, Micro-
Novel processes Bio-

Measurement and control


Fig. 1.7. The integrated approach for sustainable PRE.

Life cycle analysis is a tool assisting decision making in the engineering process.
LCA includes the information on the history of the materials used, and the dif-
ferent process and raw material alternatives, as well as the final product require-
ments. LCA is an instrument driven by environmental considerations against a
background of technical and economic specifications, and involves the so-called 3-
P concept (people, planet, and profit). The LCA-based PRE methodology (Figure
1.8) [22] leads to an optimization of all the parameters involved and a reduction of
the costs. This seems to be contradictory at first sight, but integrating all the as-
pects often leads to cost reductions. In our view, the use of this approach will lead
to a ‘‘sustainable integrated PRE’’.

The Disciplines in Polymer Reaction Engineering

The different disciplines involved in PRE can be represented using the academia–
industry dichotomy (Figure 1.9). The interests of the two types of players are not
identical: the differences are similar to the differences in their mission statements.
Nevertheless, we can observe that a great overlap is present in the middle zone,
10 1 Polymer Reaction Engineering, an Integrated Approach


Specification Balances
recycling raw material
- technical - energy
- economic - material
- ecologic - emission
- safety product use synthesis - waste
- sewage


Evaluation leads to closed

loop assessment of costs

Fig. 1.8. Life cycle analysis of parts, methods, products, and systems.

where interests, tools, and knowledge are similar, thus providing a strong basis for
As stated above, PRE is composed of a large number of disciplines, which are
described in more detail in the following chapters of this handbook. These disci-
plines are interconnected by a synergetic and multidisciplinary approach, and com-

Fundamental Novel
kinetics technologies

Molecular Measurement
modeling and control

Reactor design Environment

Polymer physics

Polymer chemistry Safety


Applied Market
modeling Process
modeling economics

Fig. 1.9. Overlap of industrial and academic disciplines.
1.5 The Disciplines in Polymer Reaction Engineering 11


Fig. 1.10. Product-driven PRE, based on an orthogonal

relationship between science and engineering.

mercial products are the final achievement resulting from this methodology. This
could be expressed by an orthogonal representation (Figure 1.10) where polymer
sciences are linked with engineering sciences. Every type of polymerization will
have its own specific features, models, and engineering aspects involved. From Fig-
ure 1.10 it will be obvious that only teamwork, bringing together several fields of
expertise, can lead to the final objective.

Polymerization Mechanisms

Polymerization reactions can be classified depending on the reaction mechanism

involved and can be either step-growth or chain-growth. These mechanisms differ
basically with the time scale of the process. In step-growth polymerization (like
polycondensation), the polymer chain growth proceeds slowly from monomer to
dimer, trimer, and so on, until the final polymer size is formed at high monomer
conversions. Both the chain lifetime and the polymerization time are often in the
order of hours. In chain-growth polymerization (like ionic or free-radical polymer-
ization), macromolecules grow to full size in a much shorter time (seconds being
the order of magnitude) than required for high monomer conversion. High molec-
ular weights are already obtained at low monomer conversion, which is in great
contrast to step-growth polymerizations. Also, unlike step-growth polymerization,
chain-growth polymerization requires the presence of an active center.
Condensation polymers are the result of a condensation reaction between mono-
mers, with or without the formation of a condensation by-product (Chapter 3). Ex-
amples of polymers produced by condensation are polyamide[6.6], (Nylon 6,6) the
result of the intermolecular condensation of hexamethylenediamine and adipic
acid, and polyamide[6], (Nylon 6) which is the product of intramolecular condensa-
tion of a-caprolactam. This type of reaction is generally sensitive to thermodynamic
equilibrium and requires the removal of the by-product, which is often volatile.
12 1 Polymer Reaction Engineering, an Integrated Approach

The polymers produced by condensation reactions can be either linear or non-

linear, depending on the number of functional groups per monomer. The polymer-
ization process can be performed in bulk (liquid or solid state) or as an interfacial
Free-radical polymerization (FRP) can be performed homogeneously (in bulk, so-
lution, or suspension; Chapters 4 and 5) or heterogeneously (emulsion, precipita-
tion; Chapter 6). The active site is always a radical that can be unstable (classical
FRP) or stabilized as in pseudo-living FRP. Radicals can be formed by the homo-
lytic bond rupture of initiators (molecules sensitive to homolytic cleavage, such as
peroxides, photosensitive molecules, or bisazo compounds) or by complex mecha-
nisms creating radicals from monomer units using thermal or high-energy
sources, such as X-rays, g-irradiation, or UV. This type of polymerization usually
comprises several steps: initiation, propagation, various transfer mechanisms,
and termination.
In ionic polymerizations a cation or anion is the active site (Chapter 7). A heter-
olytic process leads to charged parts of molecules that can induce the polymeriza-
tion by nucleophilic or electrophilic processes. These reactions generally evolve at
low temperatures (even as low as 120 C) due to the high reactivity of ions. Also,
they are very sensitive to impurities present in the monomer or solvent. These re-
actions are not always terminated, so lead to living polymerization. This process is
often used to build tailor-made copolymers.
Coordination polymerizations require a transition metal catalyst (Chapter 8). Poly-
olefins are often produced by this kind of reaction where the catalyst (Ziegler–
Natta, for example) acts as the active site but also as the steric regulator, which
makes it possible to build polymers with a defined tacticity. Nowadays a great re-
search effort is devoted to the synthesis of new transition metal-based catalysts,
such as metallocenes, to produce new products.

Fundamental and Engineering Sciences

Apart from the various polymerization mechanisms involved, a large number of

other disciplines will have to be involved, according to the matrix depicted in
Figure 1.10.
Thermodynamics is essential to understand the physicochemical properties of
the individual reactants, solvents, and products involved (Chapter 2). Also, it pro-
vides information on the interaction between the various components present in
the reaction mixture, from which phenomena such as phase behavior and parti-
tioning can be derived. This information is usually accessible by using the appro-
priate equation of state for a given system studied. A close collaboration between
chemical physicists, chemists, and chemical engineers is required to take full ad-
vantage of this fundamental knowledge.
Polymer solutions (solid, bulk, solution, complex media) have to be characterized
by several specific analytical tools (Chapter 20). Techniques such as NMR, ESR,
1.5 The Disciplines in Polymer Reaction Engineering 13

electron microscopy, chromatography, electrophoresis, viscometry, calorimetry, and

laser diffraction are widely used to determine polymer properties, often in combi-
nation. The main characteristics being analyzed are the chain length distribution,
degree of branching, composition, tacticity, morphology, particle size, and chemical
and mechanical properties. Polymer mechanics (Chapter 14) usually concerns the
final product rather than the polymerization reaction. Nevertheless, as polymers
are usually judged on their end-use properties (Chapter 13), the final product
needs specific and often customer-based analysis. This is described more specifi-
cally for two application areas, namely the use of thermosets for coating applica-
tions (Chapter 16) and the production of polymeric fibers (Chapter 17).
Measurement and control are indispensable to achievement of a robust and safe
process (Chapter 12). Since the early 1990s, a tremendous effort has been observed
in the development of new in-line analytical techniques, including spectroscopy
(UV, IR, Raman, laser, and so on), ultrasonic sensing, chromatography, and diffrac-
tion or electrical methods. New control schemes appear where the reaction is per-
formed just below the constraint limits, independently of the reaction kinetics. All
these techniques tend to lead to safer and more robust processes while increasing
productivity and product quality at the same time.
Safety cannot be treated as a separate discipline as it is already integrated from
the early chemistry and process development (Chapter 11). Safety deals with a wide
variety of technological aspects with respect to the environment (water, air, soil,
and living species). However, economic aspects are usually taken into consider-
ation also. Modern process development intrinsically includes safety and environ-
mental aspects in all stages of the development.
Modeling is probably the tool of excellence for engineers (Chapter 9). It is
used to simulate the reaction and the process system in order to shorten the time
for development. It is based on models that can be physical or chemical, semi-
empirical or empirical, descriptive or more fundamental. To describe the develop-
ment of the molecular weight distribution upon reaction, moment methods or
equations based on population balance are often used.
Scaleup is a widely used term to define the methodology that allows scaling up
of a process from small to larger scale (Chapter 10). Often the scaleup process be-
gins with a scaledown approach in order to have reliable and representative equip-
ment already at the laboratory scale. Scaleup is always dependent on the system
studied and requires a proper understanding of the performance of process equip-
ment involved at different scales. In polymer reaction engineering, heat transfer
and mixing can be considered as two major issues in this perspective. Modern
computing techniques such as computational fluid dynamics and process simula-
tion become more and more important in the optimization of process parameters
and the equipment hardware.
Volatile organic compound (VOC) content in the final product is related to prod-
uct properties and legislation (FDA approval in the USA, for example). All the pro-
cesses aiming to lower the residual VOC content in the product are denoted as ‘‘re-
movable’’ (Chapter 18). These processes can differ from each other, depending on
14 1 Polymer Reaction Engineering, an Integrated Approach

the techniques involved. Devolatilization, post-process reaction, and extraction are

some of the methodologies employed for this purpose.
Stability and degradation of polymers (Chapter 15) become relevant especially
during post processing or moulding processes. Temperature, oxidation and me-
chanic stresses are the main contributors to product degradation.
Currently, there is a strong emphasis on the synthesis of novel functional poly-
mers shaped on a nano scale (Chapter 19) and the development of sustainable
production processes (Chapter 21). The latter includes process intensification as a
methodology, the use of ‘‘green’’ solvents, and the use of renewable resources.
Many of the new processes under development are focusing on one or more of
these topics, for which the use of supercritical fluids is currently being imple-
mented on an industrial scale.

The Future: Product-inspired Polymer Reaction Engineering

Innovation times in industry have shown a steady decrease since the 1970s. Classic
thinking is that process development becomes increasingly important as industry
matures [23]. This is due to the fact that in an early phase of the lifetime of an in-
dustry, when product concepts are still being created, the rate of product innova-
tion exceeds the rate of process innovation. This period continues until a dominant
design has emerged and opportunities for radical product innovation decrease. In
this phase, the shift is toward process innovations to reduce cost price.
The half-time of product innovation (time-to-market) in the early 1970s was
about ten years. Currently, two years is often considered long. This acceleration of
innovation time is the result of competitive pressure in the market. As a rule of
thumb, the first company to enter the market with a new product can get up to
60% of market share, so there is a high reward for being first.
As has been discussed above, chemical engineering has been the basis for poly-
mer reaction engineering in the past. In recent discussions, however, it has been
emphasized that a need exists to refocus chemical engineering toward product-
driven process engineering [24, 25]. The thinking about a process should then start
with the customer or consumer: which of the two depends on the structure of the
supply chain. The wishes of the consumer and consumer-perceived product prop-
erties have to be translated into physical and chemical product properties. In this
way, the main physical attributes of a product are determined, including an idea
about the microstructure. Next, a functional analysis is performed to determine
the lowest number of transformations needed to create the product; this is fol-
lowed by a morphological analysis [26]. Finally, a conceptual process design exer-
cise is performed to generate possible process routes to achieve the desired product
properties. This sequence of events is the core of product-inspired polymer reac-
tion engineering. A key characteristic of this approach is the fact that it avoids the
classical ‘‘unit operation trap’’, because it does not fix the mindset to consider only
traditional reactor design and separation process steps to build a process.
References 15

Concluding Remarks

In the foregoing we have presented a general framework for sustainable polymer

reaction engineering. Its most important characteristic lies in the concerted multi-
disciplinary approach, rather than focusing on individual competencies. Given the
volume of polymer production, it will be of major importance that environmental
and safety issues become an integral part of the development process. In combina-
tion with tools such as life cycle analysis and product-inspired PRE, this will allow
the development of sustainable new polymer processes.


1 Association of Plastics Manufacturers Kivits, L. J. IJzendoorn, M. T.

in Europe, Annual report, 2002. Rispens, J. C. Hummelen, R. A. J.
2 H. Staudinger, Chem. Ber., 1920, 53, Janssen, Adv. Funct. Mater., 2002, 12,
1073. 665.
3 W. Kuhn, Berichte de Deutschen Chem. 15 W. Kaminsky, H. Sinn (Eds.) Transi-
Gesellsch., 1930, 63, 1503. tion metals and organometallics as
4 W. H. Chalmers, J. Am. Chem. Soc., catalysts for olefin polymerisation,
1934, 56, 912. Springer Verlag, Berlin, 1987.
5 H. Dostal, H. Mark, Trans. Faraday 16 J. M. Benedikt, B. L. Goodall,
Soc., 1936, 32, 54. Metallocene-catalyzed polymers, B.F.
6 G. V. Schulz, Z. Physik. Chem., 1935, Goodrich, Brecksville, 1998.
B30, 379. 17 J. Sinke, Appl. Comp. Mater., 2003, 10,
7 P. J. Flory, J. Am. Chem. Soc., 1936, 293.
58, 1877. 18 E. R. Howells, Chem. Ind., 1982, 508.
8 W. H. Carothers, US Patent 2 130 19 A. Penlidis, Can. J. Chem. Eng., 1994,
948, 1937. 72, 385.
9 E. W. Fawcett, R. O. Gibson, J. 20 A. Sapre, J. R. Katzer, Ind. Eng.
Chem. Soc., 1934, 386. Chem. Res., 1995, 34, 2202.
10 K. G. Denbigh, Trans. Faraday Soc., 21 A. Azapagic, Chem. Eng. J., 1999, 73,
1947, 43, 648. 1.
11 K. Ziegler, E. Holzkamp, H. Breil, 22 P. Eyerer, J. Polym. Eng., 1996, 15,
H. Martin, Angew. Chem., 1955, 67, 197.
541. 23 W. J. Abernathy, J. M. Utterback,
12 P. W. Morgan, S. L. Kwolek, Technol. Rev., 1978, 80, 40.
Macromolecules, 1975, 8, 104. 24 E. L. Cussler, G. D. Moggeridge,
13 H. Sasabe, T. Wada, Polymers for Chemical product design, Cambridge
electronic applications, in: Compre- University Press, Cambridge, 2001.
hensive Polymer Science, vol. 7, S. L. 25 E. L. Cussler, J. Wei, AIChE J., 2003,
Aggarwal (Ed.), Pergamon Press, 49, 1072.
Oxford, 1989. 26 C. J. King, AIChE Monograph Ser.,
14 J. K. J. van Duren, J. Loos, F. 1974, 70, 1.
Morrissey, C. M. Leewis, K. P. H.

Polymer Thermodynamics1

Theodoor W. de Loos


The phase behavior of polymer solutions plays an important role in polymer pro-
duction and processing. Many polymers are produced by solution polymerization.
Solvent choice, solvent recovery and the removal of traces of solvent from the poly-
mer product are important factors in these processes. An example is the produc-
tion of linear low-density polyethylene (LLDPE), which is a copolymer of ethylene
and a 1-alkene. Hydrocarbons are used as solvents in this process. The reactor con-
ditions are limited at high temperature by the onset of a liquid–liquid phase split,
characterized by a lower critical solution temperature, and at low temperature by
crystallization of LLDPE. The pressure must be high enough to keep the ethylene
in solution. Another well-known example is the production of low-density polyethy-
lene (LDPE). In this process ethylene is compressed together with an initiator and
the LDPE is formed by radical polymerization. The reactor pressure chosen must
be high enough to dissolve the polymer in its monomer. In practice reactor pres-
sures are higher than 200 MPa. Since the conversion of ethylene to LDPE is incom-
plete, LDPE has to be separated from unreacted ethylene, which is recycled to the
reactor. To save energy this is done by pressure reduction in two steps, which in-
volve a high-pressure vapor–liquid flash.
From a thermodynamic point of view polymer solutions are complicated solu-
tions. A polymer is not a single component but a multicomponent mixture charac-
terized by a molecular weight distribution or by average molecular weights, such as
the number-average molecular weight Mn or the weight-average molecular weight
Mw . In the case of a copolymer different types of copolymers are possible, for ex-
ample random copolymers and block copolymers, and the comonomer content
may vary. Because of their asymmetric nature the entropy of mixing of polymer
solutions is much lower than in the case of a mixture of two low molecular weight
compounds; also, the pure solvent and the polymer have rather different free

1) The symbols used in this chapter are listed at

the end of the text, under ‘‘Notation’’.

Handbook of Polymer Reaction Engineering. Edited by T. Meyer, J. Keurentjes

Copyright 8 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
ISBN: 3-527-31014-2
18 2 Polymer Thermodynamics

volumes. Because of this, polymer–solvent systems often show a liquid–liquid

phase split.
In this chapter it is not possible to give an in-depth treatment of the thermody-
namics of polymer solutions. For further reading, see Refs. 1–6.

Thermodynamics and Phase Behavior of Polymer Solutions

Thermodynamic Principles of Phase Equilibria

The equilibrium conditions for phase equilibria can be derived in the simplest way
using the Gibbs energy G. According to the second law of thermodynamics, the
total Gibbs energy of a closed system at constant temperature and pressure is a
minimum at equilibrium. If this condition is combined with the condition that
the total number of moles of component i is constant in a closed system [Eq. (1),
where nai is number of moles of component i in phase a], the equilibrium condi-
tions given by Eq. (2) can easily be derived for a system of p phases and N compo-
nents [7].

nai ¼ constant ð1Þ

mai ¼ mib ¼    ¼ mpi for i ¼ 1; 2; . . . ; N ð2Þ

The chemical potential of component i in phase a is defined by Eq. (3), where g is

the molar Gibbs energy.
0 X 1
q nai g a
@ i A
mai ¼ ð3Þ
P; T; nj0i

Fugacity and Activity

In the thermodynamic treatment of phase equilibria, auxiliary thermodynamic

functions such as the fugacity coefficient and the activity coefficient are often
used. These functions are
C closely related to the Gibbs energy. The fugacity of com-
ponent i in a mixture, fi , is defined by Eq. (4a) together with (4b).
dmi 1 RT ln fi at constant T ð4aÞ
lim ¼ 1 ð4bÞ
P!0 Pi
2.2 Thermodynamics and Phase Behavior of Polymer Solutions 19
According to this definition, fi is equal
C to the partial pressure Pi in the case of an
ideal gas. The fugacity coefficient fi is defined by Eq. (5) and is a measure of the
deviation from ideal gas behavior.
C fi
fi ¼ ð5Þ

The fugacity coefficient can be calculated from an equation of state by Eq. (6) or (7)
C ð P RT

RT ln fi ¼ Vi  dP ð6Þ
0 P
0X 1
ð "  # n i RT
qP RT @ A
RT ln fi ¼   þ RT ln i ð7Þ
y qn i V; T; nj0i V PV

According to Eq. (4), the equilibrium relation in Eq. (2) can be replaced by Eq. (8).
fi a ¼ fi b ¼    ¼ fi p for i ¼ 1; 2; . . . ; N ð8Þ
The activity a i is defined as the ratio of fi and the fugacity of component i in the
standard state fi 0 at the same P and T [Eq. (9)].
fi ðP; T; xÞ
ai 1 0 ð9Þ
fi ðP; T; x 0 Þ

An ideal solution is defined by Eq. (10).

a iid 1 x i ð10Þ

The activity coefficient of component i, gi [Eq. (11)], is a measure of the deviation

from ideal solution behavior, so the fugacity of a nonideal liquid solution can be
written as Eq. (12).

gi 1 ð11Þ
a iid
fi ¼ x i gi fi 0 ð12Þ

The activity coefficient gi can be calculated from a model for the molar excess
Gibbs energy g E, Eq. (13).
0 X 1
B q n i g C
@ i A
RT ln gi ¼ ð13Þ
qn i P; T; nj0i
20 2 Polymer Thermodynamics

In this approach the standard state fugacity fi 0 of a liquid component is usually the
fugacity of the pure liquid component, and is closely related to the vapor pressure
Pisat of that component. On the vapor-pressure curve of a pure component, we have
conditions according to Eqs. (14).

fi L ðPisat ; TÞ ¼ fi V ðPisat ; TÞ ¼ fVi ðPisat ; TÞPisat ð14Þ

From Eq. (4a) we obtain Eq. (15), where viL is the molar volume of pure liquid i.
ðP ! ðP !
dmiL viL
fi L ðP; TÞ ¼ fi L ðPisat ; TÞ exp ¼ fi L ðPisat ; TÞ exp dP ð15Þ
Pisat RT Pisat RT

Combining Eqs. (14) and (15), we get Eq. (16).

ðP !
L viL
fi ðP; TÞ ¼ fVi ðPisat ; TÞPisat exp dP ð16Þ
Pi sat RT

At low pressure the fugacity coefficient and the exponential term are close to 1, so
Eq. (17) holds.

fi L A Pisat ð17Þ

Equilibrium Conditions

For low-pressure vapor–liquid equilibria (VLE) the equilibrium condition of Eq.

(18) is usually written as Eq. (19), in which fi 0 is calculated with Eq. (16) or (17)
and ji with Eq. (6) or (7); x i and yi are the liquid-phase and vapor-phase mole frac-
tion of component i, respectively.
fi L ¼ fi V ð18Þ
0 C
x i gi fi ¼ yi ji P ð19Þ

The truncated virial equation or the ideal gas equation is often used as the equa-
tion of state. In the latter case all fugacity coefficients are 1, so the simplest form
of Eq. (19) is Eq. (20).

x i gi Pisat ¼ yi P ð20Þ

In the case of low-pressure liquid–liquid equilibria (LLE) the equilibrium condition

is given by Eq. (21).
fi 0 ¼ fi 00 or ðx i gi fi 0 Þ 0 ¼ ðx i gi fi 0 Þ 00 ð21Þ
2.2 Thermodynamics and Phase Behavior of Polymer Solutions 21

The two liquid phases are indicated by prime ( 0 ) and double prime ( 00 ). If for both
liquid phases the standard fugacity is chosen as the pure liquid component, Eq.
(21) reduces to Eq. (22).

ðx i gi Þ 0 ¼ ðx i gi Þ 00 ð22Þ

For the calculation of high-pressure vapor–liquid equilibria or liquid–liquid equi-

libria an equation of state is always used for both phases and the equilibrium con-
dition used is given by Eq. (23).
fi a ¼ fi b or ðx i ji PÞa ¼ ðx i ji PÞb or ðx i ji Þa ¼ ðx i ji Þb ð23Þ

Low-pressure Vapor–Liquid Equilibria

Since polymers have no vapor pressure and as a consequence the vapor phase does
not contain polymer, the equilibrium conditions for low-pressure vapor–liquid
equilibria of polymer solutions as given by Eq. (20) are only applicable to the
solvent s as in Eq. (24), or in a case where the weight fraction of polymer wp is
used as a composition variable as in Eq. (25), where Ws is the weight fraction based
activity coefficient of the solvent.

P ¼ xs gs Pssat ð24Þ
P ¼ ð1  wp ÞWs Pssat ð25Þ

The relation between a mole fraction based activity coefficient gi and a weight frac-
tion based activity coefficient Wi of component i is given by Eq. (26), where wi and
Mi are the weight fraction and molecular weight of component i, respectively. gs
and Ws can be obtained from a correlation of experimental data using a suitable
model or from a predictive model (see Section 2.3).

gi wi XN
¼ ¼ Mi ð26Þ
Wi x i j¼1
M j

Flory–Huggins Theory and Liquid–Liquid Equilibria

The Flory–Huggins theory of polymer solutions [1] is based on a rigid lattice

model in which a polymer molecule is assumed to consist of r segments of the
size of a solvent molecule. The Flory–Huggins expression for the Gibbs energy of
mixing Dmix G for Ns moles of solvent and Np moles of polymer is given by Eq. (27).

Dmix G
¼ Ns ln js þ Np ln jp þ js jp ðNs þ rNp Þw ð27Þ
22 2 Polymer Thermodynamics

The first two terms on the right-hand side of Eq. (27) represent the so-called com-
binatorial entropy of mixing, and the third term is an approximation for the en-
thalpy of mixing. In practice w, the Flory–Huggins interaction parameter, is used
as an adjustable, temperature-dependent parameter. js and jp are the segment
fractions of solvent and polymer, respectively, defined by Eq. (28).

Ns rNp
js ¼ and jp ¼ ð28Þ
Ns þ rNp Ns þ rNp

According to the original Flory–Huggins theory w is given by Eq. (29), in which

ess ; epp , and esp are the interaction energies between two solvent molecules, two
polymer segments, and a solvent molecule and a polymer segment, respectively.

ð2esp  ess  epp Þ

wz ð29Þ

Often r is approximated by the ratio of the molar volumes of pure liquid polymer
and pure solvent. The segment fractions of solvent and polymer are then equal to
the volume fractions of solvent and polymer, fs and fp . Equation (27) is derived us-
ing many assumptions and approximations (for a discussion, see Ref. 8), but on
the basis of this rather simple expression many features of the phase behavior of
polymer solutions can be explained. The expressions for the mole fraction based
activity coefficients of solvent and polymer are Eqs. (30) and (31), respectively.

jp 1
ln gs ¼ ln ð1  jp Þ þ þ 1  jp þ jp2 w ð30Þ
r r

ln gp ¼ ln½ð1  jp Þr þ jp   ðr  1Þð1  jp Þ þ rð1  jp Þ 2 w ð31Þ

As can be seen from Eq. (30), the activity coefficient of the solvent is strongly de-
pendent on r for low values of r, but at high values of r gs becomes practically inde-
pendent of r. This implies that the equilibrium pressure of low-pressure VLE for
polymer–solvent systems is hardly dependent on the molecular weight of the
In the original formulation of the Flory–Huggins theory Dmix G increases with
decreasing temperature, which leads to a liquid–liquid phase split with an upper
critical solution temperature. Per mole of lattice sites, the entropy of mixing of a
polymer–solvent system is much less than for a solvent–solvent system; because
of this, polymer–solvent systems show a stronger tendency to demix than solvent–
solvent systems. This is illustrated in Figure 2.1, which shows schematically the
influence of r on the location and shape of the liquid–liquid two-phase region.
The figure shows that with increasing r, the two-phase region increases in size
and becomes more asymmetric. The critical point, the temperature maximum of
2.2 Thermodynamics and Phase Behavior of Polymer Solutions 23








0.00 0.10 0.20 0.30 0.40 0.50 0.60

Fig. 2.1. Liquid–liquid equilibria for polymer–solvent systems
at different polymer chain lengths r. Calculated from the Flory–
Huggins theory.

the two-phase region, shifts with increasing r to lower values of w, higher tempera-
ture, and lower values of jp . At r ¼ y the critical point is found at a limiting value
of w ¼ 12 , T ¼ y, and jp ¼ 0. This follows directly from the critical point conditions
[Eq. (32), where G is the Gibbs energy per mole of lattice sites].
! !
qG q2G
¼ ¼0 ð32Þ
qjp qjp2
P; T

This leads to Eqs. (33) and (34) for the Flory–Huggins expression [Eq. (27)].
crit 1 1 2
w ¼ 1 þ pffiffi ð33Þ
2 r

jpcrit ¼ pffiffi ð34Þ
1þ r
24 2 Polymer Thermodynamics

The theta temperature y is the highest possible upper critical solution temperature
(UCST) within the framework of the Flory–Huggins theory.
Many polymer–solvent systems, at temperatures above the UCST, show a second
region with liquid–liquid immiscibilty (Figure 2.2), which is characterized by the
occurrence of a lower critical solution temperature (LCST) [9]. This phenomenon
can not be explained on the basis of the original Flory–Huggins theory. Delmas
and Patterson [10] calculated binary critical curves of polymer–solvent systems us-
ing the Flory equation of state [11, 12]. According to these authors the reason for
the occurrence of this LCST is the large difference in thermal expansion of pure
liquid polymer and pure solvent, which leads to an increasing difference in free
volume with increasing temperature and an LCST-type liquid–liquid phase split at
temperatures below the critical point of the solvent. If the molecular weight of the
polymer is increased, the UCST and LCST approach each other. In some cases, this
can lead to the phase diagram of Figure 2.3, in which the liquid–liquid region has
the shape of an hourglass, as in the polystyrene–acetone system (Figure 2.4) [13].
However, another possibility is that the UCST and LCST never meet and that each
liquid–liquid region shows a (different) y temperature for r ¼ y. This type of be-
havior is found for polystyrene–methylcyclohexane [14]; see Figure 2.5.
Another type of LCST is connected to specific interactions like hydrogen bond-
ing. This type of LCST is found at temperatures below the UCST. The resulting
phase diagram is presented in Figure 2.6, which shows a closed-loop region of
liquid–liquid immiscibility. This type of phase diagram is not predicted by the
original Flory–Huggins theory, either.



Solvent φp Polymer
Fig. 2.2.Liquid–liquid equilibria showing a low-temperature
two-phase region with a UCST and a high-temperature two-
phase region with a LCST.
2.2 Thermodynamics and Phase Behavior of Polymer Solutions 25




Solvent φp Polymer
Fig. 2.3. Hourglass-shaped two-phase liquid–liquid equilibria.

High-pressure Liquid–Liquid and Vapor–Liquid Equilibria

Figure 2.7 shows schematically the phase behavior of a binary polymer solution at
constant composition in a P; T diagram. In the case of curve a, the mixture at low
temperature shows a liquid–liquid region characterized by UCST behavior and at
high temperature a liquid–liquid region characterized by LCST behavior. Both two-
phase regions are bounded at low pressures by a three-phase liquid–liquid–vapor
curve, which separates the liquid–liquid region from a vapor–liquid region. The
curve that separates a liquid–liquid region from the one-phase liquid region is
often called a cloud-point curve. The liquid–liquid–vapor curve is found in many
cases very close to the vapor-pressure curve of the pure solvent because the poly-
mer concentration in the vapor phase and in the solvent-rich liquid phase is low.
Curve b shows the case where the low-temperature liquid–liquid region and the
high-temperature liquid–liquid region are merged into a single liquid–liquid re-
gion. In this case the cloud-point curve shows a pressure minimum. At pressures
below this minimum, hourglass-shaped two-phase regions are found. At higher
pressures the phase behavior is as represented by Figure 2.2. An example of this
behavior is shown by the polystyrene–acetone system [15] (see Figure 2.8). In the
case of curve c the cloud point curve is found at much higher pressure and no
longer shows a minimum.
In going from case a to case c, the mutual solubility of polymer and solvent de-
creases. Stronger polymer–polymer or solvent–solvent interactions lead to a lower
mutual solubility, while stronger polymer–solvent interactions lead to a higher mu-
tual solubility [15]. As discussed in Section 2.2.5, increasing molecular weight of
26 2 Polymer Thermodynamics

Fig. 2.4.Liquid–liquid equilibria in the polystyrene–acetone

system at the indicated polystyrene molecular weights
(in g mol1 ). Reproduced with permission from Ref. 13.

the polymer also leads to a decrease in mutual solubility for the same polymer–
solvent system. An example of this effect is given by Zeman and Patterson [16] for
the polystyrene–acetone system.
A similar effect is observed when the chain length of the solvent is increased.
Ehrlich and Kurpen [17] determined cloud-point curves of 5 wt.% LDPE in ethane,
propane, butane, and pentane. These results are reproduced in Figure 2.9, which
shows that with increasing molecular weight of the solvent the cloud-point pres-
sure increases and dP=dT of the cloud-point curves change sign from positive to
negative. Since ethylene is a worse solvent than ethane, the cloud point pres-
sures for the system ethylene þ LDPE [18, 19] are higher than for the system
2.2 Thermodynamics and Phase Behavior of Polymer Solutions 27

Fig. 2.5. Liquid–liquid equilibria in the polystyrene–

methylcyclohexane system at the indicated molecular weights
of polystyrene (in g mol1 ). Reproduced with permission from
Ref. 14.

ethane þ LDPE. For this type of systems it is not possible to make a clear dis-
tinction between vapor–liquid equilibria and liquid–liquid equilibria. The poly-
mer-rich phase can be considered to be a liquid phase, but the solvent-rich phase
is liquidlike at low temperature, but vaporlike at high temperature. De Loos et al.
[20] measured cloud-point curves for LLDPE systems plus hexane, plus heptane,
and plus octane (see Figure 2.10). In this case also, the polymer is more soluble
in the higher molecular weight solvent, which is shown by the shift of the LCST-
type cloud-point curve to higher temperatures. An increased degree of branching
of the polymer leads to a better mutual solubility [21].
28 2 Polymer Thermodynamics


φp Polymer
Fig. 2.6. Closed-loop liquid–liquid equilibria.

P c


Fig. 2.7.P; T phase diagram for a constant cloud pressure; (c) system in which the low-
composition polymer–solvent system: (a) and high-temperature regions of demixing have
system with separate low- and high- merged, without a minimum cloud pressure.
temperature regions of demixing; (b) system in The full curves are cloud-point curves; the
which the low- and high temperature regions broken curves are liquid–liquid vapor curves.
of demixing have merged, showing a minimum
2.3 Activity Coefficient Models 29

Fig. 2.8. Liquid–liquid equilibria in the polystyrene–acetone

system at the indicated pressures. M ¼ 20400 g mol1 .
Reproduced with permission from Ref. 16.

The addition of an anti-solvent such as a supercritical gas also produces changes

in the phase diagram to one like Figure 2.7 [22–25]. However, in this case the
liquid–liquid–vapor equilibrium is represented by a region instead of by a curve.

Activity Coefficient Models

In practice the original Flory–Huggins theory is not accurate enough for a quanti-
tative representation of polymer–solvent phase equilibria. To improve this situation
a concentration-dependent w-parameter can be introduced and also the tempera-
ture dependence of w can be made more complicated than in Eq. (29). The terms
representing the combinatorial entropy of mixing, the first two terms in Eq. (27),
can be replaced by results from more accurate theories [4, 26] and the difference
30 2 Polymer Thermodynamics

Fig. 2.9. Cloud-point isopleths of LDPE in various n-alkane

solvent at 5 wt.% polymer. The short dash–long dash curve is
the solidification boundary of LDPE. Reproduced with
permission from Ref. 17.

in the free volume of polymer and solvent can be accounted for by adding a free
volume contribution [4]. Further, Eq. (27) can be extended to account for the poly-
dispersity of the polymer.
The system-dependent parameters in these models can be adjusted to experi-
mental data or predicted from a group contribution approach [4, 26].

Flory–Huggins Theory

For a solution of a polydisperse polymer with m polymer components in one sol-

vent, the Flory–Huggins expression for the Gibbs energy of mixing per mole of lat-
2.3 Activity Coefficient Models 31

Fig. 2.10. Cloud-point curves of poly(E-co-1-octene)–n-alkane

systems at P ¼ 3 MPa. Mn ¼ 33 kg mol1 , Mw ¼ 124 kg mol1 ,
Mz ¼ 420 kg mol1 . The systems show LCST phase behavior.
Reproduced with permission from Ref. 20.

tice sites can be written as Eq. (35). The sum is only over the polymer components,
ji is the segment fraction of polymer component i, and jp ¼ ji ¼ ð1  js Þ is the
overall polymer segment fraction.

Dmix G Xm
¼ js ln js þ ji ri1 ln ji þ js jp g sp ð35Þ
RT i¼1

g sp ¼ f ðT; jp Þ ð36Þ

To avoid confusion with the polymer–solvent interaction parameter, the symbol g sp

[Eq. (36)] is used instead of w, which is independent of concentration. Equations
(37a)–(37c) are examples of expressions used for the temperature and composition
dependence of g sp [27–29].

g sp ¼ a þ þ cT þ djp þ ejp2 ð37aÞ

g sp ¼ a þ T ð37bÞ
1  djp
32 2 Polymer Thermodynamics

aþ þ c ln T
g sp ¼ T ð37cÞ
1  djp

We can derive Eq. (38) for the solvent, and Eq. (39) for polymer component i,
where rn is number-average chain length of the polymer, which is defined by Eq.
(40) and which is proportional to the number-average molecular weight of the
1 qg sp 2
lnðxs gs Þ ¼ ln js þ 1  j þ g sp  js j ð38Þ
rn p qjp p
ri jp qg sp 2
lnðx i gi Þ ¼ ln ji þ 1   ri js þ ri g sp  jp j ð39Þ
rn qjs i

n i ri
rn ¼ ð40Þ

Using Eqs. (37) and (38), vapor–liquid and liquid–liquid equilibria can be calcu-
lated as discussed above. Reference 5 gives details of liquid–liquid equilibrium cal-
culations in polydisperse polymer–solvent systems. In Figure 2.11 the experimen-
tally determined phase behavior of three PEG–water systems is compared with the
calculated phase behavior of these systems using Eq. (37c) to represent g sp as a
function of temperature and polymer segment fraction. The parameters were fitted
to the data [29].
The description of the influence of pressure on polymer–solvent phase behavior
is also possible using a pressure-dependent g sp [27, 30–32]. However, this approach
is purely empirical.

Hansen Solubility Parameters

According to the regular solution theory of Hildebrand the w-parameter can be ap-
proximated by Eq. (41) [8], where vs is the molar volume of the solvent and ds and
dp are the solubility parameters of solvent and polymer, respectively. Since these
solubility parameters are pure component parameters, Eq. (41) combined with Eq.
(27) results in a predictive model. However, since many simplifications are in-
volved, the results of this model can be considered as only a rough estimate. Fol-
lowing the slogan ‘‘like dissolves like’’, a good solvent for a polymer is a solvent for
which ds and dp have similar values.

w¼ ðds  dp Þ 2 ð41Þ
2.3 Activity Coefficient Models 33

Fig. 2.11. Closed-loop liquid–liquid equilibria in the PEG–

water system. Symbols: experimental data, M ¼ 3:35 kg mol1
(0), M ¼ 8 kg/mol (5), M ¼ 15 kg/mol (4); curves fitted using
Eq. (37c). Reproduced with permission from Ref. 29.

Hansen suggested refining the solubility parameter theory by the introduction of

contributions from dispersive interactions (d), polar interactions ( p) and hydrogen
bond formation (hb), as in Eq. (42) [33].

w¼ ½ðds; d  dp; d Þ 2 þ ðds; p  dp; p Þ 2 þ ðds; hb  dp; hb Þ 2  ð42Þ

Recently, Lindvig et al. [34, 35] showed that Eq. (42) systematically overestimates
the infinite dilution activity coefficient of the solvent and proposed an alternative
expression, Eq. (43).

w¼a ½ðds; d  dp; d Þ 2 þ 0:25ðds; p  dp; p Þ 2 þ 0:25ðds; hb  dp; hb Þ 2  ð43Þ
34 2 Polymer Thermodynamics

These authors showed that for a number of polymer–solvent system with a ¼ 0:6
this method performs similarly to group contribution methods using volume frac-
tions to represent the segment fractions in the Flory–Huggins model. Values of
solubility parameters are tabulated by Barton [36].

Correlation of Solvent Activities

The combination of Eq. (35) with one of the Eqs. (36) or similar empirical equa-
tions is not very successful for describing the phase behavior of polymer–solvent
systems with strong interactive species and of systems which only differ in mo-
lecular weight. Since the mid-1990s activity coefficient models have been proposed
based on a combination of the Flory–Huggins type of expression for the combina-
torial entropy of mixing and segment-based local composition models to account
for the contribution from energetic interactions (the residual contribution to the
activity coefficient).
In 1993 Chen [37] proposed a correlative model that used a combination of the
Flory–Huggins expression for the combinatorial entropy of mixing and the non-
random two-liquid (NRTL) theory [38]. The same approach was followed by Wu
et al. [39]. These authors used Freed’s expression [41, 42] for the combinatorial
entropy of mixing, which is more accurate than the Flory–Huggins expression, in
combination with the NRTL theory. The nonrandom factor model of Haghtalab
and Vera [42] was modified by Zafarini-Moattar and Sadeghi [43] for polymer solu-
tions. In 2004, Pedrosa et al. [44] suggested use of the UNIQUAC theory [45] in-
stead of the NRTL theory and tested various combinations of combinatorial con-
tributions and residual contributions using a database of 70 low-pressure VLE
systems. These authors have concluded that the combination of a segment-based
NRTL or UNIQUAC residual term with a good combinatorial term is able to pro-
duce the best correlations of experimental VLE data and can also be used to predict
the influence of the molecular weight of the polymer on polymer–solvent VLE.
As an example we will give here the expression for the activity coefficient of the
solvent for a model which combines the p free-volume combinatorial term with a
segment-based Wu-NRTL residual term [44]. The p free-volume combinatorial
term combines combinatorial contributions and free-volume contributions [45].
The activity coefficient of the solvent is given by Eq. (44).

ln gs ¼ ln gscombfv þ ln gsres ð44Þ

The combinatorial/free-volume term is given by Eq. (45).

fsfv f fv
ln gscombfv ¼ ln þ1 s ð45Þ
xs xs

The free-volume fraction fsfv is calculated from Eq. (46).

2.3 Activity Coefficient Models 35

fv xiv
fi ¼ X i fv ð46Þ
xj vi

The free volume of a component is defined by Eq. (47), where vi is the molar
volume of liquid component i, vi; vdW is the hard-core volume or van der Waals
volume of this component and can be calculated using the Bondi tables [46], and
p is a correction factor, which is calculated from Eq. (48) in the p free-volume
model [45].

vi ¼ ðvi  vi; vdW Þ p ð47Þ

p¼1 ð48Þ

The residual term is given by Eq. (49) together with Eqs. (50), where asp and a ps are
adjustable interaction parameters, a is the NRTL-nonrandomness parameter,
which was fixed by Pedrosa et al. at 0.4.
" #
tps Gps tsp Gsp2
ln gsres ¼ qs X p2 þ ð49Þ
ðXs þ X p Gps Þ 2 ðX p þ Xs Gsp Þ 2
a ij
tij ¼ exp and Gij ¼ expðatij Þ ð50Þ

X i is the effective mole fraction of segments of species i and qi is the effective seg-
ment number of species i, which are given by Eqs. (51) and (52); rs ¼ 1 and rp ¼ r.

x i qi
Xi ¼ X ð51Þ
xj q j
qi ¼ ri 1  2a 1  ð52Þ

Group Contribution Models

Thermodynamic properties can be predicted from group contribution methods.

In these models molecules are divided in functional groups. Group contribution
models for activity coefficients consider the interactions between functional groups
rather than between molecules. Since the number of functional groups is much
lower than the number of possible molecules composed of these groups, only a
limited number of group interaction parameters have to be known to describe a
large number of systems. These group interactions are obtained from regression
36 2 Polymer Thermodynamics

of experimental data. This makes the group contribution methods purely predic-
tive. However, since details of molecular structure are not considered, group contri-
bution methods for activity coefficients are in general less accurate than correlative
Oishi and Prausnitz [47] proposed writing the solvent activity coefficient for
polymer–solvent systems as the sum of three terms [Eq. (53)].

ln gs ¼ ln gscomb þ ln gsfv þ ln gsres ð53Þ

The combinatorial contribution gscomb accounts for differences in size and shape
of the molecules. The free-volume contribution gs accounts for changes in free
volume due to mixing, caused by the large difference between the free volumes of
pure solvent and polymer. For ordinary liquid mixtures in far from critical condi-
tions, this term is usually negligible. The residual contribution gsres accounts for
energy interactions. In the approach of Oishi and Prausnitz, the combinatorial con-
tribution is represented by the Staverman–Guggenheim expression, a modification
of the Flory–Huggins equation, also used in the UNIFAC group contribution
model [48]; for the residual contribution also, the corresponding expression of the
UNIFAC model is used. The free-volume contribution is calculated from the Flory
equation of state [49]. This group contribution model and those of Chen et al. [50]
and Danner and High [51] are discussed in Ref. 4. Here we will discuss the en-
tropic free volume model of Elbro et al. [52] and Kontogeorgis et al. [53], and the
group contribution Flory (GC–Flory) model of Bogdanic et al. [54, 55], which is a
modification of the model of Chen et al. [50].
In the entropic-free volume model, the activity coefficient of the solvent is given
by Eqs. (44)–(48) with p ¼ 1 [52]. The residual contribution is represented by the
residual contribution of the UNIFAC model with temperature-dependent interac-
tion parameters [53]. The liquid molar volumes needed for the calculation of the
free volume of a component can be taken from experiment or calculated from the
Tait equation [4] or by the group contribution method of Elbro et al. [56]. This
model is relatively easy to use.
In the GC–Flory model, the solvent activity coefficient is calculated from Eq.
The combinatorial term is calculated by means of the original Flory–Huggins ex-
pression, Eq. (54).

fi f
ln gicomb ¼ ln þ1 i ð54Þ
xi xi

The free-volume and residual terms are calculated from a modification of the orig-
inal Flory equation of state, Eq. (55), where ~v is the reduced volume, defined by Eq.

RT v~1/3 þ C E attr
P¼  ð55Þ
v v~1/3  1 v
2.3 Activity Coefficient Models 37

v~ ¼ ð56Þ

The molar hard-core v  is calculated from the pure component hard-core molar
volumes vi using a linear mixing rule [Eq. (57)]. The same type of mixing rule is
used for the number of external degrees of freedom parameter C [Eq. (58)].
v ¼ x i vi ð57Þ

C¼ x i Ci ð58Þ

The pure component hard-core volumes and C parameters are calculated from the
group contribution expressions in Eqs. (59) and (60),

vi ¼ 21:9662 uðiÞ
m Rm ð59Þ

X    X
1 1 R
Ci ¼ uðiÞ
n CT0 ; n þ CT; n  þ X n Cn0 ð60Þ
T T0 n Rm

where un is the number of groups of type n in molecule i and T0 is a reference
temperature. R n is the normalized van der Waals volume of group n as used in
the UNIFAC method. The attraction term E attr is related to the UNIFAC model
by Eq. (61), z is the lattice coordination number, chosen to be equal to 10, and qi ,
the surface area of molecule i, is given by Eq. (62)
2 X 3
yj expðDeji /RTÞDeji
X1 6
6 j 7
E attr ¼ zqi x i 6
4eii þ
X 7
2 yk expðDeki /RTÞ 5
v~  1 X X ðiÞ 
 3R ln xi un CT; n ð61Þ
v~ i n

qi ¼ vðiÞ
n Qn ð62Þ

Q n is the normalized van der Waals surface of group n, as in UNIFAC. The inter-
action energy parameters eji and Deji are given by Eqs. (63), in which eij0 is calcu-
lated from a group contribution expression [Eq. (64), together with Eq. (65)].

eij ¼ and Deij ¼ eij  eii ð63Þ
38 2 Polymer Thermodynamics
eji0 ¼ yðiÞ
m yðn jÞ enm ð64Þ
m n

enm ¼ ðenn emm Þ 1/2 þ Denm ð65Þ

In these expressions the volume fraction fi of molecule i, the segment fraction yi

of molecule i, and the segment fraction yðiÞ
n of n in molecule i are defined by Eqs.

fi ¼ X i  ð66Þ
xj vj

x i qi
yi ¼ X ð67Þ
xj q j

u Q
n ¼
Xn n ð68Þ
m Qm

Note that indices m and n refer to groups m and n, and i and j to molecules i
and j.
The resulting expressions for the free-volume contribution and the residual con-
tribution in the activity coefficient are Eqs. (69) and (70).

fv v~i  1 v~i
ln gi ¼ 3ð1 þ Ci Þ ln 1/3  Ci ln ð69Þ
v~  1 v~
1 4 1 X
ln gires ¼ zqi ½eii ð~ vi Þ þ 1  ln
vÞ  eii ð~ yj expðDeji /RTÞ
2 RT j
X yj expðDeji /RTÞ
 X 5 ð70Þ
j yk expðDe ki /RTÞ

To apply this method, the liquid molar volumes of the mixture and of the pure
components need to be known. At a given pressure and temperature these values
can be calculated from the equation of state. However, since eji according to Eq.
(63) is volume-dependent, this involves an iterative procedure similar to that
described by Danner and High [4] for the method of Chen et al. [50]. Figure
2.12 shows the experimental solvent activity of the system poly(propylene oxide)–
benzene at 347.85 K compared to the correlation by UNIQUAC and predictions by
the GC–Flory model. The result of the correlation is almost perfect. The predicted
solvent activities by the GC–Flory model are also very close to the experimental
values. In Figure 2.13 a comparison is shown of experimental solvent activity
2.4 Equation of State Models 39

Fig. 2.12. Activity of benzene in the poly(propylene oxide)–

benzene system at 347.85 K. Mn ¼ 500 kg mol1 . Symbols:
experimental data; curves: UNIQUAC correlation and GC–Flory
prediction. Reproduced with permission from Ref. 54.

coefficients at infinite dilution and predictions by the GC–Flory model [54] for
homopolymer–solvent systems, demonstrating that the results of predictions are
good. The predictions for copolymer solutions are slightly worse [55].

Equation of State Models

High-pressure phase equilibria in systems of polymers, solvents, and supercritical

gases are in almost all cases modeled using equations of state. A review of equa-
tions of state for polymer systems, including a discussion of their theoretical back-
ground, has been given by Lambert et al. [6]. One of the first equations of state that
was used to model the high-pressure phase behavior of polymer–solvent systems
was the Flory equation of state [11, 12]. Patterson and Delmas [10] showed that
this equation of state can be used to describe both LCST and UCST phase behavior.
The perturbed hard-chain theory (PCHT) was developed by Prausnitz and co-
workers [57–59]. It can be considered as an improvement of the approach of Flory
40 2 Polymer Thermodynamics

log Ω∞cal

log Ω∞exp
Fig. 2.13. Comparison of calculated activity coefficients at
infinite dilution using the GC–Flory model with experimental
values for many polymer–solvent systems. Reproduced with
permission from Ref. 54.

and co-workers and can be used to model the phase behavior of mixtures of small
and large molecules, including polymers, over a wide range of pressure and tem-
perature [57–60]. Here, we will discuss only the two equations of state methods:
the relatively simple Sanchez–Lacombe (S–L) lattice fluid model [61, 62], and the
statistical associating-fluid theory (SAFT) [63, 64], which has now become one of
the standard equations of state for polymer solutions.

The Sanchez–Lacombe Lattice Fluid Theory

Like the Flory–Huggins model, the Sanchez–Lacombe lattice fluid theory is based
on the assumption that segments of solvent molecules and polymer molecules oc-
cupy the lattice sites of a rigid lattice, but vacant lattice sites are also allowed. The
number of vacant lattice sites, and as a consequence the total number of lattice
sites, are pressure-dependent, and in this way compressibility is introduced.
2.4 Equation of State Models 41

The resulting equation of state for a pure component is given by Eq. (71).

p~v~ 1 1 1
¼  1 þ v~ ln 1   ð71Þ
T r v~ v~T~

The reduced volume v~, the reduced pressure p~, and the reduced temperature T~ are
defined by Eqs. (72a)–(72c).

v n0 þ rn
v~ ¼ ¼ ð72aÞ
v rn

P Pv 
p~ ¼ 
¼  ð72bÞ
P e

T~ ¼  ð72cÞ

The parameters with an asterisk ( ) are the so-called characteristic parameters. In

practice r, the number of segments per molecule, e  , the interaction energy param-
eter, and v  , the molar volume of a lattice site, are used as the independent pure-
component parameters; n is the number of moles of the component and n0 is the
number of moles of vacant lattice sites. If v is the volume per mole of segments,
the total volume of the system is given by V ¼ nrv. For high molecular weights r
is large, so it can be concluded from Eq. (71) that the density of polymer melts is
not very dependent on molecular weight and that the PVT behavior of polymer
melts follows the corresponding states principle (see Figure 2.14) [62].
For mixtures the same equation of state is used, but the characteristic parame-
ters r; e  , and v  are composition-dependent. Neau [65] gives an overview of differ-
ent mixing rules proposed in the literature. Often-used mixing rules are given in
Eqs. (73)–(75). In Eq. (75) k ij is an adjustable binary interaction parameter which
equals zero for i ¼ j and which can fitted to binary experimental data.

r¼ x i ri ð73Þ
v ¼ ji vi ð74Þ
XX qffiffiffiffiffiffiffiffiffi
e ¼ ji jj eii ejj ð1  k ij Þ ð75Þ

The segment fraction ji of component i is given by Eq. (76).

x i ri
ji ¼ X ð76Þ
x i ri
42 2 Polymer Thermodynamics

Fig. 2.14. Corresponding states behavior of various liquid–

polymer PVT data according to the Sanchez–Lacombe model.
Symbols represent experimental data; curves are calculated
from Eq. (71). Reproduced with permission from Ref. 62.

Neau [65] also showed that the expressions for the chemical potential used in
earlier literature to calculate phase equilibria are thermodynamically inconsistent.
According to Neau, the correct expression for the fugacity coefficient for the SL
model is Eq. (77).
2.4 Equation of State Models 43
   !"   #

1 1 z1 nr qv
ln j^i ¼ ln z þ ri 2  ln 1  þ X
v~T~ v~ xj r j v  qn i nj 0n i
"   #
1 nr qe 
v~T~ e  qn i nj 0n i

The compressibility factor z is given by Eq. (78).

Pv p~v~
z¼ ¼ r ð78Þ

In Figure 2.15 [66] experimental isothermal cloud-point curves of the linear low
density polyethylene þ hexane system are compared with the results of a fit of
these data using the Sanchez–Lacombe equation of state. The pure component
parameters of hexane were calculated from the critical point of hexane and its
acentric factor [67]. The pure component parameters of the polymer were obtained
from a simultaneous fit of PVT data and the data presented in Figure 2.15. The
equations solved were those described by Koak and Heidemann [68]. The binary
interaction parameter was linearly dependent on temperature. The polymer was


P / MPa

450 K
470 K
490 K
2 Critical Point
Cloud Point
1 Spinodal
Critical Point
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
Weight fraction LLDPE
Fig. 2.15.Isothermal cloud-point curves of LLDPE þ n-hexane.
Symbols: experiments; curves: modified Sanchez–Lacombe fit.
44 2 Polymer Thermodynamics

represented by 36 pseudo-components. The SL theory is very well able to describe

the experimental phase behavior.

Statistical Associating-fluid Theory

The statistical associating-fluid theory (SAFT) developed by Chapman et al. [63, 64]
is based on the thermodynamic perturbation theory of Wertheim [69]. Since it
first appeared, many different versions of SAFT have been published. The different
SAFT versions and their application have been reviewed by Muller and Gubbins
For polymer solutions the SAFT version of Huang and Radosz [71, 72] is the
most widely used. In 2000, a promising new version of SAFT for polymer solutions
called PC-SAFT (perturbed chain-SAFT), was proposed by Gross and Sadowski
[73]. Here we will restrict ourselves to these two SAFT versions.
The basics of both equations of state are equal, and can be written as separate
contributions to the molar Helmholtz energy a. The molar residual Helmholtz en-
ergy a res, which is the difference between the molar Helmholtz energy of the sys-
tem and the molar Heltmholtz energy of the same system in the ideal gas state in
the same conditions of temperature, pressure, and composition, is calculated as
the sum of the contributions of a hard chain term a hc , a dispersion term a disp , and
an association term a assoc [Eq. (79)].

a res ¼ a  a ig ¼ a hc þ a disp þ a assoc ð79Þ

From this expression the pressure and chemical potential can be derived [Eqs.
(80) and (81), respectively] using standard thermodynamic relationships
(A ¼ n i a).
p¼ ð80Þ
qV T; n
mi ¼ ð81Þ
qn i V; T; nj0i

For polymer solutions the association term is normally not used. The two SAFT
versions discussed here do not explicitly account for polarity and differ only in the
way the dispersion contributions are calculated. SAFT and PC-SAFT Hard Chain Term

In the molecular picture behind SAFT a chain consists of mi hard-sphere seg-
ments. These hard-sphere segments are bonded by covalent bonds. The hard-
sphere term of both SAFT versions is the sum of two contributions: a hard-sphere
contribution and a term due the connectivity of these hard-sphere segments, as
2.4 Equation of State Models 45

given by Eq. (82), where m is the average chain length of the molecules in the mix-
ture [Eq. (83)] [64].

a hc a hs a chain
¼m þ ð82Þ
m¼ x i mi ð83Þ

The hard-sphere contribution a hs is represented by the Boublik–Mansoori hard-

sphere equation of state for mixtures of hard spheres, Eq. (84) [74, 75].
"  3  #
a hs 6 z23 þ 3z1 z2 z3  3z1 z2 z32 z2
¼ þ 2  z0 lnð1  z3 Þ ð84Þ
RT pr z3 ð1  z3 Þ 2 z3

In this equation zk is given by Eq. (85), where r is the number density, dii is
the temperature-dependent hard-sphere diameter obtained from Eq. (86); mi , the
hard-sphere diameter si, and the energy parameter ei are pure component pa-

p X
zk ¼ r x i mi ðdii Þ k ð85Þ
dii ðTÞ ¼ si 1  0:12 exp ð86Þ

The chain term a chain is given by Eq. (87), where gðdii Þ hs is the so-called hard-
sphere radial distribution function at close contact [Eq. (88)] [74, 75].

a chain X
¼ x i ð1  mi Þ ln½gðdii Þ hs  ð87Þ
1 didj 3z2 d i d j 2 3z22
gijhs ¼ þ þ ð88Þ
ð1  z3 Þ d i þ d j ð1  z3 Þ d i þ d j ð1  z3 Þ 3 SAFT Dispersion Term

In the Huang and Radosz version of SAFT [71, 72] the Chen–Kreglewski disper-
sion term is used. This term is obtained from a fit to the physical property data of
argon and is given by Eq. (89) [76], where t is a constant equal to 0.74048. The con-
stants Dij are given by Chen and Kreglewski [76].

 i   j
a disp X 4 X 9
u z3
¼ Dij ð89Þ
RT i¼1 j¼1
kT t
46 2 Polymer Thermodynamics

For mixtures, the van der Waals one-fluid mixing rules or the volume fraction mix-
ing rules can be used.
The van der Waals mixing rules are given by Eq. (90), where the combining rules
are Eqs. (91) and (92), in which k ij is an adjustable binary interaction parameter.

uij 3
x i xj d
u i j
kT ij
¼ XX ð90Þ
kT x i xj dij3
i j

di þ dj
dij ¼ ð91Þ
uij ¼ ð1  k ij Þ uii ujj ð92Þ

The volume fraction mixing rules are given by Eq. (93), with volume fractions de-
fined by Eq. (94).

u XX uij
¼ fi fj ð93Þ
kT i j

x i mi d 3i
fi ¼ X ð94Þ
xj m j d 3j

The combing rule for uij is again given by Eq. (92). The PC-SAFT Dispersion Term

In SAFT the dispersion term represents the interactions between individual seg-
ments, while in PC-SAFT the dispersion term represents the interactions of chains
of segments. The expression for a disp derived by Gross and Sadowski is Eq. (95),
where the terms on the right-hand side are defined by Eqs. (96) and (97).

a disp a1 a2
¼ þ ð95Þ
  X 6
a1 u
¼ 2prm 2 s3 a i ðmÞh i ð96Þ
RT kT i¼0
" #1
a2 8h  2h 2 20h  27h 2 þ 12h 3  2h 4
¼ prm 1 þ m þ ð1  mÞ
RT ð1  hÞ 4 ð1  hÞ 2 ð2  hÞ 2
 2 X
2 u
m s3 bi ðmÞh i ð97Þ
kT i¼0
2.4 Equation of State Models 47

The reduced fluid density h is defined by Eq. (98), in which NAv is Avogadro’s

h¼ rmd 3 ð98Þ

The parameters a i and bi are dependent on m, as Eqs. (99) and (100) state.

m1 ðm  1Þðm  2Þ
a i ðmÞ ¼ a 0i þ a 1i þ a 2i ð99Þ
m m2

m1 ðm  1Þðm  2Þ
bi ðmÞ ¼ b0i þ b1i þ b2i ð100Þ
m m2

The constants a 0i –b2i are fitted to the thermophysical properties of n-alkanes and
are given by Gross and Sadowski [73]. SAFT and PC-SAFT Applications

Both SAFT and PC-SAFT contain pure component parameters: the energy parame-
ter e or u, the hard-sphere diameter s, or the hard-sphere volume v 00 , and the num-
ber of segments m per molecule. For small (solvent) molecules these parameters
are obtained from a fit of vapor pressure data and saturated liquid volume data.
Since they do not have a vapor pressure, this fit is not possible for polymers, and
the pure component polymer parameters are obtained from a fit to PVT data of the
molten polymer or from a fit to PVT data and binary phase equilibrium data. For
the description of a mixture one needs one binary interaction parameter k ij per
binary, which has to be fitted to phase equilibrium data. If necessary, k ij can be
made temperature-dependent. In general, phase equilibria are very sensitive to
the k ij value.
In Figure 2.16 [77], experimental results from a system of ethylene þ HDPE are
compared with modeling results using SAFT and Sanchez–Lacombe models. In
both cases k ij is taken to be linearly dependent on temperature. Due to the polydis-
persity of the polymer, the cloud-point curves show a dip in which the critical point
is located. If in the modeling the polymer is assumed to be monodisperse, this
behavior cannot be reproduced. The figure shows that both SAFT and Sanchez–
Lacombe models give a reasonable description of the experimental phase behavior,
although at high and low polymer concentrations the deviations become larger.
The same system was modeled by Tumakaka et al. [78] using SAFT and PC-
SAFT. The results are presented in Figure 2.17, which clearly shows that in this
case PC-SAFT gives a better result than SAFT. In the same paper these authors
present PC-SAFT modeling results for LDPE þ solvent systems at constant poly-
mer concentration. The pure LDPE parameters were fitted to the experimental
data of ethane þ LDPE. These parameters were subsequently used to describe the
LDPE þ ethane, þ propene, þ propane, þ butane, and þ 1-butene systems, using a
48 2 Polymer Thermodynamics

Fig. 2.16. Isothermal cloudpoint curves of the

HDPE þ ethylene system. Mn ¼ 43 kg mol1 , Mw ¼ 118
kg mol1 , Mz ¼ 231 kg mol1 . Symbols: experimental data;
curves: modeling results: (a) SAFT model; (b) Sanchez–
Lacombe model. Reproduced with permission from Ref. 77.

temperature-independent k ij . The results are very good (see Figure 2.18), but it
should be kept in mind that the modeling is restricted to one polymer concen-
tration. Extension to Copolymers

For the modeling of the phase behavior of copolymer–solvent systems, the copoly-
mer can be treated as a homopolymer with effective pure component parameters.
Examples of this approach are given by McHugh and co-workers [79, 80]. The dis-
advantage of this approach is that the pure component polymer parameters depend
on the type and composition of the copolymer. Pure component polymer parame-
ters are obtained from binary polymer–solvent phase equilibrium data. With these
parameters it is possible to model the phase behavior of the same polymer with
another solvent.
A better approach is the copolymer SAFT approach of Radosz and co-workers
[81–83], in which the copolymer parameters are estimated on the basis of the mo-
lecular weight and structure only. For an AB-type copolymer there are three binary
interaction parameters, the interaction parameters between A segments and seg-
ments of the solvent molecule, the interaction parameter between B segments
2.4 Equation of State Models 49

Fig. 2.17. Isothermal cloud-point curves of the

HDPE þ ethylene system. Mn ¼ 43 kg mol1 , Mw ¼ 118
kg mol1 , Mz ¼ 231 kg mol1 . Symbols: experimental data;
curves: modeling results. Reproduced with permission from
Ref. 78.

and segments of the solvent molecule, and the interaction parameter between A
and B segments. The first two binary interaction parameters can be obtained from
the phase behavior of the two homopolymer systems, while the third has to be
fitted to some copolymer–solvent data. Once these parameters are known, predic-
tions can be made for copolymer–solvent systems with the same type of copolymer
but with a different copolymer composition. The same approach was followed by
Gross et al. [84] for the PC-SAFT model. The result is known as copolymer PC-
SAFT. The two parameters that characterize the polymer structure are the fraction
of type A segments in the polymer molecule and the bonding fraction which gives
the fraction of bonds between segment types A and B. The original literature gives
Figures 2.19 and 2.20 show experimental cloud-point curves and PC-SAFT mod-
eling of the ethylene þ poly(E-co-EA) and ethane þ poly(E-co-BA) systems, respec-
tively [85], at a polymer concentration of 5 wt.%. The model correctly predicts
the change in the location of the cloud point with changing comonomer concentra-
tion in the polymer. Especially interesting is the ethylene þ poly(E-co-EA) system,
in that the curve of cloud-point pressure as a function of the EA concentration in
the polymer shows a minimum. This behavior is correctly described by the model.
50 2 Polymer Thermodynamics

Fig. 2.18. Constant composition cloud-point curves of LDPE–

solvent systems at 5 wt.% polymer. Symbols: experimental
data; curves: modeling results using PC-SAFT with one
temperature-independent k ij for each system. Reproduced with
permission from Ref. 78.


