Chemical Thermodynamics For Industry

Chemical Thermodynamics for Industry
Chemical Thermodynamics
for Zndustry
Edited by
Trevor Letcher
University of Natal, Durban, South Africa
RSC
advancing the chemical sciences
ISBN 0-85404-591-0
A catalogue record for this book is available from the British Library
0 The Royal Society of Chemistry 2004
All rights reserved
Apart from fair dealing for the purposes of researchfor non-commercial purposes or for pri-
vate study, criticism or review, as permitted under the Copyright, Designs and Patents Act
1988 and the Copyright and Related Rights Regulations 2003, this publication may not be
reproduced, stored or transmitted, in any form or by any means, without the prior permission
in writing of The Royal Society of Chemistry, or in the case of reproduction in accordance
with the terms of licences issued by the Copyright Licensing Agency in the UK, or in accor-
dunce with the terms of the licences issued by the appropriate Reproduction Rights
Organization outside the UK. Enquiries concerning reproduction outside the terms stated
here should be sent to The Royal Society of Chemistry at the address printed on this page.
Published by The Royal Society of Chemistry,

Thomas Graham House, Science Park, Milton Road,
Cambridge CB4 OWF, UK
Registered Charity Number 207890
For further information see our web site at www.rsc.org
Typeset by Macmillan India Ltd, London, UK

Printed by Athaneum Press Ltd, Gateshead, Tyne and Wear, UK
Foreword
The International Association of Chemical Thermodynamics (IACT) was estab-
lished in 2002 to continue the activities of the former IUPAC Commission 1.2 on
Thermodynamics. The challenge ahead for IACT is to provide leadership in acade-
mia and industry by maintaining the highest standards in research, teaching and
application in all areas of chemical thermodynamics. At the crossroad of quite dif-
ferent areas, the communication among such a diversified audience of scientists,
teachers and engineers includes publication of reference data collections, recom-
mendations for using and applying thermodynamics and related techniques, publi-
cations of selected reviews in refereed journals as well as monographs and collective
works. In each case, the responsibility of IACT is the identification and selection of
the best authors/contributors. In this context, I would like to point out that IACT con-
tinues the successful series “Chemistry for the 21st Century” with contributions
from international experts, which may be profitably used even by readers not spe-
cialized in the fields covered. Following the volume “Chemical Thermodynamics”,
the present one on “Chemical Thermodynamics for Industry” deals with a large
spectrum of applications of thermodynamics in modern chemical industry. It is a col-
lection of contributions showing the importance of thermodynamics for pushing
frontiers of efficiency and economics towards increasingly better-quality products
and services for the benefit of mankind.
The book assembles 23 chapters authored by leading thermodynamicists.The ini-
tiative for this work originated with Professor Trevor Letcher, the editor, who invested
considerable energy and patience to select, collect and organize these timely high-
quality contributions. Undoubtedly, the book will find its place within IUPAC’smajor
publications. As President of IACT, it is my honour and pleasure to thank Professor
Letcher for his guidance in such an enterprise. My thanks also extend to the authors
for making their knowledge available and for the time they devoted to this work.
Jean-Pierre E. Grolier
January 2004
Preface
The history of chemical thermodynamics goes back 150 years and the subject has, for
over a century, been the foundation for much of chemistry. This historical importance
is often associated with an attitude that thermodynamics has little relevance to modern-
day chemistry and will have little importance in the future development of chemistry.
To counteract this, the W A C Commission 1.2 on Chemical Thermodynamics, pub-
lished a volume entitled “Chemical Thermodynamics for the 21st Century” in 1999. It
consisted of 27 chapters, all focusing on the applications of thermodynamics to very
recent developments in chemistry. The aim was to highlight the role that thermody-
namics was playing at the forefront of chemical research at the dawn of the 21st cen-
tury. Three years later, in 2002, the International Association of Chemical
Thermodynamics, IACT, the successor to Commission 1.2, decided to publish a collec-
tion of essays on applied chemical thermodynamic topics under the same editorship.
This new collection is aimed not only at those working in a specific area of
Chemical Thermodynamics but also at the general chemist, the prospective
researcher and those involved in funding chemical research.
The new volume is dedicated to the chemical industry and will show that thermody-
namics is not only helping us to understand the world we live in but is also helping
us to create a better world.
The topics for this volume have been chosen to include the very latest thermodynamic
development in not only the traditional areas but also in newly established areas and
potentially important new areas of chemistry. The traditional areas covered include
aspects of:
calorimetry
microcalorimetry
transport properties
crystallization
adsorption
electrolyte systems
transport fuels.
The newly established areas include:

multiphase modelling
reactive distillation
...
Vlll Preface
non-equilibrium thermodynamics
spectro-calorimetry.
Topics covered, which are considered potentially important to the chemical industry
in the future, include:
ionic liquids
0
0 new materials
ab initia quantum chemistry

0
nano particles
polymer recycling
clathrates
the economic value of applied thermodynamics.
The topics have all been written by specialists in the field and include some of the
most important names in chemical engineering, thermodynamics and industrial
applications of applied thermodynamics.
I wish to record my thanks to authors and to the publishers in producing a book of

this breadth and importance.
T.M. Letcher
Durban, January 2004
Contents
Chapter 1 Non-EquilibriumThermodynamicsfor Industry 1
S. Kjelstrup, A. R4sjorde and E. Johannessen
1 What Can Non-Equilibrium Thermodynamics Offer? 1
2 Developments and Status of NET 2
3 NET Applied to Phase Transitions 2
4 NET Equations for Distillation Columns 5
5 The Second Law Optimal Path of Operation 6
6 Second Law Optimal Distillation Columns 7
7 Second Law Optimal Chemical Reactors 8
8 Conclusions 10
Acknowledgement 10
References 10
Chapter 2 A Modelling Technique for Non-equilibrium Metallurgical

Processes Applied to the LD Converter 12
M. Modigell, A. Traebert, I? Monheim and K. Hack
1 Introduction 12
2 Process Model Development 12
3 Modelling Tool 16
4 Simulation Results 17
5 Conclusions 21
6 List of Symbols 21
References 22
Chapter 3 Multiphase Thermodynamicsof Pulp Suspensions 23

I? Koukbri, R. Pajarre and E. Rasanen
1 Introduction 23
2 Fibres in Aqueous Solution 24
3 Fibre-Ion Interactions 25
3.1 Donnan Equilibrium 25
3.2 Specific Interactions 26
4 Solid and Gaseous Phases 27
X Contents
5 General Gibbs Energy Model 27
6 Discussion 31
References 32
Chapter 4 Reactive Distillation 33

H.G. Schoenmakers and W Arlt
1 Introduction 33
2 Thermodynamics 34
2.1 Phase Equilibrium 34
2.2 Chemical Equilibrium 36
3 Technical Application of Reactive Distillation 38
3.1 Introduction 38
3.2 Process Synthesis 38
3.3 Process Design and Optimization 39
3.4 Limitations of the Methods for Synthesis
and Design, the Scale-up Problem 39
3.5 Choice of Equipment 40
4 Conclusions 41
References 42
Chapter 5 Thermodynamic Properties from Quantum Chemistry 43

S.I. Sandler
1 Introduction 43
2 Quantum Mechanical Computation of Force Fields
for Molecular Simulation 44
3 Approximate Quantum Mechanical Calculation of
Thermodynamic Properties 53
4 Conclusions 55
References 55
Chapter 6 Thermodynamics of Natural Gas Clathrate Hydrates 57

E.D. Sloan
1 Introduction 57
2 What are Natural Gas Clathrate Hydrates? 58
3 Development of Hydrate Thermodynamics 59
3.1 Discovery and Academic Interest 60
3.2 Thermodynamics Driven by Discovery in
Flowlines 60
3.3 Thermodynamics Driven by Hydrates
Outside Flowlines 62
4 Hydrate Thermodynamics 65
4.1 The Hydrate Phase Diagram 65
4.2 The Most Important Advance: The Statistical
Thermodynamic Model 67
Contents xi
5 Hydrate Challenges for the Future: Reversible and
Irreversible Thermodynamics 72
References 72
Chapter 7 Ionic Liquids in Separation Processes 76
J. Gmehling
1 Introduction 76
2 Thermal Separation Processes Using a Selective
Solvent as Separating Agent 77
2.1 Extractive Distillation 77
2.2 Liquid-Liquid Extraction 78
2.3 Absorption 79
3 Thermodynamic Fundamentals 79
3.1 Vapor-Liquid Equilibrium (VLE) 79
3.2 Liquid-Liquid Equilibrium (LLE) 80
3.3 Gas-Liquid Equilibrium (GLE) 80
4 Selection of Selective Solvents 80
5 Thermophysical Properties Required for Selective
Solvents 81
6 Ionic Liquids 82
7 Results 83
8 Conclusion 86
Acknowledgement 87
References 87
Chapter 8 Spectrocalorimetric Screening for Complex Process
Optimization 88
E Dan and J-RE. Grolier
Reaction Calorimetry: General Aspects 88
Instrument 89
Particularities of Anionic Polymerization of
Lactams in Organic Media 91
Spectro-Calorimetric Investigation Under
Isothermal Conditions 92
4.1 The Manner of Contacting the Catalytic Species 92
4.2 Effect of the Catalyst Nature 98
4.3 Stirring Level 100
Non-Isothermal Reaction Calorimetry 101
Conclusion 102
References 103
Chapter 9 Microcalorimetry for the Pharmaceutical Industry 104
A.E. Beezer
1 Introduction 104
2 Stability Determinations 105
xii Contents
3 Characterisation of Physical FodAmorphous Content 107
4 Drug/Receptor Studies by Titration Calorimetry (ITC) 108
5 Multiple Miniature Calorimeters and their Potential 108
6 Conclusions 109
References 109
Chapter 10 Isothermal Flow-Microcalorimetry:Principles and

Applications for Industry 110
M.A.A. 0 'neill
1 Introduction 110
2 Theory 111
3 Flow Calorimeter Design 111
3.1 Solution-Phase Flow Calorimeters 112
3.2 Gaseous-Phase Flow Calorimeters 113
4 Flow Calorimetric Equations 113
5 Validation of Flow-Through Microcalorimeters 115
6 A Simple Flow Calorimetric Experiment 116
6.1 Experimental Protocol 117
6.2 Determination of Thermodynamic and Kinetic
Parameters from Calorimetric Data 118
7 Further Applications of Flow Microcalorimetry 119
8 Conclusions 119
References 120
Chapter 11 Transport Properties and Industry 122
WA. Wakeham and M.J. Assael
1 Introduction 122
2 Measurements 124
2.1 Viscosity 125
2.2 Thermal Conductivity 126
2.3 Diffusion Coefficient 127
3 Theoretical Predictions 127
4 International Dimension 129
5 Databases 129
6 Future 130
References 131
Chapter 12 Micro- and Nano-particlesProduction

Using Supercritical Fluids 132
E. Reverchon and I. De Marc0
1 Introduction 132
2 Experimental Apparatus for SAS Processing 133
2.1 Experimental Procedure 133
3 SAS Fluidodynamics, Mass Transfer and
Thermodynamics 134
Contents xiii
4 Relationship between Particles Morphology and
High-Pressure VLEs 135
4.1 Experimental Evidences 135
5 System Modifications Induced by the Presence of
Solute 140
6 Conclusions 142
References 142
Chapter 13 CalorimetricMeasurements of Thermophysical

Properties for Industry 144
J-RE. Grolier and E Dan
1 Introduction 144
2 Heat Capacities 145
2.1 Heat Capacities of Gases 145
2.2 Heat Capacities of Liquids 147
3 Bulk Thermomechanical Properties 149
3.1 In Extended Ranges of T and p 149
3.2 In the Vicinity of the Critical Point 151
4 Phase Transition Thermal Properties 152
4.1 Fusion/Crystallization 152
4.2 Glass Transition Temperature 154
5 Conclusion 157
References 158
Chapter 14 Plastic Recycling 159

W Arlt
1 Introduction 159
2 Cost of Recycling 160
3 Thermodynamic Considerations 160
4 Polymer Production 160
5 Recycling Schemes 162
6 Energetic Considerations 165
References 165
Chapter 15 Industry Perspective on the Economic Value of Applied
Thermodynamics and the Unmet Needs of AspenTech
Clients 166
C-C. Chen, S. Watanasiri, 19 Mathias and V V De Leeuw
1 Economic Impact of Applied Thermodynamics 166
2 Applied Thermodynamics System 167
3 Elements of an Applied Thermodynamics System 167
3.1 Calculation Engine 168
3.2 Analysis Tools 169
3.3 Programmable Interfaces 169
3.4 Deployment Tools 169
xiv Contents
4 Industrial Applications 170

4.1 Phase Equilibrium Modeling for Nylon-6
Process Simulation 170
4.2 Quickly Screen Solvents for Organic Solids 170
5 Quantitative Cost-Benefit Assessments 171
6 Unmet Needs of AspenTech Clients 172
6.1 Estimation Methods for Pure Compound
Properties 172
6.2 Cubic Equations-of-State 173
6.3 Activity Coefficient Models 174
6.4 Modeling Electrolyte Systems 174
6.5 Modeling Polymer Systems 175
6.6 Data Packages 175
6.7 Equilibrium Calculations 176
6.8 Transport and Interfacial Properties 177
6.9 Model Deployment 177
7 Conclusion 177
References 178
Chapter 16 Thermodynamics of New Materials 180
M. Schroeder and M. Martin
1 Introduction 180
2 Ceramic Membranes for Oxygen Separation and
Catalytic Partial Oxidation of Hydrocarbons 180
2.1 Transport and Defect Thermodynamics in
Membrane Materials 181
2.2 Oxygen Permeation 184
2.3 Stability of Membrane Materials 186
2.4 Conclusion 187
3 Solid Oxide Fuel Cells 187
3.1 Fuel Cells 187
3.2 Electrolytes for Solid Oxide Fuel Cells 189
3.3 Lanthanum Gallates 191
3 Conclusion 193
References 194
Chapter 17 Thermodynamic Prediction of the Formation and
Composition Ranges of Metastable Coating
Structures in PVD Processes 197
f?J. Spencer
1 Introduction 197
2 (Ti,Al)N Coatings 198
2.1 Metastable + Stable Structural Transformation
Energies for Nitride and Carbide Phases 199
3 Al,O,-A1N Coatings 206
Contents xv
4 summary 208
References 208
Chapter 18 Thermodynamics of the Nano-Sized Particles 209
I: Tanaka, J. Lee and N. Hirai
1 Introduction 209
2 Gibbs Energy in Small Particle Systems 210
3 Calculation of Nano-Sized Phase Diagrams of
Binary Alloys 213
4 Concluding Remarks 217
References 218
Chapter 19 Thermodynamics of Electrolyte Systems of Industry 219
K. Thomsen
1 Standard States and Chemical Potentials 219
1.1 Ionic Mean Properties 220
2 Act: rity Coefficient Models 221
3 Equ librium Calculations 222
3.1 Speciation Equilibrium 222
3.2 Vapor-Liquid Equilibrium 223
3.3 Solid-Liquid Equilibrium 223
3.4 Liquid-Liquid Equilibrium 224
4 Pha ;e Diagrams 224
5 Density 226
6 Viscosity 228
References 228
Chapter 20 Thermodynamics of Crystallization 230
A.S. Teja and R.W Rousseau
1 Introduction 230
2 Pure Component Phase Behavior 230
3 Binary Solid-Liquid Equilibria 23 1
3.1 Systems that Form Solid Solutions 232
3.2 Systems that Exhibit Eutectic Behavior 233
4 Ternary Solid-Liquid Equilibria 236
4.1 Systems that Form Solid Solutions 236
4.2 Eutectic Systems 238
5 Solid-Fluid Equilibria 239
6 Kinetics of Dissolution and Crystal Growth 24 1
7 Summary 242
References 242
Chapter 21 Thermodynamics of Adsorption 243
A.L. Myers
1 Introduction 243
2 Adsorption Isotherm and Equation of State 244
xvi Contents
3 Thermodynamic Functions 246
4 Mixtures 248
5 Example 249
6 Summary 253
References 253
Chapter 22 Mesoscopic Non-equilibrium Thermodynamicsof

Polymer Crystallization 254
D. Reguera and J.M. Rubi
1 Non-Equilibrium Thermodynamics and the
Mesoscopic Level of Description 254
1.1 Meso-Structures 254
1.2 Mesoscopic Non-Equilibrium Thermodynamics
(MNET) 255
2 Polymer Crystallization 256
3 The Mesoscopic Non-Equilibrium Thermodynamics
Approach to Polymer Crystallization 257
3.1 Mesoscopic Non-Equilibrium Thermodynamics
of Activated Processes 258
3.2 Polymer Crystallization 259
4 Conclusions 261
References 261
Chapter 23 Applied Thermodynamicsfor Petroleum Fluids in the

Refining Industry 262
D. Ramjugemath and R. S h a m a
1 Introduction 262
2 Composition of Petroleum 263
3 Characterization of Petroleum and its Fractions 263
3.1 Measured Data 263
3.2 Calculated Parameters 265
3.3 Pseudo-Components Characterization 267
4 Thermo-Physical Property Prediction Methods 267
4.1 Enthalpy 267
4.2 P-V-T Relationship 269
4.3 Vapour-Liquid Equilibria (Including Vapour
Pressure) 270
5 Codes and Data Sources 27 1
6 Future Trends 27 1
References 272
Subject Index 274

Contributors xvii
Contributors
W. Arlt Professor, Dr, Technische Universitat Berlin, Thennodynamik und Thennische
Verfahrenstechnik,Building TK7, 0-10623 Berlin, Germany; Tel.: +49 30 314 22755;
e-mail: w.arlt@vt.tu-berlin.de.
MJ. Assael Professor, Dr, Chemical Engineering Department, Aristotle University,
Thessaloniki 54124, Greece; Tel.: +30 6937 158044; e-mail: assael@auth.gx
A.E. Beezer Professor, Dr, Medway Sciences, University of Greenwich at Medway, Central
Avenue, Chatham Maritime, Kent ME4 4TB; Tel.: +44 1634 883362; e-mail:
a.beezer@gre.ac.uk.
C-C. Chen Dr, Aspen Technology Inc., I0 Canal Park, Cambridge, MA 02141 USA; Tel.:
+I 617 941 1202; e-mail: chauchyun.chen@aspentech.com.
F. Dan Dr, Macromolecular Chemistry Department, “Gh. Asachi” Technical University of
lasi, 71A, Mangeron Avn., P.O. 10, P.B. 2007, 66OO-lasi,Romania; Tel.: +40 232 27 86 83/
2332; e-mil: jlorin.dan@ch.tuiasi.ro.
V.V. De Leeuw Dr, STS Associks, 16 rue Jacques Tati, 91042 Evry, France.
I. De Marco Dr, Dipartimento di Ingegneria Chimica ed Alimentare, Universita degli studi di
Salemo, Via Ponte Don Melillo, 84084, Fisciano (SA),Italy.
J. Gmehling Professor, Dr, Carl von Ossietzky Universitat Oldenburg,Fakultat V Technische
Chemie, 0-2611I Oldenburg, Germany; Tel.: +49 441 7983831; e-mail:
gmehling @ tech.chem.uni-oldenburg.de.
J-P.E. Grolier Dr, Laboratoire de Thermodynamiques des Solutions et des Polymdres,
Universite‘ Blaise Pascal, Clermont-Ferrand, 24, Avenue des Landais, 63177 Aubidre,
France; e-mail:j-pierre.grolier@ univ-bpclermont.f x
K. Hack Dr, Managing Director, GTT-Technologies, Kaiserstrasse 100, 0-52134
Herzogenrath, Germuny; Tel.: +49 2407 59533; e-mail: kh@gtt-technologies.de.
N. Hirai Dr, Department of Materials Science and Processing, Graduate School of
Engineering, Osaka University, Osaka, 565-0871, Japan.
E. Johannessen Dr, Department of Chemistry, Norwegian University of Science and
Technology, N-7491 Trondheim,Norway.
S. Kjelstrup Professor, Dr, Department of Chemistry, Norwegian University of Science and
Technology, N- 7491 Trondheim, Norway; Tel.: +47 73594179; e-mail:
Signe.Kjelstrup@nt.ntnu.no.
P. Koukkai Dr, VTT Processes, PO.Box, FIN-02044 VlT, Finland; Tel.: +358 9 456 6366;
e-mail: Pertti.Koukkuri@vtt.ji.
J. Lee Dr, Department of Materials Science and Processing, Graduate School of Engineering,
Osaka University, Osaka, 565-0871, Japan.
M. Martin Professor, Dr, Institute of Physical Chemistry,Aachen University of Technology,
Templergraben 59, 0-52056 Aachen, Germany; Tel.: +49 241 80 94712; e-mail:
martin@rwth-aachen.de.
xviii Contributors
P. Mathias Dr, Aspen Technology Inc., 10 Canal Park, Cambridge, MA 02141 USA.
M. Modigell Dr, lnstitut fur Verfahrenstechnik,RWTH Aachen, Germany.
P. Monheim Dr, SMS Demag AG, Duisburg, Germany.
A.L. Myers Professor, Dr, Chemical and Biomolecular Engineering, University of
Pennsylvania, Philadelphia, PA, I9104, USA; e-mail: amyers @ seas.upenn.edu.
M.A.A. O'neill Dr, Medway Sciences, University of Greenwich at Medway, Central
Avenue, Chatham Maritime, Kent, ME4 4TB, UK; Tel.: +44 1634 883362; e-mail:
moneiE127@tiscali.co. uk.
R. Pajarre Dr, VTT Processes, P.0. Box, FIN-02044 VI?: Finland; Tel.: +358 9 456 6366.
D. Ramjugernath Professor, Dr, School of Chemical Engineering, University of Natal,
Durban 4041 South Afr?ca; Tel.: +27 2603128; e-mail: ramjuger@nu.ac.za.
E. Rasiinen Dr, VTT Processes, €!O.Box, FIN-02044 Finland; Tel.: +358-9-456 6366.
A. Rmjorde Dr, Department of Chemistry, Norwegian University of Science and Technology,
N-7491 Trondheim, Norway.
D. Reguera Dr, Department de Fisica Fonamental, Facultat de Fisica, Universitat de
Barcelona Diagonal 647, 08028 Barcelona, Spain.
E. Reverchon Professor, Dr, Dipartimento di Ingegneria Chimica ed Alimentare, Universita
degli studi di Salerno, Via Ponte Don Melillo, 84084, Fisciano (SA), Italy; e-mail:
ereverchon@ unisa.it.
R.W. Rousseau Dr, School of Chemical & Biomolecular Engineering, Georgia Institute of
Technology,Atlanta, GA 30332-0100, USA.
J.M. Rubi Professor, Dr, Department of Chemistry and Biochemistry, UCLA 607 Charles E.
Young East Drive, Los Angeles, CA 90095-1569, USA; e-mail: mrubi@fi.ub.es.
S.I. Sandler Professor, Dr, Center for Molecular and Engineering Thermodynamics,
Department of Chemical Engineering, University of Delaware, Newark, DE I971 6 USA;
Tel.: +I 302-831-2945; e-mail: sandler@udel.edu.
H.G. Schoenmakers Dr, BASF AG, Process Engineering, Building L540 0-67056
Ludwigshafen, Germany; Tel.: +49 621 60 56047; e-mail: hartmut.schoenmakers@basf-
ag.de.
M. Schroeder Dr, Institute of Physical Chemistry Aachen University of Technology
Templergraben 59, 0-52056 Aachen, G e m n y .
R. Sharma Dr, School of Chemical Engineering, University of Natal, Durban, 4041, South
Africa.
E.D. Sloan Professor, Dr, Gaylord and Phyllis Weaver Distinguished Professol; Center for
Hydrate Research, Colorado School of Mines, Golden, Colorado 80401, USA; Tel.:
+I 303 273 3723; e-mail: esloan@gashydrate.mines.edu.
PJ. Spencer Dr, The Spencer Group, P.O.Box 393, Trumansburg, NY 14886, USA; e-mail:
ps @spensergroupintl.com.
T. Tanaka Professor, Dr, Department of Materials Science and Processing, Graduate School
of Engineering, Osaka University, 2-1 Yamadaoka, Suita, Osaka 565-0871, Japan; Tel.:
+81 6 6879 7504; e-mail: tanaka@mat.eng.osaka-u.ac.jp.
A. Traebert Dr, Institut fur Verfahrenstechnik,RWTH Aachen, G e m n y .
A.S. Teja Professor, Dr, School of Chemical & Biomolecular Engineering, Georgia Institute
of Technology, Atlanta, GA 30332-0100, USA; Tel: +I 404 894 3098; e-mail:
amyn.teja @ chbe.gatech.edu.
K. Thomsen Associate Professor Dr, Department of Chemical Engineering, Technical
University of Denmark, Building 229, DK 2800 Kongens Lyngby, Denmark. Tel.:
+45 4525 2860, e-mail: kth @ kt.dtu.dk.
W.A. Wakeham Professor, Dr, University of Southampton, Highfield, Southampton SO1 7
IBJ, UK; Tel.: +44 2380 592901; e-mail: vice-chancellor@soton.ac.uk.
S. Watanasini Dr, Aspen Technology Inc., I0 Canal Park, Cambridge, MA 02141 USA; Tel.:
+ I 617 949 1202.
CHAPTER 1
Non=Equilibrium
Thermodynamicsfor Zndustry
SIGNE KJELSTRUP, AUDUN RDSJORDE
AND EIVIND JOHANNESSEN
1 What Can Non-Equilibrium Thermodynamics Offer?

Non-equilibrium thermodynamics (NET) offers a systematic way to derive the local
entropy production rate, 0, of a system. The total entropy production rate is the inte-
gral of the local entropy production rate over the volume, V, of the system, but, in a
stationary state, it is also equal to the entropy flux out, J:, minus the entropy flux
into the system, J;:
The entropy flux difference and the integral over CJ can be calculated independently,
and they must give the same answer. The entropy production rate governs the trans-
port processes that take place in the system. We have
o = z JiXi > 0
1
where Ji and Xiare conjugate flux-force pairs. Each flux is a linear combination of
all forces:
J i = C L r- XJ I. (3)
i
This means that NET gives flux equations in agreement with the second law of ther-
modynamics, and that the theory offers a possibility, through Equation (l), to check
for consistency in the models that are used.
The usefulness of NET in describing industrial problems has been questioned,
because these problems are frequently non-linear. It is then important to know that the
flux-force relations in Equation (3) also describe non-linear phenomena. The phe-
nomenological coefficients Lii can, for instance, be functions of the state variables. By
2 Chapter I
including internal variables in the thermodynamic description, one can extend the
application of NET to activated processes; see Chapter 2. For this reason, NET appears
today as a non-linear and versatile theory that applies to many practical condition^.'-^
It is a misunderstanding that flux equations need to be linear on the global level.lo
The total entropy production rate times the temperature of the environment is
equal to the exergy destruction rate in a process. Processes with small losses in
exergy have a high second law efficiency. A high second law efficiency, or exergy
efficiency, is seldom a specific aim in process design. An increasing worldwide con-
cern with CO, emission may change this. Multiobjective optimisation, with small
entropy production as one target, may then be interesting in chemical engineering
design.
2 Developments and Status of NET

Non-equilibrium thermodynamics was founded by Onsager.11J2The theory was fur-
ther elaborated by de Groot and Mazuri3and Prigogine.14 The theory is based on the
hypothesis of local equilibrium: a volume element in a non-equilibrium system is in
local equilibrium when the normal thermodynamic relations apply to the element.
Evidence is emerging that show that many systems of interest in the process indus-
try are in local equilibrium by this Onsager prescribed that each flux be
connected to its conjugate force via the extensive variable that defines the
Onsager assumed that the variables and the rate laws were the same on the macro-
scopic and the microscopic level; this is the so-called regression hypothesis. Also
using the assumption of microscopic reversibility, he proved the reciprocal relations:
These assumptions restrict the validity of NET, but as stated above, they have a
wide range of validity. It has long been known that the Navier-Stokes equations are
contained in NET.13 More recently, NET has been extended to deal with transport
across surfaces: quantum mechanical system^,^ and mesoscopic systems;a see
Chapter 2. We have chosen to illustrate NET with cases of transport through surfaces
in the following sections.
3 NET Applied to Phase Transitions

Phase transitions are central in many industrial problems, for instance, in distillation,
absorption, condensation, manufacture of liquid natural gas, and multiphase flow.
We shall use the liquidhapour transition to illustrate the basic hypotheses and use-
fulness of NET for surfaces, a relatively new application. For applications to homo-
geneous phases, we refer the reader to the basic literature; see Section 2.
R@sjordeet aZ? studied the phase transition in a pure fluid using non-equilibrium
molecular dynamics simulations (NEMD). The NEMD method solves Newton's
equations of motion for several thousand particles in an imaginary box; see
Hafskjold16 for a review. The particles interacted with a Lennard-Jones-type pair
Non-Equilibrium Thermodynamicsfor Industry 3
potential. The phase diagram and the surface tension as a function of temperature of
the system were first determined.2The ends of the box were then thermostatted to give
an enormous temperature gradient: for argon-like particles, the gradient was lo8K-'
m. The temperatures of the ends were also chosen so that a liquid-vapour interface
established itself. The position of the surface was then determined according to Figure
1. The surface started where the equation of state for the vapour was no longer valid
and ended where the density in the liquid phase started to vary in a linear manner.
It was established that the surface was in local equilibrium: the surface tension
calculated for the system with a gradient was indeed the same as for the system in
global equilibrium at a given temperature; see Figure 2. The fluxes for heat and mass
across the surface can be written as
J=-Ip,A - -I,,Rln- P
with 1, being the surface transfer coefficients. The coefficients obey Equation (4).
Equation (6) indicates that evaporation or condensation takes place when there is a
thermal force (a difference in inverse temperature) and/or when there is a chemical
potential difference (the pressure of the vapour differs from the saturation pressure
at the temperature of the liquid). Equation (5) can be rewritten as
Here, A is the stationary-state thermal conductivity and q* is the heat of transfer. The
results of the simulations for A and q* are shown as a function of surface tension in
Figures 3 and 4. The transfer coefficients decrease in magnitude as we move from the
critical point in the phase diagram (zero surface tension) to the triple point (maximum
surface tension). It is reasonable that the surface becomes more resistive at high surface
-0 10 20 30 40 50 60 70
Figure 1 Density variation across a non-isotheml NEMD cell with Lennard-Jones spline
particles. The extension of the su@ace is indicated by vertical bars
4 Chapter I
2 -
1 -
Figure 2 Surface tension as afunction of surface temperature in equilibrium (whole line) and
in a temperature gradient (points)for Lennard-Jones spline particles
.,-loo0
-1500
t . 0 Kinetic theory
0 2 4 6
y/ N/m
Figure 3 Heat of transfer for the Lennard-Jones spline surface and from kinetic theory
according to R#sjorde et aL9
tensions. In the same figures, the results are compared to computations from kinetic the-
ory. Kinetic theory predicts the properties of hard spheres and dilute gases. Indeed, we
observe from the figures that the closer one comes to the triple point, the closer the
agreement between theory and simulations. The heat of transfer can be significant; its
minimum value here is 20%of the heat of evaporation. The last term is frequently neg-
lected in thermal modelling of interface transport. In this case, it was found that the flux
equations were linear in the forces, even for large gradients. It was thus not necessary
to invoke a nonlinear dependence on the forces, as is done in statistical rate tl1e0ry.l~
Little systematic information is available on surface transfer coefficients. There
are indications that the surface transfer coefficients may depend on the range of par-
ticle interaction.18The entropy production rate per m2 of area can be high for the sur-
face, implying that the surface conductivity per m is smaller than the conductivity of
the homogeneous phases. But since the homogeneous phases are thicker than the
surface, these phases will still give the dominating total contribution to the entropy
production of the system.*9
Non-Equilibrium Thermodynamics for Industry 5
OCP 0 NEMD simulations

6 - 0 Kinetic theory
cu
Y
4
E 4 -
0
0 .
$
0
8
\
i 3 -
v)
o0 8
x
2-
't
0 2 4 6 a
y / N/m
Figure 4 Thermal conductivity of a knnard-Jones spline surface and from kinetic theory
according to R#sjorde et aL9
4 NET Equations for Distillation Columns

Distillation columns are made to separate at least two different components. There
are several different types of columns. The assumption of equilibrium between the
liquid and vapour that leave each tray is still in common use to model tray distilla-
tion. The work of Krishna and Wesselingh20shows, however, that non-equilibrium
models give results that are very different from those obtained with the equilibrium
assumption.
A description consistent with NET is the Maxwell-Stefan equations.20 In the
application of Maxwell-Stefans equations, one frequently neglects the heat of trans-
fer, q*. This assumption may be good for gases. It is not so good for liquid mix-
tures.16According to Section 3, it is likely that the heat of transfer plays an important
role in flux equations of interface transport. Olivier21showed that failure to include
the heat of transfer leads to an error of up to 20%in the heat flux calculated for some
typical phase transition conditions.
The entropy production rate and the complete set of equations that follows can
be most conveniently written for the liquid film, the interface, and the vapour film
in series.22 Film layers of thicknesses aL and 6" in the liquid and the vapour
are illustrated in Figure 5. With constant fluxes (in stationary states), the integration
is easy. The approach was called the integrated interface approach.22 For the
three layers, the integrated overall force is the sum of the integrated force across
each layer:
6 Chapter 1
Figure 5 Variationin chemicalpotential and temperature across the liquid luyel; the intelface,
and the vapour layer during evaporatiodcondensation in a two-component system
With constant fluxes of energy and mass through the interface and adjacent films,
the overall resistance coefficients were defined through the parentheses in the fol-
lowing equations:
With these relations, the combined films and interface is regarded as an “effective
interface”. There is no need to assume phase equilibrium between liquid and vapour.
The entropy production rate can alternatively be expressed by the measurable heat
flux in the vapour and fluxes of mass.18p22This set of flux equations was used to
explain the entropy production in tray distillation columns. However, it has not yet
been used for predictive purposes. Much work remains to be done to include these
equations in a software that is useful for industrial purposes.
5 The Second Law Optimal Path of Operation

A process, natural or industrial, follows a path between states. In industry, this is
called “the path of operation”. PrigogineI4investigated the dissipation of energy in
a system out of equilibrium, and found that the entropy production rate was at a min-
imum in the stationary state close to global equilibrium. Also S ~ h e c h t e rgave~~ a
variational formulation of the problem by solving the partial differential equations
that resulted from substituting the flux equations into the conservation equations,
and confirmed that the entropy production was minimum in the stationary state, pro-
vided the phenomenological coefficients were constant. This may have strengthened
the belief that processes need to be globally l i n e d for NET to be useful, an assump-
tion that would limit the applicability of the theory seriously.
Non-Equilibrium Thermodynamicsfor Industry 7
Rather than asking for the general nature of a system that dissipates energy at the
stationary state, like Prigogine and Schechter did, we find it more appropriate, at
least for industrial purposes, to ask about the nature of the operating path, given cer-
tain constraints. A typical example would be: how can we minimise the entropy pro-
duction of a system while maintaining a given production (or heat exchange)? The
solution to this minimisation problem can be found for conditions far from global
equilibrium, and is different from the solution found by Prigogine14or S ~ h e c h t e r . ~ ~
The question of constrained optimisation is answered in a standard manner by
Euler-Lagrange optimisations. By formulating the problem with optimal control
theory, Johannessen and K j e l ~ t r u explained
p~~ that the “Hamiltonian” of the problem
was constant in a study of chemical reactors. The total entropy production for a plug
flow reactor was written as a function of a position-dependent state variable vector
x(z) and the control variable u(z):
The Hamiltonian for the problem was constructed as
The solution to the problem is given by 2(rn+2) differential equations for the tern-
perature, pressure, degrees of conversion, and for Lagrange multipliers (T, P,ki, ;iT,
Ap, and Ak), with partial derivatives of H , where rn is the number of reactions
between the component^.^^ The constant Hamiltonian of this problem was reduced
to a solution with constant entropy production, 0,in the case of a heat exchange
process. Using NET, it was also found that this solution was approximated by a solu-
tion with constant driving force, Xi.25 How to realise this in practice, remains to be
solved.
6 Second Law Optimal Distillation Columns

For an adiabatic distillation column separating a binary mixture, the path of opera-
tion is shown through a McCabe-Thiele diagram in Figure 6. The compositions in
the liquid and vapour streams at different trays are shown. Linear operating lines
result when the heats of vaporisation of the two components are equal. For a diabatic
distillation column,26however, the operating lines are curved and tend to follow the
equilibrium line, as shown in the right part of Figure 6. This kind of operation is pos-
sible if heat is added or withdrawn on each tray. As a consequence, the entropy pro-
duction rate in the column plus heat exchangers decreases.
A reduction of 38% in the entropy production rate was obtained for a binary sep-
aration of benzene and toluene in an equimolar mixture. The heat added on each tray
in the optimal diabatic column and the corresponding vapour flows are shown in
Figure 7.
The results for the corresponding adiabatic column are also shown for compari-
son. The most important trays for distributed heating were the trays closest to the
reboiler and condenser.
8 Chapter I
X 1
Figure 6 Examples of operating lines of an adiabatic and a diabatic column
Figure 7 The vapourjlow (mol-'s) and heat exchanger load (kW)in an optimised diabatic
column and an adiabatic column. Both columns separate an equimolar mixture of
toluene and benzene
A new design that takes the varying vapour flows into account was proposed.27The
effect of the changing hydrodynamic conditions has not yet been explored. It is still
too early to conclude on the precise outcome of these optinisation studies, since the
assumption of equilibrium on each tray was also used in the model. Progress in the
methods of Section 4 may lead to improvements in the future. It is documented that
the lost work can be reduced, but the increased investment costs are not yet clarified.
Nevertheless, it is important to understand the thermodynamic conditions for optimal
performance, independent of technical-economic considerations.
7 Second Law Optimal Chemical Reactors

The entropy production of a sulfur dioxide oxidation (exothermic) reactor with heat
exchangers was minimised in two different case^.^^,^^ Case 1 was a four-bed reactor
with intermediate heat exchangers of a given total area,2gsee Figure 8. The entropy
production rate was calculated from the entropy balance over the system. All inlet
and outlet flow conditions were kept constant, except the pressure at the outlet. The
problem was to find the inlet and outlet temperatures of the intermediate heat
exchangers. These were found, and shown to give a saving of 16.7%in lost exergy.
Case 2 was a typical plug flow reactor with temperature profiles as shown in
Figure 9.29The entropy production rate was calculated from NET. The coolant
temperature profile that gave the minimum entropy production rate was determined.
Inlet and outlet conditions were kept constant, and so was the area of heat exchange.
The bold lines in Figure 10 represent a case with 11% smaller exergy loss. In the
next case, three constant boiling liquids were used as coolant. The length and tem-
perature of the heat exchangers were set free to vary. The result is indicated by the
grey lines in Figure 10. The lengths of the coolers are given by the vertical lines in
the figure, and their temperature is shown on the ordinate axis. The lost work was
now reduced by about 7%.29
Product A T5
Figure 8 The four-bed reactor with five intermediate heat exchangers according to de
Koeijer et a1.28
-T
_ _ _ T,
I
I
900
600I
0 1 2 3 4 5 6
Position/m
Figure 9 Temperature of the reaction mixture, I: and of the coolant, Ta, in the plug flow reac-
tor for sulfur dioxide oxidation according to Johannessen and K j e l ~ t r u p ~ ~
10 Chapter 1
1000 5
-T
___ Ta
600 I
0 1 2 3 4 5 6
Positiodm
Figure 10 Temperatures T and T, in the reactor with minimum entropy production. The cases
of a continuous and a discrete coolant proBle are shown according to Johannessen
and K j e l ~ t r q ? ~
8 Conclusions
By referring to a few selected examples, we have discussed the following:
Entropy production is important for a proper definition of fluxes and forces in

non-equilibrium systems that are relevant to the industry. A higher accuracy in
the fluxes can be obtained using NET.
NET can be used to describe transport of heat and mass in the presence of sub-
stantial gradients in systems far from global equilibrium.
Transport of heat and mass across interfaces can be described without using
assumptions of equilibrium at the phase boundary.
Expressions for the entropy production rate can serve as meaningful object
functions in optimisations. NET offers an insight into the nature of the solution
by characterising the system's H , 0,or Xi.
These properties of NET may prove useful for modelling and in the design of
more energy-efficient chemical processes in the future.
Acknowledgement
The Research Council of Norway and the Statoil Vista Program are thanked for
financial support.
References
1. B. Hafskjold and S. Kjelstrup Ratkje, J. Stat. Phys., 1995,78,463.
2 . A. Rosjorde, D. W. Fossmo, S. Kjelstrup, D. Bedeaux and B. Hafskjold, J. Colloid Inter$
Sci., 2000, 232, 178.
3. S. Kjelstrup and D. Bedeaux, Elements of Irreversible Thermodynamics for Engineers,
International Centre of Thermodynamics, ITU, Istanbul, Turkey, 200 1.
4. D. Bedeaux, Adv. Chem. Phys., 1986,64,47.
5. D. Bedeaux and P.Mazw, Physica A, 2001,298,81.
6. D. Reguera and J.M.Rubi, Phys. Rev. E, 2001,64,61106.
7. A. Perez-Madrid, J. M. Rubi and P. Mazw,Physica A, 1994,212,231.
8. J. M. G. Villar and J. M. Rubi, PNAS, 2001,98, 11081.
9. A. Rbsjorde, D. W.Fossmo, S. Kjelstrup, D. Bedeaux and B. Hafskjold, J. Colloid Inte$
Sci., 2001,240, 355.
10. Y.Demirel, Nonequilibrium Thermodynamics. Transport and Rate Processes in Physical
and Biological Systems, Elsevier, Amsterdam, 2002.
1 1. L. Onsager, Phys. Rev., 193 1,37,405.
12. L. Onsager, Phys. Rev., 1931,38,2279.
13. S. R. de Groot and P. Mazur, Non-Equilibrium Thermodynamics, Dover, London, 1984.
14. I. Prigogine, Thermodynamics of Zrreversible Processes, Charles C.Thomas, Springfield,
1955.
15. B. D. Coleman and C. Truesdell, J. Chem. Phys., 1960,53,28.
16. B. Hafskjold, in Thermal Nonequilibrium Phenomena in Fluid Mechanics, W. Kohler and
S. Wiegand (eds), Springer, Berlin, 2002.
17. G. Fang and C. A. Ward, Phys. Rev. E, 1999,59,441.
18. S. Kjelstrup, T. Tsuruta and D. Bedeaux, J. Colloid Znte@ Sci., 2003, 256,451.
19. G. M. de Koeijer and S. Kjelstrup, Chem. Eng. Sci., 2003,58, 1147.
20. R. Krishna and J. Wesselingh, Chem. Eng. Sci., 1997,52, 861.
21. M.-L. Olivier, Proceedings of ECOS 2003, Copenhagen, Denmark, 2003.
22. D. Bedeaux and S. Kjelstrup, Chem. Eng. Sci., 2003,59, 109.
23. R. S. Schechter, The Variational Method in Engineering, McGraw-Hill, New York, 1967.
24. E. Johannessen and S. Kjelstrup, Proceedings of ECOS 2002, Berlin, Germany, 2002.
25. E. Johannessen, L. Nummedal and S. Kjelstrup, J. Heat Mass Trans$, 2002,45,2649.
26. R. Rivero, Energy, 2001,26,561.
27. G. M. de Koeijer, Energy efficient operation of distillation columns and a reactor apply-
ing irreversible thermodynamics, Ph.D. Thesis, Department of Chemistry, Norwegian
University of Science and Technology, Trondheim, 2002.
28. G. M. de Koeijer, E. Johannessen and S. Kjelstrup, Energy, 2004,29, 525.
29. E. Johannessen and S. Kjelstrup, Energy, 2004, in print.
CHAPTER 2
A Modelling Technique for

Non-equilibrium Metallurgical
Processes Applied to the LD
Converter
M. MODIGELL, A. TRAEBERT, P. MONHEIM AND K. HACK
1 Introduction
In the LD converter process, pure oxygen is blown on a molten iron bath for refin-
ing purposes. Elements dissolved in the molten iron, e.g. C, Si, Mn, P, etc., but also
part of the molten iron, are oxidised. They either form a slag phase covering the hot
metal or, in the case of C, gas bubbles containing CO and CO,. Several reaction
zones can be identified in Figure 1. In the hot spot, the oxygen directly reacts with
iron and dissolved elements. Due to the impact of the oxygen jet, iron droplets are
dispersed in the slag phase as well as slag droplets in the metal bath. The metul-slug
dispersion is mixed further by CO and CO, bubbles and serves as the main reaction
zone. A third zone contains the hot metal that is not dispersed in the slag, but forms
the bath underneath. Droplets from the dispersion fall back into this bath.
2 Process Model Development

The task of developing a suitable converter model will be discussed based on the
decarburisation reaction of the iron melt. Obviously, the rate of the decarburisation
reaction depends on the reaction rate of
and the transport conditions in the converter. Thereby, these are determined by oxy-
gen blowing conditions and CO formation. On the other hand, CO formation is influ-
enced by the decarburisation reactions.
A Modelling Technique for Non-equilibrium Metallurgical Processes 13
Slab
\
Figure 1 LD converterprocess
Intermediary formation of FeO in the hot spot is the main oxygen source for the
decarburisation. FeO is dispersed in the slag and the metal bath as well as Fe-C
droplets, which are accelerated by the oxygen jet. The subsequent reaction of FeO
with carbon dissolved in iron occurs in both phases. In the slag phase, dissolved
FeO reacts with dispersed Fe-C droplets, and in the metal bath FeO droplets form
the disperse phase. For the decarburisation, several reaction routes can be formu-
lated:
(FeO) +[C]+ [Fe]+ { CO} (2)
(FeO) + { CO} + [Fel+ { CO,} (3)
Reaction (2) is kinetically limited.' The time for the complete reduction of an
FeO droplet with a diameter of 1 mm (equalling 4.7X1Op5 mol) in an Fe-C melt
is between 30 and 175 s. The same amount of FeO reacts with pure CO according
to reaction (3) in only 0.2 s. The consecutive reaction (4) takes place in only
2x10-3 s.
From these data, it can be concluded that the direct reaction between FeO and C
contributes little to the decarburisation due to its kinetic limitation. However, reac-
tions (3) and (4) can only occur spontaneously when CO, bubbles come in contact
with Fe droplets in the slag phase or CO bubbles with FeO droplets in the metal
phase, respectively. The probability for such contacts depends on the transport con-
ditions in the slag and metal bath, respectively.
14 Chapter 2
In principle, the conditions close to the phase boundary of the droplets can be
modelled by transport and reaction equations. In a simple two-layer model, a trans-
port equation according to Fick’s law can be stated for every component x:
nx=Kx, elTA(cx, metal-'^, phase boundary) (5)
The transport coefficients Kx,effare functions of the macroscopic and the local micro-
scopic fluid movement, which is caused by the stirring of the gas bubbles.
Additionally, the increase of the diffusion boundary layer during the reaction due to
the formation of a pure Fe phase has to be taken into consideration for all three reac-
tion routes (2)-(4). For the reaction rate, the following holds:
However, there is no reliable information on transport coefficients, reaction rate con-

stants and concentration ratios that are needed for solving Equations (5) and (6).
Hence, a process model based on these equations is not promising. Instead, the
process will be modelled on a less detailed level using a cell model, which is based
on the concept of local equilibrium.
As shown in Figure 2, according to the main reaction zones, the converter is
divided into three sections. These are treated assuming complete (local) equilibrium:
the hot spot, where the reaction between oxygen and iron melt takes place, the
metal-slag zone, where the conversion of the FeO from the hot spot with melt
droplets takes place, and the metal bath. The fourth zone, the slag, as shown in
Figure 2, is needed for the temporal discretisation of the model. It is assumed that
+ Bath reactor
Figure 2 Cell model for the LD converter process

A Modelling Techniquefor Non-equilibrium Metallurgical Processes 15
the total amount of FeO formed in the hot spot is transported to the metal-slag zone,
where it reacts with carbon-containing iron melt and already available slag. This
assumption is reasonable as reactions taking place in the bath zone can be transferred
to the metal-slag zone.
The individual ideal reaction zones are interlinked by material streams in a way
that a circular flow through these zones results. This flow models the circulation of
the material in the converter and the stream of metal droplets through the slag phase
as they have been observed in experimental studies. With a given oxygen blow rate
and lance height, the circulation rate between bath reaction zone and hot spot reac-
tion zone can be defined. This also determines the mass flow between the assumed
reaction zones, for which a boundary condition can be stated. Assuming the experi-
mentally proven full conversion of oxygen in the hot spot leads to
mFe, Hot Spot
2nO25 (7)
MFe
Between the bath and the metal-slag zone, mass transfer of Fe droplets takes place.
During a major part of the process, the decarburisation reaction is limited by the oxy-
gen supply. After reaching the critical point, it is limited by the carbon transport to
the reaction zone. Therefore, at the critical point, we have
Nv m ~ eM
. etal-slag
2n0 5 -co=hVco= CO
2 t’ P Fe
These mass flows are a function of the oxygen blow rate and conditions as well as
gas production, which are constant over the main process time. Towards the end of
the process, gas formation depends on carbon concentration and, therefore, the mass
transfer conditions change.
As the kinetic and transport limitations of the process are modelled by the interlink
of the reaction zones, the zones themselves can be treated assuming thermochemical
equilibrium. Although derived with regard to the decarburisation reaction, the present
model is also valid for other chemical components with similar behaviour present in
hot metal and slag. Processes determined by other kinetic phenomena, such as the
melting of scrap and the dissolution of lime, need to be modelled separately.
The dissolution of fluxes, especially lime, is limited by the formation of a dical-
cium silicate layer around the particles for a major part of the process. This behav-
iour is mainly dependent on the composition and temperature of the slag and mixing
between metal and slag.2 A dissolution function was incorporated into the model
regulating lime participation in reactions leading to the formation of slag. With this
function, the heating up of the lime prior to its dissolution and the withdrawal of
energy from bath and slag due to this heating up are also treated.
A similar problem exists for the addition of steel scrap. Scrap melting not only
depends on temperature but also on carbon concentration. Diffusive carbon transport
to the scrap/metal bath interface is therefore an important factor for melting behav-
iour. A function that treats the melting as well as the heating up of the scrap prior to
melting was implemented.
Not only material streams to and from ideal reaction zones are defined in the
model, but also energy streams. Energy losses are calculated and appropriate
16 Chapter 2
amounts of enthalpy are withdrawn from the ideal reaction zones throughout the
simulation. Thus, radiation losses from the converter mouth as well as radiation and
convection losses through the converter walls are considered. Furthermore, enthalpy
losses due to the hot off-gas stream are directly taken into account in the reaction
zone model as the hot off-gas stream is removed throughout the process.
A temporal discretisation of the described model is needed for simulation pur-
poses. According to a chosen time increment At and the determined mass flows
between the ideal reaction zones, in every time step the mass exchange between the
reaction zones and the reaction progress is calculated. Hence, a step-by-step trans-
port and conversion of different mass streams results.
3 Modelling Tool
The process modelling tool SimuSage is a combination of the thermochemical pro-
grammers library ChemApp3 and a library of additional graphical components for
Borland’s programming environment Delphi. Initially developed by Mannesmann
Demag Metallurgy as an in-house tool under the name ProMoSys, it is now co-
developed by GTT-Technologies and SMS-DEMAG under the new name
SimuSage. It consists of flow-sheeting components, which add to Borland Delphi’s
powerful programming language the capability to easily generate stand-alone flow-
sheeting models with a sound thermodynamic basis. The elements of the reaction
zone model shown in Figure 2, for example material streams and equilibrium reac-
tors, are readily available in SimuSage as graphic components. This has made it
possible to set up the converter model and to test modifications and additions to the
model very quickly.
As indicated above, the ChemApp programmers library3 (in the form of a
Dynamic Link Library, DLL) has been used to form the thermodynamic backbone
of SimuSage. The general concept is represented in Figure 3. The ChemApp library
permits the easy use of full complex equilibrium calculations within a software by
way of a set of interface routines. These interface routines are used to
define the conditions for a “local” equilibrium calculation,

execute the equilibrium calculation and
extract information from the calculated equilibrium state (such as phase
amounts or phase internal concentrations) that is needed for the process model
to proceed.
The equilibrium calculations in ChemApp are performed by the same Gibbs energy
minimisation code as in the well-known interactive software ChemSage,4 now
F a ~ t S a g eand
, ~ are thus of proven reliability.
The thermodynamic data used in the present modelling have been taken from the
extensive file store of GTT-Technologies. One hundred and nine phases and alto-
gether 202 species have been included in the calculations. The gas phase (60 species)
has been treated as ideal, while the liquid Fe phase (dilute solution approach, 14
species) and the liquid slag (Gaye-Kapoor-Frohberg model, eight species) have been
treated as non-ideal chemical solutions.
SimuSage program
Process parameters Results
I
I
I I
I
I
I
I
I
E E
l I
I
I
I
I
I
I
I
I
I
I
I
I
Data handling and phase
I
I
equilibrium calculation module
I
I
I
I
Figure 3 Integration of ChemApp into SimuSage
4 Simulation Results
For the simulation calculations, the following process conditions are assumed
according to Asai and Muchi?
150 t converter
oxygen blow rate of 10.87 kg O,/s for 0 to the 15th minute, and 7.99 kg O,/s
from the 15th minute
5 t scrap input
4 10 t CaO input
From the experimental results6 shown in Figure 4 and the corresponding process con-
ditions stated above, model parameters can be calculated. The lower limit for
theexchange ratio between the bath and the hot spot reaction zone defined as
mHotSpot/mtotdis calculated using Equation (7); it amounts to 0.023. The exact value for
this exchange ratio is derived by comparing the measured and calculated decarburisa-
tion results. Other reactions, for example of Si, and the temperature development also
have to be considered. From the comparison of experimental and calculated results, the
exchange ratio between the bath and hot spot reaction zone is found to be 0.04.
In Figure 4, the critical point, where the change from a constant decarburisation rate
to a rate controlled by carbon concentration takes place, is found to be at 13.5 min at
a carbon concentration of 0.006. From this, the exchange ratio between the bath and
the metal-slag reaction zone defined as mMetalSlag/mtotal is calculated to be 0.7.
The results of the calculation with these parameters are shown in Figure 5. The
decarburisation reaction, the critical point and carbon content are modelled correctly,
18 Chapter 2
0.045
0.04
5 0.035
0 2 4 6 8 10 12 14 16 18
Time [min]
Figure 4 Carbon content in hot metal
0.0035
0.003
C-content of hot metal
Figure 5 Decarburisation rate
as can be seen from the calculation of the decarburisation rate. With the calculated
exchange ratio between bath and metal-slag reaction zone, a critical carbon content
of 0.006 results, as expected.
As discussed above, Equation (8) is valid for process conditions up to the critical
point. The decrease in decarburisation rate from the critical point onwards results in
a decrease in CO production and, therefore, less stirring is caused by the gas bub-
bles. Hence, mass transfer in the form of metal droplets in the slag decreases.
In Figure 6, this effect is taken into account and modelled by a decrease of the
bath/metal-slag reaction zone exchange ratio from the critical point onwards. After
reaching the critical point, the exchange ratio is set to 0.35. Obviously, the decrease in
decarburisation rate can be modelled quite well with this set of parameters. The calcu-
lated carbon content of the metal phase shows good agreement with measured values.6
At the beginning of the process, the oxidation of Si from the hot spot is limited by
the transport of Si to the phase boundary. Hence, decarburisation starts, although
0.045 0.0005
0.04 0.00045
-A- C-content calculated 0.0004
0.035
-0- Si-content calculated
0.00035
0.03 4 Mn-content calculated
Y 0.0003 2
g 0.025
0.00025 ’g
4I
0.02
0.0002 0
5 0.015 0.00015
w 0.01 o.Ooo1
0.005 O.ooOo5
0 0
0 2 4 6 8 10 12 14 16 18
Time [min]
Figure 6 Components in hot metal
-
Simulation Results Components in Slag Phase
25000 1
-
4-CaO calculated
Si02 calculated
4 MnO calculated
0
5 10 15
Time [min]
Figure 7 Components in slag phase
normally Si is oxidised preferentially to carbon. The limitation of transport to the

phase boundary is modelled by a limited mass flow through the hot spot reaction
zone and, therefore, by a limited amount of Si in this reaction zone. The resulting
excess oxygen reacts to form CO and FeO-rich slag. This slag facilitates the disso-
lution of lime and forms the main reaction zone for further reactions. The oxidation
of Si, therefore, not only takes place in the hot spot but also in the metal-slag reac-
tion zone. The amount of Si calculated in the metal phase shows good agreement
with measured value^.^ Additionally, the content of Mn and 0 in the metal phase was
calculated. Both components show qualitatively good agreement with measured val-
u e ~Mn. ~ is oxidised towards the end of the process, and the oxygen content of the
metal phase increases.
Figure 7 shows the composition of the slag phase, which shows qualitatively good
agreement with measured values7 The FeO content rises towards the end of the
process, resulting in iron losses.
20 Chapter 2
The calculated gas production is shown in Figure 8. The shift of the main reaction
zone from the hot spot to the metal-slag dispersion in the course of the process is
modelled well, as can be seen from the gas production decreasing in the hot spot
while increasing in the metal-slag reaction zone. The decreasing gas production at
the end of the process is caused by the change of the oxygen blow rate and the slower
decarburisation.
As discussed above, the ideal reaction zones in the cell model are not only cou-
pled by defined mass transfer but also by heat transfer. Energy losses from the con-
verter are calculated at every time step and enthalpy is withdrawn from various
reaction zones, mainly from the metal-slag reaction zone. In Figure 9, the evolution
of the temperature in the hot spot and of the mean temperature of the hot metal bath
1600 I I Hot spot