Polymer–solvent systems behave in many respects in the same way as systems of

low molecular weight components. Differences between polymer–solvent systems
and low molecular weight systems are mainly caused by the fact that polymers
have no vapor pressure, that polymers are composed of many components of
different molecular weights, and that there is a large difference between the free
volumes of the solvent and of the polymer.
The thermodynamic models for polymer–solvent systems are less advanced than
for systems of low molecular weight compounds. In general low-pressure vapor–
liquid equilibria can be described very well with a variety of models. Once the ad-
justable parameters in these models are fitted to experimental data, reliable predic-
tions can be made for other conditions, for example at a different temperature or
for a system with the same solvent and the same type of polymer with a different
2.5 Conclusions 51

Fig. 2.19. Constant composition cloud-point the range 113–157 kg mol1 . Symbols:
curves for poly(E-co-EA)–ethylene systems with experimental data; curves: PC-SAFT
different repeat-unit compositions at 5 wt.% calculations. Reproduced with permission from
polymer. The copolymer molecular weight is in Ref. 85.

molecular weight. Vapor–liquid-equilibria can be predicted with different group

contribution models. Most recent models give reliable predictions of comparable
accuracy for different polymer–solvent systems. A disadvantage of this type of
model is that the parameters for the groups of interest should be available.
In general the thermodynamic modeling of low-pressure liquid–liquid equilibria
is more difficult than for vapor–liquid equilibria. This also holds for polymer–
solvent systems. Reliable prediction methods are not available. The correlation of
liquid–liquid data using the extended Flory–Huggins type models [27–29] gives
reasonably good results. Again, once the adjustable parameters in these models
are known, predictions can be made for other conditions.
High-pressure fluid-phase equilibria can only be modeled using equations of
state. However, the equation of state models contain adjustable binary interaction
parameters that have to be fitted to data. Small variations in these parameters in
general have a large influence on the predicted phase equilibria. The most promis-
ing models for high-pressure phase equilibria of polymer solutions are the SAFT
and the PC-SAFT ones.
52 2 Polymer Thermodynamics

Fig. 2.20. Constant composition cloud-point the range 155–283 kg mol1 . Symbols:
curves for poly(E-co-BA)–ethylene systems with experimental data; curves: PC-SAFT
different repeat-unit compositions at 5 wt.% calculations. Reproduced with permission from
polymer. The copolymer molecular weight is in Ref. 85.



a activity, molar Helmholtz energy

A Helmholtz energy
C number of external degrees of freedom
d temperature-dependent hard-sphere diameter
DC constant
f fugacity
g molar Gibbs energy; interaction parameter; radial distribution
G Gibbs energy
Gij NRTL parameter
G Gibbs energy per mole of lattice sites
k Boltzmann’s constant
k ij binary interaction parameter
m number of polymer components; number of segments
M molecular weight
n number of moles
Notation 53

N number of components
p correction factor
P pressure
q effective segment number; surface area
Q normalized van der Waals surface
r number of segments in a molecule
rn number-average chain length
R gas constant
Rm ; Rn normalized van der Waals volume
T absolute temperature
u energy parameter
v molar volume
V volume
w weight fraction
x; y mole fraction
X segment fraction
z lattice coordination number; compressibility factor


a NRTL nonrandomness parameter

a; b; p phase
g activity coefficient (mole fraction basis)
d solubility parameter; density
e energy parameter
y theta temperature; segment fraction
jC segment fraction; volume fraction
f fugacity coefficient
m chemical potential
W activity coefficient (weight fraction basis)
s temperature-independent hard-sphere diameter
t NRTL interaction parameter; constant in SAFT
z density-related variable [Eq. (85)]
h reduced density
u number of groups
w interaction parameter


d dispersion
hb hydrogen bonding
i; j component
mix mixing
p polymer; polar
s solvent
vdW van der Waals
54 2 Polymer Thermodynamics


0 standard state

hard core, indicates characteristic parameter
@ reduced
assoc association
attr attraction
chain chain formation contribution
comb combinatorial
crit critical
disp dispersion
E excess
fv free volume
hc hard chain
hs hard sphere
id ideal mixture
ig ideal gas
L liquid
res residual
sat saturated
V gas; vapor


LCST lower critical solution temperature

LDPE low-density polyethylene
LLDPE linear low-density polyethylene
LLE liquid–liquid equilibria
NRTL nonrandom two-liquid
PCHT perturbed hard-chain theory
poly(E-co-BA) poly(ethylene-co-BA)
poly(E-co-EA) poly(ethylene-co-EA)
SAFT statistical associating-fluid theory
S–L Sanchez–Lacombe lattice fluid model
UCST upper critical solution temperature
VLE vapor–liquid equilibria


1 P. J. Flory, Principles of Polymer 3 H. Fujita, Polymer Solutions, Elsevier,

Chemistry, Cornell University Press, The Netherlands, 1990.
USA, 1953. 4 R. P. Danner, M. S. High, Handbook
2 H. Tompa, Polymer Solutions, Butter- of Polymer Solution Thermodynamics,
worth, UK, 1956. DIPPR/AIChE, USA, 1993.
References 55

5 R. Koningsveld, W. H. Stockmayer, 24 N. Koak, Th. W. de Loos, R. A.

E. Nies, Polymer Phase Diagrams, Heidemann, Fluid Phase Equilib. 1998,
Oxford University Press, Oxford, 2001. 145, 311.
6 S. M. Lambert, Y. Song, J. M. 25 M. H. ter Horst, S. Behme, G.
Prausnitz in Equations of State for Sadowski, Th. W. de Loos, J.
Fluids and Fluid Mixtures, J. V. Supercrit. Fluids 2002, 23, 181.
Sengers, R. F. Kayser, C. J. Peters, 26 N. Pedrosa, J. Gao, I. M. Marrucho,
H. J. White (Eds.), Elsevier, The J. A. P. Coutinho, Fluid Phase
Netherlands, 2000, Chapter 14. Equilib. 2004, 219, 127.
7 J. M. Smith, H. C. Van Ness, M. M. 27 S. Vanhee, R. Koningsveld, H.
Abbott; Introduction to Chemical Berghmans, K. Solc, J. Plolym. Sci.
Engineering Thermodynamics; 6th ed., B., Polym. Phys. Ed. 1994, 32, 2307.
McGraw-Hill, USA, 2001. 28 R. Koningsveld, L. A. Kleintjens,
8 J. M. Prausnitz, R. N. Macromolecules 1971, 4, 637.
Lichtenthaler, Edmundo Gomes de 29 Y. C. Bae, J. J. Shim, D. S. Soane,
Azevedo; Molecular Thermodynamics of J. M. Prausnitz, J. Appl. Polym. Sci.
Fluid Phase Equilibria, 3rd ed. 1993, 47, 1193.
Prentice-Hall Inc., Englewood Cliffs, 30 Th. W. de Loos, R. N.
USA, 1999. Lichtenthaler, G. A. M. Diepen,
9 J. S. Rowlinson, F. L. Swinton, Macromolecules 1983, 16, 117.
Liquids and Liquid Mixtures, 3rd ed., 31 C. Qian, S. J. Mumby, B. E.
Butterworth, UK, 1982. Eichinger, Macromolecules 1991, 24,
10 D. Patterson, G. Delmas, Trans. 1655.
Faraday Soc. 1969, 65, 708. 32 S. Enders, Th. W. de Loos, Fluid
11 P. J. Flory, R. A. Orwoll, A. Vrij, J. Phase Equilib. 1997, 139, 335.
Am. Chem. Soc. 1964, 86, 3507. 33 C. M. Hansen, J. Paint Technol. 1967,
12 P. J. Flory, J. Am. Chem. Soc. 1965, 39, 104.
87, 1833. 34 T. Lindvig, M. L. Michelsen, G. M.
13 K. S. Siow, G. Delmas, D. Patterson, Kontogeorgis, AIChE J. 2001, 47,
Macromolecules 1972, 5, 29. 2573.
14 S. Saeki, N. Kuwahara, S. Konno, 35 T. Lindvig, M. L. Michelsen, G. M.
M. Kaneko, Macromolecules 1973, 6, Kontogeorgis, Fluid Phase Equilib.
246. 2002, 203, 247.
15 C. F. Kirby, M. A. McHugh, Chem. 36 A. F. M. Barton, Handbook of
Rev. 1999, 99, 565. Solubility Parameters and Other
16 L. Zeman, D. Patterson, J. Phys. Cohesion Parameters, CRC Press, USA,
Chem. 1972, 76, 1214. 1990.
17 P. Ehrlich, J. J. Kurpen, J. Polym. 37 C.-C. Chen, Fluid Phase Equilib. 1993,
Sci. Pt. A. 1963, 1, 3217. 83, 301.
18 P. Ehrlich, J. Polym. Sci. Pt. A. 1965, 38 H. Renon, J. M. Prausnitz, AIChE J.
3, 131. 1968, 14, 135.
19 T. Swelheim, J. de Swaan Arons, 39 Y.-T. Wu, Z.-Q. Zhu, D.-Q. Lin, L.-H.
G. A. M. Diepen, Recl. Trav. Chim. Mei, Fluid Phase Equilib. 1996, 121,
Pays-Bas 1965, 84, 261. 125.
20 Th. W. de Loos, J. de Graaf, J. de 40 M. G. Bawendi, K. F. Freed, J. Chem.
Swaan Arons, Fluid Phase Equilib. Phys. 1988, 88, 2741.
1996, 117, 40. 41 J. Dudowicz, K. F. Freed, W. G.
21 Th. W. de Loos, W. Poot, R. N. Madden, Macromolecules 1990, 23,
Lichtenthaler, J. Supercrit. Fluids 4803.
1995, 8, 282. 42 A. Haghtalab, J. H. Vera, AIChE J.
22 M. A. McHugh, T. L. Guckes, 1988, 34, 803.
Macromolcules 1988, 34, 9. 43 M. T. Zafarani-Moattar, R.
23 E. Kiran, W. Zhuang, Y. L. Sen, J. Sadeghi, Fluid Phase Equilib. 2002,
Appl. Polym. Sci. 1993, 47, 895. 34, 803.
56 2 Polymer Thermodynamics

44 N. Pedrosa, J. Gao, I. M. Marrucho, 65 E. Neau, Fluid Phase Equilib. 2002,

J. A. P. Coutinho, Fluid Phase 203, 133.
Equilib. 2004, 219, 127. 66 Th. W. de Loos, I. Nagy, R. A. Krenz,
45 G. M. Kontogeorgis, P. Coutsikos, R. A. Heidemann, Proceedings 7th
D. Tassios, A. Fredenslund, Fluid International Symposium on Super-
Phase Equilib. 1994, 92, 35. critical Fluids, Trieste, Italy, June 13–16,
46 A. Bondi, Physical Properties of 2004, ISASF, 2004.
Molecular Crystals, Liquids and Glasses, 67 K. Gauter, R. A. Heidemann, Ind.
Wiley, New York, 1968. Eng. Chem. Res. 2000, 39, 1115.
47 T. Oishi, J. M. Prausnitz, Ind. Eng. 68 N. Koak, R. A. Heidemann, AIChE J.
Chem. Process Des. Dev. 1978, 17, 333. 2001, 47, 1219.
48 A. Fredenslund, J. Gmehling, M. L. 69 M. S. Wertheim, J. Chem. Phys. 1987,
Michelsen, P. Rasmussen, J. M. 87, 7323.
Prausnitz, Ind. Eng. Chem. Process 70 E. A. Muller, K. E. Gubbins, Ind.
Des. Dev. 1977, 16, 450. Eng. Chem. Res. 2001, 40, 2193.
49 P. J. Flory, Discuss. Faraday Soc. 1970, 71 S. H. Huang, M. Radosz, Ind. Eng.
49, 7. Chem. Res. 1990, 29, 2284.
50 F. Chen, A. Fredenslund, P. 72 S. H. Huang, M. Radosz, Ind. Eng.
Rasmussen, Ind. Eng. Chem. Res. Chem. Res. 1991, 30, 1994.
1990, 29, 875. 73 J. Gross, G. Sadowski, Fluid Phase
51 M. S. High, R. P. Danner, AIChE J. Equilib. 2000, 168, 183.
1990, 36, 1625. 74 T. Boublik, J. Chem. Phys. 1970, 53,
52 H. S. Elbro, A. Fredenslund, P. 471.
Rasmussen, Macromolecules 1990, 93, 75 G. A. Mansoori, T. W. Leland, N. F.
2927. Carnahan, K. E. Starling, J. Chem.
53 G. M. Kontogeorgis, A. Freden- Phys, 1971, 54, 1523.
slund, D. Tassios, Ind. Eng. Chem. 76 S. S. Chen, A. Kreglewski, Ber.
Res. 1993, 32, 362. Bunsenges Phys. Chem. 1977, 81, 1048.
54 G. Bogdanic, A. Fredenslund, Ind. 77 N. Koak, R. M. Visser, Th. W. de
Eng. Chem. Res. 1994, 33, 1331. Loos, Fluid Phase Equilib. 1999, 158–
55 G. Bogdanic, A. Fredenslund, Ind. 160, 835.
Eng. Chem. Res. 1995, 34, 324. 78 F. Tumakaka, J. Gross, G. Sadowski,
56 H. S. Elbro, A. Fredenslund, P. Fluid Phase Equilib. 2002, 194–197,
Rasmussen, Ind. Eng. Chem. Res. 541.
1991, 30, 2576. 79 B. M. Hasch, M. A. McHugh, J.
57 S. Beret, J. M. Prausnitz, Polym. Sci. B 1995, 33, 715.
Macromolecules 1975, 6, 878. 80 S. H. Lee, B. M. Hasch, M. A.
58 S. Beret, J. M. Prausnitz, AIChE J. McHugh, Fluid Phase Equilib. 1996,
1975, 21, 1123. 117, 61.
59 M. D. Donohue, J. M. Praunitz, 81 M. Banaszak, C. K. Chen, M.
AIChE J. 1978, 24, 849. Radosz, Macromolecules, 1996, 29,
60 D. D. Liu, J. M. Pruasnitz, Ind. Eng. 6481.
Chem. Process Des. Dev. 1980, 19, 205. 82 C. Pan, M. Radosz, Ind. Eng. Chem.
61 I. C. Sanchez, R. H. Lacombe, J. Res., 1998, 37, 3169.
Phys. Chem. 1976, 80, 2352. 83 C. Pan, B. Folie, M. Radosz, Ind.
62 I. C. Sanchez, R. H. Lacombe, Eng. Chem. Res., 1999, 38, 2842.
Macromolecules 1978, 11, 1145. 84 J. Gross, O. Spuhl, F. Tumakaka, G.
63 W. G. Chapman, K. E. Gubbins, G. Sadowski, Ind. Eng. Chem. Res., 2003,
Jackson, M. Radosz, Fluid Phase 42, 1266.
Equilib. 1989, 52, 31. 85 F. Becker, M. Buback, H. Latz, G.
64 W. G. Chapman, K. E. Gubbins, G. Sadowski, F. Tumakaka, Fluid Phase
Jackson, M. Radosz, Ind. Eng, Chem. Equilib. 2004, 215, 263.
Res. 1990, 29, 1709.


Mário Rui P. F. N. Costa and Rolf Bachmann

Basic Concepts

A Brief History

The concept of producing larger molecules starting from smaller ones containing
suitable groups (such as carboxyls and hydroxyls), by reaction to create a bond be-
tween inert chemical moieties by splitting off a smaller molecule (a condensation),
can be traced back to the middle of the nineteenth century [1]. It was discovered
when preparing oligomers of ethylene glycol, but, in spite of the good impact the
work had at that time, it fell into oblivion a little later. Phenol–formaldehyde (in
the form of resins and varnishes) was the first commercial product (in 1907) con-
sisting of an entirely synthetic polymer [2]; its formation was also an example of
polycondensation; but this early development was purely empirical and contrib-
uted very little to scientific progress.
Staudinger’s concept of the existence of macromolecules came two decades later
[3]. It paved the way to the systematic study aimed at producing synthetic fibers
carried out by Carothers at DuPont in the late 1920s and early 1930s [4], from
which stems the still-used concepts of addition and condensation polymerizations.
In his famous book, Flory [5] justly criticized this terminology, since closely
related polymerization mechanisms involving reactions of groups with active hy-
drogen atoms with epoxides and isocyanates do not lead to the splitting of a by-
product; he proposed a classification into ‘‘step growth’’ and ‘‘chain growth poly-
merization’’. These expressions became widely used, even if they are quite vague;
much better would be ‘‘random’’ and ‘‘sequential’’ polymerization [6].
A survey of important examples of polycondensation reactions can be found in
Table 3.1.

1) The symbols used in this chapter are listed at

the end of the text, under ‘‘Notation’’.

Handbook of Polymer Reaction Engineering. Edited by T. Meyer, J. Keurentjes

Copyright 8 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
ISBN: 3-527-31014-2
58 3 Polycondensation

Tab. 3.1. A survey of technically important polycondensation reactions.

Functional groups or monomers Connecting group By-product


Carboxylic acid hydroxyl ester
Ester hydroxyl ester
Carboxylic acid primary amine amide
Phosgene/chloroformate hydroxyl C
Na2 S/aSNa aCl aSa NaCl
Di-sodium sulfide chloro end group (as in sulfide
monomer/sodium sulfide ClC6 H4 Cl)
end group
NaOa aCl aOa NaCl
Sodium alkoxide/phenoxide, chloro end group, e.g., ether
e.g., O
CH3 Cl S Cl

aNbCbO aOH O none

Isocyanate hydroxyl C
aNbCbO aNH2 O none
Isocyanate primary amine C
CH CH2 aNHa OH none
O secondary amine N CH2 CH
1,2-epoxide b-hydroxy-t-amine
aCH2 OH (NH2 )2 CO/NH2 CONHa O H2 O
Methylol urea/N-alkylurea C
N,N 0 -dialkylurea
AraCH2 OH Ar 0 aCH2 OH AraCH2 aAr 0 HCHO
Aromatic methylol aromatic methylol

In some polycondensations, new functional groups are created by reactions of

the monomers. Thus, methylol end groups are formed by the initial reactions of
formaldehyde with other monomers, and their condensation reactions shown
in the last two rows of Table 3.1 occur in later stages of formaldehyde/urea and
formaldehyde/phenol polymerizations (see Section 3.3.4).
3.1 Basic Concepts 59

Molecular Weight Growth and Carothers’ Equation

Carothers was mainly interested in polyesterifications, and later in polyamidations,

of compounds with the structures AXA and BYB, or sometimes AXB, in which X
and Y are inert chemical moieties whereas A and B are the active carboxyl and hy-
droxyl or amine groups, giving rise to a connecting ester or amide group Z and
splitting off a by-product (water) W [Eq. (1)].

AþB!ZþW ð1Þ

If reacting groups A and B do not belong to the same molecule, and thus do not
create a ring, this reaction leads to a decrease of one polymer molecule (compris-
ing the starting monomers).
A count of the polymer molecules and their mass in the system can easily be
carried out with the help of p, the conversion of reference groups A. Starting with
1 mol of monomer AXB, the remaining moles of polymer are 1  p. Polymer mass
per mole of AXB is equal to its molecular weight MAXB , minus the mass lost in by-
product formation, p  MW , yielding Eq. (2), allowing the number-average molecu-
lar weight (mass of polymer/moles of polymer) to be predicted.

Mn ¼ ð2Þ
1 p

With two monomers, on introduction of the initial stoichiometric ratio

r ¼ ½B0 /½A0 between mole concentrations of groups A and B, a similar reasoning
leads to Eq. (3).


Mn ¼ ð3Þ
1 þ r  2p

High molecular weights (say, above 20 000, needed for use as fibers) are possible
only by reaching high conversions when monomer molecular weights are a few
hundreds. With two different monomers, stoichiometric ratios very close to unity
are also needed. Well-known relationships for number-average degree of polymer-
ization xn are obtained from Eqs. (2) and (3) by replacing the molecular weights of
monomers by 1 and the molecular weight of by-product by 0.
The effect of ring formation on Mn of open chain molecules can be taken into
account by introducing pc , the conversion of end groups yielding rings, in Eqs.
(2) and (3). For instance, when there are two monomers AXA and BYB, Eq. (4) is

MAXA ð1  pc Þ þ ðr  pc ÞMBYB  2ð p  pc ÞMW

Mn ¼ ð4Þ
1 þ r  2ð p  pc Þ
60 3 Polycondensation

Fig. 3.1. Average degrees of polymerization and sol fraction in

XA f polycondensations without intramolecular reaction, versus
conversion p of A groups with monomer functionalities f ¼ 2
and f ¼ 3, for equilibrium or batch reaction starting from

Prediction of pc and the average molecular weight of rings is not a trivial task.
Carothers has stated that the fraction of rings in bulk polycondensations is negligi-
ble, except when five- or six-membered rings (or higher in the case of rings con-
taining Si, P, and similar heavier elements lower in the periodic table) can result.
Recent experimental results show this concept is not valid in the case of irrevers-
ible polycondensations (see Section 3.1.7).
Monomers with three or more functional groups lead to the formation of
branched polymers. A network of macroscopic dimensions (a gel) appears for
high enough conversions and monomer branching. In Figure 3.1 are presented
classical predictions (see Section 3.4.3), of average degrees of polymerization and
weight fraction of finite molecules (sol fraction) for the single monomer polycon-
densations of XA3 and XA2 . Notice that, in contrast with linear polycondensations,
which require high conversions for obtaining high values of x n , low-value maxima
(4 if f ¼ 3) at gel point are observed with these nonlinear polycondensations.
Close to full conversion, the sol is nearly pure unconverted monomer.
Stockmayer [7] has obtained the distribution for the fractions of the initial mono-
mer units present in each finite polymer molecules (Eq. (5), where Px is the set of
isomers with x repeating units X).

f ð fx  xÞ!
x½Px /½X  ¼ xp x1 ð1  pÞ xð f 2Þþ2 ð5Þ
x!ð fx  2x þ 2Þ!
3.1 Basic Concepts 61

Fig. 3.2. Equilibrium distributions of monomer units x ½Px =½X 

for XA f polycondensations without intramolecular reaction
versus conversion p of A groups with monomer functionalities
f ¼ 2 and 3.

If the functionality f ¼ 2, the even more notorious Schulz–Flory [8] distribution is

obtained [Eq. (6)].

x½Px /½X  ¼ xð1  pÞ 2 p x1 ð6Þ

This latter distribution has a relative maximum for x G x n, in contrast to the gen-
eral Stockmayer distribution, which decreases monotonically (see Figure 3.2).
More complex chemical systems lead to molecular weight distributions that are
qualitatively similar: linear polycondensations have most often geometrical or
nearly geometrical distributions with polydispersities M w /M n close to 2, whereas
nonlinear polycondensations lead to extremely broad distributions near the gel
point, but always dominated in number and even in weight by the lower molecular
weight species.
This chapter concentrates on processes for making either high molecular weight,
mainly linear polymers, or else low molecular weight, branched oligomers which
will be later used to form networks. In any case, consideration of initial stoichiom-
etry, and possible change of functionality due to side reactions, is essential to pre-
dict average molecular weights, and to avoid premature gelation in the processes
described here.
It is also possible to prepare low molecular weight mixtures of ring molecules for
further reactive processing (using, for instance, anion polymerization), starting
62 3 Polycondensation

with monomers [8] or even with recycled polymer [9], by dilution with inert sol-
vents and a suitable catalyst (see also Section 3.1.7). The discussion of these other-
wise interesting processes falls outside the scope of this chapter.

The Principle of Equal Reactivity and the Prediction of the Evolution of Functional
Group Concentrations

Carothers’ successful synthesis of high molecular weight, aliphatic polyesters could

only be carried out thanks to his care in reducing the effect of the hydrolysis reac-
tion through use of high vacuum. Further research carried out by his former col-
laborator P. J. Flory in order to assess the effect of molecular weight on chemical
kinetics would eventually lead to the establishment of the principle of equal reac-
tivity. A parallel research, which also has contributed to the establishment of the
same concept and to the prediction of equilibrium chain length distribution
(CLD), has been led by G. V. Schulz [10] in Germany.
Assuming there is no mutual interference of end groups in the same molecule
(which does often happen with groups attached to aromatic rings), the specific rate
of consumption, RA , of end groups A by an irreversible reaction reaction with
groups B can be written as Eq. (7).

RA ¼ k½A½B ð7Þ

Equation (7) states that a single apparent second-order rate constant k should de-
scribe all intermolecular reactions of the involved groups A and B for a certain con-
centration of catalyst(s) (which may be A or B themselves) and solvent, regardless
of the polymer molecule to which they belong. A change of composition of the re-
action medium will often cause k also to change.
If the reaction is reversible, a more general expression [Eq. (8)] for RA should be

RA ¼ k½A½B  k Z ½Z ð8Þ

In the case where, as often happens, a by-product W exists, the apparent first-order
reverse rate constant k Z is related, through Eq. (9), to apparent equilibrium ratio K
and by-product concentration ½W .

k Z ¼ k½W /K ð9Þ

Effect of Reaction Media on Equilibrium and Rate Parameters

Polycondensations must be carried out in bulk or with a concentration of solvent

as low as possible (viscosity being the limiting factor) in order to minimize the for-
mation of rings. If the end groups have polarities and/or hydrogen acceptor or do-
3.1 Basic Concepts 63

nor affinities very different from the bonds, this is likely to lead to changes in the
value of apparent constant k, which will be observed as a dependence on conver-
sion or stoichiometric ratio.
Known examples are the apparent increase in reactivity with conversion in ester-
ifications and the strong effect of water concentration on apparent equilibrium and
rate constants in amidation systems (see Sections 3.3.1 and 3.3.3 respectively).
Should some thermodynamic model of the chemical system be available, knowl-
edge of activity coefficients g of chemical groups A; B; Z and by-product W would
allow the apparent equilibrium ratio K to be related to the thermodynamic K 0 (a
true constant, a function only of temperature) through Eq. (10).

gA gB ½Z½W 
K ¼ K0 ¼ ð10Þ
g Z gW ½A½B

This kind of computation can be done using the UNIFAC contribution group
method [11], but its empirical character and lack of available parameters at high
temperatures limit drastically its usefulness.
Solvent or bulk media effects can be modeled by taking advantage of group con-
tribution methods [12]. Multiple kinetic experiments with different media compo-
sitions are needed in order to compare the free energy of interaction involving the
hypothetical transition state group (see Ref. 13 for an example). No such studies
have ever been tried with polymerization reactions.
Progress in computational quantum mechanics should ultimately make the
practical use of these correction methods more widespread.
In nonlinear polycondensations, it is possible to observe a limiting conversion,
the topological limit [14] when unreacted groups in network are too far apart to
meet. Rate constants will fall before reaching the topological limit if glass transi-
tion temperature Tg is attained (the glass effect). This situation occurs with many
thermosetting systems of practical interest, epoxies being the most studied, as dis-
cussed in the reviews by Mita and Horie [15] and Dušek [14].
A new theory for describing this phenomenon, leaving aside the long-used ‘‘free
volume’’ concept, was recently put forward by Corezzi et al. [16]. It is based upon
the Adam–Gibbs [17] entropy theory of glass transition. Formation of more cova-
lent bonds by polymerization decreases the number of available statistical configu-
rations of the chemical system, thus decreasing configurational entropy, suggest-
ing Eq. (11) relating structural relaxation time tr to number-average degree of
polymerization x n .

tr ¼ t0 exp½b g ðTÞx n 
b g0 ð11Þ
bg ¼
T  T0

Function b g ðTÞ is empirical and could be different from that in Eq. (11) [16]. There
is for the moment no published use of this theory to fit apparent kinetic constants.
64 3 Polycondensation

A possible suggestion is the following (untested) correlation with yet another two
parameters, k 0 and Cg , based on the Rabinowitch equation for diffusional effects
on chemical reactions [Eq. (12)].

1/k ¼ 1/k 0 þ Cg tr ð12Þ

In Dušek’s terminology, diffusion control of chemical reactions may be specific,

when the bond formation rate depends on the mobilities of individual reacting
molecules or substructures, or else nonspecific, when the reaction rate depends
only on the average properties of the system.
Experimental evidence points toward prevalence of nonspecific rate control: for
instance, gel conversion in epoxide resin cure is the same regardless of glass effect
Lacking a better alternative, an empirical dependence of rate parameters on con-
version, for a given starting composition, must be tried. Examples of this approach
will be discussed along with their respective chemical systems (bulk polyester and
polyamide formation). It is seen that the glass effect could be taken into account
using a similar procedure.

Polycondensation Reactions with Substitution Effects

When the reaction of a functional group changes the reactivity of its neighbors, a
substitution effect exists.
A first-shell substitution effect (FSSE) is the simplest kind of departure from
ideal behavior. It is the situation where reactivity is affected only by the reaction of
the functional groups attached to the same monomer unit. FSSEs are encountered
in many polycondensations, as reactivities of functional groups in monomers are
often different from the reactivities of end groups in polymers because of mutual
steric, resonance, or electrostatic interactions.
For instance, a glycol HOaXaOH shows no substitution effect in an esterifica-
tion reaction if the reactivity of the hydroxyls in aCOOaXaOH is the same: a single
kind of group aOH needs to be considered and the description of the reaction
scheme is much simpler.
A second-shell substitution effect (SSSE) occurs when the reactivity of a func-
tional group is affected by the reaction of all the groups attached not only to the
same monomer unit, but also to the units linked to that one. This behavior has
been recognized in urea/formaldehyde formation (see Section
Linear polycondensations AXA þ BYB with FSSEs in both root units X and Y are
examples found in some important processes, such as the esterification of tereph-
thalic acid and ethylene glycol leading to poly(ethylene terephthalate) (PET). Four
different rate constants ki are needed to describe forward reactions, and possibly an
equal number of apparent rate constants (kiZ , i ¼ 1 . . . 4) are required to describe
reverse reactions:
3.1 Basic Concepts 65
k1 k2
k1Z k2Z
k3 k4
k3Z k4Z

Recall that apparent first-order rate constants of the reverse reactions kiZ are related
to the apparent second-order rate constants of the forward reactions ki through the
apparent equilibrium ratios Ki and the concentrations of by-product:

Ki ¼ ð14Þ
kiZ ½W 

If there is no by-product, as happens with isocyanate þ hydroxyl reactions, ½W  ¼ 1

(dimensionless) should be used in Eq. (14).
This example illustrates several difficulties encountered in modeling reversible
polycondensation reactions with substitution effects. A major problem is the possi-
ble change in the nature of bonds because another bond connecting a different
unit has been destroyed. There is no closed set of rate equations in terms of
the concentrations of functional groups ½AXA; ½BYB; ½ZD ; ½ZA ; ½ZB , and ½ZP ,
even with the help of the two stoichiometric restrictions ½X  ¼ ½AXA þ ½ZD  þ
½ZA  þ ½ZP  and ½Y  ¼ ½BYB þ ½ZD  þ ½ZB  þ ½ZP . Introduction of more complex
chemical entities does not alleviate the problem.
For instance, sequence ZB XZB (a particular kind of monad, a repeating unit X
with its pendent chemical groups) is formed by destruction of either of the sequen-
ces ZB XZP YZP or ZB XZP YZA (which are dyads, sets of two connected repeating
units) by reverse reactions at the rightmost Z group as written above. In order to
predict the concentrations of monads, one needs the concentrations of dyads; this
demands knowledge of the concentrations of triads, and so on. This hierarchy of
equations can not be broken without obtaining the whole chain length distribu-
tion; for further details see Section 3.4.5.
In fact, the above kinetic model considers SSSEs for the reverse reaction.
An FSSE model taking into account reverse reaction must consider equality
of rate constants of reverse reactions k1Z ¼ k2Z ¼ k3Z ¼ k4Z ¼ k Z . Since ½ZA  ¼
2½AXZA YZA XA þ ½AXZA YZP  and ½ZB  ¼ 2½BYZB XZB YB þ ½BYZB XZP , rates of
formation of groups can be written as shown in Eqs. (15).

RAXA ¼ 4k1 ½AXA½BYB  2k2 ½AXAð½ZD  þ ½ZB Þ þ k Z ð½ZD  þ ½ZA Þ

RBYB ¼ 4k1 ½AXA½BYB  2k3 ½BYBð½ZD  þ ½ZA Þ þ k Z ð½ZD  þ ½ZB Þ
RZD ¼ 4k1 ½AXA½BYB  ½ZD ½2k2 ½AXA þ 2k3 ½BYB
þ k4 ð2½ZD  þ ½ZA  þ ½ZB Þ  k Z ð½ZD   ½ZA   ½ZB Þ
RZA ¼ 2k2 ½AXAð½ZD  þ ½ZB Þ þ 2k4 ½ZD 2  ½ZA ð2k3 ½BYB þ k4 ½ZB Þ  2k Z ½ZA 
RZB ¼ 2k3 ½BYBð½ZD  þ ½ZA Þ þ 2k4 ½ZD 2  ½ZB ð2k2 ½AXA þ k4 ½ZA Þ  2k Z ½ZB 
66 3 Polycondensation

In conclusion, descriptions of substitution effects not only require knowledge of

more kinetic parameters, but also may lead to added mathematical difficulties,
even for just the prediction of basic characteristics of the chemical system, such
as the concentrations of functional groups.

Exchange Reactions

Exchange reactions without formation of by-products, such as the acidolysis and

aminolysis reactions present in polyamidations, do not increase number-average
molecular weight, but can be an important cause of relaxation of chain length dis-
tributions toward the equilibrium. Kotliar has presented a general review of these
reactions [19].
Direct bond exchanges in the complete absence of by-product (such as amide–
amide exchange) are at best very slow (if they indeed exist) and, as these reactions
can be safely neglected, only exchange reactions involving an end group and a
bond should be of any importance. The exchange of bonds Z1 and Z2 (necessarily
formed with a common by-product W), can be described through Eqs. (16).

k1 k2
XA þ YC T XZ1 Y þ W XB þ YC T XZ2 Y þ W
k1Z k2Z
XA þ X 0 Z2 Y T XZ1 Y þ X 0 B

These reactions modify the counts of functional groups of similar chemical nature
(for example, distinguishable kinds of amides/amines/carboxylic acids). If there
is only a single kind of bond, there is no net creation or destruction of func-
tional groups or bonds, but a reshuffling of pieces of the reacting molecules takes
In order not to violate the condition of chemical equilibrium, rate constants
of forward and reverse exchange reactions can be related to equilibrium constants
of the condensation reactions in which the same functional groups are involved.
In this example, the equilibrium ratio of the exchange reaction must be in accord
with Eq. (17).

k12 K1
¼ ð17Þ
k21 K2

This operation will reduce the number of kinetic parameters in the kinetic scheme.
Also, several of the various equilibrium constants can also be related through a rea-
soning due to Gordon and Scantlebury [20] for the XA f polycondensation, which
consists of checking the formation of the same final products through successive
equilibria but with different intermediate stages.
3.1 Basic Concepts 67

Ring-forming Reactions

Prediction of the entropy of cyclization for the gaussian chain molecular model
by Jacobson and Stockmayer [21] could relate rate and equilibrium constants of
intramolecular ring-forming reactions to their analogous intermolecular counter-
parts. Their theory has later been refined [22, 23] as it provides a good test for
statistical–mechanical treatments of chain molecules.
Measured ring concentrations for thermodynamically controlled (reversible) poly-
condensations in bulk are usually only a few percent, so this is not a major factor
in those systems.
For irreversible polycondensations, recent studies by Kricheldorf [24, 25] using
MALDI mass spectroscopy have in several circumstances detected quite an appre-
ciable concentration of ring molecules. Unfortunately, dependence of ionization
and thus of instrumental response factor of polymer molecules on the nature of
the end groups [26] prevents a quantitative exploration of those findings. But it
can be concluded that extension of kinetic modeling in order to take into account
the presence of rings is more important than was previously acknowledged.
The rationale for this unexpected concentration of rings is the irreversible char-
acter of their formation. Unless there is a slight excess of some of the end groups,
presence of both kinds of end groups in the same molecule will inevitably lead to
competition of ring formation with respect to polycondensation. The limiting fac-
tor for molecular weight increase becomes ring formation, not closeness to stoichi-
ometry of the reagents.
In what follows, Cn is a ring with n bonds and L n is a linear molecule containing
n  1 bonds. The formation of that ring moiety can occur by intramolecular reac-
tion of functional groups, or by a ‘‘back-biting’’ reaction, analogous to the exchange
reactions discussed in the previous section, according to the schemes represented
in Eqs. (18) and (19).
k c ðnÞ
L n ! Cn þ W ð18Þ
k aE
L m þ Cn ! L mþn ð19Þ
cE k ðnÞ

Except for small rings, rate constants k cZ , representing breakage of bonds in the
ring, and k aE , describing the addition of rings to the end of chains through the ex-
change reactions, should be identical to their open-chain counterparts, k Z and k E
respectively. The reverse constants depend on the size of the ring. According to
the Jacobson–Stockmayer theory, valid for gaussian chains (and therefore only for
rings with at least some tens of bonds, and bulk media or with low solvent concen-
tration) k c and k cE should be related to the corresponding parameters for intermo-
lecular relations through Eqs. (20)–(21).

kjc ðnÞ ¼ kj k c ðnÞ ð20Þ

68 3 Polycondensation

kijcE ðnÞ ¼ kijE k c ðnÞ ð21Þ

k c ðnÞ ¼ ð3/2pnb 2 Þ 3/2 ð22Þ

In Eq. (22), the small contributions of the distances of the ends of functional
groups b were assimilated to the length of a bond, and NAL is the Avogadro–
Löschmidt number.
If there is only one kind of bond, an equilibrium constant of cyclization K c ðnÞ
can be defined from the equilibrium in Eq. (23).

½Cn ½L m 
K c ðnÞ ¼ ¼ k c ðnÞ/n ð23Þ
½L nþm 

Modeling of Polymerization Schemes

Methods for predicting molecular weight distributions at chemical equilibrium and

for irreversible polycondensations are presented with some detail below; see Sec-
tions 3.4.3 and 3.4.4.
Systems at chemical equilibrium are amenable to description by the domain of
calculus of probabilities known as the theory of branching processes [27]. Batch re-
actions starting from monomers can sometimes be described by the same solu-
tions, increasing the practical importance of such an approach. Exchange reactions
also drive molecular weight distributions toward equilibrium. Many reactions of
great technological interest, such as melt polyesterifications, can be tackled using
this approach.
Applicability of statistical–probabilistic methods far from chemical equilibrium
cannot be guaranteed, and the kinetic approaches described in Sections 3.4.4 and
3.4.5 are a better choice.
Use of statistical–probabilistic methods for describing polycondensations started
with Flory, who used them to compute the equilibrium chain length distribution of
linear systems and later was able to predict gelation conditions for multifunctional
monomers. Stockmayer [7] could extend this method to the computation of chain
length distributions and average molecular weights of nonlinear polymers.
Gordon [20, 28] recognized that nonlinear polymers in the absence of intramo-
lecular reaction can be described by a Galton–Watson branching process. Relatively
complex chemical systems, presenting substitution effects, became tractable, and
even some properties of polymer networks relevant to rubber elasticity theory [29]
and average radius of gyration [30] as well as other polymer properties [31–33]
could be computed.
Alternative approaches to Gordon’s branching theory have later been developed,
mainly the so-called recursive method [34, 35], which is in a number of ways
simpler to understand and use, but lacks the power of the older theory in many
3.2 Mass Transfer Issues in Polycondensations 69

Reversible polycondensations can be tackled using the concept of molecular frag-

ments introduced in Section 3.1.5. It is possible to establish closed sets of rate
equations for those fragments in many important cases (the main difficulty being
the presence of higher-order substitution effects). For a more detailed discussion,
together with a short analysis of the much simpler case of linear polycondensa-
tions with a single kind of bond (a single monomer AXB or two monomers with
different groups AXA þ BYB), see Section 3.4.5.
It is noteworthy that CLD in those chemical systems (and probably in similar al-
ternating polyesterifications and polyamidations) is nearly always fairly close to
equilibrium, the main discrepancies occurring as far as the concentrations of the
first linear oligomers are concerned. So, a method which predicts those concentra-
tions correctly is what is really needed in practice. The main factor controlling the
CLD should be the presence of multifunctional impurities, which should be care-
fully tracked by the kinetic model.
Prediction of the whole CLD should be possible only with Monte Carlo methods,
with their usual drawbacks [36]. Nevertheless, polycondensations are easier to sim-
ulate by Monte Carlo than other polymerizations, since all reactions have more or
less the same time scale.
These simulations are invaluable for investigating the effect of space correlations
between reacting groups and their effect on ring formation and elastic properties
[37]. Prediction of molecular size distribution (which hopefully can be determined
by size exclusion chromatography) is also a useful result [38] which is difficult to
obtain otherwise.

Mass Transfer Issues in Polycondensations

This section deals with situations where mass transfer effects of by-products (devo-
latilization, solid-state polymerization) and monomers (interfacial polymerization)
become so important as to cause a spatial change in polymer molecular weight

Removal of Volatile By-products

Devolatilization in irreversible polycondensations is carried out in the later stages

of the process and is similar to other polymerizations. For reversible polycondensa-
tions, however, it differs from the devolatilization of other polymers in a number of

 Reaction and removal of by-product are intimately connected.

 Monomers may be removed in substantial quantities.
 Removal of volatiles takes place during the whole course of the reaction.
70 3 Polycondensation

In the first stages of the reaction, reversible polycondensates readily form boiling
liquids and care has to be taken to avoid removal of monomers, foaming, or prod-
uct entrainment.
Equilibrium constants are of the order of unity for polyesters and polycarbon-
ates, and of the order of hundreds for polyamides. In order to obtain end-group
conversions above 0.99, this implies a concentration of by-product ½W  below 104
times the concentration of bonds in the former two systems. The high temperature
of the processes increases the vapor pressure of the by-products and their equilib-
rium solubility decreases, but even so partial pressures of some millibars for poly-
ester and polycarbonate processes are required; they may be obtained by using a
combination of vacuum and inert stripping gas. These low partial pressures and
consequently nearly infinite dilution often justify the application of Henry’s law to
describe the vapor–liquid equilibrium according to Eq. (24), in which PW is the
equilibrium vapor pressure of by-product W, HW is its Henry constant for the poly-
mer melt (in the usual range of 10–1000 bar), wW is mass fraction in the melt, PW
is the vapor pressure of pure W, and WW is its weight fraction activity coefficient.

PW ¼ HW wW ¼ PW WW wW ð24Þ

The second part of Eq. (24) is useful below the critical temperature of the by-
product, when some thermodynamic model is available. For instance, given the
densities of melt and by-product respectively as rP ; rW , Flory–Huggins theory leads
to Eq. (25).

WW ¼ expð1 þ wÞ ð25Þ

Similar expressions hold for the volatile monomers, which at high dilutions can be
treated independently. At low conversions, the infinite dilution is not applicable
and the multicomponent Flory–Huggins equation [Eq. (26)], for example, should
be used instead. Here the fYi are the volume (or segment) fractions of the various
components, i ¼ 1 up to NY being the volatile components and the polymer being
component NY þ 1; the yYi are their mole fractions, VYi the molar volumes, and
the wij the binary interaction parameters.
0 1
rP B X
N Y þ1 f
Yj X
N Y þ1
f NX X
Y þ1 NY þ1 C
W Yi ¼ expB
@VYi þ fYi wij  Yi yYi fYk wjk C
A ð26Þ
rYi j¼1
V Y j j¼1
yY i j¼1 k¼ jþ1

At high conversions, mass transfer resistance to by-product removal is a key factor,

not because of low diffusivities D of the by-products (in the range 109 to 1011
m 2 s1 ), and consequent mass transfer coefficients, but because of low interfacial
Because of the huge increase in viscosity of the melt when the reacting mixture
changes from the initial mixture of monomers/oligomers to high polymer, a possi-
3.2 Mass Transfer Issues in Polycondensations 71

ble solution is the use of different continuous stirred reactors (CSTRs) in series,
with different kinds of stirrers. Other processes consist of simple unstirred bubble
In either case, these reacting mixtures at the beginning of the processes resem-
ble boiling liquids. Stirring helps the formation of bubbles by cavitation and breaks
existing bubbles, increasing interfacial area. Inert gas injection will also help (but
loss of volatile monomers becomes more difficult to counterbalance). These mech-
anisms become inefficient for high enough molecular weight with the consequent
higher viscosity (above 200 Pa s), and special film-forming equipment must be
Equipment for devolatilization of residual monomers and solvents without
chemical reaction is used in most polymerization processes, vented extruders be-
ing a common choice. They are also used in the final stages of reversible polycon-
densations [39], as well as other specially developed devices that we discuss next.
The basic design of these latter systems consists in a partially filled horizontal
cylindrical vessel in which the reaction mixture moves axially with the help of a
screw or a rotor with blades. The impeller periodically extracts part of the bulk
liquid, leaves it exposed as a thin film to the gas phase for a short time t f , and re-
mixes it again with the main stream as it flows toward the outlet.
According to the mechanism of creation of film, three classes of devices can be
distinguished [40–42]:

 Wiped-film reactors (WFRs) [43], where the impeller blades throw the polymer
melt against the inner surface of the vessel (Figure 3.3), subjecting it to a high
shear rate (in the range 10 3 –10 4 s1 ) in the gaps between their tips and the
wall. This high shear is favorable to conveying pseudoplastic, high-viscosity poly-
mers, but can conversely bring about problems for chains which break under
shear, such as polybutylene terephthalate. Vented extruders work using a similar
principle in their central section (where deeper screws lead to partial filling of
the channel) and their reactor models can be considered a variant of this class.
 Rotating disk contactors (RDCs) (Figure 3.4), in which the disks periodically dip
into the pool of reacting mixture. The polymer film they carry is exposed to the
gas phase before being mixed again with the bulk liquid.
 Falling-strand or falling-film evaporators.

Fig. 3.3. Diagrammatic representation of a wiped-film reactor.

72 3 Polycondensation

Fig. 3.4. Diagrammatic representation of a rotating disk contactor.

The currently used description of homogeneous diffusion of volatile by-products in

polymer media during reversible polycondensations is due to Secor [44]. It consid-
ers polymer molecules immobile. The flux of small molecules has a negligible con-
vective contribution; only the diffusional flux with respect to the polymer needs
be considered, and the microscopic mass balance of a generic volatile component
(usually, but not always, a by-product) Yi and a group A i belonging to the polymer
may be written, neglecting density variations, as in Eq. (27).

¼ DYi ‘ 2 ½Yi  þ RYi
q½A i 
¼ RA i

As the resistance to mass transfer in the gas phase may be neglected, the diffusive
flux of evaporation N_ Yi , in moles per unit area and unit time, will be obtained with
the help of a mass transfer coefficient kfYi . There are two equivalent alternative def-
initions [Eq. (28)], one using a driving force in terms of mole concentrations, the
other in terms of activities (asterisks mark a concentration or activity value at the
interface, considered to lie at y ¼ 0).

N_ Yi ¼ DYi ¼ kfYi ð½Yi   ½Yi   Þ ¼ kfY
ðaYw  aYi Þ ð28Þ
qy jy¼0 i

Combining this expression with the microscopic mass balances in the polymer
film, it is possible to predict mass transfer coefficients for simple geometries and
to take into account possible coupling with chemical reactions. For instance, if the
polymer film is immobile, has a constant depth L, and mass transfer occurs along
the y direction with a plane geometry, time-averaged mass transfer coefficient of
3.2 Mass Transfer Issues in Polycondensations 73

volatile species kfi after an exposure time t f would be computed by solving Eq. (27)
with initial and boundary conditions as given in Eqs. (29).

½Yi jt¼0 ¼ ½Yi 0 ; ½A i jt¼0 ¼ ½A i 0

½Yi jy¼0 ¼ ½Yi   ; ¼0 ð29Þ
qy jy¼L
ð tf
q½Yi ð y; tÞ
DYi dt
0 qy jy¼0
kfYi t f ¼
½Yi 0  ½Yi  

In view of the kinds of gas–liquid contact described above, an obvious model is

provided by the penetration theory (infinite depth L), where a portion of fluid with
uniform concentration profile is suddenly exposed to the gas phase, and is replaced
by fresh fluid after an exposure time t f . For an infinite film depth L and negligible
chemical reaction, this yields the well-known result [Eq. (30)] for the time-averaged
mass transfer coefficient.
kfYi ¼ 2 ð30Þ
pt f

Pell and Davis [45] were the first to actually measure a diffusion coefficient for a
volatile by-product of a polycondensation, using PET formation in films of varying
depth (although obtaining values of D much higher than those nowadays ac-
cepted). An early example of discussion of coupling of diffusion/chemical reactions
in these systems may be found in Ref. 46.
The first model associating the axial transport along the reactor (direction z) with
the cross-flow transfer of volatile by product (direction y) (see Figure 3.5) is due to
Amon and Denson [47] (Ault and Mellichamp [46] considered that all the polymer
was in the film) and was developed for WFRs. It assumes plug flow in the pool,
which implies a negligible hold-up of the liquid in the film. A time-averaged mass

Volatile by-product

Low M JT
polymer inlet Polymer outlet

z=0 z z=L y
Fig. 3.5. Simple model for a WFR.
74 3 Polycondensation

transfer coefficient is obtained at each axial position by solving Eqs. (27) and (29),
and the mass balance of the pool is written as Eqs. (31), where u z is the axial su-
perficial velocity (volumetric flow rate of polymer Q divided by cross-section area of
the pool) and a v is the film area per unit volume.

uz ¼ RYi  kfYi a v ð½Yi   ½Yi   Þ

d½A i 
uz ¼ RA i ð31Þ

½Yi jz¼0 ¼ ½Yi 0

Since the main mass transfer area is the film on the barrel wall, the exposure time
would be calculated [47] through Eq. (32), where d T is the inner barrel diameter, L x
is the film perimeter (L x G pd T , if the nip is small) and n_ is the screw rotational
speed in rotations per unit time.

tf ¼ ð32Þ
pd T n_

A better model [48] takes into account the movement of the film along the wall
with velocity u x ¼ pd T n_ between coordinates x ¼ 0 and x ¼ L x and therefore adds
a convection term, leading to Eqs. (33), where the ½A i f and ½Yi f are the concentra-
tions of the species in the film. There is no need to consider cylindrical geometry
for the film, since its thickness is low.

q½Yi f q½Yi f q 2 ½Yi f q½A i f q½A i f

þ ux ¼ DYi þ RYi þ ux ¼ RA i
qt qx qy 2 qt qx
q½Yi  q½Yi  ux
þ uz ¼ R Yi þ ½Yi f ðt; yÞ dy  u z ½Yi 
qt qz L 0
q½A i  q½A i  ux
þ uz ¼ RA i þ ½A i f ðt; yÞ dy  u z ½A i  ð33Þ
qt qz L 0

½Yi jz¼0 ¼ ½Yi 0 ðtÞ ½A i jz¼0 ¼ ½A i 0 ðtÞ

½Yi f jx ¼0 ðt; y; zÞ ¼ ½Yi ðt; zÞ ½A i f jx ¼0 ðt; y; zÞ ¼ ½A i ðt; zÞ

q½Yi f
¼0 ½Yi f ðt; 0Þ ¼ ½Yi  
qy jy¼L

With some modifications, a model of single-screw vented extruders can also be de-
veloped. We will present here a slightly extended version of the treatment by Rob-
erts [49] and Biesenberger and Sebastian [50]. Fundamental studies on fluid me-
chanics and mass transfer without reaction are reported in Refs. 51 and 52.
3.2 Mass Transfer Issues in Polycondensations 75

Unwrapped views:


Fig. 3.6. Scheme of flow and mass transfer in single-screw vented extruders.

A scheme of flow and mass transfer in single-screw vented extruders, also illus-
trating some key geometrical parameters, is shown in Figure 3.6.
As polymer flows inside the channel along a trajectory in a helix, with a total
length LB /sin y, in which y is the angle of the screw, the coordinate z will be taken
along this helicoidal path. Dimensions of the channel will be LW (width) by H
(depth), with a fraction fL filled with liquid. As the movement of screw pushes
the polymer pool, a fraction passes through the space between the barrel and the
screw, forming the desired evaporating film. If the fluid is Newtonian, the film
width is h/2, where h is the clearance. The average transverse velocity of the screw
dragging the film, vT , is given by Eq. (34).

vT ¼ pd T n_ sin y ð34Þ

The film which is wiped from the channel re-enters at a distance d upstream of its
departure point given by Eq. (35) [50] and the time of exposure of the film t f is
given by Eq. (36).

d ¼ pd T fL cos y ð35Þ
t f ¼ ð1  fL Þ/n_ ð36Þ

The concentrations ½A i f ðzÞ and ½Yi f ðzÞ respectively for nonvolatile and volatile
components in the film, back-mixed at axial position z, are obtained by solving
Eq. (27) for t ¼ t f with initial conditions described by Eqs. (37) (a more exact model
would consider convection as in Eq. (33)):

½Yi jt¼0 ¼ ½Yi ðz þ dÞ; ½A i jt¼0 ¼ ½A i ðz þ dÞ ð37Þ

76 3 Polycondensation

Assuming plug flow in the channel (a trivial change would be to add an axial dis-
persion coefficient), the mass balances in the channel taking into account also the
devolatilization from the pool (mass transfer coefficients kfpYi ) thus becomes those
given in Eqs. (38).

d½Yi  vT h kfpYi
uz ¼ RYi þ ð½Yi f  ½Yi Þ  ð½Yi   ½Yi   Þ
dz fL LW H fL LW

d½A i  vT h
uz ¼ RA i þ ð½A i f  ½A i Þ ð38Þ
dz fL LW H

½Yi jz¼0 ¼ ½Yi 0 ½A i jz¼0 ¼ ½A i 0

The presence of the delay d may be circumvented by using a Taylor expansion

about z [50] in order to obtain a system of second-order ODE. Also according
to Refs. 49 and 50, the exposure time of the pool may be estimated through Eq.

t fP ¼ ð39Þ
pd T n_ sin y

Another integration of Eq. (27) for t ¼ t fP with a trivial modification of Eq. (30) will
provide an estimation of the mass transfer coefficients of the pool kfpYi .
Twin-screw extruders have the advantage of being self-cleaning and can work
with extremely high viscosity, above 10 6 Pa s [53], making them a good choice for
polyamide and polyurethane final stages of reaction, thanks to the possibility of
using reduced space times. Their detailed modeling is more difficult than with
single-screw devices, and few fundamental studies [54] have been carried out.
Rates of mass transfer have been predicted for a co-rotating twin screw using pen-
etration theory and experiments have been done with a transparent device (for
observing whether bubbles were present or not in the devolatilization zone). Ob-
served results of kf a v were proportional to the speed of rotation to the power of
0.5, as penetration theory predicts, but were three times lower than theoretical pre-
dictions, which has been attributed to the very low liquid hold-up and consequent
lack of coverage.
In contrast to WFRs and vented extruders, use of staged models for describing
rotating disk contactors is a natural choice [55], since the J compartments with liq-
uid hold-ups Vm j can be approximated as CSTRs. Taking into account the possibil-
ity of back-flow, the overall volume flow rate leaving the jth CSTR, Q j , will be di-
vided into a fraction b j going back to CSTR j  1 and 1  b j going to tank j þ 1,
except for the first CSTR, in which b 1 ¼ 0. Also, Q Jþ1 ¼ 0, and the gas phase will
be considered well mixed with uniform temperature (see Figure 3.7). Thus, mass
balances at a steady state of volatile and nonvolatile components in the jth com-
partment ( j ¼ 1, J) may be written as in Eqs. (40).
3.2 Mass Transfer Issues in Polycondensations 77

Fig. 3.7. Staged model of an RDC.

Q j1 ð1  b j1 Þ½A i j1 þ Q jþ1 b jþ1 ½A i jþ1 ¼ Q j ½A i j  RA i Vm j

Q j1 ð1  b j1 Þ½Yi j1 þ Q jþ1 b jþ1 ½Yi jþ1 ð40Þ

¼ Q j ½Yi j  RYi Vm j þ kfYij avj Vm j ð½Yi j  ½Yi   Þ

Mass transfer coefficients can be computed using penetration theory as described

Murakami et al. [56] obtained correlations for hold-up and mixing time in RDCs.
Fractional dead space (between 0 and 18% for their experimental data) can be pre-
dicted from the mixing time. This dead space should obviously be kept to a mini-
mum, because polymer staying there will degrade due to secondary reactions
to possibly discolored or gelled material, and product quality may be seriously
harmed. Local film thickness and hold-up have been correlated to physical proper-
ties (surface tension, viscosity) and geometrical parameters [57]. Residence time
distribution has been shown to become narrower when viscosity grows [58].
Interfacial area [58] can be correlated with relative filling level H/d T and the
number of disk rings per unit length ( J/LT ) [Eq. (41)].

J 1:72  1:87H/d T
av ¼ ð41Þ
LT 0:085 þ 0:955H/d T

An experimental study on mass transfer in disk-ring contactors of diameter dR

using low-viscosity acrylamide solutions as a model fluid [60] has led to the corre-
lation of Eq. (42), but this correlation should only be used as a first approximation
[40], in view of the complexities introduced by polymer viscoelasticity.
 2 0:5
kf dR d n_
Sh ¼ ¼ 1:59 R ð42Þ

Falling-film or falling-strand devolatilizators receive increasing attention as they re-

quire no heavy and expensive machinery: Polymer is simply pumped through
small slits or holes. Efficient surface renewal is achieved by shear thinning during
the fall [59]; strands and films become very thin and may be completely depleted of
78 3 Polycondensation

volatiles. Depletion or possible tearing determine the optimal height of strands/

films for a given viscosity and initial film diameter. Reaction considerably increases
the efficiency of these devices (by a factor of up to ten) as it replenishes the volatile
by-product. Guides to the strands/films, such as wires, thin rods, or grids, will fur-
ther enhance the role of the reaction while reducing the effect of shear thinning
and tearing.
In all these devices, an additional component of interfacial area is provided by
the gas bubbles, which can result from sparging with inert gas (widely used for
devolatilization without chemical reaction) or from boiling. Observed rates of
mass transfer are often several times higher than predicted [50] and this discrep-
ancy has been linked to the presence of bubbles.
The Laplace–Kelvin equation predicts that an isolated gas bubble should redis-
solve if its size is below a critical threshold, and conversely it should grow if its ra-
dius r b exceeds the critical value given by Eq. (43) [50], where s is the surface ten-
sion, Pme is the local pressure over the bubble (it may be simply the hydrostatic
pressure, but in other circumstances it may be controlled by medium elasticity
[61]), and Pb is the pressure inside the bubble, equal to the sum of the partial vapor
pressures due to inert gases and volatile components if physical equilibrium and
gas-phase ideality hold. So, it is possible that no bubbling occurs if the pressure is
high enough or the content of volatiles is too low, but this is often not the case.

rbc ¼ ð43Þ
Pb  Pme

Bubble nucleation in polymer media is usually heterogeneous [62]. It starts when

shear stress manages to detach the small bubbles which are stuck in cracks and
crevices of vessel walls, internals, or suspended dust. Once the source of heteroge-
neous nucleation is spent, only high supersaturation will restart boiling. Of course,
in many practical situations there are already small air bubbles in the polymer, and
they will start foaming too. Favelukis et al. [63] have confirmed this view, and have
developed a theory for bubble growth which may explain the higher mass transfer
rates obtained with vented extruders and similar devices for high rotation speeds.
A predictive theory for bubble nucleation was developed in this same research [64].
Gestring and Mewes [65] have studied polymer degassing both with and without
bubbling using a transparent drum with a rotating blade (similar to the screw of a
vented extruder). Measured values of mass transfer rates without bubbling are
three times lower than predicted by penetration theory, because neither pool nor
film is well stirred – which could explain the failure of predictions in Ref. 54. Rates
of mass transfer in the presence of foaming were about 40 times higher than in
the bubble-free regime.
Trace devolatilization with the help of a stripper agent has a greatly enhanced ef-
ficiency. An important recent result is that, when it forms, foam grows until a
limiting volume is reached, regardless of initial volatile content and presence of
stripper gas [66], and thus the devolatilization section in vented extruders should
3.2 Mass Transfer Issues in Polycondensations 79

have enough room for that expansion, and residence time should also be sufficient
to allow the final density to be reached.
A patent [67] for enhancing mass transfer in this class of reactors proposes
the introduction of inert gas into the polymer in order to force the formation of
bubbles (the forced gas sweeping process). An experimental and theoretical model
has also been presented [68] and will be briefly summarized here.
The reactor was a rotating disk contactor for making bisphenol A (BPA) polycar-
bonate. The reactor model uses a staged approach, and the crux is the prediction of
the mass transfer rate of by-product (phenol). The two relationships expressed in
Eqs. (44) for the volume of a gas bubble Vb as a function of the gas flow rate Q g
[69] and of its rising velocity u b in a laminar regime [70] were the key data.

4p 15mQ g 3/4
Vb ¼
3 2rg
2gd b Q g d b 1/2
ub ¼ 1þ ð44Þ
Cd u b Vb

Cd ¼ þ1

In these equations m; r, respectively, are the viscosity and density of the liquid, g is
the acceleration of gravity, d b is the bubble diameter, and Cd is the drag coefficient.
The mass transfer coefficient was predicted using penetration theory, and the expo-
sure time was computed [Eq. (45)] as the rising time of a bubble in the melt (at
height hR above the gas injection point).

tf ¼ ð45Þ

The interfacial area per unit volume was obtained from the number of bubbles Nb
and the melt volume Vm [Eqs. (46)].

a v ¼ pdb2
Q g tf
Nb ¼

The observed bubble frequency agreed with the theoretical predictions, as also did
the profiles of x n versus reaction time.
It is interesting to finish this complex section with such a case study, suggesting
that for some problems ‘‘classical’’ Chemical Engineering of the 1960s can still
help in present-day industrial and scientific problems.
80 3 Polycondensation

The design and operation of most polycondensation reactors for devolatilization

of volatile by-products (namely vented extruders) is clearly a very difficult problem
because of the complexity of the flows (with or without foaming). Further progress
is likely to require a heavy use of computational fluid dynamics, as simplified
models seem to have arrived at their limits.

Solid-state Polycondensation

Current industrial processes for the production of high molecular weight, linear,
aromatic polyesters and polyamides, for use as plastics and fibers, use solid-state
polycondensation (SSP) for their last stages. This is the kind of process that will
be treated in this section: the polycondensation of semi-crystalline, low molecular
weight polymers to high molecular weight ones, occurring below the melting tem-
perature of high polymer, but well above the glass transition temperature. We will
disregard polycondensation of crystalline monomers, which is also covered in the
review by Pilati [71].
Because of the need to provide enough interfacial area to allow the removal of
volatile by-products, the polymer has to be in a powdery form. One of the several
optimization problems of these processes is to specify an economical starting par-
ticle size.
A practical difficulty is the possible tendency of the particles to stick, which will
make the process unfeasible. It is counteracted by starting with polymers with suf-
ficiently high crystallinity, and by adding glass beads. Another problem may be the
sublimation of oligomers, as in nylon-6, which may clog the bed. A precrystalliza-
tion step for PET, to make it attain at least 40% crystallinity before SSP starts, is
present in industrial processes since early 1970’s.
The main reason for using SSP is the achievement of molecular weights higher
than would be possible in melt polycondensation, owing to the selectivity gain of
polycondensation with respect to degradation reactions. Therefore, in a batch SSP,
a maximum in molecular weight versus time is expected, and this maximum will
occur at shorter times and will lead to lower molecular weights as the temperature
is increased.
The key assumptions made in order to interpret SSP are due to Gostoli, Pilati et
al. [72, 73] (see Figure 3.8):

 All chain end groups belong to the amorphous regions.

 No reactions occur in crystalline regions.
 Chemical reactions follow the same kinetics as in melt, taking into account
the change in the volume of the reaction media affecting functional group
 Equilibrium CLD holds locally.

The overall polydispersity of polymer will be greater than the equilibrium value (2
3.2 Mass Transfer Issues in Polycondensations 81

Fig. 3.8. Phase separation in solid-state polycondensation;

A, B are the end groups, W is volatile by-product.

for linear polycondensations) because of the radial profile of Mn in the diffusion-

controlled regime.
In the same way as in heterogeneous catalysis, effective diffusion coefficients
DYie (for fluxes with respect to the total geometric area) are decreased relative to
the values in the melt DYi because of the obstruction due to the crystalline phase
at a volume fraction fcr and because of the tortuosity factor tD (which depends on
the structure of the solid, values of 1.5 to 3 being common); the relationship is
given in Eq. (47).

1  fcr
DYie ¼ DYi ð47Þ

Notice that volume and mass fractions wcr of the crystalline phase are different be-
cause of the slight difference in density with respect to the amorphous phase.
An unavoidable complication is thus the description of the build-up of the crys-
talline phase, which affects mass transfer and chemical reactions by increasing
functional group concentrations in the amorphous phase. The rate of crystalliza-
tion is often described by the Avrami equation [Eq. (48)].

wcr ¼ 1  expðkcr t ncr Þ ð48Þ

The exponent ncr is a function of nucleation and growth type, which is not constant
for the entire course of crystallization. Mallon and Ray [74] have put forward in-
82 3 Polycondensation

stead of Avrami equation an empirical rate law in terms of residual amorphous

phase volume fraction.
The microscopic mass balance of polymer functional groups and volatile compo-
nents given in Eqs. (27) has to be modified in order to take into account the vari-
able reaction volume due to polymer crystallization. Here we do not follow the
notation in Ref. 74; rather, we use concentrations and rates of reaction per unit
volume of amorphous phase instead of concentrations per unit volume of particle
½A i p ¼ ½A i /ð1  fcr Þ, and so on, in order to use the same kinetic rate laws [Eqs.
(49)] as in the melt.
q½ð1  fcr Þ½Yi  1  fcr
¼ ‘ DYi ‘½Yi  þ ð1  fcr ÞRYi
qt tD
q½ð1  fcr Þ½A i 
¼ ð1  fcr ÞRA i

Examples of the use of this approach with PET and nylon-6,6, including a success-
ful comparison with available experimental data, can be found in Ref. 74.
Industrial-scale SSP is carried out in moving packed bed, fluidized bed, and
stirred bed reactors; Mallon and Ray have published a brief discussion of idealized
models of these reactors [75]. Fluidized beds have a serious drawback because of
the high consumption of gas needed to keep the bed in a fluidized state, and resi-
dence time distribution is unfavorable to high conversions. Stirred beds in series
are a possible solution, depending on economic details.
A model for SSP of nylon-6,6 in a moving bed reactor, considering its complex
geometry and its start-up and shutdown operation [76], can serve as a guide for
dealing with more complex real-life situations.

Interfacial Polycondensation

Typical chemical systems are fast reactions between two difunctional monomers,
AXA þ BYB. The first monomer (diamine, bisphenolate) is dissolved in a water so-
lution (in alkaline media in both cases), and the other monomer, with low water
solubility (acid chloride, phosgene), is usually dissolved in an organic solvent. Ei-
ther the neutral form of AXA is in an appreciable amount (in the case of amines),
or a phase transfer catalyst is needed (as in polycarbonate synthesis), since ionized
forms will not dissolve in the organic phase. A decrease in the pH is often used to
quench interfacial polyamidation.
The chain extension of water dispersions of isocyanates with water-dissolved
amines, in order to make polyurea dispersions, shares some similarities with the
former (amine þ acid chloride) systems.
Another common feature among these chemical systems is the presence of a
parasite reaction consuming end groups B by reaction with water.
3.2 Mass Transfer Issues in Polycondensations 83

Addition of a monofunctional chain stopper to the organic phase is advisable in

order to control the final M n .
In most cases (namely polyamides and polyureas), the polymer is insoluble in
monomer BYB or in its solvent, and precipitates as soon as it forms, often yielding
a shell through which AXA has to diffuse. Polycarbonates are an exception, as they
are completely soluble in the methylene chloride solvent chosen for their produc-
tion. Also, the organic phase is continuous in this latter case, which is a rather ex-
ceptional situation for interfacial polycondensations.
Also with the exception of polycarbonates, interfacial polycondensation is mostly
used in the production of specialty products, such as membranes [80–82] and mi-
crocapsules [83–87].
Early important contributions on interfacial polycondensations are described by
Morgan [77] as well as in Refs. 78 and 79. These studies show that the reaction
occurs in a layer close to the interface, on the organic side; the adjective ‘‘interfa-
cial’’ is thus rather misleading. Reaction is very fast and mass transfer resistance
is an important factor.
Models have for a long time concentrated on describing the velocity of con-
sumption of water-soluble monomer and consequent rate of film growth [83–87].
Although a possible framework for describing the simpler case of polycarbo-
nate formation has been presented by Mills [88], the first attempt at predicting
molecular weight distributions for more typical interfacial polycondensation sys-
tems is due to Karoda, Kulkarni et al. [89, 90], based on experimental data by
Johnson [91]. This work will be the basis of the analysis next presented in this
Assuming an apparent second-order rate constant in the organic phase k of the
order 10 2 to 10 4 m 3 /kmol s [77] and a concentration of functional groups B in
the bulk organic phase ½Bb ¼ 1 kmol m3 , the characteristic reaction time for con-
sumption of monomer AXA is of the order of 104 to 102 s. In the absence of
polymer, the diffusion coefficient of AXA in organic solvent þ BYB, assuming a
molecular weight of up to a few hundreds, should be of the order of 1010 to
1011 m 2 s1 , and this yields a characteristic width of reaction zone LR ¼ 108 to
107 m. So, polymerization occurs in a thin shell beneath the film of precipitated
polymer. There will be no functional groups B at the water interface and a steep
gradient of concentration of monomer BYB will be rapidly established inside the
organic phase.
Instead of solving microscopic mass balances for the concentration profiles, Kar-
ode et al. [89, 90] use average concentrations in the reaction zone and consider its
thickness LR constant (see Figure 3.9).
Furthermore, the rate of movement of the interface between the organic phase
and the precipitated polymer film is considered to be slow, so that a pseudo-steady
approximation for the diffusion of AXA through the polymer film and for diffusion
of BYB inside the organic phase should be valid. Mass balances of monomers in
the reaction zone and in bulk aqueous and organic phases resulting from these as-
sumptions are given in Eqs. (50).
84 3 Polycondensation

He [AXA]blk

Hi [AXA]


Bulk Swollen Bulk

aqueous polymer Reaction zone organic

phase film phase

Fig. 3.9. Model for interfacial polycondensation.

d½AXA He ½AXAblk  Hi ½AXA

dt Lp

LR ¼ LR RBYB þ kf BYB ð½BYBblk  ½BYBÞ
d½AXAblk He ½AXAblk  Hi ½AXA
Vaq ¼ av; aq DAXA
dt Lp

Vorg ¼ av; org kf BYB ð½BYBblk  ½BYBÞ

There are two partition coefficients He and Hi for monomer AXA:

 He is the ratio of bulk concentration in the aqueous phase ½AXAblk to concentra-

tion in the outer interface of polymer film (thickness L p ).
 Hi is the ratio of monomer concentration in the polymer film to concentration in
the organic phase.

Vaq and Vorg are the volumes of aqueous and organic phases and av; aq ; av; org are
their interfacial areas per unit volume (trivially related). No mass transfer resis-
tance is assumed to exist outside the polymer film (although it can be easily in-
cluded), but it is considered inside the organic phase, with the help of a mass
transfer coefficient kf BYB (which can be obtained by penetration theory). Introduc-
3.3 Polycondensation Processes in Detail 85

ing the rates of reaction of polymer species (see Section 3.4.4), their mass balances
can be written as Eq. (51).

¼ RPnIJ  k nuc; n ð½PnIJ   ½PnIJ sat Þ I; J ¼ A; B; C ð51Þ

Polymer precipitation is taken into account through the model of Kamide et al. [92]
with a phenomenological rate of nucleation k nuc; n (nil for n ¼ 1, taken as indepen-
dent of molecular weight for n > 1) [93]. The ‘‘saturation’’ concentrations of poly-
mer species are taken as the values of their concentrations in the lower branch of
the spinodal curve for the liquid–liquid equilibrium with organic solvent. The ear-
lier paper by Karode et al. [89] considered only spinodal decomposition.
A coherent film is predicted to form when the sum of projected areas for all
phase-separated polymer nuclei (assumed spherical) is equal to the interfacial
area. Film thickness is predicted through the overall mass balance of precipitated
Besides thermodynamic data on the liquid–liquid equilibria of polymer/solvent
and hydrophilic monomer polymer/solvent, this model needs the rate of nuclea-
tion k nuc , considered as an adjustable parameter, it also tries to fit the diffusion co-
efficient of monomer in the polymer film DAXA , and, as it postulates a constant
width of the reaction zone, it has also to fit the kinetic parameters.
In spite of its limitations in predictive power (a natural consequence of the com-
plexity of the phenomena involved), this approach has given an important new in-
sight on these processes. Film permeation properties should depend on the mode
of phase separation, nucleation giving better crystallinity. Molecular weight de-
pends on the competition between reaction and precipitation of polymer: a more
powerful solvent should lead to higher molecular weight. The existence of a sharp
maximum of average molecular weight as a function of the concentration of mono-
mer in the organic phase when the two monomer fluxes toward the reaction zone
are balanced, remarked upon by P. W. Morgan, could at last be explained by this
model – 40 years later.

Polycondensation Processes in Detail

Polyesters and Production Processes

Two kinds of polyesters will be discussed in this section:

 Linear, crystalline, high molecular weight (M n between 15 000 and 100 000), used
as plastics and fibers, the most important being poly(ethylene terephthalate)
(PET) and poly(butylene terephthalate) (PBT);
86 3 Polycondensation

 Unsaturated low molecular weight (Mn between 1000 and 10 000), often
branched, used as macromonomers for synthesis of thermosets (polyester res-
ins), or thermosetting materials by themselves (alkyd resins). They are prepared
from several monomers, namely phthalic and maleic anhydrides, adipic acid, iso-
phthalic acid, natural fatty acids or triglycerides, and a great variety of multifunc-
tional alcohols. In a few special cases, they may be saturated and/or linear for
use as macromonomers in the production of polyurethanes or other polymers. Acid-catalyzed Esterification and Alcoholysis

The two main reactions in these processes are esterification and alcoholysis, which
share similar mechanisms, with a hypothetical tetrahedral intermediate.
Reactions between anhydrides and alcohols are much faster than carboxyl þ
alcohol reaction, and full conversion of anhydrides (unless they are in excess) will
occur in less than one minute while being melted and blended with the rest of
the mixture. This stage is slightly exothermal and care is needed to avoid sudden
boiling of the reacting mixture. Kinetic parameters of this reaction (apart from
selectivities with respect to hydroxyls) are usually not needed.
Acidolysis reactions have a different mechanism, involving mixed anhydrides
[94]. Their rate is comparable to the esterification reactions only above 250 C.
Esterification also occurs in high-temperature alcoholysis of aromatic esters, as
carboxyl end groups are formed by side reactions, and should also be considered
in the kinetic modeling of these processes.
In spite of being the first reaction ever studied [95], esterification has been under
investigation ever since, and much knowledge has accumulated, even if some
points are still less clear. The basic kinetic model for polyesterification was estab-
lished by Flory and is summarized in his classic book [5]. Esterification was shown
to be acid-catalyzed, it is first-order with respect to hydroxyls and, with respect to
carboxyls, its order is either one in the presence of foreign strong protic acids, or
two in their absence [Eq. (52)].

R COOH ¼ kscat ½OH½COOH 2 ðno foreign acidÞ

R COOH ¼ kfcat ½OH½COOH½Cat-H ðstrong foreign acid Cat-HÞ

However, these rate laws can only be observed at low concentrations of hydroxyl
and carboxyl groups; otherwise a higher dielectric constant, association through
hydrogen bonds, and generic nonidealities will introduce changes in the apparent
values of the kinetic constants in Eq. (52). Experimental verification of these rate
laws is more delicate than it seems at first sight (the effect of reverse hydrolysis re-
action must be adequately taken into account or eliminated by the experimental
set-up) and has been repeated by Hamann et al. [96], but proposal of other kinetics
has continued. In their extensive review, Fradet and Maréchal [97] have found that
the overall order of esterification in the chemical literature is claimed to vary from
zero to six!
It is nevertheless useful to have some relationship, even empirical, that could ex-
tend the validity of Eq. (52) to the whole range of concentrations of functional
3.3 Polycondensation Processes in Detail 87

groups, and such a relationship (Eq. (53) where p is carboxyl conversion) has been
proposed by Chen and Wu [98, 99].

k ¼ kA expðapÞ ð53Þ

In Eq. (53) the empirical parameter a is a function of the initial stoichiometric ratio
r. It was theoretically linked to the dependence of the dissociation equilibrium con-
stant of the carboxylic acid on the dielectric constant of the medium and to the de-
pendence of the latter on the carboxylic acid concentration through conversion p. A
good fit of experimental data has been obtained with the parameter a varying in the
range 0.2 to 1.2 both for the self-catalyzed and the foreign acid-catalyzed esterifica-
tions of adipic acid with different diols.
In their modeling of unsaturated polyester resin formation, Beigzadeh et al.
[100] and Zetterlund et al. [101] have also found the Chen–Wu relationship more
useful than the rather cumbersome empirical models of Paatero et al. [102] or Leh-
tonen et al. [103]. Zetterlund et al. [101] have also presented interesting experimen-
tal data on the simultaneous self- and cross-catalysis by two carboxyl groups,
namely those formed by the addition of maleic and phthalic anhydride to 1,2–
propanediol; they show there is an anti-synergistic effect: that is, the total rate reac-
tion of the two carboxyl groups is lower than the sum of the rates of reaction of the
individual acids with the same concentration and at the same temperature. Catalysis by Metallic Compounds

Self-catalyzed esterification is often too slow to be of practical use, especially be-
cause hydroxyl-terminated polymers are either sought or are a consequence of the
process (aromatic polyesterifications are carried out with a large excess of hydroxyls
at the beginning of the process, because of the low solubility of the diacid), and
strong protic acids are not advisable, as they would catalyze polymer hydrolysis if
allowed to remain with the polymer. Even volatile catalysts such as methanesul-
fonic acid are avoided. Therefore, metallic salts are currently used as catalysts,
both for esterification and alcoholysis. Strong bases, such as lithium hydroxide,
can also be used, but for alcoholysis only (as in polycarbonate formation).
Metals in metallic complexes can catalyze esterification and alcoholysis through
two distinct mechanisms [117]:

 Metals of groups II–III (such as Zn, Mn, Ce, Pb), usually introduced as carboxy-
lates, complex the oxygen in carbonyl esters preferentially.
 Metals of groups III–VI (namely Ti, Sb, Ge, Bi), usually introduced as alkoxides,
dialkyltin oxides R2 SnO and carboxylates such as dibutyltin dilaurate, coordinate
with the acylic oxygen of esters.

Their activity with respect to esterification and alcoholysis has been compared by
Habib and Málek [105, 106] and Chung [107], who found volcano-shaped relation-
ships with different optima of activity of the several metals in terms of metal elec-
tronegativity according to Tanaka [108] for the glycolysis of dimethyl terephthalate
88 3 Polycondensation

(DMT) and for the polycondensation of bis(2-hydroxyethyl) terephthalate (BHET).

A much more careful analysis [117] has considered the catalyst concentrations
and the differences between the catalyzed reactions.
There are striking differences [109, 110] between the two groups of metals as
regards sensitivity to inhibition by carboxyls (which poison the first group) or by
hydroxyls (which poison the second group). Titanium has the best balance of prop-
erties, because it is little inhibited by carboxyls and hydroxyls (which poison Sb)
and also efficiently catalyzes esterification. It is also active at surprisingly low con-
centrations [117].
The choice of catalyst also depends on secondary reactions (Ti causes yellowing),
toxicity (a problem for Sb) and price (expensive Ge will nevertheless yield a white
The metal responsible for the catalysis is usually involved in several complexes.
An FTIR/NMR study of alcohol exchange with titanates in bulk or concentrated
solutions in apolar solvents [111] has shown that these compounds are present in
stable dimers, trimers, and other associations. These complexes exchange alkoxide
groups very rapidly, even at temperatures below ambient.
Kinetic models have to take into account both the existence of these polynuclear
complexes and the poisoning phenomena, and are more complicated than for acid-
catalyzed reactions. Alcoholysis of DMT, BHET, and other aromatic and aliphatic
esters is now becoming better understood. A key step was to recognize that in sys-
tems such as DMT þ ethylene glycol (EG) there are two reversible alcoholysis reac-
tions, one consisting in the attack of methyl esters either by EG or the hydroxyethyl
end groups, the other in the attack of the hydroxyethyl ester groups which produ-
ces oligomers detected by HPLC [Eq. (54)].

Besnoin and Choi [112] were the first to actually use experimentally measured
oligomer concentrations to validate this kinetic scheme for Zn catalyst, as was
soon confirmed and perfected by others [113–117].
These reactions are first order with respect to the esters and hydroxyls, but the
order with respect to the catalyst becomes zero at catalyst concentrations over a
few millimoles per gram (no power rate law [114]). Moreover, the ratio kEG /kHE
much depends on the catalyst (it may vary from 1 to 5) and even on the catalyst
concentration. Mixtures of divalent metal catalysts can have considerable synergis-
tic effects [117], which cannot be explained unless polynuclear complexes partici-
pate in the reaction.
It is noteworthy that, in the similar system dimethyl 2,6-naphthalenedicarboxy-
late þ 1,3-propanediol, the reactivity of hydroxyl groups in monomer and in hy-
droxypropyl chain ends is the same [118].
There are also studies for DMT þ 1,4-butanediol with Ti and divalent metal cata-
lysts, for which an order of one with respect to Ti and the hydroxyl and ester
3.3 Polycondensation Processes in Detail 89

groups has been reported [119]. The analysis of rate data is more difficult because
of the relatively high importance of the secondary reaction leading to THF forma-
tion (see below).
Models for Ti-catalyzed esterification are still more complex. Titanates are hydro-
lyzed by water and form oligomeric a(RO, R 0 O)aTiOa structures (unless the hy-
droxyl excess is large). These structures are more active than the monomeric tita-
nate [120]. Too much water will lead to a drop in activity [110], probably due to
formation of insoluble products. Order one was found with respect to acid, hy-
droxyl, and Ti (at a concentration of a few parts per million) [121], but there is a
slight inhibition effect by the ester groups.
The increase in activity brought about by vestiges of water has also been ob-
served both for Ti and Zr (this latter is even more active) in the model reaction of
octadecanoic acid þ octadecanol [122]. No simple first-order rate law with respect
to hydroxyls and carboxyls was found in that research.
Dialkyltin catalysts (such as dibutyltin dilaurate) have catalytic properties for es-
terification and alcoholysis similar to Ti and Zr [123]. The SnaC bond is fairly sta-
ble, but the upper acceptable temperature limit is around 220 C. On the other
hand, thanks to the added flexibility provided by the organic group and the possi-
bility of oligomerization, it is possible to prepare catalysts that are quite insensitive
to deactivation by vestiges of humidity [124]. Nowadays these catalysts are often
used in alkyd resin production. Side Reactions in Aromatic Polyester Production

The important side products formed in nonthermooxidative secondary reactions
in PET formation are acetaldehyde, vinyl end groups, and diethylene glycol (DEG)
units, together with carboxyl end groups. DEG units decrease the melting point,
crystallinity and thermal stability. Acetaldehyde, even at parts-per-million level, is
detectable by its flavor in drink bottles. Vinyl esters and acetaldehyde lead to chro-
mophoric products. Carboxyl end groups promote hydrolytic and thermal instabil-
ity. Additionally, these side reactions lead to a decrease in molecular weight, which
is particularly undesirable for product use in markets such as tire cord and soft
drink bottles.
Side reactions can be minimized to a certain extent by choice of operating condi-
tions and catalysis. Knowledge of this chemistry has obvious economic advantages,
and many details (for example, the influence of catalysts and stabilizers) cannot be
found in the open literature.
Thermal scission of ethylene diester groups is a likely source of carboxyl and
vinyl end groups [125] (Scheme 3.1):

Scheme 3.1. Vinyl end group formation from PET.
90 3 Polycondensation

This is a first-order reaction, with no effect of catalysts or additives.

Further alcoholysis of vinyl esters yields acetaldehyde [Eq. (a)].



DEG units are observed to form mainly in the first stages of the process. The pres-
ence of a hydroxyl ester group is indispensable, as it can be shown experimentally
that DEG forms in negligible amounts if ethylene glycol is heated alone in the ab-
sence of acid catalysts [126]. Reimschuessel [127] has suggested the attack of ethyl-
ene glycol or a hydroxyethyl end group as a possible source of the DEG moieties
(see Scheme 3.2). The analogous intramolecular etherification is the source of
side product dioxane [125] (see Scheme 3.3).

C CH2 O + H O CH2

C + CH2 O
Scheme 3.2. Formation of DEG moieties in PET.

C C + O O
Scheme 3.3. Dioxane formation in PET.

Poly(butylene terephthalate) is also subject to analogous side reactions [128],

the formation of 1-butenyl end groups (Scheme 3.4) followed by splitting-off of
1,3-butadiene or tetrahydrofuran (Scheme 3.5). This latter reaction is not affected
by the metallic catalyst, but is catalyzed by the carboxyl end groups, making pro-
duction of PBT by direct esterification of terephthalic acid and 1,4-butanediol dis-
advantageous. Side Reactions in the Formation of Unsaturated Polyesters

A recent review on the chemistry of unsaturated polyesters has been published by
Malik et al. [129]. Besides cis–trans isomerization [130], the other important side
reaction is Ordelt reaction [131–133] (see Scheme 3.6) which increases branching
and consumes double bonds. Both reactions are reversible and acid-catalyzed.
3.3 Polycondensation Processes in Detail 91

C + H2C CH CH2

Scheme 3.4. Vinyl end group formation from PBT.

C + CH2 CH2
Scheme 3.5. THF splitting-off from PBT.

Scheme 3.6. Ordelt reaction. Modeling of Processes of Aromatic Polyester Production

Continuous processes are currently used in the manufacture of PET. Several mod-
els have been developed, with the aim of contributing to a better design and opera-
tion. A brief discussion of their main assumptions and predictive capacities fol-
lows. Since polycondensation in solid-state and in film-forming devices has been
analyzed previously, only the specific aspects of the initial process stages still needs
to be covered.
Earlier models for continuous processes based on DMT [134, 135] study the in-
fluence of variables, such as the initial stoichiometric ratio, reactor temperatures,
and average residence times for CSTRs in series connected with distillation col-
umns, on process performance. They show it is possible to optimize conversion
and minimize formation of side products. Even if kinetics and physical equilibria
are now much better known, their qualitative conclusions should still hold.
PBT production, for which no published process models have been found,
should be described by a similar approach.
Yamada et al. [136] and more recently Kang et al. [137] have presented models of
the direct esterification process of terephthalic acid (TPA) with ethylene glycol (Fig-
ure 3.10). As TPA has a low solubility because of its high melting point, the first
reactor in the train (‘‘esterification’’ reactor) is operated at a higher pressure and
92 3 Polycondensation

Fig. 3.10. Simple model for WFR used for the direct
esterification process of terephthalic acid (TPA) with ethylene

temperature than the ‘‘pre-polycondensation’’ reactor in order to counterbalance

this lack of solubility. The calculations show it is also possible to optimize conver-
sion and minimize side reactions by a choice of average reaction times, tempera-
tures, and pressures. There are, however, no actual plant data to validate these con-
clusions. The kinetic model uses a ‘‘fragment’’ approach similar to what was
recommended in previous sections, although without taking into account the influ-
ence of alcoholysis and acidolysis on monomer concentrations. Also, no thermo-
dynamic model has been used for predicting the solubility of TPA, only an inter-
polation between values in pure ethylene glycol and different oligomers.
An integrated view of recent processes for PET has presented by Yao and Ray
[138]. Most of the DEG units are shown to be produced essentially in the esterifi-
cation and pre-polycondensation reactors, with very little change afterwards. Mini-
mizing vinyl ester formation, thus improving molecular weight and product qual-
ity, is achieved by decreasing residence time in the ‘‘finishing’’ wiped-film reactor,
but increasing the residence time in the solid-state polymerization reactor, which is
operated at a lower temperature under a stream of inert gas. Modeling of Processes for Unsaturated Polyester Production

Nava gives a brief description of the industrial processes [139]. Polyester resins
should ideally be produced with a certain predefined viscosity in their solution in
acrylic/vinyl monomers and with a known, reproducible, distribution of double
bonds. Trans double bonds are much more reactive in free-radical polymerization,
and the amounts of each one should be known. Carboxyl end groups are in some
processes further converted to metal carboxylates (typically, by adding MgO) in or-
der to thicken the solution, so it is also important to control their concentration.
Therefore, it is worth developing models for these processes, and a few recent
studies have appeared in this area [101, 104], but for now they only aim at predict-
ing the concentrations of functional groups, which is not a minor task in view of
the large number of parameters needed. Modeling of the full process requires con-
sideration of the losses of volatile monomers (lack of reliable vapor–liquid equilib-
ria data is a problem), and the aforementioned problems of taking into account the
low solubility of isophthalic and terephthalic acids are also pending.
3.3 Polycondensation Processes in Detail 93

Description of the branched structure of the resins is less problematic than in

the case of alkyds, because the conversion of double bonds by Ordelt reaction is in
the range 10 to 20%, so the molecules are almost comb-like and the approxima-
tions used by Yang and Pascault [140] should be reasonable.
Modeling of alkyd resin production is a rather formidable task because of the
high number of distinguishable chemical groups, the branched structure of the
polymer, a nonnegligible amount of intramolecular reaction, and side reactions of
the double bonds in fatty acids. The usual problems found in previously discussed
polyesterifications, namely lack of data for liquid–vapor and liquid–solid equilibria
and associated mass transfers, are also present.
Industrial processes [141–143] use batch stirred reactors, connected to partial
condensers in order to recover volatile glycols. Azeotropic distillation with xylene
(for instance) is often used. Monomers may be added in several steps to over-
come solubility problems. Vacuum and inert gas sparging is also used in different
stages. Catalysts (mostly alkyltin salts) are used to convert most of the carboxyl
groups, as specifications often require less than 1 mg g1 KOH acid value. Long
reaction times are sometimes avoided through different techniques for eliminating
residual carboxyls.
The same plant usually produces different varieties of resins, according, for in-
stance, to fatty acid content. New compositions are often sought in order to im-
prove end use properties or to compensate for fluctuations of raw material prices.
Therefore, it is often necessary to look for initial amounts of monomers which sat-
isfy stoichiometric constraints (such as mole ratio and fatty acid content), will not
lead to gelation, and meanwhile keep an acceptably high M n (for instance).
An early review of the foundations of the macromolecular chemistry of alkyds
has been presented by Kienle [143] and simple methods for predicting gelation
conversion have been reviewed by Jonason [144]. These predictions are not very ac-
curate, but good data on intramolecular reactions and differences of reactivity of
functional groups will be needed if better control of physico-chemical properties
is sought.

Polycarbonates General Introduction

Polycarbonates are polyesters of carbonic acid formed by reaction of diols (aro-
matic, aliphatic or a mixture of both) with a derivative of carbonic acid. The first
preparations of polycarbonates were reported by Einhorn in 1898 [155], by reaction
of phosgene with resorcinol or hydroquinone in a pyridine solution. Bischoff and
van Hedenström in 1902 [156] obtained the same aromatic polycarbonates via
transesterification with diphenyl carbonate (DPC). Thus the main routes to poly-
carbonates were established early, but the properties of the products seemed unin-
teresting. Around 1930 aliphatic polycarbonates were studied by Carothers and van
Natta [157]. These carbonates have low melting points and thermal resistance and
are not commercially interesting as stand-alone thermoplastics. Low molecular
94 3 Polycondensation

weight, aliphatic polycarbonates with hydroxy end groups, however, are widely
used as a diol component for the synthesis of polyurethanes and polyurethane–
urea elastomers.
Following the work of Whinfield and Dickson [158], who in 1941 succeeded
in preparing high molecular weight, high melting polyesters, the chemistry of
polycarbonates was re-examined. This led to the preparation of a linear, high
molecular weight polycarbonate derived from bisphenol A [BPA, or 2,2-di(4-
hydroxyphenyl)propane] by Schnell at Bayer [159, 160] and shortly afterwards by
Fox at General Electric [161]. BPA polycarbonate (BPA-PC) proved to be an out-
standing engineering thermoplastic that differs from the other polyesters in that
it is noncrystalline with a high glass transition temperature (about 150 C) and re-
tains its high transparency and toughness after molding. It has thermal stability up
to over 300 C as well as excellent mechanical, optical, and electrical properties, in-
herent fire resistance, and food compatibility. Improvements of several polymer
properties such as heat resistance, melt flow, and birefringence were achieved
with different (co)monomers [163, 164]. However, BPA-PC remains the commer-
cially most important polycarbonate.
The original processes – phosgenation in pyridine solution and melt
transesterification – were soon replaced by interfacial polycondensation with phos-
gene, which proceeds at low temperatures and allows the easy production of high
molecular weight polymer. It still remains the predominant production process al-
though interest in the simpler, non–phosgene-based and environmentally more at-
tractive transesterification process revived in the 1990s; problems with the earlier
melt carbonates, in particular the poor resin color, could be overcome. Polycarbon-
ate demand has enjoyed steady growth and total production capacity in 2003 was
about 2.5 million tons per year. Approximately 12% is produced by transesterifica-
tion and the percentage is expected to increase. The presentation here focuses on
the engineering aspects of BPA homopolymer production. Detailed reviews of the
synthesis and application of polycarbonates are given in Refs. 164–171. Interfacial Polycondensation

In the interfacial process, BPA and phosgene react at the boundary of two immis-
cible liquids, an aqueous alkaline BPA solution and an organic phase containing
phosgene. The overall reaction is shown in Scheme 3.7.

-2n NaCl
n NaO C ONa + n COCl2 O C O
CH3 n
Scheme 3.7. Overall stoichiometry of bisphenol A polycarbonate formation.

The synthesis proceeds in two steps: phosgenation of BPA forming oligomeric

carbonates with phenolic and chloroformate end groups; and polycondensation of
the oligomers (see Scheme 3.8). For the phosgenation, BPA is first dissolved in an
3.3 Polycondensation Processes in Detail 95

* ... C Cl + * ... Na * ... C ... * + NaCl
m n m n

* ... C Cl + NaO * ... O + NaCl
m m

* ... C Cl + 4 NaOH * ... Na + NaCl + Na2CO3 + H2O
m m
Scheme 3.8. Formation of polycarbonate by interfacial polycondensation.

aqueous alkaline solution as sodium bisphenolate, and phosgene is dissolved in a

solvent of chlorinated hydrocarbons (such as dichloromethane and monochloro-
benzene) which also dissolves polycarbonate. The reaction is started by dispersing
the two phases. Alternatively, liquid–gas phosgene (boiling point 4 C) is fed into a
slurry of BPA in the presence of an organic solvent. At high BPA concentrations a
fourth solid phase may be present [172]. Part of the phosgene is hydrolyzed with
NaOH to NaCl and Na2 CO3 and an excess of phosgene (10–20 mol%) is required
to compensate for hydrolysis and provide an excess of chloroformate end groups
for the following reaction step. Reaction temperatures are between 20 and 50 C
and the pH is kept between 9 and 13 by the addition of NaOH.
In the polycondensation step, a monofunctional phenol (such as 3–5 mol% phe-
nol, p-tert-butylphenol, p-cumylphenol) is added as a chain terminator to control
the molecular weight of the final polycarbonate. Reaction partners are now end
groups (chloroformate and phenolic aOH; see above) and reaction rates decrease.
The final polycondensation stages are catalyzed by tertiary amines. The amines re-
act with the chloroformate end groups to form intermediate quaternary acylium
salts which then react with phenolate to form carbonate and OH ions, hydrolyz-
ing the chloroformate end groups, or to form a urethane in a side reaction [164].
Detailed mechanistic studies of the catalyst reaction were performed by Aquino et
al. [173] and Kosky et al. [174].
Numerous variations of the interfacial process have been published. The reac-
tions can be carried out in batch in stirred tank reactors or continuously in series
of CSTRs and tubular reactors. Intensive mixing with dispersion and redispersion
is required throughout the reaction stages. After the reaction is complete, the brine
phase is separated and the polymer solution washed to remove residual amine and
base. Several processes for devolatilization are in use, including solventless precip-
itation, steam precipitation, spray drying, falling-strand devolatilization, and vac-
uum extrusion in devolatilizing extruders.
Phosgenation is generally mass transfer-limited [175, 176] and its rate depends
on mixing as well as on pH and the volume ratio of the organic and aqueous
phases. Although the polycondensation reaction of end groups is slower, both rates
still show the same dependencies due to the interfacial nature of the reaction. The
following phenomena contribute to the reaction process:
96 3 Polycondensation

 Dispersion of the two phases. Rates always depend on mixing. Effective kinetic
rate constants can be formulated as a function of energy dissipation or interfacial
 The type of emulsion. Both types of emulsions, oil in water (o/w) as well as water
in oil (w/o), can be found.
 The partition of phenols between the phases and its pH dependence [177]. Silva
and Kosky [178] studied the reaction of hydrolysis taking into account the differ-
ent phases and the partitioning of BPA between them. Monofunctional phenols
with better solubility in the organic phase show a better efficiency as chain termi-
 Mass transfer to and across the boundaries.
 Intrinsic (kinetic) reaction rates.

Due to the low reaction temperature and the use of chain terminator, the molecu-
lar weight distribution in interfacial synthesis is kinetically controlled and may be
far from thermodynamic equilibrium. In two parametric studies Mills [179] and
Munjal [180] have tried to model the full molecular weight distribution of polycar-
bonate. Varying the ratio of mass transfer/kinetic rates, they show how mass trans-
fer limitations can lead to a higher polydispersity or a higher oligomer content. Transesterification
The melt process is based on the transesterification of diphenylcarbonate (DPC)
and BPA (see Scheme 3.9).

n HO OH + n OCO

O OC + 2n OH
Scheme 3.9. Formation of bisphenol A polycarbonate by transesterification.

The melt process requires no solvent and polycarbonate is produced directly

without the need for cleaning, drying, and devolatilization. The only by-product,
phenol, is removed and can be recycled to the production of DPC or BPA. The pro-
cess, however, requires high temperatures (190–320 C) so that the polymer re-
mains molten and the viscosity can still be handled. With improved quality of the
starting materials and improved high-viscosity reactors these temperatures no lon-
ger affect polycarbonate quality and most grades can now be produced via the melt
process. The transesterification is a reversible reaction with an equilibrium con-
stant close to unity, so that the by-product phenol has to be removed for the reac-
tion to proceed. Pressures below 0.1 hPa may have to be applied for the final stage.
3.3 Polycondensation Processes in Detail 97

Small amounts of basic catalyst (< 0.01 mol% of alkyl-ammonium, or

phosphonium-salts) are mixed together with the DPC and BPA and the reaction
is started at the lower temperature. Conversion is driven by the removal of phenol
and the pressure is successively reduced while the temperature is increased. Sim-
ple reactors such as CSTRs, tubular reactors or falling-film evaporators can be used
for the first stages. Viscosity increases dramatically from 5 mPa s up to 1000 Pa s
as conversion increases and special reactors are needed for dealing with the high
viscosity at high conversions [149–151].
The melt process depends on the reaction rates, the thermodynamic phase equi-
librium, and mass transfer between the phases. Detailed mechanistic studies of the
Li-catalyzed melt process have been published by Choi et al. [152], and of the solu-
bility of DPC and phenol in polycarbonate by Webb [153]. The liquid–gas equilib-
rium has to take into account at least two components: phenol and DPC.
For a semi-batch operation for the first stages, optimal variations of pressure and
temperature can be calculated based on the above relationships plus the assump-
tion of phase equilibrium, or on a simple relationship for the mass transfer of
each volatile component Yi (Eq. (55), with the mass transfer rates per unit volume
Ji of component Yi , mass transfer coefficient of component i kfi , interface area per
unit volume a v , and equilibrium concentration ½Yi   at the interface).

Ji ¼ kfi a v ð½Yi   ½Yi   Þ ð55Þ

Mass transfer coefficients may be obtained by fitting to process data. Including

DPC loss, production capacity (reaction time), and unwanted side products in a
cost function, the optimization leads to a balancing of mass transfer and reaction
rate. This means that an optimal process is neither entirely mass-transfer nor ki-
netically controlled. To avoid side reactions that impair product quality, the lowest
temperature that kinetics and mass transfer allow is chosen. The results of the
semi-batch optimization can be transferred to the design of a staged, continuous
process [154].
The balancing of reaction versus mass transfer rate can similarly be applied for
the last stages. The relative DPC loss compared to phenol removal is quite high for
these stages. A high fraction of phenol end groups compared to OH end groups is
desirable for product quality. Therefore, excess DPC has to be used at the begin-
ning of the reaction. Unfortunately, a high fraction of phenol end groups reduces
the concentration of one reaction partner and hence reaction rates [146].
Mass transfer and hence the speed with which phenol can be removed are re-
duced with increasing viscosity. Reactors are needed that provide a high rate of
surface renewal. These may be wiped-film evaporators, single- or twin-screw ex-
truders, reactors of the rotating disk type or falling-strand evaporators [145, 149,
150]. Whereas the first three reactor types create surface actively, in falling-strand
evaporators the melt is simply pumped through small holes. High surface renewal
rates are achieved by small holes, but also by shear thinning of the falling strands
[59]. The effect of shear thinning films can also be achieved in rotating disk reac-
tors with disks that contain holes. Apart from surface renewal, mass transfer is de-
98 3 Polycondensation

termined by diffusion. Measurement of diffusion coefficients is difficult because of

the parallel reaction. An alternative is the determination of diffusion constants us-
ing molecular dynamics simulations [148].

Polyamides Introduction
Since their discovery by Carothers [181], aliphatic polyamides such as nylon-6,6
and nylon-6 are important textile fibers and plastics. Similar polyamides produced
by melt reaction of aliphatic diacids þ diamines, or by hydrolytic polymerization of
lactams, have some interest as engineering plastics, and will also be discussed in
this section.
Aliphatic, low molecular weight, branched polyamides of dimer fatty acids with
di- and triethylenediamine are among the most widely used curing agents for
epoxide resins. The chemistry of their formation is similar to that observed for
the aforementioned crystalline linear polyamides.
Aromatic polyamides are specialty products [182], used for high-performance fi-
bers and composites, which are produced by solution or interfacial processes from
acid chlorides and amines. Kinetic Modeling

The only published kinetic data on nylon-6,6 reaction are due to Ogata [183, 184],
and there is in general a great scarcity of information on the amide þ acid reaction.
Much more data are available on nylon-6 formation, such as in Refs. 185–187.
Because of the obvious identity in average chemical group composition of both
polymers, it should be possible to reconcile both kinetic laws [188]. However,
nylon-6,6 data have been obtained at high water concentrations and relatively low
temperatures, and data on nylon-6 are in the opposite situation. Degradation
reactions of nylon-6,6 through adipic acid chain ends (see Section compli-
cate the interpretation of kinetic data at higher temperatures. Schaffer, McAuley
et al. [189] have recently obtained experimental data on the nylon-6,12 system,
which does not present these problems, and these kinetics are now much better
Apparent equilibrium and kinetic constants in these systems are seen to depend
on water concentration. It obviously changes the activity coefficients of functional
groups, but no thermodynamic model has ever been used to completely describe
the mixture. The activity of water can be directly measured from knowledge of its
vapor pressure, and it has been claimed that a correlation based upon a Flory–
Huggins model can predict it [190], but no model exists for taking into account
the group interactions. Some interesting ideas can be found in Ref. 191, but no
actual thermodynamic model has been developed, concentrations of hypothetical
species having been used throughout that paper.
The crux of the treatment by Schaffer et al. [189] lies in the empirical correlation
3.3 Polycondensation Processes in Detail 99

expressed by Eq. (56), where bA and mA are assumed to be constant parameters;

activity coefficients are based on mole fractions.

¼ bA þ mA ½H2 O ð56Þ

As the overall composition of the system may be described by two variables, such
as [H2 O] and [COOH], a dependence on [COOH] might be added in Eq. (56).
However, this is not needed, as no effect of the mole ratio [COOH]/[NH2 ] on the
apparent equilibrium constant has ever been detected. Insertion of the above rela-
tionship into the mass action law provides a relationship between apparent equilib-
rium constant K a and true thermodynamic constant K 0 [Eq. (57)].

K a ¼ K 0 gH2 O ðbA þ mA ½H2 OÞ ð57Þ

It is seen that parameter bA is absorbed by the unknown thermodynamic constant,

and only the ratio g A ¼ mA /bA can actually be found from experimental data. The
correlations found for the activity coefficient of water are given in Eqs. (58).
gH2 O ¼ exp 9:624  ðNylon 6,12Þ
gH2 O ¼ exp 6:390  ðNylon 6,6Þ

Choosing a reference temperature T0 ¼ 549 K, Eq. (57) is rewritten as Eq. (59).

1 þ g A ½H2 O DH 1 1
K a ¼ K a0 exp  
gH2 O /gH2 O ðT0 Þ R T T0
K a0 ¼ exp  þ
gH2 O ðT0 Þ RT0 R

Parameters describing equilibrium (DH; K a0 , and g A ) have been fitted simulta-

neously with a mass transfer time constant for water k m, appropriate to their exper-
imental set-up, and activation energy Ec and pre-exponential factors kc0 or ku0 de-
scribing forward reaction through a third- or second-order rate law:

Ec 1 1 ½H2 O½CONH
R CONH ¼ kc0 exp   ½COOH ½NH2 ½COOH 
R T T0 Ka
Ec 1 1 ½H2 O½CONH
R CONH ¼ ku0 exp   ½NH2 ½COOH  ð60Þ
R T T0 Ka
100 3 Polycondensation

Tab. 3.2. Equilibrium and rate parameters for nylon formation.

Parameter Units Estimate 95% confidence interval

DH kcal mol1 1.82 G1.42

K a0 – 63.1 G6.2
gA g mmol1 2:03  102 G0:45  102
Ec kcal mol1 22.6 G7.7
kc0 g 2 mmol2 h1 3:19  104 G0:71  104
km h1 24.3 G15.4

The optimum values of parameters for the third-order model are shown in Table
3.2. Their approximate correlation matrix can be found in the same reference. It
shows that parameters describing equilibrium are highly correlated with g A , and
the pre-exponential factor is highly correlated with k m . Further, there is little
change in parameters for the second-order model, which yields ku0 ¼ 2:64  10 7
mg mol1 h1 and has a similar sum of weighed squared residuals. Thus, it is not
yet possible to determine the reaction order without performing experiments with
excess of diamine or diacid, which have not yet been reported at the time of writ-
ing. The authors state the parameters thus obtained for nylon-6,12 should hold for
the other aliphatic polyamides, just correcting the equilibrium constant, which is
lower by a factor of 0.4 to 0.5 in nylon-6,12 relative to nylon-6,6.
There are several studies concluding that the apparent order of reaction changes
from two to three as conversion grows. The reason might be a nonideality effect
similar to the one observed in esterifications.
Miller [192] has carried out an experimental study on aminolysis and acidolysis
reactions using carefully dried model compounds. As in esterifications, acidolysis
is slower and is explained by a mechanism involving the formation of intermediate
anhydrides. Its activation energy is 27 kcal mol1 , whereas aminolysis has the
much smaller activation energy of 13 kcal mol1 . The rate of aminolysis was
shown to be first order in carboxylic acid.
These results are valuable not only for dealing with block polymers, but also in
kinetic modeling, particularly with cyclic lactams: a narrow CLD will not occur be-
cause of the reorganization brought about by aminolysis.
Hydrolysis of caprolactam is autocatalytic. Mallon and Ray [191] have suggested
that its initial rate is determined by the presence of impurities.
The same authors have also remarked that the rate constant of addition of capro-
lactam to amine end groups is about what would be expected for an aminolysis
It should also be possible to predict the concentrations of higher cyclic
oligomers, but the only usable data concern the equilibrium concentrations, and
experimental confirmation has not yet been possible [191]. Nonoxidative Thermal Degradation Reactions

The main degradation reaction of nylon-6 is decarboxylation through interaction
of a carboxyl end group and caprolactam or an amide in the polymer chain (see
3.3 Polycondensation Processes in Detail 101

Scheme 3.10) [193, 194]. A slow deamination reaction has also been shown to

k1 O CH2
OH CH2 k2 C
- H2 O H2C CH2 - CO
CH2 2
Scheme 3.10. Nonoxidative thermal degradation of nylon-6.

Nylon-6,6 also degrades in the absence of oxygen, to a much greater extent than
nylon-6, and it eventually gels, as explained by the simplified mechanism in
Scheme 3.11 [195, 196].


NH Y k1
O O C k2 O
C CH2 CH2 C + NH2 Y

- CO2
X k3 X
+ 2 NH2 Z Z Z + 2 NH3

Scheme 3.11. Nonoxidative thermal degradation of nylon-6,6.

Cyclopentanone units are created, with a decrease in molecular weight, either

from carboxyl end groups or by intrachain reaction. Losses of CO2 and NH3 lead
to an imbalance of end groups, a nuisance for dyeing, and also to crosslinking. Process Modeling

Recent progress in kinetic and reactor modeling make it possible to assist the reac-
tor design and process operations with unprecedented exactness. A recent analysis
of the nylon-6,6 process [138] looked for improvements based on expansion of the
solid-state polymerization, but the gain was minor compared to what happened
with PET. The reason is the much reduced sensitivity of polyamides to by-product
removal as compared to polyesters, since the equilibrium constant of the former is
much greater.
Since the early 1970s, many researchers have modeled nylon-6 processes [197],
as reviewed by Kumar and Gupta [198]. There is an interesting optimization prob-
lem inherent to this process, which consists in adding just as much water as is
needed to start polymerization, and to get rid of it in the later stages. Because of
102 3 Polycondensation

the autocatalytic nature of the caprolactam hydrolysis, back-mixing increases con-

A simple and widely used reactor is the VK column (simplified continuous col-
umn), essentially a vertical tube at atmospheric pressure [193, 199], stirred by the
boiling action of water leaving the reactor mixture in the top zone.
Extensive pilot-plant tests carried by Jacobs and Schweigman [199], and further
data from industrial plant, have lead to a simple model of the VK column, consist-
ing in one or two CSTRs in series, followed by a plug-flow reactor.
Other designs [200] have improved performance by preventing water evapora-
tion at the top of the column through the use of above-atmospheric pressure.
The bottom one-third of the columns has a homogenizing function, not only
physical but also chemical, through the aminolysis reaction [201].
Vacuum stripping with a low residence time is used to eliminate most of the
large amount of caprolactam which remains because of the chemical equilibrium
of the back-biting reaction. An alternative is a hot-water extraction step, which will
also extract higher cyclic oligomers.

Polymerizations with Formaldehyde: Amino Resins (Urea and Melamine) and
Phenolics Formaldehyde Solutions in Water

Formaldehyde is a gas at room temperature. It may be sold as a low molecular
weight, solid polymer (paraformaldehyde), and more conveniently as 37% or 55%
water solutions, which usually contain some methanol. Under such conditions,
nearly all the formaldehyde is transformed into methanediol and higher oligomers
(see Scheme 3.12), usually end-capped by methanol, in order to reduce the average
molecular weight and prevent precipitation of paraformaldehyde.


(CH2O)xOH + HOCH2OH (CH2O)x+1OH + H2O



Scheme 3.12. Formaldehyde/water/methanol equilibria.

These reactions are not very fast at room temperature: characteristic reaction
times are of the order of minutes.
The various equilibrium constants have been measured using NMR [202] and a
model describing the vapor–liquid equilibrium in that system has been developed. Amino Resins

Reaction between a water solution of formaldehyde with urea, melamine, and sim-
ilar molecules (such as acrylamide) leads to hydroxymethylation of the nitrogens
3.3 Polycondensation Processes in Detail 103

[Eq. (b)] and further condensations produce the so-called amino resins [Eq. (c)], of
which 80% are based on urea [203], the rest being nearly all produced from mela-
mine (1).

XaNH2 þ HCHO ! XaNHaCH2 OH ðbÞ

XaNHaCH2 OH þ YaNH2 ! XaNHaCH2 Y þ H2 O ðcÞ



1 NH2
Scheme 3.13. Melamine.

Since melamine is made from urea and ammonia, it is more expensive. Mela-
mine resins are therefore chosen when one can get an appreciable benefit from
their better hydrolytic or thermal resistance.
Urea–formaldehyde (UF) resins are mainly used as adhesives for wood. Lami-
nated sheets (tables and counter tops) are a major application for melamine resins,
which stay in the outer decorative surface. Molding compounds, their first big ap-
plication, is still a major market, taking advantage of their extreme hardness and
heat resistance. Coatings, textile finishing, paper additives, leather tanning and
foundry binders, for which methanol- or butanol-etherified resins are usually em-
ployed, are important markets discussed in Ref. 203.
A major problem with the use of UF resins is their formaldehyde emission due
to hydrolysis. Formation of melamine resins is much less reversible and therefore
food contact with them is allowed.
Besides earlier classic data on the kinetics of reactions between urea, formalde-
hyde, and UF oligomers by de Jong and de Jonge [204–206], only experiments by
Price et al. [207] at higher temperatures in a sealed reactor are of immediate use to
establish a kinetic model of the chemical system. Kumar and Sood [208] have pro-
posed an FSSE model for the early stage of this polycondensation. A modified ver-
sion of that model introduces the groups presented in Scheme 3.14, where their
five urea monads U0 . . . U4 have been kept but three formaldehyde monads have
been used instead of two. Formation of tetrasubstituted urea is known to be negli-
Both de Jong and de Jonge, and Price et al., have considered that hydrolysis
reactions are unimolecular. Kumar and Sood [208] have considered it could be bi-
molecular, which seems to be reasonable. Available experimental data could not de-
cide for any of the alternatives, since the water concentration was always the same;
this matter needs to be solved, because higher initial concentrations of formalde-
104 3 Polycondensation

U0 = C U1 = C U2 = C
U3= U4 = C

F0 = C F1 = -CH2OH F2 = -CH2-
Scheme 3.14. Monads in the FSSE model of urea/formaldehyde polycondensation.

hyde are often used nowadays. Chemical transformations according to this new
model are shown in Scheme 3.15.

k1 k6
→ →
U 0 + F0 U 1 + F1 + W U 0 + F1 U 1 + F2 + W
 ←


h1 h2

k k

U 1 + F0 U + F1 + W U 1 + F1 U + F2 + W
 2 ←

 2
h1 h2

k k

U 1 + F0 U 3 + F1 + W U 1 + F1 U 3 + F2 + W
 ←


h1 h2

k4 k
→ →9

U 2 + F0 U 4 + F1 + W U 2 + F1 U + F2 + W
 ←

 4
h1 h2

k5 k10
→ →
U 3 + F0 U 4 + F1 + W U 3 + F1 U 4 + F2 + W
 ←


h1 h2

Scheme 3.15. Kinetic scheme of FSSE model of urea/formaldehyde polycondensation.

A possible simplification consists in distinguishing only rate constants of for-

ward reactions according to the number of hydrogens involved, and therefore
k1 ¼ 2k2 ¼ 4k5 and k3 ¼ k4 , k6 ¼ 2k7 ¼ 4k10 and k8 ¼ k9 , reducing the number of
unknown parameters.
Since the only information in the experimental data of Price et al. [207] is form-
aldehyde concentration versus time, it is not possible to estimate rate constants k6
to k10 from them. Although only at low temperatures, rate constants k1 ; k3 , and kh1
(this one measured from the rate of formation/hydrolysis of both monomethylol
urea and dimethylolurea) are nevertheless available from other sources, such as
Ref. 209. A crucial check of the FSSE hypothesis is the equality of the first-order
hydrolysis constants of methylol groups in mono- and dimethylolurea: the rate con-
stants per mole of the chemical substances should be double for dimethylolurea.
The value of this ratio is 1.68, standard deviation 0.42, for 15 values reported by
Landqvist with different buffers (pH 6, 7, 9.2, 10) at temperatures 20, 30 and 40 C.
3.3 Polycondensation Processes in Detail 105

However, that ratio is 6.9 according to de Jong and de Jonge; the activation energy
of the hydrolysis is the same (20 kcal mol1 ), though, according to both research
The equilibrium constants of the first and second hydroxymethylations of urea
are, respectively, 990 and 253 at 35 C, and there is a decrease of about 3 in the for-
ward rate constants of the successive substitutions. Interestingly, the rate con-
stants for the reaction of methylenediurea with formaldehyde or monomethylo-
lurea are identical, respectively, to those observed for urea þ formaldehyde and
urea þ monomethylolurea [210, 211].
So, there seems to be enough evidence to take FSSEs into account for urea, but
unfortunately it seems there might be no such thing as a single ‘‘aCH2 OH’’ group,
and SSSEs should be needed for fully describing this chemistry.
For simplicity, we will keep using the above model in the discussion.
Rate constants are known to depend on pH (although not much between pH 4
and 9); there is catalysis by OH and Hþ , so these constants should be written as
in Eq. (61) [204].

ki ¼ ki0 þ kiOH bOH c þ kiH bHþ c ð61Þ

Rate constants k6 to k10 in this scheme can be estimated from data of reactions in
acid media between urea, mono- and dimethylolurea [206]. In the same work, re-
action between methylols was found negligible (the temperature was at most
50 C). For these reactions, the terms ki0 ; kiOH can be neglected; the reactions are
very slow at pH > 4.
More recent work has concentrated on analysis supported by quantitative 13 C
NMR [212–216] and size exclusion chromatography [217] has completed this view
of the chemistry. At high temperatures and alkaline pH, methylol groups form
methylol ether bridges and uron rings 2 (Scheme 3.16).

2 C C C + H2O

Scheme 3.16. Reactions at alkaline pH in urea/formaldehyde polycondensation.

The preparation and possible industrial uses of uron UF resins may be found in
Ref. 218. NMR shows that these intra- or intermolecular ether bonds are destroyed
with liberation of formaldehyde at acid pH.
106 3 Polycondensation

Another question, that should not be overlooked, is the incomplete solubility of

the polymer in water. Except at low conversions, the reaction medium resembles a
colloidal dispersion [219]. Gelation may be physical, before or instead of being
chemical [220].
Amino resins are nearly always made in batch processes, consisting of thermo-
stated reaction kettles connected to a condenser. The current method of synthesis
has three stages [221]:

 synthesis of methylolated oligomers at pH 8 to 8.5;

 acid condensation at pH 4 to 5;
 addition of urea to decrease the final stoichiometric ratio to 1 to 1.3.

The goal is to reduce formaldehyde emissions [222], while keeping the properties
of products (such as wood panels) at an acceptable level. Since the formaldehyde/
urea molar ratio had to be decreased, the process has also become more difficult to
Modeling of the process has increased its potential importance in this context,
but the difficulties of putting it into practice are considerable, because of the daunt-
ing complexity of the chemistry. Notice that it should be integrated with the mod-
eling of the cure stage (a complex combination of heat, mass, and mechanical
modeling, in the case of wood panel manufacture). Cure is performed with ammo-
nium chloride as catalyst, which also acts as a formaldehyde scavenger. Its chemis-
try is not fully understood, because of the much higher temperatures than in resin
synthesis, which lead possibly to ladder structures.
Melamine resins have a similar chemistry, the main difference being the re-
duced importance of hydrolysis reactions. They are prepared using two stages, al-
kaline addition of formaldehyde, and acid polycondensation.
The initial stage of the melamine/formaldehyde reaction has been studied by
Nastke et al. [223], who succeeded in providing evidence not only of hydroxymethy-
lation, but also of the formation of methylene and methylene ether bridges using
polarography. The functionality of melamine is six, with a negative substitution ef-
fect of about 40% [224, 225].
As with UF, formaldehyde addition is acid- and base-catalyzed (sensitive to pH),
and equilibrium constants are 100 to 200. Methylene bridges are also formed only
at acid pH. Direct analysis of methylene ether bridges has also been performed by
NMR [226].
A simplified model (no reaction reversibility) of melamine–formaldehyde forma-
tion based on Tomita’s kinetic scheme [225] has been presented [227] and after-
wards extended to reaction in a CSTR [228], also considering reaction reversibility.
There is no experimental validation, but it is noteworthy for the use of a program
to calculate the CLD of a nonlinear reversible polycondensation in order to over-
come the astronomical number of reaction possibilities in the rates of formation
of individual oligomers.
3.3 Polycondensation Processes in Detail 107 Resins
Two subclasses have to be distinguished [229]:

 resols, which are highly branched, low molecular weight (150–1500) polymers
with a formaldehyde/phenol stoichiometric ratio between 1.2 to 3, formed at al-
kaline pH;
 novolacs, made at acid pH, with a formaldehyde/phenol stoichiometric ratio be-
tween 0.5 and 0.8, which have a different and much less branched structure than
resols. They are low molecular weight (500–5000) thermoplastics, further cross-
linked with hexamethylenetetramine.