1400
1200
800
600
400
200
0
0 1.5 3 4.5 6 7.5 9 10.5 12 13.5 15 16.5 18
Time [min]
Figure 8 Gas production
2600
2400
-
E
2200
-5 2000
2
b
8 1800
1600
1400
1200
0 5 10 15
Time [min]
Figure 9 Temperature evolution

A Modelling Techniquefor Non-equilibrium Metallurgical Processes 21
are shown. The temperature in the hot spot is well described. Measurements give val-
ues of 2300-2400 "Cfor the first 8 min of the process.8 The mean temperature shows
very good quantitative agreement with the experimenk6
5 Conclusions
The aim of the research presented in this paper is the evaluation of a technique for
modelling metallurgical processes. In the long term, the process models developed
using this technique will be used to improve process design. Special focus is set on
modelling non-equilibrium phenomena that are caused e.g. by transport limitations
or dissolution processes. The objective is to model the process with a relatively sim-
ple model structure and only a few model-specific parameters.
The LD converter process was modelled according to this intention. Based on
experimental results for the decarburisation reaction and possible reaction routes, a
simple cell model that well describes the transport limitations of the decarburisation
was developed. Simple relationships that define the mass transfer between the cells
(reaction zones) and thus the model parameters were deduced. Only a slight further
adjustment was needed. Energy losses from the process are calculated throughout
the simulation. The resulting temperatures of the hot spot show good agreement with
real process data. With respect to the reactions, the model not only describes the
decarburisation but also the behaviour of other elements dissolved in metal and slag.
The temperature of the metal bath could also be reproduced by the model.
Altogether, these facts show that, although only a simple model structure was
employed, the physical and chemical conditions in the converter process are handled
correctly.
6 List of Symbols
A area of phase boundary
Cx, Ci concentration of component x, i
CO critical carbon concentration
kX
reaction rate constant
Kx,efi transport coefficient
- -
mtotal - mHot Spot -k mMetal-Slag - mBath
mtotal total mass flow rate through the converter
mHot Spot mass flow rate through the hot spot
mMetal-Slag mass flow rate through the metal-slag
reaction zone
mBath mass flow rate through the bath
mX
mass flow rate of component x
M X
molar mass of component x
nx molar flow rate of component x
N number of droplets
N droplet flow rate
RX reaction rate
22 Chapter 2
t’ droplet residence time in the slag
V volume of droplet
[I dissolved in liquid iron
0 dissolved in slag
{I in gas phase
At time increment
P density
References
1. I. Barin, M. Modigell and F. Sauert, Metall, Trans. B, 1987, ISB, 347-354.
2. V. I. Baptizmanskii et al., Steel in the USSR, 1973,8, 634-638.
3. G. Eriksson, K. Hack and S. Petersen, ‘ChemApp - A Programmable Thermodynamic
Calculation Interface’, in Werkstomoche ’96, Symposium 8: Simulation, Modellierung,
Informationssysteme, J. Hirsch (ed), DGM InformationsgesellschaftVerlag, 1997.
4. G. Eriksson and K. Hack, Metall. Trans. B, 1990, 21B, 1013-1023.
5. C. W. Bale, P. Chartrand, S. A. Degterov, G. Eriksson, K. Hack, R. Ben Mahfoud, J.
Melancon, A. D. Pelton and S . Petersen, CALPHAD, 2002,26(2), 189-228.
6. S . Asai and I. Muchi, Trans. ISU, 1970,10, 250-263.
7. A. I. van Hoorn, J. T. van Konyenburg and P.J. Kreyger, in Role of Slag in Basic Oxygen
Steelmaking, W.-K. Lu (ed), McMaster University Press, Hamilton, Ontario, 1976.
8. K. Koch, W. Fix and P. Valentin, Archivfur das EisenhUttenwesen, 1976,47(10),583-588.
CHAPTER 3
Multiphase Thermodynamics of
Pulp Suspensions
PERTTI KOUKKARI, RISTO PAJARRE AND ERKKI e S m N
1 Introduction
By the year 2000, the global production of paper and cardboard products was
exceeding 300 Mt annually. In the 1990s, the average growth rate of the use of
printing paper and related commodities was 3.2%. Although the emergence of new
information and communication technology is expected to replace some of the tra-
ditional uses of paper, the development is likely to lead to new improved paper
products, such as qualities suitable for digital printing. Trees and other fibrous
plants also provide a renewable resource with a long industrial tradition of utilisa-
tion, which has also prepared the ground for future development of new fibre-based
materials.
Pulp fibres made of wood are the basic raw material of the present pulp and paper
industry. In order to obtain a final product with desired properties, various inorganic
and organic chemicals are added to pulp at different stages of the manufacturing
process. The interactions between the pulp fibres and these added chemicals deter-
mine the properties and quality of the resulting product. They also have a major
influence on the stability and economics of the industrial process and have a preva-
lent effect on the maintenance of the equipment by ruling the conditions of their
wear and corrosion. Therefore, the chemistry of pulp suspensions has gained con-
siderable interest during the past decades. The importance of a proper understanding
of the chemical state of pulp suspensions has been further increased by the often
environmentally driven changes in pulp- and paper-making processes since the
1990s: Reducing the amount of water used in processes has increased the amount of
metal ions circulating with process water. Transition metals like manganese, iron and
copper catalyse the decomposition of oxygen-containing bleaching chemicals that
have replaced chlorine gas, while magnesium has been noted to have a beneficial
effect during alkaline oxygen or hydrogen peroxide pulp bleaching. Thus, the
control of metal distribution in bleaching suspensions has become an important
24 Chapter 3
industrial practice. Similarly, the increasing use of CaCO, as a filler and coating
material in paper-making has made the proper understanding and controlling of cal-
cium chemistry in fibre suspensions more important. A great deal of the relevant
knowledge of fibre-ion interactions has been on a purely empirical basis; this paper
discusses work based on general thermodynamic principles aimed at a better under-
standing of the chemical interactions on a more fundamental level and using that
knowledge to improve the current pulp-and paper-making process.
2 Fibres in Aqueous Solution

Pulp fibres are composed of cellulose microfibrils, and gel-like hemicelluloses and
lignin. A schematic picture of pulp fibres in aqueous solution is shown in Figure 1.
The water in the system is divided between two sub-volumes: the volume of water
surrounding the fibres and the volume of water inside the fibre walls. The amount of
water enclosed by the fibre walls depends on the type of pulp and to some extent on
the chemical composition of the suspension liquid, typical values being around
1-1.5 Vkg-l of dry weight. The cellulose microfibrils form the backbone of the fibre,
whereas hemicelluloses and lignin contain acidic groups that can dissociate giving
the fibre an anionic charge. The chemical nature and amounts of these functional
groups depend on the wood raw material and the processes used to extract the fibres
from the wood and to bleach the fibres,' typical total amounts of these acids varying
in the range of about 50-300 mmoVkg-'. In order to neutralise the negative charge,
mobile cations are enriched in the part of solution inside the fibre walls. The differ-
ence between ionic concentrations between the fibre wall exterior and interior
increases with increasing pH, as more anionic groups are formed from dissociating
organic acids, and decreases with increasing ionic concentrations, when a smaller
fraction of the total mobile ion amounts is needed to balance the charge. In addition
to simple ionic solutes, the system contains organic macromolecules and colloidal
substances (thermochemistry of which will be mostly beyond the scope of this dis-
cussion) and often inorganic or organic metal salts.
c1-
Na'
cr Ca2+
Na'
CI- CI-
Fiber wall External suspension solution
Figure 1 Schematic presentation of a pulp fibre in an aqueous electrolyte solution

Multiphase Thermodynamics of Pulp Suspensions 25
3 Fibre-Ion Interactions
The interactions between the charged fibres can be divided into ion non-specific
and ion-specific parts. Ion non-specific part is the electrostatic interaction or repul-
sion between the ions and the charged matrix of the fibre wall. The strength of this
interaction is assumed to depend only on the charge of the ionic species. The ion-
specific part of the interaction contains the specific complexation of metal ions with
the bound acidic groups and the adsoqtion of ions into polarised surfaces. By its
very nature, the ion-specific interactiom can be used to describe a more complex
pattern of observed phenomena than non-specific interactions, but they also require
a larger number of individually fitted numerical parameters in a model.
3.1 Donnan Equilibrium

Electrochemical systems consisting of two parts where one or more charged species
is restricted to one part of the system by a semipermeable membrane were first
investigated by Donnan in the early twentieth century.2In such systems, the equilib-
rium condition for all solute species is that their electrochemical potentials are equal
on both sides of the membrane:
p;+RT In a[+z,Fy' = p;+RT In a;.'+z,Fy" (1)
From Equation (1) the relation for the activities of ionic solute species on both sides
is
where A is called the distribution coefficient. If the immobile ions are located on the
side marked with superscript" and are negatively charged, the potential difference in
Equation (2) will be positive and the coefficient k l . Mobile cations will be
enriched on the side of the bound ions, whereas mobile anions will show the oppo-
site behaviour. Shortly after its initial publication, the Donnan theory was applied to
describe polyelectrolyte systems where anionic groups were confined to one part of
an aqueous solution by chemical bonds instead of a semipermeable membrane.3 The
theory was used by Neale ( 1929)4and Farrar and Neale ( 1952)5to characterise elec-
trolyte interactions with cellulose fibres. In 1996 Towers and Scallan6published their
mathematical model, based on the Donnan theory, which could be used to calculate
the ionic distribution of mixtures of mono and divalent cations and monovalent
anions in pulp suspensions. In their model equation (2) (assuming ideal solutions, so
that the activites could be equated with corresponding concentrations), together with
charge neutrality conditions for the two aqueous phases, equilibrium coefficients for
the bound acidic groups and mass balances were used to derive an equation for A at
a given pH, cation and acid amounts and water volumes. It was assumed that anionic
groups formed from the bound acids did not form any specific complexes with any
cations except the hydrogen ion. A good agreement could be found between the
experimental and calculated distributions for Na', Ca2+,Mg2+ and Mn2+ ions in
Chapter 3
0.9
0.8
0.7
0.0 ; 3 4 5 6
I
7 8 9 1
1
0
pH of external solution
Figure 2 Fraction of the total metal content present in the external solution in a pulp sample
treated with chelating agent DTPA, redrawn based on the data in ref. 1
pulp suspensions of 1%consistency over a pH range of 2 to 10. Later, the model was
extended to describe multivalent anions, hydroxyl complexes and chelating agents. l v 7
An example of modelling results in such systems is given in Figure 2. The formation
of complexes of metal ions with hydroxide ions and chelating agents can strongly
affect the cation distribution as negatively charged complexes will be repelled by
charged fibres. The Donnan theory has also been used to determine the distribution
of dissolved Cu2+and A13+ and it has been shown to be useful in describing
the removal of anionic macromolecular components during pulp washing. lo. It has
also been combined with models for displacement flow of aqueous solution through
an immobile fibre network and dissolution and ion exchange kinetics to construct a
model for pulp washing and chelation processes.' While most of the experimental
work in the area has been carried out with pulps with 1-2% consistency, the Donnan
theory has been successfully applied to suspensions up to 10%consistency.loJ1It is
still to be expected that in industrial applications with increasing consistency of the
pulp, for example in pulp washing and bleaching operations, the increasing inho-
mogeneity of the suspension will make the combination of the equilibrium model
with suitable mass transfer kinetics' more and more important.
3.2 Specific Interactions

Because of its simplicity, the concept of Donnan equilibrium is an attractive
approach for the modelling of fibre-ion interactions. However, despite its general
success in many cases, the Donnan theory alone is not sufficient to describe the
experimentally observed distributions. In those cases, it is necessary to use ion-
specific complexation equilibria with the acidic groups as a part of the model to
describe the observed phenomena satisfactorily. Such behaviour has been shown to
exist, for example, with chemically modified highly charged fibres.l 2 An apparent
systematic deviation in the ionic distribution from the predictions of the Donnan the-
ory has also been noted with common oxygen-delignified haft pulps,13 partly, but
Multiphase Thermodynamics of Pulp Suspensions 27
not entirely, explicable with the assumption of ideality of the aqueous solutions used
with most Donnan-based models.
4 Solid and Gaseous Phases

Not all the metal ions present in pulp suspensions are fully dissolved. Experiments
with oxygen delignified haft pulps have demonstrated that Ca2+and Mg2+ are pres-
ent in them predominantly as carbonate precipitates." CaCO, can also be used in
paper-making as a filler or pigment. Under other conditions, when the amounts of
carbonates in the system are less, the hydroxides are expected to dominate the
species.14 Other solids that are known to form in various parts of pulp-and paper-
making processes include calcium sulfates and oxalates, various silicates and alu-
minium salts. In addition to simple stoichiometric salts, Mn2+ and Fe2+ are known
to co-precipitate with Mg2+ in pulp suspensions, the phenomenon having practical
importance because of the detrimental effect of dissolved divalent manganese and
iron on oxygen-based bleaching chemicals. Mn2+has been shown to form solid solu-
tion hydroxides with Mg2+ and to co-precipitate on MgCO, surfaces. l4 Dissolution
kinetics of any solid precipitate should also be taken into account when modelling
the chemical state of a pulp suspension system;' the stability of solid precipitates
making it harder to quantitatively model the chemistry of ions such as Cu2+ and
Fe3+. Of the various possible gaseous species, the most important is CO,. While the
effects of its dissolution or release are difficult to model under actual process condi-
tions, the gas-aqueous equilibrium can be shown to influence significantly the
observed pH in laboratory experiment^.^^
5 General Gibbs Energy Model

Computer programs that calculate the equilibrium state of a system for given initial
composition based on the minimisation of the overall Gibbs energy of the system have
been in use for several decades. They offer the possibility of easily carrying out calcu-
lations with fast and robust codes with chemical systems containing a large selection
of possible phases, including non-ideal mixtures. A thermodynamic description of a
pulp suspension system, including the Donnan equilibrium, that is compatible with an
existing commercial thermodynamic software has been developed.16A description of
the structure of the multiphase model of a pulp suspension is given in Table 1.
As in a thermodynamic system description used for a normal solubility equilib-
rium calculation, the system contains a gas phase, if considered relevant for the prob-
lem at hand, an aqueous solution phase (external to the fibres), and a number of solid
phases, which appear either with fixed stoichiometry or as solid solutions. The fibres
are described as a separate aqueous phase. The thermodynamic data and stoichiom-
etry for the solute species inside the fibre phase are identical to those describing the
species in the external solution volume, with the exception that the charge of the
species in the two aqueous phases must be defined separately. This will ensure
that, given valid input values, charge neutrality will apply to both aqueous phases
individually in the equilibrium composition calculated by Gibbs energy
28 Chapter 3
Table 1 Structure of an example thermodynamic multiphase system
0 H N C Na Cl Ca Mn EDTA H,Ojb, Acid, e- e,&
Gas
External solution
H2O 1 2
H+ 1 -1
OH- 1 1 1
Na+ 1 -1
c1- 1 1
Ca2+ 1 -2
c0;- 3 1 2
HCO, 3 1 1 1
co2 2 1
Mn2+ ^ ^ ^
1 -2
H4EDTA 4 1
H,EDTA- 3 1 1
H2EDTA2- 2 1 2
H 1 EDTA3- 1 1 3
EDTA4- 1 4
MnEDTA2- 1 1 2
Fibres
H2O 1
H+ 1 -1 -1
OH- 1 1 1 1
Na+ 1 -1 -1
c1- 1 1 1
Ca2+ 1 -2 -2
c0:- 3 1 2 2
HCO; 3 1 1 1 1
co2 2 1
Mn2+ 1 -2 -2
H,EDTA 4 1
H,EDTA- 3 1 1 1
H,EDTA2- 2 1 2 2
HlEDTA3- 1 1 3 3
EDTA4- 1 4 4
MnEDTA2- " "
1 1 2 2
Acid,- 1 1 1
Acid,H 1 1
Solids
CaCO, 3 1 1
MnCO, 3 1 1
Mn(OH), 2 2 1
Multiphuse Thermodynamics of Pulp Suspensions 29
minimisation. It should also be noted that the program used to run the calculations
does not need to be coded to use electrochemical potentials (Equation (1)) or to
explicitly consider the Donnan distribution (Equation ( 2 ) )while carrying out the cal-
culations. Rather, since the Donnan equilibrium theory can be derived from the prin-
ciple of charge neutrality of any macroscopic part of a thermodynamic system, the
correct equilibrium composition that is consistent with the Donnan theory will be
achieved using normal formulas for chemical potential without the electrostatic
term. In a calculated equilibrium state, the following equality is valid for any solute
species (superscript s denotes the aqueous solution external to the fibres and super-
script f denotes the solution inside the fibre walls):
p; +RT In u: =,up +RT In af-zjnc (3)
where n stands for the chemical potential of the component differentiating the unit
charge in the fibre phase from the charge in the external solution. Comparing
Equations (1) and (2) with Equation (3) the following two equalities can be obtained:
A=..pi -+j
Distribution coefficient A is also readily evaluated from any calculated equilibrium
composition using Equation ( 2 ) with the calculated activities of any mobile ion in
the system.
Another necessary modification of the fibre phase data is related to the fact that the
amount of water inside the fibres cannot be allowed to vary freely during calculations.
The simplest solution is to keep its amount constant. This is not precisely true as the
swelling of fibres is affected both by pH and the ionic strength of the solution. However,
the calculated ionic distribution is not particularly sensitive to moderate changes in fibre
phase volume and the same assumption of a fixed amount of fibre phase water has been
commonly made with other Donnan theory-based models as well.
Unlike the mobile ions and neutral solutes, the bound acidic groups are included
only in the fibre phase. As no reactions that would change the total amount of these
groups are assumed to take place, the chemical potential of the undissociated forms
of these groups can be set to zero, while the chemical potential of the anionic forms
can be calculated based on the thermodynamic relation
AGj=-RT In K j
where the acidic dissociation constants and the corresponding molar amounts can be
determined experimentally by potentiometric or conductometric titration. In a simi-
lar manner, species corresponding to explicit association between bound anionic
groups and metal cations can be included in the model if experimental results show
that the non-specific Donnan equilibrium alone is not sufficient to explain the ionic
distribution with the fibre type in question.
The multiphase model can be used to cover various fibre types or mixtures, once
their acid-base characteristics (dissociation constants Ki)have been determined by
titration. The model is independent of the valences of both anions and cations in the
30 Chapter 3
system. All such solute species, whose chemical potentials in terms of their mutual
concentrations are known, can be included. Figure 3 shows the pH-dependence of
Ca2+concentration in bleached eucalyptus sulfate pulp (ESA) and thermomechani-
cal softwood pulp (TMP) suspensions. The pH adjustment has been performed with
sulfuric acid and potassium hydroxide (ESA) and with carbon dioxide (TMP). The
respective data input is shown in Table 2.
The multiphase thermodynamic model described above has been implemented as
a computational tool using the ChemSheet thermodynamic software18 (part of the
widely used Solgasmix/ChemSage/ChemAppprogram family) that supports the use
of the Pitzer formalism in describing ionic interactions in solutions and a number
non-ideal solution models for solid phases. Despite the fact that the thermodynamic
ESA/CaC03/H2S04
0.6 1
A
0.5 -
m
E 0.4 -
2
0.3 -
E
3 0.2 -
0.1 -
0 2 4 6 8 10 12
TMP/Ca2+/C02
0.8 I
0.7 -
0.6 -
0.5 -
0.4 -
0.3 -
0.2 -
0.1 -
,,,',--.
I
'.*.
0 : I I I I I
0 2 4 6 8 10 12
PH
A [Cals measured
Figure 3 Calcium concentrations in the suspension of eucalyptus sulfate (ESA) and t h e m -

mechanical sofrwood (TMP) pulps. Experimental data are @om re$ I7
Multiphase Thermodynamicsof Pulp Suspensions 31
Table 2 Data input for the ESA and TMP models

Pulp Water Acidic group content CaCO, Ca(OH),
consistency content
of the
jibre wall pK=3.6 pK=4.1 pK=9.1
(%) (dm3kg-') (moVkg-') (moVkg-l) (moVkg-') (moVkg-') (rnoL4kg-l)
ESA 1 1 0.067 0.05
TMP 1 1.5 0.162 0.068 0.1 15
data used in calculations could still be refined in many ways and that the current
knowledge of the relevant kinetically controlled processes is rather limited, the
model has already been found to be useful in process design in the paper industry.19
6 Discussion
The present thermochemical model describes the acid-base, ion exchange and solu-
bility characteristics of a homogeneous pulp suspension. The important feature of
the thermodynamic multiphase approach is that it provides the possibility to incor-
porate specific interactions of practically unlimited number of constituents into the
system. Due to its general thermodynamic basis, the multiphase method can be
applied both in the fibre line processes in pulp production and in the wet end chem-
istry of paper-making.
The thermodynamic method is used to control the formation or dissolution of pre-
cipitates in different process stages, to model pulp washing efficiencies as a function
of pH and to determine pH and alkalinity in the wet end suspensions of paper
machines. An illustrative application is the utilisation of carbon dioxide in pH-con-
trol of both fibre line and paper machine processes. The practical paper machine
stock system consists of the aqueous solution, a mixed fibre phase of different pulps
together with some recycled suspension, which contains both fibres and calcium
compounds. Solid CaCO, is introduced as a filler for opacifying and callipering the
sheet. Here, the conventional pH-control with sulfuric acid has been extensively
replaced with the more 'gentle' treatment with gaseous carbon dioxide or with a non-
stoichiometric mixture of CO, and sodium hydroxide solution.20The latter forms an
effective pH-buffer, and prevents the unwanted dissolution of solid calcium carbon-
ate. The dosage of the buffering mixture is determined with the multiphase calcula-
tions to maintain the process pH within k0.2 pH-units. Other paper machine
applications include the calculation of the impact of fresh water reduction on pH an
free calcium levels of the stock suspension and estimation of wet end process con-
ditions in systems including both sulfate and carbonate species.
The multiphase method provides a practical 'screening tool' for industrial
process research and development, even though under many circumstances the non-
equilibrium effects such as supersaturation of solutions, retarded mass transfer or
reaction kinetics and inhomogeneity of suspensions limit the applicability of the
thermodynamic calculations. When the thermodynamic multiphase models are
developed towards process simulation tools, one should incorporate such methods
that include the effects of these non-equilibrium factors. They must be based on
32 Chapter 3
laboratory and process information and by necessity involve data from kinetic
measurements. The approach towards the equilibrium state still works as a guiding
principle, and in several cases the equilibrium state can be considered to be at least
a useful approximation of the true state of the system.
The thermodynamic method provides a quantitative basis for the acid-base,
ion-exchange and solubility behaviour of the fibre stock suspensions. Many impor-
tant measurable process indicators, such as concentrations, pH, alkalinity and Ca-
hardness, can be directly calculated using the thermodynamic treatment, while
additional information, concerning, for example, conductivity and colloidal stability,
can be deduced from the simulated chemical state with the help of additional physic-
ochemical and empirical relations.
References
1. E. Rasanen, Modelling Ion Exchange and Flow in Pulp Suspensions, Doctoral Thesis,
Helsinki University of Technology,V?T Publications 495, Espoo, Finland, 2003.
2. F. G . Donnan and A. B. Harris, J. Chem. SOC., 191 1,99, 1554.
3. H. R. Proctor and J. A. Wilson, J. Chem. SOC., 1916, 109, 307.
4. S. M. Neale, J. Textile Znst., 1929, 20, 373.
5. J. Farrar and S. M. Neale, J. Colloid. Sci., 1952, 7 , 186.
6. M. Towers and A. M. Scallan, J. Pulp Pap. Sci., 1996,22, 332.
7. E. Rasanen, P. Stenius and P. Tervola, Nord. Pulp Pap. Res. J., 2001, 16, 130.
8. J. Lindgren, L. Wiklund and L.-0. Ohman, Nord. Pulp Pap. Res. J., 2001, 16, 24.
9. J. Lindgren, Experimental Studies of the Acimase Properties and Metal Ion Affinities of
Wood Fibres, Doctoral Thesis, Umet University, Dept. Chemistry, Inorganic Chemistry,
Umei, Sweden, 2001.
10. R. Anderson, J. LidCn and L.-0. Ohman, 7th International Conference on New Available
Technologies, Stockholm, Sweden, 2002,42.
11. C. Sjolander, J. LidCn and L.-0. Ohman, International Pulp Bleaching Conference,
Halifax, Canada, Oral presentations, 2000, 273.
12. J. Lindgren, P. Person and L.-0. Ohman, Nord. Pulp Pap. Res. J., 2001,16, 225.
13. K. Athley and P Ulmgren, Nord. Pulp Pap. Res. J., 2001, 16,204.
14. J. Liden, L. Wiklund and L.-0. Ohman, 11th International Symposium on Wood and
Pulping Chemistry, Vol 1, Nice, France, 2001, 1, 193.
15. P. Koukkari, R. Pajarre and E. Rasiinen, 12th International Symposium on Wood and
Pulping Chemistry, Madison, Wisconsin, 2003,3, 59.
16. P. Koukkari, R. Pajarre and H. Pakarinen, H., J. Solution Chem., 2002,31,627.
17. H. Maenpaa, ‘Paperikoneen kuituseosten ioninvaihtomallit’ (‘Ion-exchange models
of the pulp mixtures at paper machines’, in Finnish), Masters Thesis, Helsinki University
of Technology, Laboratory of Forest Products Chemistry, Espoo, Finland, 2001.
18. P. Koukkari, K. Penttila, K. Hack and S . Petersen, in Microstructures, Mechanical
Properties and Processes, Euromat 99, Vol 3. Y. Brechet (ed), Wiley-VCH Publishers,
Weinheim, 2000,323.
19. K. Viitikko and A. Kalliola, Wet End Technology, Scandinavian Paper Symposium,
Helsinki, Finland, 2002.
20. A. Weaver, A. Kalliola and P. Koukkari, Third Major Pira International Conference of
Scient$c & Technical Advances in Wet End Chemistry, Vienna, Austria, 2002.
CHAPTER 4
Reactive Distillation
HARTMUT G. SCHOENMAKERS AND WOLFGANG ARLT
1 Introduction
Reactive distillation, as the name implies, refers to a distillation process that incor-
porates a reaction and a separation step within a distillation column. The technique
offers a key opportunity for improving the structure of a process.'V2 It is a so-called
hybrid process, i.e. it merges two different unit operations in a single apparatus,
namely reaction and distillation. But the combination of distillation and reactions is
possible only if the conditions of both unit operations can be combined. This means
that the reactions have to show reasonable data for conversions at pressure and tem-
perature levels that are compatible with distillation conditions. Because of the lim-
ited hold-up in distillation column, those reactions having a conversion half-time of
10-30 min are preferred. So, the judicious use of the chemical equilibrium constant
is the basis for the design of reactive distillation processes.
The advantages of reactive distillation as compared to a reaction plus a separate
distillation process are:
a reduction of the investment,

a simpler process,
the use of the heat of reaction (if present) in situ,
ease of control of the reaction temperature (evaporating system), and
the possibility of overcoming azeotropes.
A typical reactive distillation set-up is depicted in Figure 1.

The high boiling reactant is fed as feed 1 and the low boiling reactant as
feed 2. Between the two feeds, there is the reaction zone. As a special application,
feed 1 can serve as an extractive agent, e.g. in the case of the production of methyl
acetate, acetic acid serves as an entrainer for the binary azeotropic mixture
methanol and methylacetate. The ensemble is then a reactive extractive distillation
column.
The low boiling product leaves the column at the top and the section between feed
1 and the reflux serves as a rectifying section. The high boiling product is withdrawn
34 Chapter 4
1
Feed 1
Catalyst
e distillation ZOI
Feed 2
Product 2
Figure 1 The basic elements of a reactive distillation column
L 40 1
ca
g 35
2 30
'=0 25
8 20
2 15
.c
O 10
ti
2 i13m.lllllllllllllllllllllllln
z
1965 1970 1975 1980 1985 1990 1995
year
Figure 2 A summary of the publications in the field of extractive distillation between I965
and 1995
at the bottom; the section between feed 2 and the evaporator is known as the strip-
ping section.
Reactive distillation is becoming increasingly popular and in the three decades
from 1965 to 1995, there was a significant increase in the number of papers pub-
lished in the literat~re.~The figures of BeBling3are graphed in Figure 2.
2 Thermodynamics
As in most separation processes, thermodynamics play a key role in designing reac-
tive distillation equipment. The process is governed, on the one hand, by the ther-
modynamic properties related to separation by distillation and on the other hand, by
the equilibrium constants related to the reactions.
2.1 Phase Equilibrium

The general condition for phase equilibrium is the equality of chemical potentials.
For practical reasons this is replaced by the equality of the fugacity f of each of the
Reactive Distillation 35
compounds in the liquid (L) and vapour (V) phase:
One of the two ways to describe the fugacity is by using activity coefficients yin
the liquid phase and fugacity coefficients q in the vapour phase. Activity coeffi-
cients require the definition of a standard state. In distillation, the two terms are
related by
XiXJ+ = yiqivp
where + refers to the standard state, xi and yj are the mole fractions of the liquid and
vapour phases, P is the pressure of the system and q is the fugacity coefficient. The
standard state is chosen to be the pure substance at the temperature and pressure of
the mixture. Pkv is the vapour pressure of the pure compound (subscript ‘0’)and is
usually given in the form of an Antoine-type equation. The exponential expression
relates to the change in going from the vapour pressure to the pressure of the system
as a function of the pure liquid volume. In many cases, the difference of pressures is
small enough such that the exponential expression tends to be unity:
The fugacity coefficients of the pure substance in Equation (3) and of the mixture in
Equation (2) are of similar value for many compounds, so they often cancel.
However, for organic and inorganic acids (very often feed 2 in Figure 1) and some
other compounds that involve the formation of dimer and higher n-mers formation
in the vapour phase, the fugacity coefficient must be taken into account.
Activity coefficients can be derived from databank^,^ correlation model^,^ struc-
ture interpolation methods6 or a priori Modern methods allow the cal-
culation of multicomponent behaviour on the basis of binary systems. In this case,
only binary data are needed. Very often, however, experimental data for the binary
mixture that forms the reactants are unknown. As an example let us consider the
equation:
A+B H C+D (4)
described in Figure 3. Phase equilibrium data of the binary pair A,B will be rare.
This is because the reaction makes it most difficult to measure the phase equilibrium.
A comprehensive discussion is given by Ti~chmeyer.~ Only in the case when the
reaction rate is slow is a fast determination of the phase equilibrium possible with a
standard apparatus. In many other cases, special equipment must be used. A flow
apparatus enables to resolve the reaction time as a reaction locus. So, an appropriate
flow will make it possible to make the measurements.
Roederer et aZ.I0 investigated an isomeric reaction (alpha- and beta-isophorone)
at 388,423 and 493 K. They experimented in a stirred batch reactor in the form of
36 Chapter 4
D
A
C
Figure 3 Concentration space of the reaction A +B H C+ D
a closed-loop reactor. For VLE experiments it was possible to use the experimental
set-up as a flow-type cell. The compositions of the phases are obtained on-line with
a Fourier-transform-infrared spectrometer. Alsmeyer et a2.l have set up a flow-
through apparatus for VLE measurements in reactive systems that also allows
determination of reaction kinetics in the liquid phase. The reactive mixture is
forced into a steady state that does not represent its chemical equilibrium but
reaches phase equilibrium between vapour and liquid. Both phases are then
analysed in situ by mid-infrared (MIR) spectroscopy. While the gas phase is meas-
ured by transmission, attenuated total reflection (ATR) is used for the liquid phase
measurements.
Another flow-type apparatus has been operated by Reichl et a2.12It was set up as
a non-recycle flow still and, due to the short residence time, it was used to determine
isobaric VLE data of thermally unstable components and of reactive mixtures. Two
esterification systems (methyl formate and ethyl acetate) and one etherification sys-
tem (tert-amyl methyl ether) were investigated.
Arlt13 proposed a new type of static still. The temperature and pressure are
recorded at a given binary start mixture over time. By extrapolation to the zero time,
pressure and temperature for this point can be derived. The method works if the reac-
tion half-times are longer than 10 min, which is a prerequisite for the use of reactive
distillation columns (see above).
The binary data of the mentioned apparatus are used to determine binary interac-
tion parameters of @-models or of equations of state.
2.2 Chemical Equilibrium

In formulating and understanding the problems inherent in reactive distillation, it
is necessary to take the chemical equilibrium into account.
Starting from the fundamental equation of thermodynamics for one component:
Equation (6) can derived. The free enthalpy G, which should reach a minimum
within the field of the fundamental variables P,T and compositions, is related to the
chemical potentials p of the species i. Here, the moles n of the compounds in the
reaction are replaced by the variable 1that uses the information of the reaction kinet-
ics, namely the stoichiometric coefficients v:
dnA - dnB - dnC - &D

dL= -
vA vB 'C D
'
The expression in Equation (6) can be calculated using the model of activity coeffi-
cients yor of the fugacityf. In terms of activity coefficients, the following equation
can be derived:
xixf& vi
K=l$$)vi=n(,,)
The product of activity coefficients yand the mole fraction x is often called activity
a . It is noteworthy that the chemical equilibrium constant can only be calculated if
the acitivity coefficients are known. In the case of reactive distillation, this informa-
tion is available. It should be mentioned that the equilibrium constant can be calcu-
lated from the pure free enthalpies of formation that have to be corrected to the
reaction state depending on the model (activity coefficients or fugacity) used.
It is further important to know that the kinetics must also be formulated in activ-
ity coefficients. too. The following equation gives the scheme for the above-men-
tioned equilibrium reaction:
The k's refer to rate constants.

If the reaction kinetics are expressed in concentrations only, an ideal solution is
assumed. Furthermore, under chemical equilibrium conditions the two reaction rates
(to products and from products) should be the same. Because the chemical equilibrium
constant K must be expressed in activity coefficients, an inconsistency might occur.
In general, the illustration in Figure 3 is not suitable to present data in a plane area.
Several proposals have been made by (a) Zharov, (b) Barbosa and Doherty and (c)
Ung and Doherty14and Espinosa et aZ.15The latter will be used here.
For a given concentration y the following rule is applied to calculate the new coor-
dinate Y in reference to the component k:
38 Chapter 4
3 Technical Application of Reactive Distillation

3.1 Introduction
In reactive distillation, the type of the catalysis is important. Homogeneous cataly-
sis is possible in most cases but needs a separation step to recycle the catalyst. This
can be avoided in heterogeneous catalysis, but here special constructions are neces-
sary to fix the catalyst in the reaction zone. If everything harmonizes, considerable
advantages arise as can be seen with reference to the Eastman-Kodak Chemicals
process for the production of methyl acetate. As can be seen in Figure 4 only one
column is needed if reactive distillation is used as opposed to nine and a reactor if it
is not used.
Is this an exception or does it stand for a type of process with a comparable poten-
tial for improvement? The questions are:
- how can processes be identified that exhibit potential for reactive distillation,
- how can such a process be developed and
- how can the equipment be designed?
Some answers to this question will be offered from the viewpoint of an industrial
process development by simulation, experimental validation and the choice and
design of suitable equipment.
3.2 Process Synthesis

A vital aim of process synthesis is to reduce the complexity of the problem in order to
enable simple solutions to be recognized quickly. In order to systematically analyse
processes involving reversible reactions, a comprehensive process synthesis strategy has
-TI-
er
1 SOlVenI I
water
catalyst
Figure 4 Production of methyl acetate at Eastmun-Kodakwithhithout reactive distillation

been developed. One element of this strategy is the analysis of reactive distillation lines
that enable the feasibility of reactive distillation processes to be examined simply.
To achieve this, various authors have developed transformation m e t h o d ~ . ~ J ~ ~ ~
These transformation methods enable reactive distillation to be described by a sys-
tem of equations that is known from conventional distillation; see Equation (9). The
method is limited by the fact that it is possible only for equilibrium reactions and is
more an estimation than a design.
3.3 Process Design and Optimization

While process synthesis gives qualitative reference points, for industrial
implementation we need quantitative results. Therefore, tools for rigorous process
simulation including all effects are needed. In practice, the application of staged
models with increasing complexity can be recommended but tools with this com-
plexity are not yet on the market, so that in many cases reactive distillation cannot
be simulated so far that a process design is possible without experiments.
3.4 Limitations of the Methods for Synthesis and Design,

the Scale-Up Problem
Another important question is how main and secondary reactions are described. In
practice, it is frequently the case, for time and cost reasons, that a complete kinetic
model cannot be developed and experiments in a reaction column are needed to ver-
ify the achievable purities and yields. With the results of these experiments, a scale-
up problem exists. This is because in these experiments, reaction and mass transfer
interact with one another. The product removal capability has to be designed with
respect to separation performance and the property of converting the starting materi-
als that show the characteristics of a reactor performance with different scale-up rules.
If an engineer has to design an industrial column purely on the basis of mini-plant
experiments, he has to maintain not only the separation performance but also the ratio
of separation performance/reactor performance so that main and secondary reactions
proceed to a comparable extent in the industrial-scale reaction column. One way in
which this can be achieved is in terms of construction by separating reaction and
product separation from one another both in mini-plant tests and on an industrial
scale. This is possible, for example, using a heterogeneous catalyst in the downcomer
or with side reactors on the column. An alternative is to use a structured packing with
well-defined paths for the liquid flow. In general, the problem has not yet been solved,
the main reason being the lack of reference columns on an industrial scale.
The way we see is either:
- to develop scale-up methods or

- to carry out reactive distillation experiments on an industrial scale.
The latter possibility, which represents a step back into the process engineering
stone age in the design of columns, will in many cases mean the end of the line for
40 Chapter 4
reactive distillation, since a company will only seldom be prepared, for time and cost
reasons, to build integrated experimental plants on an industrial or semi-industrial
scale. Thus, developing an appropriate scale-up procedure for reactive column inter-
nals is a key task.
3.5 Choice of Equipment

Up to now it was assumed that reaction and distillation can favourably be combined
in a column - in a normal distillation column in the case of homogeneous catalysis,
and in a column with special internals or an additional exterior volume in the case of
heterogeneous catalysis. This was discussed in the previous chapter under the aspect
of scale-up in connection with separation and reaction performance. However,
columns are an appropriate solution only for reactions that are so fast as to achieve
considerable conversions in the residence time range of such columns. The question
is whether the full potential for combining reaction and distillation can be found and
industrially implemented using columns only.
Thus two operating conditions can be distinguished:
- The range in which the conversion is influenced mainly by the concentration of

the component to be separated; this range is called ‘controlled by distillation’.
- The range in which the conversion is influenced mainly by the residence time
and the reaction constant; this range is called ‘controlled by kinetics’.
Industrial process design should aim at operating conditions within these two
ranges: just the sufficient residence time and only the necessary expenditure for the
distillation.
So we have a second scale-up problem: what is the suitable equipment for com-
plying with these demands? This means more precisely: how can the reactor per-
formance be achieved over a broad range of reaction velocities? Different equipment
may be chosen to combine reaction and distillation within the limiting conditions of
reaction velocity, relative volatility and catalysis.
Equipment in question are:
- vessels stirred or with fixed beds,

- cascades of vessels, both with or without columns and
- reaction columns.
Additional volume can be provided for all of them.

The next considerations concentrate on homogeneous catalysis. Similar consider-
ations apply to heterogeneous catalysis, and this will be commented on later.
At first, a slow reaction is considered. ‘Slow’ means that the reaction rate is slow,
compared with the residence times typical for separation equipment such as distil-
lation trays. For residence time reasons a stirred vessel or, better, a cascade of
stirred vessels is needed. Each vessel is supplied with energy to evaporate the com-
ponent to be separated. If the relative volatility of this component is very high, a
one-stage evaporation is sufficient. At a lower relative volatility, the separation
Figure 5 Principles of the choice of equipment (homogeneous catalysis)
requires more stages, i.e. a column has to be placed on top of the vessel. If the sep-
aration is even more difficult, a stripping section must be added to the column and
a reboiler is necessary. In the limiting case of a very low relative volatility, each
stage of the cascade can be operated as a countercurrent stage. The first stage is
additionally provided with a fractionating column to enrich the component to be
separated. Such a set-up is equivalent to a reaction column with a large holdup on
each reaction stage.
Fast reactions do not demand long residence times. In this context ‘fast’ means
that the reaction reaches equilibrium in the residence time range that is typical for
column internals. The equipment therefore may be selected under the aspect of sep-
aration efficiency. The procedure to select the equipment is comparable to the pro-
cedure described above and leads to a list of possible options.
As a result of such considerations a chart of process alternatives can be drawn2I
for homogeneous catalysis - see Figure 5.
As mentioned earlier, heterogeneous catalysis can also be treated in a similar
manner. There is, however, one main difference and that is that additional, separate
reaction volumes are necessary to retain the catalyst. These volumes can be arranged
either within the equipment or in a side position,22coupled by recycle systems.
4 Conclusions
In the case of conventional distillation, a sufficient body of knowledge has, thanks
to intensive development, been acquired to enable many tasks in process synthesis,
design and scale-up to be solved quickly.23For reactive distillation, the methods of
design and synthesis have been developed largely with the aid of many analogies to
A major focus of research and development in future years should be
the scale-up of reaction columns. This is where great deficiencies still remain. In
addition, the methodology in choosing the best equipment will have to be improved.
In the expert community, columns are often seen as the only possible choice. Work
will have to be carried out to ensure that the benefit of combining reaction and dis-
tillation can be enjoyed to the full by employing the most suitable equipment.
42 Chapter 4
References
1. M. Doherty, F. Michael and G. Buzad, Trans. Znst. Chem. Eng.,1992,70,448.
2. J. L. DeGarmo, V. N. Parulekar, V. Pinjala, Chem. Eng. Progr., March 1992, 88 (3), 43.
3. Befiling, B., Ph.D.Thesis (Dissertation), Universitat Dortmund, 1998.
4. J. Gmehling, U. Onken, W. Arlt, R. Grenzheuser, B. Kolbe, U. Weidlich, and J. R. Rarey,
Vapor-Liquid Equilibrium Data Collection, Vol. I, DECHEMA Chemistry Data Series,
Frankfurt, 1977.
5. H. Renon and J. M. Prausnitz, and A.1.C.h.E. J., 1968, 14 (l), 135-144.
6. J. Gmehling, R. Wittig, J. Lohmann and R. Joh, Znd.Eng.Chem.Res., 2001, 40,
5831-5838.
7. A. Klamt and G. Schiiiirmann, J. Chem. SOC.Perkin Trans., 1993,2,799-805.
8. I. Clausen, W. Arlt, Chem.-1ng.-Tech., 2000,72,727.
9. M. Tischmeyer, Thesis, Erlangen, 2004.
10. A. Roederer, Ph.D. Thesis, ETH Zurich, No. 13169, 1999.
11. F. Alsmeyer, Ph.D. Thesis, Aachen 2003.
F. Alsmeyer, W. Marquardt and G. Olf, Fluid Phase Equilibr., 2002, 203, 3 1-5 1.
12. A. Reichl, U. Daiminger, A. Schmidt, M. Davies, U. Hoffmann, C. Brinkmeier, C. Reder
and W. Marquardt, Fluid Phase Equilibr., 1998, 153 (l), 113-134.
13. W. Arlt, Fluid Phase Equilibr., 1999, 158-160, 973-977.
14. S. Ung and M.F. Doherty, Chem. Eng. Sci., 1995,50,2348.
15. J. Espinosa,; P. A. Aguirre and G. A. Perez, Chem.-hg.-Tech., 1995, 43, 541-550.
16. B. Bessling, J.M. Lohning, A. Ohligschlager, G. Schembeckerand K. Sundmacher,Chem.
Eng. Technol., 1998, 21, 393400.
17. D. Barbosa and M. Doherty, Proc. Roy. SOC. Lond., 1987, A 413,459.
18. J. Stichlmair, ChemJng. Tech., 1988,60, 747-754.
19. J. Stichlmair, H. Offers and R.W. Potthoff, Znd. Eng. Chem. Res., 1993,32, 2438-2445.
20. D. Barbosa, Ph.D. Thesis, University of Massachusetts, Order Number 8727018.
21. U. Block, Chem. Zng. Tech., 1977,49, MS 4-44/77.
22. H. Schoenmakers and W, Buehler, Gel: Chem. Eng., 1982,5,292.
23. G. Ondrey, Chem Eng. 1996,6, Newsfront.
24. Jones, European Patent 461855, 1991.
25. L. A. Smith and M. N. Huddleston, Hydrocarbon Process, 1982,3,121.
26, J. Krafczyk and J. Gmehling, Chem. Zng. Tech., 1994,66, 1372.
27. U. Hoffman and K. Sundmacher, Chem Zng. Tech., 1997,69,613.
28. A. Gorak and L. Kreul, German Patent DE 197 01 045 Al, 1997.
CHAPTER 5
Thermodynamic Properties from