Phenolic resins are mainly used as wood adhesives, laminates, molded parts, insu-
lating varnishes, abrasives, and rigid foams.
Novolacs can be made using either strong acid catalysts (sulfuric acid is pre-
ferred) or at pH 4 to 7 using carboxylates of divalent metals (such as Zn, Mn,
Mg). These catalysts complex phenol and methanediol and lead to formation of o-
methylolphenol in a first step. Subsequent addition steps may also be ortho-
directed (Scheme 3.17), or more randomly distributed (in the case of Zn).

- M2+ - M2+ CH2
CH2 CH2OH ...
OH - H2 O - H2 O

+ OH

Scheme 3.17. Formation of ortho-directed methylene bridges.

Novolacs prepared with acid catalysts have a more random structure. The
branching density is low because nonterminal rings are less reactive. This is
caused by molecular coiling, which is especially important in high ortho-novolacs.
Phenol groups tend to associate through hydrogen bonds and change the mole-
cular conformation (Scheme 3.18). Nonterminal units are often assumed to stay


Scheme 3.18. Hypothetical intramolecular hydrogen bonds in novolac resins.

108 3 Polycondensation

preferentially inside the molecular coils and decrease their reactivity because of
Mathematical models describing formation of novolacs both in batch and contin-
uous reactors have been developed by Frontini et al. [230] and Kumar et al. [231].
They consider the existence of at most one methylol group per molecule, and lump
together all isomers with the same unit counts. A Monte Carlo method [232] can
also be used in order to obtain a more detailed description at molecular level.
The initial addition of formaldehyde to phenol in alkaline media could for the
first time be successfully described, thanks to Zavitsas and collaborators [233]. Re-
sol formation in further reactions is a complex process, owing to the several differ-
ent aromatic reaction sites and substitution effects [234]; a total of 19 can be distin-
guished [235, 236]. Concentrations of fragments have been computed, assuming
irreversible reactions. Number-average and weight-average molecular weights have
been predicted using the ‘‘recursive’’ approach. Experimental determination of the
necessary structural and kinetic parameters is a huge task, which requires exten-
sive use of 13 C NMR and synthesis of model compounds. The research informa-
tion [237, 238] allows modeling of these reacting systems to be elaborated with
better chemical support. Fairly good agreement of model and experimentally mea-
sured functional group concentrations in resol formation at various stoichiometric
ratios has been claimed [239] and it is expected that a trustworthy quantitative de-
scription of these systems may eventually be achieved. Validation of these predic-
tions is plagued by experimental difficulties, and has mainly been carried out
through measurement of individual oligomer and functional group concentrations.
Reaction is exothermal (DH ¼ 80 kJ mol1 ). Heat of reaction is removed using
water reflux, sometimes with a small amount of inert solvent (aromatics are inert
only if they carry deactivating groups), and relatively small batch reactors ( 2 to 10
m 3 ) are usually preferred.
Resol reactors have been the subject of studies, such as Ref. 240, concerning op-
eration in accident situations. In novolac production, formaldehyde can be fed con-
tinuously in order to increase safety.

Epoxy Resins

The three-membered cyclic ether group oxirane, a 1,2-epoxide, or an epoxy group

reacts with substances containing an active hydrogen group, such as amines, phe-
nols, and carboxylic acids, or can be polymerized with anion or cation initiators,
therefore yielding a great variety of potentially useful polymers. Nearly all of them
are thermosetting, and used as coatings and adhesives. Their cure processes,
which should be considered in close connection with the method of producing
the final material, will not be discussed in this section. Instead, a brief review
of processes to make the macromonomers containing epoxy groups, the so-called
epoxy resins, will be presented.
The most widely used epoxy resins are formed through the reaction between
epichlorohydrin (ECH; 3 in Scheme 3.19) and bisphenol A.
3.3 Polycondensation Processes in Detail 109

+ -NaCl

Scheme 3.19. Formation of epoxy resin by reaction of bisphenol A with epichlorohydrin.

The so-called ‘‘taffy process’’ consists in the two-phase reaction of an alkaline so-
lution of bisphenol A with ECH in stoichiometric excess [241]. The main reaction
as described above is accompanied by side reactions, such as hydrolysis and alco-
holysis of chlorine and epoxides in ECH. These reactions create molecules with
functionality one or even zero, and must of course be minimized.
The kinetics has been studied by Enikolopyan et al. [242] and Gao [243], among
others. Branching formation occurs to a low extent and can usually be neglected;
the reaction can be described as a linear irreversible polycondensation AXA þ BYC,
with A ¼ aOH, B ¼ aCl, and C ¼ epoxide (ECH and oligomers have different
The similar epoxidation of novolacs with ECH has been additionally studied by
Oyanguren and Williams [244]. In this system, intramolecular ring formation has
been measured.
The so-called ‘‘advancement process’’ consists in the melt reaction of bisphenol
A (or a similar monomer) with a bifunctional epoxy resin in the presence of a cat-
alyst, with the goal of producing a higher molecular weight, bifunctional, epoxy
resin. This process leads to branching, due to the reaction of the pendent hydroxyl
group with epoxide, and eventually gelation occurs. Its kinetics has recently been
studied by Smith and Ishida [276]. The activation energy of the branching reaction
was found to be higher (20 as compared to 18 kcal mol1 ) than that of chain exten-
sion, and both constants have been determined both for the catalyzed and noncata-
lyzed reaction.

Polyurethanes and Polyureas

Urethane polymers were discovered by Baeyer in 1937 [246, 247], using the addi-
tion of alcohols to isocyanates leading to carbamates (or urethanes), 4 in Scheme

C O R'
H 4
Scheme 3.20. Addition of alcohols to isocyanates.
110 3 Polycondensation

The analogous, much faster, reaction with primary amines produces N-

substituted ureas 5 (Scheme 3.21).

R NCO + H2N R' R N
H 5
Scheme 3.21. Addition of primary amines to isocyanates.

Reaction with water produces an amine and carbon dioxide through an unstable
carbamic acid intermediate (Scheme 3.22).

R NCO + H2O R N R N + CO2
Scheme 3.22. Reaction of water with isocyanates.

As the amine reacts again with isocyanate, this reaction will lead to branching, as
will consecutive reactions with carbamates, leading to allophanates 6 (Scheme
3.23) and with ureas, leading to biurets 7 (Scheme 3.24).

C O R' C O R'
H C N R 6
Scheme 3.23. Reaction of urethanes with isocyanates leading to allophanates.

H C N R 7
Scheme 3.24. Reaction of ureas with isocyanates leading to biurets.

Trimerization of isocyanates, leading to the thermostable isocyanurate ring 8,

occurs with basic catalysts (multifunctional amines, carboxylates, alkoxides, and
so on) through allophanate intermediates [252–254] (Scheme 3.25).
Isocyanate groups also react to form uretdiones 9. This is an equilibrium reac-
tion, which is mainly important at high isocyanate concentration (Scheme 3.26).
Most urethane polymers are thermosets [247, 248]. They are mainly used in the
production of foams, taking advantage of the reaction of water and of the pre-
cipitation of insoluble ureas, which stabilize the foam even before the system
3.3 Polycondensation Processes in Detail 111

O R' O
R' N N
R N + R NCO R N O + R' OH
Scheme 3.25. Formation of isocyanurates.

O 9
Scheme 3.26. Formation of uretdiones.

gels chemically. Amine- or hydroxyethyl-capped branched polyols of molecular

weight of the order of a few thousands are preferred. They are most often star-
shaped poly(oxypropylene), in combination with shorter polyols such as glycerine,
additives such as surfactants and demolding agents, blowing agents, and water.
Rigid foams use more branched polyols than do flexible foams.
Another important use is the fabrication of plastics by RIM (reaction injection
molding) (see Figure 3.11), which exploits the very fast reaction which can be

Fig. 3.11. Scheme for an RIM machine with a jet impingement

mix-head (on a very exaggerated scale), in its recirculation and
injection modes.
112 3 Polycondensation

achieved either with amines or with hydroxyls, in the presence of catalysts such as
dibutyltin carboxylates, tertiary amines, or a combination of the two [250]. The ex-
cellent book by Macosko [249] extensively covers this technology, which is particu-
larly well adapted to the fabrication of big, flat, molded parts, and also to complex
elastomeric objects (viscosity is very low during mold filling). The whole produc-
tion cycle can take less than one minute from injection to demolding.
As with the other thermosets in this chapter, these processes will not be dis-
cussed in this section. We will nevertheless give a few hints of the reaction engi-
neering of the production of polymers and macromonomers based on this chemis-
try; other relevant uses are adhesives, binders, coatings, thermoplastic elastomers,
and fibers.
The catalysis of isocyanate reactions has been extensively studied because of
its critical importance in many of these processes. Noncatalyzed (or rather, self-
catalyzed) reactions may sometimes be fast enough in practice: isocyanate reac-
tions with amines are so fast that only recent studies using stopped-flow methods
could lead to useful data [255, 256], metallic or tertiary amine catalysts being inef-
fective in this case.
As often happens with polymerization reactions, simple rate laws can seldom
describe the whole course of reaction because of catalysis or inhibition by the
urethane groups formed or by the initial reagents. Self-association of metallic cata-
lysts, or their loose complexation by products or reagents, also prevents correlation
of rates of reaction by simple proportionality or even power-law relations. Catalysis
of isocyanate reaction with hydroxyls [250, 251] is by far the best understood.
It might be thought that modeling of polyurethane processes would be relatively
straightforward, given that reactions are mostly irreversible and the methods de-
scribed in Section 3.4.4 should deal with them without difficulties.
In reality, allophanate and biuret formation is reversible at temperatures above
130 C [252]. Formation of isocyanurates causes a reorganization of the CLD akin
to what happens in reversible polycondensations because of exchange reactions.
Many polyurethanes are block polymers prepared with a diisocyanate, a short
diol such as 1,4-butanediol or 1,6-hexanediol, or a diamine (the chain extender),
and a diol with molecular weight between 500 and 4000, based on a polyether,
polyester, polycarbonate, poly(butadiene) or other. Most often, the preparation is
performed in two steps: firstly, reaction of the longer polyol with the isocyanate,
then with the chain extender in the second stage.
An important and desirable feature of polyurethanes and polyureas is the phase
separation of the small isocyanate/chain extender blocks, which is possible pro-
vided the thermodynamics is favorable: this means a high enough concentration
and chain length of ‘‘hard’’ blocks. These ‘‘hard’’ blocks act as physical crosslinks
at a temperature lower than their melting point, and thermoplastic elastomers (in-
cluding elastic fibers) can therefore be obtained.
From the point of view of the prediction of structure, this brings about a complex
problem, which shares some characteristics with solid-state polycondensation, and
has been tackled through the use of Monte Carlo methods [257, 258, 259].
Reactors for these processes range from simple batch or continuous stirred tank
3.4 Modeling of Complex Polycondensation Reactions 113

reactors for very low molecular weight or solvent-based processes, to tubular reac-
tors with static mixers or extruders. Owing to the strong exothermicity of the main
reaction, cooling has to be used in order to prevent side reactions from becoming
too important.
Many processes at low temperature and in homogeneous phase can nevertheless
be analyzed through the methods described in Section 3.4.4, as they now stand.

Modeling of Complex Polycondensation Reactions


Rate equations allowing the prediction of concentrations of reactive groups, includ-

ing more complex molecular fragments (monads, dyads and so on) from mass bal-
ance equations are established in Section 3.4.2 for irreversible reactions or systems
with at most FSSEs if reversible reactions exist. Stoichiometric coefficients are in-
troduced in order to obtain a fairly general formalism, later exploited in Section
Knowledge of the distributions of numbers of bonds connecting repeating units
makes it possible to describe molecular structure at chemical equilibrium. In Sec-
tion 3.4.3, the main results concerning molecular weights and network properties
of chemical systems verifying FSSEs are presented, using the theory of branching
processes. The presence of rings is also allowed. The results of Section 3.4.2 are
useful in order to predict average numbers of bonds for each monad, needed for
predicting average molecular weights.
Irreversible polycondensations can be tackled quite easily using a general kinetic
approach developed in Section 3.4.4, allowing prediction of average molecular
weights before or after gelation, and even molecular weight distributions (lumping
together isomers with the same numbers of groups).
Prediction of molecular weight distributions for reversible, linear, alternating
polycondensation is discussed in Section 3.4.5. Mathematical difficulties grow con-
siderably in the presence of SSSEs. There seems to be no alternative to Monte
Carlo methods for dealing with reversible nonlinear polycondensations or even lin-
ear polycondensations where more than two kinds of bonds are present.

Description of Reactions in Polycondensations of Several Monomers with
Substitution Effects

The goal of this section is to present a general nomenclature of chemical entities

and reactions which provides a concise form for writing the rate equations and
mass balances of chemical species. The existence of substitution effects seriously
complicates the task, because the same molecular entity has to be labeled in a dif-
114 3 Polycondensation

ferent way, not because it has intrinsically changed, but because its neighbors have
We introduce the following nomenclature for the species (monomer units or
functional groups) and vectors h; g; e storing their indices:

Xh i monomer units, h i A 1 . . . NX
A gi ; A giþ functional groups which react forming bond ZiR ; indices gi ; giþ A
1 . . . NA
We i by-products, e i A 1 . . . NW
Z iR bonds, i A 1 . . . NR
Zei ; Zeþi ‘‘half-bonds’’, ei ; ei A 1 . . . NZ

The NA functional (or end) groups A i will be distinguished (even if chemically sim-
ilar) according to the monomer unit Xh i where they are attached. h is yet another
vector of indices, with size NA .
NR reactions involving pairs of functional groups create connections between
monomer units, possibly (with the well-known exceptions of epoxides and isocya-
nate reactions) also forming by-products. A total number NW of by-products Wi
will be considered for the sake of generality. By-product We i is formed by the reac-
tion between functional groups A gi and A giþ creating the bond ZiR .
Vectors e; gþ and g have sizes NR . This definition allows for more than one pos-
sible kind of bond between two given monomer units, as happens for instance if a
carboxylic acid reacts with glycerol, which possesses distinguishable primary and
secondary hydroxyls.
In some of these NR reactions, such as in the case of self-condensations of sila-
nols in silicone formation and of methylols in formaldehyde polymerizations, an
end group may react with itself. The number of such reactions in this subset will
be defined as NRs .
For each bond, it is useful to define a positive sense in the direction of the unit
with higher or equal index, which will be coincident for the NRs reactions consid-
ered above. So, there is a total number of NZ ¼ 2NR  NRs kinds of directed bonds
Z i , incident on monomer units Xzi ; the set of monomer units and bonds are the
vertices and edges of a directed graph (or digraph). Vector z, of size NZ , therefore
contains the indices of the repeating units to which each directed bond points.
There is a one-to-one correspondence between directed bonds and half-bonds
hanging from the repeating units at each side.
The directed bonds associated with ZiR will be named Zei and Zeþi , using yet an-
other pair of vectors of indices, eþ and e , of sizes NZ . Hence, the vectors z and h,
defining respectively their incident and adjacent monomer units, will be related
through hgiþ ¼ zeþi and hgi ¼ zei .
In order to avoid multiple levels of indexing in equations, the notation given by
Eqs. (62) (loosely inspired by indirect addressing of computer assembly languages)
will be used hereafter.

A giþ 1 A½iþ Zeþi 1 Z½iþ We i 1 W½i Xh i 1 X½iA Xzi 1 X½iZ ð62Þ

3.4 Modeling of Complex Polycondensation Reactions 115

For the NR reactions which create new bonds from functional groups and also (very
often) a by-product, we will introduce apparent second-order rate constants of the
forward reaction (ki , i ¼ 1; NR ), as well as apparent first-order rate constants of the
backward reactions (kiZ ), related to the former through the equilibrium ratios Ki
and the concentration of the corresponding by-product, if it exists, according to
Eq. (63).

kiZ ¼ ki ð63Þ

This may look rather artificial, but it helps in situations such as the thermal de-
composition of urethanes or ureas, which are first-order reactions. By convention,
in such cases we will introduce a nil by-product W0 (assuming that the indices are
counted starting from one) with a constant unit concentration.
If the groups react independently, by definition there is no substitution effect. The
condensation reaction creating a bond (the same as two half-bonds) can be written
as Eq. (64).
A½i þ A½iþ T Z½i þ Z½iþ þ W½i ð64Þ

To account for first-shell substitution effects (FSSEs), a more general expression

[Eq. (65)] can be written, using stoichiometric coefficients n:

ðninA þ ninAþ ÞAn þ nW
in Wn þ ðninZ þ ninZþ ÞZn ¼ 0 ð65Þ
n¼1 n¼1 n¼1

ninA ; ninZ correspond, respectively, to functional groups An and oriented bonds Zn

coming out of the monomer unit in which stood A½i , and a similar convention is
used for ninAþ ; ninZþ.
Aþ Zþ
No FSSE means that, for every reaction i, nigA Z
 ¼ n þ ¼ 1, nie ¼ n þ ¼ 1, and,
igi iei
i i
moreover, all other stoichiometric coefficients are nil. Likewise, it is useful to intro-
duce the stoichiometric coefficients of by-products, nW in .
This trick only works with FSSEs. Higher-order substitution effects (see Sections
3.1.5 and 3.4.5) are much more difficult to describe.
In the absence of FSSEs, the exchange of bonds ZiR and ZjR , necessarily forming
the same by-product (e i ¼ ej ), can be described through Eqs. (66).
X  A½i þ X þ Z½ j Z½ jþ X þþ !  þ
 X Z½i Z½iþ X þ A½ j X
E kji
 þ þþ !  þ þþ
X A½iþ þ X Z½ j Z½ jþ X  X Z½i Z½iþ X þ A½ jþ X


In the presence of FSSEs, they would be written with an algebraic notation as in

Eqs. (67).
116 3 Polycondensation

ðnijn þ nijn þ njin ÞAn þ ðnijn þ nijn þ nijn ÞZn ¼ 0
n¼1 n¼1
AEþ AEþþ AEþþþ ZEþ ZEþþ ZEþþþ
ðnijn þ nijn þ nijn ÞAn þ ðnijn þ nijn  nijn ÞZn ¼ 0
n¼1 n¼1

Stoichiometric coefficients nijn ; nijn are the changes in numbers of functional
groups An and bonds Zn , respectively, connected to the root unit X  where either
AEþþ ZEþþ AEþþþ ZEþþþ
A½i or A½iþ were attached, and nijn ; nijn ; nijn ; nijn are the changes in
numbers of functional groups An and bonds Zn , respectively, connected to the root
unit X þþ where stood the living group A½ j or A½ jþ. The other stoichiometric coef-
AEþ ZEþ AEþþ ZEþþ
ficients nijn ; nijn ; nijn ; nijn are the changes in the numbers of groups
attached to root unit X þ which gets connected to the unit where stood the attack-
ing group, and they are nil if there are no substitution effects.
The above reactions modify the counts of functional groups of similar chemical
nature (for example, distinguishable kinds of amides/amines/carboxylic acids). If
there is only a single kind of bond, there is no net creation or destruction of func-
tional groups or bonds, as shown by the cancellation of the stoichiometric coeffi-
cients in Eq. (67), but a reshuffling of pieces of the reacting molecules takes place.
Rates of production by chemical reaction of the various groups are obtained us-
ing Eqs. (68)–(70).

R An ¼ ðninA þ ninAþ Þðki ½A½i ½A½iþ   kiZ ½Z½iþ Þ

þ de ij ½kijE ½A½i ½Z½ jþ ðnijn
þ nijn EAþþ
þ nijn Þ
i¼1 j¼iþ1

þ kjiE ½A½ j ½Z½iþ ðnjin

þ njin EAþþ
þ njin Þ ð68Þ

R Zn ¼ ðninZ þ ninZþ Þðki ½A½i ½A½iþ   kiZ ½Z½iþ Þ

þ de ij ½kijE ½A½i ½Z½ jþ ðnijn
þ nijn ZAþþ
þ nijn Þ
i¼1 j¼iþ1

þ kjiE ½A½ j ½Z½iþ ðnjin

þ njin ZAþþ
þ njin Þ ð69Þ

RWn ¼ k i nW Z
in ð½A½i ½A½iþ   ki ½Z½iþ Þ ð70Þ

The reaction volume changes mostly because of by-product removal, and little be-
cause of density changes. The relative rate of change of reaction volume RV can be
3.4 Modeling of Complex Polycondensation Reactions 117

estimated as the sum of products of the molar volume of by-products by their rate
of elimination by phase change.
The mass balances of the functional groups in a batch reactor can thus be writ-
ten as Eqs. (71).

¼ RAn  RV ½An 
¼ RZn  RV ½Zn 

A convenient way of computing the concentrations of groups at chemical equilib-

rium consists in integrating the system of ODE [Eq. (71)] until close to steady

Equilibrium Polycondensations with Several Monomers

Instead of the elegant but often error-prone Gordon’s notation, we will introduce a
more straightforward description, hopefully easier to translate into computer pro-
grams. The goal is to obtain a set of formulae for predicting average molecular
weights, molecular weight distributions, and other polymer properties, valid for ge-
neric chemical systems.
These computations assume there is some way of predicting how molecular
fragments (usually monads, unless higher-order substitution effects have to be
tackled) are mutually connected. More specifically, it is necessary to know the dis-
tributions of the numbers of bonds connecting the fragments, for each kind of frag-
ment. In fact, one may need only some of the moments of the aforementioned dis-
tributions for making a few calculations.
A kinetic method may be used for this prediction, but mainly as a means to
avoid solving equations derived from mass action laws for concentrations of frag-
ments. This has been one of our motivations for presenting the formalism in Sec-
tion 3.4.2.
Each directed bond Z i is supposed to start a pendent chain Vi ðxZ ; xA Þ with
counts of end groups and directed bonds xA and xZ , respectively. Notice that the
molecular graphs have to be considered as digraphs, otherwise xZ would be mean-
ingless: it would be impossible to know the counts of the monomer units accord-
ing to their chemical nature.
All isomeric trees with the same counts of groups are lumped into the same
chemical species leading to vector count distributions with NZA ¼ NZ þ NA inde-
pendent variables.
Vectors of dummy Laplace variables sA and sZ will be associated with the counts
of unreacted groups and directed bonds. Variables sA and sZ will be often ranged
together as subvectors of a vector s, of size NZA .
118 3 Polycondensation

Vector xX containing the counts of the monomer units can be obtained from xZ
through Eq. (72).

xX ¼ ðZ ZX Þ t xZ ð72Þ

ZZX is a matrix containing the incidence vectors z defined by Eq. (73).

1 if j ¼ zi
ZijZX ¼ ð73Þ
0 if j 0 zi

The chemical system is further described through knowledge of the molar frac-
tions of monomers or monomer units, yX i , (summing to 1) and of the molecu-
lar weights of monomer units, unreacted groups, and half-bonds, respectively
MX i ; MA i and MZ i . Hence, the molecular weight M½Vi ðxZ ; xA Þ of a tree Vi ðxZ ; xA Þ
can be computed through Eq. (74).

M½Vi ðxZ ; xA Þ ¼ ðMX½ jZ þ MZ j ÞxZ j þ MA j xA j ð74Þ
j¼1 j¼1

X i ðxZ ; xA Þ is, according to the concept introduced in Section 3.1.5, a monad with
vectors of group counts xZ and xA . Its concentration, normalized by the concentra-
tion of repeating units X i , can be thought of as a probability: the probability that a
certain repeating unit is attached to those counts of groups, as stated in Eq. (75).

PfX i is connected to xZ ; xA groupsg ¼ ½X i ðxZ ; xA Þ/½X i  ð75Þ

Let F X i ðsZ ; sA Þ; F A i ðsZ ; sA Þ and F Z i ðsZ ; sA Þ be the probability generating functions

(pgf ) of the counts of the different kinds of connecting and unreacted functional
groups directly linked to a unit X i , an unreacted group A i or a directed bond Z i
[Eq. (76)].

y X
y X
y X
F J ðsZ ; sA Þ ¼  
xZ1 ¼0 xZN ¼0 xA1 ¼0 xAN ¼0

xZ xAN
PfJ is connected to xZ ; xA groupsgsZ1 1    sAN A
xAN ¼0

J ¼ Xi; Ai; Zi ð76Þ

These pgf values are mutually related through Eqs. (77) and (78), in which 1N
means a vector with N components equal to 1.

qF X½ jA qF X½ jA X X
F ðsÞ ¼ s1
Aj ¼ s1 ½ jA ½ jA
A j LA j ðsÞ/lA j ð77Þ
q log sA j q log sA j js¼1
3.4 Modeling of Complex Polycondensation Reactions 119

qF X½ jZ qF X½ jZ X X
F Z j ðsÞ ¼ s1
Zj ¼ s1 ½ jZ ½ jZ
Z j LZ j ðsÞ/lZ j ð78Þ
q log sZ j q log szj js¼1

Notice the use of L with lower indexes for the derivatives of the pgf values with
respect to the logarithms of dummy Laplace parameters, as well as of l for their
moments, a useful convention which will be encountered often in the rest of this
If a pgf relative to the count of monomer units is desired, for a vector of dummy
Laplace variables sX , it can be found by obtaining the pgf with respect to the counts
of directed bonds with Eq. (79).

sZ ¼ ZZX sX ð79Þ

The average numbers of bonds Z j ; lZX ij , and of unreacted functional groups A j ; lAX ij ,
attached to a monomer unit X i , will be often needed, and can be obtained using
Eqs. (80) and (81).

qF X i
lZX ij ¼ ð80Þ
q log sZ j js¼1

qF X i
lAX ij ¼ ð81Þ
q log sA j js¼1

The mass of polymer per mole of monomer units, MP , can therefore be computed
using Eq. (82).

" #
MP ¼ y X i MX i þ lZX ij MZ j þ lAX ij MA j ð82Þ
i¼1 j¼1 j¼1

For the classic self-polycondensation of XA f , with equal and independent groups

A; p, the conversion of A groups, is introduced in Eqs. (83).

F X ¼ ½ð1  pÞsA þ psZ  f

F Z ¼ F A ¼ ½ð1  pÞsA þ psZ  f 1

lZX ¼ f p lAX ¼ f ð1  pÞ

MP ¼ MX þ f pMZ þ f ð1  pÞMA

Expressions for the less trivial case XA f þ YBg C illustrated diagrammatically

in Figure 3.12, A reacting with B or C, B not reacting with C (such as adipic
acid þ glycerol, f ¼ g ¼ 2), with constant reactivity of end groups A; B, or C, are
presented in Table 3.3.
120 3 Polycondensation

Fig. 3.12. Example of proposed notation: polycondensation XA2 þ YB2 C.

Tab. 3.3. Probability generating functions F describing polycondensation XA f þ YBg C and its
gelation condition.
" # " #
X F XX ½ð1  pA ÞsA þ pAB sZAB þ pAC sZAC  f
F ¼
F XY ½ð1  pB ÞsB þ pB sZBA  g ½ð1  pC ÞsC þ pC sZCA 
2 3 2 3
F ZAB ½ð1  pB ÞsB þ pB sZBA  g1 ½ð1  pC ÞsC þ pC sZCA 
6 ZAC 7 6 ½ð1  pB ÞsB þ pB sZBA  g 7
6F 7 6 7
FZ ¼6 7 6 f 1 7
4 F ZBA 5 4 ½ð1  pA ÞsA þ pAB sZAB þ pAC sZAC  5
f 1
F ZCA ½ð1  pA ÞsA þ pAB sZAB þ pAC sZAC 
2 AA 3 2 3
F ½ð1  pA ÞsA þ pAB sZAB þ pAC sZAC  f 1
6 A 7 6 7
F ¼4 F B 7
A 6
5 4 ½ð1  pB ÞsB þ pB sZBA  g1 ½ð1  pC ÞsC þ pC sZCA  5
F AC ½ð1  pB ÞsB þ pB sZBA 
Gelation condition gpBg pACg þ ðg  1Þ pBg pABg þ pCg pABg ¼
f 1

Group counts have multinomial distributions for these simple systems and can
be easily related to conversions of functional groups (which are the probabilities of
The theory of branching processes leads to a system of NZ algebraic equations
[Eqs. (84)] for the pgf of pendent trees of the different kinds:

Vi ðsZ ; sA Þ ¼ sZ i F Z i ½VðsZ ; sA Þ; sA  i ¼ 1; . . . ; NZ ð84Þ

The vector v of the NZ probabilities of extinction (the fractions of finite pendent

chains) v ¼ Vð1NZA Þ contains the solutions of system (84) for s ¼ 1NZA .
Gelation occurs when system (84) has a double root v ¼ 1NZA for s ¼ 1NZA, imply-
ing that its Jacobian becomes nil [Eq. (85)].
" #
qF Z i

 I ¼ j½LzZj i ð1NZA Þ  Ij ¼ 0 ð85Þ

For the polycondensation of a single monomer, this leads to the well-known result
of Eq. (86).
3.4 Modeling of Complex Polycondensation Reactions 121

pg ¼ ð86Þ
f 1

Prediction of the gel point for XA f þ YBg C is also presented in Table 3.3.
The pgf values of trees starting with a prescribed monomer unit X i , an unreacted
group A i , or a directed bond Z i , are obtained from the theory of branching pro-
cesses through Eqs. (87) and (88).

GYi ðsZ ; sA Þ ¼ sYi F Yi ½VðsZ ; sA Þ; sA  Y ¼ Z; A ð87Þ

G X i ðsx ; sA Þ ¼ sX i F X i ½VðZZX sX ; sA Þ; sA  ð88Þ

The chain length or molecular weight distribution of polymer is usually described

primarily with the help of the molar concentration of species with x monomer
units, ½Px , in which x is the degree of polymerization. Since we are dealing with
several kinds of units, an overall degree of polymerization can be defined, which is
the sum of the degrees of polymerization corresponding to the various monomer
units (the components of vector xX ). Notice that vector xZ with the counts of di-
rected bonds has more information about the molecular composition.
There is a basic difficulty when trying to use Eqs. (87) and (88): they provide pgf
values of monomer units or groups, not directly molar concentrations of polymer
molecules, even as generating functions. These latter will have to be computed rel-
ative to the molar concentrations of the various groups or monomer units, and
thus will come multiplied also by the number of those groups in the distributions.
For instance, if there is only one kind of monomer unit and so just one kind of
directed bond, G X as computed by Eq. (88) will provide the generating function of
x½Px /½X  with respect to x; for an alternating polycondensation of two monomers,
G X 1 ðsZ1 ; sZ2 Þ as computed by Eq. (88) will provide the generating function of
x1 ½Pðx1 ; x2 Þ/½X1  with respect to x1 and x2 .
Fractions of units and groups of the various kinds in finite molecules (in sol),
after gelation, ySX i ; ySZ i and ySA i , can be computed by replacing s ¼ 1NZA in the
above equations to give Eq. (89).

ySYi ¼ F Yi ðv; 1NA Þ Y ¼ X; Z; A ð89Þ

Prediction of the elastic properties of networks using rubber elasticity theory is

based upon the knowledge of concentrations of elastically active network junctions
(EANJs) and chains (EANCs), respectively me and ne [260, 261]. EANJs are the inter-
section of at least three chains leading to the gel, whereas EANCs are the chains
linking EANJs (see Figure 3.13).
These concentrations can be easily predicted, given the probabilities of extinction
and the moments with respect to the numbers of pendent chains as previously de-
fined. Defining xZy as the count of infinite pendent chains and the correspondent
dummy Laplace variable as sZy , its pgf for the chains stemming out of a monomer
unit X i is F X i ½v þ sZy ð1NZ  vÞ; 1NA , and so me and ne can be computed using Eqs.
(90) and (91).
122 3 Polycondensation

Finite pendent chain

Infinite pendent



Chain connecting to gel

Fig. 3.13. Elastically active and inactive junctions and chains in a polymer network.

me ¼ ½X i  1  F X i ðv; 1NA Þ  nj ð1  nj ÞLZXji ðv; 1NA Þ
i¼1 j¼1
 nj nk ð1  nj Þð1  nk ÞLZ j Zk ðv; 1NA Þ ð90Þ
2 j¼1 k¼1
ne ¼ ½X i  ½lZX ij  nj ð1  nj ÞLZXji ðv; 1NA Þ
2 i¼1 j¼1
 nj nk ð1  nj Þð1  nk ÞLZXjiZk ðv; 1NA Þ ð91Þ
j¼1 k¼1

For instance, assuming the so-called ‘‘phantom network’’ model, shear modulus
Ge would be predicted for gaussian chains to be given by Eq. (92).

Ge ¼ RTðne  me Þ ð92Þ

In the presence of gel, it is convenient to introduce the pgf of finite pendent chains,
named V^i ðsZ ; sA Þ, and the pgf of counts of finite pendent chains connected to units
or groups, which can be found using Eqs. (93) and (94).
3.4 Modeling of Complex Polycondensation Reactions 123

F^Yi ðsZ ; sA Þ ¼ F Yi ðn1 s Z1 ; . . . ; nNZ sZNZ ; sA Þ/ySYi ¼ F Yi ðv5sZ ; sA Þ/ySYi

Y ¼ X; Z; A ð93Þ
V^i ðsZ ; sA Þ ¼ sZ i F^Z i ½V
^ ðsZ ; sA Þ; sA  i ¼ 1; NZ ð94Þ

So, the various pgf values with respect to the different kinds of groups in the
^ X i ðsZ ; sA Þ; G
molecules of the sol, G ^Z i ðsZ ; sA Þ; G
^A i ðsZ ; sA Þ, can be computed using
Eqs. (95).

^Yi ðsZ ; sA Þ ¼ F^Yi bV

G ^ ðsZ ; sA Þ; sA c Y ¼ X; Z; A ð95Þ

After computing the probabilities of extinction, the moments in Eq. (96) can be

l^YZij ...Zk ¼ LYZi j ...Zk ðv; 1NA Þ/ySYi Y ¼ Z; A ð96Þ

Pgf values of finite pendent chains with respect to molecular weight, for which the
dummy Laplace variable associated with molecular weight is sM , can be found from
Eq. (97) (notice the conventional use of a power of a scalar to a vector):
V^Mi ðsM Þ ¼ V^i ðsM 1 X½1Z ; . . . ; sM Z ; sM 1 ; . . . ; sM A
Þ ¼ V^i ðsM z ; sM Þ

Prediction of average molecular weights is now possible by introducing GM ðsM Þ,

the pgf values of the mass fractions of the polymer molecules in the sol, wi , with
respect to their molecular weight Mi , defined below; the index i in the infinite sum
sweeps all finite polymer molecules.

GM ðsM Þ ¼ sM wi ¼ wYi sM i F^Yi ½V
^ M ðsM Þ; sM  ð98Þ
i¼1 Y ¼X; Z; A i¼1

The mass fractions of the units and groups in Eq. (98) above are relative to the
mass of the sol.
The weight fraction of the sol wS , relative to the overall mass of the polymer
computed by Eq. (82), is therefore given by Eqs. (99)–(102).
wS ¼ y X i ySX i MX i þ l^ZX ij MZ j þ l^AX ij MA j MP ð99Þ
i¼1 j¼1 j¼1

y X i ySX i MX i
wX i ¼ ð100Þ
MZ i y X j ySX j l^Z ji
wZ i ¼ ð101Þ
124 3 Polycondensation

MA i y X j ySX i l^A ji
wA i ¼ ð102Þ

Before carrying out the evaluation of distributions and average molecular weights,
a special reasoning must be carried out in order to compute number-average mo-
lecular weight and degrees of polymerization. When it is taken into account that,
with the reaction of every pair of end groups in finite molecules, one polymer mol-
ecule is consumed, the number of moles of polymer molecules per mole of mono-
mer units before gelation is given by Eq. (103).

yP ¼ 1  yXi lZX ij ð103Þ
2 i¼1 j¼1

After gelation, the more general expression [Eq. (104)] is needed.

yP ¼ y X i ySX i 1 l^X i ð104Þ
2 j¼1 Z j

The number-average molecular weight of the sol can thereafter be computed

through Eq. (105).

y X i ySX i MX i þ MZ i l^ZX ij þ MA i l^AX ij
i¼1 j¼1 j¼1
Mn ¼ ! ð105Þ
y X i ySX i 1 l^X i
2 j¼1 Z j

An expression for the number-average degree of polymerization of the sol follows

from Eq. (105) by setting the molecular weights of monomer units equal to one
and the molecular weights of bonds and unreacted groups equal to zero [Eq. (106)].

y X i ySX i
xn ¼ ! ð106Þ
y X i ySX i 1 l^X i
2 j¼1 Z j

The moments with respect to molecular weight lM ; lMM , and so on, can be now
obtained through differentiation of Eq. (98) with respect to log sM and setting
sM ¼ 1. First of all, the systems of linear algebraic equations (107)–(113) must be
3.4 Modeling of Complex Polycondensation Reactions 125

½miZ  ¼ ½di  l^ZZ ij 1 ½MZ i þ MX½iZ  ð107Þ
½miA  ¼ ½di  l^ZZ ij 1 ½l^ZAji ½MA i  ð108Þ
" #
½miZZ  ¼ ½di  l^ZZ ij 1 2ðMZ i þ MX½iZ Þ mZl l^ZZli þ Z ^Z i
mZl mm lZl Zm ð109Þ
l ¼1 l ¼1 m¼1
" #
j Z i 1
½miZA  ¼ ½di ^
 lZ j  ðMZ i þ MXX½iZ Þ ^
MAm lAk þ Z ^
m l MAm lZl Am ð110Þ
m¼1 l¼1 m ¼1
" #
½miAA  ¼ ½di  l^ZZ ij 1 mlA MAm l^ZZliAm þ MAl MAm l^AZliAm ð111Þ
l¼1 m¼1 l¼1 m¼1
½miMZX  ¼ ½di  l^ZZ ij 1 ½ðMX½iZ þ MZ i Þ 2  ð112Þ
½miMA  ¼ ½di  l^ZZ ij 1 ½MA2 i  ð113Þ

Weight-average and z-average molecular weights are now obtained explicitly

through Eqs. (114) and (115).

lM ¼ ¼ Mw
q log sM jsM ¼1
" #
A ^Yi Y
¼ w Yi M Yi þ Z
ðmj þ mj ÞlZ j þ MA j l^A ji ð114Þ
Y ¼X; Z; A i¼1 j¼1 j¼1

lMM ¼ Mw MZ
( " !
¼ w Yi MY2i þ 2MYi ðmjZ þ mjA Þl^YZij þ MA j l^YZij
Y¼X; Z; A i¼1 j¼1 j¼1

þ ðmjZ þ mjA ÞðmkZ þ mkA Þl^YZij Zk
j¼1 k¼1

þ ðmjZZ þ 2mjZA þ mjAA þ mjMXZ þ mjMA Þl^YZij

þ2 ðmjZ þ mjA ÞMAk l^YZij Ak
j¼1 k¼1
þ MA j MAk l^AYjiAk þ MA2 j l^AYji ð115Þ
j¼1 k¼1 j¼1

Except for extremely simple polycondensations, formulae for predicting average

molecular weights are very cumbersome and numerical evaluation is a must.
126 3 Polycondensation

The weight-average and z-average degrees of polymerization, x w and x z , are

obtained in the same way as for number-average degrees of polymerization: the
molecular weights of the monomer units are set equal to one and the molecular
weights of bonds and unreacted groups are set equal to zero. For the polyconden-
sation of a single monomer XA f , this leads to Eqs. (116) and (117).

1 þ pð2v  1Þ
xw ¼ ð116Þ
1  p½1 þ vð f  2Þ

2f pv þ f p 2 vð f  1Þð1  p þ pvÞ f 2
xz ¼ 1 þ
ð1  p þ 2pvÞ½1  pð f  1Þð1  p þ pvÞ f 2 

ð f  1Þð f  2Þ p 2 v 2
þ ð117Þ
ð1  p þ 2pvÞ½1  pð f  1Þð1  p þ pvÞ f 2  2

The number-average degree of polymerization x n is obtained through a stoichio-

metric reasoning as previously discussed [Eq. (118)].

1  p þ pv
xn ¼ ð118Þ
1  p þ pvð1  f /2Þ

Values of average degrees of polymerization versus conversion of end groups p

shown for f ¼ 2 and f ¼ 3 in Figure 3.1 have been computed using the above
Pgf values of the various distributions with respect to the counts of monomer
units, bonds, or unreacted functional groups can be obtained from Eq. (98) by set-
ting equal to one the molecular weight of the species in question and to zero the
molecular weights of all the other species. Analytical inversion by computing deriv-
atives with respect to dummy Laplace variables on s ¼ 0 is feasible with the sim-
plest chemical systems, for which the resulting recurrence formulae are not too
complex, as happens with the Stockmayer distribution.
Numerical inversion of generating functions of the concentration distributions is
usually a better way to predict them. Unless the cost of evaluating GðsÞ ¼ s x ½Px 
x ¼0
is too high and more sophisticated methods based upon Laplace transform inver-
sion are needed, an accurate evaluation can be obtained using the method inde-
pendently developed by Mills [262] and ourselves [263] (see also Ref. 264 for a thor-
ough analysis of round-off and truncation errors of similar approaches). It consists
in computing the inversion contour integral on a circle C in a complex plane cen-
tered on the origin with radius jsj slightly below 1, using the trapezium rule and
Fast Fourier Transform for evaluating the sums [Eq. (119)].

jsjx X
N 1
2pimx Xy
½Px  ¼ exp  Gðsm Þ  ½PxþNn jsj Nn ð119Þ
N m¼0 N n¼1
3.4 Modeling of Complex Polycondensation Reactions 127

The second term in Eq. (119) can be neglected for large enough N or small enough
jsj. A recent surge on this approach has led to exploitation of other more complex
but hopefully more efficient methods [265], mainly developed for Laplace trans-
form inversion, and so more adapted to high average molecular weights. The error
of those formulae is more complex to control, unlike Eq. (119).
Besides molecular weight distribution, it is also possible to access readily some
information about the distribution of molecular sizes and other polymer proper-
ties, such as the angular dependence of light scattering intensity. The evaluation
of averages involving the distances of every pair of monomer units is required,
and a starting point for that purpose is the evaluation of the trail generating func-
tions [31–33], allowing the counting of path lengths.
Equilibrium polycondensation of a single monomer XA f taking FSSEs into
account has been analyzed by Gordon and Scantlebury [268] using TBP, and exper-
imental results concerning the POCl3 /P2 O5 system have been successfully de-
scribed. More general calculations are better carried out using the method de-
scribed by Kuchanov et al. [267], summarized below for the polycondensation of
XA f (but only in the absence of gel).
Chemical equilibrium will be attained in two hypothetical stages:

1. All the rings are formed, but no fused rings, such as naphthalene, are allowed
and molecules look like ‘‘cactus’’ [266]. The fraction of repeating units X in
rings of size n ðn ¼ 1; yÞ at the end of this stage is assumed to be y Xc ðnÞ, sum-
ming 1  yc , which will be found afterwards from mass action laws. No other
reactions of functional groups A occur.
2. The unreacted monomer and the rings start a polycondensation with an infinity
of monomers with a single group A and different functionalities, which are f for
the unreacted monomer coming from stage 1 and nð f  2Þ  1 for the rings.
Defining sCn as the dummy Laplace variables associated to the counts of rings
(including sC0 for the count of units X in the chains connecting rings) and sZl
as the variable counting the bonds not belonging to rings, the generating func-
tion of the trees with either a ring or a unit not belonging to any ring can be
found from TBP as shown in Eqs. (120).

V ¼ F Zl ðVÞ ¼ b 0 sC0 ½sA ð1  ar Þ þ ar V f 1 þ b n sCn ½sA ð1  ar Þ þ ar V nð f 2Þ1

G ¼ ð1  yc ÞsC0 ½sA ð1  ar Þ þ ar V f þ y Xc ðnÞsCn ½sA ð1  ar Þ þ ar V nð f 2Þ

f ð1  yc Þ ð f  2Þ y Xc ðnÞ
b0 ¼ ; bn ¼ ð120Þ
f  2yc f  2yc

A pgf for the counts of repeating units X will result from substituting sC0 in Eq.
(120) by sX and sCn by sXn. Variable ar in the expressions (120) is the probability of
128 3 Polycondensation

reaction of the unreacted functional groups after ring formation in stage 1. It is

necessary to eliminate it, and to introduce the mass action laws for the rings. This
last step is not straightforward, as it requires the application of graph theory in or-
der to compute the concentrations of linear polymer molecules ½L n  in Eq. (23).
The final result, which reduces to the distribution found by Jacobson and Stock-
mayer in their classic paper [21] for f ¼ 2, is Eq. (121).
½xð f  1Þ!arx1 ð1  ar Þ xð f 2Þþ2 f ð1  yc Þ x
½Px  ¼ ½X ð f  2yc Þ
x!½xð f  2Þ þ 2! f  2yc

nK c ðnÞ n f ð f  1Þar ð1  yc Þ
y Xc ðnÞ ¼ q ; q¼
½X  f  2yc

f ð p  ar Þ
nK c ðnÞq n ¼ ½X  yc ; yc ¼ ð121Þ
2ð1  ar Þ

A solution is also known for the analogous system of the alternating polycondensa-
tion XA f þ YBg [267].
Dilution with an inert solvent will make [X] decrease without affecting the cycli-
zation constants much, and the fraction of rings will increase, until it becomes
practically unity, a phenomenon which has experimental support obtained using
polysiloxanes as model polymers. In bulk systems, the fraction of rings is usually
only a few per cent.
With nonlinear polycondensations of aliphatic monomers, gel points in bulk are
affected by a few per cent due to cyclizations, and elastic properties are also af-
fected (that effect using TBP has been modeled by Dušek et al. [269]). Application
of the approach described above could lead to improved modeling where small
numbers of rings are present.
Taking cyclizations into account raises the question of introducing information
about the spatial location of atoms in models of network formation. The classic ge-
lation theory described here considers a uniform distribution in space of all the
chemical properties. Stauffer, one of the main contributors to the progress of per-
colation theory, has strongly criticized this view, claiming that Gordon’s theory is
inapplicable in the vicinity of gel point [270]. Gordon has not accepted that argu-
ment [271], claiming his theory to be universal and capable of describing gel for-
mation in any homogeneous system.
A new theory, starting with classic gelation theory and using TBP as a particular
case, has been developed by Kuchanov [272, 273], and considers the molecular
graphs to be embedded in ordinary three-dimensional space – not in a lattice, as
is usually done in percolation theory. Generating functions are replaced by generat-
ing functionals of the ensemble of positions of the functional groups and repeating
units. Dependence of space coordinates is eventually eliminated by averaging, so
the algebraic equations become integral equations. In spite of its power, it has not
become a widely used instrument for dealing with ‘‘real’’ complex chemical sys-
3.4 Modeling of Complex Polycondensation Reactions 129

tems, and we leave it here only for reference, as even a basic but comprehensible
description would be too extensive.

Kinetic Modeling of Irreversible Polycondensations

In his pioneering study of nonlinear polycondensation [7], Stockmayer has already

checked his statistical solution [Eq. (5)] by solving the mass balance equations in a
batch reactor for the concentrations of functional groups A and the set of isomeric
polymer molecules Px with x repeating units X [Eqs. (122)].

d½Px  1Xx1
¼k ½ yð f  2Þ þ 2½ðx  yÞð f  2Þ
dt 2 y¼1
þ 2½Py ½Pxy   ½A½xð f  2Þ þ 2½Px  ð122Þ

¼ k½A 2 ; ½Px jt¼0 ¼ ½X ; ½Ajt¼0 ¼ f ½X 

The two solutions are identical. Hence, for a long time no importance was attrib-
uted to the use of a kinetic approach for describing batch polycondensations start-
ing from monomers, and the statistical approach was preferred. Of course, chemi-
cal engineers had to deal with semi-batch and continuous stirred tank reactors
where the statistical approach, although possible, is cumbersome and error-prone,
so a few papers appeared in the 1960s dealing with kinetically controlled linear
polycondensations [274–276].
In reality, Kuchanov [277, 278] has shown that,with polycondensations present-
ing FSSEs, kinetic and statistical approaches give distinct results for average mo-
lecular weights and gel points. Dušek [279] has pointed out that this behavior is
even more visible when dealing with polyadditions (linear polyaddition leads to a
Poisson CLD, whereas a simplistic use of a statistical approach would lead to a
geometrical/Schulz–Flory CLD) and has confirmed this result using Monte Carlo
simulation of XA f polycondensation with FSSEs [280]. Sarmoria and Miller [35]
have tried to extend the ‘‘recursive approach’’ to systems presenting FSSE by con-
sidering network building starting from dyads and larger fragments. But chemical
systems can be found in which this latter idea does not provide useful results
[281], so that use of statistical approaches outside the description of chemical equi-
librium now seems more like a waste of time.
Kuchanov’s kinetic approach divides polymer molecules into classes PðxÞ having
a vectorial count of groups x. In this approach, ‘‘groups’’ An should include not
only the unreacted functional groups, but also the bonds and repeating units, and
even larger molecular fragments when needed. We will use NXZA as the number of
kinds of groups in that generalized sense.
130 3 Polycondensation

It is possible to obtain a rate equation for the members of each class by adding
the contributions of the various condensation reactions, leading to a version of
Smoluchowski’s coagulation equation.
Ring-forming reactions involving functional groups in the same monomer can
be described by a simple extension of the preceding FSSE model, just by consider-
ing unimolecular reactions and new fake functional groups, which are pairs of
groups [Eq. (123)].

ki X

A½i   ! nin An i ¼ 1; NR ð123Þ


The rate of formation of groups requires a modification of Eq. (68), since it is no

longer possible to include breakage or exchange reactions. Adding the unimolecu-
lar reactions defined above, the new general rate equation becomes Eq. (124).

R An ¼ ðn
in þ n þ
in Þki ½A½i ½A ½iþ  þ nin ki ½A½i    ð124Þ
i¼1 i¼1

The multiple sums in the rate of formation of PðxÞ, which will not be presented,
are simplified through consideration of its generating function [Eqs. (125), (126)]

qG qG qG qG
GRP ðsÞ ¼ ki G þ
ni Gni  ½A½iþ   ½A½i 
q log s½i q log s½iþ q log s½i q log s½iþ

X qG
þ ki ðG   1Þ ð125Þ
q log s½i   ni


Gni ¼ snin J ¼ þ; ;  ð126Þ

Equation (125) replaces a similar expression in Ref. 282 with the advantage of
considering only the reactions actually taking place, and not every combination of
pairs of unreacted groups, which does not make sense now, as repeating units are
also An moieties.
Insertion of Eq. (125) into mass balance equations, such as a continuous stirred
tank reactor (CSTR) with constant volume, leads to a nonlinear first-order PDE
[Eq. (127)].
3.4 Modeling of Complex Polycondensation Reactions 131
qG qG qG qG
¼ ki G G þ
ni ni  ½A½iþ   ½A½i 
qt i¼1
q log s½i q log s½iþ q log s½i q log s½iþ

X qG GF ðsÞ  GðsÞ
þ ki ðGni  1Þ  RV G þ
q log s½i 

GðsÞjt¼0 ¼ G0 ðsÞ ð127Þ

Solution of the above equation by the method of characteristics [283] is described

in Ref. 282, earlier examples being found in Refs. 284–286. They will not be repro-
duced here, for the sake of brevity. If GðsÞ is to be evaluated, in order to take advan-
tage of the numerical inversion formula Eq. (119), or if average degrees of polymer-
ization in the presence of gel have to be predicted, numerical solution of Eq. (127)
leads to a two-point boundary solution problem with twice as many unknowns
as the number of derivative terms log sn (the number of active groups in the
A much simpler problem, as Galina was apparently the first to remark [287], is
the prediction of the moments of ½Px  with respect to the counts of groups, lmn... , in
the absence of gel. Differentiation of Eq. (126) with respect to log sn and setting
s ¼ 1NZXA leads to an ODE system with known initial conditions, which has a
straightforward numerical solution [Eq. (128)].

dljkP X
¼ k m ½ðn þ  þ
mk þ nmk Þðnmj þ nmj Þ½A½m ½A½mþ 
dt m¼1

þ ðn þ P P
mj þ nmj Þðl½mk ½A½mþ  þ l½mþk ½A½m Þ

þ ðn þ P P P P P P
mk þ nmk Þðl½m j ½A½mþ  þ L½mþ j ½A½m Þ þ l½m j l½mþk þ l½mþ j l½mk 

X ljkPF  ljkP
þ km ðl½m
P  P   
  j nmk þ l½m  k nmj þ ½A½m   nmj nmk Þ þ  RV ljkP ð128Þ

As there is no gel, a rate equation for the overall concentration of polymer ½P
(zeroth-order moment) is obtained from Eq. (126), setting s ¼ 1NZXA [Eqs. (129)].

q½P XNR ½Pf  ½P

¼ ki ½A½i ½A½iþ   RV ½P þ
qt i¼1

½Pjt¼0 ¼ ½P0 ð129Þ

Number-average and weight-average molecular weights are found by the well-

known expressions (130).
132 3 Polycondensation

MAn ½An  MAm MAn lmn
n¼1 m¼1 n¼1
Mn ¼ Mw ¼ ð130Þ
½P X

MAn ½An 

This approach can only deal with ring-forming reactions either for a limited num-
ber of the smallest rings, or alternatively, for linear polycondensations. The impor-
tant practical case of the irreversible polycondensation of AXA þ BYB þ BYC (C
being an inert group) leads to the rate laws in Eqs. (131) for the molecules with
the six possible combinations of end groups PnAA ; PnAB ; . . . PnCC and rings Cn , where
index n counts the most frequent kind of repeating units in the molecule:

n1 X
RPnAA ¼ k 4 ½PmAA ½Pnm
 þ2 ½PmAA ½Pnm
  2½PnAA ½B
m¼1 m¼1
n1 X
RPnBB ¼ k 4 ½PmBB ½Pnm
 þ 2 ½PmBB ½Pnm
  2½PnBB ½A
m¼1 m¼1
" #
n X
RPnAB ¼ k 4 ½PmAA ½Pnmþ1
 þ ½PmAB ½Pnm
  ½PnAB ð½A þ ½BÞ  kc ðnÞ½PnAB 
m¼1 m¼1

RCn ¼ kc ðnÞ½PnAB 
n X
RPnAC ¼ k 2 ½PmAA ½Pnmþ1
 þ ½PmAB ½Pnm
  ½PnAC ½B
m¼1 m¼1
n1 X
RPnBC ¼ k 2 ½PmBB ½Pnm
 þ ½PmAB ½Pnm
  ½PnBC ½A
m¼1 m¼1

RPnCC ¼ k ½PmAC ½Pnm


RA ¼ RB ¼ k½A½B  kc ðnÞ½PnAB  ð131Þ

Modern computers will not have much difficulty with the ‘‘brute force’’ approach
of solving the mass balances after inserting the above relationships for n ¼ 1 up
to an upper value N of a few hundreds or even thousands (notice that this implies
11N 2 þ OðNÞ multiplications and sums every time this set of rates of reaction is
evaluated). A relatively large value of N is needed if the mole ratio is close to one,
in order that extrapolation of CLD and evaluation of the infinite sum using the last
3.4 Modeling of Complex Polycondensation Reactions 133

equation of Eqs. (131) may be done accurately. A more elegant method uses gener-
ating functions [288] and avoids the problem of the lack of closure of the above
equations for any finite N.
When the kinetic approach was at an early stage, it was thought that it could pro-
vide no information about polymer or network properties. More recently, a descrip-
tion of batch polycondensation using TBP which is rigorously equivalent to the one
obtained by the kinetic approach was found [289], taking into account the times of
birth of molecules, so that the fundamental restriction does not hold. Through
more elementary reasonings, it is nevertheless possible to estimate probabilities
of extinction and thus network elastic properties [282] or average radius of gyration

Kinetic Modeling of Linear Reversible Polycondensations

The reversible alternating polycondensation with FSSEs in both monomers (see

Section 3.1.5), disregarding exchange reactions, is a convenient case study for dis-
cussing problems of modeling this kind of systems. It can be described by the rate
laws of Eqs. (132) and (133).

RP1AA ¼ 4k1 ½P1AA ½P1BB   2k2 ½P1AA ð½P1AB  þ ½ZB Þ þ k1Z ½P1AB  þ k2Z ½ZA 

RP1BB ¼ 4k1 ½P1AA ½P1BB   2k3 ½P1BB ð½P1AB  þ ½ZA Þ þ k1Z ½P1AB  þ k3Z ½ZB 

RP1AB ¼ 4k1 ½P1AA ½P1BB   ½P1AB ½2k2 ½P2AA  þ 2k3 ½P1BB  ð132Þ
þ k4 ð2½P1AB  þ ½ZA  þ ½ZB Þ  k1Z ½P1AB  þ 2k2Z ½P2AA  þ 2k3Z ½P2BB 

þ 2k4Z ð½ZA  þ ½ZB   2½P2AA   2½P2BB Þ

n b 2:

RPnAA ¼ 2k2 ½P1AA ½Pn1
 þ 2k4 ½PmAB ½Pnm
  2½PnAA ½2k3 ½P1BB  þ k4 ð½P1AB  þ ½ZB Þ
" #
 ½PnAA ½2k2Z þ 2k4Z ðn  2Þ þ k3Z ½PnAB  þ k4Z ½ZA   ð½PmAB  þ 2½PmAA Þ

RPnBB ¼ 2k3 ½P1BB ½Pn1
 þ 2k4 ½PmAB ½Pnm
  2½PnBB ½2k2 ½P1AA  þ k4 ð½P1AB  þ ½ZA Þ
m ¼1
" #
 ½PnBB ½2k3Z þ 2k4Z ðn  2Þ þ k2Z ½PnAB  þ k4Z ½ZB   ð½PmAB  þ 2½PmBB Þ
134 3 Polycondensation

RPnAB ¼ 4k2 ½P1AA ½PnBB  þ 4k3 ½P1BB ½PnAA 

n1 X
þ k4 ½PmAB ½Pnm
 þ4 ½PmAA ½Pnmþ1

m¼1 m¼2

 ½PnAB ½2k2 ½P1AA  þ 2k3 ½P1BB  þ k4 ð2½P1AB 

þ ½ZA  þ ½ZB Þ þ k2Z þ k3Z þ k4Z ð2n  3Þ þ 2k2Z ½Pnþ1

 þ 2k3Z ½Pnþ1

" #
n X
þ k4Z ½ZA  þ ½ZB   2 ½PnAB   2 ð½PmAA  þ ½PmBB Þ ð133Þ
m¼2 m¼2

It may be observed from these expressions that, in general, it is not possible to ob-
tain a closed finite set of rate equations for the first oligomers – rate equations for
oligomers with chain length n always depend on concentrations of oligomers with
chain length n þ 1. Only when the rate constants of reverse equations are equal
do the above expressions simplify (FSSE for reverse reaction), and the above-
mentioned difficulty disappears.
On introduction of generating functions of the CLD, defined in Eq. (134), the
concentrations of end groups, trimers, and tetramers conform with Eqs. (134) and

y X
y X
GAA ¼ s n2 ½PnAA  GBB ¼ s n2 ½PnBB  GAB ¼ s n2 ½PnAB  ð134Þ
n¼2 n¼2 n¼2

½ZA  ¼ 2GAA ð1Þ þ GAB ð1Þ ½ZB  ¼ 2GBB ð1Þ þ GAB ð1Þ
½P2AA  ¼ GAA ð0Þ ½P2BB  ¼ GBB ð0Þ ½P2AB  ¼ GAB ð0Þ

The rate equations in terms of those generating functions become Eqs. (136).

RGAA ¼ 2k2 ½P1AA ðsGAB þ ½P1AB Þ þ k4 s 2 GAA GAB  2GAA ½2k3 ½P1BB 

þ k4 ð½P1AB  þ ½ZB Þ  2k2Z GAA  2k4Z s þ k3Z GAB


þ k4Z

RGBB ¼ 2k3 ½P1BB ðsGAB þ ½P1AB Þ þ k4 s 2 GBB GAB  2GBB ½2k2 ½P1AA 

þ k4 ð½P1AB  þ ½ZA Þ  2k3Z GBB  2k4Z s þ k2Z GAB


þ k4Z
3.4 Modeling of Complex Polycondensation Reactions 135

RGAB ¼ 4k2 ½P1AA GBB þ 4k3 ½P1BB GAA þ k4 ½2½P1AB ðsGAB þ ½P1AB Þ
þ s 2 GAB
þ 4sGAA GBB   ½2k2 ½P1AA  þ 2k3 ½P1BB  þ k4 ð2½P1AB  þ ½ZA  þ ½ZB Þ

qGAB ½ZA  þ ½ZB   2ðGAB þ GAA þ GAB Þ

þ k2Z þ k3Z þ k4Z GAB  2k4Z s þ k4Z
qs 1s


þ 2ðk2Z  k4Z Þ þ 2ðk3Z  k4Z Þ ð136Þ
s s

Rate laws for ½ZA  and ½ZB  become Eq. (137):

RZA ¼ 2k2 ½P1AA ð½ZB  þ 2½P1AB Þ  2k3 ½P1BB ½ZA  þ k4 ð2½P1AB 2  ½ZA ½ZB Þ
 k2Z ð½ZA  þ 2½P2AA Þ  k3Z ð½ZB   2½P2BB Þ
þ k4Z ½X þ ½Y þ ð½ZA  þ ½ZB Þ  ½P1AA   ½P1BB   2½P1AB 

RZB ¼ 2k2 ½P1AA ½ZB  þ 2k3 ½P1BB ð½ZA  þ 2½P1AB Þ þ k4 ð2½P1AB 2  ½ZA ½ZB Þ
 k2Z ð½ZA   2½P2AA Þ  k3Z ð½ZB  þ 2½P2BB Þ
þ k4Z ½X þ ½Y þ ð½ZA  þ ½ZB Þ  ½P1AA   ½P1BB   2½P1AB  ð137Þ

A rate law for by-product [Eq. (138)] is usually required.

RW ¼ 4k1 ½P1AA ½P1BB  þ 2k2 ½P1AA ½ZA  þ 2k3 ½P1BB ½ZA  þ k4 ½ZA ½ZB   k1Z ½P1AB 

 k2Z ½ZA   k3Z ½ZB   k4Z ½ð2n  4Þð½PnAA  þ ½PnBB Þ þ ð2n  3Þ½PnAB 

¼ 4k1 ½P1AA ½P1BB  þ 2k2 ½P1AA ½ZA  þ 2k3 ½P1BB ½ZA  þ k4 ½ZA ½ZB 

 k1Z ½P1AB   k2Z ½ZA   k3Z ½ZB 

 k4Z ½X  þ ½Y   ð½ZA  þ ½ZB Þ  ½P1AA   ½P1BB   2½P1AB  ð138Þ

Insertion of these rate laws in mass balances of ideal reactors (batch/plug flow or
transient CSTR) leads to systems of semi-linear, first-order, partial differential
equations, with a single family of characteristics [Eq. (139)].

¼ 2sk4Z ð139Þ

The relationship of Eq. (138) allows integration along characteristics, relating ini-
tial values GAA ð0; s 0 Þ; GAB ð0; s 0 Þ, and GBB ð0; s 0 Þ to GAA ðt; sÞ; GAB ðt; sÞ, and GBB ðt; sÞ,
136 3 Polycondensation

but a serious problem remains: integration along characteristics and prediction of

concentration of first oligomers have to be done at the same time.
Only the solution with equal reactivity has been published [291]. A sketch of an
iterative method to solve this more complex problem is described below (no actual
calculations have been published at the time of writing).

1. Two auxiliary characteristics sþ ðtÞ ¼ s ðtÞ, starting at sþ ¼ es js0 j and s ¼

es js0 j, respectively, for t ¼ 0, are introduced in order to provide an accurate nu-
merical estimation of oligomer concentrations through Eq. (140).

GI ðsþ Þ þ GI ðs Þ
GI ð0Þ ¼ þ Oðes2 Þ I ¼ AA; BB; AB ð140Þ

Parameter es should be a small value (say 106 to 108 ). Owing to Eq. (139), it is
easy to relate the auxiliary characteristics to the ‘‘main’’ characteristic sðtÞ start-
ing with s ¼ s0 through Eq. (141).

sþ ðtÞ ¼ s ðtÞ ¼ es sðtÞ ð141Þ

2. The resolution of a system of ODE comprising Eq. (139), the balances of end
groups ZA , ZB , by-product W and differential equations for GAA ðt; sÞ, GAB ðt; sÞ,
GBB ðt; sÞ, GAA ðt; sþ Þ, GAB ðt; sþ Þ, GBB ðt; sþ Þ, GAA ðt; s Þ, GAB ðt; s Þ and GBB ðt; s Þ,
will eventually allow the evaluation of all unknown variables of this problem.
As s0 is to be found iteratively in order that the characteristic starting with
s ¼ s0 will attain the prescribed value of s at the prescribed time t, this is a
boundary value problem, but very straightforward to solve.

If there are more than two functional groups per repeating unit, the approach de-
scribed above is no longer useful, as there is no simple way of describing forma-
tion of chemical species through breaking of larger ones, owing to their branched
Nevertheless, in most situations resembling industrial practice, such as CSTRs
in series, the CLD of these chemical systems will stay close to equilibrium (see
Ref. 291). The distributions of monads depend only on the equilibrium ratios and
it is not necessary to use models with SSSEs for the reverse reactions (the source of
the most challenging mathematical problems), as the same distributions will be
found using equal values of the rate constants of the reverse reaction.


1N vector with N components equal to one

av film area per unit volume [m1 ]
b length of a Kuhn segment connecting two half-bonds in the same
repeating unit [m]
Notation 137

bA empirical parameter in Eq. (56) (dimensionless)

Cd drag coefficient (dimensionless)
Cg parameter in Eq. (12) [s1 ]
D binary diffusivity [m 2 s1 ]
db bubble diameter [m]
dR diameter of ring (in a RDC devolatilizer) [m]
dT inner diameter of tube, or of the barrel of an extruder [m]
Ec activation energy of amidation reaction [J mol1 ]
ei index of by-product formed by reaction leading to bond Z iR ; no by-
product if it is nil
i ; ei indices of ‘‘half-bonds’’ of bond Z iR
F X i ðsZ ; sA Þ; F A i ðsZ ; sA Þ; F Z i ðsZ ; sA Þ probability generating functions (pgf ) of the
counts of the different kinds of connecting and unreacted functional
groups directly linked to, respectively, a monomer unit X i , an un-
reacted group A i , or a half-bond Z i
f monomer functionality (number of functional groups it contains)
Ge equilibrium zero strain shear modulus of polymer network [Pa]
GðsÞ moment generating function of molar concentrations of polymer mol-
ecules with respect to their counts of monomer units and unreacted
functional groups
Gni generating functions of stoichiometric coefficients of reaction
i ð J ¼ þ; ;  Þ, as defined by Eq. (126)
G ðsX ; sA Þ; G A i ðsX ; sA Þ; G Z i ðsX ; sA Þ pgf of trees starting, respectively, with a pre-

scribed monomer unit X i , an unreacted group A i , or a half-bond Z i

GM ðsM Þ pgf of mass fractions of polymer molecules in a sol
g acceleration due to gravity [m s2 ]
gA ratio mA /bA [m 3 mol1 ]
gi ; gi indices of functional groups A gi ; A giþ reacting to create bond Z i
H depth of liquid in the channel of a vented extruder [m]
He ; Hi partition coefficients of hydrophilic monomer in interfacial polymer-
ization relating, respectively, the bulk concentration in the water
phase and the concentration in the outer interface of the polymer
film, and the concentrations in organic phase in the two interfaces of
polymer film (dimensionless)
HW Henry constant of component W in terms of pressure versus weight
fraction [Pa]
h clearance between the screw and barrel in a vented extruder [m]
hi index of a repeating unit carrying functional group A i
hR height of the liquid above the gas injection point in an RDC [m]
J number of compartments in a staged model of an RDC devolatilizer
Jj mass transfer rate per unit volume of component j [mol m3 s1 ]
K apparent equilibrium ratio of bond formation from end groups
K0 thermodynamic equilibrium constant of bond formation from end
groups (dimensionless)
138 3 Polycondensation

Ka apparent equilibrium ratio of amidation (dimensionless)

K a0 reference value of the apparent equilibrium ratio of amidation
K c ðnÞ apparent equilibrium ratio of cyclization [mol m3 ]
k apparent second-order rate constant [m 3 mol1 s1 ]
k0 parameter in Eq. (12) [m 3 mol1 s1 ]
kA empirical parameter in Chen–Wu relation, Eq. (53) [mol m3 s1 ]
k aE rate constant of breakage of a bond in a ring molecule through reac-
tion with a functional group [mol m3 s1 ]
k c ðnÞ first-order rate constant of L n cyclization [s1 ]
k cE ðnÞ rate constant of formation of Cn by the back-biting cyclization reaction
[s1 ]
kcr rate coefficient of crystallization in Avrami equation [sncr ] (see below)
k cZ apparent first-order rate constant of breakage of a bond in a ring
molecule [s1 ]
kc0 third-order pre-exponential factors of amidation rate constant
[kg 2 mol2 s1 ]
kfYi mass transfer coefficient of species Yi in terms of molar concentra-
tions [m s1 ]
mass transfer coefficient of species Yi in terms of activities
[mol m2 s1 ]
ki0 noncatalytic contribution to the rate constant in Eq. (61) [mol m3 s1 ]
kijE rate constant of exchange reaction forming bond Z iR and destroying
bond Z jR
kijE ; kijEþ rate constants of exchange reaction forming bond Z iR and replacing
bond Z jR , when the attacking group is, respectively, A½i or A½iþ
kiH acid-catalyzed reaction constant in Eq. (61) [m 6 mol2 s1 ]
kiOH catalytic term of hydroxyls in Eq. (61) [m 6 mol2 s1 ]
km water mass transfer rate constant [s1 ]
k nuc; n rate coefficients of nucleation of polymer species with chain length n
[s1 ]
kscat ; kfcat rate constants of self-catalyzed and foreign acid-catalyzed esterification
[m 6 mol2 s1 ]
ku0 second-order pre-exponential factor of amidation rate constant
[m 3 mol1 s1 ]
kZ apparent first-order rate constant of bond group breakage [s1 ]
kiZ apparent first-order rate constant of breakage of bond group ZiR [s1 ]
ki first-order rate constant of intramolecular reaction i destroying group
A½i   [s1 ]
L film thickness [m]
Lp thickness of polymer film in interfacial polymerization [m]
LR thickness of reaction zone [m]
LW width of the channel in an extruder [m]
Lx film perimeter [m]
Notation 139

M relative molecular weight

Mi relative molecular weight of generic polymer species i, where i is an
 y index
y P
Mn ¼ Mi ½Pi  ½Pi  number-average molecular weight of polymer
i¼1 i¼1
MP mass of polymer per mole of monomer units [kg mol1 ]
MW relative
 ymolecular weight of by-product W
Py P
Mw ¼ Mi2 ½Pi  Mi ½Pi  weight-average molecular weight of polymer
i¼1 i¼1

y P
Mz ¼ Mi3 ½Pi  Mi2 ½Pi  z-average molecular weight of polymer
i¼1 i¼1
mA empirical parameter in Eq. (56) [mol1 m 3 ]
NA number of functional (or end) groups
NAL Avogadro–Löschmidt number [mol1 ]
Nb number of bubbles
N_ j flux of component j [mol m2 s1 ]
NR number of distinctive chemical bonds
NRs number of symmetrical chemical bonds
NX number of kinds of monomer units
NW number of by-products
NY number of volatile (or precipitating) components
NZ number of half-bonds
NZA ¼ NZ þ NA overall number of kinds of functional groups and half-bonds
NZXA ¼ NZ þ NX þ NA overall number of kinds of functional groups, monomer
units, and half-bonds
n_ screw rotational speed (as rotations per unit time) [s1 ]
ncr Avrami exponent
Pb pressure inside a bubble [Pa]
Pme local pressure over a bubble [Pa]
PW vapor pressure of W [Pa]
PW vapor pressure of pure W [Pa]
p conversion of reference functional groups A
pc conversion of functional groups A having given rise to rings
Q volumetric flow rate of polymer [m 3 s1 ]
Qg volumetric flow rate of gas [m 3 s1 ]
R ideal gas constant [J mol1 K1 ]
Re Reynolds number for the rising movement of a bubble
Rj rate of formation by chemical reaction of species j [mol m3 s1 ]
RV combined relative rate of change of reaction volume by chemical reac-
tion and transfer of by-product [s1 ]
r stoichiometric ratio in alternating polycondensations
rb radius of a bubble [m]
s vector of dummy Laplace variables associating sA and sZ (defined
140 3 Polycondensation

sA vector of dummy Laplace variables associated with the counts of un-

reacted functional groups
sCn dummy Laplace variables associated with the counts of rings (includ-
ing sC0 for the count of units X in the chains connecting rings) for the
equilibrium polycondensation of a single monomer XA f
sM dummy Laplace variable associated with molecular weight
sZ vector of dummy Laplace variables associated with the counts of half-
s Zy dummy Laplace variable associated with xZy
sZl dummy Laplace variable associated with the counts of bonds not be-
longing to rings for the equilibrium polycondensation of a single
monomer XA f
T absolute temperature [K]
T0 reference temperature [K]
t time [s]
tf exposure time [s]
t fP exposure time of liquid in the pool of a vented extruder [s]
ub bubble rising speed [m s1 ]
ux ; uz average velocity in directions x; z [m s1 ]
V volume of reacting mixture [m 3 ]
Vb volume of a bubble [m 3 ]
Vi ðsZ ; sA Þ pgf of pendent chains Vi
V^i ðsZ ; sA Þ pgf of finite pendent chains Vi
Vj molar volume of species j [m 3 mol1 ]
VMi ðsM Þ pgf of finite pendent chains Vi with respect to molecular weight
Vm overall melt volume [m 3 ]
Vmj volume of liquid in compartment j in a staged model of an RDC devo-
latilizer [m 3 ]
v vector of the NZ probabilities of extinction (the fractions of finite pen-
dent chains)
vT average transverse velocity of the screw dragging the film [m s1 ]
wcr weight fraction of crystalline material
wj weight fraction of j
wS weight fraction of the sol, relative to the overall mass of the polymer
x degree of polymerization, the number of repeating units in a polymer
x coordinate along the perimeter, when discussing mass transfer from a
liquid flowing over a cylindrical wall [m]
xA vector storing the counts of functional groups
xi degree of polymerization of generic polymer species i, where i is an
 y index
y P
xn ¼ x i ½Pi  ½Pi  number-average degree of polymerization of polymer
i¼1 i¼1

Py Py
xw ¼ xi2 ½Pi  x i ½Pi  weight-average degree of polymerization of polymer
i¼1 i¼1
Notation 141

xX vector containing the counts of monomer units

xZ vector storing the counts of half-bonds
xZ y number of infinite pendent chains connected to a junction
y Cartesian coordinate normal to the interface [m]
yc overall molar fraction of repeating units X in rings for a single mono-
mer polycondensation
yP number of moles of polymer molecules per mole of monomer units
before gelation
ySX i ; ySZ i and ySA i fractions of units and groups of the various kinds in finite
molecules (in a sol), after gelation
y Xc ðnÞ molar fraction of repeating units X in rings of size n for a single
monomer y X i molar fractions of monomers or monomer units
ZZX matrix defined in Eq. (73) from incidence vectors z defined below
z Cartesian coordinate parallel to the interface [m]
zi index of the monomer unit to which half-bond Z i points


a empirical parameter of Chen–Wu relation [Eq. (53)] (dimensionless)

ar probability of reaction of unreacted functional groups after ring for-
mation in the first stage (absence of open-chain molecules) of a hypo-
thetical two-stage equilibration of XA f
b g ; b g0 parameters in Eq. (11) describing the glass effect [K]
bj fraction of volume flow rate in compartment j going to the preceding
compartment j  1
gj activity coefficient of species j
wij Flory–Huggins parameter (multicomponent mixture)
DH enthalpy of reaction [J mol1 ]
DS entropy of reaction [J mol1 K1 ]
d upstream distance of re-entry of film wiped by the screw in a vented
extruder [m]
es small starting value of s in characteristic line
y screw angle
k c ðnÞ ratio of rate constants of similar intramolecular and intermolecular
reactions [mol m3 ]
q . . . qF I
LIJ...K ðsÞ ¼ derivatives of pgf F I ðsÞ (or other moment gener-
q log sJ . . . q log sK
ating functions) with respect to logarithms of dummy Laplace
I q . . . qF I P
y Py
lJ...K ¼ ð1N Þ ¼ ... xJ . . . xJ IðxÞ moments of distribu-
q log sJ . . . q log sK x1 ¼0 xN ¼0
tions, where F ðsÞ is the moment generating function of vector distri-
bution IðxÞ
142 3 Polycondensation

lM ; lMM first and second moments of molecular weight of distribution of

m dynamic viscosity [N m1 s1 ]
me concentrations of elastically active network junctions (EANJ)
[mol m3 ]
n stoichiometric coefficient (the change in number of moles caused by
chemical reaction)
ninA ; ninZ stoichiometric coefficients of functional groups An and oriented bonds
Zn coming out of the monomer unit in which A½i was situated for
the reaction forming Z iR
Aþ Zþ
nin ; nin stoichiometric coefficients of functional groups An and oriented bonds
Zn coming out of the monomer unit in which A½iþ was situated for
the reaction forming Z iR
nijn ; nijn changes in numbers of functional groups An and bonds Zn , respec-
tively, connected to the root unit X  where either A½i or A½iþ were
attached, for the exchange reaction forming Z iR at the expense of Z jR
AEþþ ZEþþ AEþþþ ZEþþþ
nijn ; nijn ; nijn ; nijn changes in numbers of functional groups An
and bonds Zn , respectively, connected to the root unit X þþ where the
living group A½ j or A½ jþ was situated, for the exchange reaction
forming Z iR at the expense of Z jR
AEþ ZEþ AEþþ ZEþþ
nijn ; nijn ; nijn ; nijn changes in numbers of groups attached to the
root unit X þ which gets connected to the unit where the attacking
group was situated, for the exchange reaction forming Z iR at the ex-
pense of Z jR
nin stoichiometric coefficients of by-products for the reaction forming Z iR

nin stoichiometric coefficient of group An for the ith intramolecular
ne concentration of elastically active network chains [mol m3 ]
rj density of j [kg m3 ]
s surface tension [N m1 ]
t space time based upon feed flow rate, Vm /Q F for a CSTR [s]
tr structural relaxation time [s]
t0 parameter in Eq. (11) [s]
tD tortuosity for by-product diffusion in a solid polymer particle
fcr volume fraction of crystalline material
fj volume fraction of j
fL fraction of channel extruder filled with liquid
w Flory–Huggins parameter (binary mixture) (dimensionless)
WW weight fraction activity coefficient of W

Chemical symbols

A; B functional groups
A gi 1 A½i ; A giþ 1 A½iþ functional groups reacting in bond Z iR formation
Notation 143

Aj active component (molecular species or fragment) j; nonvolatile,

Cn ring molecule with n bonds
Ln linear polymer molecule with n  1 bonds and a pair of active end
LT length of tube
Pi generic polymer species, where i is an arbitrary index
Pn linear polymer molecule with end groups I and J and n repeating
units of the most frequent kind
½U  molar concentration of species U
Vi ðxZ ; xA Þ pendent chain starting with a half-bond Z i and containing vector
counts of bonds and functional groups, respectively, xZ and xA
W by-product of condensation reaction
We i 1 W½i by-product of condensation reaction leading to bond Z iR
X; Y monomer units
Xh i 1 X½iA monomer unit carrying functional group A i
Xz i 1 X½iZ monomer unit to which each half-bond Z i points
Yj volatile (or precipitating) component j
Z bond group; Z iR is the ith distinctive bond in the chemical system
ZA bond group in fragment AXZA YZ (polycondensation AXA þ BYB)
ZB bond group in fragment BYZB XZ (polycondensation AXA þ BYB)
ZD bond group in dimer AXZD YB (polycondensation AXA þ BYB)
Zei 1 Z½i ; Zeþi 1 Z½iþ ‘‘half-bonds’’ of bond Z iR
ZP bond group in fragment ZXZP Z (polycondensation AXA þ BYB)


aq aqueous
b of a bubble
blk bulk
c critical
cr crystalline
F in feed
g at gel point
org organic
sat at saturation


* used for describing values of concentration at the interface

x average value of x


BHET bis(2-hydroxyethyl) terephthalate

BPA bisphenol A
144 3 Polycondensation

CLD chain length distribution

CSTR continuous stirred tank reactor
DEG diethylene glycol
DMT dimethyl terephthalate
DPC diphenyl carbonate
EANC elastically active network chains, the chains connecting EANJs (see
EANJ elastically active network junctions, the intersection of at least three
chains leading to a gel
ECH epichlorohydrin
EG ethylene glycol
FSSE first-shell substitution effect
ODE ordinary differential equation
PBT poly(butylene terephthalate)
PDE partial differential equation
PET poly(ethylene terephthalate)
pgf probability generating function
RDC rotating disk contactor
RIM reaction injection molding
SSP solid-state polycondensation
SSSE second-shell substitution effect
TBP theory of branching processes
TPA terephthalic acid
UF urea–formaldehyde polymer
WFR wiped-film reactor


1 A. V. Lourenço, Comptes Rendus 8 D. J. Brunelle, J. E. Bradt, US

Hébd. Séances Acad. Sciences 1859, 49, Pat. 5 039 783 to General Electric, 1991.
619; A. V. Lourenço, Comptes Rendus 9 R. B. Burch, S. R. Lustig, M. Spinu,
Hébd. Séances Acad. Sciences 1860, 50, Macromolecules 2000, 33, 5053.
91. 10 G. V. Schulz, Z. Phys. Chem. 1938,
2 L. H. Baekeland, J. Ind. Eng. Chem. A182, 127.
1909, 1, 149; reprinted in Chemtech, 11 M. R. P. F. N. Costa, J. Villermaux,
1976, 6, 40. Ind. Eng. Chem. Res. 1989, 28, 711.
3 H. Staudinger, Ber. 1920, 53, 1073. 12 M. H. Abraham, Pure Appl. Chem.
4 W. H. Carothers, J. Am. Chem. Soc. 1985, 57, 1055.
1929, 51, 2548. 13 E. A. Macedo, R. C. Gonçalves, A. N.
5 P. J. Flory, Principles of Polymer Simões, C. M. Ventura, L. C.
Chemistry, 1st ed., Cornell University Albuquerque, Ind. Eng. Chem. Res.
Press, Ithaca, New York, 1953. 1995, 34, 1910.
6 J. L. Stanford, R. F. T. Stepto, J. 14 K. Dušek, Polym. Gels Networks 1996,
Chem. Soc., Faraday Trans. 1975, 71, 4, 383.
1292. 15 I. Mita, K. Horie, J. Macromol. Sci.
7 W. H. Stockmayer, J. Chem. Phys. Rev. Macromol. Chem. Phys. 1987, C27,
1943, 11, 45. 91.
References 145

16 S. Corezzi, D. Fioretto, D. Puglia, Henze, W. C. L. Wu, U.S. Pat. 3 522

J. M. Kenny, Macromolecules 2003, 36, 214, 1970.
5271. 40 K. Ravindranath, R. A. Mashelkar,
17 G. Adam, J. Gibbs, J. Chem. Phys. Chem. Eng. Sci. 1986, 41, 2969.
1965, 43, 139. 41 R. Albalak, An Introduction to
18 S. Luňák, J. Vladyka, K. Dušek, Devolatilization, in Polymer
Polymer 1978, 19, 931. Devolatilization, R. J. Albalak (Ed.),
19 A. M. Kotliar, J. Polym. Sci.: Marcel Dekker, New York, 1996.
Macromol. Revs. 1981, 16, 367. 42 P. S. Mehta, An Overview of
20 M. Gordon, G. R. Scantlebury, Devolatilizers, in Polymer
Trans. Far. Soc. 1964, 60, 604. Devolatilization, R. J. Albalak (Ed.),
21 H. Jacobson, W. Stockmayer, J. Marcel Dekker, New York, 1996.
Chem. Phys. 1950, 18, 1600. 43 F. Widmer, Adv. Chem. Ser. 1973, 128,
22 P. J. Flory, J. A. Semlyen, J. Am. 51.
Chem. Soc. 1966, 88, 14. 44 R. M. Secor, AIChE J. 1969, 15, 861.
23 P. J. Flory, U. W. Suter, M. Mutter, 45 T. M. Pell Jr, T. G. Davies, J. Polym.
J. Am. Chem. Soc. 1976, 98, 5733. Sci. B: Polym. Phys. 1973, 11, 1671.
24 H. R. Kricheldorf, M. Rabenstein, 46 J. W. Ault, D. A. Mellichamp, Chem.
M. Maskos, M. Schmidt, Eng. Sci. 1972, 27, 2233.
Macromolecules 2001, 34, 713. 47 M. Amon, C. D. Denson, Ind. Eng.
25 H. R. Kricheldorf, S. Böhme, G. Chem. Fundam. 1980, 19, 415.
Schwarz, Macromolecules 2001, 34, 48 B. R. Choi, H. H. Lee, Ind. Eng.
8879. Chem. Fundam. 1996, 35, 1550.
26 C. Puglisi, F. Samperi, R. Alicata, 49 G. W. Roberts, AIChE J. 1970, 16,
G. Montaudo, Macromolecules 2002, 878.
35, 3000. 50 J. A. Biesenberger, D. H. Sebastian,
27 T. E. Harris, The Theory of Branching Principles of Polymerization Engineering,
Processes, Springer, Berlin, 1963. chap. 6, Wiley, New York, 1983.
28 M. Gordon, Proc. Roy. Soc. 1962, 51 G. A. Latinen, Adv. Chem. Ser. 1962,
A268, 240. 34, 235.
29 G. R. Dobson, M. Gordon, J. Chem. 52 R. W. Coughlin, G. P. Canevari,
Phys. 1965, 43, 705. AIChE J. 1969, 15, 560.
30 M. Gordon, G. N. Malcolm, Proc. 53 W. A. Mack, R. Herter, Chem. Eng.
Roy. Soc. 1966, A295, 29. Prog. 1976, 72(1), 64.
31 K. Kajiwara, J. Chem. Phys. 1971, 54, 54 G. P. Collins, C. D. Denson, G.
296. Astarita, AIChE J. 1985, 31, 1288.
32 K. Kajiwara, M. Gordon, J. Chem. 55 K. Ravindranath, R. A. Mashelkar,
Phys. 1973, 59, 3623. in Frontiers in Chemical Engineering,
33 W. Burchard, Adv. Polym. Sci. 1982, Vol. 1, p. 652, L. K. Doraiswamy, R.
48, 1. A. Mashelkar (Eds.), Wiley Eastern,
34 C. W. Macosko, D. R. Miller, Delhi, 1984.
Macromolecules 1976, 9, 199. 56 Y. Murakami, K. Fujimoto, K.
35 C. Sarmoria, D. R. Miller, Kakimoto, M. Sekino, J. Chem. Eng.
Macromolecules 1991, 24, 1833. Jpn. 1972, 5, 257.
36 K. Platkowski, K. H. Reichert, 57 K. Vijayaraghvan, J. P. Gupta, Ind.
Polymer 1999, 40, 1057. Eng. Chem. Fundam. 1982, 21, 333.
37 B. E. Eichinger, O. Akgyray, chap. 8 58 M. Dietze, H. Kuhne, Chemiefasern
in Computer Simulation of Polymers, 1969, 3, 19.
E. A. Colbourn (Ed.), Longman, 59 H. Brod, Chem. Eng. Technol. 1988,
Harlow, UK, 1994. 21, 937–942.
38 H. Tobita, N. Hamashima, J. 60 D. D. Ravetkar, D. D. Cale, Chem.
Polym. Sci. B: Polym. Phys. 2000, 38, Eng. Sci. 1981, 36, 399.
2009. 61 R. L. Denecour, A. N. Gent, J. Polym.
39 J. E. Crawford, R. W. Edwards, E. D. Sci. A-2 1968, 6, 1853.
146 3 Polycondensation

62 S. T. Lee, J. A. Biesenberger, Polym. 85 S. K. Yadav, A. K. Suresh, K. C.

Eng. Sci. 1989, 29, 782. Khilar, AIChE J. 1990, 36, 431.
63 M. Favelukis, Z. Tadmor, R. Semiat, 86 S. K. Yadav, K. C. Khilar, A. K.
AIChE J. 1999, 45, 691. Suresh, AIChE J. 1996, 42, 2616.
64 A. Yarin, D. Lastochkin, Y. Talmon, 87 M. Kubo, Y. Harada, T. Kawatsu, Y.
Z. Tadmor, AIChE J. 1999, 45, 2590. Yonemoto, J. Chem. Eng. Jpn. 2001,
65 I. Gestring, D. Mewes, Chem. Eng. 34, 1506.
Sci. 2002, 57, 3415. 88 P. L. Mills, Chem. Eng. Sci. 1986, 41,
66 C.-T. Yang, T. G. Smith, D. I. Bigio, 1045.
C. Anolick, AIChE J. 1997, 43, 1861 89 S. K. Karode, S. S. Kulkarni, A. K.
and 1874. Suresh, R. A. Mashelkar, Chem. Eng.
67 K. H. Song, S. H. Lee, K. H. Park, Sci. 1997, 52, 3243.
Korean Pat. Appl. 1999-73 161, 1999. 90 S. K. Karode, S. S. Kulkarni, A. K.
68 B.-G. Woo, K. Y. Choi, K. H. Song, Suresh, R. A. Mashelkar, Chem. Eng.
Ind. Eng. Chem. Res. 2001, 40, 3459. Sci. 1998, 53, 2649.
69 J. F. Davidson, B. O. G. Schüler, 91 E. D. Johnson, A study of the kinetics
Trans. Inst. Chem. Eng. 1960, 38, S105. and mechanism of interfacial polymeri-
70 P. Snabre, F. Magnifotcham, Eur. zation, Ph.D. dissertation, Carnegie-
Phys. J. 1998, B4, 369. Mellon University, Pittsburgh, 1985.
71 F. Pilati, Solid-State Polymerization, 92 K. Kamide, H. Ijima, S. Matsuda,
in Comprehensive Polymer Science, p. Polym. J. 1993, 25, 1113.
201, G. Allen, J. C. Bevington 93 L. J. Gonzalez-Ortiz, J. M. Asúa,
(Eds.), Pergamon Press, Oxford, 1989. Macromolecules 1996, 29, 4520.
72 C. Gostoli, F. Pilati, G. C. Sarti, B. 94 J. Huang, J.-P. Leblanc, H. K. J.
Di Giacomo, J. Appl. Polym. Sci. 1984, Hall, J. Polym. Sci. A: Polym. Chem.
29, 2873. 1992, 30, 345.
73 F. Pilati, C. Gostoli, G. C. Sarti, 95 M. Berthelot, L. Péau de Saint-
Polym. Proc. Eng. 1986, 4, 303. Gilles, Ann. Chim. Phys. 1863, 68, 225.
74 F. K. Mallon, W. H. Ray, J. Appl. 96 S. Hamann, D. H. Solomon, J.
Polym. Sci. 1998, 69, 1233. Swift, J. Macromol. Sci. Chem. 1968,
75 F. K. Mallon, W. H. Ray, J. Appl. A-2, 153.
Polym. Sci. 1998, 69, 1775. 97 A. Fradet, E. Maréchal, Adv. Polym.
76 K. Z. Yao, K. B. McAuley, D. Berg, Sci. 1982, 43, 51.
E. K. Marchildon, Chem. Eng. Sci. 98 S.-A. Chen, K. C. Wu, J. Polym. Sci.
2001, 56, 4801. A: Polym. Chem. 1982, 20, 1819.
77 P. W. Morgan, Condensation Polymers: 99 C. T. Kuo, S.-A. Chen, J. Polym. Sci.
by Interfacial and Solution Methods, A: Polym. Chem. 1989, 27, 2793.
Wiley, New York, 1965. 100 D. Beigzadeh, S. Sajjadi, F. A.
78 J. H. Bradbury, P. J. Crawford, A. Taromi, J. Polym. Sci. A: Polym. Chem.
N. Hambly, Trans. Faraday Soc. 1968, 1995, 33, 1505.
65, 1338. 101 P. B. Zetterlund, W. Weaver, A. F.
79 F. MacRitchie, Trans. Faraday Soc. Johnson, Polym. React. Eng. 2002, 10,
1969, 65, 2503. 41.
80 V. Enkelmann, G. Wegner, 102 E. Paatero, K. Narhi, T. Salmi, P.
Makromol. Chem. 1976, 177, 3177. Nyholm, K. Immonen, Chem. Eng.
81 J. Cadotte, R. King, R. Majerle, R. Sci. 1994, 49, 3601.
Petersen, J. Macromol. Sci. Chem. 103 J. Lehtonen, T. Salmi, K. Immonen,
1981, A15, 727. E. Paatero, P. Nyholm, Ind. Eng.
82 R. J. Petersen, J. Membrane Sci. 1993, Chem. Res. 1996, 35, 3951.
83, 81. 104 J. Lehtonen, K. Immonen, T. Salmi,
83 R. G. Pearson, E. L. Williams, J. E. Paatero, P. Nyholm, Chem. Eng.
Polym. Sci. 1985, 23, 9. Sci. 1996, 51, 2799.
84 M. Sirdesai, K. C. Khilar, Can. J. 105 O. M. O. Habib, J. Málek, Coll. Czech.
Chem. Eng. 1988, 42, 2616. Chem. Commun. 1976, 41, 2724.
References 147

106 O. M. O. Habib, J. Málek, Coll. Czech. Gorbunova, Kin. Cat. 1990, 32,
Chem. Commun. 1976, 41, 3077. 541.
107 J. S. Chung, Kor. J. Chem. Eng. 1989, 124 J. Otera, N. Dan-oh, H. Nozaki, J.
6, 354. Org. Chem. 1991, 56, 5307.
108 K. Tanaka, A. Ozaki, J. Catalysis 1967, 125 L. H. Buxbaum, Angew. Chem. Int. Ed.
8, 1. Engl. 1968, 7, 182.
109 S. C. Hovenkamp, J. Polym. Sci. A: 126 A. V. Popoola, J. Appl. Polym. Sci.
Polym. Chem. 1971, 9, 3617. 1991, 43, 1875.
110 J. Otton, S. Ratton, V. Vasnev, G. D. 127 H. K. Reimschuessel, Ind. Eng. Chem.
Markova, K. M. Nametov, V. I. Prod. Res. Dev. 1980, 19, 117.
Bakhtumov, S. V. Vinogradova, 128 F. Pilati, P. Manaresi, B. Fortu-
V. V. Korshak, J. Polym. Sci. A: nato, A. Munari, V. Passalacqua,
Polym. Chem. 1989, 27, 3535. Polym. 1981, 22, 1566.
111 L. I. Komarova, N. N. Lapina, B. V. 129 M. Malik, V. Choudhary, I. K.
Lokshin, V. I. Bakhmutov, G. D. Varma, J. Macromol. Sci. Rev.
Markova, V. A. Vasnev, Bull. Acad. Macromol. Chem. Phys. 2000, C40, 139.
Sci. USSR Div. Chem. Sci. 1991, 39, 130 J. Grobelny, Polymer 1995, 36, 4215.
1767. 131 Z. Ordelt, Makromol. Chem. 1963, 63,
112 J.-M. Besnoin, G. D. Lei, K. Y. Choi, 153.
AIChE J. 1989, 35, 1445. 132 Z. Ordelt, Makromol. Chem. 1963, 68,
113 G. D. Lei, K. Y. Choi, Ind. Eng. Chem. 166.
Res. Sci. 1992, 31, 769. 133 A. Fradet, E. Maréchal, Makromol.
114 M. J. Barandiarán, J. M. Asúa, J. Chem. 1982, 183, 319.
Polym. Sci. A: Polym. Chem. 1989, 27, 134 K. Ravindranath, R. A. Mashelkar,
4241. J. Appl. Polym. Sci. 1982, 26, 471.
115 E. Santacesaria, F. Trulli, L. 135 K. Ravindranath, R. A. Mashelkar,
Minervini, M. Di Serio, R. Tesser, J. Appl. Polym. Sci. 1984, 29, 437.
S. Contessa, J. Appl. Polym. Sci. 1994, 136 T. Yamada, Y. Imamura, O. Maki-
54, 1371. mura, Polym. Eng. Sci. 1985, 25, 788.
116 M. Di Serio, R. Tesser, F. Trulli, 137 C.-K. Kang, B. C. Lee, D. W. Ihm,
E. Santacesaria, J. Appl. Polym. Sci. D. A. Tremblay, J. Appl. Polym. Sci.
1996, 62, 409. 1997, 63, 163.
117 B. Apicella, M. Di Serio, L. Fiocca, 138 Z. Yao, W. H. Ray, AIChE J. 2001, 47,
R. Po, E. Santacesaria, J. Appl. 401.
Polym. Sci. 1998, 69, 2423. 139 H. Nava, Polyesters, Unsaturated, in
118 U. Stier, F. Gähr, W. Opermann, J. Kirk-Othmer Encyclopedia of Chemical
Appl. Polym. Sci. 2001, 80, 2039. Technology (on-line version), John
119 F. Pilati, P. Manaresi, B. Wiley, New York, 2004; http://www.
Fortunato, A. Munari, V. mrw.interscience.wiley.com./kirk/
Passalacqua, Polym. 1983, 24, 1479. articles/polysell.a01/sect4.html
120 S. I. Chervina, E. G. Maksimenko, 140 Y. S. Yang, J. P. Pascault, J. Appl.
R. S. Barshtein, N. V. Shabanova, Polym. Sci. 1997, 64, 133 and 147.
A. Kh. Bulai, Yu. I. Kotov, I. Ya. 141 F. N. Jones, Alkyd Resins, in
Slonim, Kin. Cat. 1987, 28, 947. Ullmann’s Encyclopedia of Industrial
121 J. Otton, S. Ratton, V. Vasnev, G. D. Chemistry, Wiley-VCH, Weinheim,
Markova, K. M. Nametov, V. I. 2001.
Bakhtumov, L. I. Komarova, S. V. 142 N. Tuck, The Chemistry of Alkyds, in
Vinogradova, V. V. Korshak, J. Waterborne and Solvent Based Alkyds
Polym. Sci. A: Polym. Chem. 1988, 26, and their End User Applications, chap.
2199. I, John Wiley, New York, 2000.
122 F. Leverd, A. Fradet, E. Maréchal, 143 R. H. Kienle, Ind. Eng. Chem. 1949,
Eur. Polym. J. 1987, 23, 699 and 705. 41, 726.
123 A. Kh. Bulai, I. Ya. Slonim, B. M. 144 M. Jonason, J. Appl. Polym. Sci. 1960,
Arshava, R. S. Barshtein, V. G. 11, 129.
148 3 Polycondensation

145 S. Kühling, K. Schebesta, U. 165 H. Vernaleken, Polycarbonates,

Hucks, H. Schuchardt, R. Interfacial Synthesis, in Polymer
Bachmann, T. Fischer, K. Applications and Technology, Vol. 2,
Kohlgrüber, F. F. Rhiel, G. Zaby, p. 65, F. Millich, C. E. Carrahers
EP 0 798 329 to Bayer-AG, 2001. (Eds.), Marcel Dekker, New York, 1977.
146 P. J. McCloskey, A. Nisoli, W. W. 166 D. W. Fox, Polycarbonates, in Kirk-
Reilly, US Pat. 6 399 739 to General Othmer Encyclopedia of Chemical
Electric Company, 2002. Technology, 3rd ed., Vol. 18, p. 479,
147 K. Komiya, Y. Kawakami, H. Okamoto, John Wiley, New York, 1982.
US Pat. 5 596 067 to Asahi, 1997. 167 D. Freytag, U. Grigo, P. R. Müller,
148 O. Hahn, D. Mooney, F. Müller- W. Nouvertné, Polycarbonates, in
Plathe, K. Kremer, J. Chem. Phys. Encyclopedia of Polymer Science and
1999, 111, 6061. Engineering, 2nd ed., Vol. 11, pp. 651–
149 K. Tominari, A. Kanezawa, T. 692, H. Mark, G. Menges, N.
Sakashita, K. Miura, T. Shimoda, Bikales, C. G. Overberger, J. I.
US Pat. 5 525 701 to General Electric Kroschwitz (Eds.), Wiley, New York,
Company, 1996. 1988.
150 S. Kühling, U. Hucks, G. Zaby, C. 168 D. C. Clagett, S. J. Shafer,
Wulff, US Pat. 5 652 313 to Bayer- Polycarbonates, in Comprehensive
AG, 1997. Polymer Science; The Synthesis,
151 K. Komiya, S. Fukuoka, EP 0 892 001 Characterization, Reactions, and
to Asahi, 2002. Applications of Polymers, Vol. 5, p. 345,
152 B.-G. Woo, K. Y. Choi, W. H. Song, G. Allen, J. C. Bevington (Eds.),
S. H. Lee, J. Appl. Polym. Sci. 2001, 80, Pergamon Press, Oxford, 1989.
1253. 169 V. Serini, Polycarbonates, in
153 S. W. Webb, AIChE J. 1993, 39, 701– Ullmann’s Encyclopedia of Industrial
706. Chemistry, 5th ed., p. 207, B. Elvers,
154 U. Hucks, R. Bachmann, F. F. S. Hawkins, G. Schulz (Eds.), VCH,
Rhiel, T. Fischer, S. Kratschmer, Weinheim, 1992.
WO 02/077 066 A1, 2002. 170 D. G. LeGrand, J. T. Bendler (Eds.),
155 A. Einhorn, Liebigs Ann. Chem. 1898, Handbook of Polycarbonate Science and
300, 135. Technology, Marcel Dekker, New York,
156 C. A. Bischoff, A. van Hedenström, 2000.
Ber. 1902, 35, 3431. 171 D. J. Brunelle, in Encyclopedia of
157 W. H. Carothers, F. J. van Natta, J. Polymer Science and Engineering, 3rd
Am. Chem. Soc. 1930, 52, 314. ed., Vol. 7, pp. 396–426, H. Mark,
158 J. R. Whinfield, J. D. Dickson, Br. N. Bikales, J. I. Kroschwitz (Eds.),
Pat. 578 079, 1946. Wiley, New York, 2003.
159 H. Schnell, L. Bottenbruch, H. 172 P. G. Kosky, J. M. Silva, E. A.
Krimm, Belg. Pat. 532 543 to Bayer Guggenheim, Ind. Eng. Chem. Res.
AG, 1954. 1991, 30, 462.
160 H. Schnell, Angew. Chem. 1956, 68, 173 E. Aquino, W. J. Brittain, D. J.
633. Brunelle, Polymer Int. 1994, 33, 161.
161 D. W. Fox, Australian Pat. 221 192 to 174 P. G. Kosky, E. P. Boden, J. Polym.
General Electric Company, 1959. Sci. A: Polym. Chem. 1990, 28, 1507.
162 D. W. Fox, US Pat. 3 153 008 to 175 J.-T. Gu, C. S. Wang, J. Appl. Polym.
General Electric Company, 1964. Sci. 1992, 44, 849.
163 D. Freytag, G. Fengler, L. 176 C. Wulff, H. Kauth, G. Weymans, Z.
Morbitzer, Angew. Chem. Int. Ed. Kricsfalussy, W. Alewelt, J. Heuser,
Engl. 1991, 30, 1598. EP 0 517 044 to Bayer AG, 1995.
164 H. Schnell, The Chemistry and Physics 177 Ya. I. Korenman, A. A. Gorokhov, J.
of Polycarbonates (Polymer Reviews, Vol. Appl. Chem. USSR, 1974, 46, 2571.
9), Interscience Publishers, New York, 178 J. M. Silva, P. K. Kosky, Ind. Eng.
1964. Chem. Res. 1991, 30, 468–474.
References 149

179 P. L. Mills, Chem. Eng. Sci. 1985, 41, 199 H. Jacobs, C. Schweigman, in
2939. Proceedings of the 5th European/2nd
180 S. Munjal, Polym. Eng. Sci. 1994, 34, International Symposium on Chemical
93–101. Reaction Engineering, Amsterdam, May
181 W. H. Carothers, G. T. Berchet, J. 2–4, p. B7.1, 1972.
Am. Chem. Soc. 1930, 52, 5289. 200 German Pat. 1 167 021 to Vereinigte
182 J. N. Weber, Aromatic, Polyamides in Glanzstoffe Fabriken, 1962.
Kirk-Othmer Encyclopedia of Chemical 201 C. S. Wang, N. X. Huang, Z. L.
Technology (on-line version), John Tang, K.-D. Hungenberger,
Wiley, New York, 2002; http://www. Macromol. Theory Simul. 2002, 11, 687.
mrw.interscience.wiley.com./kirk/ 202 M. Albert, I. Hahnenstein, H.
articles/genewebe.a01/sect10-fs.html Hasse, G. Maurer, AIChE J. 1996, 42,
183 N. Ogata, Makromol. Chem. 1960, 42, 1741.
52. 203 L. L. Williams, Amino Resins and
184 N. Ogata, Makromol. Chem. 1961, 43, Plastics, in Kirk-Othmer Encyclopedia of
117. Chemical Technology (on-line version),
185 F. Wiloth, Z. Phys. Chem. (München) John Wiley, New York, 2002; http://
1955, 4, 66. www.mrw.interscience.wiley.com./
186 C. Giori, B. T. Hayes, J. Polym. Sci. A: kirk/articles/aminwill.a01/sect1.html
Polym. Chem. Ed. 1970, 8, 335 and 351. 204 J. I. de Jong, J. de Jonge, Rec. Trav.
187 K. Tai, H. Teranishi, Y. Arai, T. Chem. 1952, 71, 643.
Tagawa, J. Appl. Polym. Sci. 1979, 24, 205 J. I. de Jong, J. de Jonge, Rec. Trav.
211. Chem. 1952, 71, 661.
188 D. D. Steppan, M. F. Doherty, M. F. 206 J. I. de Jong, J. de Jonge, Rec. Trav.
Malone, J. Appl. Polym. Sci. 1987, 33, Chem. 1953, 72, 139.
2333. 207 A. F. Price, A. R. Cooper, A. S.
189 M. A. Schaffer, K. B. McAuley, M. F. Meskin, J. Appl. Polym. Sci. 1980, 25,
Cunningham, E. K. Marchildon, 2597.
Ind. Eng. Chem. Res. 2003, 42, 2946. 208 A. Kumar, A. Sood, J. Appl. Polym.
190 M. A. Schaffer, E. K. Marchildon, Sci. 1990, 40, 1473.
K. B. McAuley, M. F. Cunningham, 209 N. Landqvist, Acta Chem. Scand.
Polym. Eng. Sci. 2003, 43, 639. 1956, 10, 244.
191 F. K. Mallon, W. H. Ray, J. Appl. 210 J. I. de Jong, J. de Jonge, Rec. Trav.
Polym. Sci. 1998, 69, 1213. Chem. 1953, 72, 213.
192 I. K. Miller, J. Polym. Sci. A: Polym. 211 J. I. de Jong, J. de Jonge, Rec. Trav.
Chem. Ed. 1976, 14, 1403. Chem. 1953, 72, 207.
193 H. K. Reimschuessel, J. Polym. Sci., 212 R. M. Rammon, W. E. Johns, J.
Macromol. Rev. 1977, 12, 65. Magnuson, A. K. Dunker, J. Adhes.
194 H. K. Reimschuessel, G. J. Dege, J. 1986, 19, 115.
Polym. Sci.: Polym. Chem. Ed. 1970, 8, 213 M. G. Kim, L. W. Amos, Ind. Eng.
3265. Chem. Res. 1990, 29, 208.
195 D. D. Steppan, M. F. Doherty, M. F. 214 M. G. Kim, J. Polym. Sci. A: Polym.
Malone, J. Appl. Polym. Sci. 1991, 42, Chem. Ed. 1999, 37, 995.
1009. 215 M. G. Kim, J. Appl. Polym. Sci. 2000,
196 M. A. Schaffer, E. K. Marchildon, 75, 1243.
K. B. McAuley, M. F. Cunningham, 216 M. G. Kim, J. Appl. Polym. Sci. 2001,
J. Macromol. Sci. Rev., Macromol. 80, 2800.
Chem. Phys. 2000, C40(4), 233. 217 S. Katušcák, M. Tomáš, O. Schiessl,
197 H. K. Reimschuessel, K. J. Appl. Polym. Sci. 1981, 26, 381.
Nagasubramanian, Chem. Eng. Sci. 218 C. Soulard, C. Kamoun, A. Pizzi, J.
1972, 1119. Appl. Polym. Sci. 1999, 72, 277.
198 A. Kumar, S. K. Gupta, J. Macromol. 219 T. J. Pratt, W. E. John, R. M.
Sci. Rev., Macromol. Chem. Phys. 1986, Rammon, W. L. Plagemann, J. Adhes.
C26(2), 183. 1985, 17, 275.
150 3 Polycondensation

220 A. K. Dunker, W. E. John, R. M. 238 C. C. Riccardi, G. Astarloa-Aierbe,

Rammon, B. Farmer, S. J. Johns, J. J. M. Echeverria, I. Mondragon,
Adhes. 1986, 19, 153. Polym. 2002, 43, 1631.
221 K. Siimer, T. Pehk, P. Christjanson, 239 L. B. Manfredi, C. C. Riccardi,
Macromol. Symp. 1999, 148, 149. O. de la Osa, A. Vázquez, Polym. Intl.
222 G. E. Myers, Holzforschung 1990, 44, 2001, 50, 796.
117. 240 E. Kumpinsky, Ind. Eng. Chem. Res.
223 R. Nastke, K. Dietrich, G. Reinisch, 1994, 33, 285.
G. Rafler, J. Macromol. Sci.–Chem. 241 H. Lee, K. Neville, Handbook of Epoxy
1986, A23, 579. Resins, 1st ed., McGraw-Hill, New
224 M. Gordon, A. Alliwell, T. Wilson, York, 1967.
J. Appl. Polym. Sci. 1966, 10, 1153. 242 N. S. Enikolopyan, M. A. Marke-
225 B. Tomita, J. Polym. Sci. A: Polym. vitch, L. S. Sakhonenko, S. Z.
Chem. Ed. 1977, 15, 2347. Rogovina, V. G. Oshmyan, J. Polym.
226 M. L. Scheepers, P. J. Adriansens, Sci. A: Polym. Chem. Ed. 1982, 20,
J. M. Gelan, R. A. Carleer, D. J. 1231.
Vanderzande, N. K. de Vries, P. M. 243 J. G. Gao, J. Appl. Polym. Sci. 1993, 48,
Brandts, J. Polym. Sci. A: Polym. 237.
Chem. Ed. 1985, 33, 915. 244 P. A. Oyanguren, R. J. J. Williams,
227 S. K. Gupta, A. K. Gupta, A. K. Polym. 1992, 33, 2736.
Ghosh, A. Kumar, Modeling of 245 M. E. Smith, H. Ishida,
Melamine Formaldehyde Polymeriza- Macromolecules 1994, 27, 2701.
tion, in Frontiers in Chemical Engineer- 246 O. Baeyer, Angew. Chem. 1947, A59,
ing, p. 824, L. K. Doraiswamy, R. A. 257.
Mashelkar (Eds.), Wiley Eastern, 247 H. Ulrich, Urethane Polymers, in
Delhi, 1984. Kirk-Othmer Encyclopedia of Chemical
228 A. Kumar, J. Appl. Polym. Sci. 1987, Technology (on-line version), John
34, 1367 and 1383. Wiley, New York, 2000; http://www.
229 P. W. Koft, Phenolic Resins, in Kirk- mrw.interscience.wiley.com./kirk/
Othmer Encyclopedia of Chemical articles/uretulri.a01/sect2.html
Technology (on-line version), John 248 G. Oertel, Polyurethane Handbook,
Wiley, New York, 2003; http://www. 2nd ed., Hanser, Münich, 1993.
mrw.interscience.wiley.com./kirk/ 249 C. W. Macosko, RIM, fundamentals of
articles/phenkoft.a01/abstract-fs.html Reaction Injection Molding, 1st ed.,
230 P. M. Frontini, T. R. Cuadrado, Hanser, Münich, 1989.
R. J. J. Williams, Polymer 1982, 23, 250 A.-C. Draye, J. J. Tondeur, J. Mol.
267. Cat. A. Chem. 1999, 138, 135.
231 A. Kumar, S. K. Gupta, N. Somu, 251 K. Schwetlick, R. Noack, F. Stebner,
B. Kumar, Polymer 1982, 23, 215. J. Chem. Soc. Perkin Trans. 2, 1994, 599.
232 S. Ishida, Y. Tsutsumi, K. Kaneko, J. 252 M. Sˇpı́rková, M. Kubı́n, P. Sˇpacek, I.
Polym. Sci. A: Polym. Chem. Ed. 1981, Krakovský, K. Dušek, J. Appl. Polym.
19, 1609. Sci. 1994, 52, 895.
233 A. A. Zavitsas, R. D. Beaulieu, J. R. 253 M. Sˇpı́rková, M. Kubı́n, P. Sˇpacek, I.
LeBlanc, J. Polym. Sci. A: Polym. Krakovský, K. Dušek, J. Appl. Polym.
Chem. Ed. 1968, 6, 2541. Sci. 1994, 53, 1435.
234 S. K. Gupta, A. Kumar, P. K. Pal, 254 K. Schwetlick, R. Noack, J. Chem.
Polymer 1981, 22, 1691. Soc. Perkin Trans. 2 1995, 395.
235 P. K. Pal, A. Kumar, S. K. Gupta, 255 M. C. Pannone, C. W. Macosko, J.
Brit. Polym. J. 1980, 12, 21. Appl. Polym. Sci. 1987, 34, 2409.
236 M. G. Kim, L. W. Amos, E. E. Barnes, 256 M. C. Pannone, C. W. Macosko,
Ind. Eng. Chem. Res. 1990, 29, 2032. Polym. Eng. Sci. 1988, 28, 660.
237 G. Astarloa-Aierbe, J. M. Echeverria, 257 T. A. Speckhard, J. A. Miller, S. L.
J. L. Egiburu, M. Ormaetxea, I. Cooper, Macromolecules 1986, 19,
Mondragon, Polymer 1998, 39, 3147. 1558.
References 151

258 J. A. Miller, T. A. Speckhard, S. L. 277 S. I. Kuchanov, Methods of Kinetic

Cooper, Macromolecules 1986, 19, 1568. Calculations in Polymer Chemistry,
259 J. A. Miller, T. A. Speckhard, J. G. Khimiya, Moscow, 1978 (in Russian).
Homan, S. L. Cooper, Polymer 1987, 278 S. I. Kuchanov, Ye. S. Povolotskaya,
28, 758. Vysokomol. Soyed. 1982, A24, 2190;
260 L. C. Case, J. Polym. Sci. 1960, 45, 397. Polym. Sci. USSR 1982, 24, 2512.
261 J. Scanlan, J. Polym. Sci. 1960, 43, 279 K. Dušek, J. Sˇomvársky, Polym. Bull.
501. 1983, 13, 313.
262 P. L. Mills, Comp. Chem. Eng. 1986, 280 J. Mikeš, K. Dušek, Macromolecules
10, 399. 1982, 15, 93.
263 M. R. P. F. N. Costa, J. Villermaux, 281 S. I. Kuchanov, A. G. Kholostyakov,
Ind. Eng. Chem. Res. 1988, 27, 421. J. Polym. Sci. A: Polym. Chem. 1999,
264 J. Abate, W. Whitt, Oper. Res. Lett. 37, 2145.
1992, 12, 245. 282 M. R. P. F. N. Costa, R. C. S. Dias,
265 A. Brandolin, M. Asteasuain, C. Chem. Eng. Sci. 1994, 49, 491.
Sarmoria, A. López-Rodriguez, K. S. 283 R. Courant, D. Hilbert, Partial
Whiteley, B. del Amo Fernández, Differential Equations (Methods of
Polym. Eng. Sci. 2001, 41, 1156. Mathematical Physics, Vol. II), pp. 97–
266 S. I. Kuchanov, S. V. Korolev, M. G. 105, Interscience Publishers, New
Slin’ko, Polym. J. 1983, 15, 785. York, 1962.
267 S. I. Kuchanov, S. V. Korolev, M. G. 284 L. M. Pis’men, S. I. Kuchanov,
Slin’ko, Vysokomol. Soyed. 1982, A42, Polym. Sci. USSR 1971, 13, 890.
2160; Polym. Sci. USSR 1982, 24, 285 S. I. Kuchanov, L. M. Pis’men,
2477. Polym. Sci. USSR 1972, 14, 147.
268 M. Gordon, G. R. Scantlebury, Proc. 286 S. I. Kuchanov, L. M. Pis’men,
Roy. Soc. 1966, A292, 380. Polym. Sci. USSR, 1972, 14, 985.
269 K. Dušek, M. Gordon, S. B. Ross- 287 H. Galina, A. Szustalewicz,
Murphy, Macromolecules 1978, 11, Macromolecules 1989, 22, 3124.
236. 288 F. D. Magalhães, M. R. P. F. N.
270 D. Stauffer, Pure Appl. Chem. 1981, Costa, in 3rd Berlin International
53, 1479. Workshop on Polymer Reaction
271 M. Gordon, J. A. Torkington, Pure Engineering, p. 311, K. H. Reichert,
Appl. Chem. 1981, 53, 1461. W. Geiseler (Eds.), DECHEMA
272 S. I. Kuchanov, Dokl. Akad. Nauk. Monographs, VCH, Weinheim, 1989.
SSSR 1987, 294, 633. 289 S. I. Kuchanov, T. V. Zharnikov, J.
273 S. I. Kuchanov, S. V. Korolev, S. V. Stat. Phys. 2003, 111, 1273.
Panyukov, Adv. Chem. Phys. 1988, 72, 290 M. R. P. F. N. Costa, R. C. S. Dias, in
115. 6th International Workshop on Polymer
274 W. H. Abraham, Ind. Eng. Chem. Reaction Engineering, DECHEMA
Fundam. 1963, 2, 221. Monographs, Vol. 134, pp. 197–205,
275 H. Kilkson, Ind. Eng. Chem. Fundam. K. H. Reichert, H.-U. Moritz (Eds.),
1964, 3, 281. Wiley-VCH, Weinheim, 1998.
276 H. Kilkson, Ind. Eng. Chem. Fundam. 291 M. R. P. F. N. Costa, J. Villermaux,
1968, 7, 354. Ind. Eng. Chem. Res. 1989, 28, 702.

Free-radical Polymerization: Homogeneous1

Robin A. Hutchinson

Free-radical polymerization (FRP) is one of the most important commercial pro-

cesses for preparing high molecular weight polymers. It can be applied to almost
all vinyl monomers under mild reaction conditions over a wide temperature range
and, although requiring the absence of oxygen, is tolerant of water. Multiple mono-
mers can be easily copolymerized via FRP, leading to the preparation of an endless
range of copolymers with properties dependent on the proportion of the incorpo-
rated comonomers. This chapter will provide an overview of the kinetics and mech-
anisms, and the techniques used to construct mathematical representations of
bulk and solution FRP. The description will also serve as a good base for the sus-
pension and emulsion FRP chapters that follow.

FRP Properties and Applications

Polymers produced via free-radical chemistry include the following major families:

 Low-density polyethylene (LDPE) and copolymers, used primarily in films and

packaging applications. LDPE has density of <0.94 g cm3 , and is produced via
high-pressure free-radical polymerization; polyethylenes of higher density (and
polypropylene) are produced via transition metal catalysis, as described else-
where (see Chapter 8).
 Poly(vinyl chloride) and copolymers, used primarily to produce pipe and fittings,
flooring material, and films and sheet.
 Polystyrene and its co- and terpolymers with acrylonitrile and butadiene. Homo-
polymer is used for packaging and containers, while the acrylonitrile-containing
polymers are used for various molded products in the appliance, electronics, and

1) The symbols used in this chapter are listed at

the end of the text, under ‘‘Notation’’.

Handbook of Polymer Reaction Engineering. Edited by T. Meyer, J. Keurentjes

Copyright 8 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
ISBN: 3-527-31014-2
154 4 Free-radical Polymerization: Homogeneous

automotive industries. Styrene–butadiene is the most widely used synthetic

 Acrylic- and methacrylic-based polymers. Poly(methyl methacrylate) (PMMA),
due to its transparency and weatherability, is used extensively in signs, lighting
fixtures, and windows. Polyacrylates and copolymers are used extensively in the
adhesives and coatings markets, and are also combined with acrylonitrile to
make acrylic fibers.
 Poly(vinyl acetate) and copolymers, used extensively in adhesives, coatings, and
paper and textile treatment.
 Fluoropolymers including polytetrafluoroethylene and copolymers, used widely
in the wire and cable industry. They also have many specialty applications as
coatings due to their inertness and low-surface tension.

In addition to these major product families, there exist many other smaller-volume
(but often high-value) polymeric products synthesized via free-radical chemistry.
These products are manufactured via heterogeneous (emulsion, suspension) or
homogeneous (bulk, solution) polymerization in a wide range of reactor configura-
tions ranging from tubular to well-mixed tanks (and everything in between) in pro-
cesses that may be continuous, batch, or semi-batch. FRP kinetics together with re-
actor design and operating conditions controls the composition and architecture of
the polymer produced. As with other chemistries, the final product consists of a
mixture of polymer chains with varying length, composition, and structural proper-
ties such as branch points and branch lengths. The processing and end-use proper-
ties of the polymer depend upon the distributions of these characteristics, not just
the average values. Thus it is necessary to develop methods to relate the fundamen-
tal kinetic mechanisms of free-radical polymerization to these distributed charac-

Chain Initiation

Free-radical polymerization, like other chain growth mechanisms, involves the se-
quential addition of vinyl monomers to an active center. The defining characteristic
of FRP is that the active centers are free radicals. Each growing polymeric radical
increases in size rapidly; a typical chain is initiated, grows to high molecular
weight, and is terminated in the time scale of, at most, a few seconds. When the
macromolecule stops growing it cannot generally react further (barring side reac-
tions), and is considered ‘‘dead’’. These dead chains have a residence time of mi-
nutes or hours in the polymerization reactor, and the final polymer product is an
intimate mixture of chains formed under time and/or spatially varying conditions.
The free radicals that initiate polymerization are in most cases generated by ther-
mal or photochemical homolytic cleavage of covalent bonds. Commercial initiators
include azo and peroxy compounds. The driving force for the dissociation of azo
4.2 Chain Initiation 155

initiators is the formation of the stable nitrogen molecule and resonance stabilized
tertiary radicals, as shown for 2,2 0 -azobisisobutyronitrile (AIBN) in Scheme 4.1.