Quantum Chemistry
STANLEY I. SANDLER
1 Introduction
The basis of computational quantum mechanics is the equation posed by Erwin
Schrodinger in 1925 that bears his name. Solving this equation for multielectron sys-
tems remains as the central problem of computational quantum mechanics. The dif-
ficulty is that because of the interactions, the wave function of each electron in a
molecule is affected by, and coupled to, the wave functions of all other electrons,
requiring a computationally intense self-consistent iterative calculation. As compu-
tational equipment and methods have improved, quantum chemical calculations
have become more accurate, and the molecules to which they have been applied
more complex, now even including proteins and other biomolecules.
However, ab initio quantum mechanical calculations increase dramatically in
scale with N , the number of basis functions (e.g.,Gaussian functions) used to repre-
sent the wave functions of all electrons in a system (a molecule or a collection of
molecules).' Hartree-Fock methods (which do not adequately account for the corre-
lation between electrons) scale between N2 and P,and it is higher for levels of the-
ory that introduce increasing extents of correlation: Moller-Plesset (MP)
second-order perturbation theory scales as P to N4, more accurate coupled cluster
(i.e., CCSD(T)) methods scale as fl to fl,and full configuration interaction (Full
CI) methods, which should be almost exact, scale as N-' to I@. Because of this high-
order scaling (referred to as the O(N) problem), the size of the molecules and the
number of molecules that can be considered simultaneously, using high-accuracy
quantum calculations are limited.
The earliest and continuing successful application of quantum chemical methods
has been to properties of a single molecule, such as the ideal gas heats of formation,
bond energies, spectral frequencies, heat capacities and other energy properties.
Heats of formation can now be calculated to what is referred to as a chemical
accuracy of 2 2 kcal mol-', but with considerable effort.' However, much of the
44 Chapter 5
chemistry occurs in dense phases, and the accurate calculation of thermodynamic
properties in a liquid is much more difficult, which is the subject of this review. For
such a calculation of an ensemble average over many states, a sufficient number of
interacting molecules to represent a bulk phase are required, and this is carried out
by molecular simulation. This requires the energy of interaction of each configura-
tion of a collection of molecules, which is difficult to calculate accurately due to the
O(N) scaling. For example, a single full CI calculation for 100 molecules requires
approximately 1007 greater computer resources than a single molecule calculation.
This problem will be discussed here, along with the parameterized density functional
theory and quantum mechanical polarizable continuum models. Semi-empirical
methods will not be reviewed as they do not yield accurate interaction energies. A
thorough discussion of the various quantum mechanical methods and the computer
packages available can be found e l ~ e w h e r e . ~ * ~
One method to deal with the O(N) problem, although at the price of decreased
accuracy, is to use the Density Functional Theory (DFT). In ab initio methods, the
goal is to solve for the wave functions of the electrons, while in DFT, which is also
based on single-electron wave functions, one calculates the total electron density dis-
tribution, as the ground state electronic energy and other system properties can be
determined from this distribution function. The advantages of this method are its low
cost and reasonable reliability, and consequently, it is widely used in molecular mod-
eling to calculate ground-state energies and related properties. However, DFT is
semi-empirical in that it contains parameters that have been adjusted to experimen-
tal data; unlike ab initio methods, there is no systematic way of improving predic-
tions (e.g., by increasing the level of theory), and it is not of sufficient accuracy for
the calculation of van der Waals forces.
2 Quantum Mechanical Computation of Force Fields

for Molecular Simulation
In molecular level (Monte Car10 or molecular dynamics) ~imulation,~ as is also dis-
cussed elsewhere in this book, typically hundreds or thousands of molecules are con-
sidered. Because of the O(N) scaling and the number of molecules and
configurations that must be considered, high-level quantum mechanics calculations
cannot be carried out. At present, there are two ways to proceed.
The first method is that of Car and par in ell^,^ in which one performs molecular
dynamics simulations for a large number (hundreds) of molecules, calculating the
forces and energy from quantum mechanics for the whole assembly of molecules at
each step. However, in order to be tractable, the calculations are carried out with
DFT, but even then only for small molecules and requiring greater computational
resources than are usually available. Also, the use of DFT restricts this method to
systems in which electrostatic forces dominate; ab initio methods could be used
instead, but only at very large supercomputer centers.
The second more common and somewhat less computationally intensive way of
using quantum mechanics is the temporal hybrid method in which interaction ener-
gies are first computed for pairs (and rarely triplets) of molecules at a large number
Thermodynamic Properties from Quantum Chemistry 45
of relative positions and orientations. Such calculations are usually carried out using
the perturbation theory (such as MPn, where n=2, 3 or 4, indicates the level of the-
ory) and a large basis set. When only weak interaction forces (van der Waals forces)
are present, high-level calculations are needed to obtain a reasonable accuracy, as is
also the case for very specific interactions, such as hydrogen bonding. This infor-
mation is thtn fit to an analytic interaction potential or force field by adjusting
parameters in site-site potentials using either atom-based or united-atom functional
group-based (i.e., -CH3 or -CH,) sites, and perhaps off-atom sites (usually along
bonds). While the two-parameter Lennard-Jones 6 - 1 2 potential has been used fre-
quently, other more complicated potentials are preferred for better accuracy. Also, it
can be useful to add site-site Coulomb interactions as atoms and/or functional
groups in neutral molecules can have a nonzero charge. These charges are computed
as part of the quantum mechanics calculation. The force field is subsequently used
in simulations to compute thermodynamic properties.
As the interaction energy in quantum mechanics is usually computed as the dif-
ference in energies of the dimer and the two separated molecules, two complications
arise. First is that since the interaction energy is orders of magnitude less than the
total energy of each molecule, each energy must be calculated to high accuracy to
obtain a good value for this difference. Second, since the calculated energies depend
on the number of wave functions, a Basis Set Superposition Error (BSSE)6 arises if
different basis sets are used for each of the separated molecules and the dimer. This
error is reduced by computing the energy of each molecule and of the dimer with the
basis set of the dimer.
The second virial coefficient (B,) can be exactly calculated from the ab initio-
based pair potential using multidimensional integration, and such results provide a
test of its quality. Some results obtained in this way are shown in Figures 1 and 2 for
methyl f l ~ o r i d ehydrogen
,~ chloride* and methan01,~respectively. From these fig-
ures, we can see that the predicted second virial coefficients (B,) for CH3F and HCl
are remarkably good for complete predictions (there are no adjustable parameters fit
to the experimental data), but not for methanol. The B, values suggest that the inter-
action potential for methanol is not sufficiently attractive. This is probably because
a higher level of quantum mechanics calculations is needed for this hydrogen-bond-
ing fluid than could be done with the computational resources available.
To calculate dense phase properties, molecular simulation must be used, and pair-
wise additivity is usually assumed, i.e., that the total interaction energy of a collection
of molecules can be computed as the sum of the interactions between all pairs of mol-
ecules as if each pair of molecules was isolated from all others. [Note that this assump-
tion of pairwise additivity in not used in the Car-Parinello method as the energy is
calculated for the entire collection of molecules.] Using the painvise additivity
assumption in Gibbs ensemble Monte Car10 (GEMC) simulation,1° we obtained the
results shown in Figures 3 to 5 , where we can see that the combination of a quantum
mechanics-based potential and simulation leads to excellent predictions of the vapor
pressure, phase boundaries and densities for CH3F and HC1, but not for methanol. The
results suggest that the pairwise additivity assumption is reasonable for nonhydrogen-
bonding molecular fluids, and that for such fluids, it is possible to use a temporal-
hybrid method to predict thermodynamic properties of reasonable accuracy.
46 Chapter 5
-100
n
F
1 -a00
E
c3
E
$ -a0
-400
-500
200 300 400 500
Temperature (K)
0
-100
-200
Y
-0
I
? -300
Y
O0" -400
-500
-600
100 200 300 400 500 600 700
Temperature (K)
Figure 1 The predicted second virial coeficient for methyl fluoride and hydrogen chloride
(lines) compared to experimental data (points)
There are two sources of error for hydrogen-bonding fluids: (1) that a higher level
of theory and/or larger basis sets may be needed to obtain an accurate pair potential
than for nonhydrogen-bonding fluids; and (2) that the assumption of painvise addi-
tivity may not be accurate. We can examine each of these separately. First, we can
assume that because of the use of too low a level of theory and/or a basis set of insuf-
ficient size, the interaction potential for methanol is inaccurate in a systematic man-
ner, and we can multiply the potential obtained by a single scale factor adjusted to
reproduce the second virial coefficient. From Figure 2, we see that using a multi-
plicative factor of 1.2 gives excellent agreement for B, of methanol over the entire
temperature range. The use of this scaled potential in simulation gives more accurate
0
unscaled
: :
-500
--
-
-1000 b
r,
E
m
6
: L
-1500
v
mi
-2000
6
I Tl= 1.20
I
c
I
-2500 - 1 ,
I
I
c
I I
-3000
250 350 450 550
Temperature (K)
Figure 2 The predicted second virial coeficient for methanol without and with scaling (as
described in the text) compared to experimental data (points)
vapor pressures and phase boundary for methanol than obtained previously (Figure
5); however, the agreement is still not of engineering accuracy, nor as good as that
for the nonhydrogen-bonding fluids.
The simplest method to correct for the inadequacy of the pairwise-additivity
assumption is to use a classical polarization model" to allow for the induction of one
or more dipoles in each molecule due to the electric field of the surrounding mole-
cules. The details of such calculations are given elsewhere. Because of the multibody
nature of such a calculation, and the need for iteration over all the molecules in the
system, this adds a factor of 3-10 to the computational load of an MC simulation.
We have used a single-site dipole polarization model in some of our simulations
and found that polarization (i. e., departures from pairwise additivity) had relatively
little effect on the predicted thermodynamic behavior of methyl fluoride and hydro-
gen chloride, as expected. For hydrogen-bonding methanol, the simulation results
with polarization (but not scaling) are significantly improved, although the results
are still not satisfactory. However, the combination of both point polarization (to
account for multibody effects) and scaling (to account for the use of an incomplete
basis set) yields accurate vapor pressures, although the liquid density and critical
point property predictions are still not as good as with nonhydrogen-bonding fluids.
This suggests that for accurate properties predictions for hydrogen-bonding and
other associating fluids, a high level of quantum theory and large basis sets must be
used to compute the pair potential, and that corrections must be made for the
assumption of painvise additivity.
The extension of the temporal-hybrid method to mixtures is conceptually straight-
forward. What must be done for a binary mixture is to separately compute and fit the
1-2, 2-2 and 1-2 interactions, and use the potentials obtained in simulation.
48 Chapter 5
100
10
h
5
n
Y
2
u)
2
a 1
0.1 I I
6.0 5.0 4.0 3.0
1000/Temperature (K-')
2
300
350 I @
I
3
c
$
.
250
dI
ez-
200
0.0 0.2 0.4 0.6 0.8 1.0
(b) Density (g an3)
Figure 3 Vapor pressure ( a ) and vapor-liquid coexistence diagram ( b )for methyl fluoride
from GEMC NVT simulations. The solid line indicates the experimental data and
the symbols and dashed line are simulation results based on the ab initio pair
potential. The closed symbol is the estimated critical point
The results of one such example are shown in Figure 6 using the scaled methanol
potential and quantum-mechanically derived potentials for acetonitrile and their
cross interaction. These results are true predictions from the combination of quan-
tum mechanics and molecular simulation. No parameters have been adjusted to
100
10
0.1
2 4 6 8
1OOO/Temperature (K-’)
350
r
300
x
h
A A
2
3
+
$ 250 -A
a
E
I- L A
200
t
AL A
w
150 I 1
0.0 0.5 1.o
Density (g ~ m - ~ )
Figure 4 Vapor pressure and vapor-liquid phase boundary of hydrogen chloride as a func-
and simulation A
tion of temperature from experiment (line or .)
mixture behavior (only a scaling of the pure methanol force field to experimental
second virial coefficient data.) Indeed, there are no mixture parameters to adjust as
all the interaction information comes from quantum mechanics. This example shows
the subtlety of phase behavior that can be predicted with a completely “first princi-
ples” approach of quantum mechanics and molecular simulation.
The use of quantum mechanics for large molecules, polymers, proteins or repeat-
ing small molecule structures (such as zeolites, hydrates or graphite sheets) is more
difficult because of the O(N) problem. There are two typical approaches to deal with
this problem. The first is to “divide and conquer”, i.e., to split the molecule into seg-
ments (usually capping broken covalent bonds with hydrogen atoms), and then
50 Chapter 5
100
1
no scalin
no polarizagion
\
\ n
z
n
Y
10 r
polarization//
no scaling $//’/
y’N =1.20
n
2
72
1:
\“ ,/ d//
non-polarization
/Lo
0.1 :
/ polarization
a)
3.8 3.4 3.0 2.6 2.2
9
450 A
s
2
3
c
2
Q
Q,
f 0
t- 350
25 0
0.0 0.2 0.4 0.6 0.8
Density (g ~ r n - ~ )
Figure 5 Vapor pressure and vapor-liquid phase boundary for methanol as a function of
temperature from experiment (lines) and from simulation with combinations of scal-
ing and polarization
adding up the contributions. The second is to recognize (e.g., in protein docking) that
there may be a few interaction sites that are especially important, while distant parts
of the molecule are less so. The important regions of the molecule are then treated
with a high level of theory, and the distant regions with less accurate methods, or
I
0.0 0.2 0.4 0.6 0.8 1.0
0.9
0.8
0.7
0.6
0.5
1
0.4 I I I I
0.0 0.2 0.4 0.6 0.8 1.0
(b) Xrnethanolp Yrnethanol
Figure 6 Vapor-liquid equilibrium for the methanol + acetonitrile system. The unfilled
points indicate experimental data, and the$lled points indicate the predictionsfrom
simulation
with the density functional theory, and the results are combined. Examples of this
approach include the ONIOM method l2,I3 and others.
Computational packages such as Gaussian, l 4 Jaguar15and others are available for
the quantum mechanical calculation of the interactions energies, but it is very inten-
sive, and requires the advice of an expert, as the results obtained depend upon the
level of theory and basis set used, and it is not always obvious as to which set of
52 Chapter 5
choices is an optimal compromise between the computational resources required and
accuracy. Although there are some molecular simulation packages available com-
mercially,16 most simulators use programs that they have written.
An alternative to the computationally intensive method of developing force fields
from quantum mechanics has been to use empirical potentials, either transferable
potentials that are meant to be used for many compounds (e.g., 0PLS,l7AMBER,'*
TraPPE,19 etc.) or specialized potentials for specific compounds (e.g., SPC,20
TIP4P2I and later models for water). Such empirical potentials have been fit (fre-
quently using simulation) to some experimental data. The results in Figure 7 for
methanethiol illustrate the potential inaccuracies of using transferable potentials.22
There, we see that using a potential function fit to the quantum-mechanically
l!
3.4 3.2 3.0 2.8 2.6 2.4 2.2 2.0
1OOOflemperature (K-')
0.0 0.2 0.4 0.6 0.8

Density (g ~ r n - ~ )
Figure 7 The vapor pressure and liquid-vapor coexistence curves (including estimated crit-
ical points)for methanethiol from simulation and experiment. The square symbols
are obtained using the quantum-mechanical-based potential in simulation, the tri-
angles from using the OPLS-UA potential, and the solid line shows the smoothed
experimental data. The filled symbols are measured and estimated critical points
(re$ 22)
calculated energies yields quite good predictions of the phase behavior and vapor
pressure of this fluid, while the results using the generalized OPLS potential in sim-
ulation are significantly in error. When such a potential fails for a system, there is no
recourse except to use additional experimental data and refit, perhaps adding more
terms to the potential.
3 Approximate Quantum Mechanical Calculation of

Thermodynamic Properties
The completely “first principles” way of computing thermodynamic properties
described in the preceding section results in the O(N) scaling problem if a pair of
large molecules are considered. A way around this is to use theories or models that
require only a single molecule quantum mechanics calculation. [A high-level energy
calculation for a single molecule takes less than one-hundredth (actually lD7) the
time of a two-molecule calculation, and hundreds of two-molecule calculations are
needed to develop a pair potential; thus, the computational load is increased by a fac-
tor of 10,000 or more.]
One of the most well developed of the class of theories that require only a single-
molecule quantum mechanics calculation is the polarizable continuum model
(PCM).23-25The goal of these models is to calculate the free energy of solvation
(sol), i.e., the free energy change of bringing a molecule at fixed orientation from
vacuum into a solvent that is treated as a polarizable continuum. This free energy
change is computed as the sum of three terms:
where AGcavis the free energy required to create a cavity in the solvent to accept the
solute, AGvdwis the contribution from the repulsion and dispersion forces between
the solute and the continuum solvent, and AGelecis the electrostatic contribution that
is computed from quantum mechanics. The last term is small for nonpolarizable
molecules ( i e . , alkanes), but very important for polar solutes; it depends on the
dielectric constant of the solvent, and is typically computed using the density func-
tional theory. Once AG,,, is known, the vapor pressure and some other pure compo-
nent thermodynamic properties can be computed directly; also, if AG,,, is known for
inserting the solute both into its own pure liquid and into a (pure or mixed) solvent,
mixture properties such as the activity coefficient of the solute can be obtained, from
which the phase behavior can be computed.
While the quantum mechanical calculation of the AGelecis reasonably well estab-
lished, the term with the greatest uncertainty, which therefore limits the accuracy of
the method, is the cavity formation term AGcav.Various approximations have been
made for this term, including the use of spheres and nonspherical shapes to repre-
sent the solute with atomic radii empirically adjusted or based on crystal structure.
However, as this term is less important for mixture properties (as AGcavfor a solute
in its pure liquid and in the solvent to some extent cancel), PCM has been more use-
ful for the calculation of activity coefficients, octanol-water partition coefficients
54 Chapter 5
T= 323.15k
- -- ---.
.-.------------
41 A
30 -COSMO-SAC
Expt 323.15K
I T=308.15k
350 1 1
2 200
E
E
150
100
50
0
0 0.2 0.4 0.6 a8 1
x1 J1
Figure 8 ( a ) Comparison of vapor-liquid equilibrium predictions from COSMO-SAC, UNI-

FAC and mod$ed UNIFAC models for water(1) + 1,4-dioxane(2)mixtures at tem-
peratures 308.15 and 323.15 K, and (b) vapor-liquid equilibrium prediction from
COSMO-SAC for benzene(l)/n-methylformamide(2)at temperatures of 31 8.15 and
328.15 K
and other mixture properties than for the calculation of pure component properties.
The most successful PCM model is COSMO-RS (Conductor-like screening model
for real solvents) of Klamt and co-workers26-28and others,29in which AGeI, is cal-
culated from DFT for transferring a molecule from vacuum to a perfect conductor
(infinite dielectric constant), and then an analytic correction is made for the dielec-
tric constant of the actual solvent. A significant advantage of this approach is that the
quantum calculation is for the solute in a perfect conductor regardless of the solvent,
so that it needs to be done only once for each solute, and stored for later use with
any pure or mixed solvent. Two examples of phase behavior predictions based on
this method are given in Figure 8. This method is new and is still being developed,
and will probably replace group contribution methods such as UNIFAC30 and mod-
ified UNIFAC3' in the future; a commercial product is already available.32Most
importantly, one does not have to be an expert in quantum mechanics to use this soft-
ware, and presently it is the most useful and efficient method of using quantum
mechanics to directly compute phase behavior.
4 Conclusions
Computational quantum mechanics continues to be a rapidly developing field, and
its range of application, and especially the size of the molecules that can be studied,
progresses with improvements in computer hardware. At present, ideal gas proper-
ties can be computed quite well, even for moderately sized molecules. Complete
two-body force fields can also be developed from quantum mechanics, although
generally only for small molecules, and this requires the study of pairs of molecules
in a large number of separations and orientations. Once developed, such a force field
can be used to compute the second virial coefficient, which can be used as a test of
its accuracy, and in simulation to compute phase behavior, perhaps with corrections
for multibody effects. However, this requires major computational effort and expert
advice. At present, a much easier, more approximate method of obtaining condensed
phase thermodynamic properties from quantum mechanics is by the use of polariz-
able continuum models based on COSMO calculations.
References
1. K. Morokuma, in Applications of Molecular and Materials Modeling. P. R. Westmoreland
et al. (eds), Kluwer Academic, New York, 2002.
2. P. v. R. Schleyer, N. L. Allinger, T. Clark, J. Gasteiger, P. A. Kollman, H. F. Schaefer III and
P. R. Schreiner, The Encyclopedia of Computational Chemistry, Wiley, New York, 1998.
3. A. R. Leach, Molecular Modelling: Principles and Applications, 2nd edn, Prentice Hall,
New York, 200 1.
4. D. Frenkel and B. Smit, Understanding Molecular Simulation, 2nd edn, Academic Press,
New York, 2002.
5. R. Car and M. Paninello, Phys. Rev. Lett., 1985, 55, 2471.
6. N. R. Kestner and J. E. Combariza, in Reviews in Computational Chemistry, Vol 13,
K. B. Lipkowitz and D. B. Boyd (eds), Wiley-VHC, New York, 1999.
7. S. I. Sandler, S. T. Lin and A. K. Sum, Fluid Phase Equalibr., 2002, 194,61.
8. P. K. Naicker, A. K. Sum and S. I. Sandler, J. Chem. Phys., 2003,118,4086.
9. A. K. Sum, S. I. Sandler, R. Bukowski and K. Szalewicz, J. Chem. Phys., 2002,116,7627.
10. A. Z. Panagiotopoulos, Molec. Phys., 1987,61, 813.
11. C. J. F. Bottcher, Theory of Electric Polarization 2nd edn, Elsevier, Amsterdam, 1973.
12. F. Maseras and K. Morokuma, J. Comp. Chem., 1995, 16, 1170,
13. T. Vreven and K. Morokuma, J. Chem. Phys., 1999,111, 8799.
14. M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman,
V. G. Zakrzewski, J. A. Montgomery, R. E. Stratmann, J. C. Burant, S.Dapprich, J. M.
56 Chapter 5
Millam, A. D. Daniels, K. N. Kudin, M. C. Strain, 0. Farkas, J. Tomasi, V. Barone, M.
Cossi, R. Cammi, B. Mennucci, C. Pomelli, C. Adamo, S . Clifford, J. Ochterski, G. A.
Petersson, P. Y. Ayala, Q. Cui, K. Morokuma, D. K. Malick, A. D. Rabuck, K.
Raghavachari, J. B. Foresman, J. Cioslowski, J. V. Ortiz, B. B. Stefanov, G. Liu, A.
Liashenko, P. Piskorz, I. Komaromi, R. Gomperts, R. L. Martin, D. J. Fox, T. Keith, M.
A. Al-Laham, C. Y. Peng, A. Nanayakkara, C. Gonzalez, M. Challacombe, P. M. W. Gill,
B. G . Johnson, W. Chen, M. W. Wong, J. L. Andres, M. Head-Gordon, E. S. Replogle and
J. A. Pople, Gaussian 98, Revision A. 9; Gaussian, Inc: Pittsburgh, PA, 1998.
15. Jaguar 5.0 User Manual, available at http://www.schrodinger.com/docs/jaguar5.0/pdf/
j 50-user-manual.pdf
16. Accelrys Materials Design Studio, information available at http://www.accelrys.
com/mstudio/
17. W. L. Jorgensen and J. Tiradorives, J. Am. Chem. SOC., 1988, 110, 1666.
18. W. D. Cornell, P. Cieplak, C. I. Bayly, I. R. Gould, K. M. Merz, D. M. Ferguson, D. C.
Spellmeyer, T. Fox, J. W. Caldwell and P. A. Kollman, J. Am. Chem. Soc., 1996, 118,
2309; and later papers by Kollman et al.
19. B. Chen, J. J. Potoff and J. I. Siepmann, J. Phys. Chem. B., 2001, 105, 3093; and previ-
ous papers by J. I. Siepmann et al.
20. H. J. C. Berendsen, J. P. M. Potsama, W. F. van Gunseren and J. Hermans, in Intermolecular
Forces, B. Pullman Reidel (ed), Dordrecht, 1981.
21. W. L. Jorgensen, J. Chandrasekhar, J. D. Madura, R. W. Impey and M. L. Klein, J. Chem.
Phys. 1983, 79,926; and later papers by Jorgensen et al.
22. S. Garrison, private communication, 2003.
23. J. Tomasi and M. Persico, Chem. Rev. 1994, 94, 2027.
24. C. Amovilli and B. Mennucci, J. Phys. Chem. B, 1997,101,1051.
25. C. J. Cramer and D. G. Truhlar, J. Computer-Aided Molecular Design, 1992,6,629; and
later papers by the same authors.
26. A. Klamt, J. Phys. Chem., 1995,99,2224.
27. A. Klamt, V. Jonas, T. Briiger and J. C. Lohrenz, J. Phys. Chem. A, 1998,102,5074.
28. A. Klamt and F. Eckert, Fluid Phase Equalibr, 2000, 172, 43; and later papers by these
authors.
29. S.-T. Lin and S.I. Sandler, Znd. Eng. Chem. Res., 2002,41, 899.
30. A. Fredenslund, R. L. Jones and J. M. Prausnitz, A.Z.Ch.E. J., 1975, 21, 1086; and the
many publications that followed.
31. J. Gmehling, J. Li and M. Schiller, Znd. Eng. Chem. Res., 1993, 32, 178; and the many
publications that followed.
32. http://www.cosmologic.de/products.html
CHAPTER 6
Thermodynamics of Natural Gas

Clathrate Hydrates
E. DENDY SLOAN, JR.
1 Introduction
Natural gas will be the premium fuel for this century for three reasons: (1) while we
wish to use hydrogen as a fuel, we are economically restricted from obtaining it by
reforming light hydrocarbons (with attendant inefficiencies) until we learn to effi-
ciently electrolyze water, (2) gas bums cleanly, causes few pollution problems and,
relative to oil or coal, produces less carbon dioxide, and (3) liquid fuels are better
used as feedstocks for petrochemicals.
Yet when we consider natural gas as a resource, we invariably find it associated
with water. For example, when natural gas is produced from a Gulf of Mexico deep-
water well, flowline design is driven by concurrent water production to prevent
gas+ water from combining into a solid clathrate hydrate, which blocks the pipeline.
As a second example, when dead plant and animal matter decompose on the bottom
of the ocean, they form methane, which combines with water to form a clathrate
solid. The amount of enclathrated methane in the bottom of the ocean is between
1015 and 1017 m3 (STP) - two orders of magnitude greater than the conventional
worldwide gas reserve.
Important application questions arise due to the profusion of natural gas hydrates:
1 . What defines economic flow assurance in deepwater energy production?

2. Is it possible that in situ gas hydrate reserves can be an economic energy
resource?
3. Can subsea stability be affected by hydrate dissociation, affecting not only sub-
sea structures but also climate change, if hydrated methane evolves to the
atmosphere?
4. Is it possible to storehransport hydrated gas to onshore gas distribution, from
a gas reserve that is too small to justify liquefaction, or too far from a pipeline?
5. What role do hydrates play in the environments of other planets?
58 Chapter 6
Answers to such difficult questions can be found in applied thermodynamics - in
terms of measured, macroscopic values of pressures, temperatures, compositions,
volumes, enthalpies, etc. This chapter provides an overview of natural gas clathrate
hydrates - structures, phase diagrams, and thermodynamic predictions/measurements
that guide our understanding in dealing with such questions. The hydrate historical
perspective provides an example of how knowledge advances in a technical field. At
the conclusion of the chapter, future thermodynamic challenges are presented.
2 What are Natural Gas Clathrate Hydrates?

Clathrate hydrates form when small (<0.9 nm) non-polar molecules contact water
at ambient temperatures (typically <300 K) and moderate pressures (typically
>0.6 MPa). On a molecular scale, single small guest molecules are encaged
(enclathrated) by hydrogen-bonded water cavities in these non-stoichiometric
hydrates. Guest repulsions prop open different sizes of water cages, which combine
to form three well-defined unit crystals shown in Figure 1.
Cubic structure I predominates in the earth's natural environments with small
(0.4-0.55 nm) guests and cubic structure I1 generally occurs with larger (0.6-0.7
nm) guests in mostly man-made environments. Hexagonal structure H may occur in
either environment, but only with mixtures of both small and the largest (0.8-0.9
nm) molecules. The smallest hydrated molecules (Ar, Kr, 0, and N2) with diameters
46 HZO
/
- 16
51262
___)
8
Structure I
136 H2O
512 \ 3 '
5 264
+ Structure I
- -
2 1
34 H2O
435663 51268
Structure H
Figure 1 The three common clathrate hydrate structures, including the constituent cavities.
Nomenclature: 5'268 indicates 12 pentagonal and 8 hexagonal sides in a cavity;
numbers along lines indicate the number of cavities in each unit crystal structure.
The rightmost numbers indicate the water molecules per crystal structure
Thermodynamics of Natural Gas Clathrate Hydrates 59
Table 1 Geometry of Cages in Three Hydrate Crystal Structures in Figure 1
Hydrate Crystal Structure Z zz H
Cavity Small Large Small Large Small MediumLarge
Description 512 51262 512 51264 512 435663 51268
Number of cavitiedunit cell 2 6 16 8 3 2 1
Average cavity radius, A 3.95 4.33 3.91 4.73 3.91" 4.06" 5.71"
Coordination numbe9 20 24 20 28 20 20 36
Number of waterdunit cell 46 136 34
aEstimates of structure H cavities from geometric models.
bNumber of oxygens at the periphery of each cavity.
less than 0.4 nm form structure 11. Most of hydrate science, and thus most applica-
tions, concentrate on structure I (sI) and structure I1 (sII), with structure H (sH) in
anecdotal occurance. While this work will emphasize sI and sII applications, many
analogs occur for sH hydrates.
Table 1 presents properties of the three common unit crystals shown in Figure 1.
Water molecules hydrogen bond in a basic building block for both SI and sII, called
the 512 (pentagonal dodecahedra) because there are 12 faces of pentagonally bonded
water molecules in that cavity. Within the cavity, small guest molecules are
enclathrated, with limited translation motion, but substantially more rotation and
vibration, almost that of the gas. The 512 building blocks are joined to other 512cav-
ities either through the vertices (sI) or via the 512 faces (sII).
Yet, because the 512 basic building blocks cannot fill space without severe hydro-
gen bond strain and breakage, interstices between the 512 cages are filled with other
cavities that relieve the strain by incorporating additional hexagonal faces with the
12 pentagonal faces - two hexagonal faces in the 512ti2cavity of SI and four hexago-
nal faces in the 51264of sII, in addition to the 12 pentagonal faces in each cavity. Thus,
the SI and sII cages can contain larger guest molecules in decreasing order: 51264(in
sII), 51262(in sI), and 512 (almost the same size in both SI and sII). The cages form
basic repeating unit crystals with ratios of 2 X 512+6X 51262 with 46 H 2 0 molecules in
sI, and 16X512+8X51264and 136 H 2 0 molecules in the sII unit crystal. In all three
common structures, typically only one guest molecule forms within each cage.
When all hydrate cavities are filled, the three crystal types (I, 11, and H) have
remarkably high and similar concentrations of components: 85 mol % water and 15
mol % guest(s). Hydrate formation is most probable at the interface of the bulk guest
and aqueous phases because the hydrate component concentrations greatly exceed
the mutual fluid solubilities. The solid hydrate film at the interface acts as a barrier
to further contact of the bulk fluid phases, and fluid interface renewal is required for
continued, rapid clathrate formation. Three phase interfacial hydrate formation
occurs with equilibrium of gas, liquid, and hydrate phases in artificial situations such
as laboratories or in man-made processes.
3 Development of Hydrate Thermodynamics

After their discovery and attendant interest, hydrate thermodynamic advances have
been driven by practical concerns. In this section, motivating factors are shown with
60 Chapter 6
historical advances in three periods: (1) initial discovery and academic interest, (2)
discovery of hydrates in pipelines, and (3) discovery of hydrates in nature. The appli-
cations of each of the three periods still exist as foundations for hydrate thermody-
namic advances.
3.1 Discovery and Academic Interest

While the clathrate hydrate discovery was announced in Davy’s Bakerian Lecture,’
Priestley2 apparently discovered SO, hydrates some years before, but failed to
repeat, or pursue the original experiment. Davy’s hydrates of oxymuriatic acid (now
known as chlorine) were succeeded by hydrate discoveries of methane, ethane, and
propane through 40 years of efforts by Villard and deF~rcrand.~ These early
researchers were intrigued by two aspects of these unusual hydrates: (a) the ratio of
guest-to-water molecules (as high as 15:85) was very high relative to most other
combinations of hydrocarbons and water, and (b) the fact that the ratio varied with
each experiment - the hydration number changed from fractional values between 6
and 8, or higher for larger molecules such as propane. In addition, this era concen-
trated on finding a comprehensive list of chemicals that formed these unusual non-
stoichiometric hydrates.
More recently, academic hydrate research has moved from gentlemanly pursuits
of science to being pragmatically driven. The rare research work stresses hydrate
interest due to insights into water potential^.^ Currently, academic research is driven
by funding vehicles, to provide answers to applications such as the five mentioned
in the introduction. Fortunately, while pursuing single-minded goals, intellectually
challenging research discoveries occur, usually providing the best driving force for
hydrate advances.
Consider the notable example of hydrate researcher D.N. Glew, who first pub-
lished on ethylene oxide hydrates5 with Dow of Canada in 1959. Upon his retire-
ment, his interest increased, as did his productivity in hydrates. In 2003, Glew
published his 19th hydrate article6 in a series that provided the foundation for many
new advances, including the new methane +water phase diagram discussed in
Section 4.
3.2 Thermodynamics Driven by Discovery in Flowlines

The discovery in 1934 of hydrates blocking energy flowlines led to a much more
intensive period of hydrate research. Without knowledge of the hydrate structures,
researchers generated methods to predict and to inhibit the conditions of hydrate
formation in flowlines and to prevent them. Six major efforts, driven by pipeline dis-
covery, hallmark this period.
1. In 1934, Hammer~chmidt,~ having discovered hydrate formation in pipelines,

applied colligative physical chemistry to quantify prevention chemical
effects. His discovery of alcohols, glycols, and other inhibitors and his
equation for hydrate formation temperature suppression still serve as indus-
trial standards.
2. The first systematic phase equilibrium study of hydrates was carried out at the
US Bureau of Mines by Deaton and Frosts for a number of natural gas hydrate
systems. Although the separation methods were somewhat crude (e.g.,a distinc-
tion could not be made between butane isomers), their hydrate formation pres-
sure and temperature data continue to be foundations for prediction comparisons.
3. In the 1940s, Katz had the insight that hydrates were ideal solid solutions rather
than stoichiometric compounds. With the ideal solid solution concept, Katz and
his students generated two experimentally based prediction m e t h o d ~ ~for t'~
hydrate formation conditions, without the knowledge of different hydrate
structures.
4. The first phase diagrams of hydrocarbons and water that contained hydrates
were generated by Kobayashi and Katz for methane+water," and for
propane +water. l 2 These diagrams were substantial advances in our under-
standing of water +hydrocarbon systems. Yet, the diagrams specified single,
stoichiometric hydrate concentrations because the non-stoichiometric nature of
hydrate structure was unknown. The intuitive insights from such diagrams
were accompanied by a pragmatic, useful thermodynamic summary of hydro-
carbon+water systems in Chapter 5 of the Handbook of Natural Gas
Engineering by Katz et al.13
5. In a heroic effort after World War 11, von Stackelberg and his students at the
University of Bonn determined powder X-ray diffraction patterns for many
hydrates, particularly sI and sII. Using von Stackelberg's diffraction pat-
t e r n ~ the
,~~unit crystals of sI and sII were rapidly determined in the early
1950s by C l a u s ~ e n ' ~and
> ' ~by Pauling and Marsh.17 It is interesting that, von
Stackelberg's hydrate classification forced all of his data into the sI and sII
paradigms, yet his data contained indications of hydrate structure H, dis-
counted because they did not conform to the sI or sII formats.
6. The idea of a fixed crystal structure in which single cages contained at most
one guest proved irresistible to statistical thermodynamicists. After an initial
effort by Powell,l8 Royal Dutch Shell workers van der Waals and PlatteeuwI9
generated a method that still stands today as a principal, regular industrial use
of statistical thermodynamics. However, the model was not suitable for man-
ual calculations (as were the methods of Katz in item 3 above), but required
access to then-scarce computers, which limited its application to large compa-
nies or major universities. Widespread adoption of the model awaited the pro-
liferation of personal computers.
The van der Waals and Platteeuw model represents the largest major advance in
the thermodynamic history of hydrate technology. With this model, single hydrate
formation data could not only be interpolated and extrapolated but also they could
be extended to accurate mixture formation condition predictions. In Section 4, the
details of this model are considered in light of modern hydrate phase measurements
and modeling.
62 Chapter 6
In retrospect, the discovery of hydrates blocking energy flowlines motivated sig-
nificant advances in clathrate hydrate thermodynamics, which may be typical of
knowledge advances for many new compounds, with the following stages:
1. observation of initial phenomena with methods of prevention,

2. measurement of macroscopic formation temperatures and pressures,
3. generation of macroscopic prediction methods to interpolate and extrapolate
data,
4. placement in the thermodynamic knowledge base via construction of phase
diagrams,
5 . measurement and determination of microscopic crystal structure, and
6. construction of a statistical thermodynamics method to connect the micro-
scopic and macroscopic domains.
The discovery of hydrates inside flowlines provided insights into hydrate physics
and chemistry, which allowed application outside the pipelines. Even today, ques-
tions in flow assurance provide the most constant vehicle for hydrate research.
However, the discovery of hydrates outside of flowlines has provided the most recent
motivations for advances in hydrate science.
3.3 Thermodynamics Driven by Hydrates Outside Flowlines

The Russian discovery of hydrates in nature spawned the third wave of advances in
hydrate research. Ten major applications and basic research advances occurred in the
modern time.
1. Makogon20 discovered in situ hydrates in porous media, leading to an

instance of Siberian gas production from hydrates associated with normal gas
reservoirs. Beginning in 1980, Kvenvolden and colleagues*' published sur-
veys of in situ hydrates, independent of Makogon, but arriving at a consensus
figure of 1015m3 (STP) of methane in hydrates. The intellectual heirs of such
early workers are Ginsburg and Soloviev22in Russia, Dillon and C01lett~~ at
the United States Geological Survey (USGS), and Hyndman and D a l l i m ~ r e ~ ~
at the Canadian Geological Survey.
2. Kobayashi and c o - ~ o r k e r began
s ~ ~ to use the van der Waals and Platteeuw
theory in the 1960s to predict hydrate formation in ternary systems. Parrish
and Prausnitz26extended the method to prediction of hydrate incipient for-
mation in natural gas systems causing the widespread industrial adoption of
the van der Waals and the Platteeuw statistical method. Many academic (e.g.,
Holder and c o - ~ o r k e r s ~and
~ ) commercial programs (e.g., D.B. Robinson,
and Associates2*)enabled the gas and oil industry to predict thermodynamic
conditions at the incipient formation point, and thereby to prevent hydrate
formation in industrial processes. All of the errors in the van der Waals and
Platteeuw theory were placed in the solid phase. It may be argued that the the-
ory's unusual success in prediction inhibited motivations for advances in
hydrate phase measurements.
3. Yet, beginning in the 1960s at the Canadian National Research Council
(NRC), the group initiated by Davidson and extended by R i p m e e ~ t e rhas
~~
made the best modern measurements in hydrate science, with particular
emphasis on nuclear magnetic resonance (NMR) spectroscopy, neutron dif-
fraction, calorimetry, and molecular modeling. An abbreviated overview of
the major advances in hydrate science at NRC is presented30in Table 2.
4. NMR success motivated other spectroscopic studies to measure the hydrate
phase directly. This work represented an experimental departure, because pre-
viously only the fluid phases (vapor and liquid(s)) were measured, and any
experimental error was incorporated in the solid-phase model of van der
Waals and Platteeuw. However, with modern solid-phase measurements, the
errors in the van der Waals and Platteeuw model could be clarified and cor-
rected. Raman spectro~copy~~ and diffraction (X-ray and neutron, supple-
mented by Rietveld analysis32) have been successful: the first method to
measure the relative occupation of single guest cages, and the second to
extend the work to hydrate isothermal, adiabatic, and isobaric compressibili-
ties. As shown in Section 4, these measurements combine with spectroscopic
hydrate phase measurements to enable improvements of the model.
5. For computer simulations, in 1972, using early super-computing power at the
USA Los Alamos National Laboratory, Tester and c o - ~ o r k e r performed
s~~
the first Monte Car10 hydrate simulation, followed in 1983 by the first molec-
ular dynamics simulation by Tse at NRC,34as computing power became more
readily available. The most thorough, systematic simulations of hydrates have
been performed by Tanaka and colleagues, who published 40 computer sim-
ulation articles on hydrates since 1984. Building on a foundation of thermo-
dynamic simulations, since 1989 Rodger and c o - ~ o r k e r shas
~ ~extended the
hydrate simulation studies to time-dependent and inhibitor-development stud-
ies. Recently, Klauda and Sandler36 and Trout and c o - ~ o r k e r shave ~~
extended ab initio studies to hydrates.
Table 2 Major Advances in Hydrate Science at NRC Canada

1960-1980 - Dielectric and NMR measurements of hydrate guest and host dynamics; many
new hydrate sI and sII guests identified
1982 - Diffraction studies identify small guests as sII occupants
1983 - 016/018isotope ratio measurements for laboratory hydrate
1983 - First molecular and lattice dynamic calculations of hydrates
1983 - Confirmation of low thermal conductivity values of hydrates
1986 - First measurements of hydrate index of refraction
1987 - First reports of sH hydrate from powder diffraction and N M R
1988 - 13CMAS N M R determined as means of sI and sII hydrate identification
1990 - First published inventory of sH hydrate formers
1992 - First studies of phase equilibria in porous media
1992 - First inelastic neutron scattering of methane hydrate
1999 - Discovery of new hydrate structure - complex polytype of sII and sH, suggesting that
interpenetrating lattices may exist naturally
2000 - New hydrate structure discovered for dimethyl ether hydrate
200 1 - First single-crystal X-ray diffraction for experimental composition measurement
64 Chapter 6
6. Of paramount importance is the reproducible generation of high-quality
hydrate samples. In 1996, Stern et ~ 1 at USGS . ~ independently
~ generated a
reproducible sample preparation method from melting ice, previously deter-
mined by Handa,39 that had gone unnoticed. This was a major advance
because previously, when hydrates formed at the vapor-liquid interface, the
solid hydrate barrier prevented further contact, so that the hydration number
varied between samples, a sample variability that had plagued researchers
since before Villard and deF~rcrand.~
7. A major question arises from in situ hydrates detection, “Is energy recovery
from hydrates viable?’ Two recent drillings - the 2002 Mallik well in
Canadian permafrost (Dallimore et uL40) and the Hydrate Ridge ocean
drilling (Trehu et ~ 1 . ~-’ document
) the technical viability of hydrate recov-
ery. However, except in special circumstances (e.g., national security of very
high gas prices), energy recovery from hydrates alone is not economical from
an energy company’s per~pective,~~ principally due to the dispersed, low-
energy density of the resource. However, energy recovery is viable in special
circumstances, as when hydrates overly a gas reservoir and can replenish the
gas reservoir when the pressure is below the hydrate decomposition condi-
tion. A decade of such production was reported by M a k ~ g o in n ~the
~ Siberian
Messoyaka Field in the 1970s. The Japanese hydrate program led by Japanese
National Oil Companyu will attempt to produce energy from hydrates in the
Nankai Trough by 2015.
8. The discovery of large amounts of in situ hydrates has led to two further
research motivators. First, the community is concerned about seafloor stabil-
ity, and the possibility of endangering the foundations of pipelines and sub-
sea structures, as discussed in the monograph by Paul1 and Dil10n.~~ One
major component of USA Department of Energy’s hydrate research program
concerns hydrate developments in the Gulf of Mexico, whose results can be
extended to other giant oil field sites such as offshore Africa or Brazil.
9. The most recent major application concerns hydrates and climate change.
Using isotopic record data by Kennett et aZ., Dickens and colleagues proposed
that the Late Paleocene Thermal Maximum (LPTM) was caused by gas evo-
lution from hydrates. Most recently, Kennett et aZ.46 have proposed the
“hydrate gun hypothesis,” which suggests that as late as 15,000 years ago,
methane evolution from hydrates may have contributed to global warming.
Such hypotheses, extended to more recent times, have motivated research
programs in hydrate-induced climate change. Much more hydrate research is
needed before such an hypothesis can advance to the stage of being an accept-
able theory.
10. The discovery of hydrates in the natural environments initiated other hydrate
monographs, with diverse topics of carbon dioxide seq~estration,~~ per-
mafrost and oceanic and seafloor stability and climate
change.49However, hydrate research proliferation has led to state-of-the-art
records of the international hydrate community at three-year intervals at New
York50in 1993, T o u l o ~ s ine ~1996,
~ Salt Lake City52in 1999, a n d Y ~ k o h a m a ~ ~
in 2002. The fifth international meeting is scheduled on June 12-16, 2005 in
Trondheim, Norway. The most recent hydrate science is recorded in the four
hydrate sessions of the Tenth International Conference on Physics and Ice.54
Any hydrate researcher will benefit from a review of such literature.
The modem era of hydrate research is marked by the industrial adoption of the van
der Waals and Platteeuw statistical thermodynamics model for the hydrate phase.
The spectroscopic measurement of the hydrate phase, abetted with molecular simu-
lation, led to accuracy improvements, and industrial applications to energy, seafloor
stability, and climate change. With the above historical advances, consider modem
thermodynamics of the hydrate phase itself.
4 Hydrate Thermodynamics
Due to limitations of space, this section emphasizes only two major aspects of hydrate
thermodynamics - namely the phase diagram and the hydrate prediction method. For
examples of several counter-intuitive spectroscopic measurements results that impact
the thermodynamic perspective, the reader is referred to a recent review,55presented
with an overview of the 2002 Fourth International Hydrate Conference.
4.1 The Hydrate Phase Diagram

One might think that a review of a single hydrocarbon+water phase diagram would
be simplistic. Yet, discussion of such a diagram leads to both new and essential
results to understanding of industrial gas mixture hydrate equilibria. In industry, the
hydrocarbon or gas phase is almost always present in large excess for economic
reasons, so that the hydrocarbon gas composition does not change on hydrate
formation, but maintains a fixed composition, adding credibility to the study
simpler diagrams. Here, we discuss the methane+ water diagram56for structure I.
The reader may wish to supplement this discussion with more phase diagram
details.57-58
Kobayashi and Katz" provided the first methane+water phase diagram, as a
major advance in fundamental thermodynamics understanding. However, because
these two researchers knew neither the hydrate structure nor their non-stoichiomet-
ric nature, their diagram did not indicate a variable composition for hydrates. This
early advance was in violation of the phase rule because it suggests that only the
specification of temperature or pressure is required to fix the hydrate composition in
a two-phase condition.
Recently, Raman spectroscopy was used to probe the methane+ water phase dia-
gram further. A 4% decrease in methane hydrate concentration was measured by
Huo et U Z . , ~ ~at the two-phase (H-V) condition, relative to two-phase (LurH)
hydrates formed with dissolved methane in water at the same temperature and pres-
sure. The result is a non-stoichiometric hydrate area in the methane+water isobaric
T-y-z diagram shown in Figure 2, rather than the previously thought single hydrate
composition, shown as a vertical line. Such solid solution areas are common in met-
allurgical phase diagrams. A similar non-stoichiometric hydrate region has been
66 Chapter 6
proposed59for the phase diagram of CO,+H,O, in which CO, also occupies both
cavities of the sI hydrate lattice.
The system is constrained to obey the Gibbs Phase Rule? In the most common
application, with methane (or a fixed gas composition) and water, the two compo-
nents will form the three phases of hydrates, vapor, and liquid with one degree of
freedom. Fixing the temperature (or the pressure) determines the pressure (or the
temperature) at the three-phase condition.
For example, in Figure 2 the horizontal line connecting the three phase V+L,+ H
(vapor+liquid water+ hydrate) phases indicates a single equilibrium temperature for
the three phases in the isobaric diagram. This univariance allows one to apply the
Clapeyron equation to obtain the useful relation of heat of hydrate dissociation from
phase measurements of pressure and temperature, as verified by ~alorirnetry.~~
In contrast, with two-phase formation, without a free water (or a free vapor)
phase, an additional degree of freedom, such as temperature, pressure, or phase
composition, must be applied. In Figure 2, the two-phase areas (V+H, &+H,
&+H, M + H , or H+I) indicate that two variables (temperature of the single-phase
composition, in addition to the isobar of the diagram) must be specified to fix the
system state. Hydrate formation from a water-saturated vapor counters the common
myth that a free water phase must be present to have hydrates. That there is no ther-
modynamic necessity for a free water phase is perhaps shown most clearly by a
study of the methane+water phase diagram.
~ ~~
V vapor
Lw: liquid water
H: hydrate
M: solid methane
LM:liquid
methane
I: ice
v-L,
Previous hydrate line
H,O CH4.?H,0 CH4

Figure 2 The phase diagram of CH, +H,O. Note that the non-stoichiometric hydrate region
replaces the previously proposed vertical line for stoichiometric hydrate concen-
tration
In Figure 2, there are no scales on either axis, and the diagram is only qualitatively
shown. In the figure, the smaller concentration regions have been magnified for the
sake of clarity. For example, at 298 K, the concentration of methane dissolved in
water is typically 8.3 X and the concentration of water vaporized in the gaseous
methane is 1X In a quantitatively correct Figure 2, both solubility lines would
be vertical, at either extreme of the abscissa. So the qualitative Figure 2 is presented
to illustrate the principles, trusting that the reader will turn to a reliable phase
method as shown in the following subsection to quantify the diagram.
Also, in Figure 2, hydrate formation from the two-phase regions is clearly per-
mitted, and in fact comprises a large portion of the phase diagram. However, due to
the small concentrations of water in the methane phase, or even smaller concentra-
tions of methane in the water phase, it may be unlikely that kinetics will allow suf-
ficient hydrate formation, for example, to plug a flow channel. Therefore, the
restriction on a free water phase to form a hydrate plug is a kinetic restriction rather
than a thermodynamic one.
The phase diagram in Figure 2 has implications for mesoscopic structures. Kuhs61
has recently reported a crystalline morphology difference; porous hydrates formed
with excess methane to the right of the non-stoichiometric region, while non-porous
hydrates formed to the left.
4.2 The Most Important Advance: The Statistical Thermodynamic

Model
The major advance in hydrate thermodynamics was the generation of the van der
Waals and Platteeuw model bridging the normal macroscopic and microscopic
domains. Only a brief overview is given here to provide a basis for model improve-
ments; the reader interested in more details should refer to another source.62The
essence of the van der Waals and Platteeuw model is the equation for the chemical
potential of water in the hydrate phase:
where pwH the chemical potential of water in hydrate gwpthe molar Gibbs free energy
of water in the empty hydrate, k the Boltzmann’s constant vi the number of type i cav-
ities per water molecule for sI: v,= 1/23, vL1=3/23; sII:vs=2/17, vL2=1/17; and e,. I
the occupancy of a hydrate cavity of type i with a molecule of type m.

Equation (1) indicates that when a guest fills the cavities of a hydrate, the chemi-
cal potential of water in the cage is lowered, thereby stabilizing the hydrate phase.
In principle, Equation (1) solves the hydrate prediction dilemma - that is, the hydrate
formation conditions are determined by the pressures and temperatures that cause
equality between the hydrate chemical potential of water in Equation (l), and the
chemical potential of water in the other phase(s), as determined by separate equa-
tions of state. There are two important terms on the right: gwB the molar Gibbs free
energy of water in the empty hydrate, and Qmi the filling of the cavity as a function
of temperature and pressure. These details are not vital to the subsequent discussion,
but are detailed elsewhere.62
68 Chapter 6
The van der Waals and Platteeuw principal equation (1) has five assumptions,
which enable refinement of the theory:
1. No quantum effects are needed; classical statistics are valid.