N2 + 2 C CH3


Scheme 4.1. Decomposition of AIBN.

Many types of peroxides (R-O-O-R 0 ) are also utilized, including diacyl peroxides,
peroxydicarbonates, peroxyesters, dialkyl peroxides, and inorganic peroxides such
as persulfate, the latter being used mainly in water-based systems. The rate of per-
oxide decomposition as well as the subsequent reaction pathway is greatly affected
by the nature of the peroxide chemical structure, as illustrated for tert-butyl perox-
yesters in Scheme 4.2. Pathway (a), the formation of an acyloxy and an alkoxy rad-
ical via single bond scission, is favored for structures in which the carbon atom in
the a-position to the carbonyl group is primary (for example, tert-butyl peroxyace-
tate, R ¼ CH3 ). Pathway (b), concerted two-bond scission, occurs for secondary
and tertiary peroxyesters (for example, tert-butyl peroxypivalate, R ¼ C(CH3 )3 ) [1,
2]. The tert-butoxy radical formed in both pathways may decompose to acetone
and a methyl radical, or abstract a hydrogen atom to form tert-butanol.
The exact nature of the decomposition pathway is important, as it plays a role in
the efficiency of the primary radicals in initiating new polymer chains. When an
initiator decomposes, the primary radicals are nearest neighbors surrounded by a
‘‘cage’’ of other molecules through which they must diffuse to escape from each
other before they recombine. Once one radical leaves the cage it is extremely un-
likely that the pair will encounter each other again. In-cage reactions for pathway
(a) in Scheme 4.2 will lead to the re-formation of the original initiator molecule,
while the in-cage reactions for pathway (b) lead to the formation of an ether which
is not capable of forming free radicals by another dissociation step.

(a) O (b) .
R O +O CH3 R O CH3 R + CO2 + O CH3


Scheme 4.2. Decomposition of tert-butyl peroxyesters. The
preferred pathway depends on the nature of the R substituent.

In general, peroxide decomposition can yield both carbon- and oxygen-centered

primary radicals that add to the carbon–carbon double monomer bond to form a
new propagating chain, abstract hydrogen atoms from other molecules (including
156 4 Free-radical Polymerization: Homogeneous

polymer) in the system, or recombine to form inactive compounds. The relative

rates of these processes are dependent on both the nature of the primary radicals
and the monomer system. The mechanistic pathway followed will influence end-
group structures of the polymeric products, affecting final properties such as ther-
mal stability. Further details on initiator decomposition kinetics and mechanistic
pathways can be found in the excellent monograph by Moad and Solomon [3].
The above discussion highlights the fact that the simple initiation process of
most polymer texts is the exception rather than the rule, with the fraction of pri-
mary radicals that initiate a new polymer chain a complex function of the reaction
system. Nonetheless, the kinetic treatment is usually simplified by the introduction
of a fractional initiator efficiency ( f ), formally defined by Eq. (1), where n is the
number of moles of primary radicals generated per mole of initiator; n ¼ 2 for
most common initiators.

Initiation Rate of Propagating Chains

f ¼ ð1Þ
nðRate of Initiator DisappearanceÞ

The initiator efficiency is normally in the range 0.4–0.9, with a low value indicating
inefficient usage of the initiator and potentially a high formation rate of undesired
The kinetic descriptions in this chapter are developed for thermal unimolecular
scission of a compound to yield two radicals, as this is the most common means of
generating radicals in industrial systems. Thermal initiation of monomers is an ad-
ditional mechanism capable of forming primary radicals at higher temperatures, as
discussed for styrene (see Section Other initiation systems are also avail-
able, and bear a brief mention at this point. Initiators with multiple peroxide link-
ages ðn > 2Þ have been the subject of recent academic study [4]. Photoinitiators
that produce radicals by ultraviolet irradiation are commonly used to initiate cross-
linking and curing reactions in polymeric systems, as the rate of initiation can be
controlled through the intensity and location of the light source. And finally, a re-
dox (reduction–oxidation) process is often used to initiate chains in emulsion poly-
merization (see Chapter 6).

Polymerization Mechanisms and Kinetics

It is important to achieve an understanding of how the basic FRP mechanisms

control polymerization rate and average polymer chain length. This section starts
with the derivation of appropriate kinetic expressions for a single monomer
system. Complicating (but industrially important) secondary reactions are then
introduced, followed by the extension to multi-monomer systems. Dispersed
throughout are up-to-date estimates for important free-radical polymerization rate
coefficients, and descriptions of how they are obtained experimentally.
4.3 Polymerization Mechanisms and Kinetics 157


Many commercial polymers, including polystyrene, PMMA, and LDPE, are synthe-
sized via homogeneous free-radical polymerization of a single monomer. Homo-
polymer properties are controlled by average chain length and chain-length distri-
bution as well as, in some cases, structural characteristics such as branching level. Basic Mechanisms

The basic set of FRP mechanisms includes initiation, propagation, termination,
and transfer to monomer and solvent or transfer agent, as shown in Scheme 4.3.

Initiator Decomposition I 
→2 f I ∗

Chain Initiation I ∗ + M 
→ P1

Chain Propagation Pn + M → Pn +1

Chain Termination
By Combination Pn + Pm 
→ Dn+m

By Disproportionation Pn + Pm 
→ Dn + Dm

Chain Transfer

k mon
Pn + M 
→ Dn + M *
To Monomer k mon
M * + M 
→ P1

k sol
Pn + S →
Dn + S *
To Solvent or Agent k sol
S * + M →
Scheme 4.3. Basic free-radical homopolymerization mechanism.

The subscript n denotes the number of monomeric units in growing polymer

radicals (Pn ) and dead polymer chains (Dn ). Each reaction has an associated kinetic
rate law expression and a specific rate coefficient. The free-radical initiator (I) un-
imolecularly decomposes (with rate coefficient kd ) to form two primary radicals (I  )
with efficiency f. Chain initiation occurs when the primary radical adds to mono-
mer M, and chain growth continues via successive addition of monomer units to
the radical center (chain propagation, with rate coefficient k p ). Bimolecular cou-
pling of two growing chains results in the loss of two radicals from the system
and the formation of either one (termination by combination, k tc ) or two (termina-
tion by disproportionation, k td ) dead polymer chains. Chain stoppage may also oc-
cur via a transfer mechanism, where the growing radical abstracts a weakly bonded
158 4 Free-radical Polymerization: Homogeneous

atom (usually hydrogen) from monomer or other molecules (solvent or chain-

transfer agent, denoted by S) in the system to generate a dead polymer chain as
well as a new radical that initiates another polymer chain.
A key assumption implicit in the formulation of Scheme 4.3 is that the rates
of propagation, transfer, and termination reactions are independent of n, the
length(s) of the radical(s) involved. It is known that propagation and likely transfer
reactions involving very short chains (n ¼ 1; 2; 3) are faster by a factor of 10 than
addition to long-chain radicals [5], but this effect diminishes rapidly with chain
length and has a negligible effect on the overall kinetics of the system. Chain ter-
mination, the coupling of two polymeric radicals, is a very fast chemical process
that is controlled by the rate at which the two radicals find each other in the reac-
tion system. The nature of this diffusion control makes termination the most com-
plex reaction in the polymerization process since the apparent rate coefficient can
vary greatly with system conditions such as monomer conversion and solution vis-
cosity (see Section 4.3.3). Although termination may also exhibit some chain-
length dependence, most engineering treatments of FRP neglect this complex de-
pendence. For further discussion of the individual mechanisms of Scheme 4.3 and
their rate coefficients, see Section
The set of rate laws that can be written from Scheme 4.3 is given by Eqs. (2)–(6).

Initiator Decomposition R d ¼ kd ½I ð2Þ

Chain Initiation R init ¼ 2f kd ½I ð3Þ
Chain Propagation R p ¼ k p ½M½Ptot  ð4Þ
2 2
Chain Termination R term ¼ ðk tc þ k td Þ½Ptot  ¼ k t ½Ptot  ð5Þ
Chain Transfer R tr ¼ ðktrmon ½M þ ktrsol ½SÞ½Ptot  ð6Þ

Here Ptot represents the concentration of all polymer radicals in the system [Eq.

½Ptot  ¼ ½Pn  ð7Þ

In some literature the right-hand side of the termination rate expression [Eq. (5)] is
written as 2k t ½Ptot  2 . The mode of termination – combination or disproportionation
– has no effect on the overall termination rate, and thus the two events can also be
expressed by the nomenclature of Eq. (8), where d is the fraction of the termination
events that occur by disproportionation.

k td
k t ¼ k tc þ k td ; d¼ ð8Þ
k tc þ k td

The following assumptions are widely accepted and usually valid in FRP systems:
4.3 Polymerization Mechanisms and Kinetics 159

 The small radical species I  ; M  , and S  are not consumed by side reactions and
do not accumulate in the system, but are converted to polymeric radicals with
100% efficiency. Thus, the total rate of polymer radical formation is given by
(R init þ R tr Þ. The net formation of polymeric radicals is R init , since transfer
events both consume and create a polymeric radical species.
 With a continuous source of new radicals in the system, an equilibrium is
achieved instantaneously between radical generation and consumption, such
that R init ¼ R term . This characteristic, proven to be true for almost all FRP condi-
tions [6], is a result of the fast dynamics of radical reactions compared to that of
the overall polymerization system. Often referred to as radical stationarity or the
quasi-steady-state assumption (QSSA), it leads to the well-known analytical expres-
sion for total radical concentration [Eq. (9)].
 1=2  1=2
R init 2f kd ½I
½Ptot  ¼ ¼ ð9Þ
kt kt

 The consumption of monomer by chain-initiation or transfer events is negligible

compared to that by propagation. This result, called the long-chain hypothesis
(LCH), must be true if high molecular weight polymer is being produced. Thus
the rate of polymerization (disappearance of monomer) can be taken as equal to
the rate of propagation (R pol ¼ R p ) with the rate of heat generation proportional
to the rate of this exothermic reaction.

Under these (generally valid) assumptions, the classic expressions for rate of poly-
merization (R pol ), kinetic chain length (u, the average number of monomer units
on a living chain), and instantaneous degree of polymerization (DPninst , the average
number of monomer units on a dead polymer chain formed at any instant) are
given in Eqs. (10)–(12), respectively.
2f kd ½I
R pol ¼ k p ½M½Ptot  ¼ k p ½M ð10Þ

Rp k p ½M
u¼ ¼ ð11Þ
R term þ R tr k t ½Ptot  þ ktrmon ½M þ ktrsol ½S

k p ½M
DPninst ¼ ð12Þ
ðk td þ 0:5k tc Þ½Ptot  þ ktrmon ½M þ ktrsol ½S

The difference between Eqs. (11) and (12) arises because termination by combina-
tion yields a single polymer chain such that the chain length of dead polymer
formed (DPninst ) is greater than the chain length of polymer radicals (u) in the sys-
tem at the same instant.
Table 4.1 lists the range of concentration and rate coefficient values typically en-
countered in homogeneous free-radical polymerization systems at low conversion.
These can be combined with Eqs. (9)–(12) to illustrate the tradeoffs involved be-
160 4 Free-radical Polymerization: Homogeneous

Tab. 4.1. Typical values of coefficients and concentrations in

low-conversion homogeneous FRP systems.

Coefficient/Concentration Typical range

kd [s1 ] 106 –104

f 0.4–0.9
k p [L mol1 s1 ] 10 2 –10 4
k t [L mol1 s1 ][a] 10 6 –10 8
ktrmon =k p 106 –104
ktrsol =k p 106 –103
[I] [mol L1 ] 104 –102
[M] [mol L1 ] 1–10
[S] [mol L1 ] 1–10
[a] At low conversion.

tween the desire for high throughput (R pol ) and the need to produce high MW
(DPn ) polymer.

 The denominator of Eq. (12), the rate of dead chain formation, must be of the
order 105 –108 chains L1 s1 in order to produce polymer with a DPn of
10 2 –10 4 . Individual polymer radicals exist on average only for a fraction of a sec-
ond, as calculated by the expression u=ðk p ½M). Thus after the first few seconds of
polymerization, the concentration of dead polymer chains is higher than that of
polymeric radicals, and by the end of a typical polymerization the concentration
of dead chains is orders of magnitude higher than [Ptot ]. Final polymer MW and
MWD (molecular weight distribution) are controlled by how the concentrations
and kinetic coefficients in Eq. (12) vary with polymer conversion.
 The theoretical MW limit for a system is controlled by transfer events, and is cal-
culated by setting [Ptot ] to zero in Eq. (12). For bulk FRP with no solvent, limiting
values of DPn are 10 4 –10 6 . However, this theoretical limit can only be ap-
proached in homogeneous FRP by reducing rates of polymerization to extremely
low levels. For most systems both termination and transfer events play an impor-
tant role in controlling polymer MW.
 To produce high MW polymer (DPn of 10 2 –10 4 ) it is necessary to keep total rad-
ical concentration low, so that [Ptot ] is of the order 108 –106 mol L1 . This dic-
tates the choice of initiator such that R init (the product of kd and [I]) is also of
order 108 –106 mol L1 s1 .
 Transfer can occur to monomer, solvent or any other species in the system. In
some cases, chain-transfer agents are added deliberately to limit and control poly-
mer DPn . These agents are generally chosen such that the rate of abstraction is
much higher than that which occurs with monomer or solvent (k tr =k p ¼ 103 –
10 0 ) and thus can be added in trace quantities (f 1 mol L1 ). The use of transfer
agents allows for independent manipulation and control of R pol and DPn , but is
only possible if the desired MW is less than that achieved for the corresponding
transfer-free reaction.
4.3 Polymerization Mechanisms and Kinetics 161

 Initial rates of polymerization (monomer consumption rates) at low conversion

are of order 104 –102 mol L1 s1 . Approximately 10 3 –10 5 s are required to
take a batch system to complete conversion. Faster rates can be achieved by in-
creasing R init , but at the expense of decreased polymer MW. Achievable values
of R pol are also often limited by the heat removal capabilities of the reactor sys-
tem, as the heat released by monomer addition is of the order 50–100 kJ mol1 .

It should be cautioned that these statements are generalities for a typical FRP sys-
tem. Rate coefficients vary with monomer, initiator, and solvent choice (see Section as well as polymerization conditions, and the kinetic treatment is compli-
cated by the occurrence of side reactions (Section and the variation of k t
with conversion and other system conditions (Section 4.3.3). These factors necessi-
tate the use of more-powerful modeling techniques to quantitatively describe FRP
systems (Section 4.4). Nonetheless, Eqs. (9)–(12) provide an idea of the factors con-
trolling rate and MW, and are very useful for a qualitative examination of FRP
systems. Coefficients
The coupling of polymer MW and polymerization rate is further illustrated via re-
arrangement of Eq. (12) to give Eq. (13).

1 0:5k tc ½Ptot  k td ½Ptot  ktrmon ktrsol ½S

¼ þ þ þ
DPn k p ½M k p ½M kp k p ½M

0:5k tc R pol k td R pol ktrmon ktrsol ½S

¼ þ þ þ ð13Þ
kp2 ½M 2 kp2 ½M 2 kp k p ½M

R pol [Eq. (10)] and DPn are both dependent on k 2p =k t , with DPn also a function of
mode of termination (disproportionation versus combination) and chain transfer.
Although R pol and DPn are easily measured experimentally, it is not possible to re-
solve the quantities into estimates for individual rate coefficients. Even the estima-
tion of the ratio k 2p =k t from R pol requires independent knowledge of initiator char-
acteristics f and kd , and the assumption that radicals are not being consumed or
retarded by adventitious impurities in the system. These factors have led to consid-
erable scatter in rate coefficients reported in the literature (for example, Ref. 7), es-
pecially for individual rate coefficients [8, 9]. Yet individual values, and knowledge
of how they vary with temperature, are required for model development and an ac-
curate representation of multi-monomer systems. The emergence of specialized
experimental techniques since 1988 has greatly improved this situation and led to
an improved understanding of free-radical polymerization kinetics. The following
discussion highlights some of these advances, as well as summarizing key FRP
rate coefficients as expressed by the Arrhenius law [Eq. (14)].

k i ¼ A i expððEi þ 0:1DVi PÞ=RTÞ ð14Þ

162 4 Free-radical Polymerization: Homogeneous

Activation energies (Ei ) and volumes (DVi ) are reported with units of kJ mol1 and
cm 3 mol1 respectively, with T in K and P in bar. All second-order rate coefficients
are reported with units of L mol1 s1 .

Initiation Thermal scission of an initiator is the most common means of generat-

ing radicals in FRP (see Section 4.2). This unimolecular reaction is characterized
by a first-order rate coefficient (kd , s1 ) so that, for a constant-volume batch system,
Eq. (2) may be integrated to yield Eq. (15), with ½I0 the initial concentration at
t ¼ 0.

½I ¼ ½I0 expðkd tÞ ð15Þ

The decomposition kinetics is often expressed by the half-life of the initiator, the
time needed for the concentration to decrease to half of its initial value [Eq. (16)].

t 1=2 ¼ ðln 2Þ=kd ð16Þ

The requirements for an initiator can vary widely: an initiator with a half-life of 10
h might be used in an academic study so that [I] does not change significantly dur-
ing the course of an experiment, while an initiator used for high-pressure ethylene
polymerization typically has a half-life on the order of a few seconds. Activation en-
ergies for thermal homolysis of peroxide and azo compounds are in the range of
100–150 kJ mol1 , and thus decomposition rates are very temperature-sensitive:
the t 1=2 of benzoyl peroxide drops from 13 h at 70 C to 0.4 h at 100 C. The depen-
dence on pressure is much less, in the range of 0 to 15 cm 3 mol1 [1, 2], but still
important for high-pressure LDPE systems. Special care must be taken in handling
and transport of thermal initiators, especially those with faster decomposition
Values of kd at a particular temperature and pressure can be determined by mea-
suring initiator concentration as a function of reaction time using a technique such
as infrared spectroscopy. The experimental difficulty increases as half-life shortens,
where special care must be taken to eliminate transient nonisothermal effects. De-
composition kinetics are summarized for a wide range of initiators in the Polymer
Handbook [7] and in trade literature available from commercial suppliers. Of spe-
cial note is the recent work of Buback and co-workers that systematically examines
not only how Ed and DVd vary with alkyl substituent for the peroxyester family
(Scheme 4.2), but also how the substituent choice affects the decomposition path-
way and initiator efficiency [1, 2].

Propagation In general, chain growth or propagation proceeds in a highly selec-

tive manner to yield a polymer chain consisting of head-to-tail linkages (Scheme
4.4). In order for high polymer to be formed the propagating radical must not be
too stable: addition must occur at a sufficiently high rate in comparison with com-
peting transfer and termination events.
4.3 Polymerization Mechanisms and Kinetics 163

R1 R1 R1 R1

CH2 C + H2C CH2 C CH2 C

R2 R2 R2 R2
Scheme 4.4. Free-radical chain propagation.

A number of nonsteady polymerization rate techniques have been introduced to

measure k p , many prone to significant error [8, 9]. The introduction of a new
method, that couples pulsed-laser-induced polymerization (PLP) with size exclu-
sion chromatographic (SEC) analysis of the resulting polymer [10], has greatly im-
proved the reliability of k p data. In this technique, a mixture of monomer and pho-
toinitiator is illuminated by short laser pulses separated by a time of t 0 , typically
0.01–0.2 s. Propagation and termination (but no initiation) occur between pulses,
with the radical concentration and the rate of termination both decreasing with
time according to Eq. (5). Those growing macroradicals formed by a laser pulse
which escape termination will all have the same chain length (within a narrow
Poisson distribution) that increases with time. There is a high probability that
these surviving radicals will be terminated at t 0 , when a new population of short
radicals are generated by the next laser pulse. Thus, a significant fraction of dead
chains formed have a chain length DP0 corresponding to a lifetime of t 0 seconds
[Eq. (17)].

DP0 ¼ k p ½Mt 0 ð17Þ

Since radicals have a certain probability of surviving the laser flash and of terminat-
ing at a later pulse, the relative concentration of polymer with chain lengths
2DP0 ; 3DP0 ; . . . is also increased. As a result, PLP produces a well-structured
MWD with peaks at chain lengths of DP0 and its multiples, as shown in Figure
4.1. With known values for t 0 and [M] (kept constant, as samples are pulsed only
long enough to allow a conversion of 1–3%), Eq. (17) yields a direct estimate of k p
from the experimentally-determined value of DP0 .
PLP-SEC has proven to be a very simple and robust experimental technique for
determining k p and its temperature dependence, provided adequate care is taken
with SEC analysis. Data collected from various research laboratories around the
world have been compiled in a series of papers for styrene [11] and methacrylates
[12–14]. These papers provide benchmark k p data for these monomer systems, il-
lustrate the good agreement between facilities (typically 10–20%), and make rec-
ommendations for best experimental practices.
Table 4.2 summarizes the Arrhenius parameters for free-radical propagation of
a wide range of monomers, as determined by PLP-SEC. An extensive review by
Beuermann and Buback [15] contains further discussion and data for additional
monomers. The data for acrylates are limited, due to measurement difficulties aris-
ing from additional secondary mechanisms that occur (see Section What
164 4 Free-radical Polymerization: Homogeneous



wt log(MW) 0.6 DP0



3.5 4 4.5 5 5.5 6 6.5
Log MW
Fig. 4.1. A typical PLP-generated MWD, as measured by SEC
for poly(butyl methacrylate) produced by PLP of bulk butyl
methacrylate at 25 C and a laser repetition rate of 10 Hz. DP0
is determined from the inflection point of the peak, obtained by
differentiating the distribution.

is striking is not only the large variation in k p and activation energies between
monomer families, but also the similar behavior within a monomer family. All
methacrylates exhibit a similar temperature (Ep of 21 to 23 kJ mol1 ) and pressure
(DVp of 15 to 17 cm 3 mol1 ) dependence, and the k p values for alkyl methacry-
lates at 50 C increase with increasing size of the alkyl ester group (methyl to do-
decyl) by a factor of less than 2. The alkyl acrylate family exhibits exactly the same
trends, although the activation energies (17 to 18 kJ mol1 ) and volumes (10 to
12 cm 3 mol1 ) are lower than for methacrylates, and the values of k p at 50 C

Tab. 4.2. Arrhenius k p parameters for various monomers determined by PLP-SEC.[a]

Monomer Ep DVp Ap ˚
k p at 50 C/1 atm
[kJ molC1 ] [cm 3 molC1 ] [L molC1 sC1 ] [L molC1 sC1 ]

Ethylene 34.3 27.0 1:88  10 7 54

Styrene 32.5 12.1 4:27  10 7 238
Methyl methacrylate 22.4 16.7 2:67  10 6 648
Butyl methacrylate 22.9 16.5 3:78  10 6 757
Dodecyl methacrylate 21.0 16.0 2:50  10 6 995
Glycidyl methacrylate 22.9 15.0 6:19  10 6 1230
Cyclohexyl methacrylate 23.0 16.2 6:29  10 6 1204
2-Hydroxypropyl 20.8 n.d. 3:51  10 6 1504
Vinyl acetate 20.7 10.7 1:47  10 7 6625
Methyl acrylate 17.7 11.7 1:66  10 7 22 900
Butyl acrylate 17.4 n.d. 1:81  10 7 27 900
Dodecyl acrylate 17.0 11.7 1:79  10 7 32 000
[a] All values taken from Ref. 15; n.d., not determined.
4.3 Polymerization Mechanisms and Kinetics 165

are 30 to 40 times greater. The propagation behavior is affected more significantly

by electron-donating or -withdrawing substituents present in functional monomers
such as glycidyl and 2-hydroxypropyl methacrylates. The PLP-SEC method has also
been utilized to show that there is little to no solvent influence on propagation ki-
netics of most monomers [15]. However, the propagation kinetics of acrylic acid in
water is a strong function of concentration, a result that may be explained by a
change in local monomer concentration around the propagating radical center
With accurate estimates of k p now available, the possibility of correlating the co-
efficient with the structural characteristics of the propagating radicals and mono-
mers is receiving attention. While some progress has been made in relating activa-
tion energy to reaction enthalpy corrected for polar and resonance factors [17],
more work is required for quantitative predictions.

Termination With independent measures of k p available, k t can now be estimated

from the lumped ratio of k 2p =k t . Specialized techniques involving pulsed-laser-
induced polymerization have also been developed to yield accurate estimates of
the ratio k p =k t [15]. Termination rates in FRP are always diffusion-controlled so
that the apparent value of k t depends on the conditions under which it has been
measured, including the lengths of the radicals involved in the reaction. Nonethe-
less, the assumption of chain-length independence is usually made for modeling
of commercial FRP systems, since the errors introduced are not large. For a more
detailed discussion of the diffusion-controlled nature of the termination reaction,
including its strong conversion dependence, see Section 4.3.3.
Most measurements of k t have been carried out at low monomer conversion,
and thus it is useful to tabulate available data in this regime. Significant scatter,
as much as an order of magnitude, is found in the data contained in the Polymer
Handbook [7], as reflected in the vinyl acetate k t data in Table 4.3. The uncertainty

Tab. 4.3. Arrhenius kt parameters for various monomers.[a]

Monomer Et DVt At ˚
k t at 50 C/1 atm
[kJ molC1 ] [cm 3 molC1 ] [L molC1 sC1 ] [L molC1 sC1 ]

Ethylene 4.6 15.6 1:6  10 9 2:9  10 8

Styrene 6.5 14 2:2  10 9 2:0  10 8
Methyl methacrylate 4.1 15 4:3  10 8 9:4  10 7
Butyl methacrylate 4.1[b] 15[b] 8:5  10 7 1:9  10 7
Dodecyl methacrylate 4.1[b] 15[b] 2:8  10 7 6:2  10 6
Vinyl acetate – – – (5–50)  10 7 [c]
Methyl acrylate 6.7 20 6:0  10 9 5:1  10 8
Butyl acrylate 4.0 16 5:1  10 8 1:2  10 8
Dodecyl acrylate 1.7 21 2:1  10 7 1:1  10 7
[a] Allvalues taken from Ref. 15, unless otherwise indicated.
[b] Assumed equal to MMA value.
[c] From Ref. 7.
166 4 Free-radical Polymerization: Homogeneous

arises from a number of measurement and interpretation factors [18]. The rest of
the data in Table 4.3 are based on PLP studies, and have a higher level of accuracy
(within a factor of 2); these are taken from the recent review by Beuermann and
Buback [15], adjusted to conform to the convention used in Eq. (5) for the ter-
mination rate. Note the large difference in magnitude between values for sty-
rene, methyl methacrylate, and methyl acrylate (k t of 1  10 8 –5  10 8 L mol1 s1 )
and values for dodecyl acrylate and dodecyl methacrylate (@0:5  10 7 –1  10 7
L mol1 s1 ). This has been attributed to differences in segmental diffusion rates
and/or steric hindrance of the large ester side groups. The reported activation
energies and volumes are consistent with the diffusion-controlled nature of the
The mode of termination is also of importance, as it affects the molecular archi-
tecture of the polymer formed, and thus some of its properties. Polymer polydis-
persity (Mw/Mn ) is 1.5 if all chains are terminated by combination of two radicals,
and 2 if chains are terminated by disproportionation or chain transfer. As shown in
Scheme 4.5, termination by combination results in head-to-head linkages in the
polymer chain, whereas disproportionation results in the formation of an unsatu-
rated end group that may undergo further reaction.

R1 R1

R1 R1
R2 R2
CH2 C + C CH2

R2 R1 R1

R2 R2
Scheme 4.5. Free-radical chain termination by combination and disproportionation.

The relative importance of termination by disproportionation versus combina-

tion expressed by d [Eq. (8)] is difficult to measure. The value depends largely on
the structure of the monomer: d is in the range of 0.5–0.8 for a-methylvinyl mono-
mers such as the methacrylates and 0.05–0.2 for styrene and acrylates [3]. A weak
temperature dependency has been proposed, but is difficult to verify within the
scatter of the data [19].

Chain transfer Chain transfer in radical polymerization involves the transfer of

the radical center from a polymeric radical to another molecule. It can occur to all
of the substances present in the polymerization system, and always causes a reduc-
tion in u and DPninst . Scheme 4.3 includes only transfer to monomer and solvent (or
added transfer agent), but transfer to initiator and dead polymer can also occur.
The former is often neglected since [I] is small relative to the concentration of
other species, while the latter can be an important reaction impacting MWD and
4.3 Polymerization Mechanisms and Kinetics 167

final polymer properties (see Section As well as reducing chain length, the
fragments from the transfer reactions (I  and M  in Scheme 4.3) are incorporated
as end groups in the final polymer product.
The rate and MW equations in Section were derived assuming that the
radicals formed by transfer reinitiate new polymeric radicals quickly, within about
the same time period as that required for a propagation step (k imon A k isol A k p ).
This assumption is valid for transfer to most species, and can be verified by exam-
ining whether addition of the transfer agent has an effect on polymerization rate.
If the low-conversion polymerization rate is significantly decreased, the species is
classified as a retarding agent or inhibitor, and Eqs. (9)–(13) no longer describe
the kinetics of the system; see Section
The chain-transfer ability of a compound can be studied by varying its concentra-
tion while holding all other variables fixed. By stopping the reaction at very low
conversion such that concentrations and diffusion-controlled k t values (and thus
R pol and DPninst ) are kept constant, Eq. (13) can be rearranged as Eq. (18).

1 1 k sol ½S
¼ þ tr ð18Þ
DPn ðDPn Þ0 k p ½M

A plot of 1=DPn against ½S=½M should yield a straight line with slope k trsol =k p and
intercept 1=ðDPn Þ0 , where ðDPn Þ0 is the average chain length measured in the ab-
sence of transfer agent. The form of this equation leads to the definition of a chain-
transfer constant C for each species, including monomer, in Eq. (19).

ktrsol ktrmon
Ctrsol ¼ ; Ctrmon ¼ ð19Þ
kp kp

It is these ratios and their temperature dependence that are tabulated in standard
references (for example, Ref. 7).

Solvent/transfer agent For some homogeneous FRP systems, solvent is added not
for its chain-transfer ability, but as a diluent to control viscosity and/or heat trans-
fer. In other systems, a small amount of chain-transfer agent (CTA) is added spe-
cifically to control and reduce the MW of the polymer. Thus values for Ctrsol may
vary widely from a low of 106 –104 to a high of 101 –10 1 , depending on the
number of weakly bonded atoms (generally hydrogen or halogen) on the transfer
agent and their ease of abstraction. For very active compounds (for example, thiols
and halogenated compounds) with Ctrsol > 101 it is necessary to account for the
consumption of CTA during the course of the polymerization, since a changing ra-
tio of ½S=½M will cause a corresponding drift in polymer MW. In such cases care-
ful control of CTA addition to the system is required. If it is added at higher con-
centrations, telomerization (formation of low-MW species) will occur rather than
Table 4.4 summarizes values for a few typical solvents and CTA compounds at
50 C. Most organic compounds have low transfer rates (Ctrsol of 106 –104 ) so that
168 4 Free-radical Polymerization: Homogeneous

Tab. 4.4. Transfer constants (k trsol =k p ) to solvents and transfer agents at 50 C.[a]

Styrene Methyl Methyl Vinyl Ethylene[b]

methacrylate acrylate acetate

Benzene 2  106 4  106 2  105 1  104 9  104 [b]

Toluene 1  105 2  105 1  104 2  103 1:3  102 [b]
Ethyl acetate 5  104 1  105 6  105 2  104 5  103 [b]
Triethylamine 5  104 8  104 4  102 4  102 1:8  102 [b]
CCl 4 1  102 2  104 2  104 0.8 1.0[b]
1-Butanethiol 20 0.6 1.5 50 6.0[b]
[a] Representative values at 50 C [7]; there can be an order of
magnitude range in literature data.
[b] Ethylene values at 130 C and 1360 atm [20].

transfer is important only when they are present in high concentration (that is, as
solvent). They tend to be more reactive toward a radical such as ethylene or vinyl
acetate than a resonance-stabilized radical such as that of styrene. For a given
radical type, aliphatic compounds that yield tertiary radicals will be more effec-
tive transfer agents than those that produce secondary radicals. Halomethanes
and thiols, used as MW-controlling agents due to their high transfer rates, re-
act faster with nucleophilic radicals such as styrene or vinyl acetate than with
(meth)acrylates. Abstraction reactions from organic solvents generally have higher
activation energies than propagation, with (Etr  Ep ) in the range of 20–50 kJ mol1
[21, 22]. The Polymer Handbook [7] and the monograph by Moad and Solomon [3]
provide a summary of available data for a wider range of monomer–CTA (solvent)

Monomer Transfer to monomer cannot be avoided, and the maximum upper limit
of chain length that can be achieved in a polymerization is Ctrmon , assuming the ab-
sence of all other transfer and termination events. The details of the mechanism
depend upon the specific monomer involved. For monomers that contain aliphatic
hydrogens such as vinyl acetate and (meth)acrylates, the transfer process involves
H-atom abstraction to form an unsaturated new radical, as shown for vinyl acetate
in Scheme 4.6. The polymer chain that grows from this radical thus contains an
unsaturated end group that may undergo further reaction. Transfer to monomer
rates are generally very low, and are difficult to measure experimentally since the
ratio with propagation (R trmon =R p ) is independent of [M]. Table 4.5 summarizes

CH2 C + H2C C CH2 CH2 + H2C C
Scheme 4.6. Free-radical chain transfer to monomer (vinyl acetate).
4.3 Polymerization Mechanisms and Kinetics 169

Tab. 4.5. Transfer constants (k trmon =k p ) to monomer.[a]

C trmon E tr C E p [kJ molC1 ] Reference

Styrene 5  105 – 7
Methyl methacrylate 2  105 23.7 23
Butyl acrylate 6  105 15.2 24
Vinyl acetate 2  104 – 7
Ethylene[b] 4  104 [b] 43.9 25
[a] Representative values at 50 C, unless otherwise noted; values
reported in Ref. 7 can show an order of magnitude range.
[b] At 250 C and 2000 bar.

the range of values found in the literature. Ctrmon increases with temperature, with
(Etrmon  Ep ) in the range of 10–40 kJ mol1 . Additional Mechanisms

The basic mechanisms of Scheme 4.3 are common to and occur in every FRP
system. Other mechanisms are more system specific, dependent not only on the
choice of monomer but also the process operating conditions. These additional
mechanisms complicate the kinetic analysis and often play an important role in
controlling polymerization rate and polymer structure under typical industrial op-
erating conditions.

Thermal initiation Primary radicals in most FRP systems are generated by scis-
sion of an added initiator. Free-radical polymerizations can also be initiated by the
monomer itself, or by reactions involving trace impurities in the system. Generally
the rate of radical generation by these processes is very low – negligible compared
with R init from the added initiator. Styrene, however, exhibits significant thermal
polymerization at temperatures of 100 C or more. Indeed, it is not necessary to
add initiator to styrenic systems at high temperatures, as the rate of thermal initia-
tion is sufficient to make an industrially viable process [26]. Acrylates and metha-
crylates have also been reported to undergo thermal polymerization, but at a signif-
icantly slower rate than styrene.
The mechanism behind the thermal polymerization of styrene is still under de-
bate, but is believed to involve the reversible formation of a dimeric species by a
Diels–Alder reaction, followed by the subsequent hydrogen transfer to a third sty-
rene molecule to form two radicals that can initiate polymerization (Scheme 4.7).
This complex mechanism can be approximated by a third-order dependency on sty-
rene concentration [Eq. (20)] [27].
k therm
3M ! 2I 
R therm ¼ k therm ½M 3 ; k therm ¼ 2:2  10 5 expð114:8=RTÞ ð20Þ
mol 2  s
170 4 Free-radical Polymerization: Homogeneous




Scheme 4.7. Thermal initiation of styrene.

This expression from the mid-1970s has stood the test of time, although the pre-
exponential factor has been increased by a factor of 2–3 to fit 260–340 C data ob-
tained in a more recent study [26].

Retardation and inhibition Some substances retard or suppress free-radical poly-

merization by reacting with primary or polymer radicals to yield nonradical prod-
ucts or radicals that are too stable to add further monomer. By decreasing the
concentration of reactive radicals in the system, polymerization rate is slowed (re-
tardation) or stopped completely (inhibition).
Phenolic inhibitors (for example, hydroquinone, monomethylhydroquinone) are
commonly added to monomers at parts-per-million levels in order to prevent poly-
merization during transport and storage by rapidly and effectively scavenging any
radicals (Eq. (21), in which R  represents I  or Pn ) that may form.
R  þ Z ! dead products; R inhib ¼ kinhib ð½Ptot  þ ½I  Þ½Z ð21Þ

Ideally an inhibitor (Z) stops all polymerization until completely consumed

(R inhib g R p ), at which time polymerization proceeds at a normal rate, as shown
schematically in Figure 4.2. In many academic studies, monomer is distilled or
passed over a column to remove inhibitor before polymerization. This purification
is not done in industry if the concentration of inhibitor in the monomer is much
lower than the concentration of initiator added to the system, since the excess of
initiator-generated radicals quickly consumes the inhibitor.
Retardation, the slowing of polymerization rate by consumption of radicals, can
take different forms and thus must be considered on a case-by-case basis. If R inhib
is of similar magnitude to R p , polymerization will proceed at a lower rate until all
of the retarding species is consumed (curve a in Figure 4.2). The kinetic ex-
pressions (Section cannot be applied since termination is no longer the
4.3 Polymerization Mechanisms and Kinetics 171




Fig. 4.2. Conversion–time plots for normal, retarded, and
inhibited free-radical polymerizations. Retardation cases a and
b are described in the text.

sole mechanism of radical consumption; application of radical stationarity yields

R init ¼ R term þ R inhib .
A different form of retardation occurs when a radical species formed from trans-
fer (S  in Scheme 4.3) reinitiates at a slow rate. In addition to the slower reaction
rate with monomer to form a polymer radical, the termination of S  with other
radicals in the system may also need to be considered (Scheme 4.8). Explicit bal-
ances must be written for S , and the extra mechanisms must be included when
deriving expressions for [Ptot ], R pol , and DPn . As solvent/transfer agent is generally
not completely consumed, the retardation effect will last the duration of the poly-
merization (curve b in Figure 4.2). The degree of retardation depends on the
value of k isol , which can vary with monomer type; many carbon-centered radicals
show much lower reactivity toward vinyl esters (for example, vinyl acetate) than
(meth)acrylates [3].

k sol
Pn + S →
Dn + S *
k sol
S *+ M →
k sol
S *+ Pn →
Scheme 4.8. Retardation of polymerization by solvent or transfer agent.

Oxygen inhibits or retards vinyl polymerizations by the formation of peroxy rad-

icals that are generally unreactive. However, subsequent monomer addition can oc-
cur in some systems; this copolymerization forms polymeric peroxides that affect
the thermal stability of the final polymer product. Good practice requires the re-
moval of air by sparging or freeze–thaw cycles before a reaction is started, and ex-
clusion of air during polymerization by operating under an inert atmosphere (for
example, nitrogen) or refluxing solvent.
172 4 Free-radical Polymerization: Homogeneous

Depropagation The addition of a radical to a double bond – the propagation

reaction – is potentially a reversible process [Eq. (22)].
Pn þ M Ð Pnþ1 ; R dep ¼ kdep ½Ptot  ð22Þ

The relative importance of the reverse reaction is governed by the free energy
change, DGp ¼ DHp  TDSp , where DHp represents the enthalpy and DSp repre-
sents the entropy change upon propagation; polymerization can only occur sponta-
neously when DGp is negative. With most common monomers, depropagation is
negligible at typical reaction temperatures. For some systems, however, the en-
thalpy and entropy terms for propagation are delicately balanced so that depropa-
gation has a significant effect on polymerization rate (R pol ) at higher temperatures
[Eqs. (23)].

R pol ¼ R p  R dep ¼ ðk p ½M  kdep Þ½Ptot  ¼ kpeff ½M½Ptot 

kdep ð23Þ
kpeff ¼ k p 

Examining the reaction from a thermodynamic viewpoint leads to the definition of

½Meq , the monomer concentration for a particular temperature at which the effec-
tive propagation rate coefficient (k p ) and polymerization rate become zero [Eq.
kp DGp DHp DSp 1
K eq ¼ ¼ exp ¼ exp  ¼ ð24Þ
kdep RT RT R ½Meq

This equation can also be rearranged to Eq. (25), to define the ceiling temperature
Tc at which, for a given monomer concentration, the polymerization rate becomes

Tc ¼ ð25Þ
DSp þ R ln½M

It also provides a link between the thermodynamic and kinetic coefficients accord-
ing to Eqs. (26).
DHp ¼ Ep  Edep ; DSp ¼ R ln þ R lnð½MÞ ð26Þ
A dep

Values of DHp for common monomers are tabulated in Table 4.6; an expanded list
may be found in Ref. 7. DSp is more difficult to measure experimentally, but is typ-
ically in the range of 100 to 140 J mol1 K1 . Table 4.6 also contains estimates
of Tc calculated at ½M ¼ 1 mol L1 . Ethylene, vinyl acetate, and acrylates have very
high values, and depropagation does not occur under viable polymerization condi-
4.3 Polymerization Mechanisms and Kinetics 173

Tab. 4.6. Depropagation behavior of common monomers.

CDH p [kJ molC1 ] [a] ˚

Tc [ C][b]

a-Methylstyrene 35 19
Styrene 73 335
Methyl methacrylate 56 194
Methyl acrylate 80 394
Vinyl acetate 88 460
Ethylene 102 577
[a] From Ref. 7.
[b] Calculated for ½M ¼ 1 mol L1 , assuming DSp ¼ 120 J mol1 K1 .

tions. Styrene has a slightly lower ceiling temperature and depropagation must be
considered above 250 C. This temperature range, while higher than that generally
employed for styrene polymerization, is used for some commercial processes [26].
The addition of a methyl group to a monomer greatly reduces Tc , as seen by
comparing values for a-methylstyrene to styrene and methacrylates to acrylates.
Figure 4.3 is a plot of k p measured for dodecyl methacrylate (bulk monomer) by
the PLP-SEC technique [28]. At 180 C, the highest temperature examined, k p is
1 1 1 1
4000 L mol s , less than half of the k p value of 9400 L mol s extrapolated
from the lower-temperature data. With known Arrhenius coefficients for propaga-
tion (Table 4.2), the temperature dependence of kdep may be estimated directly from


log[kpeff (L/mol-s)]





2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6
1000/T (K -1)
Fig. 4.3. Effective propagation rate coefficient for bulk dodecyl
methacrylate measured by PLP. The line indicates the forward
rate coefficient (k p ) calculated according to the Arrhenius
parameters of Table 4.2. Experimental data are taken from
Ref. 28.
174 4 Free-radical Polymerization: Homogeneous

these data. The corresponding values for DHp (54 kJ mol1 ) and DSp (123
J mol1 K1 ) agree well with thermodynamic estimates. The depropagation behav-
ior of many methacrylate monomers is similar [28], and depropagation must be
considered at temperatures above 120 C, especially for systems with low monomer
concentration [29].

Long-chain branching All of the mechanisms presented to this point do not

change the basic linear architecture of the polymer chains, with each repeat unit
linked to two others. Branched polymers, those in which the repeat units are not
linked solely in a linear array, can have significantly different physical properties
than their linear counterpoints, depending upon the number and distribution of
the branches along the polymer backbone as well as their length. Understanding
the mechanisms by which these branches are formed, therefore, is a key compo-
nent to manipulation and control of polymer structure. Branches can be formed
either by intramolecular reactions, discussed in the next subsection, or intermolec-
ular reactions. The reaction of a polymeric radical with another polymer chain, or
long-chain branching, can occur by three distinct mechanisms. The common feature
is the re-activation of a dead polymer chain. As well as creating branched struc-
tures, these reactions significantly broaden the polymer chain-length distribution.

Intermolecular transfer to polymer The transfer of a radical center from a polymeric

radical to another polymer chain is shown in Scheme 4.9. Addition of monomer
to the mid-chain radical produces a polymer with a branch point, with the length
controlled by the number of propagation events before chain transfer or radical–
radical termination occurs. An additional subscript can be added to track the num-
ber of long-chain branches formed [Eq. (27)].
mk tr pol pol
Pn; b þ Dm; c ! Dn; b þ Pm; cþ1 ; R tr ¼ k tr ½Ptot m1 ð27Þ

This reaction does not change the number of monomer units that have been poly-
merized or the number of polymer chains in the system, and thus has no effect on
DPn . The rate expression in Eq. (27) is written assuming that all repeat units on all

R1 R1 R1 R1

CH2 C + CH2 C CH2 CH2 CH2 + CH2 C CH2



+ H2C CH CH2 C CH2
R1 H2C
Scheme 4.9. Long-chain branch formation by chain transfer to polymer.
4.3 Polymerization Mechanisms and Kinetics 175

dead polymer chains have an equal probability of reaction, as represented by m1, the
total concentration of polymerized monomer units in the system [Eq. (28)].

y X
X y
m1 ¼ n½Dn; b  ð28Þ
n¼1 b¼0

Note that the reaction is not proportional to the number of chains in the system,
but to the number of repeat units contained in the chains. This has two important
consequences. First, the importance of transfer to polymer increases with polymer
conversion (x p ) in the system, as seen in Eq. (29) by looking at the ratio of branch-
ing to monomer consumption.

pol pol pol

R tr k ½Ptot m1 k tr x p
¼ tr ¼ ð29Þ
R pol k p ½Ptot ½M k p ð1  x p Þ

Thus polymerizations operating at high monomer conversion will have signifi-

cantly higher branching than low-conversion systems. Second, longer polymer
chains – those with more repeat units – are more likely to participate in a branch-
ing reaction than short chains. Since the re-activated chains increase in length
through subsequent propagation, this leads to a broadening of the molecular
weight distribution reflected by an increase in the weight-average MW (Mw ); even
low levels of branching can increase polydispersity values (Mw/Mn ) to 5–15 com-
pared to 2–3 for linear polymers.
As well as conversion, the importance of transfer to polymer depends upon the
monomer system. The reaction can be important in systems with very reactive rad-
icals such as ethylene [30–32], vinyl acetate [33–35], and acrylate [36, 37] polymer-
izations, but seldom occurs in styrene and methacrylate systems. Transfer to poly-
mer usually occurs via abstraction of a methine hydrogen as shown in Scheme 4.9,
but may also involve other easily abstracted H-atoms, such as the acetate methyl
pol pol
hydrogens on poly(vinyl acetate). Transfer constants to polymer (Ctr ¼ k tr =k p )
are not as readily determined as other transfer constants because the process does
not decrease DPn . Long-chain branching (LCB) levels are usually quite low, less
than 2 per 1000 repeat units, making it difficult to employ NMR. Indirect methods
such as multi-detector SEC [32, 38] are often used, leading to a significant scatter
in reported Ctr values [7]. Like other transfer events, the relative importance in-
creases with temperature.

Reaction with unsaturated polymer chains Termination by disproportionation, trans-

fer to monomer, and chain scission (discussed later in this section) create polymer
chains with terminal unsaturation (denoted in Eq. (30) by D ¼ ). These reactive
chains, sometimes called macromonomers, can add to a growing radical to form a
long-chain branch.
¼ ¼
Pn; b þ Dm; c ! Pnþm; bþcþ1 ; R pol pol
p ¼ k p ½Ptot ½Dtot  ð30Þ
176 4 Free-radical Polymerization: Homogeneous

R1 R1 R1
CH2 C + H2C C CH2 C CH2 C

R2 R2

Scheme 4.10. Long-chain branch formation by addition to a

terminally unsaturated polymer chain.

This reaction is fundamentally different than transfer to polymer in several re-

spects. It is an addition rather than a transfer (H-abstraction) reaction (Scheme
4.10). The reaction rate is dependent on the number of unsaturated chain ends
rather than the number of repeat units in the chains so that its importance relative
to propagation is controlled by the ratio ½Dtot =½M according to Eq. (31).
pol pol ¼
Rp k p ½Ptot ½Dtot 
¼ ð31Þ
R pol k p ½Ptot ½M

Unlike transfer to polymer, the mechanism combines two polymer chains (and
all of their repeat units) into one, affecting both DPn and DPw . Reaction with
terminally unsaturated chains can be important in vinyl acetate [33] and higher-
temperature methacrylate [3] polymerizations.

Reaction with multifunctional monomers Another way to introduce branching is

through addition of a multifunctional monomer to the polymer system. An exam-
ple is the addition of small levels of ethylene glycol dimethacrylate (EGDMA) to
methyl methacrylate (MMA), as shown in Scheme 4.11. Reaction of the first dou-

CH2 C + H2C C C CH2 C
CH2 C + H2C C CH2 C CH2 C
Scheme 4.11. Addition of ethylene glycol dimethacrylate to a
methyl methacrylate radical. The pendent double bond is
attacked by another polymer radical to form a crosslink.
4.3 Polymerization Mechanisms and Kinetics 177

ble bond incorporates EGDMA in the PMMA chain, with the second reactive site
remaining as a pendent double bond. Reaction of this pendent bond with another
polymer radical creates a branched structure termed a crosslink, since the branch
is tetrafunctional rather than trifunctional in nature.
The addition of a small amount of difunctional monomer such as EGDMA to
MMA [39, 40] or divinylbenzene to styrene is a means to increase polymer MW
without decreasing radical concentration. Addition of higher levels leads to an in-
terconnected branched, or network, polymer.

Short-chain branching (intramolecular transfer to polymer) Intramolecular H-atom

abstraction, often called back-biting, occurs via the formation of a six-membered
ring, as shown in Scheme 4.12 for butyl acrylate (nBA). Monomer addition to the
resulting interior radical leaves a short-chain branch (SCB) consisting of two repeat
units. This mechanism has long been known important for high-pressure LDPE
production at 150–300 C [30], where the number of short-chain branches is in
the range of 20–50 per 1000 ethylene repeat units. This level of SCB significantly
decreases the polyethylene crystallinity and gives LDPE some of its unique proper-
ties; LDPE density is 0.92 g cm3 compared to 0.98 g cm3 for linear polyethylene.
Multiple back-biting events lead to other SCB structures; in addition to the com-
mon butyl branch in LDPE, ethyl and 2-ethylhexyl branches have been identified
using 13 C NMR [30, 41, 42].


Scheme 4.12. Formation of a mid-chain radical by
intramolecular chain transfer to polymer. Monomer addition to
the new radical structure creates a short-chain branch in the

In ethylene/nBA copolymerization it is observed that the methine hydrogens on

nBA units in the polymer chain are much more susceptible to abstraction than hy-
drogen abstraction from a CH2 unit in the backbone [43]. This result, along with
the observed high concentration of branch points formed at conditions of very low
polymer concentration during nBA homopolymerization [36], indicates that the in-
tramolecular (back-biting) mechanism is also important in acrylate polymeriza-
tions. There is strong evidence to suggest that monomer addition to the resulting
mid-chain radical, due to its higher stability, is much lower than addition to a
chain-end radical [44]. Thus the mechanism affects nBA polymerization rate as
well as polymer structure [37]. The same type of back-biting reaction has also
178 4 Free-radical Polymerization: Homogeneous

been shown to occur during styrene polymerization at high temperatures (260–

340 C) [26].

Chain scission The radical structure formed by intra- or intermolecular transfer

to polymer is less reactive than a chain-end radical. Under higher-temperature con-
ditions, the radical can undergo b-fragmentation (chain scission) as shown in
Scheme 4.13 for butyl acrylate [45]. The scission can occur in either direction,
yielding a short-chain radical or a short-chain unsaturated trimer species. As well
as lowering polymer MW, the scission event produces an unsaturated chain end
that can react further, as discussed previously (see Scheme 4.10).




+ H2C
Scheme 4.13. b-Scission of butyl acrylate mid-chain radical.

Scission events can occur in any system where mid-chain radicals are formed.
However, scission is more temperature-activated than H-abstraction and thus be-
comes important only at elevated temperatures. The reaction is not believed to oc-
cur during butyl acrylate polymerization at 75 C [37], but is shown to be important
at 140 C [29, 45]. Scission is a dominant mechanism in styrene polymerizations
at 260–340 C [26], and also occurs during LDPE production [30]. Scission of mid-
chain radicals formed via intermolecular transfer to polymer can have a significant
effect on the breadth and the shape of polymer MWD [46].
Kinetic treatment of these more complex mechanisms is often difficult. Equa-
tions (32)–(34) are a network of reactions developed to treat intramolecular trans-
fer, short-chain branch formation, and b-scission for butyl acrylate polymerization
Pn ! Q n ; R bb ¼ kbb ½Ptot  ð32Þ
Q n þ M ! Pnþ1 þ SCB; RSCB ¼ kptert ½M½Q tot  ð33Þ
Q n ! Dn2 þ P2
0:5kb ð34Þ
Q n ! Pn3 þ D3¼ ; R b ¼ kb ½Q tot 

The kinetic coefficients are estimated by measuring the level of quaternary carbons
(equated to short-chain branch level) and terminal unsaturations (D ¼ ) by NMR.
4.3 Polymerization Mechanisms and Kinetics 179

Termination of the mid-chain radical (Q n ) is also considered in the scheme. While

the mechanism provides improved understanding of this complex system, many
questions still remain: does the mid-chain radical terminate at the same rate as
chain-end radicals; what is the reactivity of the unsaturated chain end; and does
the mid-chain radical scission with equal probability in each direction? The reac-
tion engineering challenge is to consider the set of mechanisms required to repre-
sent the basic rate behavior and polymer architecture of the system without intro-
ducing unneeded complexity.


Reaction of two or more monomers by free radical polymerization is an effective

way of altering the balance of properties of commercial products. Addition of the
polar monomer acrylonitrile to styrene (or methyl acrylate to ethylene) produces a
polymer that combines the strength of the base homopolymer with much im-
proved oil and grease resistance. The adhesive and cohesive properties of coatings
resins are balanced by controlling the mix and relative proportions of monomers in
the recipe.
FRP leads to the formation of statistical copolymers, where the arrangement of
monomers within the chains is dictated purely by kinetic factors. However, reactiv-
ity of a monomer in copolymerization cannot be predicted from its behavior in ho-
mopolymerization. Vinyl acetate polymerizes about 30 times more quickly than
styrene (see Table 4.2), yet the product is almost pure polystyrene if the two mono-
mers are copolymerized together in a 50:50 mixture. a-Methylstyrene cannot be ho-
mopolymerized to form high-MW polymer due to its low ceiling temperature (see
Table 4.6), yet is readily incorporated into copolymer at elevated temperatures.
These and other similar observations can be understood by considering copolymer-
ization mechanisms and kinetics. Basic Mechanisms

In copolymerization, the presence of more than one type of monomer adds an ex-
tra degree of complexity to the kinetics. The different monomers form different
radical structures, and the relative rates of chain growth depend on the structure
of both monomer and radical. It is these propagation mechanisms that control
polymer composition (the relative amounts of each monomer unit incorporated
into the copolymer) and sequence distribution (the way in which these monomer
units are arranged within the chain). Developing a set of mechanisms to describe
how radical structure affects termination and transfer rates is required to represent
copolymer chain length and molecular weight distributions.
The most common treatment of free radical copolymerization kinetics assumes
that the reactivity of the polymer radical depends solely on the nature of its termi-
nal monomer unit; that is, that the identity of the penultimate unit on the radical
does not affect its reactivity. This assumption provides a good representation of
polymer composition and sequence distribution, but not necessarily polymeriza-
180 4 Free-radical Polymerization: Homogeneous

tion rate (see Section This so-called terminal model is widely used to model
free-radical copolymerization according to the set of mechanisms in Scheme 4.14.

Initiator Decomposition I 
→2 f I ∗

Chain Initiation I ∗ + M j 
→ P1 j

Chain Propagation Pni + M j →
Pn +j 1

Chain Termination
By Combination Pni + Pmj →
Dn + m

By Disproportionation Pni + Pmj →
Dn + Dm

Chain Transfer
To Monomer Pni + M j 
→ Dn + P1 j

Pni + S →
Dn + S *
To Solvent or Agent k isol
S * + M j →
P1 j
Scheme 4.14. Basic free-radical copolymerization mechanism,
assuming terminal radical kinetics.

In this scheme monomer-j (denoted by Mj ) adds to the initiator primary radical

to form a polymer radical of type-j and unit length. The dead polymer and radical-i
chains of length n (Dn and Pni ) are made up of a mixture of the monomer types in
the system, with their relative amounts governed by the copolymer composition
equation developed below. Chain growth occurs by addition of Mj to radical-i (Pni )
with the propagation rate coefficient k pij dependent on both radical and monomer
type. The rate coefficients for transfer and termination reactions can also be depen-
dent on the nature of the radical center, as indicated by subscripts. However, since
radical–radical termination is a diffusion-controlled reaction, the rate coefficient
can usually be assumed to be independent of radical type, such that kt ¼ k tij for
all i and j.
Most of the kinetic coefficients in Scheme 4.14 are binary parameters, dependent
on the radical and monomer type. Thus the polymerization behavior of three or
more monomers can be estimated reliably from knowledge of the corresponding
binary copolymerizations. For a two-monomer system assuming the long-chain hy-
pothesis, the consumption rates of the two monomers are written as in Eqs. (35).

1 2
R pol1 ¼ k p11 ½Ptot ½M1  þ k p21 ½Ptot ½M1 
1 2
R pol2 ¼ k p12 ½Ptot ½M2  þ k p22 ½Ptot ½M2 
4.3 Polymerization Mechanisms and Kinetics 181

where Ptot represents the concentration of all polymer radicals of type-i in the sys-
tem [Eq. (36)].

½Ptot ¼ ½Pni  ð36Þ

The ratio of the two consumption rates dictates the instantaneous composition
) of the polymer being formed [Eq. (37)].

R pol1 k p11 ½Ptot 2
½M1  þ k p21 ½Ptot ½M1 
¼ ¼ 1 2 ½M 
R pol2 k p12 ½Ptot ½M2  þ k p22 ½Ptot 2

Application of the quasi-steady-state assumption yields the ratio of the radical types
as in Eq. (38), and substitution and rearrangement leads to the well-known poly-
mer composition equation [Eq. (39)], where f1 and f2 are the mole fractions of M1
and M2 in the monomer mixture, and monomer reactivity ratios r1 and r2 are de-
fined as k p11 =k p12 and k p22 =k p21 .

½Ptot  k p21 ½M1 
2  ¼ k ½M 
p12 2

r1 f12 þ f1 f2
Fpinst ¼ ð39Þ
r1 f12 þ 2 f1 f2 þ r2 f22

Equation (39) defines the composition of the copolymer formed at any instant dur-
ing polymerization, and is dependent only on the ratios of the propagation rate co-
efficients and not on their absolute values.
Copolymer properties depend on the distribution of the monomer units along
the chain as well as the average composition. Reactivity ratios also control copoly-
mer sequence distribution. The probability P11 that an M1 unit follows an M1 unit
in the copolymer is equal to the rate of M1 M1 formation divided by the sum of the
rates of all additions to radical-1 [Eq. (40)].

k p11 ½Ptot ½M1  r1 f1
P11 ¼ 1 1 ½M 
¼ ð40Þ
k p11 ½Ptot ½M1  þ k p12 ½Ptot 2 r1 f1 þ f2

The probability P12 that an M2 unit follows an M1 is given by Eq. (41).

P12 ¼ 1  P11 ¼ ð41Þ
r1 f1 þ f2

Similar expressions, Eqs. (42) and (43), can be derived for addition to radical-2:
182 4 Free-radical Polymerization: Homogeneous

k p22 ½Ptot ½M2  r2 f2
P22 ¼ 2 2 ½M  ¼ ð42Þ
k p22 ½Ptot ½M2  þ k p21 ½Ptot 1 r2 f2 þ f1

P21 ¼ 1  P22 ¼ ð43Þ
r2 f2 þ f1

These probabilities can be used to calculate NðM1 ; n i Þ, the fraction of all M1 se-
quences that are exactly n i units long. This is simply the probability of having
ðn i  1ÞM1 M1 linkages followed by an M1 M2 linkage, according to Eq. (44).

n i 1 n 1
NðM1 ; n i Þ ¼ P11 P12 ; NðM2 ; nj Þ ¼ P22j P21 ð44Þ

Thus the fraction of M1 sequences that consists of an isolated M1 unit is P12 , the
fraction that consists of isolated M1 M1 diads is P11 P12 , the fraction of triads is
P11 P12 , and so forth. The number-average length of M1 sequences (N 1 ) is given by
Eq. (45) and the fraction of all M1 units contained in a sequence of length n i is
n i 1 2 2
n i P11 P12 ; that is, the fraction of M1 contained in isolated diads is 2P11 P12 .

1 1 1 1
N1 ¼ ¼ ; N2 ¼ ¼ ð45Þ
1  P11 P12 1  P22 P21

Implicit in these expressions is the approximation, valid for long-chain polymer,

that the number of M1 M2 linkages in a chain is equal to the number of M2 M1 link-
ages. Thus the ratio of M1 to M2 units in the chain must equal the ratio of the re-
spective average sequence lengths [Eq. (46)].

1 N1
¼ ð46Þ

Substitution and rearrangement of this equation yields the polymer composition

equation, Eq. (39). Thus it is possible to estimate reactivity ratios for binary copoly-
mers from triad distributions measured by NMR analysis.
While copolymer composition and sequence distribution are only functions of
the reactivity ratios, the same is not true for polymerization rate. The overall rate
of monomer consumption is given by Eq. (47), where [Mtot ] indicates the total
monomer concentration (½M1  þ ½M2 ).

1 2 1 2
R pol ¼ k p11 ½Ptot ½M1  þ k p21 ½Ptot ½M1  þ k p12 ½Ptot ½M2  þ k p22 ½Ptot ½M2 
XX2 2
¼ k pij fri fj ½Ptot ½Mtot  ð47Þ
i¼1 j¼1

The fraction of radical-i in the system, fri [Eq. (48)] can be calculated from Eq. (38).
4.3 Polymerization Mechanisms and Kinetics 183

Ptot Pi
fri ¼ 1 2
¼ tot ð48Þ
Ptot þ Ptot Ptot

[Ptot ], the total radical concentration, is calculated from an overall radical balance
similar to Eq. (9) and given by Eq. (49).

R init 1=2 2f kd ½I 1=2
½Ptot  ¼ cop ¼ cop ð49Þ
kt kt

The form of Eq. (47) is analogous to the homopolymerization rate expression [Eq.
(10)], with a copolymer-averaged rate coefficient for propagation (generalized for a
system with Nmon different monomers) defined in Eq. (50).

Nmon X
X Nmon
k pcop ¼ k pij fri fj ð50Þ
i¼1 j¼1

For a two-monomer system, application of the QSSA [Eq. (38)] and simplification
lead to Eq. (51).

r1 f12 þ 2f1 f2 þ r2 f22

k pcop ¼ ð51Þ
ðr1 f1 =k p11 Þ þ ðr2 f2 =k p22 Þ

Using Eqs. (47)–(51), the copolymerization reaction rate can be analyzed as for
homopolymerization (see Section, with k p now a function of monomer
composition. Kinetic Coefficients

The traditional method for determining reactivity ratios involves determination of
copolymer composition for a range of monomer feeds at very low conversion; that
is, measuring Fp1 as a function of f1 . NMR measurement of sequence distributions
provides additional information about chain microstructure, but suffers from
greater experimental noise and signal assignment uncertainty from tacticity ef-
fects. There is a large body of published r1 –r2 data for monomer pairings, summar-
ized in Ref. 7. The scatter for these ratios is much less than found in k p and k t
data, but care still must be exercised when extracting values from this and similar
compilations. Some error can arise from the methodology used to estimate r1 and
r2 from Fp1 versus f1 data. The estimation is best accomplished by nonlinear least
squares techniques [48, 49], and a statistical analysis also provides a guide to the
optimal monomer compositions at which experimentation should be performed
to improve the estimates [48]. Error can also occur if polymer conversion is suffi-
ciently high for f1 to deviate significantly from the zero-conversion value (in which
case an integrated form of Eq. (39) must be used [50]), or if the experimental sys-
tem does not remain homogeneous (often observed with acid monomers).
184 4 Free-radical Polymerization: Homogeneous

Tab. 4.7. Monomer reactivity ratios at 50 C[a], r1 is tabulated horizontally and r2 vertically.

Monomer-1 Monomer-2

Styrene Methacrylate Acrylate Vinyl acetate

Styrene – 0.6 0.8 40

Alkyl methacrylate 0.4 – 2.2 20
Alkyl acrylate 0.2 0.4 – 6
Vinyl acetate 0.02 0.03 0.03 –
[a] Representative values Ref. 7.

There are only a few systems for which the terminal model upon which Eq. (39)
is based does not provide a good description of copolymer composition. The result-
ing polymers exhibit an alternating structure (for example, styrene–maleic anhy-
dride), or the observed reactivity ratios in the system vary with monomer concen-
tration or solvent choice. Since they often include a polar monomer with strong
electron-withdrawing or electron-donating properties, alternative kinetic models
that include the formation of monomer complexes have been developed to repre-
sent these systems [3, 51]. The majority of systems, however, are well behaved
and well represented by the terminal model. Table 4.7 summarizes values of reac-
tivity ratios for styrene, alkyl methacrylate, alkyl acrylate, and vinyl acetate systems.
Reactivity ratios for alkyl methacrylates and acrylates (for example, methyl, butyl,
dodecyl) exhibit a family type behavior, with the composition data of various sys-
tems fitted well by a single curve [52]. Values for ethylene copolymer systems at
typical production conditions are summarized in Table 4.8: While r1 (1 ¼ ethylene)
values for alkyl methacrylates, alkyl acrylates, acrylic acid, and methacrylic acid
could not be distinguished within experimental error [53], there were significant
differences in the r2 values [54].
It is informative to consider some of the implications of these values. In Figure
4.4 the relationship between polymer and monomer composition is plotted for var-
ious copolymer systems:

Tab. 4.8. Monomer reactivity ratios for ethylene (monomer-1)

systems at 240 C and 2000 bar [53, 54].

Monomer-2 r1 r2

Alkyl methacrylates 0.058 7

Methacrylic acid 0.058 11
Alkyl acrylates 0.058 4
Acrylic acid 0.058 8
Vinyl acetate [a] 1.0 1.0
[a] From Ref. 7.
4.3 Polymerization Mechanisms and Kinetics 185




Fp 1



0.2 MMA-MA
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Fig. 4.4. Relationship between monomer composition f1 and
corresponding instantaneous polymer composition Fpinst
various monomer pairings (MMA ¼ methyl methacrylate,
Sty ¼ styrene, MA ¼ methyl acrylate, VAc ¼ vinyl acetate,
Eth ¼ ethylene).

 For the case where r1 ¼ r2 ¼ 1:0 (ethylene–vinyl acetate; methacrylate–

methacrylate), the monomers have equal reactivity to propagation and will be
incorporated into polymer at the same ratio as they are in the monomer phase
(Fp1 ¼ f1 ).
 Where r1 > 1 and r2 < 1, the copolymer will always be richer in monomer-1 than
in the monomer phase, so that monomer-1 will become depleted in a batch poly-
merization. The further the reactivity ratios deviate from unity, the greater the
deviation between polymer and monomer composition. Systems that exhibit
this behavior include methacrylate–acrylate polymerizations, and styrene, meth-
acrylates, or acrylates polymerized with vinyl acetate or ethylene.
 With both r1 and r2 less than unity (styrene–acrylate, styrene–methacrylate),
cross-propagation is favored over homopropagation and the copolymer tends
toward an alternating structure. The system has an azeotropic composition at
which the copolymer composition is exactly equal to the monomer composition.

Reactivity ratios exhibit a weak temperature dependence that is often difficult to

measure. With increasing temperature, the ratios tend to approach unity as dem-
onstrated for styrene–butyl acrylate [55], butyl acrylate–methyl methacrylate [56],
and ethylene copolymer systems [53, 54]. The temperature dependence of the latter
agrees well with activation energies reported for addition of monomers to small
radicals [57].
186 4 Free-radical Polymerization: Homogeneous


(L/mol-s) 1000.0


0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
f 1 (MMA)
Fig. 4.5. Relationship between monomer composition f1 and
corresponding copolymer-averaged propagation rate coefficient,
k p . Lines are calculated assuming terminal-model kinetics and
points are experimental data for: MMA–nBA (—, e) [62] and
MMA–styrene (---, C) [59] at 20 C.

While the terminal model effectively represents copolymer composition for most
systems, it does not provide as good of a description of polymerization propaga-
tion behavior. Rate abnormalities have previously been attributed to termination
reactions and represented by introducing physically unrealistic cross-termination
mechanisms [58]. Careful experimental work by Fukuda and co-workers [51], how-
ever, demonstrated that the rate deviations are due to the inadequacy of the ter-
minal model to describe propagation, with further evidence obtained by applica-
tion of the PLP-SEC technique to measure k p . Many common systems show
deviation from terminal propagation kinetics, including styrene–methacrylate [51,
59], styrene–acrylate [60, 61], and acrylate–methacrylate [52, 62] systems. As
shown in Figure 4.5, the measured k p values can be higher or lower than the ter-
minal model predictions, with the deviation substantial in some cases.
The ‘‘implicit penultimate unit effect’’ model, which accounts for the influence
of the penultimate monomer unit of the growing polymer radical on the propaga-
tion kinetics [51], provides a good representation of this behavior [Eqs. (52)].

r1 f12 þ 2f1 f2 þ r2 f22

k pcop ¼
ðr1 f1 =k p11 Þ þ ðr2 f2 =k p22 Þ
k p ½r1 f1 þ f2  k p ½r2 f2 þ f1 
k p11 ¼ 111 k p22 ¼ 222
r1 f1 þ ½ f2 =s1  r2 f2 þ ½ f1 =s2 
4.3 Polymerization Mechanisms and Kinetics 187

The extra parameters s1 and s2 , called radical reactivity ratios, capture the effect of
the penultimate unit on the addition rate of monomer [Eqs. (53)].

k p211 k p122
s1 ¼ s2 ¼ ð53Þ
k p111 k p222

A value of greater than unity for si indicates that a comonomer unit-j in the penul-
timate position increases the addition rate of monomer-i to radical-i compared to
the homopolymerization case.
The kinetics of diffusion-controlled termination in copolymerization is also diffi-
cult to study. The original interpretation of low-conversion rate data was based on a
chemically controlled model utilizing a cross-termination factor [Eq. (54)].

Nmon X
X Nmon
k t 12
kt ¼ k tij fri frj ; F¼ ð54Þ
i¼1 j¼1
k t 11 k t22

Assuming terminal propagation kinetics, the best fit for F was found to be much
greater than unity for styrene–methacrylate, styrene–acrylate, and other systems
such that kt was greater than either homotermination value. When the deviation
of propagation kinetics from the terminal model is taken into account, however,
the estimates for kt become well behaved and bounded by the homotermina-
tion values [51]. A penultimate model [51, 63] that accounts for the influence
of polymer composition on segmental diffusion is required to fit recent acrylate–
methacrylate low-conversion kt data [63, 64]. In order to use this representation
(Eq. (55), with the cross-termination coefficients k t 12; 12 and k t21; 21 fitted to experi-
mental data), it is necessary to track the four possible penultimate radical types in
the system, with frij ði; j ¼ 1; 2Þ the penultimate radical fraction.

ðkt Þ 0:5 ¼ kt0:5 f þ kt0:5
11; 11 r11
f þ kt0:5
21; 21 r21
f þ kt0:5
22; 22 r22
12; 12 r12

While the terminal model and reactivity ratios provide a good description of copoly-
mer composition, the kinetic studies summarized in this section indicate that ad-
ditional parameters and penultimate radical fractions are required to represent k p
and kt . (For a discussion of transfer to monomer in copolymer systems, see Ref.
65.) These mechanistic complexities are often not considered when developing
FRP models for polymer reaction engineering applications. It is expected that this
situation will change as more data become available. Additional Mechanisms

The secondary mechanisms presented for homopolymerization in Section
also occur in multi-monomer systems. The kinetics of depropagation, chain trans-
fer to polymer, and chain scission are strongly influenced by not only the nature of
the monomer and terminal radical, but also the penultimate unit on the polymer
188 4 Free-radical Polymerization: Homogeneous

~~~ M 1 ⋅ + M 1 

 → ~~~ M 1M 1 .
kdep 11
~~~ M 1 ⋅ + M 2 → ~~~ M 1M 2 .
~~~ M 2 ⋅ + M 2 
→ ~~~ M 2 M 2 .
~~~ M 2 ⋅ + M 1 
→ ~~~ M 2 M 1 .
Scheme 4.15. Depropagation in copolymerization for the case
where M2 does not depropagate and M1 depropagates only
when an M1 -unit is in the penultimate position.

Depropagation It is not possible to produce high MW homopolymer close to the

ceiling temperature of the monomer (see Eq. (25) and Table 4.6). Addition of a
non-depropagating monomer (M2 ) to the system disrupts this behavior, as illus-
trated by Scheme 4.15. Depropagation of radical-1 becomes a competitive process
with addition of monomer-2, with monomer-1 units at the radical end becoming
irreversibly trapped in the growing chain as soon as monomer-2 adds. Further-
more, depropagation of monomer-1 will occur only if an M1 unit is also in the pen-
ultimate position; depropagation to the less-stable M2 radical can be assumed to be
unlikely (kdep21 A 0). Thus MMA can be successfully copolymerized with ethylene at
290 C [57], and a-methylstyrene with various comonomers at temperatures well
above its ceiling temperature [66, 67].
Depropagation in copolymerization affects polymer composition, sequence prob-
abilities, and polymerization rate. With Scheme 4.15, the relative rates of monomer
consumption are given by Eq. (56) (compare with Eq. (37) in the absence of depro-

k p11 ½Ptot 2
½M1  þ k p21 ½Ptot ½M1   kdep11 ½P 1 
R pol1 fr11 þ fr21 tot
¼ ¼ 1 ½M  þ k ½P 2 ½M 
R pol2 k p12 ½Ptot 2 p22 tot 2

The ratio fr11 =ð fr11 þ fr21 Þ accounts for the fraction of radical-1 that ends in a 11
diad. The effect of depropagation becomes larger as total monomer concentration
decreases and as the fraction of the depropagating monomer in the system in-
creases. Lowry [68] first derived the composition expressions for the situation
where only one monomer depropagates, and general expressions have been devel-
oped for the situation where all four of the propagation reactions are reversible
[66, 69]. Depropagation must be considered when examining the kinetics of
starved-feed semi-batch copolymerization involving methacrylates at the higher-
temperature conditions typically used to produce acrylic coatings [29, 70].

Chain transfer to polymer The rate of chain transfer to polymer is dependent on

both the reactivity of the radical and the abstractability of the hydrogen atom on
the monomer unit in the polymer chain. For the case of intermolecular chain
transfer (long-chain branching), this is represented by Eq. (57), where m1 repre-
sents the total concentration of polymerized monomer-j units in the system and
4.3 Polymerization Mechanisms and Kinetics 189

nj represents the number of monomer-j units on a particular chain of length n

[Eqs. (58)].

mj k tr
ij j pol
Nmon X
X Nmon
pol j
Pn;i b þ Dm; c ! Dn; b þ Pm; cþ1 ; R tr ¼ i
k trij ½Ptot m1 ð57Þ
i¼1 j¼1

y X
y X
m1 ¼ nj ½Dn; b ; n¼ nj ð58Þ
n¼1 b¼0 j¼1

Active radicals (ethylene, acrylate, vinyl acetate) are more likely to abstract from a
polymer chain than styrenic or methacrylate radicals, and acrylate and vinyl acetate
monomer units on a chain are more likely to have an H-atom abstracted. Thus it is
not uncommon for the intermolecular transfer to polymer rate of one pairing (for
example, acrylate radical to acrylate monomer unit) to dominate the system, with
the overall transfer rate R tr decreasing rapidly with increasing content of the less-
reactive monomer.
The situation is more complicated in the case of intramolecular transfer, which
occurs through the formation of a six-membered ring. In the case of acrylate (1)/
methacrylate (2), it can be assumed that the methacrylate radical is not reactive
enough to back-bite and that the acrylate radical can only abstract hydrogen if the
antepenultimate unit on the chain is also an acrylate unit. Thus back-biting can oc-
cur only for two monomer sequences (M1 M1 M1 and M1 M2 M1 ) at the radical end,
as shown in Scheme 4.16 [70]. The overall back-biting rate must be corrected for
the sequence probabilities [Eqs. (40)–(43)] at the chain end, according to Eq. (59).



Scheme 4.16. Back-biting reaction in high temperature
copolymerization of butyl acrylate and butyl methacrylate
(R ¼ H or CH3 ).
190 4 Free-radical Polymerization: Homogeneous

R bb ¼ kbb ½Ptot ðP11 P11 þ P12 P21 Þ ð59Þ

Back-biting mechanisms have also been examined for styrene/acrylate [71] and
ethylene copolymer systems [43].

Chain scission There is evidence that the rate of b-fragmentation of a mid-chain

radical is affected by the nature of the units adjacent to the mid-chain radical. Har-
ada et al. [72] studied the copolymerization of cyclohexyl acrylate and methyl meth-
acrylate at high temperatures. As in Scheme 4.16, H-abstraction only happens to
the acrylate unit. If the adjacent units are acrylate and methacrylate, two types of
fragmentation with different reaction rates are possible, as illustrated in Scheme
4.17. It was observed that the number of the unsaturated end groups per chain is
increased by increasing the methacrylate content in the polymerization. Thus it
was deduced that fragmentation greatly favors the generation of tertiary (methacry-
late chain end) radical species, a result in agreement with the high rate of fragmen-
tation observed in methacrylate macromonomer systems [73].

C C HC fast H2C CH2
+ C HC



+ HC

Scheme 4.17. Fragmentation of a mid-chain radical with
adjacent MA and MMA units (after Ref. 72).

Fragmentation after intramolecular transfer results in the formation of a long-

chain radical and trimer species or a dimer radical and an unsaturated dead chain
(Scheme 4.13). Consideration of all possible pathways and structures becomes
complex, but the resulting model requires no additional parameters from the ho-
mopolymerization back-biting/scission case and provides a good representation of
high temperature acrylate–methacrylate copolymerization [70].

Diffusion-controlled Reactions

The kinetic schemes in this chapter have been written assuming that k t is indepen-
dent of the sizes of the radicals involved in the termination reaction. This is not
4.3 Polymerization Mechanisms and Kinetics 191



kt (L/mol-s)



1.0E+04 MMA

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Fig. 4.6. Typical variation of k t with conversion for methyl
methacrylate (MMA), butyl acrylate (nBA), and dodecyl acrylate
(DA). Based upon data from [15].

strictly true, since the termination rate is limited by the rates at which the radical
ends can encounter each other. For a diffusion-controlled reaction, the apparent
rate coefficient is affected not only by pressure and temperature, but also by system
viscosity (a function of solvent, polymer concentration, and MW) and the lengths
of the two terminating radicals. This complex behavior, as well as experimental dif-
ficulties in measuring k t , has led to a large scatter in reported values, even at low
conversion [7]. Through the application of pulsed-laser experimental techniques
[15] and a critical examination of available data [18], the situation is starting to
For most commercial free radical polymerization, the errors involved by neglect-
ing the dependence of k t on radical chain length are not large. The change in k t
with conversion (increasing viscosity), however, cannot be neglected. Figure 4.6
shows the three to four orders of magnitude decrease in k t typically observed. The
shape of the curve reflects the changing nature of the rate-controlling diffusion
mechanism. The usual division is as follows:

 Low conversion: the system viscosity is still low, and the two chains diffuse to-
gether quickly. The rate of reaction is controlled by segmental diffusion, the in-
ternal reorganization of the chain that is required to bring the reactive ends to-
gether. In this region k t is of the order of 10 8 L mol1 s1 for many common
monomers (Table 4.3), with the value remaining relatively constant up to 10–
20% conversion. Solvent choice can have a significant effect on the value [18].
192 4 Free-radical Polymerization: Homogeneous

Lower values of 10 6 L mol1 s1 for dodecyl (meth)acrylate is attributed to shield-

ing of the radicals by the long-chain dodecyl ester groups; for these monomers k t
remains relatively constant up to 60% conversion [64]. Even lower k t values are
measured for termination during polymerization of sterically hindered mono-
mers such as the itaconates [74] and acrylate trimer species [44]. The variation
of low-conversion k t with polymer composition in copolymer systems can be rep-
resented by Eq. (55).
 Medium to high conversion: the large increase in system viscosity with polymer
formation leads to a change in the controlling mechanism. The rate of reaction is
controlled by how quickly the two chains find each other among the tangle of
dead polymer chains in the system. This so-called center-of-mass or translational
diffusion mechanism is complex, affected by the lengths of the reacting chains
as well as the system viscosity. The value of k t can drop by several orders of mag-
nitude in this regime.
 Very high conversion: at high conversion, the system may become so viscous that
the polymer radicals move more quickly through propagation (addition of new
monomer units) than by translation. This phenomenon, called reaction diffu-
sion, leads to a second plateau region in the k t versus conversion plot, with k t
proportional to k p ð1  x p Þ. If the glass transition temperature of the reaction
mixture exceeds the reaction temperature, the propagation reaction and apparent
initiator efficiency [75] may also become diffusion-controlled.

The overall behavior of k t with conversion is often modeled as a composite of the

various diffusional processes [Eq. (60)].

1 1 1
¼ þ ð60Þ
k t k t; SD k t; TD þ k t; RD

The subscripts SD; TD, and RD refer to segmental diffusion, translational diffu-
sion, and reaction diffusion. k t; SD can be set to the low conversion values summar-
ized in Table 4.3, and k t; RD is set proportional to propagation, with proportionality
coefficient CRD fitted to experimental data [Eq. (61)].

k t; RD ¼ CRD k p ð1  x p Þ ð61Þ

Many semi-empirical approaches have been used to model k t; TD as a function of

system viscosity, conversion, or free volume. The latter treatment relates the chain
diffusivity to the system free volume vf by Eq. (62), with parameters Ci fitted to ex-
perimental data.

k t; TD z C1 expðC2 vf Þ ð62Þ

This effect of polymer MW on system viscosity may be captured by expressing C1

as a function of Mw . It is observed experimentally that addition of a powerful
chain-transfer agent to MMA lowers polymer MW and system viscosity, thereby in-
4.4 Polymer Reaction Engineering 193








0 500 1000 1500 2000
time (s)
Fig. 4.7. Typical time–conversion plots for methyl methacrylate
and butyl acrylate batch polymerizations. The sharp increase in
rate seen for MMA, known as the gel effect, is due to the large
decrease in k t with conversion.

creasing k t , while addition of a small amount of EGDMA crosslinking agent has

the opposite effect [39]. Details and variations of this modeling approach can be
found in the literature [40, 76–79].
A good model for k t is necessary to capture the time–conversion behavior in ho-
mogeneous batch FRP systems. The large decrease in k t at intermediate conver-
sion results in an increase in radical concentration [Eq. (9)] and a corresponding
increase in R pol [Eq. (10)] that causes an upward curvature in the time–conversion
plot (Figure 4.7). This accelerated rate is accompanied by a large heat release that
can be difficult to remove from the viscous reaction system. The severity of the gel
effect is directly related to the magnitude of the decrease in k t , as seen by compar-
ing the nBA and MMA rate profiles in the figure. The decrease also leads to an in-
crease in DPninst [Eq. (13)] for systems where MW is controlled by termination.

Polymer Reaction Engineering

The design of an industrial polymerization process begins with a clear under-

standing of objectives and an appreciation of constraints. Design and operation
requirements are very different for a process manufacturing several grades of a
high-volume commodity homopolymer, and one that produces dozens of differ-
194 4 Free-radical Polymerization: Homogeneous

ent (and often new) low-volume higher-value products of varying composition

and structure. Typical product specifications for a homogeneous FRP process
may include average molecular weight/molecular weight distribution, copolymer
composition/copolymer composition distribution, degree of branching/branching
distribution, and sequence length distribution. Depending on the nature of the
product, any of these properties can simultaneously be product specifications.
However, the polymer is ultimately not sold on basic structural characteristics but
rather on end-use properties. This poses the challenge of relating structural fea-
tures to properties. Invariably the end-use properties are a product of not one but
several structural features; therefore establishing relationships is a complex task,
and the ensuing relationships are usually restricted to a narrow range of materials.
Establishing structure–property relationships remains an active area of research.
Together with an understanding of the key properties and their desired values is
a need to understand quantitatively how much variation is acceptable for each
property. In an industrial polymerization environment, there will naturally be
some degree of process variability that will translate into product variability. Know-
ing the extent to which deviations from the target value of a property affect the
manufacturer’s ability to sell that product is a critical piece of design information.
Unlike many chemical systems, off-spec polymeric material cannot be easily re-
cycled or altered by downstream unit-operations. The difficulty of characterizing
polymer structure on-line makes design of a robust easy-to-operate process espe-
cially important.
Design of a process involves several decisions such as the type of reactor used,
the flow and contacting patterns for the reagents, the choice of homogeneous ver-
sus heterogeneous process types, and so on. Heat removal and mixing issues are
two key factors that strongly influence design and operation of homogeneous FRP

Heat removal Free-radical polymerizations are highly exothermic, with adiabatic

temperature rises for bulk monomers typically @200–500 C. (Values of DHp for
common monomers are tabulated in Table 4.6.) In addition, the overall activation
energy for a radical polymerization, calculated from the activation energies of
initiation, propagation, and termination [see Eq. (10)] is on the order of 80 G 15
kJ mol1 . The sudden and dramatic increase in the heat produced in the gel-effect
region can result in loss of effective temperature control. If there is a process dis-
turbance leading to a thermal runaway condition, the heat generation rate can ex-
ceed the heat removal rate to such an extent that the reaction behaves close to adia-
batically. The resulting temperature increase can pose serious safety concerns such
as reactor overpressurization and possible explosion, requiring processes to be
designed to safely release the pressure prior to failure (rupture) of the process
Heat removal is also an important issue to consider during process scale-up. As
reactor size increases, the system dynamics become increasingly slow, and there-
fore it takes longer for desired changes (for example decreasing the reactor temper-
ature) to occur. This may in itself not be a serious problem, provided the reactor
4.4 Polymer Reaction Engineering 195

cooling system is able to maintain control, albeit at a higher than desired tempera-
ture. However, large variations in the temperature profile can translate into differ-
ences in the final product properties, particularly molecular weight distribution.
The monograph by Biesenberger and Sebastian [80] provides an excellent discus-
sion of thermal effects, including thermal runaway.

Mixing Mixing can directly affect the kinetics, molecular weight, and composition
in radical polymerizations by homogenizing local concentration gradients, but it
can also indirectly play an important role through its role in reducing thermal gra-
dients in a reactor. In small-scale experiments, most transport phenomena (heat
transfer, mixing) occur sufficiently fast for overall behavior to be dictated primarily
by reaction kinetics. However, as scale increases, kinetic and transport effects be-
come increasingly coupled. Homogeneous FRP reactions offer a particularly chal-
lenging problem because of the large increase in viscosity accompanying the
conversion from bulk monomer (@1 cp for liquids) to polymer (> 10 5 cp). The in-
crease in viscosity can greatly affect the reaction kinetics (see Section 4.3.3) as well
as the heat removal and quality of mixing in the system. Some processes are de-
signed to not require mixing. For example, PMMA can be polymerized in large
sheets. By having large surface areas available for heat transfer, adequate tempera-
ture control is achieved without the need to provide mixing during polymerization.
Other bulk polymerizations (for example, styrene) employ more than one reactor
in series, since different reactor configurations and agitators will be required as
the viscosity increases. Solution polymerizations offer low viscosity but the trend
in industry is to eliminate (or greatly reduce) the use of organic solvents. Mixing
can also be an issue at lower viscosities such as those found in LDPE systems,
where the high-temperature conditions make for very fast reactions (for example,
initiator half-life of <1 s) on the order of characteristic mixing times.

Types of Industrial Reactors

There are three major classifications of chemical processes, categorized by the

method by which the reactants are added to the reaction vessel. Varying the
contacting pattern can dramatically alter the local reaction conditions (for exam-
ple, concentrations of individual species, including monomers, initiators, chain-
transfer agents, and so on), and is therefore a potentially powerful design tool for
controlling properties such as molecular weight distribution, copolymer composi-
tion distribution, and degree of branching. Because of the ability to manipulate lo-
cal monomer concentrations, the rate of polymerization can also be controlled,
thereby providing safer and more robust operation. Batch Processes

All of the reactants are added to the reactor prior to starting the polymerization. No
material is added to or removed from the reactor during operation. When the poly-
merization is complete, the contents are discharged and the reactor prepared for
196 4 Free-radical Polymerization: Homogeneous

the next batch. Batch polymerizations are the simplest to run, but offer the least
control over the polymerization. For polymerizations with more than one mono-
mer, the relative consumption rates of the different monomers will be governed
by their respective reactivities, possibly resulting in broad copolymer composition
distributions and inhomogeneous product. Another feature of batch polymeriza-
tions is that reactant concentrations change throughout the polymerization. Molec-
ular weight distribution drift is therefore a common phenomenon and can lead to
very broad distributions in the final product. From an economic perspective, batch
polymerizations suffer from downtime between batches, although much progress
has been made in automating many of the reactant weighing, charging, and dis-
charging steps to minimize this interval. Automation has also improved reproduci-
bility of batch reactions. For operations where changes to the formulation or the
polymerization conditions are common, batch processes have the advantage of be-
ing flexible and readily adaptable to new products. Semi-batch Processes

These processes (also called semi-continuous) are similar to batch processes, ex-
cept that reactants can be added and/or products removed during the polymeriza-
tion. Usually only a portion of the total reactant charge is initially fed into the reac-
tor. The polymerization is then started, and reactants are added during reaction in
order to control a desired property (for example molecular weight distribution, co-
polymer composition distribution) or the reaction rate. Any reactant can be fed,
and it is common practice to add monomer(s), initiators, and/or chain-transfer
agents. Two of the most common applications for semi-batch operation in homo-
geneous FRP are control of copolymer composition distribution and control of re-
action rate. In a batch reaction, copolymer composition drifts according to the in-
herent reactivities of the monomers. However in semi-batch operation, drift can be
substantially reduced by maintaining a (near) constant concentration ratio of the
respective monomers in the reactor. Production of low molecular weight co- or ter-
polymers (for example coatings) is also readily done using this type of approach.
Initiator and monomers are continuously fed in the desired ratio to provide com-
position and MW control while maintaining starved conditions (low monomer con-
centration) in the reactor. This mode of operation also ensures that at any time the
monomer concentration in the reactor is low, and therefore the maximum poten-
tial hazard in the event of a thermal runaway reaction is minimized. A potential
concern with operating in a starved mode is that polymer concentrations are high,
resulting in higher rates of transfer to polymer and branching reactions. Continuous Processes

Reactants are fed, and products and unconsumed reactants are removed, continu-
ously. The process may take place in a single reactor or in a train (series) of reac-
tors in which the monomer conversion is gradually increased. Most continuous
processes are operated at ‘‘steady-state’’ conditions, meaning all reactant concentra-
tions and process conditions (temperature, pressure, and so on) are time-invariant.
This can be an enormous advantage for certain types of properties if the reactor is
4.4 Polymer Reaction Engineering 197

also well mixed (no spatial variations); because concentrations are constant, there
is no molecular weight distribution drift, and no composition distribution drift.
Thus, narrower molecular weight distributions can be produced in a well-mixed
continuous reactor for linear polymer systems than in batch systems. Branching
reactions, however, broaden the MWD in a well-mixed continuous reactor to a
greater extent than in a batch system due to the residence time distribution [80,
81]. For large-volume polymers with a limited number of variations to the polymer-
ization conditions (for example, formulation changes), continuous processes are
favored because of their low operating cost, high throughput rates, more uniform
product quality, and simplicity of operation.

Mathematical Modeling of FRP Kinetics

Mathematical modeling is a powerful methodology to improve the understanding

and operation of polymer processes. A good process model can be used to predict
the influence of operating conditions on reaction rate and polymer properties, to
guide (along with appropriate experimentation) the selection and optimization of
standard operating conditions for existing and new polymer grades, to guide pro-
cess development from lab to pilot-plant to full-scale production, to help discrimi-
nate between kinetic and physical (for example, heat and mass transfer) effects, to
perform design and safety studies, to train plant personnel, and to understand and
optimize transitions and other dynamic behavior (that is, process control).
The modeling approach and level of detail should be dictated by the application.
Whereas an empirical model linking measured inputs and outputs may be the best
solution for control of an existing industrial reactor, it would be totally inappropri-
ate for design of a new process or to choose operating conditions for a new poly-
mer grade. On the other hand, it makes little sense to develop a model that can
predict detailed polymer architecture for control purposes when the only measure
of polymer structure is a melt index value obtained from the lab two hours after
the sample was produced. While empirical modeling has its uses, the focus of
this section is the development of fundamental models based upon first-principles
descriptions of chemical and physical phenomena. Although a perfect description
of an actual process is, in the end, an unattainable goal, the attempt often leads to
valuable insights that can aid process and product development, scale-up, and opti-
mization. Techniques to model the FRP mechanisms and kinetics of Section 4.3
are presented first, followed by a discussion of issues related to reactor modeling.
Wherever possible, examples have been selected to emphasize industrial applica-
tion. Although a few historical references are included, the main focus is on advan-
ces made since the mid-1990s.
The objective of kinetic modeling is to build a description of how polymer archi-
tecture and polymerization rate depend on reaction conditions (temperature, pres-
sure) and species concentrations from a defined set of kinetic mechanisms; a dy-
namic model is required to examine how properties change as a function of time.
The mechanisms to be included in a model depend upon its end-use. For simple
198 4 Free-radical Polymerization: Homogeneous

mass and energy balances, it is only necessary to consider those that consume
monomer, initiator, and radicals – initiation, propagation, and radical–radical ter-
mination. To track polymer molecular weight, all mechanisms that include radical
transfer must also be included. Additional balances are needed to follow other mo-
lecular properties, such as the density of short- or long-chain branches, end-group
functionality, and the creation and reaction of terminal double bonds. Method of Moments

The equations in Section 4.3, while useful for examining rate of polymerization
and instantaneous chain length, are not written in an appropriate form for substi-
tuting into a generalized reactor model. In addition, they do not provide a means
of tracking higher MW averages that are strongly affected by long-chain branching
reactions. One of the challenges in modeling polymerization systems is how to re-
duce a very large number of individual species (living and dead chains with lengths
from 1 to >10 5 , often with other distributed attributes such as the number of
branch points) to a tractable solution. The classical approach to this problem is
to reduce the system of equations through definition of the principal moments
of the various distributions [82]. Construction of moment balances allows the
tracking of average polymer properties: for molecular weight this would be Mn
(number-average), Mw (weight-average), and possibly Mz , and for branched sys-
tems it is possible to track the number-average (B n ) and weight-average (B w ) num-
ber of branches per chain. Only the basics of the mathematical treatment for a ho-
mopolymerization system will be given here: more details can be found in recent
comprehensive reviews [83, 84].
Consider a free radical system that includes the basic set of mechanisms shown
in Scheme 4.3 as well as long-chain branching [Eq. (27)]. The moments for the rad-
ical (l j ) and dead (zj ) polymer distributions are defined in Eqs. (63) and (64).

lj ¼ n j ½Pn  ð63Þ

vj ¼ n j ½Dn  ð64Þ

It is also helpful to define moments for the bulk polymer (mj ), the total polymer in
the system including live radicals [Eq. (65)].

mj ¼ n j ð½Dn  þ ½Pn Þ ð65Þ

With ½Dn  g ½Pn  there is little difference in magnitude between mj and zj . Its intro-
duction, however, eliminates the moment closure problem created by the LCB
mechanism [40, 81, 85]. Many of the moments have precise physical meanings.
4.4 Polymer Reaction Engineering 199

The zeroth live moment, l 0 , is the concentration of polymer radicals in the system
(denoted by [Ptot ] in Section 4.3), and the first live moment, l1 , is the concentration
of monomer units contained in all growing radicals. Similarly, m 0 is the concentra-
tion of all polymer chains in the system, and m1 is the concentration of monomer
units bound in all polymer chains. These moment definitions collapse the infinite
set of equations for polymeric species into a manageable subset used to calculate
MW averages, where m w is the molecular weight of the monomeric repeat unit
[Eqs. (66)].

m1 m2 m3
Mn ¼ m w ; Mw ¼ m w ; Mz ¼ m w ð66Þ
m0 m1 m2

For this example, equations for the kinetic expressions for m 0 ; m1, and m 2 will be
developed for the calculation of Mn and Mw .
The first step is to formulate balances for live radicals, dead chains, and total
chains of length n, accounting for all of the consumption and generation terms
from the kinetic mechanisms [Eqs. (67)–(69)].
( )
y X
RPn ¼ 2f kd ½I þ ktrmon ½M sol
½Pj  þ ktr ½S ½Pj  dðn  1Þ
j¼1 j¼1

þ k p ½Mð½Pn1   ½Pn Þ
( )
 ktrmon ½M þ ktrsol ½S þ ðk td þ k tc Þ ½Pj  ½Pn 

þ k tr n½Dn  ½Pj   k tr ½Pn  j½Dj  ð67Þ
j¼1 j¼1
( )
1 X n1
R Dn ¼ ktrmon ½M þ ktrsol ½S þ k td ½Pj  ½Pn  þ k tc ½Pj ½Pnj 
2 j¼1

þ k tr ½Pn  j½Dj   k tr n½Dn  ½Pj  ð68Þ
j¼1 j¼1
( )
y X
RPn þDn ¼ 2f kd ½I þ ktrmon ½M sol
½Pj  þ ktr ½S ½Pj  dðn  1Þ
j¼1 j¼1

þ k p ½Mð½Pn1   ½Pn Þ
1 X n1
 k tc ½Pj  ½Pn  þ k tc ½Pj ½Pnj  ð69Þ
2 j¼1

The origin of the various terms in these balances should be evident by looking
at the mechanisms of Scheme 4.3 and Eq. (27). The Kronecker delta function
[dðxÞ ¼ 1 if x ¼ 0 and sðxÞ ¼ 0 if x 0 0 ] accounts for the generation of new poly-
meric radicals (P1 ), and the terms for transfer to polymer account for the probabil-
200 4 Free-radical Polymerization: Homogeneous

ity that transfer to a certain chain Dn is proportional to chain length n. The expres-
sion for termination by combination accounts for the possibility of creating Dn
from any combination of two smaller radical fragments whose lengths sum to n.
The next step in the procedure is to substitute these species balances into the
moment definitions in Eqs. (63)–(65). The use of generating functions [82, 86, 87]
eliminates the tedium (and possible errors) of performing the required series sum-
mations, and leads to Eqs. (70)–(78) for the moments.

Live moments:

Rl 0 ¼ 2 f kd ½I  ðk td þ k tc Þl02 ð70Þ

Rl1 ¼ 2f kd ½I þ ktrmon ½Ml 0 þ ktrsol ½Sl 0 þ k p ½Ml 0
 fktrmon ½M þ ktrsol ½S þ ðk td þ k tc Þl 0 gl1 þ k tr ðl 0 v2  l1 v1 Þ ð71Þ
Rl2 ¼ 2f kd ½I þ ktrmon ½Ml 0 þ ktrsol ½Sl 0 þ k p ½Mðl 0 þ 2l1 Þ
 fktrmon ½M þ ktrsol ½S þ ðk td þ k tc Þl 0 gl2 þ k tr ðl 0 v3  l2 v1 Þ ð72Þ

Dead moments:

R v0 ¼ ktrmon ½Ml 0 þ ktrsol ½Sl 0 þ k td l02 þ k tc l02 ð73Þ
R v1 ¼ fktrmon ½M þ ktrsol ½S þ ðk td þ k tc Þl 0 gl1  k tr ðl 0 v2  l1 v1 Þ ð74Þ
R v2 ¼ fktrmon ½M þ ktrsol ½S þ ðk td þ k tc Þl 0 gl2 þ k tc l12  k tr ðl 0 v3  l2 v1 Þ ð75Þ

Bulk moments:

Rm 0 ¼ 2f kd ½I þ ktrmon ½Ml 0 þ ktrsol ½Sl 0  k tc l02 ð76Þ

Rm1 ¼ 2f kd ½I þ ktrmon ½Ml 0 þ ktrsol ½Sl 0 þ k p ½Ml 0 ð77Þ

Rm 2 ¼ 2f kd ½I þ ktrmon ½Ml 0 þ ktrsol ½Sl 0 þ k p ½Mðl 0 þ 2l1 Þ þ k tc l12 ð78Þ

The set of moment expressions to be considered consists of either the live and dead
moments [Eqs. (70)–(75)] or the live and bulk moments [Eqs. (70), (71) and (76)–
(78)], substituting m1 and m 2 for z1 and z2 in Eq. (71). Choice of the former, while it
is common practice in the literature, suffers from the problem that the equations
for l2 and z2 depend on z3 . The Hulburt and Kutz [88] method is often used to ap-
proximate z3 , assuming that the molecular weight distribution can be represented
by a truncated series of Laguerre polynomials by using a gamma distribution
weighting function [Eq. (79)].

v3 ¼ ð2v0 v2  v12 Þ ð79Þ
v0 v 1
4.4 Polymer Reaction Engineering 201

Using the bulk moments not only eliminates the need for this approximation, but
also reduces the set of equations by one, since Eq. (72) is not required to solve for
Mw . An additional balance [Eq. (80)] can be added to either set of equations to track
the concentration of LCB formed by the transfer to polymer mechanism,

RLCB ¼ k tr l 0 m1 ð80Þ

Thus, a set of six equations [Eqs. (70), (71), (76)–(78), and (80)] can be used to col-
lapse the molecular weight distribution into its principle averages to calculate Mn ,
Mw , and LCB density.
The set of moment equations developed here can be extended to include addi-
tional complex mechanisms, such as reactivity of terminal double bonds [Eq. (30),
Scheme 4.10] [81], crosslinking (Scheme 4.11) [40], and chain b-scission following
intermolecular H-abstraction [89]. The methodology can also be extended to co-
polymerization systems, either by defining copolymer-averaged rate coefficients as
in Eqs. (50) and (54) [6, 85–87], or by defining additional moment quantities (for
examples, see Refs. 40, 81, 85). Furthermore, since it is easy to implement as part
of larger-scale reactor modeling, it is the standard methodology used in process
simulation packages (see, for example, Refs. 90, 91). For discussion regarding the
final step of model development, substitution of the kinetic expressions for the mo-
ments into reactor balances, see Section 4.4.3. Modeling of Distributions

The major limitation of models based on the method of moments is that they only
track average quantities. While adequate for most situations, more detail may be
needed if the objective of the study is to improve our knowledge of the kinetics
– for example, to examine the combined effect of chain-scission and long-chain
branching on polymer architecture, or to incorporate chain-length dependent ter-
mination kinetics into the mechanistic scheme. In such cases, the kinetic and
modeling assumptions can be tested more rigorously through a detailed compari-
son with full molecular weight distributions (MWDs) measured experimentally.
Recent advances in modeling tools now make it possible to simulate the complete
MWD, as well as how a second distributed quantity (for example, LCB) varies with
chain length.
The modeling of complete MWDs has long been possible for linear polymer sys-
tems, that is, those without any branching. In the absence of long-chain branching
and employing the QSSA, a recursive relationship can be derived for Pn and Dn
from Eqs. (67) and (68), leading to Eqs. (81) and (82) for the weight-fraction distri-
bution of polymer formed at any instant in time [92].
wninst ¼ ðt þ 0:5ðn  1Þbðt þ bÞÞðt þ bÞn ð81Þ

k td ½l 0  þ ktrmon ½M þ ktrsol ½S k tc ½l 0 

t¼ ; b¼ ð82Þ
k p ½M k p ½M
202 4 Free-radical Polymerization: Homogeneous

Assuming the formation of long-chain polymer (t þ b f 1), the values for instanta-
neous DPn and DPw are given by Eqs. (83).

1 2t þ 3b
DPninst ¼ ; DPwinst ¼ ð83Þ
t þ 0:5b ðt þ bÞ 2

The instantaneous PDI is 2 for a system where there is no termination by combi-

nation (b ¼ 0), and 1.5 if termination by combination is the only chain-ending
mechanism (t ¼ 0). These expressions can be integrated to follow the evolution of
the MWD with time or conversion and compare against experimental data, as was
done by Balke and Hamielec [92] for batch isothermal FRP of MMA.
Unfortunately, the methodology cannot be easily extended to branched systems,
due to the interaction of the polymer radical and dead polymer chain distributions
through reaction (for example, H-abstraction, terminal double bond polymeriza-
tion, crosslinking). The methods for modeling MWDs with branching can be di-
vided into three main groups. The first, utilizing Monte Carlo techniques, has
been greatly advanced through the efforts of Tobita. Assuming a given set of
mechanisms, the probabilities of connections between primary polymer molecules
(the linear chain that would exist if all of its branch points were severed) is calcu-
lated, and the resulting MWD solved using Monte Carlo techniques [93].
A second group of models is based upon the ‘‘numerical fractionation’’ concept
developed by Teymour and Campbell [94]. This seminal work identifies a succes-
sion of branched polymer generations based on the degree of complexity of
their architecture, tracking the population of chains in each generation using
the method of moments. The complete MWD is approximated by combining the
MWDs for individual generations which themselves are reconstructed from the
leading moments assuming a distributional form. Numerical fractionation was
specifically developed to examine the problem of gel formation in polymer sys-
tems. Thus, the generations were defined to follow the geometric progression in
chain length caused by connection of two molecules in the same generation: while
chains from the zeroth generation progress to the first generation by participating
in a branching event, a chain from the first generation can only progress to the sec-
ond by joining (through crosslinking, terminal double bond polymerization, or ter-
mination by combination) with another molecule from the same generation [94]. It
has been shown that this classification scheme leads, in certain cases, to errors in
the shape of the overall MWD: through comparison with distributions calculated
by rigorous numerical solution, Butté et al. [95] have shown that the definition of
generations proposed by Teymour and Campbell can create an artificial high MW
shoulder due to the accumulation of chains with a wide distribution of the number
of branches (and thus high polydispersity) in the first branched generation. The
authors conclude that a more accurate approximation is obtained by classifying
the chains according to their number of branches. Both Monte Carlo [96] and a
modified numerical fractionation technique [95] can also be used to calculate the
LCB number as a function of chain length, an important quantity often presented
4.4 Polymer Reaction Engineering 203

The commercial software Predici2 package uses yet another numerical tech-
nique, calculating MWDs using a discrete Galerkin technique with variable grid
and variable order [97]. In 2000, the package was extended to follow branch point
concentrations as a function of chain length through the introduction of balance
equations [98]. The possibility of performing these tasks – calculation of complete
MWD as well as LCB distribution – in a commercial software package is especially
noteworthy because it makes it possible for a wider range of practitioners to per-
form detailed kinetic modeling. These new modeling capabilities, in combination
with improved characterization techniques, will hasten progress to a better under-
standing and representation of complex polymer architecture [46, 99].

Reactor Modeling

The kinetic expressions of Section 4.4.2 are substituted into overall material and
energy balances to construct a model to represent an FRP process. For a well-
mixed reactor system, Eqs. (84)–(89) comprise the general system of equations for

¼ ðqin ½Min Þ  ðqout ½MÞ  VR pol ð84Þ

¼ ðqin ½Iin Þ  ðqout ½IÞ  VR d ð85Þ

¼ ðqin ½Sin Þ  ðqout ½SÞ  VRtrsol ð86Þ

dðV½l i Þ
¼ ðqin ½li; in Þ  ðqout ½l i Þ þ VRl i ð87Þ

dðV½m i Þ
¼ ðqin ½m i; in Þ  ðqout ½m i Þ þ VRmi ð88Þ

dðVrc p ðT  Tref ÞÞ
¼ ðqin ðrc p Þin ðTin  Tref ÞÞ  ðqout ðrc p Þout ðT  Tref ÞÞ

þ ðDHp ÞVR pol  Q ð89Þ

In these balances, which must be accompanied by appropriate initial conditions

and a volume balance, V is the reactor volume, qin and qout are the inlet and outlet
volumetric flow rates, r and c p are the density and heat capacity of the mixture, and
Q is the rate of heat removal from the reactor system. The various source terms
have been presented earlier: Eq. (10) for the monomer and energy balance, Eq. (2)
for the initiator balance, Rtrsol ¼ ktrsol ½Sl 0 in the solvent balance, Eqs. (70) and (71)
for the live moment balances, and Eqs. (76)–(78) for the bulk moment balances.
204 4 Free-radical Polymerization: Homogeneous Polymerization
Inflow and outflow terms are set to zero, but the change in volume with conver-
sion must be considered for the constant-mass system [Eq. (90)].

dV V dr
¼ ð90Þ
dt r dt

Volume contraction due to the difference in polymer and monomer densities can
be as high as 20% for bulk polymerization, and should not be neglected when solv-
ing the material balances. Solution of the initiator balance and substitution into
the monomer balance lead to a differential equation for conversion, Eq. (91), where
the volume contraction factor is defined as e ¼ ðrfinal  r0 Þ=rfinal where rfinal is the
system density at 100% conversion of monomer to polymer and r0 is the initial
density of the system.
! !1=2  
dx p kp 2f kd ½I0 kd t
¼ 1=2
ð1  x p Þ exp  ð91Þ
dt kt 1  ex p 2 Continuous Polymerization

Equations (84)–(89) can be solved to provide a description of the dynamic behavior
of a continuous well-stirred system. Analytical solutions may be derived for the
steady-state case by setting all derivative terms to zero. Assuming no inflow of rad-
icals, Eq. (92) results for l 0.

ðqout ½l 0 Þ ¼ Vð2f kd ½I  k t ½l 0  2 Þ ð92Þ

For most cases, the radical lifetime is much shorter than the average residence
time. Thus outflow of radicals can be neglected, resulting in the familiar expres-
sion, Eq. (93), for l 0.

2f kd ½I 1=2
½l 0  ¼ ð93Þ

[I] can be calculated from the steady-state solution of Eq. (83), namely Eq. (94),
where y ¼ V=qout is the reactor residence time and e ¼ ðrout  rin Þ=rout is the frac-
tional density change between inlet and outlet streams.

qin ½Iin  ½Iin 

½I ¼ ¼ ð94Þ
qout þ Vkd ð1 þ ykd Þð1  eÞ

Substitution of these expressions into the monomer balance leads to a nonlinear

relationship between conversion and y [Eq. (95)].
4.4 Polymer Reaction Engineering 205
! 1=2
kp 2f kd ½Iin 
x p ¼ yk p l 0 ¼ y 1=2
kt ð1  eÞð1 þ ykd Þ

Assuming no inflow of polymer into the reactor and the long-chain hypothesis, the
steady-state values for DPn and DPw in the absence of LCB (k tr ¼ 0) are given by
Eq. (83). LCB and reaction with terminal double bonds broaden the MWD signifi-
cantly [81]. Complex Flowsheets

These are often constructed to represent systems with nonideal mixing and fast re-
action. A classic example is the high-pressure high-temperature free radical pro-
duction of ethylene copolymers, generally conducted in a homogeneous phase con-
sisting largely of supercritical ethylene monomer. These conditions make for very
fast reactions (for example, initiator half-life of <1 s), promote numerous side re-
actions (long-chain branching, short-chain branching, and chain scission), and in-
troduce the potential of thermal runaway. Recent models of these systems combine
a detailed description of polymerization kinetics with a complex flowsheet of con-
tinuous well-mixed tanks in series with recycle to represent mixing in a multizone,
multifeed autoclave reactor [100–102]. Models for multifeed tubular systems also
include heat transfer and pressure drop along the length of the system [103]. The
general strategy is to ‘‘tune’’ the model (based upon a set of proposed kinetic
mechanisms captured by the method of moments) by fitting kinetic coefficients
and mixing and heat transfer parameters to a set of industrial data, and then to
use the model to interpret and optimize industrial operating conditions. Advances
in computing power have allowed the complexity of these models to increase. Computational Fluid Dynamics (CFD)

CFD simulation is emerging as an alternative and more fundamental approach to
examining polymerization systems with complex mixing and reaction. Once again,
much of the work is focused on high-pressure ethylene polymerization systems.
A major challenge is incorporating both macromixing (turbulent diffusion and
convection) as well as micromixing (molecular diffusion) into the representation
[104]. The first efforts in this area [104, 105] concern themselves with the predic-
tion of temperature and conversion profiles in the reactors; to simplify the calcula-
tional load they consider only initiation, propagation, and termination reactions.
More recently, Kolhapure and Fox [106] incorporated a more complete kinetic
scheme to allow the prediction of polymer MW, polydispersity, and average branch-
ing number. These CFD studies can point the way to improved reactor design and
operation; for example, by examining the importance of initiator distribution at the
injection point, and defining conditions for stable reactor operation [106]. An arti-
cle in 2000 discussed the implementation of CFD calculations within a process
simulation package [107]. Although not yet applied to polymerization systems,
this advancement shows enormous promise.
206 4 Free-radical Polymerization: Homogeneous Model-based Control

Model-based control of polymerization systems has also garnered its share of atten-
tion. The goal of these efforts is the development of robust strategies to guide and
control the manufacture of polymer safely and reproducibly in the face of unmea-
sured disturbances and frequent product grade transitions. The main challenge in
controlling polymerization systems is the lack of on-line measurements of polymer
structure. A review by Congalidis and Richards [108] provides a good summary of
literature focusing on this difficult issue; for further discussion, see Chapter 12 of
this book. In most cases, the implementation of detailed fundamental models is
not warranted for control application. However, simple models can often be com-
bined with limited on-line measurements (for example, in Refs. 109–112) to im-
prove control performance. Fundamental models can also be used to test empirical
models developed for control purposes [113].


Polymer reaction engineering issues can pose major challenges for industrial-scale
polymer synthesis. As reactor size increases, transport phenomena such as heat
and mass transfer become more difficult, and large-scale homogeneous FRP pro-
cesses are often limited by transport effects. In many cases, the final reaction rate,
molecular weight, and copolymer composition are determined by the coupled ef-
fects of reaction kinetics and transport phenomena. Mathematical modeling, in
conjunction with a strong experimental program, is a powerful means to improve
our mechanistic and process understanding. The use of empirical models has util-
ity in the control of polymerization reactors, and can be an important contributor
to product quality and process robustness. Continuing development of new mea-
surement techniques, together with an ability to relate measurements to functional
properties, will be a critical area of future research.
While many current free-radical polymerization processes have been in indus-
trial use for years, the next several years may see the emergence of new industrial
technologies. Promising technologies for homogeneous FRP include living radical
polymerization chemistries and polymerization in supercritical carbon dioxide.
The adoption of new technology requires identifying an application or product for
which the new technology is clearly advantaged, and successfully overcoming the
numerous scaleup challenges in converting the process to industrial scale. These
processes will require a detailed understanding of polymerization kinetics com-
bined with fundamental reaction engineering principles.


Ai pre-exponential factor for mechanism-i

CtrXXX ratio of chain-transfer rate coefficient (k trXXX , XXX ¼ mon, pol, sol) to
propagation [Eq. (19)]
Notation 207

cp heat capacity [kJ kg1 K1 ]

Dn¼ dead polymer chain of length n with terminal unsaturation
Dn; b dead polymer chain of length n with b branch points
DPn number-average degree of polymerization
DPw weight-average degree of polymerization
Ei activation energy for mechanism-i [kJ mol1 ]
mole fraction of monomer-i contained in polymer chains being formed
f initiator efficiency [Eq. (1)]
fi mole fraction of monomer-i in the monomer mixture
fri mole fraction of polymer radicals ending in monomer unit-i
frij mole fraction of polymer radicals ending in monomer unit-j with pen-
ultimate unit-i
[I] initiator concentration [mol L1 ]
K eq equilibrium propagation–depropagation constant
kbb rate coefficient for intramolecular H-abstraction [s1 ]
kd rate coefficient for initiation decomposition [s1 ]
kdep rate coefficient for depropagation [s1 ]
ki rate coefficient for primary-radical initiation [L mol1 s1 ]
k inhib rate coefficient for inhibition [L mol1 s1 ]
kp rate coefficient for propagation [L mol1 s1 ]
k pij rate coefficient for addition of monomer-j to radical-i [L mol1 s1 ]
k pijk rate coefficient for addition of monomer-k to penultimate radical-ij
[L mol1 s1 ]
kp copolymer-averaged propagation rate coefficient [L mol1 s1 ]
kp effective propagation rate coefficient, corrected for depropagation
[L mol1 s1 ]
kp rate coefficient for radical addition to a terminally-unsaturated chain
[L mol1 s1 ]
addition rate of monomer to mid-chain radical [L mol1 s1 ]
kt rate coefficient for termination [combination þ disproportionation)
[L mol1 s1 ]
kt copolymer-averaged termination rate coefficient [L mol1 s1 ]
k tc rate coefficient for termination by combination [L mol1 s1 ]
k td rate coefficient for termination by disproportionation [L mol1 s1 ]
k therm rate coefficient for thermal initiation of styrene [L 2 mol2 s1 ]
k trXXX rate coefficient for transfer to species XXX (mon, pol, sol) [L mol1 s1 ]
kb rate coefficient for radical b-scission [s1 ]
[M] monomer concentration [mol L1 ]
[M]eq equilibrium monomer concentration [mol L1 ]
Mn number-average molecular weight [g mol1 ]
Mw weight-average molecular weight [g mol1 ]
mw molecular weight of the reaction mixture [g mol1 ]
Ni number-average length of monomer-i sequences
NðMi ; nj Þ fraction of monomer-i sequences that are exactly nj units in length
Pij probability of unit-j following unit-i in a polymer chain
208 4 Free-radical Polymerization: Homogeneous

Pn polymer radical of length n

Pn; b polymer radical of length n with b branch points
[Ptot ] total concentration of radicals [mol L1 ]
[Ptot ] total concentration of polymer radicals of type-i [mol L1 ]
Qn mid-chain polymer radical of length-n (formed by intramolecular or in-
termolecular H-abstraction)
qin reactor inlet volumetric flow [L s1 ]
qout reactor outlet volumetric flow [L s1 ]
Rd rate of initiator disappearance [mol L1 s1 ]
R dep rate of depropagation [mol L1 s1 ]
R inhib rate of inhibition [mol L1 s1 ]
R init rate of radical generation from initiator [mol L1 s1 ]
RLCB rate of long-chain branch formation [mol L1 s1 ]
Rp rate of propagation [mol L1 s1 ]
Rp addition rate of terminally unsaturated polymer chains mol L1 s1 ]
R pol total rate of monomer consumption [mol L1 s1 ]
R term rate of radical–radical termination [mol L1 s1 ]
R therm rate of thermal initiation of styrene [mol L1 s1 ]
R trXXX rate of transfer to species XXX (mon, pol, sol) [mol L1 s1 ]
Rl i rate of change of moment l i [mol L1 s1 ]
ri monomer reactivity ratio for radical-i (for example, r1 ¼ k p11 =k p12 )
[S] solvent (transfer agent) concentration [mol L1 ]
si radical reactivity ratio for monomer-i in the penultimate copolymeriza-
tion model (for example, s1 ¼ k p211 =k p111 )
T temperature [K]
Tc ceiling temperature of monomer [ C]
t 1=2 initiator half-life [s1 ]
V reactor volume [L]
wn weight fraction of chains of length-n
xp fractional monomer conversion to polymer
[Z] inhibitor concentration [mol L1 ]


b ratio of chain termination (combination) to propagation [Eq. (82)]

DGp free energy change of polymerization [kJ mol1 ]
DHp heat of polymerization [kJ mol1 ]
DSp entropy of polymerization [J mol1 K1 ]
DVi activation volume for mechanism-i [cm 3 mol1 ]
d fraction of termination by disproportionation
e volume contraction factor
zk kth moment of the dead polymer chains [mol L1 ]
y average residence time [s]
lk kth moment of the polymer radicals [mol L1 ]
mk kth moment of the total (dead þ radical) polymer chains [mol L1 ]
References 209

n kinetic chain length of polymer radicals

r density [kg L1 ]
t ratio of chain termination (disproportionation) and transfer to propaga-
tion [Eq. (82)]
u average number of monomer units on a living chain
F copolymerization cross-termination factor


AIBN 2,2 0 -azobisisobutyronitrile

BMA butyl methacrylate
CFD computational fluid dynamics
CTA chain-transfer agent
DMA dodecyl methacrylate
EGDMA ethylene glycol dimethacrylate
FRP free radical polymerization
LCB long-chain branching
LCH long-chain hypothesis
LDPE low-density polyethylene
MA methyl acrylate
MMA methyl methacrylate
MW molecular weight
MWD molecular weight distribution
nBA Butyl acrylate
PDI polydispersity index (Mw/Mn )
PLP pulsed-laser polymerization
PMMA poly(methyl methacrylate)
QSSA quasi-steady-state assumption
SCB short-chain branching
SEC size exclusion chromatography
Sty styrene


1 M. Buback, S. Klingbeil, J. 4 M. Asteasuain, A. Brandolin, C.

Sandmann, M.-B. Sderra, H.-P. Sarmoria, Polymer 2004, 45, 321–335.
Vögele, H. Wackerbarth, L. 5 A. A. Gridnev, S. D. Ittel,
Wittkowski, Z. Phys. Chem. 1999, Macromolecules 1996, 29, 5864–5874.
210, 199–221. 6 H. Tobita, A. E. Hamielec, Polymer
2 M. Buback, J. Sandmann, Z. Phys. 1991, 32, 2641–2647.
Chem. 2000, 214, 583–607. 7 J. Brandrup, E. H. Immergut, E. A.
3 G. Moad, D. H. Solomon, The Grulke (Eds.), Polymer Handbook, 4th
Chemistry of Free Radical Edition, John Wiley & Sons, New York,
Polymerization, Elsevier Science, New 1999.
York, 1995. 8 M. Buback, L. H. Garcia-Rubio, R. G.
210 4 Free-radical Polymerization: Homogeneous

Gilbert, D. H. Napper, J. Guillot, 21 P. V. T. Raghuram, U. S. Nandi, J.