2. Each cavity encages at most one guest molecule.
3. The internal partition function of guest molecules is the same as that of an ideal
gas.
4. There are no interactions of the solute molecules, i.e., the energy of each
encaged guest molecule is shielded by the surrounding water cage, and is
therefore independent of the number and types of other solute molecules.
5. The host molecules’ free energy contribution is independent of cavity occupa-
tion. This assumption implies that encaged molecules do not distort the cavity.
The first major success of the theory is that Equation (1) can be used to fit single
guest hydrate dissociation pressures and temperatures. This fit of cavity occupation
(yki) provides temperature- and pressure-independent Kihara potential parameters
( E and 0)for interactions between the guests and host cages. The fit enables inter-
polation between (and extrapolation beyond) single guest data for the eight common
hydrate formers in gas and oil systems (methane, ethane, propane, iso- and normal
butane, nitrogen, carbon dioxide, and hydrogen sulfide).
The second major, more important success of the theory is that, using the fits to
single component data, the theory allows predictions of mixture conditions for any
combination of pure components. Due to the ideal solution, solid nature (assump-
tions 4 and 5 ) fits of pure component data provide acceptable predictions of all mix-
ture conditions. Since laboratory data are costly both in terms of time and funds
(e.g., US$2,000 per data point), it is very efficient to predict rather than to measure
each hydrate condition. It is important to emphasize that, in contrast to vapor-liquid
equilibria predictions, no interaction parameters are needed for mixture prediction.
The above two successes have served the energy industry well for the last 40
years. Before the modifications indicated in the next section, the hydrate formation
temperature prediction error to within 0.7 K (average absolute error for uninhibited
hydrates) was acceptable, approximately twice the experimental error. With modern
modifications of the van der Waals and Platteeuw theory, the energy industry feels
confident in making multimillion dollar hydrate formation decisions based on the
predictions, without obtaining data for important applications, such as:
1. Prediction of the temperatures and pressures of hydrate formation for an infi-

nite variety of mixtures of the above eight components. This enables specifi-
cation of flowline insulation or heating, to prevent fluids from entering the
hydrate formation region.
2. Quantification of the effect of thermodynamic inhibitors (such as methanol,
glycol, or salts) on the operation of processes. This determines the amount of
inhibitor to be inserted into the free water phase to prevent hydrate formation
from vapor and free water.
3. Specification of the water content to which a gas or liquid hydrocarbon must
be dried to prevent hydrate formation, without a free water phase.
Thennodynamicsof Natural Gas Clathrate Hydrates 69
4.2.1 Improvements to the Original Statistical Thermodynamics Theory
The van der Waals and Platteeuw theory’s success is due to the general applicability
of its five fundamental assumptions. Because the program was used mostly at low
pressures in gas transportation and processing applications, it was fairly accurate at
those pressures. However, the need for improvements was apparent at pressures
above 20 MPa. Theoretical improvements have arisen by correcting the five assump-
tions, considered here in order. Modem experimentalists wish to supplement the nor-
mal macroscopic measurements with microscopic measurements, typically
involving diffraction, Raman, and NMR spectroscopy. With the three spectroscopic
methods, measurements of the solid phase not only provide hybrid e~perirnents,~~
but they also suggest improvements to microscopic models, and mitigate the earlier
error of measuring the fluid phases and inserting all of the error in the hydrate phase.
With the theoretical corrections in this section, prediction inaccuracies are reduced
to values approaching experimental inaccuracies.
The first assumption neglecting quantum effects stands unassailed. For the second
assumption, under unusual conditions such as very high pressure, it is possible to
have multiple-cage occupancy with unusual guests. For example, Mao et aL6,
recently showed that hydrogen can form hydrates at a very high pressure with as
many as two occupants in the small cage and four in the large cage of hydrate struc-
ture 11. However, these small guests and multiple occupancies are considered an
aberration rather than the norm.
With respect to the third assumption, for a number of years, Tanaka65has shown
that the internal particle partition function of the guest molecules differ significantly
in the cages, with restricted rotation and vibration, particularly for those molecules
larger than methane. Indeed, some restriction on such motions is the basis, for exam-
ple, in Raman spectroscopic determination of differing environments of the methane
molecule, in the gas, in solution, and in the hydrate cages.66
Corrections to the last two assumptions have been the most beneficial in improv-
ing the theory. Together, the two assumptions (4. no interactions between guests, and
5. host molecules’ energy being independent of cavity occupation) comprise the
ideal solid solution theory, which has been satisfactory for almost 40 years.
However, a recent showed that a 0.5% change in the lattice parameter can
change the pressure prediction (at P-100 bar) by 15%.
The three ways in which the ideal solid solution assumptions are in error are based
upon spectroscopic measurements of effects of pressure, temperature, and guest
composition on hydrates:
1. Pressure. Using high-pressure neutron diffraction, Kuhs and colleagues32meas-

ured the isothermal compressibilities of CH4+H20 and CH4+D20hydrates to
be 9.11 and 8.21 GPa, respectively, at 271 and 273 K. While these values con-
flict slightly with Helgerud’s68measured adiabatic compressibility (9.01 GPa at
273 K for CH, and H20), they provide quantitative evidence that hydrates are
compressible.
2. Temperature and composition. Universal thermal expansivities of hydrate
structures I and I1 have been measured for many common g ~ e s t sto~ within
~,~~
70 Chapter 6
0.05% error. They showed that the hydrate lattice parameter increases signifi-
cantly in proportion to both guest size and temperature.
3. Composition. Mooijer-van den Heuvel et aL7* have confirmed that when the
larger cages (e.g., 512 64) are occupied by larger guests, the unit crystal lattice
parameter increases. This lattice size increase also increases the small 512cage
size, causing a decrease in Raman frequency - as much as 4 cm- for methane
in the 512 when the largest molecules occupy the 512 64.In addition to address-
ing the ideal solution limit of how one guest molecule size affects the environ-
ment of another, this work suggests a limit to Raman spectroscopy, so that
binary guest compositions cannot be determined a priori via integration of
guest peaks. Instead, a method like NMR spectroscopy should be used to deter-
mine the relative cage occupation for >1 type of guest molecule.
The above three errors can be seen as second-order corrections to a theory that has
served the hydrate thermodynamics community well. To account for these second-
order changes, Equation (1) must be slightly modified as follows:
where the final term (yw) is an activity coefficient of water in the solid phase, which
accounts for the change in the volume of the hydrate cavity. The best hydrate lattice
models, incorporating the above three volume changes enable the prediction of
hydrate unit cell volume to within 0.1%.
The concept of this change is indicated in Figure 3. Figure 3(a) indicates that the
free energy of water in the empty hydrate lattice is assumed to be known at a given
temperature and volume. The volume of the empty hydrate lattice (V,"') must be
equal to the volume of the filled lattice (V,), so that the only energy change in Figure
3(a) is due to occupation of the hydrate cavities. Yet, in Figure 3(b), the volume of
the hydrate lattice is allowed to change due to filling of the lattice with guest mole-
cules. As noted above, very small lattice changes (e.g., 0.5%) due to filling by guest
can result in large (e.g., 15%) errors in high-pressure predictions.
This non-ideality correction results in reducing the prediction inaccuracy sub-
stantially, For example, in comparison to all published uninhibited hydrate data
(1685 data points), the average absolute temperature error can be reduced from 0.65
to 0.4 K, and the pressure decreases from 8 to 6% average absolute error. These
hydrate prediction errors approach those of measurements, and may indicate a limit
to fruitful improvements of the model. For more details, the reader is referred to the
overview7' and to the four detailed publications of the model
4.2.2 Method Improvements in the use of the Statistical

Thermodynamics Theory
It should be noted that until the last decade, hydrate predictions using Equation (1)
were restricted to incipient hydrate formation - i.e., specifying the temperature and
pressure conditions at which the hydrate phase was initiated. The vapor-liquid equi-
libria analog of this is the determination of the dew point. However, Gupta et ~ l . , ' ~
Figure 3 Concepts of the van der Waals and Platteeuw model and improvement. (a) original
model that does not allow for distortion by guests (v,=vfT); (b) corrected model
allowing guest distortion of lattice (vH# vHMT)
in Bishnoi’s laboratory, indicated how Equation (1) could be extended to predictions

beyond the incipient hydrate formation condition, to determine the phase fractions,
in the vapor-liquid analog of a flash calculation.
Today, all modern hydrate prediction programs are carried out using this new
technique, which is based on Gibbs energy minimization. It is extremely useful to
determine phase fraction; in the aforementioned cases of two-phase hydrate forma-
tion (without a free water phase), one can use the Gibbs energy minimization to esti-
mate whether there will be a sufficient hydrate amount to form a blockage in a
process.
72 Chapter 6
5 Hydrate Challenges for the Future: Reversible and

Irreversible Thermodynamics
As we deplete the readily accessible natural gas reserves, we will encounter condi-
tions that are both more remote and more severe. We will be challenged to explore
deep ocean environments with higher pressures, permafrost environments with lower
temperatures, and gases that were previously considered non-economical, such as
those containing non-combustible components of nitrogen, carbon dioxide, and
hydrogen sulfide. Such unusual conditions also stretch the applicability limits of
hydrate phase equilibrium thermodynamics.
In addition to their terrestrial applications, hydrates are of concern in outer space.
Miller77suggested hydrates’ presence in comets, and on Jupiter, Mars, Neptune, Saturn,
Uranus, and Venus. Recently, Loveday et aZ.78suggested that methane hydrate may be
the dominant methane-containing phase in the nebula from which Saturn, Uranus,
Neptune and their major moons, including an important role in formation models of
Titan, Saturn’s largest moon. The temperature, pressure, composition, and enthalpy
ranges for space applications encompass and usually exceed those encountered on
earth. However, for normal conditions, hydrate thermodynamic predictions now rest on
a comparatively sound foundation, with an accuracy approaching that of measurements.
Yet, the largest future challenge goes beyond time-independent descriptions, to
irreversible thermodynamics, or kinetics. We know very little about a kinetic mech-
anism founded on hydrate measurements. Due to the stochastic nature of nucleation,
experimentalists have dealt with the deterministic growth process.
Kinetic measurements are changing from macroscopic to microscopic scales.
Initially, kinetics consisted of macroscopic measurements of the fluid phases associated
with hydrates - such as gas consumption rates or liquid turbidity, fundamentally in
Bishnoi’s l a b ~ r a t o r y .Subsequently,
~~ mesoscopic measurements of hydrate cry-
s t a l ~ , 8 &particle
~ ~ size distribution,83+ 84 and film growth rates are Micro-
scopic kinetic hydrate phase measurements are emerging.83.88-92 A review of
microscopic hydrate science for both kinetics and thermodynamics is presented by
~oh.93
To borrow Churchill’s phrase, proving such hydrate phase kinetics measurement
principles represents “the end of the beginning” for time-dependent experiments. The
systematic kinetic study to generate a fundamental mechanism of hydrate nucleation
and growth constitutes the major remaining challenge in hydrate physico-chemical
science.
References
1. H. Davy, Phil. Trans. Roy. SOC. (London), 1811, 101, 1.
2. J. Priestley, Experiments and Observations on Different Kinds of Aid and Other Natural
Branches of Natural Philosophy, Thos. Pearson, London, in 3 volumes, 1778, pp. 358ff.
3. R. deForcrand and P. Villard, Compt. Rendu., 1888, 106, 849.
4. J. S. Loveday, R. J. Nelmes, D. D. Klug, J. S. Tse and S . Desgreniers, Can. J. Phys., 2003,
81, 539.
5. D. N. Glew, Nature, 1959, 184, 545.
6. D. N. Glew, Can J. Chem., 2003,81, 179.
7. E. G. Hammerschmidt, Znd. Eng. Chem., 1934,26, 851.
8. W. M. Deaton and E. M. Frost, Gas Hydrates and Their Relation to the Operation of
Natural-Gas Pipe Lines, US Department of the Interior, Washington, DC, 1946, 101 pp.
9. D. B. Carson and D. L. Katz, Trans. A.Z.M.E., 1942,146, 150.
10. D. L. Katz, Trans. A.Z.M.E., 1945, 160, 140.
11. R. Kobayashi and D. L. Katz, J. Petrol. Technol., 1949, 1, 66.
12. R. Kobayashi and D. L. Katz, Znd. Eng. Chem., 1953,45,440.
13. D. L. Katz, D. Cornell, R. Kobayashi, F. H. Poettmann, J. A. Vary, J. R. Elenbaas and C. F.
Weinaug, Handbook of Natural Gas Engineering, Chapter 5 , McGraw-Hill, New York, 1959.
14. von Stackelberg, Natuwiss, 1949, 11-12, 1.
15. W. F. Claussen, J. Chem. Phys., 1951, 19, 259 (662 - Erratum).
16. W. F. Claussen, J. Chem. Phys., 1951,19, 1425.
17. L. Pauling and R. E. Marsh, Proc. US Natl. Acad. Sci., 1952,38, 1 12.
18. H. M. Powell, J. Chem. SOC (London), 1948,61.
19. J. H. van der Waals and J. C. Platteeuw, Advances in Chemical Physics, Vol 1 1, Interscience,
New York, 1959, 1.
20. Y. F. Makogon, Gaz. Prom., 1965, 5, 14.
21. K. A. Kvenvolden and M. A. McMenamin, Hydrates of Natural Gas: A Review of Their
Geologic Occurrence, Geol Surv. Circ (US) 825, Washington, DC, 1980.
22. G. D. Ginsburg and V. A. Soloviev, Proc. Ofshore Tech Conference, OC 7693, Houston,
Texas May 1-4, 1995,513.
23. M. W. Lee, D. R. Hutchinson, T. S. Collett and W. P. Dillon, J. Geophys. Res., 1996,
101(B9),20, 347.
24. A. E. Taylor, S. R. Dallimore, R. D. Hyndman and F. Wright, Proceedings ofthe Fourth
International Conference on Gas Hydrates, Yokohama May 19-23,2002, 63.
25. S. Saito, D. R. Marshall and R. Kobayashi, A.Z.Ch.E. J., 1964, 10, 734.
26. W. R. Parrish and J. M. Prausnitz, Znd. Eng. Chem. Proc. Des. Dev., 1972, 11, 26.
27. S. Y. Lee and G . D. Holder, A.Z.Ch.E. J., 2002, 48, 161.
28. F. W. Nolte, D. B. Robinson and H. J. Ng, Proceedings of the 64th Annual Gas Processing
Association Convention, San Antonio, Texas, 1985, 137.
29. J. A. Ripmeester, Ann. NYAcad. Sci., 912, 2000, 1.
30. J. A. Ripmeester, Achievements in Hydrate Science at the National Research Council of
Canada, Personal Communication, July 18, 2003.
31. A. K. Sum, R. C. Burmss and E. D. Sloan, J. Phys. Chem. B., 1997, 101,7371.
32. A. Klapproth, E. Goreshnik, D. Staykova, H. Klein and W. F. Kuhs, Can. J. Phys., 81,
2003,503.
33. J. W. Tester, R. L. Bivins and C. C. Herrick, A.Z.Ch.E. J., 1972, 18, 1220.
34. J. S. Tse, M. L. Klein and I. R. McDonald, J. Phys. Chem., 1983,87,4198.
35. C . Moon, P. C. Taylor and P. M. Rodger, Can. J. Phys., 2003, 81, 451.
36. J. B. Klauda and S. I. Sandler, J. Phys. Chem. B, 2002, 106, 5722.
37. Z. Cao, B. Anderson, J. W. Tester and B. L. Trout, Proceedings of Fourth International
Hydrate Conference, Yokohama, Japan, May 2002,3 15.
38. L. A. Stem, S . H. Kirby and W. B. Durham, Science, 1996,273, 1843.
39. Y. P. Handa, J. Chem. Thermo., 1986, 18, 915.
40. S. R. Dallimore, T. S . Collett, T. Uchida, M. Weber and H. Takahashi, Proceedings of
Fourth International Hydrate Conference, Yokohama, Japan, May 2002, 36.
41. A. Trehu, N. L. Bangs, M. A. Arsenault, G. Bohrmann, C. Goldfinger, J. E. Jonson,
Y. Nakamura and M. E. Torres, Proceedings of Fourth International Hydrate Conference,
Yokohama, Japan, May 2002,90.
74 Chapter 6
42. K. Bil, Natural Gas Hydrate in Oceanic and Pennufiost Environments, M. Max (ed),
Kluwer Academic Publishers, Dordrecht, The Netherlands, 2000, 349.
43. Y F. Makogon, Natural Gas Hydrates: The State of Study in the USSR and Perspectivesfor Its
Use,presented at the Third Chemical Congress of North America, Toronto, Canada, June 1988.
44. R. Matsumoto, Proceedings of the Fourth International Conference on Gas Hydrates,
Yokohama, May 2002, 1.
45. C. K. Paul1 and W. P, Dillon, (eds), Natural Gas Hydrates: Occurrence, Distribution, and
Detection, American Geophysical Union, Monograph 124, Washington, DC, 2001.
46. J. P. Kennett, K. G. Cannariato, I. L. Hendy and R. J. Buhl, Methane Hydrates in
Quaternary Climate Change, American Geophysical Union, Washington, DC, 2003.
47. N. Handa and T. Ohsumi, (eds), Direct Ocean Disposal of Carbon Dioxide, Terra
Scientific Publication Company, Tokyo, 1995.
48. M. D. Max, Natural Gas Hydrate in Oceanic and Permafrost Environments, Kluwer
Academic Publishers, Dordrecht, The Netherlands, 2000.
49. J.-P. Henriet and J. Mienert, Gas Hydrates: Relevance to World Margin Stability and
Climatic Change, Geological Society, London, Special Publication 137, 1998.
50. E. D. Sloan, J. Happel and M. A. Hnatow (eds), (First) International Conference on
Natural Gas Hydrates, Vol715, Annals NY Acad Sci., New York, 1994.
51. J. P. Monfort, (ed), Proceedings of Second International Conference on Natural Gas
Hydrates, Toulouse, France, 1996.
52. G. D. Holder and P. R. Bishnoi (eds), Gas Hydrates Challenges for the Future (Third
International Conference on Natural Gas Hydrates: Salt Lake City),Vol912, Annals NY
Acad Sci., New York, 2000.
53. Y. H. Mori, (ed.), Proceedings of the Fourth International Conference on Gas Hydrates,
Yokohama, Japan, 2002.
54. Tenth International Conference on Physics and Ice, Can. J. Phys., 81, (2003).
55. E. D. Sloan, J. Chem. Thermo., 2003,3541.
56. Z. Huo, K. Hester, K. T. Miller and E. D. Sloan, A.I.Ch. E. J,, 2003,49, 1300.
57. E. D. Sloan, Clathrate Hydrates of Natural Gases, 1st edn, Chapter 4, Marcel Dekker,
Inc., New York, 1989.
58. E. D. Sloan, Clathrate Hydrates of Natural Gases, 2nd edn, Chapter 4, Marcel Dekker,
59. F. Kiyono, A. Yamasaki and K. Ogasawara, Proceedings of the Fourth International
Conference on Gas Hydrates, Y. H. Mori (ed), Yokohama, Japan, 2002,272.
60. J. W. Gibbs, Collected Works, Vol I, Longmans, Green and Company, New York, 1928.
61. W. F. Kuhs, Personal Communication, St. Etienne, France, June 17, 2003.
62. E. D. Sloan, Clathrate Hydrates of Natural Gases, 2nd edn, Chapter 5, Marcel Dekker,
63. M. D. Jager, C. J. Peters and E. D. Sloan, Fluid Phase EquiZibx, 2002, 193, 17.
64. W. L. Mao, H. K. Mao, A. F. Goncharov, V .V. Struzhkin, Q. Z. Guo, J. Z. Hu, J. F. Shu,
R. J. Hemley, M. Somayazulu and Y. S. Zhao, Science, 2002,297, 2247.
65. H. Tanaka, Can. J. Phys., 2003, 81, 55.
66. T. Uchida, S. Takeya, C. A. Wilson, C. A. Tulk, J. A. Ripmeester, J. Nagao, T. Ebinuma
and H. Narita, Can. J. Phys., 2003, 81, 351.
67. R. A. Kini, Z. Huo, M. D. Jager, P. Bollavaram, A. L. Ballad, S. F. Dec and E. D. Sloan,
Proceedings of the Fourth International Conference on Gas Hydrates, Y. H. Mori (ed),
Yokohama, Japan, 2002,867.
68. M. B. Helgerud, Ph.D. Thesis, Standford University, 2001.
69. C. J. Rawn, A. H. Rondinone, B. C. Chakoumakos, S. Circone, L. A. Stern, S. H. Kirby
and Y. Ishii, Can. J. Phys., 2003, 81, 431.
70. M. M. Mooijer-van den Heuvel, R. Kini, 2. Huo, C. J. Peters and E. D. Sloan,
Proceedings of Fourth International Hydrate Conference, Yokohama, May 2002,619.
7 1. A. L. Ballard and E. D. Sloan, Proceedings of Fourth International Hydrate Conference,
Yokohama, May 2002,307.
72. A. L. Ballard and E. D. Sloan, Fluid Phase Equilibr., 2002, 194-197, 371.
73. M. D. Jager and A. L. Ballard, Sloan, Fluid Phase Equilibl:, 2003,211, 85.
74. A. L. Ballard and E. D. Sloan, Fluid Phase Equilibl:, 2004,218, 15.
75. A. L. Ballard and E. D. Sloan, Fluid Phase Equilibr., 2004, (in press).
76. A. K. Gupta, P. Bishnoi and N. Kalogerakis, Fluid Phase Equilibl:, 1991,63,65.
77. S . L. Miller, Proc. USA Nutl. Acad. Sci., 1961,47, 1798.
78. J. S. Loveday, R. J. Nelmes, M. Guthrie, S. A. Belmonte, D. R. Allan, D. D. Hug, J. S.
Tse andY.P. Handa, Nature, 2001,410, 661.
79. P. Englezos, N. Kalogerakis, P. D. Dholabhai and P. R. Bishnoi, Chem. Eng. Sci., 1987,
42, 2647.
80. E. A. Smelik and H. E. King, Amel: Minerologist, 1997, 82, 88.
81. R. Larsen, C. A. Knight and E. D. Sloan, Fluid Phase Equilbr., 1998, 150-151, 353.
82. P. Bollavaram, S. Devarakonda, M. S. Selim and E. D. Sloan, Gas Hydrates: Challenges
for the Future, Annls. NY Acad. Sci, Vol912, G. D. Holder and P, R. Bishnoi (eds), 2000,
533.
83. M. A. Clarke, and P. R. Bishnoi, Chem. Eng. Sci., 2000,55,4869.
84. A. Fiedel-Dufour and J.-M. Heni, Proceedings of the Fourth International Hydrates
Conference, Yokohama Japan, May 2002,917.
85. D. K. Staykova, T. Hansen, A. N. Salamatin and W. F. Kuhs, Proceedings of the Fourth
International Hydrate Conference, Yokohama, Japan, May 2002,537.
86. L. A. Stem, S. Circone, S. H. Kirby and W. B. Durham, Proceedings of the Fourth
87. E. M. Freer, M. S. Selim and E. D. Sloan, Fluid Phase Equilibl:, 2001, 185, 65.
88. X. Wang, A. J. Schultz and Y. Halpern, Proceedings of the Fourth International Hydrate
Conference, Yokohama, Japan, May 2002,455.
89. B. M. Freifeld, T. J. Kneafsey, L. Tomutsa, L. A. Stem and S. H. Kirby, Proceedings of
the Fourth International Hydrate Conference, Yokohama, Japan, May 2002, 750.
90. S . Subramanian and E. D. Sloan, Fluid Phase Equilbr., 1999, 158-160, 813.
91. T. Pietrass, H. C. Gaede, A. Bifone, A. Pines and J. A. Ripmeester, J. Am. Chem. SOC.,
1995,117,7520.
92. I . L. Moudrakovski, C. I. Ratcliffe and J. A. Ripmeester, Proceedings of the Fourth
93. C. A. Koh, Chem. Soc. Rev., 2002,31, 157.
CHAPTER 7
Ionic Liquids in Separation

Processes
JURGEN GMEHLING
1 Introduction
Separation processes are not only of great importance in refineries, but also in the chem-
ical, petrochemical, gas processing, and pharmaceutical industries. Although the reac-
tor can be regarded as the heart of a chemical plant, in most cases, 60-80% of the total
cost is taken up by the separation step. This step involves one or more thermal separa-
tion processes such as distillation, extraction, absorption, crystallization, adsorption,
membrane processes, etc., which are used to obtain the products at the required purity.
In the various separation processes, concentration differences between two differ-
ent streams (in most cases, two different phases) are employed to carry out the sep-
aration. The second stream is generated with the help of energy or a mass separating
agent (solvent, membrane, adsorbent, etc.). Because of the various advantages of the
distillation process (energy as separating agent, large density difference of the coex-
isting fluid phases (vapor, liquid)), 90% of all separations involve distillation.
Other separation processes can become advantageous, when separation problems
such as unfavorable separation factors (0.95 < aI24 . 0 5 ) or azeotropic points occur.
In these cases, a special distillation process (extractive distillation) may be used.
Extraction processes do not depend on a difference of vapor pressure between the
compounds to be separated but on the relative magnitudes of the activity coefficients
of the compounds. As a result, extraction processes are particularly useful in separat-
ing the different aromatic compounds (C, to CI2)from the different aliphatic com-
pounds (C, to C12).Absorption processes are ideally suited for the removal of
undesired compounds from gas streams, e.g., sour gases (H2S, CO,) from natural gas.
For extractive distillation, extraction and absorption processes, highly selective
solvents are required. The economic importance of extraction and extractive distil-
lation processes can be recognized from the fact that the worldwide production of
the BTX aromatics (benzene, toluene, xylenes, ethylbenzene) as important primary
petrochemical products for the industrial manufacturing of many chemical products
Ionic Liquids in Separation Processes 77
is approximately 60X lo6 t/a2. Aromatic compounds are mainly produced by steam
cracking and reforming processes, together with extractive distillation and extraction
processes used to separate the hydrocarbons.
In all cases, the selective solvents (entrainers) have the task of altering the parti-
tion coefficients in a way that high separation factors and selectivities for the differ-
ent phase equilibria (extractive distillation: vapor-liquid equilibrium (VLE),
extraction: liquid-liquid equilibrium (LLE), absorption: gas-liquid equilibrium
(GLE)) are achieved, resulting in a separation of compounds. The required partition
coefficients, separation factors and selectivities can be calculated with the help of
thermodynamic models ( gE-models, equations of state).
As the total cost of these separation processes is strongly influenced by the selec-
tivity of the solvent, the search for new and better solvents is important and has been
a priority for many decades.
2 Thermal Separation Processes Using a Selective

Solvent as Separating Agent
2.1 Extractive Distillation
Extractive distillation is a suitable distillation process for the separation of azeotropic
systems or systems with separation factors a,,close to unity. A typical extractive dis-
tillation process for the separation of aliphatics from aromatics is shown in Figure 1. In
extractive distillation processes, the high boiling selective solvent (entrainer), intro-
duced not far from the top of the extractive distillation column, has to alter the volatili-
ties in such a way that the separation factor a12 attains a value very different from unity.
Typical entrainers for the separation of aliphatics from aromatics are N-Methyl-
pyrrolidone (NMP) or N-Fonnylmorpholine (NFM). In the presence of NMP or NFM,
aliphatics
7 aromatics
7
Figure 1 Schematic diagram of an extractive distillation process (example: separation of

aliphatics from aromatics)
78 Chapter 7
but also in the case of most other entrainers, the aliphatic hydrocarbons become more
volatile. This means that they appear at the top, while the aromatic hydrocarbons,
together with the entrainer, leave at the bottom of the column. In the second distillation
column, the high boiling entrainer can be separated easily from the aromatics and recy-
cled to the first column. To achieve selectivities that are high enough, entrainer concen-
trations of 50-80% are required. The purity of the products is mainly influenced by the
number of theoretical stages realized and the selectivity of the entrainer. Depending on
the saturation vapor pressure of the entrainer and the feed location, traces of the
entrainer may be lost with the distillate in the first column.
2.2 Liquid-Liquid Extraction

For liquid-liquid extraction, a selective solvent is required, which shows only a par-
tial miscibility with the liquid stream to be separated. In Figure 2, a typical multi-
stage counter-current liquid-liquid extraction process is shown for the separation of
aliphatics from aromatics. In this process, with the help of the selective solvent
(extractant) the desired components (aromatics) are extracted from the feed stream.
Distillation is typically used for the recovery of the selective solvent from the extract
and raffinate stream leaving the extraction column.
The selectivity, which strongly influences the number of separation stages, together
with the capacity of the solvent, strongly influences the investment cost of a separation
process. The objective is to find entrainers with a high selectivity and a sufficient capac-
ity but unfortunately, in most cases, an increase in selectivity is linked to a decrease in
capacity. The very low capacity of water, for example, is the reason that in spite of its
high selectivity, it is not used for the separation of aliphatics from aromatics.
extraction distillation
column
ractant raffinate stream
I
Turn*\
aliphatics
1 I
aromatics
‘
I hydrocarbons
T
Figure 2 Schematic diagram of a typical extraction process (example: separation of aliphat-
ics from aromatics)
purified gas (CH,) desorbed gas ( C 0 2 , H2S)
---D
$4
f
P
c
natural gas
-+
-
+ - absorbent
Figure 3 Schematic diagram of a typical absorption process (example: removal of sour gases
from natural gas)
2.3 Absorption
The main objective of absorption processes is the removal of one or more compo-
nents from a gas stream using selective solvents. Figure 3 shows a typical absorption
process in which, with the help of a selective solvent (absorbent), the undesired com-
pounds (in this case, H,S, CO,) are removed from the raw gas (in this case, natural
gas) in a multistage countercurrent process. While the purified gas leaves the
absorber (saturated with the selective solvent), the absorbent is regenerated in a sec-
ond column (desorber) and is recycled to the absorber. In the case of absorption, one
can distinguish between physical and chemical absorption processes. In the case of
physical absorption, the absorber is operated at high pressures and low temperatures
while for the desorber the opposite conditions are used.
3 Thermodynamic Fundamentals
For the right choice of the selective solvent and for the development and design of
separation processes, a reliable knowledge of the phase equilibrium behavior
(extractive distillation: vapor-liquid equilibrium (VLE); extraction: liquid-liquid
equilibrium (LLE), absorption: gas-liquid equilibrium (GLE)) is required. This
information is available from phase equilibrium thermodynamics.
3.1 Vapor-Liquid Equilibrium (VLE)

The VLE behavior for weakly associating systems can be described with the
required accuracy by the following simplified e q ~ a t i o n : ~
80 Chapter 7
where xi refers to the mole fraction of component i in the liquid phase, yi to the mole
x
fraction of component i in the vapor phase, to the activity coefficient of compo-
nent i, P is the total pressure and is Pi”the vapor pressure of pure i.
Using Equation (l), the relationships for the partition coefficient, Ki, and the sep-
aration factor, aii,can be derived:
For the separation, only the difference, aii- 1, is used. In the case of azeotropic sys-
tems (a,=l),no separation can be achieved. For separation factors close to unity
(e.g., 0.95 < a,, < 1.05), a large number of theoretical stages are required and a spe-
cial distillation process has to be used.
3.2 Liquid-Liquid Equilibrium (LLE)

Starting from the isoactivity criterion, the partition coefficients, Ki, and selectivities,
Si,j,for extraction processes can be described using the following relations:
where E refers to the extract stream and R to the raffnate stream.

For the calculation, a reliable knowledge of the activity coefficients as a function
of composition and temperature is required.
3.3 Gas-Liquid Equilibrium (GLE)

Gas solubilities can be calculated using Henry coefficients, Hi,j,In contrast to VLE,
x*
the value of the activity coefficient approaches unity, when xi becomes zero. For
small solubilities and low gas pressure, the activity coefficient and the fugacity coef-
ficient can be neglected and Equation (7) can be simplified:
x
xi *Hi,j=yiqiP (7)
4 Selection of Selective Solvents

The selectivities, Si,j,at infinite dilution are usually used4 for the selection of selec-
tive solvents. These values can be calculated from activity coefficients at infinite
dilution (extractive distillation, extraction) or from Henry coefficients (absorption).
The required activity coefficients can either be calculated (predicted) with the help
of thermodynamic models (group contribution methods) or obtained from factual
data banks. The procedure for the selection of selective solvents for extractive dis-
tillation processes is given in refs. 4 and 5. The capacity Cj of extractants can be esti-
mated using activity coefficients at infinite dilution.
S"= -
lJ
r
3;.- (9)
For absorption processes, the partition coefficients and the separation factors at infi-
nite dilution can be calculated directly using Henry coefficients.
where 1 and 2 refer to the gases to be separated and 3 refers to the solvent
(absorbent).
5 Thermophysical Properties Required for Selective

Solvents
Typically, organic compounds are used as selective solvents. For a safe and efficient
operation, all selective solvents for separation processes should have the following
properties in common:
0 high selectivity
0
high capacity
0 no separation problems in the required regeneration step
high thermal and chemical stability
0
low melting point
0
non-toxic
0
low viscosity
0
negligible corrosivity
0
low price
0
simple purification
Additionally, in order to have little or no impurities in the distillate of the extractive

distillation column and also to avoid separation problems in the regeneration col-
umn, selective solvents for extractive distillation should have a boiling point that is
at least 40 "C higher than the boiling points of the components to be separated.
In the case of extraction, of course, it is most important that the solvent shows a
miscibility gap with the feed and the raffinate stream. At the same time, a high
82 Chapter 7
capacity Ci for the extract is desired (see Equation (10)). Also, it is desirable that the
extractant is almost insoluble in the raffinate phase. Furthermore, a sufficient density
difference between the two liquid phases and suitable values for the surface tension
are desired. In the case of absorption processes, the vapor pressure of the absorbent
should be negligible in order to avoid solvent losses. Finally, it is desirable that the
compounds to be removed should have a reasonable solubility.
A large number of organic compounds are currently used for the various separa-
tion processes.' One of the most popular ones is N-Methylpyrrolidone (NMP), used
not only for extractive distillation and for extraction but also for the various absorp-
tion processes. Since the cost of the separation step is greatly influenced by the
selectivity, new and better extractive solvents are continually being investigated.
6 Ionic Liquids
Over the past few years, ionic liquids have become very popular. These are a class of
novel solvents, which show very interesting properties such as low melting point
( 4 0 0 "C), suitable viscosity, and in particular, negligible vapor pressure and there-
fore non-flammability. The first ionic liquid, ethylammoniumnitrate with a melting
point of 12 "C, was synthesized by Walden in 1914.6Ionic liquids are good solvents
for polar and non-polar organic and also for inorganic substances. Many are liquids
at room temperature. These properties make ionic liquids interesting solvents for
industrial applications and they may replace toxic, flammable and volatile organic
solvents in future. Typically, an ionic liquid is composed of a large organic cation and
an inorganic, polyatomic anion. The various cations and anions investigated are pre-
sented in refs. 7 and 8. Since there are a large number of cations and anions to choose
from, there is virtually no limit to the number of ionic liquids that can be made. As a
result, ionic liquids may be considered as designer solvents for bi-phasic reactions or
as selective solvents (entrainers) for separation processes. In Figure 4, typical cations
and anions are shown.
Before 2001, only a few pure component properties (densities, viscosities, etc.)
and mixture properties (phase equilibrium behavior, excess properties) involving
ionic liquids were available. For a better understanding of their behavior and for the
Cations Anions
F O O F
,N\ @,N F+{-A+F
R "\ F O O F
1-R-3-methylimidazolium his( trifluoromethylsu1fonyl)imide
R = methyl, ethyl, butyl, hexyl, ..
0
II
R-0-S-0'
II
I
0
Ethylpyridinium R-sulfate
R = methyl, ethyl, methoxyethyl
Figure 4 Structural form of ionic liquids

development of thermodynamic models, reliable experimental data are required. As
a result, various research groups started systematically measuring these properties
and over the past 2-3 years, pure component properties (viscosities, heat capacities,
r,
densities, melting points, etc.), and, in particular, phase equilibrium data, such as
VLE, LLE, SLE, gas solubilities and excess enthalpies HE have been measured for
a variety of ionic liquids at different temperatures. To date, more than 6250 data
points involving ionic liquids are a~ailable.~
7 Results
r
Several research groups have measured infinite dilution activity coefficients for a
variety of solutes (alkanes, alkenes, cyclic hydrocarbons, ketones, alcohols, water,
etc.) in various ionic From the selectivities, it appears as though ionic
liquids may replace the typical entrainers (extractants) currently used for extractive
distillation or extraction processes. In Figures 5(a) and 5(b) the selectivities observed
for the separation of cyclohexane from benzene (acting as a test system for the sep-
aration of aliphatics from aromatics) for different ionic liquids are shown as a func-
tion of temperature, together with the values obtained for NMP.9 A comparison
indicates that the selectivities for ionic liquids are significantly higher than in the
case of NMP. For all solvents, the selectivity decreases with increasing temperature.
For the different imidazolium bis-trifluorosulfonylimides(R,R,IM BTI), it can be
seen that the selectivities decrease with the length of the alkyl chain from methyl to
butyl. The selectivities for 1,2-dirnethyl-3-ethylimidazolium BTI are lower than the
values for 1-methyl-3-methylimidazolium BTI (MMIM BTI ), but larger than for
EMIM BTI. For the imidazolium compounds with BF, or PF; as the anion, only the
results for the hexyl compounds are available and these also show higher selectivi-
ties than does NMP. The highest selectivity is observed for 1-ethyl-3-methylimida-
zolium ethylsulfate. In comparison to NMP, the selectivities for the latter compound
are approximately 2.5 times higher than the values for NMP. The second largest
selectivities are observed for 4-methyl-N-butylpyridinium tetrafluoroborate. Similar
promising results for ionic liquids were also obtained for the separation of systems
such as hexane-benzene and heptane-toluene. l4
In Figure 6, P-x data for benzene or cyclohexane with 1-ethyl-3-methylimidazolium
bis-trifluorosulfonylimide (EMIM BTI) at 353 K, measured with a static a~paratus,'~
together with the VLE data for the mixtures NMP+cyclohexane or +benzene, are
shown.16 From these VLE measurements, it can be seen that high selectivities are
achieved for the separation of aliphatics from aromatics using EMIM BTI. Furthermore,
it can be seen that the ionic liquid EMIM BTI forms a large miscibility gap with cyclo-
hexane. The system with benzene also shows a miscibility gap. The VLE behavior for
the system EMlM BTI-benzene is particularly interesting. A miscibility gap is observed
in spite of the system showing only a small positive deviation from Raoult's law. But
surprisingly, over a wide composition range ( 0-80 mol%), the activity coeflicients
remain nearly constant. While a value of 1.28 for the activity coefficient of benzene at
infinite dilution is obtained, a similar value leads to the miscibility gap at about 80
mol% of benzene. From Figure 6, it can also be seen that the ionic liquid is nearly insol-
uble in cyclohexane (raffmate phase in the case of liquid-liquid extraction).
84 Chapter 7
20.
18.
$ 16.
.g 14.
.*
-2 12.
10.
8.
6.
290. 300. 310. 320. 330. 340. 350.

(a) Temperature [K]
18. -
16. - 1
I
:
e
e
?? 12.
'S
v)
10.
l4.I
290. 300. 310. 320. 330. 340. 350.

(b) Temperature [K]
Figure 5 Comparison of the selectivity at injkite dilution Sz of various selective solvents for
the separation problem cyclohexane(l)/benzene(2).(a) I -methyl-3-methylimidazolium
[MMIM]' bis-trijluorosulfonylimide* [ I 01; I -ethyl-3-methylimidazolium[EMIM]+
bis-trtfluorosulfonylimide [BTI] A [ l o ] [I I ] ; 1-butyl-3-methylimidazolium
[BMIM]' bis-trijluorosulfonylimide [BTI] [ I 01; 1 -ethyl-3-methylimidazolium
[EMIM]' ethylsulfate + [lo] and NMP + [9]. (b) I,2-dimethyl-3-ethylimidazolium
bis-tnj7uorosulfonylimide [BTI] [Ill; 4-methyl-N-butylpyridiniumtetrafluorobo-
rate* [12]; I -Hexyl-3-methylimidazolium[HMIM] tetrafluoroborate A [14]; I -
Hexyl-3-methylimidazolium [HMIM] hemfluomphosphate [I31 and NMP + [ I 01
The large miscibility gap observed for an ionic liquid mixed with an aliphatic
compound (without the addition of water as in the case of NMP) can be directly used
for the separation of aromatic from aliphatic hydrocarbons by liquid-liquid extrac-
tion. Besides the miscibility gap, there are other requirements necessary for a
successful extractant, such as high selectivity, high capacity, a low solubility of the
extractant in the raffinate phase, a simple separation of the extract and the raffinate
phase, low viscosity, high chemical and thermal stability and a sufficient density dif-
ference. Nearly all these requirements are met by the ionic liquids that have so far
been investigated. However, our present knowledge of the thermal and chemical sta-
bility of ionic liquids is limited. For example, for some ionic liquids (largely depend-
ent on the anion), hydrolysis does occur.
I I
100 4 100 r
-w'''-\ T = 354.15 K -k 80
y;*= 4.81
25 60
2
T=353.15K I 2
2
L
40
20
0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
X1 X1
Figure 6 VLE data for the systems at around 353 K: 0 Cyclohexane-1 -ethyl-3-methylimida-
zolium bis-trijluorosulfonylimide [EMIM BTI] [15] (left-hand side) and
Cyclohexane-NMP [I51 (left-hand side). Benzene-I -ethyl-3-m
bis-trifluorosulfonylimide[EMIM BTI] [I51 (right-hand side) and
[16] (right-hand side)
The high solubility of aromatic compounds in ionic liquids can also be used for
the removal of the different dibenzothiophenes using liquid-liquid extraction. l7
These are the compounds that are usually difficult to convert by hydro-treating in
desulfurization processes.
Because of their negligible vapor pressure, ionic liquids are uniquely suited as
absorbents, since no contamination of the purified gas stream will occur. In Figure
7, the solubilities of carbon dioxide and methane in EMIM BTI at 50 "C are
shown.'* Additionally, the Henry coefficients derived from these data are listed in
Table 1 together with the values for NMP. It can be seen that the solubility of CO,
is much higher than the solubility of methane. Furthermore, following Equation
(12) similar separation factors (selectivities) are obtained for the solvents EMIM
BTI and NMP, but a higher capacity is observed for EMIM BTI. This means that
EMIM BTI may be used to remove CO, (one of the sour gases in natural gases')
from methane.
Table 1 Henry coeficients for CO, and CH, in [EMIM][BTIIi8 and NMP at 323.I5 K8
Gas Hi,3 (bar) [EMIM][BTI] Hi,3(bar) NMP
co* 55.5 5 1.0 105 -I 5
CH4 740 -+ 10 1120 ? 20
Table 2 Henry coeficients for CO, and CH, in [BMIh4][PF6]19andNMp8 at 298 K

Gas Hi,3 (bar) [BMIMI[PFJ Hi,3(bar) NMP
co* 53.4 -I 0. 3 70 2 2
CH4 1690 2 180 1040 2 20
'Gas solubility measurements for H,S are in progress.

86 Chapter 7
4
- 2
s2
1
0 0.2 0.4 0.6 0.8 1

x1[moVmol]
Figure 7 Solubility of CO,0 and CH, in 1 -ethyl-3-methyl-imidiazolium

bis-trijluorosul-
fonylimide EMIM BTI at 323.15 K [ I 81
Gas solubility data have also been measured by other authors, e.g., Henry coeffi-
cients for CO, and CH, in 1-butyl-3-methylimidaoliumhexafluorophosphate
[BMIM][PF,] have been determined at 298 K.19 In Table 2, the Henry coefficients
for the two gases are given together with the values in NMP.9 As can be seen from
Table 2, much better separation factors (selectivities) are obtained for [BMIM][PF,]
than for NMP and in comparison to EMIM BTI, the capacity is higher. From the
results shown in Figure 7 and listed in Tables 1 and 2, it can be concluded that ionic
liquids may be suitable absorbents for absorption processes. The removal of the sour
gas CO, from natural gas is only one example.
8 Conclusion
Recent work has shown that ionic liquids are promising selective solvents for extrac-
tive distillation, liquid-liquid extraction and for separation by absorption. For these
separation processes, high selectivities and capacities and negligible vapor pressure
are important properties, but there are also other factors, such as cost, viscosity, dif-
ficulty of purification, etc., that must be taken into account. A negligible vapor pres-
sure is particularly important for absorption processes and for the minimization of
solvent emission, but it can also cause problems. Because of the high boiling point
of an ionic liquid, their purification (e.g., the removal of high boiling compounds)
will require new separation techniques (e.g.,extraction with supercritical fluids).
For the industrial application of ionic liquids, as selective solvents for separation
processes or other applications, a reliable knowledge of the pure component proper-
ties and mixture data (phase equilibria, excess properties) is required. Although
during the past 2-3 years, various research groups have begun measuring the ther-
mophysical properties of ionic liquids, our knowledge is, today, still very limited.
For example, we do not even know if the well-known gE-modelsor equations of state
can be used to predict the behavior of multicomponent systems with ionic liquids, as
is possible for the classical solvents (organic compounds, water). Because of the vast
number of possible combinations of anions and cations, reliable predictive models
would be most desirable. Predictive models would allow the search for the best-
suited ionic liquid for a given separation problem or as a solvent for chemical reac-
tions for new processes. With the current tremendous research activity in this field
there is a good chance that the required models can be developed within the next few
years. Then, the future will indicate whether the promising properties of ionic liq-
uids will result in them replacing the established organic solvents in separation
processes and in other applications.
Acknowledgement
The author would like to thank Dr. M. Krummen, Dr. R. Kato and Dr. J. Ahlers for
their help in preparing the manuscript and Deutsche Forschungsgemeinschaft for
financial support.
References
1. J. Gmehling and A. Brehm, Grundoperationen, Wiley-VCH, Weinheim, 1996.
2. K. Jiihnisch, Erdol Erdgas Kohle, 2001, 17, 232.
3. J. Gmehling, B. Kolbe, Themodynamik, VCH-Verlag, Weinheim, 1992.
4. J. Gmehling, C. Mollmann, Znd. Eng. Chem. Res., 1998,37, 31 12.
5. B. Kolbe, J. Gmehling, U. Onken, Bel: Bunsenges. Phys. Chem., 1979, 83, 1133.
6 P. Walden, Bull. Acad. Zmper. Sci. (St. Petersburg), 1914, 1800.
7 P. Wasserscheid and T. Welton (eds), Ionic Liquids in Synthesis, Wiley-VCH, Weinheim
2003.
8 R. D. Rogers and K. R. Seddon (eds), Ionic Liquids - Industrial Applications for Green
Chemistry, ACS Symposium Series 8 18, American Chemical Society, Washington, 2002.
9 Dortmund Data Bank 2003, www.ddbst.de.
10 M. Krummen, P. Wasserscheid and J. Gmehling, J. Chem. Eng. Data, 2002,47, 1411.
11 A. Heintz, D. V. Kulikov and S. P. Verevkin, J. Chem. Eng. Data, 2002, 47, 894.
12 A. Heintz, D. V. Kulikov and S. P. Verevkin, J. Chem. Eng. Data, 2001,46, 1526.
13 T. M. Letcher, B. Soko, D. Ramjugernath, N. Deenadayalu, A. Nevines and P.K. Naicker,
J. Chem. Eng. Data, 2003,48, 708.
14 T. M. Letcher, B. Soko, P. Reddy and N. Deenadayalu, J. Chem. Eng. Data, 2003, 48,
1587.
15. M. Krummen, Ph.D. Thesis, Oldenburg 2002.
16. P. Gierycz, M.Rogalski and S . Malanowski S . , Fluid Phase Equilib., 1985, 22, 107.
17. A. Bosmann, L. Datsevich, A. Jess, A. Lauter, C. Schmitz and P. Wasserscheid, Chem.
Commun., 2001,2494.
18. J. Gmehling, unpublished data.
19. J. L. Anthony, E. J. Maginn and J. F. Brenecke, J. Phys. Chem. B, 2002,106, 7315.
CHAPTER 8
Spectrocalorimetric Screening
for Complex Process
Optimization
FLORIN DAN AND JEAN-PIERRE E. GROLIER
1 Reaction Calorimetry: General Aspects

To identify optimal operating conditions of a chemical process, knowledge of kinetic
and thermodynamic parameters for the most important main and side reactions is
needed. A conventional method for investigating a reaction during process develop-
ment is reaction calorimetry (RC).' RC is the technique accepted as the most lead-
ing method to study the process in near-to-the-industrial conditions. One of the
objectives of RC is to simulate an industrial process at bench scale, allowing a wide
spectrum of operation conditions and measurements.
Areas of application of reaction calorimetry include determination of calorimetric
data for reactions and process design, for the kinetic characterization of chemical
reactions and of physical changes, for on-line monitoring of heat release and other
analytical parameters needed in subsequent process development as well as for the
development and optimization of chemical processes with the objective, for instance,
to increase yield or profitability, control the morphology or degree of polymerization
and/or index of polydispersity, etc.
However, there is a remarkable disproportion between the three main areas occu-
pied today by RC, namely hazards, process desigdoptimization and for monitoring
the physico-chemical transformations. If there is an extensive activity in the field of
hazards, often with little contribution to increased safety, probably less than 20% of
the process development laboratories use calorimeters for process design and opti-
mization; very little interest is shown for the use of heat released as a tracer for
physico-chemical transformations since RC is still barely used in synthesis labora-
tories where reactions and process procedures are initiated.
Basically, reaction calorimeters can operate in modes so that they closely approx-
imate to isothermal, isoperibolic or adiabatic systems. Devices used to perform
Spectrocalorimetric Screening for Complex Process Optimization 89
RC measurements can be classified either as devices using jacketed vessels with
control of the jacket temperature (heat balance calorimeters, heat flow calorimeters
and temperature oscillation calorimeters) or as devices using a constant surrounding
temperature, e.g., jacketed vessels with a constant jacket temperature, (isoperibolic
calorimeters and power compensation calorimeters); such instruments may also fea-
ture single or double cells.
It should be noted that the main driving force for developing process-oriented
calorimetric instruments was the evolution of electronic hardware, which made
measurements easily possible on a (non-micro) laboratory scale.
2 Instrument
A detailed scheme of the instruments used in this study and constructed according
to the principle of scanning transitiometry is presented in Figure 1.
It consists of a calorimeter equipped with high-pressure vessels, a pVT system and a
LabVIEW-based virtual instrument (VI) software. Two cylindrical calorimetric detec-
tors (@ = 17 mm, 1 = 80 mm) made from 622 thermocouples chromel-alumel each are
mounted differentially and connected to a nanovolt amplifier. The calorimetric detectors
are placed in a calorimetric metallic block, whose temperature is directly controlled with
an entirely digital feedback loop of 22-bit resolution K), being part of the tran-
sitiometer software. The calorimetric block is surrounded by a heating-cooling shield,
and the temperature difference between the block and the heating-cooling shield is set
constant (5,10,20 or 30 K) and controlled by an analog additional controller. The whole
assembly is placed in a thermal insulation enclosed in a stainless-steel body and is
placed on a stand, which permits to move the calorimeter up and down over the calori-
metric vessels. A detailed description of the scanning transitiometer is given elsewhere.2
Figure 1 Detailed scheme of the transitiometric installation used as a reaction calorimeter

90 Chapter 8
In the case of polymer synthesis, the scanning transitiometer can be used as an
isothermal reaction calorimeter, the advancement of a polymerization reaction being
accurately monitored through the rigorous control of the thermodynamic parameter^.^
Two reaction cells - measuring and reference - are made as equal as possible to each
other and they operate in a common surroundings in a so-called twin dzferential
arrangement. The thermocouples composing the fluxmeters serve as a heat conduc-
tion path between cell walls and the block. The overall thermoelectric voltage, E, gen-
erated at the junctions provides an error-free indication of the temperature difference
that causes the heat flux along the wires; the overall heat flux (thermal power), W, is
summed up by series arrangement of the individual thermocouples. The thermoelec-
tromotive force is related to the thermal power by the following relation:
where K is a calibration constant that depends on the thermoelectric capacity, E, and

conductance, 'y, of the thermocouple wires.
To gather additional information on a reaction, reaction calorimeters are often
coupled with other analytical devices (e.g.,on-line FTIR, particle-sizing probes, tur-
bidity probes, pH or other ion selective probes, etc.). Therefore, we developed a
reaction cell that allows stirring, different dosing profiles for one or two reactants
and can accommodate a small optical probe coupled to a miniaturized spectrometer,
Figure 2.
IBI
Fiber Optic Spectrom
#I
- -
Figure 2 Illustration of the reaction cell equipped with the spectrometric probe
Spectrocalorimetric Screeningfor Complex Process Optimization 91
The performances of this instrument will be illustrated in the case of processes for
which reaction calorimetry was up to now difficult to apply, such as precipitant poly-
merization, particularly the ultrafine polyamide (PA) particles synthesis by anionic
polymerization of lactams in organic media. For all experiments, the reaction
mixture, consisting of the monomer, E-caprolactam (L6), the solvent (aliphatic
hydrocarbon) and the catalyst (lactamate anion), as well as the synthesis of the cat-
alyst (lactamate anion) were prepared in a separate reaction flask (see ref. 4); about
9 cm3 from it was transferred into the reaction cell at the reaction temperature with
a preheated syringe. As soon as a stable base line was achieved, the chain initiators
(CI) (a long-chain aliphatic isocyanate) was fed into the cell with the dosing device
following a pre-established dosing profile.
3 Particularities of Anionic Polymerization of Lactams

in Organic Media
Anionic polymerization of lactams in organic solvents is a complex process that
takes place mainly heterogeneously and allows to obtain powdered polyamide. The
particles thus obtained are virtually spherical, possessing a porous structure, a nar-
row particle size distribution and a good solvent and heat resistance. This form is
very handy for different processing procedures, such as flame spraying, electrostatic
coating, paste production, dispersions, lacquer binders and cosmetic formulations.s*6
The most important steps in polyamide formation are the following : initiation and
growth of macromolecules in a homogeneous medium, nucleation, phase separation
and aggregation of the growing chains, solidification and, finally, polymer crystal-
lization. All the above-mentioned events occur rapidly and partially over la^.^
The polymerization occurs by the activated monomer mechanism, which sup-
poses a two-step mechanism involving the acylation of the lactam anion (NaL) by
the N-acyllactam end-group followed by a fast proton-exchange with the monomer.*
In bulk polymerization, the preformed N-acyllactams or their precursors (so-
called CIS) are introduced into the system in order to avoid the slow initiation step
due to the absence of N-acyllactam groups at the beginning of polymerization:
C.I. NaL
Reaction Scheme
In the presence of organic media, the reaction does not start in the absence of CI
and, consequently, for calorimetric measurements the moment of adding the CI is
considered the “zero” time of the polymerization. In a very strong basic medium, the
92 Chapter 8
CI disappears rapidly in side reactions (Claisen condensation) and the polymeriza-
tion is prematurely ~ t o p p e dHowever,
.~ as long as the monomer remains in the reac-
tion medium, the polymerization can be continued by simply adding new amounts
of chain initiation. This is a great challenge while using RC for process monitoring
since in this way the process becomes a "controlled" one.
4 Spectro-CalorimetricInvestigation Under Isothermal

Conditions
The objective of this work is to simulate an industrial process on a bench scale,
allowing a wide spectrum of operation conditions and measurements. Taking into
account the complexity of the process as well as the fact that the calorimetric signal
reflects the sum of physicochemical changes in the reaction system, an additional
on-line mini-spectrometer (OceanOptics, S2000 spectrometer) for simultaneous in
situ measurements as well as some off-line data furnished by TMDSC and gravi-
metric measurements have been used.
The apparatus, a scanning spectro-transitiometer in such a configuration, is able
to work either under isothermal or non-isothermal conditions; consequently, the first
part deals with the isothermal RC on lactams polymerization and the second part
deals with the non-isothermal polymerization, respectively.
4.1 The Manner of Contacting the Catalytic Species

The main advantage of this process is the possibility of controlling the conversion
and orienting it toward obtaining powders with desired morphologies through a fine-
tuning of the synthesis parameters. There are many factors influencing the advance-
ment of the process and the most important ones are as follows: temperature,
solvent's quality and monomer/solvent ratio, chemical structure of catalytic species
(catalyst and chain initiator), catalyst/monomer ratio, initiatodcatalyst ratio, the
manner of contacting the catalytic species and hydrodynamic regime, e t ~In. what~
follows, the effects of three important factors on the polymerization process are dis-
cussed: the manner of contacting the catalytic components since the polymerization
starts as soon as the two species make contact with each other, the nature of the cat-
alyst and the level of stirring.
4.1.1 One-shot Technique