A. E. Hamielec, D. Hill, K. F. Polym. Sci.: Part A-1 1969, 7, 2379–
O’Driscoll, O. F. Olaj, J. Shen, D. 2385.
Solomon, G. Moad, M. Stickler, M. 22 G. A. Mortimer, J. Polym. Sci.: Part
Tirrell, M. A. Winnik, J. Polym. Sci.: A-1 1970, 8, 1535–1542.
Part C: Polym. Lett. 1988, 26, 293–297. 23 M. Stickler, G. Meyerhoff,
9 M. Buback, R. G. Gilbert, G. T. Makromol. Chem. 1978, 179, 2729–
Russell, D. J. T. Hill, G. Moad, K. F. 2745.
O’Driscoll, J. Shen, M. A. Winnik, 24 S. Maeder, R. G. Gilbert,
J. Polym. Sci., Polym. Chem. Ed. 1992, Macromolecules 1998, 31, 4410–4418.
30, 851–864. 25 M. Buback, C.-R. Choe, E.-U.
10 O. F. Olaj, I. Bitai, F. Hinkelmann, Franck, Makromol. Chem. 1984, 185,
Makromol. Chem. 1987, 188, 1689– 1685–1697.
1702. 26 J. D. Campbell, F. Teymour, M.
11 M. Buback, R. G. Gilbert, R. A. Morbidelli, Macromolecules 2003, 36,
Hutchinson, B. Klumperman, F.-D. 5491–5501.
Kuchta, B. G. Manders, K. F. 27 A. W. Hui, A. E. Hamielec, J. Appl.
O’Driscoll, G. T. Russell, J. Polym. Sci. 1972, 16, 749–769.
Schweer, Macromol. Chem. Phys. 28 R. A. Hutchinson, D. A. Paquet,
1995, 196, 3267–3280. Jr., S. Beuermann, J. H. McMinn,
12 S. Beuermann, M. Buback, T. P. Ind. Eng. Chem. Res. 1998, 37, 3567–
Davis, R. G. Gilbert, R. A. 3574.
Hutchinson, O. F. Olaj, G. T. 29 M. C. Grady, W. J. Simonsick, Jr.,
Russell, J. Schweer, A. M. van R. A. Hutchinson, Macromol. Symp.
Herk, Macromol. Chem. Phys. 1997, 2002, 182, 149–168.
198, 1545–1560. 30 P. Ehrlich, G. A. Mortimer, Adv.
13 S. Beuermann, M. Buback, T. P. Polym. Sci. 1970, 7, 386–448.
Davis, R. G. Gilbert, R. A. 31 P. Feucht, B. Tilger, G. Luft, Chem.
Hutchinson, A. Kajiwara, B. Eng. Sci. 1985, 40, 1935–1942.
Klumperman, G. T. Russell, 32 L. I. Kulin, N. L. Meijerink, P.
Macromol. Chem. Phys. 2000, 201, Starck, Pure Appl. Chem. 1988, 60,
1355–1364. 1403–1415.
14 S. Beuermann, M. Buback, T. P. 33 W. W. Graessley, R. D. Hartung,
Davis, R. G. Gilbert, R. A. W. C. Uy, J. Polym. Sci.: Part A-2 1969,
Hutchinson, A. Kajiwara, M. 7, 1919–1935.
Kamachi, I. Lacı́k, G. T. Russell, 34 D. Britton, F. Heatley, P. A. Lovell,
Macromol. Chem. Phys. 2003, 204, Macromolecules 1998, 31, 2828–2837.
1338–1350. 35 S. Grcev, P. Schoenmakers, P.
15 S. Beuermann, M. Buback, Prog. Iedema, Polymer 2004, 45, 39–48.
Polym. Sci. 2002, 27, 191–254. 36 N. M. Ahmad, F. Heatley, P. A.
16 I. Lacı́k, S. Beuermann, M. Buback, Lovell, Macromolecules 1998, 31,
Macromolecules 2003, 36, 9355–9363. 2822–2827.
17 H. Fischer, L. Radom, Angew. Chem. 37 C. Plessis, G. Arzamendi, J. R. Leiza,
Int. Ed. 2001, 40, 1340–1371. H. A. S. Schoonbrood, D. Charmot,
18 M. Buback, M. Egorov, R. G. J. M. Asua, Ind. Eng. Chem. Res. 2001,
Gilbert, V. Kaminsky, O. F. Olaj, 40, 3883–3894.
G. T. Russell, P. Vana, G. Zifferer, 38 P. Tackx, J. C. J. F. Tacx, Polymer
Macromol. Chem. Phys. 2002, 203, 1998, 39, 3109–3113.
2570–2582. 39 W.-H. Li, A. E. Hamielec, C. M.
19 M. D. Zammit, T. P. Davis, D. M. Crowe, Polymer 1989, 30, 1518–1523.
Haddleton, K. G. Suddaby, 40 R. A. Hutchinson, Polym. React. Eng.
Macromolecules 1997, 30, 1915–1920. 1993, 1, 521–577.
20 G. A. Mortimer, J. Polym. Sci.: Part 41 T. Usami, S. Takayama, Macromolec-
A-1 1966, 4, 881–900. ules 1984, 17, 1756–1761.
References 211

42 J. C. Randall, JMS – Rev. Macromol. 63 M. Buback, C. Kowollik,

Chem. Phys. 1989, C29, 201–317. Macromolecules 1999, 32, 1445–1452.
43 E. F. McCord, W. H. Shaw, Jr., R. A. 64 M. Buback, A. Feldermann, Aust. J.
Hutchinson, Macromolecules 1997, Chem. 2002, 55, 475–481.
30, 246–256. 65 H. A. S. Schoonbrood, S. C. J.
44 T. Hirano, B. Yamada, Polymer 2003, Pierik, B. van den Reijen, J. P. A.
44, 347–354. Heuts, A. L. German, Macromolecules
45 J. Chiefari, J. Jeffery, R. T. A. 1996, 29, 6717–6723.
Mayadunne, G. Moad, E. Rizzardo, 66 P. Wittmer, ACS Adv Chem Ser. 1971,
S. N. Thang, Macromolecules 1999, 32, 99, 140–174.
7700–7702. 67 N. T. McManus, A. Penlidis, M. A.
46 R. A. Hutchinson, Macromol. Theory Dubé, Polymer 2002, 43, 1607–1614.
Simul. 2001, 10, 144–157. 68 G. G. Lowry, J. Polym. Sci. 1960, 42,
47 A. F. N. Peck. MSc Thesis, Queen’s 463–477.
University, Kingston, ON, 2003. 69 C. Badeen, M. A. Dubé, Polym. React.
48 P. W. Tidwell, G. A. Mortimer, J. Eng. 2003, 11, 53–77.
Polym. Sci.: Part A 1965, 3, 369–387. 70 D. Li, MSc Thesis, Queen’s University,
49 A. M. van Herk, T. Dröge, Macromol. Kingston, ON, 2003.
Theory Simul. 1997, 6, 1263–1276. 71 C. Plessis, G. Arzamendi, J. R. Leiza,
50 V. E. Meyer, G. G. Lowry, J. Polym. H. A. S. Schoonbrood, C. Charmot,
Sci.: Part A 1965, 3, 2843–2851. J. M. Asua, Macromolecules 2001, 34,
51 T. Fukuda, K. Kubo, Y.-D. Ma, Prog. 5147–5157.
Polym. Sci. 1992, 17, 875–916. 72 T. Harada, P. B. Zetterlund, B.
52 M. Buback, A. Feldermann, C. Yamada, Polym. Prepr. 2002, 43(2),
Barner-Kowollik, I. Lacı́k, 160–161.
Macromolecules 2001, 34, 5439–5448. 73 C. L. Moad, G. Moad, E. Rizzardo,
53 M. Buback, L. Wittkowski, S. H. Thang, Macromolecules 1996, 29,
Macromol. Chem. Phys. 2000, 201, 7717–7726.
419–426. 74 P. Vana, L. H. Yee, C. Barner-
54 M. Buback, L. Wittkowski, S. A. Kowollik, J. P. A. Heuts, T. P. Davis,
Lehmann, F.-O. Mähling, Macromol. Macromolecules 2002, 35, 1651–1657.
Chem. Phys. 1999, 200, 1935–1941. 75 G. T. Russell, D. H. Napper, R. G.
55 L. K. Kostanski, A. E. Hamielec, Gilbert, Macromolecules 1988, 21,
Polymer 1992, 33, 3706–3710. 2141–2148.
56 M. Hakim, V. Verhoeven, N. T. 76 S. K. Soh, D. C. Sundberg, J. Polym.
McManus, M. A. Dubé, A. Penlidis, Sci.: Polym. Chem. Ed. 1982, 20, 1345–
J. Appl. Polym. Sci. 2000, 77, 602–609. 1371.
57 M. Buback, H. Dietzsch, Macromol. 77 B. M. Louie, G. M. Carratt, D. S.
Chem. Phys. 2001, 202, 1173–1181. Soong, J. Appl. Polym. Sci. 1985, 30,
58 A. D. Jenkins, K. F. O’Driscoll, J. 3985–4012.
Polym. Sci.: Part A: Polym. Chem. 1995, 78 D. Achilias, C. Kiparissides,
33, 1189–1190. Macromolecules 1992, 25, 3739–3750.
59 M. L. Coote, L. P. M. Johnston, T. P. 79 N. Tefera, G. Weickert, K. R.
Davis, Macromolecules 1997, 30, 8191– Westerterp, J. Appl. Polym. Sci. 1997,
8204. 63, 1649–1660.
60 T. P. Davis, K. F. O’Driscoll, M. C. 80 J. A. Biesenberger, D. H. Sebastian,
Piton, M. A. Winnik, Polym. Int. Principles of Polymerization Engineering,
1991, 24, 65–70. John Wiley & Sons, New York, 1983.
61 Y.-D. Ma, P.-S. Kim, K. Kubo, T. 81 A. Baltsas, D. S. Achilias, C.
Fukuda, Polymer 1994, 35, 1375–1381. Kiparissides, Macromol. Theory Simul.
62 R. A. Hutchinson, J. H. McMinn, 1996, 5, 477–497.
D. A. Paquet, Jr., S. Beuermann, C. 82 W. H. Ray, J. Macromol. Sci. – Rev.
Jackson, Ind. Eng. Chem. Res. 1997, Macromol. Chem. 1972, C8, 1–56.
36, 1103–1113. 83 D. S. Achilias, C. Kiparissides, J.
212 4 Free-radical Polymerization: Homogeneous

Macromol. Sci. – Rev. Macromol. Chem. 99 M. Busch, Macromol. Theory Simul.

Phys. 1992, C32, 183–234. 2001, 10, 408–429.
84 M. A. Dubé, J. B. P. Soares, A. 100 W.-M. Chan, P. E. Gloor, A. E.
Penlidis, A. E. Hamielec, Ind. Eng. Hamielec, AIChE J. 1993, 39, 111–
Chem. Res. 1997, 36, 966–1015. 126.
85 D. J. Arriola, PhD Thesis, University 101 H. Nordhus, O. Moen, P. Singstad,
of Wisconsin-Madison, 1989. J. Macromol. Sci. – Pure Appl. Chem.
86 T. Xie, A. E. Hamielec, Macromol. 1997, A34, 1017–1043.
Chem. Theory Simul. 1993, 2, 421–454. 102 P. Pladis, C. Kiparissides, J. Appl.
87 T. Xie, A. E. Hamielec, Macromol. Polym. Sci. 1999, 73, 2327–2348.
Chem. 1993, 2, 455–483. 103 A. Brandolin, M. H. Lacunz, P. E.
88 H. M. Hulburt, S. Katz, Chem. Eng. Ugrin, N. J. Capiati, Polym. React.
Sci. 1964, 19, 555–574. Eng. 1996, 4, 193–241.
89 R. C. Zabisky, W. M. Chan, P. E. 104 K. Tsai, R. O. Fox, AIChE J. 1996, 42,
Gloor, A. E. Hamielec, Polymer 1992, 2926–2940.
33, 2243–2262. 105 N. K. Read, S. X. Zhang, W. H. Ray,
90 S. Ochs, P. Rosendorf, I. Hyanek, AIChE J. 1997, 43, 104–117.
S. X. Zhang, W. H. Ray, Comp. Chem. 106 N. H. Kolhapure, R. O. Fox, Chem.
Eng. 1996, 20, 657–663. Eng. Sci. 1999, 54, 3233–3242.
91 A. Pertsinidis, E. Papadopoulos, C. 107 F. Bezzo, S. Macchietto, C. C.
Kiparissides, Comp. Chem. Eng. 1996, Pantelides, Comp. Chem. Eng. 2000,
20, S449–S454. 24, 653–658.
92 S. T. Balke, A. E. Hamielec, J. Appl. 108 J. P. Congalidis, J. R. Richards,
Polym. Sci. 1973, 17, 905–949. Polym. React. Eng. 1998, 6, 71–111.
93 H. Tobita, J. Polym. Sci.: Part B: 109 R. K. Mutha, W. R. Cluett, A.
Polym. Phys. 1993, 31, 1363–1371. Penlidis, Ind. Eng. Chem. Res. 1997,
94 F. Teymour, J. D. Campbell, Macro- 36, 1036–1047.
molecules 1994, 27, 2460–2469. 110 S.-M. Ahn, M.-J. Park, H.-K. Rhee,
95 A. Butté, A. Ghielmi, G. Storti, M. Ind. Eng. Chem. Res. 1999, 38, 3942–
Morbidelli, Macromol. Theory Simul. 3949.
1999, 8, 498–512. 111 Y. Yabuki, T. Nagasawa, J. F.
96 H. Tobita, Macromol. Theory Simul. MacGregor, Comp. Chem. Eng. 2000,
1996, 5, 129–144. 24, 585–590.
97 M. Wulkow, Macromol. Theory Simul. 112 M.-J. Park, H.-K. Rhee, Chem. Eng.
1996, 5, 393–416. Sci. 2004, 58, 603–611.
98 P. D. Iedema, M. Wulkow, H. C. J. 113 S.-S. Na, H.-K. Rhee, J. Appl. Polym.
Hoefsloot, Macromolecules 2000, 33, Sci. 2000, 76, 1889–1902.

Free-radical Polymerization: Suspension1

B. W. Brooks

Key Features of Suspension Polymerization

Many important polymers are made commercially via suspension polymerization

of vinyl monomers. These include poly(vinyl chloride), poly(methyl methacrylate),
expandable polystyrene, styrene–acrylonitrile copolymers and a variety of ion-
exchange resins and specialist materials. The annual polymer production from sus-
pension processes is very high.

Basic Ideas

There are two key requirements for any commercial free radical polymerization
process. The polymerization rate must be reasonably high and the polymer prod-
uct must have the correct molecular weight distribution. The conditions that are
necessary to achieve those goals can be predicted from the kinetics of homoge-
neous free radical polymerization (see Chapter 4). But those conditions cannot be
obtained easily in large-scale production when bulk processes are used. Uniform
mixing and temperature control are difficult to achieve because the high heat
of polymerization is combined with a large viscosity of the reaction medium. As
most polymers are poor thermal conductors, heat transfer from large reactors is
usually poor because the ratio of surface area to volume decreases as the reaction
mass increases.
In suspension polymerization, drops of a monomer-containing phase are dis-
persed in a continuous liquid phase. Monomer solubility in the continuous phase
is often low and polymer is produced inside the drops. Although the drop viscosity
increases with monomer conversion, the effective viscosity of the suspension re-
mains low and efficient agitation is possible. The ratio of surface area to volume
for small drops is relatively high and local heat transfer is good. If the continuous

1) The symbols used in this chapter are listed at

the end of the text, under ‘‘Notation’’.

Handbook of Polymer Reaction Engineering. Edited by T. Meyer, J. Keurentjes

Copyright 8 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
ISBN: 3-527-31014-2
214 5 Free-radical Polymerization: Suspension

phase is aqueous and well mixed, heat transfer from the reactor is also good. That
permits effective control of temperature and of those variables which depend on
temperature, which include reaction rates, polymer molecular weight, and copoly-
mer composition.
Suspension stability is maintained by good agitation and by the use of drop
stabilizers. Removal of the stabilizing agents, after polymerization, may not be
complete and some contamination of the final product is sometimes inevitable.
Suspension polymerization usually requires larger reactor volumes than bulk pro-
cesses because the vessels often contain about 50% of the continuous phase. Sus-
pension polymerization has been reviewed previously by Munzer and Trommsdorff
[1], Bieringer et al. [2], Warson [3], Brooks [4], Yuan et al. [5], Vivaldo-Lima et al.
[6], and Arshady [7].

Essential Process Components

The chemical events that occur inside drops of the dispersed phase are similar to
those found in bulk polymerization. The drops contain monomer (or monomers),
radical generators (often called initiators), and polymer. Sometimes chain-transfer
agents are added also. The continuous phase is often regarded as chemically inert,
but drop stabilizers are usually present in it and, in some cases, those stabilizers
participate in the polymerization process. For a discussion of stabilizer types, see
Section 5.2.1.

Polymerization Kinetics

Higher conversions of monomers can be accommodated more readily in suspen-

sion processes than in bulk processes because suspensions are more mobile than
molten polymers. Therefore, simple rate expressions may not be applicable be-
cause the values of some rate coefficients diminish at high polymer concentrations
(see Section 5.3.2).
It is often assumed that the polymerization chemistry which occurs in the dis-
persed phase is identical to that which occurs in the equivalent bulk process. That
assumption may be valid if the monomers and initiator are virtually insoluble in
the continuous phase. Then, polymerization rates, molecular weight distributions,
and copolymer compositions can be predicted from conventional kinetic schemes.
But drop stabilizers may react with species inside the drops (for example, to form
graft copolymers).
When all the monomers in a suspension polymerization are virtually immiscible
with the continuous phase, then the instantaneous copolymer composition can be
predicted from idealized relationships which apply to homogeneous systems. How-
ever, the use of those relationships is not straightforward if one, or more, of the
monomers is partially soluble in the continuous phase, because the actual compo-
sition of the drops may then be unknown. The effective monomer concentrations,
5.1 Key Features of Suspension Polymerization 215

Tab. 5.1. Apparent reactivity ratios in solution and suspension copolymerization (reproduced
with permission from Ref. 4).

Monomer 1 AS AN VC MMA
Monomer 2 S MA VA MAA

r1 (soln) 1.18 1.02 1.68 0.35

r2 (soln) 0.85 0.7 0.23 1.63
r1 (susp) 1.0 0.75 2.47 0.63
r2 (susp) 1.0 1.54 1.99 1.07
Temp [ C] 60 50 70 69

Abbreviations: AS, p-acetoxystyrene; S, styrene; AN, acrylonitrile; MA,

methyl acrylate; VC, vinyl chloride; VA, vinyl acetate, MMA, methyl
methacrylate; MAA, methacrylic acid.

which should be used in these relationships, might be predicted if the appropriate

partition coefficients for the two phases can be obtained. That is not often possible
but models that allow for water solubility of monomers have been developed for
the copolymerization of vinyl chloride and vinyl acetate [8] and for the copolymer-
ization of styrene and acrylonitrile [9].
Apparent reactivity ratios obtained directly from suspension polymerization ex-
periments may not be identical to those expected from the equivalent bulk pro-
cesses if some monomer migrates to the continuous phase. Ashady et al. [10]
found values for reactivity ratios that were not expected from results observed in
bulk or solution copolymerization. Izumi and Kitagawa [11] showed that reactivity
ratios for suspension copolymerization, of acrylonitrile and methyl acrylate, were
different from those obtained from either solution or emulsion polymerization.
Table 5.1 compares reactivity ratios obtained from solution copolymerization with
those observed in suspension copolymerization.
If the initiator in suspension polymerization is slightly soluble in water, then
simultaneous emulsion polymerization may occur when free stabilizer remains in
the continuous phase [12, 13].

Fluid–Fluid Dispersions and Reactor Type

Batch, or semi-batch, reactors are often used for suspension polymerization on an

industrial scale. Dispersions in tubular flow reactors are difficult to maintain and a
continuous stirred tank would produce drops containing partially polymerized ma-
terial that would coalesce in the receiving equipment. However, new types of flow
reactors are being developed for suspension polymerization (see Section 5.4.1).
Many of the published studies on drop behavior in agitated liquid–liquid disper-
sions are applicable to suspension polymerization. But sometimes their use is lim-
ited because they do not account for changing physical properties of drops or for
the presence of drop stabilizers.
216 5 Free-radical Polymerization: Suspension

Uses of Products from Suspension Polymerization

Advantages of suspension polymerization are not restricted to temperature control.

Some polymers, such as poly(vinyl chloride) (PVC), are immiscible with their
monomers. Subsequent polymer coagulation makes bulk processes difficult to con-
trol. In suspension polymerization, however, that problem is avoided. There, coag-
ulation is largely confined to the drop interiors and aggregation of polymerizing
drops is restricted. That is why suspension polymerization is used for the large-
scale production of PVC. In that case, the initial drop diameters, and the final par-
ticle sizes, range between 10 and 100 mm.
If a polymer product is required in particulate form, then suspension polymer-
ization is especially suitable. The energy required to disperse monomer drops is
lower than that required to break up the finished polymer. Also, drop size control
before polymerization is easier to achieve than particle size manipulation of granu-
lated polymer. Suspension polymerization provides a good route to functionalized
particles such as those used in ion-exchange resins. Expandable polystyrene beads
are also made by suspension polymerization. When a product is to be used in
‘‘bead form’’, initial drop diameters can be as large as 1–2 mm [1].

Stability and Size Control of Drops

For many applications, the size range of the final product particles is very impor-
tant. For example, bead diameters affect flow rates through ion-exchange columns.
But particle size can also be important when the polymer is to be converted to a
macroscopic object. Heat transfer rates to polymer particles during extrusion and
mass transfer rates of plasticizers in particulate polymers both depend on particle
During suspension polymerization, drop size depends on the physical properties
of the two phases, the phase ratio, the nature of the suspension flow, and the con-
dition of the phase interface. Interfacial tension and drop stability depend largely
on the nature of the drop stabilizer. If no stabilizer were used to protect the drops,
the suspension would be unstable and the final polymer particles would reach an
undesirable size.
The adsorption of stabilizer molecules at the interface between monomer and
the continuous phase reduces the interfacial tension and hence reduces the energy
required to form drops. Drop stability against coalescence depends largely on the
ability of the stabilizer to form a thin protective film at the interface. That gives
the drops better elastic properties [14]. The effect of the elastic properties is en-
hanced by increasing the concentration of the suspending agent, until a certain
surface coverage of the drops is reached. At that point, a ‘‘critical surface coverage’’
is established and a further increase in the suspending agent concentration will
have a very little effect on the drop stability [15].
5.2 Stability and Size Control of Drops 217

Stabilizer Types

Many drop stabilizers in suspension polymerization are water-miscible polymers.

These materials are sometimes called protective colloids. They include naturally oc-
curring substances, such as gelatin and pectin, and a wide range of synthetic poly-
mers such as partially hydrolyzed poly(vinyl acetate)s. Modified natural products
such as cellulose ether derivatives are also widely used. Munzer and Trommsdorff
provide a detailed list of patented stabilizers [1]. Polymeric stabilizers do not all act
in the same way but steric effects are often important [16]. Although transfer of
stabilizer molecules from the continuous phase to the drop surfaces can some-
times be fast, the development of drop stability may be slow [17]. That may arise
because rearrangement of stabilizer molecules on the drop surface is necessary.
Water-miscible polymers are not expected to be good drop stabilizers when the con-
tinuous phase is nonaqueous (see Section 5.5).
Partially hydrolyzed poly(vinyl acetate) (PVA), a common stabilizer, is often
called poly(vinyl alcohol), but that is a misnomer because not all the acetate groups
are hydrolyzed. The extent of hydrolysis (DH) has significant effects on the behav-
ior of the PVA. This is especially important in the suspension polymerization of
vinyl chloride. PVA with a DH between 70% and 80% is a good stabilizer for drops
in aqueous media. Drops retain their integrity even when agitation levels are re-
duced [18]. But when the DH is less than 60%, drop sizes become sensitive to re-
duction in agitation intensity [18]. Although PVAs with a low DH are poor drop
stabilizers in aqueous media, they are still useful because they affect product mor-
phology by influencing events inside the vinyl chloride drops (see Section 5.3.3).
Attempts have been made to measure the distribution of the stabilizer between
the two phases and the interface [19]. PVA often becomes grafted onto polymer
that is formed inside the drops, leading to the formation of a ‘‘skin’’ on the final
particle surface. This is important because subsequent removal of the skin is not
Although organic substances are commonly used as drop stabilizers, it is well
known that some particulate inorganic solids can also stabilize drops in suspen-
sion [20]. There are many reports of inorganic solids (such as calcium phosphate,
aluminium hydroxide, and zinc phosphate) being used in the suspension polymer-
ization of styrene [1]. In some cases, small amounts of surfactants (such as sodium
alkylsulfonates) are added to assist the dispersion of those solids. The effect of elec-
trolytes can also be important [21]. Inorganic stabilizers are advantageous when
only low levels of contamination are required because they can sometimes be re-
moved effectively from the final polymer particles. Also, inorganic particles are
able to stabilize relatively large drops, such as those formed in the manufacture of
expandable polystyrene beads.
Although some inorganic solids are effective stabilizers for drops in suspen-
sion polymerization, other solids are not stabilizers and may even be destabilizers.
O’Shima and Tanaka [22] suggested that the contact angle between dispersed liq-
uid and inorganic solid is a crucial factor in determining whether an inorganic
218 5 Free-radical Polymerization: Suspension

solid is a stabilizer or a destabilizer in suspension polymerization. Solids that pro-

vide a relatively large contact angle (such as aluminium hydroxide) would be stabil-
izers in aqueous media. In contrast, those which have a relatively small contact
angle (such as carbon black) would tend to be destabilizers. In practical operation,
the contact angle will probably vary with any given inorganic solid if different
monomers are used. In many cases it is not easy to measure contact angles.
Two theories, screen theory and coverage theory, have been suggested to explain the
mechanism of drop stabilization by inorganic solids [21]. According to the screen
theory, finely divided inorganic solids which are dispersed in water form a screen.
Dispersed monomer drops smaller than the mesh of the screen are free to move
through the meshes, while those bigger than the mesh cannot pass through the
meshes and are stopped from coalescing further. In the coverage theory, it is sug-
gested that dispersed inorganic solids cover the surfaces of monomer drops and
form a layer which prevents drop coalescence. Both theories are plausible but
many workers believe that the stabilization is obtained mainly by a coverage effect.
However, not all inorganic solids are adsorbed by the monomer drops. When Wol-
ters et al. [23] used hydroxyapatite and calcium carbonate in the suspension poly-
merization of styrene, the adsorption equilibrium was found to be far on the side
of desorption. Also, Wang and Brooks [24] noticed that many of their stabilizing
particles settled on the bottom of the vessel when stirring of a liquid–liquid disper-
sion ceased.
If some particles remain in the continuous phase, it is not feasible for the cover-
age theory to provide a description for every case of suspension polymerization in
which an inorganic solid is used as a stabilizer. Wang and Brooks suggested a
crowding theory which takes account of the dynamic effects of stabilizer particles
[24]. In that model, particles are not required to be permanently adsorbed on drop
surfaces but they become effective when two drops come close together. Although
crowding theory does not explain all the observations, it is compatible with a wide
range of experimental results. The addition of surfactants often improves the dis-
persion action of the fine particles, but it is possible to stabilize dispersions solely
with mono-sized spherical colloidal particles [25].

Drop Breakage Mechanisms

In an agitated suspension, the dispersed phase can be broken into small drops
when its surface is disrupted. That disruption can be caused either by frictional
forces (via viscous shear) or by inertial forces (via turbulence) [26]. The ratio of
the external disrupting force to the interfacial tension force is often expressed as
the Weber number, We. Drop deformation increases as We increases. When We ex-
ceeds a critical value, a drop will break into smaller drops.
The fluids in agitated vessels are often turbulent. If the turbulence in local re-
gions can be regarded as isotropic, a criterion for the drop breakage mechanism
can be developed [26, 27]. In turbulent flow, random eddies are superimposed on
the main flow. Eddy sizes are influenced by the location of the vessel walls and are
5.2 Stability and Size Control of Drops 219

restricted by the impeller diameter [28]. Kinetic energy is transferred to smaller

eddies in a sequential fashion, until it reaches the smallest ones. This transfer is
assumed to occur without energy dissipation. However, when the kinetic energy
reaches the small eddies, it is dissipated as heat to overcome the viscous forces.
In theories of local isotropy, it is assumed that the small eddies are statistically
independent of each other. Velocity fluctuations are determined by the local rate
of energy dissipation per unit mass of fluid (e) and by the kinematic viscosity (n).
Kolmogorov [29], by dimensional reasoning, derived an expression [Eq. (1)] for
the length of the smallest eddy (h).

h ¼ n 3=4 e1=4 ð1Þ

Rushton et al. [30] suggested that local isotropy could exist when the Reynolds
number is >10 4 . The macroscale of turbulence, L (approximately equal to the im-
peller diameter), is then much larger than the microscale of turbulence. Here, the
Reynolds number (Re) is defined by Eq. (2), where N is the impeller speed, D is the
impeller diameter, and n is the kinematic viscosity of the dispersion.

Re ¼ ND 2 =n ¼ ND 2 r=m ð2Þ

Pressure fluctuations can deform the drops and they may break if the inertial
forces exceed the interfacial tension forces. Kolmogoroff [27] and Hinze [31] de-
rived an expression for the maximum drop diameter, d max , that should be observed
when turbulence is isotropic. Here, d g h and the viscous forces may be neglected
in comparison with the inertial forces; d max , can then be related to a critical Weber
number, We crit , by Eq. (3).

We crit ¼ rc u 2 ðdÞd=s ð3Þ

The inertial forces are related to e. Thus, We crit is given by Eq. (4).

We crit ¼ ðC1 rc e 2=3 d max

Þ=s ð4Þ

Rushton et al. [30] showed that Eq. (5) holds, so that Eqs. (6) [32] and (7) follow.

e ¼ C3 N 3 D 2 ð5Þ
3=5 6=5 4=5
d max ¼ C4 ðs=rc Þ N D ð6Þ

d max =D ¼ C4 ðWeÞ0:6 ð7Þ

In order to use this expression, it is often necessary to relate d max (which is difficult
to measure) to the Sauter mean diameter, d32 (which is more readily determined).
A linear relationship between d32 and d max has been demonstrated by Sprow [33],
Coulaloglou and Tavlarides [28], Kuriyama et al. [34], and Zerfa and Brooks [35].
220 5 Free-radical Polymerization: Suspension

Volume fraction of dispersed phase


d 32 /D ( x 10 )

20 0.1


0 10 20 30
- 0.6 4
0.027(1 + 3.06 )We ( x 10 )
Fig. 5.1. Correlation of vinyl chloride drop size with volume
fraction j and Weber number We. Reproduced by permission of
Elsevier Science Ltd. From Ref. 35.

In the suspension polymerization of vinyl chloride, Zerfa and Brooks found the
relationship in Eq. (8).

d32 ¼ d max ð8Þ

Equation (6) has been verified with a large number of liquid–liquid dispersions
where the volume fraction of the dispersed phase is not high, that is, in non-
coalescing systems [28]. But, unlike many laboratory studies, real suspension poly-
merizations are operated with a high volume fraction of dispersed phase, j. Then,
drop coalescence and breakage happen simultaneously and damping of the turbu-
lent fluctuations may occur. To allow for this, a factor f ðjÞ is often introduced into
Eq. (7). Experimental expressions for f ðjÞ are reviewed by Yuan et al. [5], and by
Zerfa and Brooks [35]. The latter workers showed that, in the suspension polymer-
ization of vinyl chloride, d32 was given by Eq. (9).

d32 =D ¼ 0:027ðþ3:06jÞWe0:6 ð9Þ

Equation (9) applied when values of j ranged between 0.01 and 0.4, as shown in
Figure 5.1. An expression with similar form is given by Borwankar et al. [15].
When drop coalescence is significant, Shinnar [32] showed that there will be a
minimum drop size, d min , that is proportional to N 0:75 . The extent of drop coales-
cence is reduced by addition of an appropriate drop stabilizer. If the surface cover-
age of the drops by the stabilizer exceeds a critical value then the dispersion can
remain stable after agitation ceases [15].
It must be remembered that correlations between average drop sizes and We
only apply at steady state. The steady-state drop size distribution (DSD) can take
some time to become established [18, 36–38]. That may not be a serious problem
with nonreacting dispersions but, in suspension polymerization, the physical prop-
5.2 Stability and Size Control of Drops 221

erties change with time so that the DSD can continue to change during the process
When the diameter of the drops is less than the Kolmogorov length h, stresses
from viscous shear will be much larger than those from inertial effects. Drop
breakage is then the result of viscous shear [26]. Here, again, a drop will break if
the deformation variable, sometimes described as a generalized Weber group [31],
exceeds a critical value [39]. Taylor showed [40] that the extent of drop distortion,
before breakup, depended on the ratio of phase viscosities (Eq. (10), where md and
mc are the viscosities of the dispersed and the continuous phase, respectively).

We crit ¼ mc ðqu=qrÞðd=sÞ ¼ f1 ðmd =mc Þ ð10Þ

Shinnar and Church [26, 32] demonstrated that for locally isotropic flow where
ðqu=qrÞ 2 ¼ e=n, d max is given by Eqs. (11) and (12).

d max ¼ ðsnc1=2 =mc e 1=2 Þ f ðmd =mc Þ ð11Þ

d max z ðsnc1=2 =mc ÞN 3=2 D1 f ðmd =mc Þ ð12Þ

When viscosity-increasing agents were present in the continuous phase for the sus-
pension copolymerization of styrene and divinylbenzene, Jegat et al. [41] found
that drop breakage could occur via viscous shear when flow in the suspension was
turbulent. In that case, the Kolmogorov length would have been larger that that
found when low-viscosity aqueous solutions are used.
The form taken by f ðmd =mc Þ depends on the nature of the flow. In laminar flow,
the contributions of rotational and elongational components can be expressed via
the value of a parameter a, which ranges between 0 and 1. For simple shear a ¼ 0,
and for pure elongational (hyperbolic) flow a ¼ 1 [43]. With low a values, We crit
(now equivalent to the critical capillary number Cacrit ) passes through a minimum
as md =mc increases. After the minimum is reached, relatively small increases in
md =mc are accompanied by large increases in We crit . When a ¼ 0, We crit tends to
an infinitely high value if md =mc > 4 so that drop breakage no longer occurs [42,
43]; see Figure 5.2.
Correlations that are found in idealized laboratory studies only provide rough
guides to events that occur in commercial reactors. Local flow in large stirred ves-
sels is often ill defined. Also, drop breakage can produce more than two new drops.
Chatzi and Kiparissides [44] suggested that the formation of a daughter drop is ac-
companied by the formation of a number of smaller satellite drops. Drop stabi-
lizers influence turbulence near drop surfaces and non-Newtonian behavior leads
to further complications. Therefore, it is not surprising that the many published
studies on drop behavior in liquid–liquid suspensions lead to a variety of different
results. Leng and Quarderer [39] found that d max depended on N 0:8 . By using data
from the literature, for the suspension polymerization of methyl methacrylate, they
also showed that d max depended on mc0:5 when drop breakage occurred by viscous
shear. Presumably, the drop size was not able to respond to changes in md =mc that
222 5 Free-radical Polymerization: Suspension


1.2 A

Ca crit

0 1 2 3 4
log10 (viscosity ratio) + 3
Fig. 5.2. Critical capillary number against log10 of viscosity
ratio (md =mc ) for laminar flow. (A) a ¼ 0 (simple shear flow);
(B) a ¼ 1 (hyperbolic flow). Data from Refs. 42 and 43.

would have occurred during polymerization. However, average drop size was found
to be a linear function of interfacial tension, as expected, when viscous shear was
important. Borwankar et al. [15] suggested that the relationship between d max and
agitator speed may depend on agitator type.
When polymer-containing drops are broken, their elastic properties must be
taken into account. Arai et al. [45] derived a correlation for drop breakup from a
Voigt model to represent the elastic properties. The validity of the correlation was
confirmed experimentally using a dispersed phase with a wide range of viscosity.
Wang and Calabrese [46] carried out experiments with drops made from well-
characterized model fluids. They showed that the influence of interfacial tension
on drop breakage decreased as the dispersed-phase viscosity increased.

Drop Coalescence

The volume fraction of drops in commercial suspension polymerization reactors is

usually high and drop coalescence cannot be ignored. In liquid–liquid dispersions
the drop size distributions (DSDs) depend on the breakage and on coalescence
From experiments in which both drop breakage and coalescence occurred, Kuri-
yama et al. [34] found that drop sizes reached a steady value within an hour, when
the initial drop viscosity was low. But with a high drop viscosity, the drop size re-
duction continued for longer periods of time and the final drop size was higher.
Although model, nonreacting, fluids were used for those experiments, the results
are relevant to suspension polymerization. In their study of drop coalescence in the
suspension polymerization of styrene, Konno et al. [47] found that the Sauter
mean diameter increased as the polymer viscosity increased. They also concluded
that the stabilizer does not effectively prevent the coalescence of drops with diame-
ters larger than d max .
The overall rate of drop coalescence is related to the collision frequency of the
drops and to the coalescence efficiency. By comparing drop collisions in agitated
5.2 Stability and Size Control of Drops 223

dispersions with molecular collisions in homogeneous fluids, previous workers

have developed expressions for collision rates between drops of different sizes [48,
49]. If drops adhere for sufficient time to allow them to deform, and to permit
drainage of the continuous phase that is trapped between them, then coalescence
may occur [26]. By taking account of these events, expressions can be obtained for
the coalescence efficiency [50]. Expressions, for coalescence and for collision rates
are not easy to use because they often contain parameters that are difficult to quan-
tify. Alvarez et al. [50] constructed a model for drop breakage and coalescence, in
the suspension polymerization of styrene, which takes account of viscosity effects.
That model assumes that breakage of a drop, exposed to a turbulent flow field, is a
result of fluctuations with a wavelength equal to the drop diameter. Fluctuations
with wavelengths that are smaller or larger than the drop diameter do not lead to
drop breakage.

Drop Size Distributions

Even when expressions for drop breakage and coalescence rates are available, their
successful use must allow for variations of turbulence intensity within the reactor.
Maggioris et al. [51] describe a two-compartment population balance model for an
agitated suspension polymerization reactor. That model distinguishes between the
impeller region and the remainder of the reactor. Both regions are assumed to be
well mixed and CFD simulations predicted drop size distributions that were com-
patible with experimental results from nonpolymerizing model liquid–liquid dis-
persions. For some combinations of stabilizer type and stirrer speed, bimodal
drop size distributions were predicted by the model and found in the experiments.
Yang et al. [52] showed how the size distribution of nonreacting styrene drops
changed with agitation time. Bimodality developed and the relative size of the two
peaks in the distribution depended on the amount of drop stabilizer that was used.
Calabrese et al. [46, 53] showed that, in dilute agitated suspensions for which co-
alescence is negligible, the equilibrium drop size distribution broadened consider-
ably as dispersed-phase viscosity increased.
If no chemical reaction occurs, then the rates of drop breakage and drop coales-
cence eventually become equal and a stable DSD is obtained [54]. Considerable
periods of time might be required for that to occur [18], especially when the drops
have a high viscosity [55]. But, in the case of suspension polymerization, the phys-
ical properties of the drops change with monomer conversion. In a batch process,
early increases in drop viscosity reduce the rates of breakage and coalescence be-
fore a steady-state DSD can be established. Then, the DSD continues to change,
and the average drop size increases, as the polymer content of the drops increases.
Consequently, the final particle size distribution (PSD) is quite different from the
steady-state DSD that would be expected from nonpolymerizing drops. That was
shown to be the case by Jahanzad et al. [37] in the suspension polymerization of
methyl methacrylate. There, the DSDs of polymerizing drops were broader than
that of the DSDs in the corresponding nonpolymerizing monomer drops. But the
224 5 Free-radical Polymerization: Suspension

average drop size in the nonpolymerizing drops was only slightly smaller than that
in the polymerizing suspension when a large amount of stabilizer was used, be-
cause excess stabilizer reduced drop coalescence in both systems. Similar results
were obtained by Konno et al. [47]. With smaller amounts of stabilizer, a growth
stage exists in which drop coalescence continues when drop breakage rates are
low [50]. The increase in average drop size can be substantial and continues until
the identification point is reached when drop viscosity is too high to permit further
drop coalescence. Jahanzad et al. [37] also showed that, although the diameters of
most drops and particles ranged between 10 and 300 mm, a second peak appeared
in the particle size distribution. The average diameter of particles that accounted
for that second peak was about 1 mm. Most of those smaller particles probably de-
veloped from satellite drops which are formed during the breakage process. The
existence of small satellite drops is compatible with the ideas of Chatzi and Kiparis-
sides [44] and with the work of Sathyagal et al. [56]. But some of the very small
particles could have been formed directly in the aqueous phase because the water
solubility of the free radical initiator (lauroyl peroxide) may be high enough to in-
duce some emulsion polymerization. That can by important with monomers that
are more water-soluble, such as methyl methacrylate, vinyl acetate, and vinyl chlo-
ride. If emulsion polymerization becomes prevalent, then a significant amount of
the drop stabilizer will be adsorbed on the surfaces of the small particles. The
amount of stabilizer that is available to stabilize the drops is then reduced [57].
In batch reactors, the rate of drop coalescence is affected by changes in drop
viscosity. Therefore, any induced kinetic effects that alter the polymer molecular
weight (and, thereby, change the viscosity) could influence the coalescence rate
and the DSD. Promoting chain transfer is one way for that to happen.
When drop viscosity remains low, coalescence events can depend on the nature
of the drop stabilizer. Low viscosities may be encountered when the polymers, or
copolymers, are immiscible with their monomers. The polymers then precipitate
inside the drops and their apparent viscosity does not increase substantially until
appreciable conversions have been attained. That is the case with the suspension
polymerization of vinyl chloride. When hydrolyzed poly(vinyl acetate) (PVA) is
used as the drop stabilizer, Zerfa and Brooks [18] showed that DSDs depended on
stirrer speed. With ‘‘good’’ PVA stabilizers, that had a high degree of hydrolysis, a
reduction in stirrer speed had little effect on the DSD, as shown in Figure 5.3. But
when a ‘‘poor’’ PVA stabilizer, with a low degree of hydrolysis, was used, a reduc-
tion in stirrer speed led to a broadening of the DSD. There, the DSD became sim-
ilar to that expected when the stirrer speed was maintained at its lowest value,
as shown in Figure 5.4. Clearly, the ‘‘good’’ stabilizer protected the drops from

Drop Mixing

Most of the coalescence events described in Section 5.2.3 occur when drops have a
uniform chemical composition, but sometimes it is necessary to add material to a
5.2 Stability and Size Control of Drops 225

Drop number 200

150 A



15 35 55 75 95

Drop diameter (microns)


Drop number




15 35 55 75 95

Drop diameter (microns)

Drop number

150 C



15 35 55 75 95

Drop diameter (microns)

Fig. 5.3. Change in drop size with stirrer speed with PVA
72.5% hydrolyzed. (A) N ¼ 350 rpm (5.8 s1 ); (B) N ¼ 650
rpm (10.8 s1 ); (C) N reduced from 650 to 350 rpm. Data from
Ref. 18.

reactor in which a suspension already exists. That can happen if control of copoly-
mer composition is important. In batch operation, copolymer composition usually
changes with overall conversion because monomers react at different rates [58].
This drift in copolymer composition may be limited by adding one of the mono-
226 5 Free-radical Polymerization: Suspension


150 A

Drop number 100


15 35 55 75 95

Drop diameter (microns)


Drop number



15 35 55 75 95

Drop diameter (microns)


Drop number



15 35 55 75 95 115

Drop diameter (microns)

Fig. 5.4. Change in drop size with stirrer speed with PVA 55%
hydrolyzed. (A) N ¼ 350 rpm (5.8 s1 ); (B) N ¼ 650 rpm
(10.8 s1 ); (C) N reduced from 650 to 350 rpm. Data from
Ref. [18].

mers to the reactor incrementally (that is, by using a semi-batch procedure). But,
when new dispersible material is added to an existing suspension, the new mate-
rial and existing drops can remain segregated for a significant period of time.
Then, new drops may form with a monomer composition that differs from that of
5.2 Stability and Size Control of Drops 227

the original drops. The new monomer cannot mix with existing drops (which con-
tain the initiator and have adsorbed most of the drop stabilizer). Adding extra drop
stabilizer, with the new monomer, might not be helpful because there is the dan-
ger of stabilizing new drops with the ‘‘wrong’’ monomer composition that contain
little, or no, radical generator.
Hashim and Brooks [59] studied the addition of styrene to a suspension of stabi-
lized drops. The drops were composed of polystyrene solutions in styrene. The ini-
tial drop viscosity affected the drop size and the rate of coalescence between drops.
As the dispersed-phase viscosity increased, the drop size distribution broadened; at
some stirrer speeds, the mixing rate increased. It appears that there is a critical
drop size which determines the coalescence efficiency. Above that size, the drop
mixing rate increases as the drop viscosity decreases. Below the critical drop size,
the mixing rate is influenced noticeably by the drop size; as the drop size increases,
the coalescence rate also increases.
Drop mixing may become an issue when volatile monomers are used. The en-
thalpy of polymerization for most of the vinyl monomers that are used in suspen-
sion polymerization ranges between 30 and 90 kJ mol1 . Therefore, a high heat re-
moval rate is usually necessary to maintain a constant reactor temperature. This is
difficult to achieve by heat transfer through the reactor walls in commercial opera-
tions because large reactors have a relatively small surface area to volume ratio.
Heat removal can be improved by allowing the monomer to vaporize. The vapor
is then condensed, cooled, and returned to the reactor as a liquid. If the polymer-
ization process is to be maintained, drops of returning monomer must gain access
to initiator and drop stabilizer. With the suspension polymerization of vinyl chlo-
ride, Zerfa and Brooks [60] found that monomer returning from a reflux con-
denser formed drops that acquired initiator without the need for coalescence with
existing stabilized drops. The presence of small particles, formed by simultaneous
emulsion polymerization, appeared to provide a mechanism for transfer of initiator
(bis(4-t-butylcyclohexyl) peroxy dicarbonate) through the continuous phase. With
high reflux rates, the drop size distribution became bimodal whereas, in the ab-
sence of reflux, a monomodal DSD is expected [60]. New drops, from refluxed
monomer, had limited access to the drop stabilizer (PVA) and were larger than
the ‘‘old’’ drops.
A special drop mixing problem arises with the suspension polymerization of
vinyl chloride. Because the monomer is very reactive and has a high enthalpy of
polymerization, operators are reluctant to mix initiator in the monomer before a
suspension is formed. Therefore, as a safety precaution, the initiator is often dis-
persed in the aqueous phase of a stabilized suspension. Then the subsequent mix-
ing of monomer and initiator can be quite slow. Zerfa and Brooks [61, 62] showed
that many monomer drops remained ‘‘uninitiated’’ when monomer in other drops
had polymerized to a considerable extent. That did not happen when initiator was
dissolved in the monomer before it was dispersed: see Figure 5.5. This nonuni-
formity in drop behavior affected the final polymer properties. Also, addition of ini-
tiator via the aqueous phase promoted simultaneous emulsion polymerization and
modified the PSD. Drop mixing rates were quantified by using dyed monomer
228 5 Free-radical Polymerization: Suspension

Fig. 5.5. Effect of the method of initiator 1, 5 min; 2, 20 min; 3, 60 min. N ¼ 350 rpm
addition to vinyl chloride: (A) initiator (5.8 s1 ); j ¼ 0:1; PVA concentration ¼ 0.06%.
predissolved in drop phase; (B) initiator Reproduced by permission of John Wiley &
predispersed in continuous phase. Panels: Sons, Inc. From Ref. 62.

drops in the experiments. The mixing rate was found to depend on both the con-
centration of the PVA and the grade that was used [61].

Events at High Monomer Conversion

Suspension polymerization provides a practical method of achieving high mono-

mer conversions. Therefore, studies of polymerization and drop behavior that deal
with events at low monomer conversion may not be applicable to suspension poly-
5.3 Events at High Monomer Conversion 229

Breakage of Highly Viscous Drops

In batch reactors drop breakage and coalescence are affected by the polymerization
process because the viscosity of the polymerizing fluid often increases. Many as-
pects of the interaction of drop behavior with the polymerization process have
been discussed already (see Section 5.2).

Polymerization Kinetics in Viscous Drops

Free radical polymerization kinetics has received much attention and many aspects
of the process are well understood (see Chapter 4). Most academic investigations
have been carried out in ‘‘idealized’’ conditions where the extent of monomer con-
version is low. The classical expression for the rate of polymerization (R p ), in a
single-phase reaction, is Eq. (13), where k p is the propagation rate coefficient, CM
is monomer concentration, R i is the initiation rate and k t is the termination rate
coefficient [63].

R p ¼ k p CM ðR i =k t Þ 1=2 ð13Þ

If radicals are generated by the thermal decomposition of an added initiator then,

at steady-state conditions, Eq. (14) applies, where f is an efficiency factor, kd is the
initiator decomposition rate coefficient and CI is the initiator concentration.

R i ¼ 2f kd CI ð14Þ

Chain termination is often diffusion-controlled and the value of k t diminishes sub-

stantially as the polymer concentration increases. At high polymer concentrations,
k p decreases also [64]. The reduction in k t leads to an increase in the polymeriza-
tion rate, a phenomenon often described as a ‘‘gel effect’’.
Although most workers agree that the chain termination reaction is diffusion-
controlled, reliable quantitative relationships between the rate coefficients and
measurable properties of the reaction medium are not generally available. Radical
diffusion can depend on solution viscosity, polymer volume fraction and polymer
molecular weight. The latter three entities are interrelated in complicated ways;
meaningful experiments, which may relate them to the rate coefficients, are not
easy to devise [65]. However, Brooks et al. showed that effects of viscosity changes
on polymerization rate could be distinguished from the effects of polymer volume
fraction [66]. In some cases, the value of f may also depend on polymer content
The value of kd can often be determined independently but traditional steady-
state experiments give values for k p =k t0:5 and do not provide separate values for k p
and k t . This restriction becomes a problem for reactor modeling because, in the
presence of polymer, both k p and k t diminish (even in isothermal conditions).
230 5 Free-radical Polymerization: Suspension

Pulse laser techniques, combined with accurate molecular weight measurement,

have been used to determine independent values for k p but those techniques are
only feasible in the absence of preformed polymer [68, 69].
When a gel effect is observed, kinetic analysis can be difficult because k t be-
comes dependent on chain length [70] and the pseudo-steady state assumption
may not be valid [71]. Tefera et al. [72] showed that, when a high monomer conver-
sion is attained, a number of different models for the gel and glass effects gave an
adequate description of isothermal time–conversion data over the entire conver-
sion range for a single type and loading of initiator. But models that ignored the
effect of polymer molecular weight on the diffusion of macro radicals failed to de-
scribe the time–conversion data if the concentration of the initiator varied. It has
been shown that autoacceleration in free radical polymerization can be observed
even when low molecular weight polymer is being formed, indicating that chain
entanglements are not necessary for a gel effect to occur [73].
In practice, isothermal conditions are not always maintained in suspension poly-
merization and temperature increases can lead to a marked reduction in initiator
concentration. As the monomer conversion increases, the glass transition temper-
ature of the polymer solution also increases and the drops can become glassy. The
chain propagation reaction then becomes diffusion-controlled and the value of k p
decreases significantly. An increase in reaction temperature may become necessary
in order to achieve complete monomer conversion. Although most workers agree
that k p is reduced at high monomer conversions, Faldi et al. [74] have suggested
that diffusion control may not be the only reason for that to occur. Qin et al. [75]
developed a three-stage model to account for the gel and glass effects. The match
with experimental data, for methyl methacrylate and styrene polymerizations, up to
high monomer conversion was satisfactory for isothermal conditions. Achilias and
Kiparissides [76] showed, however, that data for polymerization of those mono-
mers were compatible with a model that did not require ‘‘break points’’ for the gel
and glass effects. The initiator efficiency, f , has been shown to depend on the size
of the initiator radicals at high monomer conversion [76].
If crosslinking or copolymer precipitation occurs, bulk polymerization may be
difficult to handle. A suspension process may then be the only feasible way in
which the copolymerization can be carried out [4]. Suspension processes also pro-
vide a means of investigating copolymerization kinetics at high conversion. The
monomer sequence in styrene–methyl methacrylate copolymers at high conver-
sion have been found to differ from those observed at low conversion [77].

Consequences of Polymer Precipitation Inside Drops

Polymers that are insoluble in their monomers will precipitate during polymeriza-
tion. The resulting fouling problems that occur in bulk polymerization are greatly
reduced by using suspension polymerization. This is one of the reasons for
choosing a suspension process for PVC manufacture, as discussed previously (see
Section 5.1.5). Manipulation of polymer precipitation, inside the drops, during
5.3 Events at High Monomer Conversion 231

Fig. 5.6. Schematic representation of polymer formation inside vinyl chloride drops.

polymerization can influence the properties of the final product, especially the po-
rosity, which in the case of PVC affects the ability of the polymer to take up plasti-
cizers. Structural changes that occur inside the drops of polymerizing vinyl chlo-
ride monomer (VCM) have been discussed by a number of authors [78, 79].
PVC starts to precipitate from the monomer phase when the conversion exceeds
0.1% conversion, forming a separate polymer-rich phase inside the drops. The pre-
cipitating polymer aggregates to form unstable microdomains, which aggregate
further to give domains. Subsequent aggregation of domains results in the for-
mation of primary particles. The primary particles grow by polymerization within
them and by buildup of microdomains or domains on their surfaces. Eventually
multiple contacts lead to the formation of a continuous network of primary par-
ticles throughout the polymer particle/monomer droplet; see Figure 5.6. At about
70% conversion the monomer-rich phase disappears and further polymerization
occurs in the polymer-rich phase [80, 81]. The final polymer grains have irregular
shapes, unlike particles that form from polymers that are completely miscible with
their monomers. Some components of the drop stabilizers are chosen because they
are miscible with the monomer and they can influence the agglomeration of pri-
mary particles [82]. Those components are sometimes called secondary suspend-
ing agents (SSAs).
As a direct consequence of the process described above, and of the density differ-
ence between PVC and VCM, PVC grains have a complicated morphology and they
can be highly porous. The particle size, the PSD, the morphological characteristics,
and the degree of porosity of PVC grains depend on polymerization conditions.
These include the agitation in the reactor, the type and concentration of suspend-
232 5 Free-radical Polymerization: Suspension

ing agent(s), the secondary suspending agents, the polymerization temperature,

the monomer conversion, and the ratio of monomer to water.
Effects of agitation on the mean particle size of PVC resins have been studied by
many researchers, using different reactor capacities, and many correlations have
been developed [83, 84]. The effects of agitation on PVC porosity have also received
much attention [80]. As with the suspension polymerization of other monomers,
the choice of suspending agent(s) affects the particle size distribution of the final
polymer. In the case of PVC, however, the suspending agents also affect the sub-
structure and porosity of the particles. The primary suspending agent system is
often based on PVA, substituted cellulose, or a mixture of the two. Wolf and
Schuessler [85] concluded that the plasticizer absorption (which depends on po-
rosity) of the resulting PVC was related to the surface activity of the suspending
agent, regardless of type. Ormondroyd [86] demonstrated the effect of PVA struc-
ture on the particle size, cold plasticizer absorption, and bulk density of the PVC
produced. Cheng [87] and Cheng and Langsam [88] used hydroxypropyl methylcel-
lulose (HPMC) as a suspending agent and analyzed the influence of molecular
weight and chemical structure of the cellulose on the particle morphology of the
resulting PVC. Cebollada et al. [81] showed how HPMC structure and concentra-
tion influenced the particle properties of PVC. SSAs are regularly used in the pro-
duction of suspension PVC to impart higher porosity to the PVC grains. That
improves the subsequent uptake of plasticizer and also promotes the removal of
unreacted VCM at the end of polymerization. PVA with a low degree of hydrolysis
and nonionic surfactants, such as sorbitan monolaurate, have been used commer-
cially as SSAs [89]. The mechanism by which such SSAs function is not entirely
clear [80, 90–92].
When nonionic surfactants, such as Span85, Span60 and Span20, are used as
SSAs, with PVA as the main stabilizer, the mean particle size of PVC resins in-
creases quickly as the concentration of nonionic surfactant increases [82]. The
degree of agglomeration of primary particles increases with polymerization tem-
perature and with conversion. A nonionic surfactant with a greater hydrophile-
lyophile balance (HLB) value is more effective in raising the particle size [82].
The increase in the mean particle size of the final PVC is probably caused by in-
creased coalescence of VCM/PVC droplets during the polymerization process, as a
result of a lowering of water-drop interfacial tension by the SSA. Nonionic surfac-
tants such as Span20, Span60, and Span85 have more affinity with VCM than the
PVA has, and they will be absorbed faster than PVA on the VCM/water interface.
Part of the interface, which would otherwise be occupied by PVA molecules, will
become occupied by nonionic surfactant molecules. Thus, the colloid protective
ability of the composite suspending agent would decrease because the nonionic
surfactants with a low molecular weight have a lower colloid protective ability
than that of PVA. In some cases, graft copolymers of PVA and PVC form a skin
(or membrane) around the polymer particles [93]. Zerfa and Brooks [60] showed
that, at low monomer conversion, PVC porosity increases if a high monomer reflux
rate is used. Porosity, surface roughness, and particle shape were found to depend
on the origin of the PVA [62]. Particle shape can also be influenced by the method
5.3 Events at High Monomer Conversion 233

CPA (g DOP/100g PVC)
0 20 40 60 80 100
Polymerisation conversion (%)
Fig. 5.7. Effect of polymerization conversion on the cold
plasticizer absorption (CPA) of PVC. Reproduced by permission
of John Wiley & Sons, Inc. From Ref. 82.

of initiator addition [62]. High conversions of VCM are not always desirable be-
cause the porosity of PVC particles usually decreases linearly with monomer con-
version, as shown in Figure 5.7 [60, 82].
When a nonionic surfactant is used as an SSA, in conjunction with PVA, the
porosity of PVC increases as the concentration of nonionic surfactant increases.
Here, increased porosity may be the result of incomplete drop coalescence creating
voids at the sites where droplets are not well contacted [82]. Nonionic surfactant
with a lower HLB value (hence with a higher affinity for VCM and a higher solubil-
ity in VCM) is more effective in raising the product porosity. The surface of pri-
mary particle aggregates becomes coarser as surfactant is added [82]. This could
be the consequence of an altering interfacial tension between the PVC-rich phase
and the monomer. That might be expected to decrease the contact deformation of
the primary particles and increase the pore space. When nonionic SSAs are used,
PVC porosity decreases linearly with an increase of polymerization temperature
[82]. The porosity increases as the concentration of nonionic surfactant increases,
and a surfactant with a lower HLB value is more effective in raising the porosity.
The increase in porosity may be caused by a combination of increased coalescence
of VCM/PVC droplets and the nonionic surfactant’s steric effect inside the droplets
[82]. Bao et al. showed that particle morphology and PVC properties can be con-
trolled by using blends of PVA suspending agents with differing degrees of hydro-
lysis [94]. The use of PVA, with a low degree of hydrolysis, as an SSA increased
particle porosity in the suspension copolymerization of vinyl chloride and propy-
lene [95]. In that case, the SSA was more soluble in the organic phase than the pri-
mary stabilizer (a cellulose ether).
In vinyl chloride polymerization, particle porosity facilitates the subsequent up-
take of plasticizers, but in other cases particle porosity is induced to enhance access
to functional groups within the polymer. That is useful when polymer beads are
required for use in ion-exchange columns, or in analytical instruments. A diene
comonomer can be added to a monomer, to promote crosslinking and phase sepa-
234 5 Free-radical Polymerization: Suspension

ration inside the drops [96–99]. Appropriate functional groups can be incorporated
in the monomers.

Reaction Engineering for Suspension Polymerization

The discussion above (see Sections 5.2 and 5.3) provides guidelines for the design
of suspension polymerization processes on an industrial scale. However, extrapo-
lating from idealized small-scale studies to full-scale commercial operation is not
straightforward. Maintenance of uniform reactor conditions cannot be guaranteed
inside large reactors and controlled heat transfer from reactors can be difficult.

Dispersion Maintenance and Reactor Choice

Many suspension polymerization processes employ stirred vessels. Oldshue et al.

[100] showed that agitator type and vessel geometry have an important influence
on internal liquid flow. This must be taken into account in the design of large ves-
sels for suspension polymerization. Maggioris et al. [51] and Vivaldo-Lima et al.
[101] devised two-compartment population balance models to account for the large
spatial variations of the local turbulent kinetic energy. These permit the prediction
of drop size evolution in a suspension polymerization reactor with a high volume
fraction of dispersed phase. Basic expressions were modified to allow for evolving
physical properties of the suspension. Maggioris et al. [51] showed that it was fea-
sible to estimate the volume ratio of the impeller and circulation regions, the ratio
of turbulent dissipation rates, and the exchange flow rate of the two compartments
at different agitation rates and continuous-phase viscosities. Although many indus-
trial suspension polymerizations are described as ‘‘batch’’ processes, they are not
genuine batch operations because some material enters the reactor after polymer-
ization has started. In some cases, substantial reflux of monomers occurs and con-
densed monomer returns to the reactor continuously. In others, a semi-batch pro-
cess is used to control product composition. In both these situations, unconverted
monomer with low viscosity is being mixed with drops of higher viscosity. The com-
plexities of these mixing processes have been discussed above; see Section 5.2.5.
Reactor configurations other than conventional stirred tanks have been proposed
for suspension polymerization. Draft tubes, or ‘‘internal loops’’, can be used for
suspension polymerization [102], but drop size changes can occur [103] and flow
patterns may be complicated [104]. Tanaka et al. [105] used a loop reactor for the
suspension polymerization of styrene. They employed a double agitation method
to control the transient droplet diameter distribution and the final particle size dis-
tribution. Ni et al. [106] developed an oscillatory baffled reactor for batch suspen-
sion polymerization of methyl methacrylate. Fluid mixing was achieved by eddies
that are generated when a fluid passes through a set of equally spaced, stationary,
orifice baffles that are located inside a tube. Periodically formed vortices were con-
5.4 Reaction Enginee