The simplest way of facilitating contact between the two catalytic species consists in
adding the whole amount of CI just at the beginning of the process. This method is the
so-called one-shot method and is well documented in the literature.5-7*10.11 The effect of
temperature on the shape of heat released is illustrated in Figure 3. Clearly, the shape of
the curves representing the heat evolution with polymerization changes from two peaks
at 75 "C to a large shoulder at 90 "C that diminishes at 105 OC, and finally disappears
at temperatures above 120 "C. The intensity of the first peak does not seem to be influ-
enced by temperature, ranging from 25 to 35 mW per gram of L6 for all experiments.
0 1000 2000 3000 4000 5000

300 1
t;i 100
0 1000 2000 3000 4000 5000 0 1000 2000 3000 4000 5000 6000
Timels
Figure 3 Eflect of the temperature on the overall heat rate for the anionic solution polymeriza-
tion of M using the one-shot method of contacting the catalytic species. In the insets,
the reaction temperatures and gravimetric conversions are given. Reaction condi-
tions: solvent, mixture of aliphatic hydrocarbons with bp. 140 - 160 "C; initial con-
centration of M, 3 mol-'L; catalyst, NaH = 3% mol mol-' M; Wcatalyst ratio = 1
On the contrary, the intensity of the second peak strongly increases with temperature,
from about 33 to 273 mW per gram of L6 at 75 and 120 OC, respectively.
The first peak corresponds to the initiation reaction between the two catalytic
species (a very fast reaction, practically not influenced by temperature in the inves-
tigated range), while the second one, which is sensitive to temperature, corresponds
to the propagation reaction. This assumption is sustained by the very sharp decline
in the intensity of the spectrometric signal observed at the very moment when phase
separation takes place, Figure 4. The drop in intensity corresponds to the occurrence
of polymer precipitation; in the thermogram, this drop corresponds to the beginning
of the second peak (see Figure 4(b)).
Taking into account the crystalline character of the aliphatic polyamides, the sec-
ond peak is given not only by the polymerization's heat but also contains the amount
of heat corresponding to the crystallization of the separated polymer. It is very diffi-
cult to discern between heat of propagation and crystallization, respectively, since
the two processes occur almost concomitantly. Consequently, additional off-line
data, i.e., gravimetric and DSC were used in order to bring about the deconvolution
of the overall heat, Figure 5.
As Figure 5(a) shows, the overall heat contains tree terms, namely heat of initia-
tion, heat of propagation (or polymerization) and heat of crystallization:
94 Chapter 8
1400
((I 1200
CI
=I
$ 1000
200
2000
1800
3 1600
1400
3 1200
E 1000
8, 800
2 600
% 400
pc 200
0
0 6M3 1200 1800 2400 3000 3600 4200
Time/s
Figure 4 On-line combination of calorimetry and spectroscopy: (a) 3 0 plots of the evolution of
spectrometric signal during the anionic polymerization of M;(b)simultaneous calori-
metric and spectroscopic signals recorded during polymerization at 90 "C (the inten-
sity of spectrometric signal was taken at 600 nm). Note the stability of the spectroscopic
signal before the phase separation and its sharp decrease at phase separation
Gravimetric data allowed the determination of L6 conversion into the high polymer,
the DSC thermogram provided the quantification of crystallization heat (supposing
that the heat of melting AHmelt for the first run is equal to the heat of crystallization
during polymerization) and the heat of initiation was calculated by deconvolution of
the heat rate using PeakFit v4.11 software (Systat Software 1nc.-SSI). In slurry poly-
merizations, changes in heat transfer properties are often neglected, as there are only
slight changes in viscosity during the course of reaction; this is why the stirring heat
was neglected in Equation (2). As shown in Table 1, the heat of melting did not
change significantly with the polymerization temperature (it increases slightly), but
the shape of the thermograms (Figure 5(b)) is strongly affected, suggesting crys-
talline and morphological changes of the PCA powders.
As is shown in the insets of Figure 3, keeping all other synthesis conditions iden-
tical, the high polymer yield increases significantly with temperature. However, due
Te rnperature/ *C
J
250
u 7 5 ' C -e-90°C -+-105"C *12OoC

0.05
(p -0.15
2 -0.35
a' -0.55
-
3; -0.75
-0.95-
-
Y -1.15-
C
-1.35-
I
-1.55-
P
-1.75' ' ' ' ' ' ' ' ' ' ' ' ' ' '
100 150 200 2 0
(b) Temperature1 ' C
Figure 5 Illustration of the method used for the quantijication of the individual contributions
to the overall heat released during the polymerization process ( a ) and TMDSC
traces (reversible curve first run) for the PCA powders obtained by the one-shot
method (b)
Table 1 Variation of the Thermal Properties of PCA Powders with the Reaction
Temperature
Reaction Melting temperature ("C) d
temperature Onset Peak Offset (Jg-' PA6)
( "C) 1st 2nd 1st 2nd
75 167.6 185.3 183.7 21 1.7 220.6 126.30
90 181.8 197.0 199.2 215.6 223.3 136.48
105 - 203.3 - 211.7 222.4 131.64
120 - 201.8 - 2 10.9a 224.8 140.55
to practical reasons the increase of the reaction temperature above 120 "C is not suit-
able due to the fouling at the reactor wall of the viscous precipitated particles. In
addition, the particles size also increases, large agglomerates being obtained at high
temperatures. An alternate route to increasing the polymer yield at lower tempera-
tures is to divide all the CI into small portions and to feed each portion just after the
baseline is reached.
96 Chapter 8
4.1.2 Successive Injections Technique
The above-mentioned manner of contacting the catalytic species is based on the
analysis of the heat released during the one-shot technique, i.e., the lifetime of the
process is a little more than 30 min. In addition, it is well known that the catalyst
(lactamate anion), in the absence of any impurities, is stable up to 48 h at the reac-
tion temperature. Figure 6 illustrates the heat released at 90 "C when the same
amount of CI, as that used in the one-shot method, was introduced by 12, 2 1 and 42
successive injections, respectively,
In all cases, almost quantitative transformation of L6 was obtained in spite of the
relatively low reaction temperature. For the greatest portion of CI (upper curve), the
overall heat released resembles the heat observed for the one-shot technique, i.e.,
passing through a maximum with time. The peak of heat released is strongly tem-
pered by a decrease in the amount of CI (lower curve in Figure 6). It should be
stressed that the shape of each individual peak corresponding to the successive injec-
tions of CI is exactly similar to the one-shot shape at the same temperature. However,
the amplitude of the individual peaks markedly decreases after a certain number of
injections, i.e., at low monomer concentrations. Of course, the intensity of each peak
and the corresponding heat are proportional to the amount of CI injected (see Figure
7(a)). Evidently, not only is the intensity of the main peak proportional to the amount
of heat associated to chain propagation but also to the amount of heat corresponding
to the initial shoulder, which reflects the initiation reaction. At the same temperature,
the ratio between the intensities of the two heat effects is almost constant,
In other experiments carried out at 83.5 OC, the injection rate of CI was progres-
sively decreased for each injection; the heat rate evolution during the first 10 min is
depicted in Figure 7(b). The slope of the heat rate corresponding to the initiation step
is proportional to the injection rate until the first peak disappears at low injection rates.
60
I I I I I
0 7200 14400 21600 28800 36000

Timats
Figure 6 Efect of the amount of CI on the overall heat for the anionic solution polymeriza-
tion of M at 105 "C using the successive injections method of contacting the cat-
alytic species
0 600 1200 1800 2400 3000

(a) Timers
3 14-
Tm 12-
5 10-
-h 0-
L 6-
I 4-
Figure 7 Successive injections method of contacting the catalytic species: ( a )the eflect of the
amount of CI on the overall heat of the individual peaks from Figure 6 Cfor all poly-
merizations, the third peak was considered; (b)influence of the injection rate of CI
on the shape of the heat released at 83.5 "C(only the$rst 10 min are represented)
The high polymer yield and the possibility of controlling the heat released during
the successive injections method suggested another method of contacting the cat-
alytic species, namely continuous injection of CI.
4.1.3 Continuous Injection Technique

According to this method the entire amount of CI is delivered over a certain time
period with a pre-established rate. Figure 8(a) shows the comparative heat evolution
between the successive injection method and the continuous injection one. The
whole amount of CI was delivered in 22 portions, 20 pl CI each, at a 20 min inter-
val, or with a constant rate, 60 pl h-' over the same time period. A good agreement
between the heat evolutions for the two processes can be noted. The effect of tem-
perature on the heat rate is illustrated in Figure 8(b). In both cases, a stationary state
was obtained but, for the same injection rate, its duration decreases with increasing
temperature (3 h and about 7 h at 105 and 83.5 OC,respectively). The first small peak
(or a shoulder) in thermograms was once again ascribed to the initiation step, i.e., it
corresponds to the reactions occurring in the homogeneous phase.
In the case of continuous addition of CI, three important indications are provided
by the spectroscopic data, (Figure 9(a)). The first one (region I in Figure 9(a))
concerns the period between the start of reaction and the beginning of polymer
98 Chapter 8
30
I
7200 14400 21600

Tlmels
19
14
-1
Figure 8 Continuous injection method of contacting the catalytic species: (a) heat evolution
versus time recorded during successive and continuous injection of the CI for the
anionic polymerization of M at I05 "C;(b)eflect of temperature on the overall heat
evolution (the CI was delivered 7 h at a rate of 60 pl h-l)
precipitation; it depends on temperature and the injection rate of CI. The second one
(region I1 in Figure 9(a)) is related to the particle size evolution with time (this is
only a qualitative information). Generally, after the phase separation the evolution of
spectroscopic signal runs parallel to the conversion. As seen in Figure 9(a), when one
approaches the end of the process, the intensity of spectrometric signal remains con-
stant. Finally, the third piece of information (also qualitative) refers to the average
size of particles. It was proved by SEM (see insets of Figure 9(a)) that the average
particle size is directly proportional to the final intensity of the spectrometric signal
(111 in Figure 9(a)).12 For the successive injections method, Figure 9(b), the spectro-
metric signal indicates the start of the phase separation. Quantitative information
corresponds to a sharp decrease of the reflected light, i.e., the time period elapsed
until the beginning of the polymer precipitation, and qualitative information, relat-
ing to the number of and size of PCA particles, is given by the absolute intensity of
the reflected light at a pre-established wavelength (I1 in Figure 9(b)).
4.2 Effect of the Catalyst Nature

The effect of the catalyst chemical nature is demonstrated using two catalysts, sodium
hydride, NaH,and sodium bis(2-metoxy-ethoxy)aluminium hydride, RedAl@,which
react with L6 giving rise to two lactamate anions with different nucleophilicity. The
4000 I
E! 3500
3000
s,
a 2500
2000
8a4 1500
Q 1000
500
0
3500
3
I0
00
! 2500
28 moo
V
z
1500
1000
500
0
0 1200 2400 3600 4800 6000
(b) TimeIS
Figure 9 Interpretation of the spectrometric data for obtaining qualitative and quantitative
information about the morphological development of PCA particles; Illustration
of reaction calorimetry as a tool for process optimization: (a) continuous injec-
tion of CI; (b) successive injection of CI. The reaction temperature was 105 "C
and the intensity of spectrometric signal was recorded at A = 600 nm. In the
insets of (a), the corresponding SEM rnicrographies of the obtained particles are
given
70
+RedAl --a- NaH
-600 600 1800 3000 4200

Timeis
Figure 10 Heat rate of process versus time for the one-shot method of contacting the cat-
alytic species using diferent catalysts. Reactions conditions are similar to those
used in Figure 3
100 Chapter 8
second lactamate anion, the so-called "reduced lactamate salt," is characterized by a
reduced nucleophilicity and a lower dissociation degree of the lactam salt.13As shown
in Figure 10, the catalyst effect is quite similar to those of temperature, i.e., with
decreasing its nucleophilicity the typical shoulder of the heat rate evolution at 105 "C
is replaced by a two-peak curve, which characterizes the process below 90 "C.
4.3 Stirring level

Considering the requirement of high turbulence to enhance the mixing of reactant
streams to the desired degree of completeness on the molecular level,14 the hetero-
geneous character of the investigated process, and the presence of the intermediary
viscous particles, where the polymerization mainly proceeds, one should expect a
strong effect of the stirring level both on the high polymer yield and the particle size
distribution. The stirring effect was investigated by the one-shot method at 120 O C ,
with stirring rates ranging from 150 to 750 rpm. As Figure 1l(a) shows, the shape of
heat rate evolution was not affected by the stirring level, but the conversion strongly
increased at the stirring level above 500 rpm, reaching a limit at about 750 rpm,
s150 rpm --8- 280 rpm u420 rpm +560 rpm +750 rpm I
245 ,
T = I 20 "C
5 195
.?
145
E
0 600 1200 1800 2400 3000 3600
1
(4
-S
Timeis
150 rprn -+-280 rpm -+- 420 rpm
200
- 560 rpm -+ 750 rpm 1
1 60
J
(D
Tm 120
.
7
g 8 0
I
B
40
0
0 600 1200 1800 2400 3000 3600
(b) Timeis
Figure 11 Eflect of stirring level on the heat rate ( a )and overall heat (b)for the anionic solu-
tion polymerization of Mat 120 "C
Figure 1l(b). Further increase in stirring level above 1000 rpm (not given here) does
not change the limit of conversion but negatively influences the particle size distri-
bution due to the collision of the viscous particles.
5 Non-Isothermal Reaction Calorimetry

As mentioned in the previous section, the apparatus is able to work in non-isother-
ma1 conditions. This is a useful feature since in the chemical industry very often the
reactions are initiated on a temperature ramp. A first series of experiments was car-
ried out by successive injections of CI with the temperature ranging from 75 to
130 "C and a scanning rate of 0.12 "C min-' (see Figure 12). The first portion of CI
was injected just after the stable dynamic base line was reached; all other injections
were initiated after the heat evolution of the previous step had ceased. As expected,
the shape of the curves representing the heat evolution with polymerization changes
from two peaks at low temperatures (up to 95 "C) to one peak with a small shoulder
at the highest injection temperature (about 105 "C). The intensity of the first peak is
not influenced by temperature, but the intensity of the second one strongly increases
with temperature, and consequently the monomer conversion also increases.
The second series of experiments deals with the continuous manner of contact-
ing the catalytic species. Here, not only was the temperature scanned but also the
injection rate was continuously changed. The temperature was scanned between 75
and 135 "C with a positive or negative ramp, and the CI profile was linearly and
symmetrically varied on either side of the constant flow rate corresponding to the
typical process. The delivered amount of CI was the same for all runs. Figure 13
illustrates the two limiting situations when both variables were concomitantly
scanned in the same direction. Compared to the typical process (see Figure 8(b)),
no stationary state was achieved with either positive ramps for temperature and
injection rate (the upper thermogram) or negative ramps (the lower thermogram).
The lower thermogram resembles that corresponding to the one-shot method (see
Figure 3 at 105 and 120 "C) because the process temperature is the highest one and
the amount of injected CI is sufficiently high just from the beginning. The opposite
- 400
- 390 Y
-.
-380 g
CI
- 370 $
n
-360
I-
- 350
-5 ! ' ' ' ' ' : ' ' ' ' ' : ' ' ' ' ' ; ' ' ' ' ' ~ ' ' ' ' ' ; I 340
0 3600 7200 10800 14400 18000

Timels
Figure 12 Non-isothermal reaction calorimetry on anionic polymerization of M using the

successive injections method of contacting the catalytic species
102 Chapter 8
49 5
v 0.12 "Clmin
39 5
5 29.5
f ':::
.I-
pjq
2.
c
2
CI
70
Volume of C.I.
g 60
50
40
"i
Temperature
10
0 t"
0 3600 7200 10800 14400 18000 21600 25200
Time/s
Figure 13 Effect of concomitant scanning of temperature and injection rate of CI on the

anionic polymerization of L6
situation is encountered in the second case where the heat rate continuously
increases and then rapidly declines to the base line after the total consumption of
the monomer. It is well known that the reaction conditions favouring pre-oriented
particle size and a narrow size distribution are not suitable from the viewpoint of
the average molecular weight or average molecular weight homogeneity. This is
why concomitant scanning of the two variables should be a very useful technique
for controlling the particles size and the size distribution as well as the molecular
weight and molecular homogeneity.
6 Conclusion
Anionic polymerization of L6 in aliphatic solvents was monitored on-line using both
isothermal and non-isothermal reaction calorimetry in combination with Vis-spec-
trometry.
The temperature governs the "calorimetric signature" of this process and the analy-
sis of the heat released, in combination with other off-line techniques, allows to discern
the sequence of the events involved and to quantify the heat released in each event.
This rapid process can be tempered either by successive or continuous addition of
the chain initiator. In the first case, the overall heat evolution is split into a number
of smaller distinct sequences and spread over the time scale. At the same tempera-
ture, each sequence presents the same shape of heat evolution with the correspon-
ding overall evolution. In the second case, a steady state of heat evolution is
achieved, whose amplitude and half-width are proportional to the feeding rate of CI.
Concomitantly scanning one or more variables should be an alternative route for
controlling both the average molecular weight and the particle size distribution. This
aspect is still under investigation.
References
1. J. Pastrk, A. Zogg, U. Fischer and K. Hungerbiihler, Org. Process Res. Develop., 2001,5,
158.
2. S . L. Randzio, Ch. Stachowiak and J-P. E. Grolier, J. Chem. Thermodynamics, 2003, 35,
639.
3. F. Dan, J-P. E. Grolier, Setaram News, 2002, 7, i3.
4. B. Stratula-Vahnoveanuand C1. Vasiliu-Oprea, Polym. Plast. Techn-Eng.,2002,41(5) 981.
5 . S. Chrzczonowicz,M. Wlodarczyk and B. Ostaszewski, Makromol. Chem., 1960,38,159.
6. C1. Vasiliu-Oprea and F. Dan, J. Appl. Polym. Sci., 1996, 62, 1517.
7. F. Dan and C. Vasiliu-Oprea, Colloid Polym. Sci., 1998,276,483.
8. H. Sekiguchi, Ring-Opening Polymerization, Vol 2, J. Iving and T. Saegusa, (eds),
Elsevier, London, 1984, 833.
9. J. Stehlicek and J. Sebenda, EM Polym. J., 1986,22, 769.
10. C1. Vasiliu-Oprea and F. Dan, J. Appl. Polym. Sci., 1997, 64, 2575.
1 1 . F. Dan and C1. Vasiliu-Oprea, J, Appl. Polym. Sci., 1998, 67, 23 1.
12. J-P. E. Grolier, F. Dan, S. Boyer, M. Orlowska, and S. L. Randzio , Znt. J. Thermophys.,
2004,25 (2), 297.
13. C. A. Veith and R. E. Cohen, ACS Polym. Preprints, 1989, 31, 42.
14. J. H. Rushton, E. W. Costish and H. J. Evert, Chem. Eng. Progr., 1950,46, 395.
CHAPTER 9
Microcalorimetry for the

Pharmaceutical Zndustry
ANTHONY E. BEEZER
1 Introduction
Calorimetry and thermal methods have been practised in the Pharmaceutical
Industry for some This chapter is, however, only concerned with applications
of isothermal heat conduction microcalorimetry in the Pharmaceutical Industry
(Chapter 10 describes Industrial Applications of Flow microcalorimetry, a topic not
covered here). A feature observed in recent years is the growth (albeit small) in the
number of Companies that offer broadly equivalent-in-performance microcalorime-
ters that operate over a range of modes (e.g., Thermometric AB, Jiirfalla, Sweden;
Setaram, Caluire, France; Calorimetry Sciences Corporation, American Fork, UT,
USA). Thus, these manufacturers all supply instruments with nanowatt detection
limits complete with a range of specialised options for titration, humidity control,
perfusion, etc. (for a general review of sensitivities of different instruments, see ref.
3). In addition, all supply solution calorimeters, high sensitivity scanning calorime-
ters and, more recently, multi-channel instruments (e.g., Thermometric offers a 48-
channel instrument). These multi-channel calorimeters generally sacrifice some of
their sensitivity and detection limit capacity in the pursuit of higher throughput rates.
This chapter will not describe in any detail the operation, design and techniques of
microcalorimetric experiments; it will, however, concentrate on the applications to
which microcalorimetry has been put. Calorimetry has several advantages that are of
significance to pharmaceuticals. It is non-invasive, non-destructive and capable of
dealing with all phases and with mixed phases. Thus, it can accommodate solids, liq-
uids, gases (and their combinations), suspensions, emulsions, creams, ointments, etc.
However it is essential, in an industrial context, to have traceable and validated
bases for the determination of target parameters whatever the experimental technique
used. Microcalorimetry is no exception and a standard test and reference reaction has
been described4that fulfils this need. The imidazole-catalysed hydrolysis of triacetin
has, as a result of an international multi-laboratory study, been selected as this test and
Microcalorimetryfor the Pharmaceutical Industry 105
reference system. The determined values for the enthalpy and (second order) rate con-
stant are: A@= -91.723.0 kJ mol-’ and k=2.80+0.10X lov6dm3 mol-’ s-l.
Calorimetric techniques have been applied across the range of issues concerned
with drug development from lead discovery to formulated medicine and in aspects
of this process - discovery, characterisation, formulation development, process
development and scaleup - that are the subject of this chapter. As most studies have
been concerned with stability and compatibility issues, it seems logical to consider
these applications in almost reverse order.
2 Stability Determinations
Conventionally, stability testing is performed5 in an accelerated reaction regime i. e.
at elevated temperatures (>323 K) and at controlled humidity (say 70% RH).
Derived reaction rate constants are used to predict, through application of the
Arrhenius’ equation, the rate constant at the proposed storage conditions of the med-
icine, This extrapolation depends on the constancy of the reaction mechanism over
the temperature range concerned. It would clearly be better to have direct determi-
nation of reaction rate constants under the storage environmental conditions.
Until around 1995, it was unusual to find kinetic data determined from isother-
mal microcalorimetric studies - and this was especially true for long, slow reaction
systems. However, the modern microcalorimeter has excellent long-term stability
( 1 pW or better over 24 h) and hence slow reactions can readily be investigated
even those for which a complete reaction is not observed. Willson et aZ.”* showed
that quantitative kinetic and thermodynamic data could be determined for reactions
that had half lives of up to 2500 years. Commencing with a conventional (here sim-
ple for clarity) rate expression such as
Rate=dxdt=k(a -x)n (1)
where k is the rate constant, a the initial quantity (it is trivial to incorporate a volume
term, however, as the enthalpy H is a molar term; it is more ~traightfonvard~*~ to out-
line the equations with quantity), x the extent of reaction to time t and, recognising
that H=q/x where q is the number Joules associated with reaction to produce x mole
of product and Q the number of Joules associated with complete reaction then
It is this calorimetric form of this equation (and the more complex equations that can
be developed and used to describe sequential, parallel and more complex reacting
systems) that can be analysed to yield values for the target parameters, n, k, H , K , G,
S and Ea. References 6 and 7 provide details on the derivation of calorimetric kinetic
equations and describe how to manipulate calorimetric data to determine equilibrium
constants, Gibbs functions, entropies and activation energies. The approach
described shows that, under normal storage conditions (“room temperature” i. e. 298
K and ambient humidity), it is possible, from only 50 h of power time (the calori-
metric output is of W vs. t ) data, through these techniques, to distinguish between a
106 Chapter 9
reaction with a (first order) rate constant of 1 X lo-" s-l and a reaction with a rate
constant of 2~ lo-" s-l.
In the most detailed studies8*10published to date raw drug, benzoyl peroxide, its
compatibility with selected excipients (see Table 1) and its presence at a range of
strengths in formulated materials have been investigated (see Table 2).
It is notable that it is possible to select, from a range of excipients from, here, only
12 h study of each system, those that contribute to greater drug stability and those that
accelerate drug degradation. Moreover, Table 2 and Figure 1 show that, over the tem-
perature range studied, there exist two regions exhibiting different activation energies.
These observations emphasise the conclusion that a simple dependence on an
Arrhenius' extrapolation from high-temperature studies can be unreliable. In the
worst case, the derived rate constant for a 1% formulation in the absence of Carbopol
is 9 orders of magnitude in error at 298 K.
Table 1 Microcalorimetric kinetic stability studies of degradation of BPO in the

presence of excipients in water at 313 K
Excipient First-order rate constant, k (s-l )
- 2.00X 10-7+7.13X lop9
2% Poly trap 1.04X 10-7t2.64X
1% Sipernat 1.11x 1.51x lo-*
0.15% Monawet 3.40X 1.62X
5% Propylene glycol: 2.5% glycerine: 0.15% monawet 8.42X 10-956.76X lo-''
5% Propylene glycol: 2.5% glycerine: 2% polytrap 3.41 X 10-723.75X lop9
Pluronic" P234 7 . 2 4 1~0 - 9 ~ 5 . 0 010-lo
x
"(PO)-2250 M(EO)-1500. T, 33.43 and MW,,,,,,,=3750. 5 mg ml-l in water cooled to 4 "C.
Table 2 Microcalorimetric studies of solid BPO and formulations (first-order rate

constants: s - l )
Temperature Solid BPO 2.5% Placebo 2.5% I%
( "C) (reference (reference(reference (reference (reference
talc) talc) talc) placebo) w/o carbopol)
15 - - 5.03X10-8 - -
52.30X lo-'
20 1.16X10-9 2.61X10-9 5.16X10-8 1.59X10-* l.llX10-9
? 1 . 0 1 ~ 1 0 - ~t1.01X10-8 +1.60x10-s 51.12x10-8 +7.51X10-10
25 4.33~ 1.36x 5.29x 2.04x10-8 2.57X
?3.59x10-9 +2.51X10-8 t8.73x10-9 ~ 7 . 7 6 X 1 0 - +6.41X10-9
~~
30 6.85~ 2.34x 5.65~ 3.93X 4.01 X
5 1.33X lo-* 5 1.64X 28.91 X lop8 2 1.38X +6.07X lop9
35 2 . 0 0 lo-*
~ 5.27x lo-* 6.94~ 4.34X lo-* 5.42X10-9
+2.83x10-* 1+_6.36x10-9 +7.14X10-8 t1.25x10-* +-3.34x10-9
40 5.84X 3.43X 1.2OX lop7 1.29X lop7 1.31X
52.53~ 22.1 1X 27.70~ + 1 . 2 2 ~ +6.52X lop8
45 1.07X10-7 7.68X10-7 1.89X10-' 1.35X1OP7 1.74X10-8
+8.49X1OP9 +1.94X10-7 +3.68X10-8 +l.04X10-9 48.90X10-'0
E, (kJ mol-') 138 (?7) 172 ( t 1 4 ) 5 (52), 72 (+9) 84 ( 5 6 )
82 (23)
First-order rate constants (s-I).
Each of the reported experiments took 12 h observation and the whole study was
completed in significantly less time (around 2 weeks) than a conventional high-tem-
perature study would require. Moreover, the studies were performed on the “as is”
formulated medicines without the need to develop assay systems (e.g. HPLC analy-
ses). Further studies of this type on, for example, solid-state systems” and solution-
phase systems1*can be found in the literature.
The problems associated with freeze drying of peptides and proteins for thera-
peutic use have also received calorimetric attention recently - particularly, attempts
to understand and interpret the dynamics of amorphous s01ids.l~Structural relax-
ation time is a measure of molecular mobility involved in enthalpy relaxation and
thus is a measure of the dynamics of amorphous (glassy) solids. These dynamics are
important in interpretation of the physicochemical properties and reactivities of
drugs in amorphous formulations. The authors conclude that “microcalorimetry may
provide data useful for rational development of stable peptide and protein formula-
tions and for control of their processing”.
3 Characterisation of Physical FordAmorphous

Content
Recent review^'^.'^ have highlighted the capacity of microcalorimetry to provide
sensitive and detailed information on the effects of processing and manipulation of
drug substances on their physicochemical properties and the extent of amorphicity
present in a sample. Using lactose as an example, the amorphous content of a sam-
ple has been determined to better than 0.5% (PXRD studies can usually only deter-
mine amorphous contents at 10%).The experimentation is simple and rapid. Surface
108 Chapter 9
energetics,16cry~tallisation,’~ dissolution, etc. are also discussed in detail in the cited
reviews. More recently, it has been shown1*that microcalorimetry can be used suc-
cessfully to study the kinetics and thermodynamics of a polymorphic transition. The
object of the study was Seratrodast (a Takeda Chemical Industries product). The con-
clusions of the study were that: microcalorimetry can, with sensitivity discriminate
between those outputs that are the sum of individual reactions associated with par-
ticular size fractions, significant saving in time can be made in such studies and no
ancillary analytical method is required to facilitate the study.
4 Drug/Receptor Studies by Titration Calorimetry (ITC)

Again, there are several review~’~-~l that treat this particular area in detail. Titration
calorimetry is unique in that it allows direct determination of ARG (= -RT In K ,
where K is the equilibrium constant); A f l and A$ from one experiment. Thus, the
characterisation of binding interactions that are important in improving understand-
ing of biomolecular recognition and informing the process of rational drug design
processes is made more amenable to study. For details, refer to these reviews but it
is worth noting here that major issues remain - these are mostly associated with data
interpretation and centre upon the role of structural changes and their impact upon
the derived thermodynamic parameters. A further limitation associated with such
isothermal titration calorimetric (ITC) studies is that the current equipment allows
only one experiment to be conducted at a time - multi-channel instruments of nor-
mal design do not exist.
5 Multiple Miniature Calorimeters and their Potential

As noted in Section 4 ITC, studies are limited in terms of throughput; thus, it is
unlikely that these types of studies will contribute significantly to high-throughput-
screening (HTS). The impact of combinatorial chemistry in drug synthesis makes
the demand on HTS ever greater. However, modern technologies (nanotechnology)
suggest that the landscape of calorimetry is changing. Whilst it is not possible to cite
industrial applications of this new technology, it is clear that its potential in indus-
trial contexts is enormous. To date, 3 Web sites have been located that offer high
throughput calorimetric experimentation and indeed one of these sites explicitly
offers HTS on a contract basis. Xensor (www.xensor.nl) offers “chip” calorimeters
- four calorimetric units on a 20 mm square silicon chip. These units are simply sold
to interested calorimetrists who can assemble them in any array and numbers
required. The chips are cheap and relatively robust. Veeco (www.veeco.co.uk) can
supply an eight bimetallic cantilever system that can be used to study, for example,
protein ligand binding. The system is based on Atomic Force Microscopy technol-
ogy and offers considerable potential for calorimetric development. Vivactiss
(www.vivactiss.com) offer 96, 384 and 1536 well differential calorimeters for con-
tract work in the area of “enzymekatalyst discovery and optimisation and physico-
chemical characterisation”. These instruments have very small heat capacities and
hence require little in the way of equilibration. They work with nanolitre quantities
Microcalorimetry for the Pharmaceutical Industry 109
of materials and, by design, will only permit the study of rapid reactions - they have
no long-term stability properties.
6 Conclusions
This overview has attempted to show that calorimetry has found applications in all
areas of the pharmaceutical industry from discovery through receptor site binding to
characterisation, on through compatibility to formulation and stability. The
nanocalorimeters should find their real application in HTS whilst the more macro (but
still micro!) calorimeters will still be needed, because of their better long-term stabil-
ity to define the physicochemical properties of the selected pharmaceutical systems.
References
1. J. L. Ford and P. Timmins, Pharmaceutical Thermal Analysis, Techniques and
Applications, Ellis Horwood, Chichester, 1989.
2. J. L. Ford and R. J. Willson, in Handbook of Thermal Analysis and Calorimetry, Vol 4,
R. B. Kemp (ed), Elsevier Science, Amsterdam, 1999.
3. L. D. Hansen, Pharm. Technol., 1996, 20, 64.
4. A. E. Beezer, A. K. Hills, M. A. A. O’Neill, A. C. Morris, K. T. E. Kierstan, R. M. Deal,
L. J. Waters, J. Hadgraft, J. C. Mitchell, J. A. Connor, J. E. Orchard, R. J. Willson, T. C.
Hofelich, J. Beaudin, G. Wolf, F. Baitalow, S. Gaisford, R. A. Lane, G. Buckton, M. A.
Phipps, R. A. Winneke, E. A. Schmitt, L. D. Hansen, D. O’Sullivan and M. K. Parmar,
Thermochim. Acta, 2001,380, 13.
5. M. E. Aulton (ed), Pharmuceutics: The Science of Dosage form Design, Churchill
Livingstone, Edinburgh, 1988.
6. R. J. Willson, Isothermal Microcalorimetry : Theoretical and Experimental Developments,
Ph.D. Thesis, University of Kent, 1995.
7. R. J. Willson, A. E. Beezer, J. C. Mitchell and W. Loh, J.Phys.Chem., 1995, 99, 7108.
8. A. E. Beezer, A. C. Morris, M. A. A. O’Neill, R. J. Willson, A. K. Hills, J. C. Mitchell and
J. A. Connor, J. Phys. Chem .B, 2001,105, 1212.
9. F. Zaman, A. E. Beezer, J. C. Mitchell, Q. Calrkson, J. Elliot, M. Nisbet and A. F. Davis,
Znt. J. Pharm., 2001, 225, 135.
10. F. Zaman, A. E. Beezer, J. C. Mitchell, Q. Clarkson, J. Elliot, A. F. Davis and R. J.
Willson, Int. J. Pharm., 2001, 227, 133.
11. M. A. A. O’Neill, A. E. Beezer, A. C. Moms, K. Urakami, R. J. Willson and J. A. Connor,
J. Therm. Anal. Cal., 2003,73,709.
12. R. J. Willson, A. E. Beezer and J. C. Mitchell, Znt. J. Pharm., 1996, 132, 45.
13. J. Liu, D. R. Rigsbee, C. Stoltz and M. J. Pikal, J. Pharm. Sci., 2002,91, 1853.
14. M. A. Phipps and L. A. Mackin, Pharm. Sci. Technol. Today, 2000,3,9.
15. G. Buckton, Pharm. Tech. Europe, Nov. 2001,44.
16. G. Buckton and A. E. Beezer, Int. J. Pharm., 1988,41, 121.
17. V, P. Lehto and E. Laine, Pharm. Res., 1997, 14, 899.
18. K. Urakami and A. E. Beezer, Znt. J. Pharm., 2003, 257, 265.
19. J. E. Ladbury and B. Z. Chowdhry (eds), Biocalorimetry, Wiley, Chichester, 1998.
20. S. E. Harding and B. Z. Chowdhry (eds), Protein-Ligand Interactions, Oxford University
Press, Oxford, 2001.
21. G. A. Holdgate, Bio Techniques, 2001,31, 164.
CHAPTER 10
Isotheimal Flow-Microcabrimetry:
Principles and Applications for
Industry
MICHAEL A. A. O'NEILL
1 Introduction
The use of isothermal microcalorimetry is now widely accepted as a powerful tool
for the study of a vast array of systems within industry. Microcalorimetry is routinely
used for compatibility studies' (active-excipient, excipient-excipient interactions),
stability213(shelf-life), degradati~n,~physical properties such as polymorphism5 and
crystallinity,6 and can be used to probe surface characteristics of solid material^;^
some of these concepts are dealt with in more detail (with particular attention to
pharmaceutical issues) by Beezer in Chapter 9. Generally, the information sought on
a given system is related to the kinetics of that system (e.g. shelf-life). It is essential
therefore that accurate/quantitative information can be derived for parameters such
as rate and the rate constant. Also important is the quantity of material undergoing
change in the system, which can be explored through the magnitude of the calori-
metric signal and calculation of the enthalpy for that reaction. Knowledge of these
parameters then allows a host of other thermodynamic and kinetic information to be
derived from the system under 8-1 A large proportion of calorimetric studies
are performed using statidbatch techniques, however, studies using batch mode
techniques are generally unsuitable for the study of reactions that have a relatively
short half-life such that the reaction is complete before the preparatory work
required for static systems is carried out3 (approximately 1 h), such as some
enzyme/substrate systems, and the activity of drugs and antibiotics on microbiolog-
ical systems. Static techniques can be problematic for systems where sedimentation
can be an issue e.g. cellular systems.'* These types of system are difficult to study
using conventional techniques and usually require invasive sampling or study under
non-relevant conditions (e.g, plating out of micro-organisms). Flow calorimetric
techniques are also necessary for the study of systems in which the surface is to be
Isothemzal Flow-Microcalorimetry: Principles and Applications for Zndustry 111
probed by passing a gas or vapour over it, such as surface characterisation of solids
for e ~ a m p l eFor
. ~ studies on systems such as those described above, it can be advan-
tageous to employ flow-through calorimetric techniques.
Microcalorimetry provides a rapid, non-invasive approach to the study of such
systems; moreover, flow calorimetry has a sensitivity that is superior to many other
traditional techniques allowing accurate, quantitative information to be derived from
the raw data.
2 Theory
Calorimeters rely on the observation that all processes, chemical and physical,
involve an exchange in heat energy to or from their surroundings. The calorimeter is
non-specific in its operation, i.e., it will monitor and record all reactions that occur.
There are several other advantages conferred by isothermal microcalorimetry. The
isothermal Thermal Activity Monitor (TAM, Thermometric AB, Jarfalla, Sweden) has
a combination of sensitivity and versatility that is superior over many other analytical
techniques. It has been reported13 that the TAM has sufficient sensitivity to monitor
slow reactions with lifetimes of up to 10,000 years. It has also been reported13that
only 50 h of data are required to discriminate between a reaction that has a first-order
rate constant of lXIO-lls-l and a reaction with a first-order rate constant of
2X 10-l s-l. The calorimeter is non-invasive to the sample under study and hence the
reaction is not influenced by the calorimeter, nor does the reactant require special
preparation before study. The sample can be in any physical state: solid, liquid, gas or
any combination. The main advantage of the calorimeter is its ability to restore a wide
variety of information that can be used to characterize a reaction system. Calorimetric
data, with the aid of the relevant equations, allow a number of parameters to be deter-
mined. These parameters can then be used to characterize a reaction in terms of the
reaction rate and mechanism. Each parameter contributes to the characterization by
defining different aspects of the kinetic, thermodynamic and mechanistic properties
of the reaction. By comparing this information between different types of reaction
schemes and mechanisms, it might be possible to determine the trends or patterns that
can be applied to unknown compounds or to facilitate the design of new compounds.
Each of the parameters determined from the calorimeter makes a unique contri-
bution to the understanding of the rate and mechanism of reaction. The mechanism
can be deduced from reaction order, n, activation energy, Ea,change in enthalpy, H,
change in entropy, AS, the quantity of material reacted and the experimental design.
The order of reaction governs the dependency of rate on the reaction components.
Determination of the reaction order with respect to individual components will yield
the dependency of the rate with the individual reactants. Ea,H and AS may yield the
information required to elucidate the process that is occurring.
3 Flow Calorimeter Design

There are a number of commercially available microcalorimeters that are routinely
employed within the chemical industry. Some are specifically designed to be operated
112 Chapter 10
solely as a flow calorimeter, such as the dedicated flow insert designed by
Therm~metric'~and those manufactured by Micro~cal.'~Others, such as the
Setaram16micro-DSC 111, have been designed such that they can be adapted for use
as flow calorimeters through application of specially modified calorimetric ampoules.
Calorimeters such as those developed by Thermometric and Setaram operate on a dif-
ferential basis, i.e. they have a twin channel configuration incorporating a reaction (sam-
ple) channel with a reference channel. This arrangement gives, as the calorimetric
output, a signal that is the differential of the sample and reference signals. This confers
the advantage of reducing noise in the signal caused by external factors such as tem-
perature fluctuations, electrical interference, etc. Moreover, if a suitable reference mate-
rial is chosen, then the observed signal will be a direct reflection of the thermal process
under investigation. Ideally, for flow calorimetric studies, one would use an inert refer-
ence solution flowing at a rate identical to the sample solution. In practice, this is diffi-
cult to achieve, therefore, one can either have the reference side filled with solvent and
kept static or have it empty and seal it off. In this instance and for calorimeters that do
not operate on the differential principle (such as those manufactured by Microscal), it
becomes necessary to record a baseline before the experiment is conducted by flowing
the solvent (or inert gas in the case of Microscal instruments) prior to the experiment.
It should be noted that, in principle, all the flow calorimeters discussed here are
capable of operating both in the gashapour phase and in the solution phase.
However, the calorimeters manufactured by Setaram and Thermometric are gener-
ally used for solution-phase studies, whereas the Microscal instruments are designed
specifically to facilitate gashapour-phase studies. For the purposes of this discus-
sion, gaseous-phase flow calorimetry will centre around a consideration of the
Microscal instruments, and solution-phase calorimetry will centre around the
Setaram and Thermometric instruments.
3.1 Solution-Phase Flow Calorimeters

There are two possible modes of operation for flow microcalorimeters used for solu-
tion-phase ~ t u d i e s . ~ J ~
3.1.1 Flow-Mix
The flow-mix mode of operation allows a fast reaction (i.e. the reaction is rapid in
comparison with the time constant of the instrument). Here, the reaction is initiated
inside the calorimetric cell by the mixing of the pre-equilibrated reagents as they enter
the cell, Figure 1. The effluent is not recycled but can be retained for further analysis.
Note that it is not possible to distinguish kinetically between different rapid (with
respect to the time constant of the instrument) reactions, i.e. the calorimetric signal
appears zero order in nature regardless of the actual reaction order.
3.1.2 Flow-Through
The flow-through cell is only suitable for relatively slow reactions with long
half-lives (i.e. the reaction is slow in comparison with the time constant of the
I s o t h e m l Flow-Microcalorimetry: Principles and Applications for Industry 113
Figure 1 Schematic of mixing cell
Thermostat
h I Calorimetric
Ce 11
To Waste
React ion#
Reaction
Vessel
Figure 2 Schematic for the flow through cell
instrument). In this mode of operation, the reactants are pre-equilibrated external to

the calorimeter and the reaction initiated also external to the calorimeter. The react-
ing solution is either allowed to flow-through to waste (or further analysis) or is
recycled to the reaction vessel, Figure 2.
In both instances, the calorimeter records the change in heat through semi-con-
ducting thermopiles positioned around the calorimetric cell. The data are recorded
as a function of time and is displayed as J s-l.
3.2 Gaseous-Phase Flow Calorimeters

The basic operation7 of the gaseous flow calorimeters is essentially identical to that
of the flow-through solution-phase calorimeters with an external gas/vapour source
that is passed, through a single calorimetric cell, across the solid of interest and the
resulting heat change measured. For these instruments, the detectors are thermistors
in direct contact with the solid under study. The form of the returned data is volts as
a function of time. The signal can be converted to J s-l via a calibration constant.
4 Flow Calorimetric Equations

As already mentioned all chemical and physical processes result in a change in heat.
Depending on whether the process is endothermic or exothermic, heat will flow
from, or to, the heat sink in order to maintain isothermal conditions. This heat
114 Chapter 10
flowlthermal power is measured as a change in heat energy (joules) per unit time
(seconds) by the thermopiles and is expressed as dqldt (or $); it has units of J s-l (or
watts).
The isothermal microcalorimeter can therefore yield two types of data: heat flow
(a kinetic term) and the time-independent reaction enthalpy change (a thermody-
namic term). It is possible, in principle, then to derive thermodynamic and kinetic
information from the raw calorimetric data.
The equations discussed here are specifically for solution-phase flow calorimetric
experiments.
Equations that describe calorimetric output can be d e r i ~ e d ~ *from
' ~ J *basic kinetic
expressions such as that seen in Equation ( 1 ) .
dx
-=k( C-xY
dt
where dxldt is the rate of reaction (mol dm-3 s-'), k is the appropriate rate constant,
C is the initial concentration of the reagent (mol dm-'), x is the quantity of the
reagent reacted to time t (mol dm-3), n is the reaction order and t is the time (s).
Calorimetric equations that describe the output for flow calorimetric data were
first derived by Beezer and Tyrell some 30 years ago.197 2o Equations (2), (3) and (4)
describe first-, second- and zero-order reactions systems, respectively.
dq kC2HVc
-= (3)
dt (1 + k t C + k z C ) ( l +ktC)
d4
-=-kVcH (4)
dt
where: dqldt is the calorimetric output (J s-'), F is the flow rate (dm3 s-l), H is the
enthalpy of reaction (J mol-'), k is the appropriate rate constant, z is the residence
time (s) and, Vc is the effective thermal volume (dm3).
Equations (2)-(4) show that elucidation of the rate constant requires prior knowl-
edge of the residence time, z (the time the reacting solution spends in the calorimet-
ric vessel) and hence the thermal volume, Vc (i.e. the operational volume of the
calorimeter). Determination of reliable values for rate constants and enthalpy
changes from experimental data (power, time data) for reacting systems, studied by
flow microcalorimetry; therefore, it requires an accurate and precise value for z at
any given flow rate. This is determined from Equation (5) through knowledge of F
and Vc.
Flowing reacting systems necessarily result in transport of heat from the detection
area. Thus, whilst the physical volume of the cell can be known, this volume may
not be the effective or thermal volume nor indeed the zero-flow-rate volume of the
cell, Figure 3.
Isothermal Flow-Microcalorimetry: Principles and Applications for Industry 115
OUT
Figure 3 Schematic for the flow-through calorimetric cell
5 Validation of Flow-Through Microcalorimeters

The determination of an effective or thermal volume is a generic problem associated
with all flow microcalorimeters operating in flow-through mode. The thermal vol-
ume, Vc,is the volume that the calorimeter detects as the medium flows through the
vessel. It is affected by several different factors but the most important is the flow
rate of the medium since, as the reacting solution flows through the vessel, a pro-
portion of the heat will be carried out of the calorimetric cell retained in the medium,
undetected21’22 (i.e. related to the heat capacity of the flowing medium). As the flow
rate is increased, more heat will be carried out by the flowing medium and hence a
smaller thermal volume will be observed. Conversely, as the flow rate is decreased,
less heat will be lost such that as the flow rate approaches zero, the apparent thermal
volume will approach a limiting value. It is clear therefore that, if quantitative ther-
modynamic and kinetic information is to be obtained from flow calorimetric data,
there is some means of validating the calorimetric output in terms of the flow rate
dependence of thermal volume.
This has been achieved,21*22 for solution-phase flow calorimeters, through the use
of a chemical test and reference reaction: the base-catalysed hydrolysis of methyl
4-hydroxy-benzoate (methyl paraben).
Note that for reactions that are rapid (with respect to the time constant of the
instrument), this validation is not necessary since the reaction is initiated inside the
calorimetric cell, and has no measurable kinetic dependence (i. e. instantaneous) and
the required thermodynamic parameters can be calculated through the use of a cali-
bration constant.
S t ~ d i e s using
~ , ~ ~the, ~base-catalysed
~ hydrolysis of methyl paraben (BCHMP) test
and reference reaction have been conducted with a variety of different solution-
phase flow calorimeters. The results obtained from these studies have shown that the
flow rate dependency of the thermal volume is different for each of the instruments
used and indeed for each experimental arrangement (e.g. sample and reference cell
set-up). The determined value for Vc can differ by as much as 15% (over a range of
experimental flow rates) from the nominal engineered volume (typically approxi-
mately lml). This effect can be minimised by careful design of the flow cell and also
by careful consideration of the sample and reference cell arrangements (more details
can be found in ref. 22 and references therein).
Knowledge of how thermal volume varies with flow rate allows the physical char-
acteristics of the flow cell to be investigated. For example, it is possible to simulate
116 Chapter 10
.l.O con"""
Figure 4 A 3 0 representation of the effect of varying theflow rate and the initial concentra-
tion, [A,], on the observed initial output for ajixed rate constant
data for a range of flow rates and investigate the interdependence of flow rate, rate
constant and concentration.
It should be noted that control is achievable over the flow rate, the concentration
and to a lesser extent over the rate constant, the latter by changing the temperature.
However, it is not possible to alter the enthalpy, assuming that it has no temperature
dependence (it is possible, in some circumstances, to magnify the enthalpy by care-
ful choice of an appropriate buffer system). The changes in the concentration and the
rate constant must also satisfy the requirement that the mechanism of reaction is not
affected as a consequence. Changing the flow rate will have no impact on the mech-
anism since it is purely a physical change to the system.
Figure 4 demonstrates how the initial signal (Do (the maximum possible signal for
a single reaction process) varies with the flow rate for fixed enthalpy and rate constant
and varying initial concentration of the reagent. If the upper and lower detection lim-
its for the calorimeter are known, then it is possible to choose the combination of Val-
ues for the concentration and the flow rate that yield an appropriate initial signal.
It is apparent from these studies that all flow microcalorimeters should be vali-
dated in this way, if quantitative information is sought. To date,23similar studies have
not been conducted for validation of gas-phase flow calorimeters. It is likely, how-
ever, that the problems associated with the removal of heat by the flowing medium
will not be as significant in such systems since the associated heat capacity is much
lower for gaseous/vapour systems and hence the amount of heat lost in this way is
significantly reduced.
6 A Simple Flow Calorimetric Experiment

It has been noted throughout this chapter that flow microcalorimetry can be used for
a wide variety of reaction systems. This section will attempt to describe a simple,
solution-phase, calorimetric experiment and discuss methods for the elucidation of
relevant thermodynamic and kinetic parameters from the raw calorimetric data.
In general, the object of most calorimetric experiments is the quantification of
thermodynamic and kinetic parameters. Several recent developments now permit the
recovery of such parameters through calculational methods. These are not discussed
in detail here; for a more detailed discussion, see refs. 8-1 1.
A good example of the range of parameters available from flow calorimetric data
can be found from the study of enzymehubstrate systems. The kinetic nature of
enzyme systems has been previously described by Michaelis and Menten.24In the
treatment discussed here, the parameters sought are the enthalpy, rate constant,
Michaelis constant and the enzyme activity. The following example describes a
study on the well-known enzyme substrate system, uredurease.
6.1 Experimental Protocol

A variety of pH 7.0 buffered, stock solutions of urea at various concentrations are
prepared. The concentrations of urea chosen are such that the complete kinetic pro-
file of the reaction can be observed (i.e. from first-order through mixed-order and
finally to zero-order kinetics).
A 50 ml aliquot of the urea solution is pre-thermostated to the operational temper-
ature of the calorimeter; for these experiments the calorimeter is housed in a constant
temperature environment and operated at 25 "C. The urea solution is then run in a
continuous loop, at a known flow rate, until a stable baseline is achieved. This solu-
tion is then inoculated with 4.55 ml of a standard, fixed concentration, urease solu-
tion (also buffered to pH 7.0 and pre-thermostated) and the resulting calorimetric
output recorded as a function of time. This is repeated for all concentrations of urea.
Figure 5 shows a selection of typical calorimetric outputs for this enzyme system.
Calorimetric data are recorded as a function of time and hence the rate of reaction
can be determined from the calorimetric output. One of the distinguishing features
of Michaelis-Menten type kinetics is the form of the plot of rate vs. substrate con-
centration. Figure 6 shows such a plot and it is clear that it has a form that is con-
sistent with that predicted by the Michaelis-Menten kinetic theory.
0.00020
o.0°016:
7
2 0.00012 -
d
.
6 O.ooOo8-
0.00004 -
0.00000 -
l , . , . , . , . , . ,
0 1000 2000 3000 4000 5000
Time (s)
Figure 5 Calorimetric outputs forfirst,-mixed- and zero-order uredurease enzyme reactions

118 Chapter 10
0.00025
0.00020
-i
01
0.00015
v
u
Y
0.ooo10 w
w
0.oooO5
0.00000 l ' l ~ l ~ l ~ l ' l ~ l ~ l ~ l

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45
substrate concentration (mol dm-7)
Figure 6 Max rate vs. substrate concentrationfor the uredurease enzyme reaction
6.2 Determination of Thermodynamic and Kinetic Parameters from

Calorimetric Data
The form of the flow calorimetric equations described earlier can also be manipu-
lated3, 25 to incorporate Michaelis-Menten parameters such as the Michaelis con-
20+
stant, KM,as well as yielding the first order rate constant and enthalpy for the overall
reaction, Equation (6).
where KMis the Michaelis constant and [atot is the total enzyme concentration (mol
dm-3).
Note that it is trivial to calculate the first-order rate constant from the calorimet-
ric data; from Equation (6), it is simply the slope of the line for a plot of In dq/dt vs.
t. However second-order rate constants require a more complicated approach and are
accessible through a new calculational approach of data pairing, this is not consid-
ered further and more details can be found in refs. 11 and 26.
Once the rate constant for reaction is known, it then becomes possible to calculate
the enthalpy for the reaction for all time points of the reaction and obtain an average
for the lifetime of the first-order reaction. Once the enthalpy is found, it is possible
to calculate the value of k[EJ,,, from a consideration of the zero-order output and
hence the Michaelis constant is now accessible from the slope of the first-order In
dqldt vs. t plot (the slope of this plot is equal to -K[EJ,,JK,).
It must be emphasized that this analysis is not possible without prior knowledge
of the thermal volume of the calorimetric cell and therefore the validation routine
described above must be conducted if accurate values for the enthalpy in particular
are to be obtained.
This new method also allows the effects of inhibitors of enzyme activity and their
mechanism of action to be studied using flow calorimetry. More details of this study,
and others, can be found in refs. 3, 11 and 26.
7 Further Applications of Flow Microcalorimetry

The examples here provide a ‘snapshot’ of the range of potential applications that
flow microcalorimetry offers.
Using the techniques described earlier, a number of systems have been explored
through flow-calorimetric techniques. Early work in this area of calorimetry centred
on the study of micro-organi~ms.~~~ 28 It was shown that, in principle, it is possible to
distinguish between different microbial species by the shape of the calorimetric
curve produced during organism growth. Quantification of the number of organisms
present in the system is possible by comparison of the areas under the curve for an
uncharacterized system with that of a known standard.
Since calorimetric data are a measure of rate, it is possible to link heat output with
the metabolic activity of human (and animal) cells. Both types of cell have been
extensively studied by flow calorimetry; one example29is the study and optimization
of Chinese hamster ovary cells producing recombinant interferon y
Flow calorimetry has also been used to investigate dosage forms. Dissolution of
tablets under various simulated conditions has been explored,30for example. Here,
the tablet is presented with various solutions designed to mimic conditions in the
gastro intestinal (GI) tract (pH 7 buffer) and stomach (pH 3 buffer for fasting and
lipid solutions for fatty meals). The rate of dissolution can then be estimated for con-
ditions in vivo.
Gas-phase flow calorimetry has been used on a number of systems to investigate
a variety of topics. Some examples are discussed below.
The use of gas flow techniques are now widely recognized as important for the
study of ad~orptiorddesorption~~~~ phenomena at solid interfaces. Such studies are
particularly important for the characterization of catalytically active sites on the sur-
faces of solids. One particular example is a of the adsorption of ammonia
onto activated carbon. It can be shown that adsorption of ammonia consists of
reversible and irreversible steps and that these steps can be attributed to physisorp-
tion for the former and adsorption on particular chemical groups for the latter. The
heats of adsorption for various sites can be measured and the data can be used to
reveal the existence of a wide distribution of acid sites that are accessible to ammo-
nia to varying degrees.
Another example33of the use of this technique can be found in the determination
of the heats of adsorption of water onto microporous carbons. Here, the effects of
different gaseous carriers on the heat of adsorption and the amount of water adsorbed
are investigated. It has been shown that adsorption of water on this type of surface
is strongly influenced by the nature of the carrier gas from which the adsorption
takes place.
8 Conclusions
Isothermal flow microcalorimetry is now an indispensable tool for the study and
characterization of a variety of reaction systems within the chemical industry. It per-
mits the study of these systems as a function of time in a non-destructive, non-inva-
sive manner and hence allows accurate, quantitative information for both the
120 Chapter 10
thermodynamic and kinetic character of the system under scrutiny and is only lim-
ited by the sensitivity of the instrument in question. This information is vital for the
determination of essential parameters such as shelf-life and compatibility. Flow
calorimetry has also been used to probe surface characteristics such as area, active
sites, pore size, etc., information that is essential for the discovery of new catalysts
and production of efficient separation systems for example.
Calculational methods for the accurate determination of thermodynamic parame-
ters (particularly for solution-phase calorimetry) have only recently become avail-
able through the BCHMP chemical test and reference reaction. It is likely therefore
that values reported for enthalpy, for example, are likely to be significantly in error
for some of the earlier work using flow calorimetry. These errors can be rectified,
however, through the calculation of the thermal volume and relevant adjustment of
the calorimetric data.
Microcalorimetry is now firmly placed as an important tool within the chemical
industry. As with all current technology, the trend is to produce ever smaller instru-
ments that allow large numbers of parallel experiments to be conducted. This is evi-
dent through the d e v e l ~ p m e nof
t ~microscale
~ ~ ~ ~ calorimeters (so-called ‘calorimeter
on a chip’); such developments, however, do not preclude the need for flow instru-
ments and indeed such microscale instruments can, and almost certainly will, be
adapted for use as flow calorimeters.
References
1. F. Zaman, A. E. Beezer, J. C. Mitchell, Q. Clarkson, J. Elliot, M. Nisbet and A. F. Davis,
Znt. J. Pharm., 2001, 225, 135.
2. S . Gaisford and G. Buckton, Thermochim. Acta, 2001,380, 185.
3. M. A. A. O’Neill, Ph.D. Thesis, University of Greenwich. 2002.
4. M. Riva, D. Fessas and A. Schiraldi, Thermochim. Acta, 2001, 370, 73.
5. K. Urakami and A. E. Beezer, Znt. J. Pharm., 2003,257,265,
6. D. Gao and J. H. Rytting, Znt. J. Phamz., 1997, 151, 183.
7. A. J. Groszek, Thermochim. Acta, 1998,312, 133.
8. R. J. Willson, A. E. Beezer, J. C. Mitchell and W. Loh, J. phys. chem., 1995, 99, 7108.
9. A. E. Beezer, A. C. Morris, M. A. A. O’Neill, R.J. Willson, A. K. Hills, J. C. Mitchell and
J. A. Connor, J. Phys. Chem., 2001,105, 1212.
10. M. A. A. O’Neill, A. E. Beezer, A. C. Morris, K. Urakami, R. J. Willson and J. A. Connor,
J. T h e m . Anal., 2003, 73, 709.
1 1 . R. J. Willson and A. E. Beezer, Thermochim. Acta, 2003,402, 75.
12. Y. H. Guan, P. C. Lloyd and R. B. Kemp, Thermochim. Acta, 1999,332,211.
13. R. J Willson, Ph.D. Thesis. University of Kent. 1995.
14. www.Thennometric.com
15. www.Microsca1.com
1 6. www.Setaram.com
17. LKB 10700-1 Flow Calorimeter Operation Manual.
18. A. K. Hills. Ph.D. Thesis, University of Kent at Canterbury, 2001.
19. A. E. Beezer, H. J. V. Tyrell, Sci. Tools, 1972, 19, 13.
20. A. E. Beezer, H. Tyrrell, and T. Steenson, Protides of the Biological Fluids, Vol 20, H.
Peeters, (ed), Pergamon Press, New York, 1972, p. 563.
21. M. A. A. O’Neill, A. E. Beezer, C. Labetoulle, L. Nicolaides, J. C. Mitchell, J. Connor,
J. Orchard, R. B. Kemp and D. Olomolaiye, Thermochirn. Acta., 2003, 399, 63.
22. M. A. A. O’Neill, A. E. Beezer, G. Vine, C. Labetoulle, L. Nicolaides, R. B. Kemp, D.
Olomolayie, P. L. 0. Volpe and D. Oliveira, Thermochim. Acta., 2003,413, 193.
23. A. J. Groszek, Personal communication, 2003.
24. L. Michaelis and H. Menten, Biochem Z., 1913, 49, 333
25. M. A. A. O’Neill, A. E. Beezer, J. C. Mitchell, J. Orchard and J. A. Connor, Thermochim.
Acta, 2003, in press.
26. M. A. A. O’Neill, Thermochim. Acta, 2003, in prep.
27. A. M. James, Thermal and Energetic Studies of Cellular Biological Systems, Wright,
Bristol, UK, 1987.
28. A. E. Beezer, R. D. Newell and H. J. V. Tyrrell, J. Appl. Bacteriol., 1976, 41, 197
29. R. B. Kemp, Thermochirn. Acta, 2001,380,229.
30. L. J. Ashby, A. E. Beezer and G. Buckton. Int. J. Pharm., 1989,51, 245.
31. G. J. Zajac and A. J. Groszek, Carbon, 1997,35, 1053
32. M. Domingo-Garcia, A. J. Groszek, F. J. Lopez-Garzon and M. Perez-Mendoza, Appl,
Catal. A , 2002,233, 141.
33. A. J. Groszek, Carbon, 1997,39, 1857
34. C. Auguet, J. Lerchner, P. Marinelli, F. Martorell, M. Rodriguez de Rivera, V. Torra and
G. Wolf, J. Therm. Anal., 2003, 71, 951.
35. C. Auguet, J. Lerchner, V. Torra and G. Wolf, J. Therm. Anal., 2003, 71,407.
CHAPTER 11
Transport Properties and Industry

WILLIAM A. WAKEHAM AND MARK J. ASSAEL
1 Introduction
The transport of mass, momentum and energy through a fluid are the consequences
of molecular motion and molecular interaction. At the macroscopic level, associated
with the transport of each dynamic variable is a transport coefficient or property,
denoted by X, such that the flux, J, of each variable is proportional to the gradient of
a thermodynamic state variable such as concentration or temperature. This notion
leads to the simple phenomenological laws such as those of Fick, Newton and
Fourier for mass, momentum or energy transport, respectively,
J=-XVY (1)
Here, Y is the appropriate state variable conjugate to the flux J and X, and depends
on the thermodynamic state of the system. These linear, phenomenological laws are
fundamental to all processes involving the transfer of mass, momentum or energy
but, in many practical circumstances encountered in industry, the fundamental trans-
port mechanisms arise in parallel with other means of transport such as advection or
natural convection. In those circumstances, the overall transport process is far from
simple and linear. However, the description of such complex processes is often ren-
dered tractable by the use of transfer equations, which are expressed in the form of
linear laws such as
J = - CAY (2)
Here, the transport coefficient, C, is not simply a function of the thermodynamic

state of the system but may depend on the geometric configuration of the system
and the properties of surfaces for example. We are concerned here with the
transport properties, X, of materials, which depend on thermodynamic state
of the material only. In practical situations, the transport coefficients, C, will
often have been expressed as correlations with a parametric dependence on the
properties, X .
Transport Properties and Industry 123
The transport properties of a material determine the rate at which an initially non-
uniform thermodynamic state evolves towards a uniform state over time. For this
reason, from an industrial point of view, within the context of equipment or process
design, the thermodynamic properties of a fluid often determine the feasibility of
what is proposed, whereas the transport properties essentially determine how large
the equipment or process unit must be or the time scale of the operation. This essen-
tial difference is responsible for the fact that the transport properties have, tradition-
ally, been less emphasized in process design activities. However, as the needs for
process integration and energy minimization grow, there is a tendency to examine
more carefully the effects of the transport properties on the process design.
In Figure 1, we show a schematic diagram of a catalytic reactor that could be used
for the synthesis of methanol. The reaction takes place at 610 K and 30 MPa. The
feed, hydrogen and carbon monoxide, enters at the bottom of the reactor vessel and
is preheated to the required temperature by the hot product gases in the preheater.
Subsequently, hydrogen and carbon monoxide react in two catalyst beds. In order to
maintain the temperature for the exothermic reaction, a water-cooled interstage unit
is employed between the two catalyst beds, from which the product emerges to heat
the incoming gases. In the configuration shown, the entire reaction zone/cooler
assembly is contained in a single high-pressure vessel. This type of configuration has
the usual benefits of intensification, including a reduction in capital costs and energy
efficiency. However, it follows that the heat exchanger design of both pre-heater and
interstage cooler must be done very carefully so that they are able to perform their
required tasks. Under-design will lead to inefficient reactor operation with little
opportunity for remedial action, whereas over-design will unnecessarily increase the
size of the overall pressure vessel. The area of the interstage heat exchanger depends
on the viscosity, thermal conductivity and density of the gas mixture passing through
it as well as the properties of the cooling water. Indeed, the design area of the
exchanger, which may be taken to represent the design variable, is approximately
PRODUCT Water
Water
Preheater
Catalyst
bed
lnterstage
Unit
Catalyst
bed
610 K,30
FEED (H2 + CO)
Figure 1 Schematic diagram of a methanol catalytic reactor

124 Chapter 11
proportional to the ratio (viscosity/thermal conductivity) to the power 1/3.2,3Hence,
and irrespective of any optimizing design procedure employed, if the gas viscosity
is underestimated by 20% and its thermal conductivity is overestimated by an equal
amount, then the area of the interstage heat exchanger required will be underesti-
mated by 25% with a corresponding overestimate of the area from the reverse cir-
cumstances. In either case, the loss of efficiency or the unnecessary capital cost may
have significant economic effects on the operation of the process. Uncertainties of as
much as 20% in the properties of gas mixtures at high pressures and temperatures
were relatively common until the latter part of the last century. Great improvements
in methods of measurement and prediction have now been achieved for many sys-
tems of practical importance but not for all, so that the example given here remains
relevant.
Indeed, many other examples of the importance of transport properties to design
can easily be found. One novel aspect of this kind of problem is revealed by the
increasing focus on the very small systems. Miniaturized systems have many advan-
tages for a wide range of purposes, for example, reducing the size of sensors reduces
their thermal inertia and permits a more rapid response while the integration of sen-
sors and actuators in single devices is essential if micro-reactor technology is to
move from laboratory to prod~ction.~ In order to transport materials in such micro-
scopic environments, fluid flow provides the simplest means. Indeed, nature is the
best example of this since aqueous solutions are the main transport media in living
organisms. Hence, nano-and micro-technologies are likely to drive fluid transport
property research into relatively unexplored areas of size and materials. For exam-
ple, in the fast-evolving area of nanotechnology, it is exceedingly important to be
able to remove the heat generated in small components effectively. For this purpose,
the heat transfer characteristics of conventional fluids have been found to be inade-
quate so that it has recently been proposed that small amounts of nanoparticles might
be suspended in the fluid to enhance the thermal conductivity. There is now some
experimental evidence that such an enhancement does take place because a 0.6% by
weight addition of carbon, multi-walled nanotubes to water results in a 40% increase
in its thermal cond~ctivity.~
2 Measurements
The vast majority of accurate transport-property measurements have been performed
on molecularly simple pure fluids under conditions close to ambient pressure
and temperature. As one moves away from this set of circumstances, the amount of
available information decays rather rapidly and its accuracy declines dramatically.
The three transport properties of the greatest concern are the viscosity, thermal
conductivity and mass-diffusion coefficients. In each case, although measurements
had been conducted over a period of at least 150 years, it was not until around 1970
that techniques of an acceptable accuracy were developed for the relatively routine
measurement of any of these properties. There is ample evidence in the literature69
of very large discrepancies among measurements made prior to that date. One rea-
son for these discrepancies lies in the conflicting requirements that, to make a trans-
port-property measurement, one must perturb an equilibrium state but, at the same
time, make the perturbation as small as possible so that the property determined does
refer to a well-defined thermodynamic state. The latter requirement implies that the
signals to be measured in any such experiment are always small and up against the
limits of resolution. A second reason for the discrepancies arises from the failure of
some experimenters to develop rigorous working equations for their instruments
using the full conservation equations of continuum mechanics. The last 25 years of
the 20th Century witnessed a very considerable refinement in measurement resolu-
tion, and the theory of instruments has now obviated many of these difficulties.
However, it is still necessary to consider whether it is possible to perform easily
all transport-property measurements that industry requires. Let us assume that we
need to measure only three properties at just 10 temperatures and 10 pressures, for
15 pure fluids and all their possible multicomponent mixtures, at five compositions
in the liquid and gas phases. Then, the total number of measurements required is of
the order of 1 O8 (3 X 10X 10X 32766 X 5 X 2). If one further assumes that one can per-
form three measurements per day, then it is obvious that even for the above program
of measurements, 90.000 man-years will be required and it considered only 15 mate-
rials from among the set of several thousand involved routinely in industry. It is clear
that measurement alone cannot serve the industrial appetite for transport property
data. Nevertheless, accurate measurements will continue to be vital for the verifica-
tion of theory and the validation of predication.
In the following sections, the most important techniques employed at present for
the measurement of the viscosity, thermal conductivity and diffusion coefficients
will be briefly presented.
2.1 Viscosity
Since 1970, two generic types of viscometer have received the greatest attention; the
first makes use of the torsional oscillations of bodies of revolution and the second is
based on the rather simpler concept of laminar flow through capillaries. Both reduce
the measurement of viscosity to measurements of mass, length and time.
In the case of torsional oscillating-body viscometers, an essentially exact descrip-
tion of the motion allows measurements of high accuracy. In such viscometers, the
characteristics of the oscillator are affected by the presence of the fluid and its prop-
erties in a way that is readily measured. Thus, oscillating-disk viscometers have
found the greatest application to both gases and liquids under relatively mild condi-
tions of temperature and pressure. They have been especially important in the deter-
mination of gases at low density and high temperature in work pioneered by Kestin
and his collaborators8 and continued by Vogel and his group.' For work on molten
metals at elevated temperatures (up to 1500 "C), instruments using oscillating cylin-
ders, with the fluid inside or outside, have been favored for operational reasons. The
work of Oye and his grouplo is an excellent example of what can be achieved.
Among the great advantages of oscillating viscometers, the fact that no bulk motion
of the fluid is required and that the measurements can be made absolute are para-
mount. Although a knowledge of the density of a fluid is necessary to evaluate the
viscosity from the measurements made, it need not be known with the same accu-
racy required of the viscosity.
126 Chapter 11
Capillary viscometers measure the time of efflux of a known volume of fluid
through a circular section tube. They intrinsically determine the kinematic viscosity,
(dynamic viscosity/density), so that evaluation of the dynamic viscosity requires a
knowledge of the density with a comparably high accuracy. Such viscometers are
most often employed for liquids at ambient conditions when a hydrostatic head pro-
vides the driving force, or for gases allowing the decay of a generated pressure dif-
ference. At elevated temperatures, the use of a fluid pump complicates the
experimental installation. Although the theory of such viscometers is superficially
simple, the details of some of the applicable corrections have only recently been
resolved.6 Furthermore, owing to the difficulty of knowing the dimensions of the
capillary tube and uniformity of bore with a very high precision, measurements are
mostly performed on a relative basis. We note the excellent work of Smith and his
collaborators9 in the dilute gas phase from 90 to 1500 K, as well as the work of
Nagashima and his co-workers on water.l0
In the case of high pressures, different types of viscometer have been employed
owing to the need to reduce the volume of fluid required. The most popular have
been falling-body viscometers and torsional-crystal viscometers.8 Neither of these
have, however, completely developed theories so that their accuracy is intrinsically
limited.8 On the other hand, the newly developed vibrating-wire vi~cometer'~~ l4 that
makes use of the damping of a transverse oscillation of a thin wire enjoys a complete
theory.
The application of optical techniques such as that involving the study of the fre-
quency and decay of surface waves on fluids (ripplons)15has great potential for the
future, especially since they require no contact with the fluid, but they are at an early
stage of development now.
It is possible to achieve an uncertainty in the measurement of the viscosity of a
fluid of a few parts in a thousand under near-ambient conditions, which deteriorates
to a few percent at extremely low and high temperatures for low densities, In very
high-pressure gases and liquids, the uncertainty achieved is at best a few percent and
frequently very much worse.
2.2 Thermal Conductivity

In the case of the thermal-conductivity, there are three main techniques: those based
on Equation (1) and those based on a transient application of it. Prior to about 1975,
two forms of steady-state technique dominated the field: parallel-plate devices, in
which the temperature difference between two parallel disks either side of a fluid is
measured when heat is generated in one plate, and concentric cylinder devices that
apply the same technique in an obviously different geometry. In both cases, early
work ignored the effects of convection. In more recent work, exemplified by the
careful work in Amsterdam with parallel plates,16and in Paris with concentric cylin-
' effects of convection have been investigated. Indeed, the parallel-plate
d e r ~ , ~the
cells employed in Amsterdam by van den Berg and his co-workers16have the unique
feature that, because the temperature difference imposed can be very small and the
horizontal fluid layer very thin, it is possible to approach the critical point in a fluid
or fluid mixture very closely (mK).
Only one transient technique has enjoyed success and that is the transient hot-wire
method, in a form pioneered by H a m a n s and subsequently developed for a wide
range of applications in gases and liquids for temperatures from 70 to 500 K and
pressures up to 700 MPa. The essential features of this technique are that it meas-
ures the transient temperature increase of a thin metallic wire (a few microns)
immersed in the test fluid following the initiation of electrical heating within it. Its
principal advantages are that the temperature increases need last for no more than
one second so that the inertia of the fluid inhibits the development of significant con-
vective heat transfer. In addition, the small magnitude of the temperature difference
applied (2 K), but the large temperature gradient (about lo6Krr-'), means that radia-
tive heat transfer is also generally insignificant.8Although the technique is unsuit-
able for work close to the critical point, it has been successfully employed over a
very wide range of conditions for different types of fluids. Generally, an uncertainty
of a few parts in a thousand in the thermal conductivity around ambient conditions
is possible for simple fluids, but this is degraded to several percent at extremes of
temperature, although relatively unaltered by pressure.
The technique has recently been applied to polar or electrically conducting fluids
with considerable success, while a sensor made of a platinum wire embedded
between two thin layers of alumina has recently been employed to measure the ther-
mal conductivity of molten metals.18 Finally, we note that a variety of new tech-
niques for measuring the thermal diffusivity of fluids have been introduced relying
largely on light scattering.8These methods have distinct advantages in special fluids
and regions of thermodynamic state, particularly close to the critical state. While
they are not yet able to achieve the same level of accuracy for simple fluids as can
be achieved by conventional techniques, the fact that they are contact free does give
them great potential for the future.
2.3 Diffusion Coefficient

The measurement of diffusion coefficients in either gases or liquids is a very slow
process and many techniques require days to attain a single result. This is largely
because of the intrinsic slowness of the process and the fact that most methods use
equipment of large scale. For this reason alone, there are relatively few experimen-
tal results and the measurements do not extend over wide ranges of temperatures. On
the other hand, there are many different techniques for the measurement of diffusion
coefficients from optical interferometric methods to NMR measurements of spin
relaxation and chromatographic flow-broadening.8 The range is too wide to treat
here but it is worthwhile noting that the most accurate interferometric techniques
yield diffusion coefficients with an uncertainty of about 0.1%, but only near-ambi-
ent conditions. Under different conditions, other techniques are usually employed
and the uncertainty is then typically a few percent.8
3 Theoretical Predictions
As noted above, the fluids of industrial concern only occasionally conform to those
for which experiments have been performed. It is therefore both apparent and
128 Chapter 11
inevitable that the burden placed on prediction and estimation of transport property
values is considerable. Hence, the need for theoretical input, in combination with a
limited set of accurate data to enable the transport-properties estimation, is unavoid-
able. The extent to which rigorous theory is available depends greatly on the ther-
modynamic state and molecular complexity as the following material indicates.
In the low-density gas, the mean-free path of the molecules is very much greater
than the molecular diameter so that free molecular motion contributes most to the
transport process and molecular collisions are relatively rare events involving only
two molecules at any one time. Such molecular collisions modify the transport by
deflecting molecules from their original course. Thus, the nature of the collision,
which is determined by the forces exerted between a pair of molecules, necessarily
contributes to the magnitude of the flux of a dynamical quantity brought about by a
particular gradient. As the density is increased, the free motion of the molecules is
progressively dominated by the transport process involving interactions among the
molecules and especially groups of them. The mean-free path becomes smaller and
the details of the interactions between the molecules therefore become less impor-
tant compared with the fact that so many interactions take place. Thus, when the liq-
uid state is attained, it seems that quite simple models of the interaction between
molecules are adequate for a description of transport properties. In the extreme case
of a fluid near its critical point, the specific intermolecular interactions become rel-
atively unimportant since the properties are determined by the behavior of clusters
of molecules and the size of these clusters rather than anything else.
At present, a rather well-developed rigorous theory exists for transport properties
in the low-density limit, which has been exploited to probe intermolecular potentials
for spherically symmetric atoms and simple polyatomic molecules. As the density of
the fluid is increased, the rigor of the theoretical description of transport properties
decreases for a variety of reasons but rises again close to the critical point and then
decays rapidly for liquid-like densities. In the moderately dense gas, the description
of transport properties is based on the Enskog theory, which first deals with gases
composed of rigid sphere molecules and assumes that the sole effect of increasing
density is an increase in the collision frequency. It also allows for the transfer of
momentum or energy across a plane without center of gravity motion of the mole-
cules. This latter phenomenon, known as collision transfer, eventually becomes the
dominant mode of transport for liquids and solids. For moderately dense gases
where this collisional effect is modest, the Enskog theory turns out to provide a reli-
able basis for means of estimating transport properties even though it is unable to
make predictions from first principles. In particular, for mixtures the theory has pro-
vided a remarkably effective means of predicting the properties of gaseous systems
at high density given information only on the pure fluids under similar conditions.
In the liquid phase also, the Enskog theory has also proved quite effective for the
representation and prediction of the properties of pure and mixed substances. In this
case, the theory has been modified using the results of computer simulations of hard
spheres, which have indicated the limitations of the assumptions of entirely random
motions at elevated densities. In combination, the Enskog theory corrected in this
way provides a good description of the properties of some pure liquids. An even
more general result of the Enskog theory is that the transport properties of a fluid or
fluid mixture depend only on the ratio of the molar volume of the fluid to a molec-
ular volume. For hard sphere molecules, this result is exact but for real molecules it
now seems that it is also valid as long as a weak temperature dependence of the
molecular volume is allowed. These observations have been made the basis of a
scheme for the representation and prediction of a wide range of fluids over a large
range of temperature and pres~ure.'~
From this analysis, it should be clear that for the dense fluids theory provides only
a framework for description and prediction rather than an absolute basis and that this
basis is most secure for spherical molecules. It follows that in the absence of experi-
mental measurements and when the molecules deviate increasingly from a rigid
spherical nature, the predictions of properties become less secure. It therefore remains
rather difficult to make predictions or estimations of the properties of complex mole-
cules for which, as was noted earlier, there are currently few measurements.
4 International Dimension
The transport properties of fluids are an integral part of the design of process plant.
For this reason, in commercial transactions across national borders involving process
plant, the design data for transport properties among others are a part of the valida-
tion of the design. As a consequence, the validation of transport property data inter-
nationally has been of considerable significance. It was to provide such validation
for transport properties that the International Union of Pure and Applied Chemistry
(IUPAC) set up a Transport Properties Subcommittee under its former Commission
on Thermodynamics. A number of representations of the transport properties of
common fluids have been published by that subcommittee and its successor the
International Association for Transport Properties.20Examples of this trade in design
and the data associated with design include conventionally fuelled power plants
where the properties of water are vital and the properties of carbon dioxide that fea-
ture in the safety calculations of certain kinds of gas-cooled nuclear reactors.21
A final point that also needs to be presented is related to quality assurance, i.e. the
requirement to satisfy regulatory frameworks concerned with safety and environ-
mental issues. Traditionally, only national regulatory bodies had to be satisfied. Now,
however, international regulatory bodies have to be satisfied while national regula-
tory bodies must, in many cases, accept the regulations of other nations. In this con-
text, quality assurance of a plant or a process may often require a demonstrable
pedigree for each number in the calculation of some aspect of the plant.
5 Databases
Individual companies can no longer afford to collect and maintain all the data for
process modeling that they may need to carry through an increasingly diverse busi-
ness. Hence, they rely increasingly on third-party software and databases. Thus,
fluid property databases are becoming increasingly important, and with them the
concept of standardized exchange protocols has evolved, so that data can easily flow
from a number of different databases to a wide range of applications. For chemical
130 Chapter 11
species, such databases often combine the physical properties of the materials
(including transport properties) with information on chemical stability, toxicity and
flammability. For many of the reasons indicated earlier, the merits of a database do
not lie simply in the wide range of compounds that it treats nor the ranges of ther-
modynamic state covered but also in the quality and pedigree of the data contained
in it. The effort required to amass large quantities of validated data is now a major
industry in its own right and the number of organizations engaged in it are rather few
internationally. The American Institute of Chemical Engineers’ main data effort is
the Design Institute for Physical Properties (DIPPR) program. DIPPR is funded by
industrial sponsors and its best-known product is the database from Project 801,
which contain evaluated and correlated pure-component data on specific key prop-
erties for over 1700 pure compounds. Another major source of evaluated fluid prop-
erty data is the Standard Reference Data program of the National Institute of
Standards and Technology (NIST). The Thermodynamics Research Center (TRC)
(recently returned to NIST from Texas A&M University4) issues a variety of data
products and maintains an enormous database, with information on approximately
100 properties for 16.000 pure components and numerous mixtures. A very impor-
tant thermophysical properties database, DETHERM, has also been developed by
DECHEMA in Germany. DETHERM incorporates 4.2 million data sets for more
then 122.000pure fluids and mixtures. Finally, the National Engineering Laboratory
in United Kingdom maintains the Physical Properties Data Service (PPDS). There
can be little doubt that in the near future, access to GRID computing technologies
can place these large databanks instantly on the desktop of every scientidengineer
in academia and industry worldwide. Then, the premium to be attached to quality of
the data and its certification then becomes ever more pressing as the data user and
its source become increasingly disconnected.
6 Future
Efficient design, energy savings and stricter environmental controls will produce a
need for transport property data that are more accurate, internationally accepted and
cover a wide range of conditions. The number of fluids for which such international
accord is necessary is likely to be relatively small and determined by international
agreement. For other materials, a small number of accurate measurements will be
made in order to provide the basis for the development of prediction schemes of
increasing sophistication. There is little doubt that in these prediction- schemes,
direct-computer simulation of fluids will play an increasing role relative to analyti-
cal theory, but it is not likely that simulation can obviate completely the need for
measurement in the next decade. It is therefore likely that the role of the experi-
mentalist will need to shift from the study of simple molecules under convenient
conditions to the study of unusual and complex materials under difficult conditions
such as molten metals at high temperatures, molten polymers. Indeed, it is quite pos-
sible that the process of measurement will be integrated within devices or process
plant so that in-situ measurements can be performed. This is most likely to be the
case for devices constructed on the micro- or nano-scale where existing theories of
fluid properties may be inadequate owing to the special conditions that pertain.
References
1. R. B. Bird, W. E. Stewart and E. N. Lightfoot Transport Phenomena, John Wiley & Sons
2. D. Q. Kern, Process Heat Transfer, McGraw-Hill International, Tokyo, 1950.
3. M. J. Assael, C. A. Nieto de Castro and W. A. Wakeham, Proceedings of CHEMPOR '78,
1978, 16.1-16.9.
4. A. H. Harvey and A. Laesecke, Chem. Engin. Prog., 2002, February, 34.
5. M. J. Assael, C. F. Chen., I. Metaxa and W. A. Wakeham, Proceedings of 15th Symposium
on Thermophysical Properties, Boulder, Colorado, 2003.
6. J. Millat, J. H. Dymond and C. A. Nieto de Castro (eds), Transport Properties of Fluids.
Their Correlation, Prediction and Determination, Cambridge University Press, London,
1995.
7. K. F. Jensen, Chem. Eng. Sci., 2001, 56,293.
8. W. A. Wakeham, A. Nagashima, and J. V. Sengers (eds), Experimental Thermodynamics
Vol. III: Measurement of the Transport Properties of Fluids, Blackwell Scientific, Oxford,
1991.
9. E. Vogel, Wiss. Zeit. Rostock, 1972,21, 169.
10. H. A. Oye and K. Torklep, J. Phys. E: Sci. Instrum., 1979, 12, 875.
11. A. G. Clarke and E. B. Smith, J. Chem. Phys., 1968,48,3988.
12. K. Kobayashi and A. Nagashima, High Temp.-High Press., 1985, 17, 131.
13. M. J. Assael, M. Papadaki, S. M. Richardson, C. Oliveira and W. A. Wakeham, High
Temp. - High Press, 1991, 23, 561.
14. M. J. Assael and W. A. Wakeham, Fluid Phase Equilibr:, 1992,75,269.
15. Y. Nagasaka, Proceedings of 16th European Conference on Thermophysical Properties,
London, 1-6 September, 2002.
16. R. Mostert, H. R. van den Berg and P, S. van der Gulik, Rev. Sci. Instrum., 1989,60,3466.
17. R. Tufeu, Ph.D. Thesis, Paris University, 1971.
18. M. Dix, I. W. Drummond, M. Lesemann, V. Peralta-Martinez, W. A. Wakeham, M. J.
Assael, L. Karagiannidis and H. R. van den Berg, Proceedings of 5th Asian
Thermophysical Properties Conference, Seoul, 1998, 133-1 36.
19. M. J. Assael, J. H. Dymond, M. Papadaki and P. M. Patterson, Int. J. Thermophys., 1992,
13, 269.
20. S. Hendl, J. Millat, E. Vogel, V. Vesovic, W. A. Wakeham, J. Luettmer-Strathmann,J. V.
Sengers and M. J. Assael, Int. J. Thermophys., 1992,15, 1.
21. M. J. Assael, The importance of thermophysical properties in optimum design and energy
saving, in Energy and Environment: Technological Challengesfor the Future, Springer-
Verlag, Tokyo, 2000.
CHAPTER 12
Micro- and Nano-particles

Production Using Supercritical
Fluids
ERNEST0 REVERCHON AND IOLANDA DE MARC0
1 Introduction
Micronization processes based on the use of supercritical fluids have been suggested
during the last few years as alternatives to traditional techniques.' Indeed, one of the
most intriguing challenges in the development of supercritical fluid (SCF)-based
applications is the micronization of solid compounds that can be precipitated at mild
temperatures and with reduced or no solvent residue. Moreover, SCF-based tech-
niques guarantee the control of particle size (PS) and distribution (PSD) in the
micrometric and nanometric range. These expectations are based on the peculiar
characteristics of gases at supercritical conditions: very fast mass transfer and the
fast and complete elimination of the SCF at the end of the micronization process.
The first supercritical fluid-based micronization process has been the rapid expan-
sion of supercritical solutions (RESS); it is based on the solubilization of the solid
to be micronized in the SCF and its subsequent precipitation by fast depressurization
of the solution. However, the use of this technique is largely limited by the low sol-
ubility in supercritical carbon dioxide (SC-CO,) of many of the solids of
Particles generation from gas saturated solutions (PGSS) has also been proposed.
In this case the product to be micronized is liquefied by heating and addition of SC-
CO,; then, the gas-liquid solution is sprayed in a low-pressure vessel, thus obtain-
ing microparticles. This technique has been successfully used to process some
polymers, but has limited applicability in the case of thennolabile compounds that
can decompose during heating.9J0
The supercritical antisolvent (SAS) precipitation has also been proposed in vari-
ous arrangements and under various acronyms.ll This technique is based on the use
of SC-CO, as the antisolvent to precipitate the solid solute from a liquid solution.
The prerequisites for the success of this technique are the complete solubility of the
Micro- and Nano-particles Production Using Supercritical Fluids 133
liquid in the SCF and the complete insolubility of the solid in it. In the batch mode
(often indicated by the acronym GAS), the precipitation vessel is loaded with a given
quantity of the liquid solution and, then, the supercritical antisolvent is added until
the final pressure is reached. In the semi-continuous mode (often indicated by the
acronym SAS), the liquid solution and the supercritical antisolvent are continuously
delivered to the precipitation vessel in co-current or counter-current mode.
Very recently, atomization processes assisted by supercritical CO, have been pro-
posed by various authors. They are based on the mixing and/or solubilization of SC-
CO, in a liquid solution and the subsequent atomization of the mixture.12-16
At present, the most promising process seems to be the supercritical antisolvent pre-
cipitation, which is the most widely a ~ p l i e d ' ~and
- ~has
~ been recently proposed on a
semi-industrial scale.30.31 However, despite the fact that many works have been pub-
lished on SAS precipitation, only a limited number of them have focused on the mech-
anisms controlling particle formation and on the role of the process parameters on the
morphology and on the dimensions of the precipitated powders. This lack of informa-
tion can be one of the main factors hampering the industrial application of this process.
Therefore, the scope of this chapter is to describe the results obtained using SAS
with particular emphasis on the production of particles with controlled PS and PSD
in the micrometric as well as in the nanometric range. We will try to understand the
role of high pressure vapor-liquid equilibria (VLEs) in determining the morphology
and particle size of precipitates.
2 Experimental Apparatus for SAS Processing

A semi-continuous SAS apparatus, as a rule, consists of two pumps used to deliver
the liquid solution and supercritical CO,, respectively. A cylindrical vessel is used as
the precipitation chamber. The liquid mixture is delivered to the precipitator through
an injector. Different arrangements have been proposed in the literature like nozzles,
capillaries, vibrating orifices and coaxial devices. Supercritical CO, is delivered
through another inlet port that can be located on the top or the bottom of the vessel.
The pressure in the precipitator is regulated by a micrometering valve located at the
exit of the chamber. A stainless-steel frit located at the bottom of the precipitator is
used to collect the produced powder. A second vessel located downstream the
micrometering valve is used to recover the liquid solvent.
Part of the experiments described in this chapter have been performed using a
SAS transparent precipitator that has two quarts windows placed along the entire
height of the vessel. This apparatus allows the visual observation of the macroscopic
features of the precipitation process.
2.1 Experimental Procedure

A SAS experiment begins by delivering supercritical CO, to the precipitator until the
desired pressure is reached. An antisolvent steady flow is then established. Then,
pure solvent is sent through the nozzle with the aim of obtaining steady-state com-
position conditions during the solute precipitation. At this point, the flow of the liq-
uid solvent is stopped and the liquid solution is delivered through the nozzle. The
134 Chapter 12
experiment ends when the delivery of the liquid solution to the chamber is inter-
rupted. However, supercritical CO, continues to flow to wash the chamber for the
residual content of liquid solubilized in the supercritical antisolvent. If the final
purge with pure CO, is not performed, solvent condenses during the depressuriza-
tion and can solubilize the collected powder.
3 SAS Fluidodynamics, Mass Transfer and

Thermodynamics
Hydrodynamic requirements of SAS are related to jet break-up; i.e., the conditions
under which small droplets are obtained to form a large surface between the liquid
and supercritical fluid that favors mass transfer. The jet break-up is critical and may
be problematic in the case of highly viscous solutions like polymer solutions.
Mass transfer is also favored by the high diffusivity that characterizes supercriti-
cal fluids. The difference in density between the liquid and the antisolvent also plays
a role in determining the direction of mass transfer. However, the SAS process does
not seem to be substantially limited by mass transfer resistance^.^^
Thermodynamic constraints to the SAS process can be summarized in the required
miscibility between the liquid solvent and the supercritical antisolvent and the insol-
ubility of the solute in the antisolvent and in the solvent-antisolvent mixture. Data are
available for various binary mixtures liquid-supercritical fluid 32 and can be described
as type I using the classification of van-Konynenburg and If jet break-up is
obtained and mass transfer is very fast, high-pressure VLEs of the ternary system liq-
uid +solute+ supercritical fluid can control the precipitation process.
Based on the previous considerations, some authors proposed thermodynamic-
based approaches to SAS. De la Fuente Badilla et aZ.34attempted to develop a ther-
modynamic-based criterion for optimum batch antisolvent precipitation (GAS)
using a definition of the volume expansion that takes into account the molar volume
of the system studied. They analyzed various binary and ternary systems and con-
cluded that the pressure corresponding to a minimum value of the liquid-phase vol-
ume expansion coincides with the pressure at which the solute precipitates. In a
subsequent work,35 Shariati and Peters further highlighted the role of SC-CO, in
GAS. It acts as a co-solvent (cosolvency effect) at lower concentrations, whereas at
higher concentrations it acts as an antisolvent.
M ~ k h o p a d h y a yproposed
~~ the relative partial volume reduction of the solvent as
the key parameter for the selection of the solvent and of the optimum GAS process
conditions. The solute solubility is proportional to the partial molar volume of the
solvent due to CO, molecules clustering around solvent molecules, which lead to the
loss of solvent power.
Both the previous approaches apply to discontinuous GAS process (in which the
pressure progressively increases) and indicate precipitation process pressures
between 50 and 70 bar, i.e., under subcritical conditions (even when the critical pres-
sure of pure CO, is considered), whereas the experimental evidences for semi-con-
tinuous SAS (steady-state pressure) indicate larger pressure values for the
precipitation of micro- and nano-particles.
Besides these thermodynamic criteria, the most common approach used in the liter-
ature is based on the operation at pressures above the binary (liquid - SC-CO,) mix-
ture critical point, completely neglecting the influence of solute on VLEs of the
system. But, the solubility behavior of a binary supercritical C0,-containing system is
frequently changed by the addition of a low volatile third component as the solute to
be precipitated. In particular, the so-called cosolvency effect can occur when a mixture
of two components solvent+solute is better soluble in a supercritical solvent than each
of the pure components alone. In contrast to this behavior, a ternary system can show
poorer solubility compared with the binary systems antisolvent+solvent and antisol-
vent +solute; a system with these characteristics is called a non-cosolvency (antisol-
vent) s y ~ t e m .In
~ ~particular,
!~~ in the case of the SAS process, they hypothesize that
the solute does not induce cosolvency effects, because the scope of this process lies in
the use of CO, as an antisolvent for the solute, inducing its precipitation.
In the light of these considerations, a different approach based on ternary system
thermodynamics could be considered. However, the phase behavior of ternary sys-
tems could be very complex and there is a considerable lack of data on ternary sys-
tems containing a component of low volatility; therefore, a possible compromise
could be to consider that the solute addition can produce the shift of the mixture crit-
ical point (MCP) (i.e., the pressure at which the ternary mixture is supercritical) with
respect to binary system VLEs and the modification of this kind of system that is
formed according to the van-Konynenburg and Scott c l a s s i f i ~ a t i o n . ~ ~ . ~ ~ * ~ ~
Therefore, the approach followed in this chapter considers pseudo-binary dia-
grams, i.e., equilibria involving the third component are, however, neglected, but
modifications due to the presence of the solute are considered on the binary system.
We will observe in the analysis of the experimental results that this approach can
provide interesting information regarding the evolution of the SAS process, and the
morphology and dimension of the precipitated particles. A rationalization of the
experimental results is also proposed.
4 Relationship between Particles Morphology and

High-pressure VLEs
In this part of the chapter, a summary of semi-continuous SAS experiments per-
formed on various materials is proposed and an attempt at providing an interpreta-
tion of the different morphologies observed in relation to the position of the process
operating point with respect to mixture critical point has been made.
4.1 Experimental Evidences

Case a : In some cases, the solute was recovered in the liquid collection vessel, i.e.,
it was completely extracted by the solution formed by the solvent and SC-CO,. An
example is represented by Medroxiprogesterone acetate (corticosteroid) processed
in acetone at 120 bar, 40 "C.
The visual observation across the windowed vessel showed that a single fluid
phase occupied the entire precipitator and no precipitation was observed. Therefore,
136 Chapter 12
the solute was extracted due to an enhancement of its solubility in the solvent-anti-
solvent mixture with respect to its solubility in the SCF alone (the cosolvency effect
described previously).
Case b: In various cases, we observed the formation of large crystals at the bot-
tom of the windowed precipitator. For example, we found large crystals of
Prednisolone (corticosteroid) precipitated from Methanol (Figure 1 ) and of
Prednisolone acetate (corticosteroid) precipitated from Tetrahydrofuran, in both
cases operating at 150 bar, 40 "C.We also observed that a relative control of the crys-
tal size down to some microns is also possible, in addition to decreasing the con-
centration in the liquid solution or increasing the ratio between the antisolvent and
the liquid solution flow rates.
In this case, the visual observation showed that two phases were formed in the pre-
cipitator. Crystals precipitated from a relatively small quantity of a liquid-rich phase
located at the bottom of the vessel, whereas no precipitation was observed from the
upper fluid phase (supercritical) that was separated from the lower one by a meniscus.
Case c: The powder was formed by crystals and amorphous particles. These mor-
phologies were observed, for example, in the case of Inulin (biopolymer) precipita-
tion from Dimethylsulfoxide (DMSO) at 150 bar, 40 "C, 20 mg/ml and Amoxicillin
(antibiotic) from DMSO under the same process conditions (Figure 2).41
The visual observation of the precipitation process showed that two phases formed
in the precipitator. The precipitation took place from both phases; the upper one
(supercritical) produced amorphous particles, whereas the lower one (liquid-rich)
produced crystals.
Case d Amorphous balloons, ix., hollow amorphous spherical particles several
microns in diameter, were observed in some cases. They are characterized by various
expansion degrees. We observed that the balloons diameter increases with
CO, density. Examples of this morphology were observed during the study of the pre-
cipitation of Yttrium acetate (YAc) (superconductor precursors) from DMSO at 120 bar
and 40 "C (Figure 3) 31 and of Cefonicid (antibiotic) from DMSO at 150 bar and 60 "C.40
Figure 1 Large needle-like crystals of Prednisolone precipitated from Methanol at 1.50 bal;
40 "C,20 mg/mL
Figure 2 Amoxicillin precipitated from DMSO at 150 bal; 40 "C,20 mg/mL: ( a ) amorphous
microparticles in the upper part of the precipitator; and (b) large crystals in the
lower part of the precipitator
Visual inspection of the precipitation process showed that a single fluid phase was
present in the precipitator.
Cuse e: Nanoparticles formation was observed, for example, in the case of Yttrium,
Samarium, N e ~ d i m i u mand~ ~Gadolinium acetates21(superconductor precursors) and
in the case of D e ~ t r a n(polymer),
~~ Tetra~ycline~~
(antibiotic) and other compounds.
This morphology was obtained at CO, densities larger than those characteristic of
case d. Nanoparticles down to a mean diameter of 50 nm were obtained in the case of
Zinc acetate20 (catalyst precursor). Microparticles as well as nanoparticles were
obtained for Amoxicillin, Rifampicin (antibiotic) (Figure 4) and other compounds.
Visual inspection of the process confirmed that also in this case, the precipitation
took place from a single fluid phase.
138 Chapter I 2
Figure 3 Yttrium Acetate amorphous microparticles precipitated from DMSO at 120 bar;
40 "C,50 mg/mL
Figure 4 Rifampicin amorphous microparticles precipitated from DMSO at 90 bar; 40 "C,

40 mg/mL
The various cases discussed all refer to pressure and temperature conditions that
are above the MCP of the binary mixture C0,-solvent; nevertheless, the presence of
solutes has evidently modified the vapor-liquid behavior with respect to the one char-
acteristic of the binary system at least in cases a-c: cosolvency and formation of two
phases have been observed. Thus, visual observation and powder analysis confirmed
our hypothesis of a shift of the MCP induced by the addition of the thirdcomponent.
We can also suppose that the formation of crystals or of amorphous particles is
controlled by the content of SCF in the liquid (and vice versa), i.e., a liquid con-
taining small quantities of SCF has a liquid-like behavior and crystals are obtained;
as the SCF content increases (SCF rich phase), we can observe the formation of
amorphous particles.
Xc0*
Figure 5 Pseudo-binary diagram and the possible operating points for SAS
A pseudo-binary diagram is shown in Figure 5 and can be used to locate the var-
ious cases discussed with respect to VLEs.
Case a can be explained in a trivial mode by the enhancement of solute solubility
in SC-CO, because of the presence of the solvent and does not need to be reported
in the diagram. Cases b and c, in our hypothesis, can be attributed to experiments
performed at pressures lower than the MCP of the ternary system. A different solute
partitioning between the two phases and their relative content in SCF can explain the
formation of crystals and/or amorphous particles. In particular, point B in Figure 5
represents an operating point producing a large liquid-rich phase, whereas point C
represents an operating point with a large gas-rich phase.
We think that case d is relative at operating conditions below the ternary MCP, but
located in the single gas phase region, in which mass transfer between the liquid
droplet and the surrounding supercritical fluid is dominated by diffusion of CO,
inside the droplet with the consequent formation of balloons of increasing dimen-
sions (point D in Figure 5). This explanation is also supported by the modelling work
performed by Werling and Debenedetti,44~45 which indicated the relative density of
the two phases as the controlling mechanism for the expansion of the droplets. Their
work confirms that droplets expansion prevails when the process is performed under
sub-critical conditions with respect to the mixture studied.
Case e is the result of precipitation under conditions above the ternary MCP and
is the true supercritical antisolvent precipitation. The formation of very small parti-
cles indicates that characteristic times for surface tension disappearance are very
short, droplets are not formed at the exit of the injector and the very small particles
are released from the fluid phase.27These aspects have been studied in the previously
cited works of Werling and Debenedetti,44t45which underline the interplay among
the different consecutive or simultaneous process that characterize SAS, mainly
mass transfer under sub-critical and supercritical conditions. The characteristic times
140 Chapter 12
of the various mechanism involved, including nucleation, have been postulated in a
subsequent work.46
5 System Modifications Induced by the Presence

of Solute
Until now, an attempt at explaining the observed modifications of VLEs induced by
the presence of the solute has been made. Some compounds do not interfere with the
binary VLEs; therefore, the one-phase region is the same as in the binary system (at
temperatures and pressures higher than MCP). But, we have seen that, in general, the
Figure 6 Cefonicid precipitated from DMSO at 120 bal; 50 "C, 10 mg/mL: (a) amorphous
microparticles in the upper part of the precipitator; and (b) amorphous balloons in
the lower part of the precipitator
presence of the solute modifies the binary VLEs even at low solute concentrations.
The simplest modification that can be presented by ternary non-cosolvency systems,
with respect to the corresponding binary system, consists of the shift of the MCP
towards higher pressures. This effect could also be a function of the concentration of
solute in the liquid solution, i.e., the higher the concentration, the larger the pressure
increase of t k MCP.
However, the shift of the MCP towards higher pressures is only the simplest pos-
sible modification of the VLEs. More complex interactions among the three compo-
nents may occur. For example, in the case of the Cefonicid precipitated from DMSO
at 150 bar and 60 "C, we observed the presence of two different morphologies: in
the upper part of the precipitator the powder was formed by sub-microparticles
(Figure 6(a), whereas in the lower part it was formed by microparticles and balloon-
like particles (expanded droplets) (Figure 6(b)). Both morphologies have been pre-
viously attributed to fluid phase precipitation.
In this case, the question is: how is it possible to obtain two gas-like phases? When
a two-phase equilibrium is obtained for type I systems, it is always characterized by
a liquid- and gas-rich phase connected by a tie line. Indeed, these systems are char-
acterized by an uninterrupted critical line that connects the critical points of the pure
components .47
If we consider a type V system, instead, as the temperature increases, an azeotrope
appears. The binary system DMSO-CO, is of type I and the presence of Cefonicid
at 40 "C does not modify this characteristic. But when the temperature is increased
at 60 OC, a type V situation is probably produced. It means that an azeotrope is
formed; the critical line is not continuous, but is formed by two lines, which start
Figure 7 High-pressure vapor-liquid equilibrium diagram at 60 "C f o r the system

DMSO-CO, in the presence of Cefonicid
142 Chapter 12
from the critical point of the component A (supercritical fluid) and from the critical
point of the component B (liquid solvent), respectively. Therefore, the precipitation
occurs from two points in equilibrium corresponding to two C0,-rich phases (gas-
like behavior) producing amorphous particles, but with different morphologies
(Figure 7).
6 Conclusions
In this chapter, we have highlighted that SAS performance is conditioned by ther-
modynamic constraints and it is possible to explain SAS products morphology
through a thermodynamic-based approach.
At this point, the question is: is the picture of SAS conditioning thermodynamic
behaviors complete? Probably not. Indeed, very complex interactions are possible in
ternary systems. The approach proposed is encouraging from another point of view.
Even if a very complex behavior develops in the ternary systems, a systematic analy-
sis of VLEs can allow the identification of the regions in which effective precipita-
tion can be obtained.
References
1. J. Jung and M. Perrut, J. Supercrit. Fluids, 2001, 20(3), 179.
2. D. W. Matson, J. L. Fulton, R. C. Petersen and R. D. Smith, Znd. Eng. Chem. Res., 1987,
26(1l), 2298.
3. R. C. Petersen, D. W. Matson and R. D. Smith, Polym. Eng. Sci., 1987, 27, 1693.
4. R. S. Mohamed, P. G. Debenedetti and R. K. Prud’homme, A.Z.Ch.E. J., 1989, 35, 325.
5. A. K. Lele and A. D. Shine, A.Z.Ch.E. J., 1992,38(5), 742.
6. E. M. Phillips and V. J. Stella, Znt. J. Pharm., 1993, 94, 1.
7. E. Reverchon, G. Donsi and D. Gorgoglione, J, Supercrit. Fluids, 1994,6, 241.
8. E. Reverchon and P. Pallado, J. Supercrit. Fluids, 1996, 9, 216.
9. E. Weidner, Z. Knez and Z. Novak, European Patent EP 0 744 992, February 1995
PatentWO 95/21688, July 1995.
10. C. Xu, B. Watkins, R. E. Sievers, X. Jing, P. Trowga, C. S. Gibbons and A. Vecht, Appl.
Phys. Lett., 1997,71(12), 1643.
11. E. Reverchon, J. Supercrit. Fluids, 1999, 15(l), 1.
12. E. Reverchon, Znd. Eng. Chem. Res., 2002, 41(10), 2405.
13. E. Reverchon and G. Della Porta, J. Supercrit. Fluids, 2003, 26(3), 243.
14. E. Reverchon and G. Della Porta, Znt. J. Pharm., 2003,258(1-2), 1.
15. R. E. Sievers, U. Karst, P. D. Milewski, S. P. Sellers, B. A. Miles, J. D. Schaefer, C. R.
Stoldt and C. Y. Xu, Aerosol Sci. Tech., 1999,30, 3.
16. S. P. Sellers, G. S. Clark, R. E. Sievers and J. F. Carpenter, J. Pharm. Sci., 2001,90, 785.
17. S. D. Yeo, G. B. Lim, P. G. Debenedetti and H. Bernstein, Biotech. Bioeng., 1993,41,341.
18. W. J. Schimtt, M. C. Salada, G. G. Shook and S. M. Speaker, A.1.Ch.E. J., 1995,41(11),
2476.
19. E. Reverchon, G. Della Porta, S. Pace and A. Di Trolio, Ind. Eng. Chem. Res., 1998,37(3),
952.
20. E. Reverchon, G. Della Porta, D. Sannino and P. Ciambelli, Powder Technol., 1999,
102(2), 127.
2 1. E. Reverchon, I. De Marco and G. Della Porta, J. Supercrit. Fluids, 2002, 23( l), 8 1.
22. G. J. Hutchings, J. A. Lopes-Sanchez, J. K. Bartley, J. M. Webster, A. Burrows, C. J.
Kiely, A. F, Carley, C. Rhodes, M. Havecker, A. hop-Gericke, R. W. Mayer, R. Schlogl,
J. C. Volta and M. Poliakoff, J. Catalysis, 2002,208(1), 197.
23. Y. Gao, T. K. Mulenda, Y. F. Shi and W. K. Yuan, J. Supercrit. Fluids, 1998, 13(1-3), 369.
24. P. M. Gallagher, M. P. Coffey, V. J. Krukonis and W. W. Hillstrom, J. Supercrit. Fluids,
1992, 5, 130.
25. R. Bodmeier, H. Wang, D. J. Dixon, S. Mawson and K. P. Johnston, Pharm. Res., 1995,
12(8), 1211.
26. D. J. Dixon, K. P. Johnston and R. A. Bodmeier, A.Z.Ch.E.J., 1993,39(1), 127.
27. T. W. Randolph, A. D. Randolph, M. Mebes and S. Yeung, Biotech. Progress, 1993,9(4),
429.
28. L. Benedetti, A. Bertucco and P. Pallado, Biotech. Bioeng., 1997, 53(2), 232.
29. S . Mawson, S . Kanakia and K. P. Johnston, J. Appl. Polym. Sci., 1997,64(1l), 2105.
30. E. Reverchon, I. De Marco, G. Caputo and G. Della Porta, J. Supercrit. Fluids, 2003,
26(1), 1.
31. E. Reverchon, G. Caputo and I. De Marco, Ind. Eng. Chem. Res., 2003,42, 6406.
32. S . Ohe, Vapor-liquid Equilibrium Data at High Pressure, Elsevier, Tokyo, 1990.
33. P. H. van Konynenburg and R. L. Scott, Philos. Trans. R. Soc., London 298 A, 1980,495.
34. J. C. de la Fuente Badilla, C. J. Peters and J. de Swaan Arons, J. Supercrit. Fluids, 2000,
17(1), 13.
35. A. Shariati and C. J. Peters, J. Supercrit. Fluids, 2002,23(3), 195.
36. M. Mukhopadhyay, J. Supercrit. Fluids, 2003,25(3), 213.
37. A. L. Scheidgen and G. M. Schneider, Fluid Phase Equilibr; 2002, 19k197, 1009.
38. F. Becker and P. Richter, Fluid Phase Equilibr; 1989,49, 157.
39. E. Reverchon, I. De Marco and G. Della Porta, Znt. J. Pharm., 2002,243(1-2), 83.
40. E. Reverchon and I. De Marco, J. Supercrit. Fluids, 2004, in press.
41. E. Reverchon, G. Della Porta and M. G. Falivene, J. Supercrit. Fluids, 2000, 17(3), 239.
42. E. Reverchon, G. Della Porta, I. De Rosa, P. Subra and D. Letourneur, J. Supercrit. Fluids,
2000, 18(3), 239.
43. E. Reverchon and G. Della Porta, Powder Tech., 1999, 106( 1-2), 23.
44. J. 0. Werling and P. G. Debenedetti, J. Supercrit. Fluids, 1999, 16(2), 167.
45. J. 0. Werling and P. G. Debenedetti, J. Supercrit. Fluids, 2000, 18(l), 11.
46. F. ChBvez, P. G. Debenedetti, J. J. Luo, R. N. Dave and R. Pfeffer, Ind. Eng. Chem. Res.,
2003,42(13), 3156.
47. J. C. G. Calado and J. N. Canongia Lopes, Pure Appl. Chem., 1999,71(7), 1183.
CHAPTER 13
Calorimetric Meas urernents

of Thermophysical Properties
for Industry
JEAN-PIERRE E. GROLIER AND FLORIN DAN
1 Introduction
Materials science is currently witnessing an impressive acceleration of activities.
New advanced materials are constantly emerging. The materials that will be used or
needed in the next decade are not yet known. Since such materials are often used in
special environments or under extreme conditions of temperature and pressure, their
careful characterization must be done not only at the early stage of their development
but also in their entire life cycle. Furthermore, their properties as a function of tem-
perature and pressure must be well established for the optimal control of their
processability. Undoubtedly, thermal and calorimetric techniques are essential in this
respect. In relating thermal as well as mechanical behavior to materials’ structures,
these techniques are perfectly adapted to provide accurate data in wide ranges of
temperature and pressure.
Typically, thermophysical properties are the most important when dealing with
materials subjected to thermal variations and mechanical constraints. The purpose of
this chapter is not to produce an exhaustive catalogue of such properties. The choice
has been made to consider the most important or the key properties. Interestingly,
materials considered here encompass gases, liquids and solids, polymers essentially
in the latter case. The selected properties are of two types: bulk properties and phase
transition properties. The bulk properties are either caloric properties, like heat
capacities Cp, and mechanical properties, like isobaric thermal expansivities a,,
isothermal compressibilities q-, and isochoric thermal pressure coefficients &. The
two main thermal properties are related to first-order transitions, fusion and crystal-
lization, and glass transitions. All these properties are now accessible thanks to
recent progress in experimental techniques, which allow measurements in the three
physical states over extended ranges of T and p , including in the vicinity of the
Calorimetric Measurements of Thennophysical Properties for Industry 145
critical point (at least in the case of gases and liquids). Measurements are also pos-
sible for polymers submitted to pressure in the presence or absence of gases. In what
follows, some of the above properties in selected materials or systems serve to illus-
trate the state of the art in both available techniques and instrument performances.
As a matter of fact, the development of the instrumentation parallels the need for
data for new materials to be used in specific applications and under further extreme
or particular conditions. For all selected examples, the industrial needs, which were
very often the incentive factor for making the measurements, will be stressed.
2 Heat Capacities
The heat capacities of pure substances or of their mixtures, irrespective of their
aggregation states, are often used to correlate the main thermodynamic properties,
including the phase equilibrium properties, over large domains of temperature. They
are now recognized as key data for engineering calculations. Heat capacities are also
used to verify the validity of existing equations of state (EOS) or even to develop
new ones. Furthermore, direct measurements of heat capacities in the vicinity of the
critical point are of practical importance, due to the spectacular changes generally
observed under conditions around this point.
2.1 Heat Capacities of Gases

Understanding the thermodynamics of natural gas is essential for optimizing the man-
agement of its resources. Correlation and prediction of thermodynamic and transport
properties of natural gas are needed at different stages during the exploration and
assessment of reserves as well as for its production, transport, and storage.
In this context, the calorimetric technique originally developed for measuring heat
capacities for liquids2 has been adapted to investigate gases.3 Basically, the method
involves recording the differential heat flux between the sample cell containing the
gas under study and the reference cell during a temperature increment AT from an
initial temperature Tito a final temperature Tf at constant pressure. The total heat Q
exchanged is obtained after integration of the heat flux between the two thermal
equilibria reached at the initial and final temperatures, respectively. In order to cal-
culate the isobaric specific heat capacities cp (J K - 1 g - l ) , two additional quantities
are necessary; the molar densities have been calculated from semi-empirical equa-
tions of state, and the inner volume of the sample cell, Vs, was calculated by cali-
brating the calorimeter with gases for which the molar densities and heat capacities
are known. The heat quantity 6Q obtained from the integration of the heat flux can
be expressed as
6Q = mcpdT-m7($) dP
where m refers to the mass and v refers to the specific volume of the gas in the sample
cell. The two terms of Equation (1) have been calculated for some gases, such as nitro-
gen, helium, and carbon dioxide at temperatures ranging from 323.15 to 423.15 K
146 Chapter 13
and pressures ranging from 5 to 25 MPa, respectively. It was found that the influence
of the pV-work (the second term on the right-hand side of Equation (1)) may not be
negligible in the case of CO, and in this case it has been taken into account in the cp
calculations.
The same technique was then used to determine the specific isobaric heat capaci-
ties of two reference gases, carbon dioxide and argon, and of five synthetic gaseous
mixtures representing different types of natural gas in the above-mentioned ranges
of pressure and temperature. For the pure gases, the results have been compared with
the calculated values obtained with the equations of state of Ely et aL4 and of Stewart
and J a ~ o b s e nrespectively.
,~ Experimental and calculated values for CO, and Ar are
given in Figures l(a) and l(b), respectively; they show the ability of the experimen-
tal technique to deliver results that can be obtained with a 2% confidence.
The experimental heat capacities of the five synthetic gaseous mixtures having the
molar compositions listed in Table 1 were used to compare the performance of var-
ious equations of state, namely: AGA 8,6 S ~ a v e Peng-Robinson,*
,~ Lee-Ke~ler,~
Simonet-B6har-Rauzy,l0 and chain-of-rotators. Numerical comparison between
experiment and EOS is given in the right-hand part of Table 1. The absolute average
deviations (in %) show that the different EOS, with the exception of the chain-of-
rotors model, agree with the isobaric heat capacities of the investigated gases within
the experimental uncertainty, i.e .=2%.
0.75
0.70
r
m
k 0.65
2
0"
0.60
0.55
5 10 15 20 25
Figure 1 Experimental heat capacities (symbols) compared with calculated values (lines) of
carbon dioxide (a) and argon (b), respectively
Table 1 Molar composition (mole/ %) of the five gaseous mixtures and the absolute
average deviations (in %) obtained using different equations of state
(EOS), over the full pressure and temperature ranges investigated
Composition (% /mole) EOS (% absolute average deviations)
CH, C,H, C,H, N , CO, AGA8 Soave P-R L-K S-B-R C-0-R
GasA 87.570 9.380 2.155 0.895 - 1.8 2 2.3 2.0 2.5 3.8
Gas B 96.072 2.077 1.433 0.418 - 0.6 1 0.6 1.3 0.5 2.6
Gas C 83.427 3.584 0.871 10.62 1.535 0.7 1.1 0.6 0.9 0.5 5.8
Gas D 85.200 14.80 - - - 0.5 - - - - -
GasE 99.000 - - - 1.000 0.8 - - - - -
Calorimetric Measurements of Thermophysical Properties for Industry 147
2.2 Heat Capacities of Liquids

The above-mentioned method was initially developed for measuring the isobaric
heat capacities of aqueous salt solutions up to 573 K and 30 MPa.2 For a typical
run, the sample cell was loaded with the sample solution and the reference cell was
loaded with a reference fluid of known heat capacity (usually water). Then, the
temperature was increased from Tito Tf, at constant pressure, and the difference Q
in the transferred heat was corrected taking into account both the cell's volumetric
dissymmetry and the differences between the densities and specific heat capacities
of the measured sample and reference fluids, respectively. Such an experiment
allows the measurement of the product pc, representing the isobaric heat capacity
divided by volume. In order to obtain the desired isobaric heat capacity, c,, of the
solution, it was necessary to know its density. For this purpose, the isobaric spe-
cific heat capacity and density were represented by polynomials in terms of tem-
perature T:
n m
c , = C c i T i and p = c r j T J
i=O j=O
The parameters rj for the density polynomial were obtained from available corre-
sponding densities. Knowing the calibration constant for each set of experiments and
the densities of the measured fluids allowed the calculation of ci and finally the spe-
cific heat capacity. The performances of the calorimetric arrangement have been
tested by measuring the heat capacity for liquid n-hexadecane as well as of NaCl
(aquous) in the molality range 0.1-2.0 mol kg-'. Measurements taken at different
temperatures and pressures compare well with the literature data. 12J3-15.
Later, the pressure-scanning technique16was used to investigate the thermophys-
ical properties, isobaric molar heat capacity C, (J K-'mol-'), and ap,over extended
T and p of several fluids or their mixtures, such as quinoline, n-hexane, l-hexa-
namine, and its binary mixtures with 1-hexanol, rn-cresol, and its binary mixtures
with quinoline, etc. As a rule, for simple liquids without strong intermolecular inter-
actions, such as n-hexane, for example, both the C,-isotherms and the pressure
effects (isotherms) on the isobaric heat capacity at pressures up to 700 MPa exhibit
minima. It is worth recalling that the pressure effect on the C, is related to the iso-
baric thermal expansibility a, by the following equation (the effect of pressure on
the a, is discussed in the next section):
When the a,-isotherms cross at a given pressure, (dad3T), changes its sign and this
property is responsible for the above-mentioned minima. In addition, as the temper-
ature increases, the minima are shifted to higher pressures. At higher temperatures,
especially when approaching the critical temperature, the minimum becomes a shal-
low one. As an example for associated liquids, some selected C,-isotherms and the
pressure effects at different temperatures on the isobaric heat capacity of rn-cresol
are presented in Figures 2(al) and (a2),17respectively.
148 Chapter 13
503K . _._.-
.... . - . . _ --.
353 K
2.15
0 100 200 300 400
-0.124 I I I 1 I I I I
0 100 200 300 400
(al) p IMPa (a2) p MPa
0
-2
-3
0 0.2 0.4 0.6 0.8
Figure 2 Isobaric heat capacities ( a l ) and pressure efects on the isobaric heat capacities
(a2)for m-cresol. Typical W-shapedexcess molar heat capacities (b)at 298.15 K of
{1,4-dioxane+n-alkanes]mixtures; numbers refer to the n-alkane carbon numbers
At lower temperatures, where the associations are strong, there are no minima,
because the isotherms of a, cross at a higher pressure and the change of sign of
(daJdQ, does not sufficiently compensate the positive value of ap’ over the pres-
sure range under investigation. At higher temperatures, where the association bonds
are weakened or even broken, the minima appear and the behavior of m-cresol
becomes similar to that of simple liquids.
As regards normal-pressure heat-capacity calorimetry, during the two decades
1973-1993, Grolier and his co-workers produced the largest database on binary
organic mixtures of different types; this appeared in more than 40 papers with a
rough estimate of 150 systems.’*Among the systems investigated, the most striking
result was obtained in collaboration with Inglese and Wilhelm; it concerned the
unusual composition dependence of excess molar heat capacity of { 1,4-dioxane+
n-alkane} binary mixtures19 as shown in Figure 2(b). Typically, the CF -curve plot-
ted against the mole fraction of the cyclic ether showed a double minimum separated
by a marked maximum, yielding a W-shaped curve. Such a behavior was assigned
to the existence of preferential orientations and aggregations of some “flexible mol-
ecules” in admixtures with similarly flexible alkanes, depending on the concentra-
tion range. Later, the W-shaped curves, including theoretical modeling, were
obtained in several laboratories worldwide, and the number of systems known for
showing W-shaped CpE -curves has increased steadily, and this behavior is now rec-
ognized as being of relatively wide occurrence in mixtures of a strongly polar sub-
stance and n-alkanes, although it is not restricted to these types of systems only.
CalorimetricMeasurements of Thermophysical Propertiesfor Industry 149
3 Bulk Thermomechanical Properties

3.1 In Extended Ranges of T andp
Nowadays, for a thermodynamicist, p W-calorimetry (further referred to as scanning
transitiometry, its patented and commercial name20) is the most accomplished experi-
mental concept. It allows direct determinations of the most important thermodynamic
derivatives; it shows how, in practice, the Maxwell relations can be used to fully sat-
isfy the thermodynamic consistency of those derivatives. Of particular interest is the
use of pressure as an independent variable; this is typically illustrated by the relatively
newly established pressure-controlled scanning calorimeters (PCSC).16J7Basically,
the isobaric expansibility a,@,7)=( l/v)(dv/d7), can be considered as the key quan-
tity from which the molar volume, v, can be obtained and therefore all subsequent
molar thermodynamic derivatives with respect to pressure. Knowing the molar volume
as a function of p at the reference temperature, To,the determination of the foregoing
pressure derivatives only requires the measurement of the isobaric expansibilities a, as
a function of p and T. Thus, it is easy to compute the total variation of the above-men-
tioned thermodynamic function over extended p, T-surfaces.
Many substances starting with gases, passing on to liquids and their mixtures, and
ending with high-molecular-weight semi-crystalline polymers have been character-
ized from the viewpoint of a, over extended ranges of pressure and temperature
using the PCSC technique; some selected examples are illustrated in Figure 3.
Measurements of the isobaric thermal expansivity a, of methane in the pressure
range from 50 to 165 MPa and along four isotherms (303,333,363, and 393 K) have
been performed in decreasing the pressure at a low constant rate, a=0.02 MPa.s-',
in such a way as to remain at thermodynamic equilibrium. Several hundreds of data
points were collected and fitted, as a function of p along each isotherm, to the fol-
lowing empirical equation:
A
a,= - (4)
@+PIC
The 32-term Modified Benedict-Webb-Rubin (MBWR) equation of state proposed
by Younglove and Ely21 was used to represent the pVT-behavior of supercritical
methane and, in particular, to compute a,. The experimental data agree within 2%
with these values at the temperature 333 K, as shown in Figure 3(a).
The same technique was used to investigate liquid substances and their binary
mixtures. A general conclusion for these investigations is that nonpolar simple liq-
uids, without strong intermolecular interactions, like n-hexane, exhibit a single
crossing point of a, isotherms at moderate low pressure [Figure 3(bl)]. For liquids
with strong intermolecular interactions, the crossing of a, isotherms, which is pres-
sure dependent, is shifted to a higher pressure and this dependence is related to the
nature of the specific molecular interactions. This is typically shown by associated
liquids, like water [Figure 3(b2)].22
Crystalline polymers, like polyethylene as an example of the stable coexistence of
crystal and amorphous phases, have been characterized by a series of apmeasurements
on polyethylenes with various crystallinities at several temperatures as a function of
150 Chapter 13
pressure, up to 300 MPa.23Clearly, all a,-isotherms of the different polyethylenes have
a tendency to converge at high pressures, but they do not cross as observed for simple
liquids. Exceptions are only for isotherms near or above the temperature of the fusion
of crystalline polymers at atmospheric pressure as seen in Figure 3(cl). From these
measurements, it was found that a, is a linear function of crystallinity over the entire
p , T-surface under investigation, Figure 3(c2). The straight lines presented in Figure
3(c2) have been obtained by the least squares method. From such approximations, a,
of both amorphous and crystal phases have been derived. Thus, in Figure 3(c3), not
only the thermal expansivities of low-, medium-, and high-density polyethylenes, but
also the estimated a, of polyethylene amorphous and crystal phases as a function of
pressure up to 300 MPa at 302.6 K are presented.
2 .a
0.90
yx 2.4
P)
=8 2.0
0.70
1.6 0.50
1.2
0.30
50 ZOO 150 200 0 100 200 300 400
p /MPa tbl) p MPa
1.20
\
MDPE
o.20' ' ' '

0 0 50 100 150 200 250 300
P (el) p lMPa
0.8, I
L I
0 0.2 0.4 0.6 0.8 1 0 50 100 150 200 250 300
ic2) Crystalinky id) p MPa
Figure 3 Transitiometric investigation of the pressure effect on the isobaric thermal expan-
sivities apof gaseous, liquid, and solid substances: ( a ) comparison between calcu-
lated and experimental data of methane at 333 K; ( b l ) quinoline as a simple fluid;
(b2) water as the associated liquid; ( c l ) efect of temperature on medium-density
polyethylene (MDPE); (c2) evolution of apwith the crystallinity (points represent
the experimental data and the lines were obtained by the least squares fitting of
data); (c3) comparison between the experimental apfor three polyethylenes with
dijferent degrees of crystallinity and the predicted values for crystal and amorphous
phases obtained by extrapolationfrom linear fitting of the experimental data
Calorimetric Measurements of Thennophysical Properties for Industry 151
3.2 In the Vicinity of the Critical Point
The above-presented pressure-scanning procedure allows direct determination of
both a, and K~ of the investigated substance. However, when approaching the satura-
tion line, especially in the vicinity of the critical point, where both apand KT tend to
infinity, achievement of a measurable pressure variation requires a very large volume
variation, and the pressure-scanning procedure becomes less precise or even unus-
able. Consequently, the isochoric thermal pressure coefficient pv= (dp/dT), =( o ! ~ / K ~ )
is hardly accessible. Thus, the volume-scanning procedure was used instead for meas-
uring the thermomechanical coefficients of n-hexane in the vicinity of the critical
point.24When the volume is varied as a linear function of time at isothermal condi-
tions with a scanning rate c, the variation of heat effect with time may be written as
where PT(V)is the thermal power generated by such a process, and S is the entropy.
The volume variations can also be expressed as a step function of time. Then, for
each step, the integration of Equation (5) permits the direct determination of the ther-
mal pressure coefficient, pv,as follows:
as aP Q
m
pV=(m)T=($F)v=
This simple relation holds, since the volume is an extensive variablz, under the con-
dition that the measured heat effect, Q, corresponds to the same amount of substance
under investigation as the change in volume AV (see ref. 24). In addition, apcan also
be obtained from heat effects, Q, and pressure variations, Ap, and similarly, K~ can
be obtained from volume and pressure variations as follows:
0.35 0.020
0.30
r
0.25
0.016
-a
P
k
2 0.20 0.012 h
h
3 0.15
0.008
-$4-
0.10
0.004
0.05
0.00 0.000
0 5 10 15 20
Figure 4 Thennomechanical coeficients a K and Pv of n-hexane at 507.2 K determined

p’ .?
by volume-controlled scanning transitlometry
152 Chapter 13
where Vin is the whole internal volume, AVe,c,, is the external volume variation cor-
rected for compressibility of the hydraulic liquid, and V, is the volume of the inves-
tigated substance. However, when approaching the critical point, the pressure
variations are too small to be measured and the derivation of both apand becomes
impossible. This is illustrated (Figure 4) by the behavior of the three thermome-
chanical coefficients for n-hexane at 507.2 K as a function of pressure. The value of
pv at 3.05 MPa (0.0406 MPa K-') is very close to the value obtained from the cor-
relation equation for the saturation pressure of n-hexane, i.e. 0.040 MPa K-' (ref.
25). Although it is much more laborious than the pressure-scanning procedure, the
volume scanning permits one to approach the critical point much closer and to deter-
mine & directly in the vicinity of the critical point with a reasonable precision.
4 Phase Transition Thermal Properties

Scanning transitiometry is an extremely useful technique for studying phase transi-
tions. Different transitions, mainly in polymer systems under various constraints of
pressure, temperature, or chemical reagents, are reported hereafter.
4.1 Fusionkrystallization
The simplest case is the first-order phase transition of semicrystalline polymers in the
presence of a chemically inert fluid as a pressure-transmitting medium. Mercury was
preferred due to its chemical inertness and due to its convenient and well-known ther-
momechanical coefficients (ap= 1.8OX K-' and K~ = 0.40X MPa-I). The
polymer sample was always in intimate contact with the pressure-transmitting fluid.
The pressure effect on the melting/crystallization temperature of an MDPE sam-
ple, in the pressure range from 50 to 200 MPa, is illustrated in Figure 5. The isobaric
temperature scans were performed at a rate of 0.833 mK s-l either in ascending or
80
60 -
ym40 -
$ 20 -
0
ii 0 -
Y
Q 20 -
I
150 MPa
-40 -
-60 -
-80 . ' ' ' i ' ' ' ' i '
390 41 0 430 450 470

Te mperaturelK
Figure 5 Heatflow curves obtained on heating and cooling at a rate of 0.833 mK s-I for the
mercury-pressurized MDPE. The base lines were shifted for the sake of clarity on
the pressure efect on the meltinghystallization temperatures
descending mode and simultaneous recording of the heat flow and volume varia-
tions. Both the melting temperature Tmand crystallization temperature Tcrhave been
obtained at the onset of the transition by the conventional method applied to the
respective peaks, i.e. from the intercept of the greatest slope of the measured signal
with the base line. Taking the fusion at 200 MPa as an example, the rates of varia-
tion with temperature of both volume and heat flow yielded the same value of the
melting temperature, i.e. 456.2 K. Details with respect to the rate of volume varia-
tions are given elsewhere.26
As expected, both fusion and crystallization are shifted toward higher tempera-
tures with increasing pressure. The obtained value of 0.297 K MPa-’ for AT,,,/Ap
(Clapeyron slope) is in fairly good agreement with the known literature values.27
Variations of volume and enthalpy in the same process were obtained by integration
of the respective curves, the corresponding values being accordingly280.0573 cm3
g-I and -88.54 J g-].
The scanning transitiometric technique was also adapted to work under supercrit-
ical conditions. In this case, the pressure-transmitting fluid is the gas that is in a
supercritical state. The investigated sample, polymer or paraffin, was placed in an
open glass ampoule supported on a spring in the reaction cell, in order to be posi-
tioned in the central active part of the detector zone and in direct contact with the
pressurization fluid. The reaction vessel was connected to a pressure gauge and to
the high-pressure pump through a stainless-steel high-pressure capillary tube; it can
also be connected to a supercritical fluid tank or compressor for filling the fluid into
the pump. Such a setup can be easily used to investigate the pressure effect on the
(solid +fluid)-to-fluid transitions.
The tetracosane/methane (C,,H5dCH,), poly(vinylidenefluoride)/1,l -difluo-
roethylene (PVDF/C,H,F,), and poly(vinylidenefluoride)/nitrogen (PVDFN,)
binary asymmetric systems have been selected to illustrate this particular
It is well known that at elevated pressures, these systems show a pronounced non-
ideal behavior because of large differences in the molecular size of components. On
the other hand, both CH, and C,H,F, are known as “good” solvents for the corre-
sponding solids, tetracosane, and PVDF, respectively; even if C2H,F, dissolves
PVDF at high pressures and temperatures (over 200 MPa and 500 K), whereas
nitrogen is inert with respect to the polymers, a comparative investigation is of
interest. Two series of calorimetric heat flow curves of melting of tetracosane and
PVDF are presented in Figures 6(al) and (a2), respectively. The isobaric tempera-
ture scans are reported at a rate of 1 rnK.s-’. The experiments were performed first
in ascending pressures. The initial pressure was automatically kept constant while
heating; then, the apparatus was cooled down at the same pressure. Next, the pres-
sure was increased and the heating and cooling procedure was repeated. As Figures
two solids (for PVDF ==

6(al) and (a2) show, despite the differences between the molecular weight of the
113,000 and % = 330,000) as well as the physical
properties of the fluids, the curves of fusion are very similar. The fusion peaks
broaden with increasing pressure, especially above 40 MPa, while the amplitude of
the peaks concomitantly decreases. One can observe that at a pressure of 180 MPa,
the transition is only slightly visible, especially in the case of the tetracosane/CH,
system.
154 Chapter 13
(81)310
I .
31I
.
310 315 330 335
I
YO
T0IWW-l-K
200
180
160 .
60
40
20
0315 320 3s 3.30 336 340
Figure 6 Pressure eflect on the temperature-inducedand gas-assisted melting of tetracosane

and PVDF (for more details and explanations, see refs. 26 and 29): ( a l ) and (a2)
heat rate evolution during fusion in the presence of supercritical CH, and C,H,F,
(VF), respectively; ( b l )fluid phase equilibria in the tetracosane/methane system;
and (b2)partial p-T phase diagramfor the PVDF- VF and PVDF-N, systems. Note
the depression in the melting/crystallization temperatures in the pressure range up
to 30 MPa for the tetracosane/CH, and PVDFNF systems
A detailed presentation of the temperature of the melting transition for the both
systems as well as the temperature of crystallization for the PVDF/C,H,F, and
PVDF/N, is given in Figures 6(bl) and (b2). Under the inert N, pressure, both melt-
ing and crystallization temperatures increase with increasing pressure, similar to the
effect observed on the mercury-pressurized MDPE (Figure 5). In the investigated
range of pressure (0.1 to 30 MPa), the slope of T,-pressure plot and T,,-pressure
plots were 0.108 2 0.002 and 0.115 k0.002 K MPa-I, respectively. On the contrary,
CH, and C,H,F, depress the melting/crystallization point by sorption of the gas in
the polymer up to 30 MPa. The antiplasticization effect of the hydrostatic pressure
of the two fluids is apparent above 30 MPa. Consequently, crystallization and melt-
ing temperatures are a compromise between the effect of hydrostatic pressure, which
increases the transition temperatures of crystallization and fusion, and the solubility
of the fluids in the solute-rich phase, which reduces these values.
4.2 Glass Transition Temperature

4.2.1 By Temperature-ModulatedDifferential Scanning Calorimetry
Differential scanning calorimetry (DSC) has been used to successfully characterize
structure and transitions (changes in structure) in plastics for more than 40 years.
Plastics are becoming more complex in order to meet the demand for lower-cost
materials and improved physical properties. Consequently, it is becoming increas-
ingly more difficult to characterize the structure and resulting physical properties of
plastics that are often polymer blends and composites. The result is that new
approaches to using DSC or new techniques like Temperature Modulated DSC@
(TMDSC) are required. TMDSC employs a modulated or sinusoidal change in the
heating rate in order to automatically separate the heat capacity-reversing contribu-
tion from the nonreversing kinetic contribution in the total heat flow signal. Apart
from traditional DSC, with TMDSC, the two signals are generated in a single exper-
iment so that each of the components to the total heat flow signal can be shown and
analyzed independently.
Total = TMDSC Reversing + TMDSC Nonreversing.

The high potentiality of TMDSC is illustrated here for the evaluation and selec-
tion of the polymers used in the NanoImprint Lithography technique (NIL).30* 31 The
nano-imprint process can be successful in a manufacturing environment only if a
good critical dimension (CD) control can be achieved across the wafer. Usually, the
nano-structures obtained by NIL [see Figure 7(al) for printing geometric parame-
ters] are studied as a function of process parameters like the type of resist, pressure,
temperature, and time, depending on the thermal properties of polymer.
Figures 7(b) and (cl) illustrate the thermal signatures of two different polymers:
mr-18030, Figure 7(b), is a polymer designed for nano-imprint lithography and NEB
22, Figure 7(cl), is a negative e-beam photoresist formulated by Sumimoto
Chemical. The glass transition temperature, Tg,is observed on the reversing curve at
-0 24
50 100 150 200 250
tb) Temporituro/*C
Figure 7 Illustration of the TMDSC technique for measuring Tg of the polymers designed for
nanolithography: ( a l ) printing geometric parameters; (a2) exumple of apperiodic
dense lines; (b)t h e m 1 signature of the polymer mr-18030; ( c l ) t h e m 1 signature of
the polymer NEB 22; and (c2) variation of the thickness of the NEB22 resist as a func-
tion of the bake temperature, showing the domains of optimal deposition temperature
156 Chapter 13
around 125 and 80 "C for mr-I8030 and NEB 22, respectively. It is worth mention-
ing that under classical DSC, such glass transition temperatures cannot be observed.
In the case of mr-18030, the decomposition of polymer is characterized by the
endothermic peak observed on the non-reversible curve at around 230 "C, showing
that the polymer has to be pressed in the temperature range 130-200 "C. In the case
of NEB 22 photo-resist, several peaks are observed on the non-reversing part of the
heat flow; the two endothermic peaks located at around 100 and 130 "C correspond
to the evaporation of the Photo Acid Generator and residual solvent. The exothermic
peak above 150 "C was attributed to the reaction between the polymer and
crosslinker. Further, NEB 22 polymer has been printed under different time and tem-
perature conditions and the results of the residual polymer thickness measurements
allowed to identify the domains of optimal deposition temperatures, Figure 7(c2).
Many other resists or their blends have been studied by TMDSC, including posi-
tive resists, which contain a protection group and no crosslinker. Typically, TMDSC
experiments were conducted under a modulation amplitude of A, = 0.8 K, a period
of p = 60 s, and an average heating rate of q = 2 K min-'. All samples analyzed had
a mass of about 1-3 mg.
4.2.2 By Scanning Transitiometry

The glass transition temperature is also affected by pressure since an increase in
pressure causes a decrease in the total volume, and an increase in Tg is expected
based on the prediction of decreased free volume. This result is important in engi-
neering operations such as molding or extrusion, when operation too close to Tgcan
result in a stiffening of the material.
Investigation of the glass transition of polymers under pressure is not a simple
problem, especially in the case of elastomers whose Tg are usually well below the
ambient temperature. In this case, the traditional pressure-transmitting fluid, mer-
cury, must be replaced since its crystallization temperature is relatively high, i.e.
235.45 K. The choice of the replacement fluid is a challenge because it should be
chemically inert with respect to the investigated sample (with respect to all its con-
stituents). Also, values of its thermomechanical coefficients, compressibility, K~ and
thermal expansivity, ap,should be smaller than those of investigated samples. An
additional difficulty in the investigation of second-order-type transformations is the
relatively weak effect measured. It is well known that the amplitude of the heat flux
at the glass transition Tgincreases with the temperature scanning rate, while the time
constant of Tian-Calvet-type calorimeters imposes temperature scan rates that are
slow compared to typical DSC scan rates. However, with the use of an ultra-cryostat
coupled to the transitiometer, we have used a temperature program that allowed the
accurate measurement of Tgat relatively high scanning rates.26In a typical run, the
temperature of the cooling liquid was lower than that of the calorimetric block dur-
ing the stabilization periods (isothermal segments) and it was higher during the
dynamic segment. In such a way, the scanning rate could be increased up to 0.7 K
min-I, always maintaining a minimal difference between the target and real temper-
atures of the calorimetric block. Since the temperature gradient between the "heat-
ing fluid" and the calorimetric block was kept constant (20 K), the power uptake of
Figure 8 Illustration of the scanning transitiometry technique for low-temperature, high-

pressure investigation of polymers glass transition. The pressure coeficient of the
glass transition temperature is given in the inset
the heating elements was quasi-constant; thus, the interference of sudden changes of
power uptake on the calorimetric signal was avoided.
Figure 8 shows the evolution of Tgof a poly(butadiene-co-styrene) vulcanized rub-
ber during isobaric scans of T with the temperature ranging 218.15 to 278.15 K at 0.4
K.min-', in the pressure range from 0.25 to 90 MPa. As can be seen in the inset of
this figure, the Tg increases linearly with pressure with a slope of 0.193 20.002 K
MPa-'. It should be noted that Tg is expressed as the temperature corresponding to
the peak of the first derivative of the heat flux (i.e.the inflexion point of the heat flux).
5 Conclusion
Modern industrial activities are characterized by more and more sophisticated
processes to exploit natural resources and produce high-quality products and serv-
ices in the best economic conditions while preserving the natural environment and
human health. The basis of these elementary rules relies on elaborated and pertinent
chemical engineering calculations. A large amount of good-quality thermodynamic
data and of thermophysical properties are a prerequisite for such calculations. In the
past and for quite a long time, equations of state and theoretical models have very
often been developed to predict or estimate thermophysical properties as well as
phase equilibria of a variety of fluid systems. Very few and poor experimental data
were available for validation of the models and most of the models concerned rather
simple molecules, or simple fluid mixtures. New advanced materials including spe-
cialty polymers and processing operations, like extrusion, over extended temperature
and pressure ranges, require accurately measured above-mentioned properties, very
often under extreme conditions of T and p . Newly developed techniques and instru-
ments, such as p VT-calorimetry, scanning transitiometry, and temperature-modu-
lated differential calorimetry (TMDSC), to cite a few, show how important thermal
158 Chapter 13
and calorimetric methods are in this scientific endeavor. Technological progress has
permitted these different techniques to be used to provide excellent, say unambigu-
ous, reference values from low to high temperatures, over very large pressure ranges
and in the critical region.
References
1. L. Dordain,J.-Y. Coxam, J. R. Quint and J.-P. E. Grolier,J. Supercritical Fluids, 1995,8,228.
2. J.-Y. Coxam, J. R. Quint and J.-P. E. Grolier, J. Chem. Thermodynam., 1991,23, 1075.
3. L. Dordain, J.-Y. Coxam and J.-P. E. Grolier, Rev. Sci. Znstrum., 1994,65, 3265.
4. J. F. Ely, J. W. Magee and W. M. Haynes, Research Report 110, Gas Processor
Association, 1987.
5. R. B. Stewart and R. T. Jacobsen, J. Phys. Chem. Ret Data, 1989,18, 639.
6. American Gas Association, Transmission measurement Committee, N8, American
Petroleum Institute MPMS, Chapter 14.2, 2nd edn, November 1992.
7. G. Soave, Chem. Eng. Sci., 1972,27, 1197.
8. D.-Y. Peng and D. B. Robinson, Znd. Eng. Chem. Fundam., 1976,15,59.
9. B. I. Lee andM. G. Kessler,A.Z.Ch.E.J., 1975, 21, 510.
10. E. BChar, R. Simonet and E. Rauzy, Fluid Phase Equilibr, 1986,21,237.
11. C. H. Chien, R. A. Greenkorn and K. C . Chao, A.Z.Ch.E. J., 1983,29,560,
12. V. Jr. Ruzicka, M. Zabransky and V. Majer, J. Phys. Chem. Ref., 1991,20,405.
13. J. A. Gates, D. H. Tillet, D. E. White and R. H. Wood, J. Chem. Thermodynam., 1987, 19,
131.
14. P. S. Z. Rogers and C. J. Duffy, J. Chem. Thermodynam., 1989, 21,595.
15. K. S. Pitzer, J. C. Pepier and R. H. Busey, J. Phys. Chem. Re$ Data, 1984, 13, 1.
16. S. L. Randzio, J.-P. E. Grolier, and J. Quint, Rev. Sci. Znstrum., 1994, 65, 960.
17. S. L. Randzio, E. A. Lewis, D. J. Eatough and L. D. Hansen, Znt. J. Thermophys., 1995,
16, 883.
18. J.-P. E. Grolier, Thermochim. Acta, 1997,300, 149.
19. J.-P. E. Grolier, A. Inglese and E. Wilhelm, J. Chem. Thermodyn., 1984, 16, 67.
20. Scanning Transitiometer ST-6VI from BGR TECH; S . L. Randzio, J.-P. E. Grolier, J.
Zaslona and J. Quint, French Patent 09227, 1991; S. L. Randzio and J.-P. Grolier, French
Patent 02612, 1998.
21. B. Younglove and J. E Ely, J. Phys. Chem. Re& Data, 1987, 16, 577.
22. S. L. Randzio, Thermochim. Actu, 1997, 300, 29.
23. L. Rodier-Renaud, S. L. Randzio, J.-P. E. Grolier, J. R. Quint and J. Jarrin, J. Polym. Sci.
Part B: Polym. Phys., 1996,34, 1229.
24. S. L. Randzio, J.-P. E. Grolier and J. R. Quint, High Temp.-High Pressures, 1998,30,645.
25. S. L. Randzio, J.-P. E. Grolier, J. R. Quint, D. J. Eatough, E. A. Lewis and L. D. Hansen,
Int. J. Thermophys., 1994, 15,415.
26. J.-P. E. Grolier, F. Dan, S. A. E. Boyer, M. Orlowska and S . L. Randzio, Znt. J.
Thermophys., 2004,25 (2), 297.
27. S. L. Randzio and J.-P. E. Grolier, Anal. Chem., 1998,70, 2327.
28. S. L. Randzio, J. T h e m . Anal. Cul., 1999, 57, 165.
29. S. L. Randzio, Ch. Stachowiak and J.-P. E. Grolier, J. Chem. Thermodyn., 2003,35, 639.
30. C. Perret, C. Gourgon, G. Micouin and J.-P. Grolier, Jpn. J. Appl. Phys., 2002,41, 4203.
31. L. Pain, C. Higgins, B. Scarfoglikre, S. Tedesco, B. Dal’Zotto, C. Gourgon, M. Ribiero,
T. Kusumoto, M. Suetsugu and R. Hanawa, J. Vac. Sci. Technol., 2000, B16(8), 3388.
CHAPTER 14
Plastic Recycling
WOLFGANG ARLT
1 Introduction
This chapter describes post-consumer plastic recycling, where “recycling” refers to
plastic material and may involve material recycling, chemical recycling or thermal
recycling.
In material recycling, the material of the original application is restored but in
another form. The recycling can involve a pure material or a mixture. Recycling of
the pure material often occurs in the production of the merchandise, e.g. the pro-
duction of insulating roofing materials. Various plastic manufacturers return rejected
parts and the trim from fabrication processes back to the production process, thus
contributing to the formation of new plastip Droducts. This procedure, known in the
plastics industry as regrinding, can be repeated numerous times, as long as the per-
centage of regrinds remains low. Material recycling in mixtures occurs when plastic
materials from the households are collected. Because most polymers, even of the
same species, do not mix well in the melt, a demixed polymer is normally not use-
ful in recycling. One example of mixed material recycling is the production of park
benches from household plastics, which were first produced in Germany in
the1990s.
Chemical recycling relates to a chemical change and usually involves a crude mix-
ture of polymeric material, e.g. for thermosetting materials; it may be of interest to
chemically treat the plastic in order to return to the initial monomers or similar mate-
rials. From the thermodynamics point of view, it should be noted that the production
of polymers from monomers is an exothermic process and as a result, the monomer
production requires work and heat. Unfortunately, in many cases, the amounts
involved are unreasonable and hence uneconomical.
Thermal recycling involves, in most cases, the simple burning of collected
material and the utilization of the resultant exothermic heat. In some cases,
the material may have a chemical use such as in the chemical reduction of iron
oxides in a furnace. In this example, the plastic replaces coal as the reducing
agent.
160 Chapter 14
2 Cost of Recycling
The purpose of plastic recycling is to save energy (crude oil) and to reduce the prob-
lems related to landfilling plastic material. When considering plastic recycling, it is
important that at least the energy balance is fulfilled. The steps in the recycling
process of mixed plastics include collection, separation and replacement.
The collection of mixed packaging material from the public and consumers
involves either curbside collection or collection from drop-off centres. Curbside col-
lection involves consumers placing designated plastics in special containers outside
their homes for pick-up by a public or private hauler, and drop-off centres refer to
consumers bringing their plastics to a centrally located collection point. In contrast
to the collection at a production site, these costs are estimated to reach 0.50 m g . ’
The cost of separating or sorting the collected recycling material has been esti-
mated to add another 1-2 €/kg to the overall cost.’
Furthermore, the recycled material has to be cleaned and shaped for the future appli-
cation. In some cases, the sorted plastics are converted into flakes or pellets, which can
then be used to manufacture new products. The cost depends on the designed applica-
tion. Recycled material cannot be used in every application, for example, strict rules
for food packaging material often prevent the use of recycled plastics.
Summing up, the recycling cost quickly adds up to over 1.5-2.5 a g . This often
exceeds the price of virgin bulk plastics. One way to increase the amount of plastic
recycling is to institute a fee for landfilling plastic material. However, in some coun-
tries in the European Union, plans are being made to eventually close landfill sites.
This will boost the amount of plastics being recycled. It should be mentioned that
most polymers are made from crude oil, so plastic recycling is one way to reduce our
dependence on oil imports.
3 Thermodynamic Considerations
From the point of view of thermodynamics, a polymer is a valuable low-entropy
product that is produced by dissipating heat to the surrounding. Any breaking of the
bonds within the recycled polymer will be costly in terms of energy. As a result, plas-
tic recycling ideally should not involve the breaking of chemical bonds.
Furthermore, the second law of thermodynamics (disorder is the preferred state)
implies that the separation (creating order) of plastic materials from a mixture of plas-
tics must involve supplying heat and work. Recycling plastics is, by nature, a costly
business and in the extreme case, total recycling requires infinite work and heat.
4 Polymer Production
Polymer production in the world is increasing and as an example, the polymer pro-
duction in Germany is reported in Table 1 for the years 1999 and 2000.
The data in Table 1 give a good impression of the magnitude of the plastics being
produced and used in Germany and of the amounts of plastic that need to be recy-
cled in the future as a result of long residence times e.g. in construction.
A detailed breakdown of the collection and recycling of plastics has been obtained
from the German trade organization, “Griiner Punkt” (“Green Dot”). The Packaging
Plastic Recycling 161
Table 1 Polymer production in Germany,2in units of 106 metric tonnes, Mt. In this
table, PVC refers to poly-vinylchloride, PP to poly-propylene, PE to poly-
ethylene (high density or low density), PS and EPS to poly-styrene and PET
to poly-ethyleneterephthalate
Year
1999 2000
Production 14.8 15.5
PVC 1.55 1.60
PP 1.43 1.46
PE-LD+LLD 1.30 1.32
PE-HD 1.06 1.08
PS +EPS 0.79 0.83
PET 0.21 0.30
Thermosetting 3.61 4.08
Trade
Export 8.47 9.39
Import 6.00 6.32
Balance 2.47 3.07
Ordinance is the legal basis for the work of the private company DSD (Duales System
Deutschland). It searches and contracts companies involved in recycling plastics, pays
them and collects a licence fee from every producer of goods. This is dependent on
the weight of the total packaging material and the number of pieces produced. The
producer of the original packaging material also has to pay a licence fee.
In the year 2000,0.570 Mt, with 64% (0.365 Mt) being mixed plastics, were col-
lected by the contractors of the Green Dot. The structure of the material is presented
in Figure 1.
EPS/ cups
2% bottles
foils
23%
mixed
plastics
64%
Figure 1 Details of the packaging material collected in the year 2000 in Germany
162 Chapter 14
I distillation I
+
naphta, gas oil
.c
fi thermal cracking
f
ethylene
I polymerisation I
+
poly -eth ylene
Figure 2 Pathway involved in the production of poly-ethylene from crude oil
Because the cups, bottles and foils are more or less of one species of polymers,
the mixed plastics component is a more serious problem. It consists mainly of poly-
olefins. As a result, much of this remaining chapter will refer to poly-olefins.
Figure 2 describes the production pathway from crude oil to poly-ethylene.
5 Recycling Schemes
There are numerous schemes for separating plastics. In this chapter, the mechanical
methods are excluded. Examples of these include a swim-float scheme and mechan-
ical handling based on the identification of the type of plastics on a conveyor belt. A
comprehensive summary of the various types of plastics is given by ref. 3.
A thermodynamic method, more fitting to this chapter, has been proposed by
Nauman et al. They claim a process for the separation of a physically mixed solid
polymers by selective dissolution. They rely on the different polymer solubility char-
acteristics. Tables of this property have been reported and are based on regular solu-
tion theory and Hildebrand solubility parameters. The core of the Nauman invention is
to find suitable solvents to dissolve particular polymers under defined temperature and
pressure conditions. A mixture of polymers is first added to one solvent, at a given tem-
perature, in order to dissolve a particular polymer. The remaining polymer mixture is
then treated at a higher temperature with the same solvent or with a different solvent.
For clarity, two examples are taken from the patent4
A mixture of PS, PVC, LDPE, HDPE, PP and PET has to be separated. If the sol-
vent tetrahydrofuran (THF) is used then:
(i) PS and PVS dissolve at room temperature and form two liquid phases of 10%
impurity in each;
(ii) LDPE dissolves from the remaining polymer at 65 "C;
(iii) HDPE and PP dissolve in a single liquid phase at 165 "C and
(iv) PET at 190 "C.
The four liquid mixtures have to be treated separately. The pressure over the solu-
tion is only slightly below the vapour pressure of the pure solvent, since a polymer
concentration of up to 20 wt% does not really influence the vapour pressure. In the
given case, the final pressure is about 17 bar.
The same polymeric mixture can be treated with the solvent xylene. In this case,
the LDPE dissolves at 75 "C,HDPE at 105 "C, PP at 118 "C and PVC at 138 "C.
The PET remains undissolved. It should be noted that the fractions are not pure in
the sense of there being only one polymer present. However, at each of the above
temperatures, the fractions include one main polymer with small amounts of other
polymers being present.
Another proposal has been made by Arlt et aL5 The basic principle of the inven-
tion is that most polymer melts tend to form two more or less pure (in terms of poly-
mers) phases. If a solvent is found that does not interfere with this, the same
behaviour should be observed at higher temperatures. It was found that the solvent
should be of a chemical nature similar to the polymers.
There is, however, a mass transfer problem of demixing at lower temperatures
caused by high viscosities. Concentrated polymer solutions tend to take hours to
form two distinct liquid phases. A solution to this problem is the use of the lower
critical solution temperature. Because of their thermodynamic nature, all poly-
mer-solvent mixtures tend to form two liquid phases ("LL") with low viscosities, at
higher temperatures (LCST) as depicted in Figure 3.
LCST-region
UCST-region mass fraction PE
Figure 3 The composition-temperature diagramfor pure poly-olefins showing UCST (upper

critical solution temperature) and LCST (lower critical solution temperature)
behaviour (because of the nature of this mixture, the UCST cannot be observed and
therefore is dotted)
164 Chapter 14
Commingled polymer mixture
T=20 "C
solvent (11) dissolution T=140 "C
unsoluble +
compounds (111) filtration T-140 "C
F (IV) shear crystallization
(V) solid-liquid separation

TsK,I-100 "C
T-100 "C
solvent (VII) partial evaporation T-100 "C
(VIII) liquid-liquid demixing TLLE- 140 "C
(IX) evaporation I (XI evaporation T-140 "C
Figure 4 Recycling scheme as invented by Arlt et al. and Lindner
Figure 3 shows a more or less pure solvent phase (left-hand side of the diagram)
and a polymer phase existing at low and high temperatures. It was important to prove
that the demixing of the melt is maintained in the LCST region, forming two liquid
phases, with one phase involving one solvent containing one polymer and the other
phase containing the solvent and the other polymer,
Because only pure phases (in terms of polymers) are required, the liquid demix-
ing was not applied to the separation of HDPE and LDPE. An alternate process of
shear crystallization was incorporated. The full scheme of the process by Arlt
et dllindner, as published in the cited patents (see refs 5-7), is shown in Figure 4.
The shear crystallization process is given in step IV in Figure 4.
The solvent-containing polymer phases are separated by evaporation (steps
VI, VII, IX and X in Figure 4) and traces of solvent can be removed by vacuum
extrusion.
Table 2 Spec@ CO, emission in kg per kg HDPE
Production of virgin polymer Polymer recycling as Diference
as depicted in Figure 2 depicted in Figure 4
1.064 0.402 0.662
6 Energetic Considerations
As indicated in the chapter, the cost of recycling the plastic must be advantageous in
terms of work and energy. Arlt et aZ.* estimated the total heat for 1 kg of recycled
polymer (average for all types of polymers) following the scheme given in Figure 4
to be 3.459 MJ and the electrical energy to be 0.250 kWh. Arlt compared this result
to the production scheme depicted in Figure 2.
Table 2 shows that the proposed recycling scheme produces 40% of the carbon
dioxide emissions of virgin material. This fact underlines that it is advantageous in
terms of energy to recycle polymers and to use crude oil for energy production. The
reason is inherent in the many energy-consuming steps leading to the virgin polymer.
References
1. J. Brandrup, M. Bittner, W. Michaeli and G. Menges, Die Wiedervenvertung von
Kunststoffen, 1995.
2. Verband Kunststofferzeugende Industrie e.V.(German society for the polmer producing
industry): <http://www.vke.de> Oktober 2001.
3. R. J. Ehrig (ed), Plastics Recycling - Products and Processes, 1992.
4. E . B. Nauman, Polymer recycling by selective dissolution, US-patent 5 198471, 1993.
5. W. Ark, B. Bungert, G. Sadowski and S. Behme, Thermisches Trennverfahren fur ver-
mischte Polymere, German patent appl. 198 06 355.5; W. Arlt, G. Sadowski, M. Seiler
and A. Thiele, Beschleunigung der Phasentrennung von Polymer-enthaltenden flussigen
Phasen durch verzweigte Losungsmittel, German patent appl. DE 199 22 944, 14.5.99;
W. Arlt, G. Sadowski, M. Seiler and S. Behme, Verfahren zur Trennung vermischter
Polyolefine, German patent appl. DE 19905029, 29.1.99; W. Arlt, B. Bungert,
G . Sadowski and S. Behme, Thermisches Trennverfahren fur vermischte Polymere, German
patent appl. DE 19806355, 10.2.98, internat. Patent appl. PCT/EP99/05809, 6.8.99.
6. W. Lindner, Verfahren zur Trennung polyolefinischer Kunststoffmischungen, German
patent application 19927523, 2000.
7. A. Keller and F. M. Willmouth, J. Macrom. Sci.-Phys., 1972, B6(3), 493-538.
8. W. Arlt and A. Korporal, unpublished data.
CHAPTER 15
Industry Perspective on the

Economic Value of Applied
Thermodynamics and the
Unmet Needs of AspenTech Clients
CHAU-CHYUN CHEN, SUPHAT WATANASIRI, PAUL
MATHIAS AND VALENTIJN V. DE LEEUW
1 Economic Impact of Applied Thermodynamics

Thermophysical properties, phase equilibria, and solution chemistries are the under-
lying physical and chemical phenomena of industrial chemical processes. Rigorous
thermodynamic modeling of such phenomena establishes a sound scientific founda-
tion for simulation of industrial chemical processes and subsequent process devel-
opment, optimization, and control.
The oil and gas, chemical, polymer, and pharmaceutical industries are striving to
create new products, catalysts, recipes, and processes. At the same time, they are
challenged to shorten the time-to-market, ensure process and plant safety, and
achieve regulatory compliance. Worldwide, the industries spend an estimated $500
billion annually’ in product and process research and development, pilot plants,
process engineering, detailed engineering, construction and startup, manufacturing
operations, and plant maintenance.
In these industries, chemists and engineers enhance the quality and efficiency of
their value-added activities and decision-making processes by using engineering
software that helps them model the complex physical and chemical phenomena of
the chemical systems and processes they investigate. These enterprise-wide, first
principles-based engineering software products are easy-to-use, effective, and accu-
rate. Underlying these software systems are the integrated and validated applied
thermodynamics systems that describe the thermophysical properties and
chemistries of the relevant chemical systems. In short, the quality and efficiency of
chemists’ and engineers’ pursuits in the process industry are profoundly impacted by
Industry Perspective on the Economic Value of Applied Thermodynamics 167
the scientific rigor and engineering accuracy of the underlying applied thermody-
namics models and tools.
2 Applied Thermodynamics System

An applied thermodynamics system includes thermophysical property calculation
methods, applied thermodynamic models, and model parameter databanks. It also
includes phase equilibrium calculation algorithms, parameter estimation and regres-
sion, and a user interface. The system is an engineering tool for studying the physi-
cal and chemical behavior of chemical systems. When used along with other
engineering programs, such as a process simulator or a heat exchanger design pro-
gram, it provides rigorous and accurate thermophysical properties that are the foun-
dation for all process calculations. Its economic impact is greatly magnified when
the system is used consistently in all engineering calculations and programs, at every
stage of life cycle modeling and throughout the enterprise. These benefits and val-
ues are summarized in Figure 1 and Table 1.
3 Elements of an Applied Thermodynamics System

The elements for an enterprise-wide applied thermodynamics system can be divided
into the following broad categories:
Calculation engine
Analysis tools
Programmable interfaces
Deployment tools
Aspen Properties@Helps Drive Chemical Product 8 Process Innovations
Finance! Conceptual Process Dettidled Construction/ Operatland

Flannlngl Design Englneerhg Engineering Start-up Mahtenancel
RBD Asset Mgf
I t t
Aspen Engineering Suite@
t t f
Yodels and Data far Physlcat Properties, Solution Chemistries, and industriai Applicatlons
T t t
Theory Experiments Simulation
Figure 1 Applied thermodynamics system helps drive chemical product and process innovations
168 Chapter 15
Table 1 Applied thermodynamics system delivers value at every stage of life cycle
modeling, and throughout the enterprise
Stage Value
Conceptual design Support and improve decision making in various process
R&D activities and tasks such as solvent selection, solvent
recovery, product purification, etc.
Process engineering Develop high-fidelity models for process design, revamp, or
optimization
Reuse and develop thermophysical property knowledge
developed during the conceptual design stage
Eliminate the use of unvalidated models and manual transfer
of model parameters
Detailed engineering Apply high-fidelity models for equipment design
Apply consistent design by leveraging the same validated
thermophysical properties knowledge and models used in
preliminary design stages
Eliminate the potential errors associated with manual transfer
of data
Plant start-up and operation Develop a high-fidelity model of the plant for operator train-
ing by using the same validated, rigorous thermophysical
property models
Reduce operating costs by debottlenecking or optimizing the
process using either offline or online models developed with
the same validated, rigorous thermophysical property models
Throughout the enterprise Improve capital and operating efficiencies by improving
communications and sharing physical property knowledge
Ensure physical property data and model integrity across vari-
ous engineering tools used at each stage of the process lifecycle
3.1 Calculation Engine

The calculation engine is the heart of the property system. The specific contents of
the engine are determined by the industries the system is designed to serve. In gen-
eral, the engine should contain the following:
A flexible thermophysical property calculation framework to support various

calculation methods.
A comprehensive library of thermophysical and transport property models.
Extensive pure component and binary system parameter databanks, including
high-fidelity data packages for important industrial processes.
Proven parameter estimation capability to fill in the gaps of missing properties
and for evaluation of new chemicals.
Comprehensive data regression capability to fit experimental data to models.
Robust and efficient mathematical algorithms for phase equilibrium calcula-
tions (e.g., VLE, LLE, VLLE, and solid precipitation).
Verified analytical derivatives of properties with respect to temperature, pres-
sure, and composition. This is a key requirement for equation-oriented process
calculations and dynamic or transient process simulation.
Robust implementation and continuity of results.
The engine should also be able to:
Extend easily to client property models and data to allow the company to use
proprietary knowledge.
Link to sources of experimental measurement data, such as the Dortmund
Databank or NIST TRC database.
Interface with third-party applications through CAPE-OPEN COM Thermo-
dynamic, OLE automation, or FORTRAN interfaces. This allows third-party
models and data to be plugged in and used.
3.2 Analysis Tools

Thermophysical property analysis tools are needed to allow chemists and engineers
to evaluate the performance of the models being used or developed for the chemical
systems and processes at conditions of interest. These analysis tools include property
tables, phase equilibrium calculations, phase envelopes, and plots of phase diagrams.
3.3 Programmable Interfaces

Programmable interfaces, whether proprietary or public, allow the calculation
engine to be used by other engineering programs. The interface must include:
Configuration (setup) of compounds and property methods to be used.

Property and phase equilibrium calculations.
Query of model details, such as parameter requirements.
Access (read) and change (write) of model parameters.
3.4 Deployment Tools

In order to maximize the value of the applied thermodynamics system throughout the
enterprise, it must be accessible to all process engineers and chemists who require
accurate thermophysical property calculations in their daily work. Web applications,
which do not require installation of the calculation engine on the user’s computer,
facilitate easy access to the system. Web applications can be designed to provide pure
component data such as normal boiling point and critical properties. They can also
provide access to the most frequently carried out calculations, such as phase equilib-
rium calculations, tabulation, and plotting of pure component properties as a function
of temperature and pressure, and mixture property calculations.
Microsoft Excel is one of the most widely used engineering programs. When coupled
with a rigorous applied thermodynamics calculation engine, it becomes a very powerful
engineering tool for various engineering applications. This can be accomplished
through the Excel Add-In feature, which facilitates access to the calculation engine for:
Selection of compounds and property methods.

Calculation of properties and phase equilibrium using an Excel Function. Such
calculations can be performed on one or more cells. The input required for
170 Chapter 15
the calculations can be defined in other cells and referenced. Calculation results
are updated and propagated automatically when these input cells are changed.
As an alternative to the Add-In, users can also develop their own Excel Macros to
call the programmable interfaces that have been published.
4 Industrial Applications
Numerous industrial applications of applied thermodynamics have been reported in the
literature for engineering analysis of wide varieties of chemical systems and processes.
For example, Chen and Mathias' reported examples of physical property modeling for
the high-density polyethylene process and for sulfuric acid plants. Here, we present
two recent examples that are illustrative of numerous applications of applied thermo-
dynamics models in the industry for various process and product development studies.
4.1 Phase Equilibrium Modeling for Nylon-6 Process Simulation

A recent study by Seavey et aL2 presented a thermodynamically consistent model
for the phase equilibrium of water/caprolactam/nylon-6 mixtures, based on the
POLYNRTL (polymer non-random two-liquid) activity-coefficient model3 in
Aspen Properties*, the commercially available applied thermodynamics system
from AspenTech. The model predicts phase equilibrium for any binary or ternary
mixture containing water, E-caprolactam, and nylon-6 at industrially relevant
temperatures and pressures with an average error of 1%.The model was validated
with operating data from commercial manufacturing process units. It was then used
to simulate a melt train and a bubble-gas kettle train for the industrial production of
nylon-6. The phase equilibrium model facilitates fundamental kinetic, thermody-
namic, and mass transfer calculations for detailed predictions about reactor prod-
ucts over wide ranges of temperature and pressure conditions. The model represents
a significant advancement toward the goal of having a unified, consistent model
that describes the phase behavior of all commercially significant nylon-6 polymer-
ization and depolymerization technologies. It forms the scientific basis for the
enterprise efforts in new polymer grade development, evaluation of new catalysts,
process development, and optimization and control, to name a few.
4.2 Quickly Screen Solvents for Organic Solids

Frank et aL4reported examples of quickly screening solvents for organic solids. In
one particular example, solubilities of aspirin in four different solvents (acetone,
ethanol, chloroform, and cyclohexane) were used to regress the Hansen solubility
parameters for the solute, aspirin. Once the Hansen solubility parameters are identi-
fied for aspirin, Frank et al. showed that one could quickly estimate the solubilities
of aspirin in any solvent or solvent mixture as long as the Hansen solubility param-
eters are also available for the solvents.
Figure 2 shows the Hansen model calculation results versus experimental data for
aspirin solubility in various solvents at room temperature. The Hansen solubility
Solubility of Aspirin (Wt%)at 25 "C

100
U
A UNIFAC Prediction
.CI A Hansen Correlation
0.1
8e
c1
0.01
0.001
0.001 0.01 0.1 1 10 loo
Literature Data
Figure 2 UNIFAC model and Hansen model results versus experimental data for aspirin sol-
ubility in various solvents at room temperature
parameters for solvents were taken from H a n ~ e nwhile, ~ the Hansen parameters for
aspirin were regressed from the experimental data. Also shown are the UNIFAC pre-
dictions. Neither the Hansen model nor the UNIFAC model provides adequate quan-
titative results. Clearly, better models for solute solubility modeling are needed. This
example shows one of the numerous opportunities for developing innovative models
to address industrial product and process development needs. Since solvents facili-
tate organic synthesis, separation via crystallization or extraction, product formula-
tion, and equipment cleaning,6 a rapid and reliable solvent selection model would
benefit the industry in many different ways, including faster production, separation
and purification, reduced solvent emission and waste, lower overall costs, and better
production processes.
5 Quantitative Cost-Benefit Assessments

AspenTech has developed and applied quantitative cost-benefit analysis for engineer-
ing systems, including applied thermodynamics. The analysis, provided by Aspen Value
ProcessTM(AVP), is a collaborative process yielding multi-level, broad commitment for
business process changes enabled by software solutions. A key outcome of AVP is the
customers' validation of the estimated value. The financial information is highly sensi-
tive and consequently there are no published cases available. Here, we present the basic
approach and give typical quantitative assessment results. One could either carry out an
extrapolation of this analysis to assess the industry-wide benefits (as is done below), or,
alternatively, carry out a series of these assessments with specific producers.
The mechanics of the value calculation involves using a company's income state-
ment, in one form or another, (for example, the cost-of-goods-sold), and identifying
the independent, major profit contributors that impact the different items in the
income statement, e.g., average sales price increase, total production increase, engi-
neering efficiency increases and cost savings, operations cost savings, and capital cost
savings. For each, a complete list of specific mechanisms that generate quantifiable
benefits is developed, by a business process. For example, having more accurate
172 Chapter 15
density estimations can improve the timing for ordering feed for a stock tank.
Improved order timing can, in turn, increase the number of inventory turns and, there-
fore, lower working capital. Line items are totaled; estimations for total potential and
a realistically reachable solution are made. For comparison, benefits of current work
practices and systems are subtracted. Solution cost is taken into account to deliver net,
additional recurrent added value. Additional public information is available at
http://www.aspentech.com/brochures/AVPBrochure.pdf.
A rough estimation for the process industries based on experiences with AVP
would be, ‘state of the art engineering systems can deliver 1-10% of turnover in net,
recurring additional benefits.’ The industry generally validates between 10 and 50%
of this ‘claim.’ A conservative estimation is that 10%is directly attributed to applied
thermodynamics. This gives a net, recurrent benefit range of 0.0145% of turnover
of the process industries, being $600 M to $30 B for a $6 trillion industry.
6 Unmet Needs of AspenTech Clients

While applied thermodynamics has contributed considerably to the work of chemists
and engineers in the process industries, numerous unmet or emerging needs for inno-
vative applied thermodynamic models remain. Some of these unmet needs, as we
understand them, are summarized below.
6.1 Estimation Methods for Pure Compound Properties

Many estimation methods are available for a wide variety of properties, such as nor-
mal boiling point, critical properties, vapor pressure, ideal gas heat capacity, and
liquid These estimation methods are based largely on the group-contri-
bution concept and were developed from databases consisting mainly of relatively
small molecules and for properties mainly of interest to small molecules. We are,
therefore, constantly forced to question the quality of the predictions, especially
when dealing with large, complex molecules. Many of these large molecules exist
as solids under normal conditions. Hence, properties for the solid phase are partic-
ularly important. As the industry shifts its focus to larger and more complex mole-
cules, we find existing property data, database, and property estimation methods to
be rather inadequate. For complex molecules, 3-D molecular configuration can
have a significant impact on the properties. To our knowledge, no estimation meth-
ods have taken into account the effects of molecular configuration. The deficiencies
with large, complex molecules exist for virtually the entire spectrum of physical
properties, from heat of formation, Cp,and density, to vapor pressure, melting tem-
perature, etc.
We particularly welcome estimation methods that incorporate available experi-
mental data, such as melting point. Engineers engage in a number of steps to quan-
tify their confidence in an estimation method and to improve the accuracy of the
prediction when working with new mole~ules.~ As shown in Table 2, a new molecule
with the structure has a molecular weight of 163.606, a measured normal boiling
point (Tb) of 237.6 OC,and a few vapor pressure data. Using an estimation method
Table 2 Estimations of vapor pressure with or without Tb experimental data

Experimental Structure only With Tb
T,, "C 237.6 273.94 237.6
T,,"C 532.8 479.2
Vapor pressure at 300 "C,bar 3.445 1.753 3.697
that relies on the structure only (the Joback method), the estimated Tb is in error by
15%. This Tb is then used to estimate critical temperature, Tc. Critical pressure, Pc,
is also estimated from the Joback method. Based on these estimated Tb, Tc,and Pc,
the Riedel method is then used to estimate liquid vapor pressure at 300 OC, with an
error of 49%. By using the experimental Tb, the estimated vapor pressure is signifi-
cantly more accurate (7% error). Chemists and engineers have a healthy, profes-
sional skepticism of results from all prediction methods, whether ab initio or
empirical. The use of experimental data not only improves the prediction results but
also raises the confidence level associated with the prediction methods.
6.2 Cubic Equations-of-State

A broad range of cubic equation of state models (EOS) are successfully used today. The
EOS range from the standard Soave-Redlich-Kwong and Peng-Robinson, which is
widely used in the hydrocarbon processing and related industries (oil & gas and petro-
chemicals), to a new class of models that extend the range of applications to chemi-
c a l ~ . New
' ~ ~models
~ ~ are continually being developed and are too numerous to cite.
For non-ideal chemical systems, given the large number of models/mixing
ruledalpha functions available, clear choices for the preferred EOS have not yet
emerged. We have found the PSRK model of Holderbaum and Gmehling'O to be
useful to our users. The PSRK model's predictive capability is based on the UNI-
FAC model and is supplemented with special parameters for light gases. At pres-
ent, it is our opinion that a model that is based on one of the CEOS/AE mixing
rules, an appropriate alpha function, and an improved density prediction will ade-
quately serve industry needs to model non-ideal chemical systems. 12-14 However,
to be useful to the industry, a successful model must have an extensive set of pure
component and binary interaction parameters available. Unfortunately, most pub-
lished models are only accompanied by a small set of such parameters. Therefore,
the deficiency in this particular area lies probably not in the model or mixing rule,
but in identifying a satisfactory model and then devoting the resources toward
developing a comprehensive library of pure component and binary interaction
parameters to support it.
174 Chapter 15
6.3 Activity Coefficient Models
Existing activity coefficient models such as NRTL and UNIQUAC have been very
successful correlative tools that provide versatile, flexible thermodynamic frame-
works to correlate available experimental data. Once properly parameterized, the
models are then used to perform both interpolation and extrapolation.
In the exploratory phases of product and process research and development, gen-
erally little or limited data are available for model parameterization. However,
chemists and engineers need to evaluate a multitude of possible molecular or process
variations. Rather than high accuracy, successful evaluation depends on the ability to
discard the least viable options and select better options for further detailed studies.
Estimation methods that give reliable results for new or unknown species are
required. The well-known group-contribution methods, like UNIFAC, have demon-
strated their value in cases where the molecules can be decomposed to functional
groups for which parameters are already available.
At present, molecular simulation consumes too much CPU time to be used directly
as a means for general-purpose property estimation, including phase equilibrium.
However, recent advances with the COSMO-RSl5 and COSMO-SAC16models allow
prediction of activity coefficients from electron density profiles that are computed from
quantum chemistry methods for molecules. If libraries of electron density profiles for
molecules can be made public, these models can provide an alternative predictive tech-
nique to UNIFAC. However, the quality of the predictions still needs to be improved.
We are currently exploring an interesting approach with the segment NRTL
method,17where three common model segments with established interaction param-
eters (e.g. ethylene segment as a model segment for all non-polar, hydrophobic
chains or groups, -OH segment as a model segment for all hydrophilic, hydrogen
bonding groups, etc.) are used to characterize any known or unknown molecules.
The model correlation parameters are the equivalent numbers of these common
model segments, i.e., hydrophobicity, polarity, and hydrophilicity, in each molecule.
Available experimental phase equilibrium data are used to determine these model
parameters. Once all molecules are characterized this way, the model becomes a
simple and flexible thermodynamic framework to provide first approximation pre-
dictions. This approach combines the correlative power of the NRTL model with the
predictive power of the UNIFAC model. The model removes the rigidity associated
with the UNIFAC model in which molecules must be defined a priori in terms of
specific functional groups. The model seems to be particularly valuable in modeling
systems with large, complex molecules. It also provides a versatile framework that
can be easily extended to systems with electrolytes, polymers, and biologicals.
6.4 Modeling Electrolyte Systems

We have seen many successful industrial applications of applied electrolyte thermo-
dynamics models. In particular, the electrolyte NRTL activity coefficient model of
Chen and Evans1*has proved to be the model of choice for various electrolyte sys-
tems, aqueous and mixed-solvent. However, there are unmet needs that require fur-
ther development.
First, we need a predictive activity coeffcient model for electrolyte systems. The
electrolyte NRTL model is correlative, and it requires extensive experimental data
sets from which NRTL binary interaction parameters can be identified. The OLI
electrolyte model, with its extensive parameter database, has been serving as a
pseudo-predictive model. However, use of the OLI electrolyte model is limited to
dilute aqueous electrolytes, its parameter database is not open to the public, and its
electrolyte speciation is not supported by experiments.
Second, we need an equation-of-state for electrolyte solutions. Equations-of-state
are needed for modeling high-pressure applications with electrolyte solutions.
Significant advances are being made in this area.19-21Given that the electrolyte
NRTL model has been widely applied for low-pressure applications, we are hopeful
that, some day, there will be an equation-of-state for electrolytes that is compatible
with the electrolyte NRTL activity coefficient model.
Modeling mixed-solvent electrolyte systems proves to be a more challenging task.
Based on our experiences in modeling mixed-solvent electrolytes, we believe
progress needs to be made on two fronts: (1) better analytics to establish the correct
solution chemistries and the resulting speciation, and (2) better activity coefficient
models that provide higher resolution in accounting for the physical interactions
between various species, and segments of species, in the liquid phase.
6.5 Modeling Polymer Systems

Significant progress has been made with polymer thermodynamics recently. The most
notable models useful for polymer process modeling are the polymer NRTL model3
and the PC-SAFT equation-of-state.22Going forward, we expect further evolution of
the PC-SAFT EOS for modeling polar and aqueous phase polymers. It is interesting
to note that, in industry, we have observed strong resistance to the use of the associ-
ation term due to the much higher level of complexity with the EOS. We are also
looking for models and methodologies that can be used to characterize polymer
microstructures such as chemical composition distribution and polymer crystallinity.
6.6 Data Packages

Data packages are those models that provide high accuracy for a particular industrial
application or technology. Both high-fidelity data packages and engineering quality
data packages are highly valuable and are much sought after for all major industrial
chemical processes and related chemical systems. In fact, a great deal of industry
resources have been spent to create and recreate data packages for various chemical
systems. Examples of successful high-fidelity data packages are the NIST data pack-
ages for the thermophysical properties of fluids (e.g., NIST/ASME Steam Database
and the NIST Thermophysical Properties of Hydrocarbon Mixtures; for further
information, visit the NIST web site at http://www.nist.gov/srd/fluids.htm). Many
engineering quality data packages have been developed and they remain confidential
properties of the owner companies.
However, at present, publicly available and proven data packages are limited to
pure fluids and relatively simple mixtures like hydrocarbons. Publicly available data
176 Chapter 15
packages are particularly lacking for complex fluids like electrolytes and polymers.
Specifically, the industry could benefit considerably from data packages for major
electrolyte systems such as nitric acid, sulfuric acid, caustic, amines gas treating, etc.
Ideally, such data packages should be developed from widely accepted, thermody-
namically consistent, thermodynamic frameworks that allow industrial practitioners
to further extend the data packages to cover additional compounds and for wider
ranges of conditions.
6.7 Equilibrium Calculations

Reliable and fast equilibrium calculations (or so-called “flash” calculations) are the
mechanism by which thermodynamic properties are used in industry. This area has
received much attention in the past. Algorithms include successive substitution with
acceleration and stability a n a l y ~ i s , ~I n~~. i~d~e - O u t and
~ ~ Interval methods,26
Homotopy continuation methods27 with application to three-phase systems,28 and
systems with simultaneous physical and chemical e q ~ i l i b r i u m An . ~ ~area of recent
focus is the flash algorithm for mixtures containing polydisperse polymers.30
However, many challenging problems remain.
The flash algorithm of Behme et aL30only addresses polydisperse homopolymers.
While the algorithm works for two phases, vapor-liquid or liquid-liquid, the algo-
rithm is not yet capable of performing vapor-liquid-liquid three-phase flash. Also,
while the algorithm computes molecular weight distributions for polymers in differ-
ent phases, the algorithm does not yet take into account the bivariate distribution
of polymer molecular weight and copolymer composition. Further challenges include
the effects of copolymer sequence structure, i.e., block vs. random vs. alternating, etc.
High-pressure systems in the vicinity of critical points, such as synthesis gas and
air separation systems, remain a challenge. Our flash algorithm has difficulty in
identifying the correct phase state, or converging to the correct vapor-liquid solu-
tions. This problem may be exacerbated by the difficulty in obtaining the equation
of state volume root in the vicinity of the critical points.31Further work to improve
the algorithm and the equation of state volume root determination is required. It is
believed that the homotopy continuation methods are probably better suited for cal-
culations near the critical points.
Liquid-liquid equilibrium calculations remain a problem area especially for sys-
tems containing non-volatile species such as strong electrolytes or high polymers.
These species have negligible or no fugacities, and, as a result, many flash algo-
rithms cannot properly account for them.
Many important industrial systems make extensive use of surfactants for various
reasons. Surfactants may dissolve in the bulk liquid phases, form distinct micelles,
or preferentially concentrate on the interfaces. Existing flash algorithms do not
address micelles or interfaces as possible reservoirs for the surfactants. The flash
results are simply not valid for systems with surfactants.
The pharmaceutical and life science industries often deal with large, complex
molecules, and separation via crystallization is an important practice. Robust flash
algorithms for solid-liquid equilibrium, particularly systems with multiple poly-
morphs, are highly desirable.
6.8 Transport and Interfacial Properties
Transport and interfacial properties are often neglected in favor of research and
development efforts directed to phase equilibrium properties. Even less attention has
been devoted to such properties for electrolytes and polymers. In industrial practice,
the needs for transport and interfacial properties are numerous, i e . , detailed design
of heat exchangers, and distillation column tray and packing sizing calculations.
Both predictive and correlative models are needed for liquid viscosity, thermal con-
ductivity, surface tension, diffusion coefficients, etc.
Some recent works on polymer v i s c ~ s i t yand~ ~ *electrolyte
~~ viscosity34 are very
promising. The simplicity and accuracy of these models make them particularly attrac-
tive for engineering applications. However, extensive successful industrial applications
are essential if any of these models are to become accepted in the industry.
6.9 Model Deployment

One of the impediments to realizing the full value of an applied thermodynamic
model is the speed and cost associated with its deployment. For a published model
to gain acceptance by practitioners, either industrial or academic, it must be used
and validated in actual applications. The cost of deploying a model includes cod-
ing, testing, validation of results, model application, and continued maintenance. If
a new model can be published in such a way as to facilitate its deployment in in-
house program or commercial process simulators, then it has a better chance of
being used and accepted. This is a difficult and complex subject. Efforts in the past
include model validation initiative35and CAPE-OPEN.36The model validation and
deployment should work in concert to ensure that the model being deployed is
sound from the point of view of thermodynamic consistency, reasonable behavior
at extreme conditions, and ability to return derived or related properties (e.g., heat
capacity or enthalpy departure). CAPE-OPEN defined a standard interface for
physical property calculations. Major process simulators support this interface stan-
dard. A property model that has been developed conforming to this standard should
be able to be used directly in these simulators. However, the standard, as currently
published, is not sufficiently precise and leaves many implementation details to
interpretation, causing potential interoperability problems (e.g., how to handle
properties of a physically non-existent phase). Nevertheless, significant progress
has been made during the past few years in reducing ambiguity, interoperability,
and performance problems in the early versions of the standard. An organization
called CO-LaN has been formed (http://www.colan.org) with the responsibility for
maintaining, promoting, and extending the standard.
7 Conclusion
Applied thermodynamics has a major and essential contributing role in the 6 trillion
dollar process industry. It is a key enabling science and technology for chemists and
engineers in their pursuit of new and better products and processes in the industry.
Continuing vigorous innovations and advances in applied thermodynamics are
178 Chapter 15
essential to address many unmet or emerging industry needs. We have shared our
personal perspectives in the applications and needs of the industry in applied ther-
modynamics from the privileged position of an industry software provider. Our
views, which are inherently limited by our personal biases, have benefited from
close collaboration with our colleagues and partners in industry and academia.
References
1. C.-C. Chen and P. M. Mathias, A.Z.Ch.E. J., 2002,48, 194.
2. K. C. Seavey, N. P. Khare, Y. A. Liu, T. N. Williams and C.-C. Chen, A New Phase-
Equilibrium Model for Simulating Nylon-6 Polymerization Processes, Znd. Eng. Chem.
Res., ASAP web release date: July 18, 2003.
3. C.-C. Chen, Fluid Phase Equilibr., 1993,83, 301.
4. T. C. Frank, J. R. Downey and S. K. Gupta, Chemical Engineering Progress, December,
1999, 41.
5. C. M. Hansen, Hansen Solubility Parameters: A User’s Handbook, CRC Press, 1999.
6. P. Kolar, J.-W. Shen, A. Tsuboi and T. Ishikawa, Fluid Phase Equilibr., 2002, 194-197,
771.
7. R. C. Reid, J. M. Prausnitz, and B. E. Poling, The Properties of Gases and Liquids, 4th
edn., McGraw-Hill, New York, 1987.
8. B. E. Poling, J. M. Prausnitz, and J. P. O’Connell, The Properties of Gases and Liquids,
5th edn., McGraw-Hill, New York, 2000.
9. E. C. Carlson, Chem. Eng. Prog, October. 1996, 35.
10. T. Holderbaum and J. Gmehling, Fluid Phase Equilibr., 1991, 70, 25 1.
11. S. H. Wong and S. I. Sandler, A.Z.Ch.E. J., 1992, 38, 671.
12. C. H. Twu, W. D. Sim and V. Tassone, Znd. Eng. Chem. Res., 2002,41,931.
13. C. H. Twu, W. D. Sim and V. Tassone, Chem. Eng. Prog., Nov. 2002,58.
14. J. Ahlers and J. Gmehling, Fluid Phase Equilibr., 2001, 191, 177.
15. A. Klamt and F.Eckert, Fluid Phase Equilibr., 2000,43, 172.
16. S.-T. Lin and S. I. Sandler, Znd. Eng. Chem. Res., 2002,41, 899.
17. C.-C. Chen, unpublished.
18. C.-C. Chen, and L. B. Evans, A.Z.Ch.E. J., 1986,32, 444.
19. W. Fiirst and H. Renon, A.Z.Ch.E. J., 1993,39, 335.
20. J. Wu and J. M. Prausnitz, Ind. Eng. Chem. Res., 1998, 37, 1634.
21. J. A.Myers, S.I. Sandler and R. H. Wood, Znd. Eng. Chem. Res., 2002,41, 3282.
22. J. Gross and G. Sadowski, Znd. Eng. Chem. Res., 2002,41, 1084.
23 M. L. Michelsen, Fluid Phase Equilibl:, 1982, 9, 1.
24 M. L. Michelsen, Fluid Phase Equilibr., 1982, 9, 21.
25 J. F. Boston, and H. I. Britt, Comp. Chem. Eng., 1978, 2, 109.
26 J. Z. Hua, R. W. Maier, S. R. Tessier, J. F. Brennecke and M. A. Stadtherr, Fluid Phase
Equilibr., 1999, 158, 607.
27 A.C. Sun and W. D. Seider, Fluid Phase Equilibr., 1995, 103, 213.
28. V. Gopal and L. T. Biegler, A.1.Ch.E. J., 1999, 45, 1535.
29. W. D. Seider and S. Widagdo, Fluid Phase Equilibr., 1996, 123, 283.
30. S. Behme, G. Sadowski, Y. Song and C.-C. Chen, A.l.Ch.E. J., 2003,49,258.
31. P. M. Mathias, J. F. Boston and S. Watanasiri, A.Z.Ch.E. J., 1984, 30, 182.
32. L. T. Novak, Znd. Eng. Chem. Res., 2003,42, 1824.
33. Y. Song, P. M. Mathias, D. A.Tremblay and C.-C. Chen, Znd. Eng. Chem. Res., 2003,42,
2415.
34. J. Jiang and S. I. Sandler, Ind. Eng. Chem. Res., 2003,42,6267.
35. R. Gani and J. P. O’Connell, Comp. Chem. Eng., 2001,25,3.
36. J-P. Belaud, B. Braunschweig, M. Halloran, K. Irons, D. Pinol, New Generation
Simulation Environment: Technical Vision of the CAPE-OPEN Standard, paper presented
at the AIChE annual meeting, Reno, Nevada, 2001.
CHAPTER 16
Thermodynamics of New
Materials
M. SCHROEDER AND M. MARTIN
1 Introduction
Progress in all fields of our technological world depends on the development and
optimization of materials. This applies to metallic materials, ceramics and polymers
as well. Research and applications related to the thermodynamics and transport
processes of new materials is therefore a vast field that cannot be covered in a sin-
gle article. As a consequence, the focus of this article will be on ceramic materials,
and within this broad field we will focus on oxide ceramics. In the following two
sections, ceramic membranes that are mixed ionic and electronic conductors and
oxygen ion conductors will be discussed. Both classes of materials exhibit interest-
ing properties and can be used for important technical applications, such as oxygen
separation and catalytic partial oxidation of hydrocarbons or conversion of chemical
energy into electrical energy in solid oxide fuel cells.
2 Ceramic Membranes for Oxygen Separation and

Catalytic Partial Oxidation of Hydrocarbons
Today, huge amounts of pure oxygen are required for the production of chemicals,
petrochemicals and primary metals, as well as for medical, military and other appli-
cations. The yearly consumption of oxygen adds up to more than 35 million tons
worldwide.' Although ubiquitously available in air, the production of oxygen is dom-
inated by cost-intensive processes, such as cryogenic distillation and pressure swing
a d ~ o r p t i o nA. ~new
~ ~ technology to generate oxygen in a more economical way would
seem to be very attractive for industrial gas-to-liquid (GLT) applications. Recently, it
was demonstrated that ceramic membranes that selectively permeate oxygen at high
temperatures (>700 "C) could be used for continuous oxygen separation from air.4.5
As various industrial oxidation processes are carried out in the same temperature
range, an integration of oxygen generation and chemical oxidation into a single-step
process would lead to a drastic reduction of investment costs in GLT processes.6The
Thermodynamicsof New Materials 181
most prominent example for such an integrated process is the conversion of natural
gas to synthesis gas by catalytic partial oxidation, which could be carried out directly
at the membrane surface. In the subsequent sections, some essential properties of
high-temperature ceramic membranes and problems related to their application in
oxygen generators and partial oxidation reactors will be discussed.
2.1 Transport and Defect Thermodynamics in Membrane Materials

In contrast to conventional permeation membranes, where separation usually takes
place by pore diffusion?J oxygen transport in high-temperature ceramic membranes
occurs by an electrochemical process. Provided the membrane is dense enough to
prevent any leakage of other feed gas components, an infinite oxygen separation fac-
tor can be achieved. The permeation is driven by a difference of the oxygen partial
pressures in the gas phases in contact with the membrane surfaces. According to
Figure 1, the overall permeation process may be divided into three steps: (a) reduc-
tive incorporation of oxygen from gas phase (') with the larger oxygen partial pres-
sure, whereby oxygen ions are formed in the membrane; (b) transport of oxygen ions
through the membrane bulk and (c) oxidation of oxygen ions and release to gas
phase (") with the lower oxygen partial pressure. To maintain local charge neutrality,
a simultaneous internal flow of electronic charge carriers (electrons or electron
holes) is necessary.
It is obvious that a highly permeable membrane material must exhibit large con-
ductivies for both ionic and electronic charge carriers. Partial conductivities of vari-
ous, so-called mixed ionic electronic conductors (MIEC), as calculated or directly
obtained from Refs. 9-21, are presented in Figure 2.
Gas phase ' Membrane Gas phase "
'2)
Carrier gas
0 2 P" ( 0 2 )
Figure 1 Working principle of a mixed-conducting oxide membrane

182 Chapter 16
Ce0.8Gdo.202-8 L%.6Sr0.4C00.8Fe0.203-~~
0
9
L%.5sr0.5c003-8
YSZ (10 d o Ti02)g
:1E-5
c
.-
:1E-6 d,
5
1E-7 *
L%.8sr0.2cro3-8 2 1E-8
I 1 1 1 1 1 1 1 11111111, 1 1 1 1 1 1 '1 1111111 111,11111 1111111 1~-9
High-performance materials are expected in the upper right section of the plot,
where mainly complex metal oxides of the general formula A, --x A: B,-, Bi,03-6, with
the dopant elements A' and B', are abundant. The B-sublattice contains transition
metal elements, which are a necessary prerequisite for electronic conduction as they
are multivalent and provide electronic defects as mobile charge carriers. For this rea-
son, MIEC oxides are predominantly electronic conductors; however, they can also
exhibit extraordinarily large oxide-ion conductivities that are comparable to, or even
exceed, those of the best-known solid oxide electrolytes. This is due to the perovskite
structure of these oxides, which is known to accommodate large quantities of unoccu-
pied oxygen lattice sites. These structural defects, known as oxygen vacancies, enable
fast oxide-ion migration. In contrast, the number of cation defects is generally small.
The deviation from stoichiometry, 6, is therefore a direct measure of the fraction of
oxygen vacancies. Controlling 6 and the average transition metal valence provides a
means to tailor the transport properties of the material and can be achieved by doping
with aliovalent ions. This will be illustrated for the case of undoped, stoichiometric
LaBO,: substitution of the acceptor ion Sr2+for La3+ requires charge compensation
either by formation of B4+or by loss of oxygen. In general, both processes take place
and 6 depends on both the dopant fraction x and the fraction v of B4+defects:
Thermodynamics of New Materials 183
On the other hand, the oxide can exchange oxygen with the gas phase. The exchange
redox reaction is an equilibrium reaction:
1
2'2 (gas) -k '(Oxide) + 2B&ide) o:6xide) + 2B:&ide)
From Equation (2), it is obvious that 6varies with the oxygen partial pressure of the
gas phase, p ( 0 , ) . In terms of the dependence of 6 on p(O,), different regions may be
d i s t i n g ~ i s h e d :at
~ ~high
. ~ ~p(O,), charge compensation of the acceptor dopant occurs
mainly by the formation of B4+ (region I). From Equation (l),it follows that v = x
and 6=0.At medium p(O,), both B4+ and vacancy defects provide charge compen-
sation (region 11) and at low p(O,), mainly vacancies are formed (v = 0 and 6 = x/2,
region 111). Thus, by going from high to low p(O,), the conductivity is expected to
change from predominantly electronic to predominantly ionic and only in region I1
the oxide should be a mixed conductor. Finally, at very low p(O,), B3+ gradually
reduces to B2+ and 6 increases further (region IV). Figure 3 illustrates the depend-
ence of 6 on p ( 0 , ) for acceptor-doped MIEC oxides with the transition metals cobalt
and iron on B - s i t e ~ .The ~ ~ cobaltate
?~~ exists only in a narrow partial pressure range,
which belongs to region 11, and is an excellent MIEC material. As expected from
equation (l), the deviation from stoichiometry increases with the dopant fraction x.
For the iron-containing oxide, all four regions are observed.
By means of defect model calculations, a quantitative description of the depend-
ence of 6 on p ( 0 , ) was obtained for this oxide, and the defect concentrations were
calculated.26 In Figure 4, the variation of the defect concentrations with p ( 0 , ) is
(.o
cr)
2,8 - 0 La,$jro.lFe03-p 1273 K
-Calculated from defect model
La,.,Sr,CoO3-~ 1173 K
2,7 - 0 x=0.2
A x=0.3
0 x=o.5
x=0.7
Figure 3 Deviation from stoichiometry as a function of oxygen partial pressure for

La,,Sr,, FeO,-, (solid line: calculated from defect model), and as a function of
p ( 0 , ) and compositionfor La,-x SrxCoO,-6 (solid lines are a guide to the eye). Data
were taken and adapted from refs. 24,25
184 Chapter 16
0 1 I I 1
re" Fe2+ 1
Figure 4 Oxygen partial pressure dependence of the defect concentrations of

Lao,,Sro,,Fe0,-6 at I273 K,as obtained from model calculations6
illustrated. The model calculations revealed that the formation of B2+ by the dispro-
portionation reaction
cannot be neglected, and that significant amounts of B2+ and B4+ defects exist even
in region 111, providing for mixed conduction. Measurements of the total electrical
cond~ctivity'~ support these results: the conductivity is largest at high p ( 0 , ) in
region I and exhibits a minimum at a p ( 0 , ) belonging to region 111, but even at this
minimum it is much higher than the ionic conductivity.
2.2 Oxygen Permeation

As outlined in the previous section, oxygen permeation through an MIEC membrane
relies on the migration of ionic and electronic defects. The transport theory of
Wagner27may be adopted to express the oxygen flux j(0,) through a membrane
exposed to a gradient of oxygen chemical potential Vp(O,),
where q and 0,are the partial ionic and electronic conductivities, respectively.
Equation (4) is valid only if the overall permeation process is limited by the bulk trans-
port of charge carriers, which implies that the oxygen exchange reaction with the gas
phase at the membrane surface (see Equation (2)) must be fast enough to maintain
equilibrium. If predominant electronic conduction is assumed, the conductivity-
dependent term in Equation (4) is approximately equal to 4. Furthermore, if a vacancy
mechanism is assumed for the bulk transport of oxygen ions, is proportional to the
vacancy mobility u, and to the molar fraction of vacancies, which is given by 6:
Themdynumics of New Materials 185
where Vm is the molar volume of the oxide. In MIEC oxides, the vacancy mobility
is often nearly independent of oxygen partial pressure, but the deviation from
stoichiometry varies with p ( 0 , ) . The oxygen flux is then obtained by the combina-
tion of Equations (4) and (5) and subsequent integration over the membrane
thickness L:
Provided the dependence of 6 on p ( 0 , ) is known, the permeation flux can be calcu-

lated as a function of the partial pressures p ' ( 0 , ) and ~ " ( 0 , )on the oxygen-rich and
the oxygen-lean side of the membrane, respectively. By evaluating the nonstoi-
chiometry data of L~,_,S~,COO,-~ given in Figure 3, it was found that 6 is approxi-
mately proportional to ~(0,)"s. Integration of Equation (6) then yields:
with 4 being the deviation from stoichiometry at p(0,) = 1 bar. n6 is dependent on

temperature and the molar fraction of the dopant x and assumes negative values in
the range - 0 . 2 4 1 n 6 1 -0.07. Recently, the oxygen permeation through
La, -,S~,COO,-~was investigated. Selected results are presented in Figure 5.
19728
Linear plots are obtained if Equation (7) is applied; however, only for membranes
whose surfaces had been modified to enhance the kinetics of oxygen exchange with
the gas phase, values of n were found to be negative and consistent with the n6
obtained from the nonstoichiometry data. For membranes without surface modifica-
tion, n>O was found in all cases, which indicates that Equation (7) is not applicable.
Apparently, sluggish surface exchange kinetics limit the permeation process through
La, -xSr,Co0,-6. Surface exchange limited permeation has also been observed for
1 I I I 1
Figure 5 Oxygen permeation flux through membranes of uncoated Lao~,Sro,,CoO,-sat 11 73

K and Luo,,Sro,,Co03~s coated with SrCoo,,Feo,,O,-s at 1183 K. The coating was
deposited on the feed side. Data were taken from re5 28
186 Chapter 16
other MIEC oxides. A detailed analysis of the influence of the surface exchange
kinetics on the permeation is beyond the scope of this paper, but can be found in sev-
eral excellent discourse^.^^-^^ However, it is stressed that data on the surface
exchange reaction kinetics are important for the design of permeation membranes.
Equation (7) implies that reducing the membrane thickness L is one way to increase
the permeation flux. This is because the bulk resistance of oxygen transport
decreases with decreasing thickness. On the other hand, the resistance due to surface
exchange does not depend on L, and below a critical membrane thickness, no gain
in permeation flux can be obtained by further reducing the membrane thickness. The
critical thickness depends on various factors (e.g. oxide composition and tempera-
ture) and ranges from several 10 pm up to several 100 pm.
2.3 Stability of Membrane Materials

In an operating partial oxidation reactor, the membrane material must tolerate harsh
conditions over long time periods. The large gradient of oxygen chemical potential
across the membrane causes a gradient of oxygen vacancy concentration. The
resulting spatial variation in lattice expansion may cause significant tress,^^.^^ and
mechanical failure after short times of operation has been observed particularly with
cobalt-containing materials.34 Furthermore, the reaction side of the membrane is
exposed to reducing gases (CO and H,) at very low oxygen partial pressures
( 10-18-10-20bar). While A-site acceptor doping and high reducibility of the B-site
transition metal ions are beneficial for the oxygen permeation performance, it
diminishes the chemical stability of the membrane. The critical p ( 0 , ) below which
reductive decomposition into a nonconductive multiphase mixture takes place is
approx. bar for undoped LaFe03-6 and bar for Even higher
critical pressures have been reported for the respective strontium-doped oxides.36
Reductive decomposition causes deterioration of the membrane permeability and is
therefore a critical problem. The partial replacement of B-site transition metals by
less reducible elements (e.g. gallium or chromium) improves the membrane stabil-
ity to some e ~ t e n t ~ ~
however,
. ~ * ; at the cost of lower permeability. Recently, a new
concept to stabilize permeation membranes has been i n t r o d ~ c e d . ~It ~relies. ~ ~ on
bilayer materials that consist of a substrate layer of an MIEC material, e.g.
Lal-xSrxCo03-6, and a thin protective coating on the low-p(0,) side. The coating
material is required to be stable towards reduction and should exhibit high oxide-ion
conductivity but low electronic conductivity. Samarium-doped ceria meets these cri-
teria. Figure 6 illustrates the effect of the coating. Because of its low electronic con-
ductivity, the coating has a higher resistance to oxygen permeation than the
substrate. If the membrane is exposed to an oxygen chemical potential gradient, a
substantial drop of oxygen potential occurs in the coating. This causes the oxygen
chemical potential ~ ~ ( at 0 the
, ) interface of substrate and coating to rise to above the
critical potential p(O,>,,, of the substrate. In this way, the reduction of the substrate
can be avoided.
Bilayer membranes have been successfully operated under partial oxidation con-
ditions for several hundred hours without any significant deterioration of the perme-
ation
MIEC MIE3C Coating
Uncoated membrane Bilayer membrane
Figure 6 Function of a bilayer membrane subjected to a large gradient of oxygen chemical

potential. The protective coating diminishes the gradient acmss the delicate MIEC
substrate, which would decompose below p(02)crir.
2.4 Conclusion
In the previous sections, it has been demonstrated that, in order to improve the prop-
erties of known MIEC membrane materials and to develop new materials, a funda-
mental understanding of their defect chemistry and the transport of defects is crucial.
This is achieved through interpretation of experimental thermochemical and trans-
port data by means of defect models.
While the introduction of a membrane separation step in catalytic partial oxida-
tion processes is most attractive for technical and economical reasons, various prob-
lems related to the material and reactor design have yet to be solved. The mechanical
and chemical stability of highly permeable MIEC oxides needs to be improved to
ensure long-term membrane stability under operation conditions and during thermal
cycling. But other issues also need to be addressed: Perovskite membrane surfaces
often favour deep oxidation of hydrocarbon^.^^ Therefore, partial oxidation catalysts
have to be used to improve the selectivity for partial oxidation. If the membrane sur-
face serves as support for these catalysts, unwanted reactions between catalyst and
membrane oxide must be avoided. The oxidation process may require additional
components, such as H20, to be present in the hydrocarbon feed stream, which also
might react with the membrane and alter its properties. On the other hand, the oxy-
gen fluxes obtained with some known MIEC membrane materials already meet the
minimum requirements for economic operation of a membrane reactor,43 which
raises hope of a breakthrough in this technology in near future. It seems that high-
temperature membrane reactor technology will continue to provide an ample field
for both applied and fundamental research,
3 Solid Oxide Fuel Cells

3.1 Fuel Cells
In a fuel cell, an electrolyte is used to divide the strongly exothermic chemical reac-
tion H2+Y2O2 t)H,O into two electrochemical reactions: a reduction and an oxida-
tion reaction. In this way, the conversion of the reaction Gibbs energy, A,G,
into electrical energy is possible, and the only reaction product is water (if pure
188 Chapter I6
hydrogen is available as fuel). The voltage, E, of this battery is given by
E= -A,GInF (where n is the number of electrons and F is the Faraday constant).The
first fuel cell was described already in 1839 by Groveu using an aqueous acid as
electrolyte. Nowadays, several fuel cell types exist for various application^,^^ and
numerous efforts are being undertaken in basic research and development with the
aim to optimize a sustainable energy conversion method.
As shown in Figure 7, there are essentially two possibilities for the operation of a
fuel cell. If protons, H+, are mobile in the electrolyte, the oxidation and reduction
reactions are H, +2H' +2e- and 2H++2e- +%O, +H,O, respectively (see Figure
7(a)). If, on the other hand, an electrolyte is used where oxygen ions are mobile, the
corresponding oxidation and reduction reactions are 02-+H, +H,O +2e- and
$,+2e- -+0,-(see Figure 7(b)).
As can be seen from Figure 7, it is necessary for the optimum operation of a fuel
cell that the two half cell reactions are separated by a membrane that does not con-
duct electrons, but only ions.
The most important fuel cells that are in use nowadays are the polymer electrolyte
membrane fuel cell (PEMFC), the molten carbonate fuel cell (MCFC), and the solid
oxide fuel cell (SOFC).45In a PEMFC, the electrolyte is a polymer membrane that
conducts protons, H+, in an MCFC the electrolyte is a carbonate melt in which oxy-
gen is conducted in the form of carbonate ions, Cog-, and in an SOFC the elec-
trolyte is a solid oxide that conducts oxygen ions, 0,-. While a PEMFC can be
operated at low temperatures of about 80 "C, an MCFC works at intermediate tem-
peratures of about 65OoC, and an SOFC needs relatively high temperatures of
800-1000 "C (see next sections).
In both PEMFCs and SOFCs, there is a strong need to improve the electrolyte
materials and to find substitutes for the traditional electrolytes, which are Nafion for
the PEMFC and yttria-stabilized zirconia (YSZ) for the SOFC. In this paper, we will
focus on the discussion of new developments in the field of SOFC electrolytes. For
recent developments concerning other fuel cell types, the reader is referred to review
articles and books.4547
a b
Figure 7 Schematic set-up of a fuel cell. ( a ) proton-conducting electrolyte, and (b) oxygen
ion-conducting electrolyte
3.2 Electrolytes for Solid Oxide Fuel Cells
The electrolyte in an SOFC is a solid, crystalline oxide in which conduction of oxy-
gen ions is possible by means of oxygen vacancies, V,. The site exchange between
oxygen ions and oxygen vacancies is a thermally activated process (the typical acti-
vation energy in YSZ48is 80-90 kJ mol-’). Since a sufficiently high electrical con-
ductivity is required to minimize the internal resistance of the electrolyte, relatively
high operating temperatures are needed in an SOFC.
The traditional oxygen ion conductor is yttria-stabilized zirconia (YSZ). Doping
ZrO, with Y203 has two effects: (i) It stabilizes a cubic structure, which is stable
from room temperature to elevated temperatures. This is very important for the pos-
sibility to change the temperature of an SOFC without destroying the electrolyte as
a result of phase transformations. (ii) Oxygen vacancies are produced, which are
necessary for a high oxygen ion conductivity. As can be seen in Equation (8), dop-
ing with Y203results in two yttrium ions occupying zirconium sites, Y;:, three oxy-
gen ions occupying oxygen sites, 0;-,and one oxygen vacancy, V,. The formation
of the oxygen vacancy is necessary to preserve the lattice structure of zirconia with
two oxygen sites per zirconium site.
Y20, + 2Y;: + 30;- + V,
As a consequence of Equation (8), the concentration of oxygen vacancies is fixed by
the dopant concentration, cvo=cY,,,. Therefore, in a typical YSZ electrolyte, e.g.
(Zr02)o~9(Y203)o~l, the oxygen vacancy fraction is as high as 10 mol%. Due to this
high concentration and the high vacancy diffusion coefficient, D,, a high oxygen
ion conductivity, a(02-)=4F2DVocvo/RT(Nernst-Einstein relation), is possible. A
typical value49for (Zr02)o.9(Y203)o,l at 1000 “C is a(02-)= 10-’ S cm-l.
It is interesting to note that the first use of doped zirconia and its high oxygen ion
conductivity goes back to Nernst,” who used this material already in 1899 in his
Nernst glower ( a lamp where the light-emitting filament is an oxygen ion conduc-
tor). It took, however, more than 40 years till the conduction mechanism was
explained in detail by Wagner.51
As mentioned above, a perfect electrolyte has a vanishing electronic conductivity.
In reality, however, an oxygen ion-conducting oxide always contains electrons, e’,
and electron holes, h’, as minority charge carriers. Their contribution to the total con-
ductivity, o,,= , o(02-) + fie’) +o(h’), is determined by the electronic transference
number, tel=(o(e’)+ o(h’))/qot. A good oxygen ion conductor is characterized by
tel< lop3.To determine the so-called ionic domain, one has to consider that the con-
centrations of electrons and electron holes depend on the external oxygen partial
pressure p ( 0 , ) . This can be observed immediately from the incorporation reaction
for oxygen from the gas phase (oxygen is reduced and occupies an oxygen vacancy)
1
- 0,(g)
2
+ 2e’ + V, H 06- (9)
and the electronic equilibrium
nil ++ e’ + h’
190 Chapter 16
1 I I I I 1 I I I I
-6
-8
.I_ - - -,-T - - _ _ _-_
ionic domain
.- I
I I I I I I I I
-10
-50 -40 -30 -20 -10 0 10 20 30 40
log(p(0,) 1bar)
Figure 8 Conductivities of oxygen ions, electrons and electron holes in Zro,9Yo,101,95

as a
function of oxygen partial pressure at diflerent temperatures (adopted from re$ 52).
The ionic domains and typical p(0,)-ranges offuel cell operation are also shown
Assuming equilibrium for both reactions and considering that the concentration of
oxygen vacancies is fixed by the high Y203fraction, cvo= cy203,we obtain:
The concentration of electrons increases with decreasing oxygen partial pressure,

while the concentration of electron holes decreases. The above results are summa-
rized in Figure 8, showing the partial conductivities of oxygen ions, electrons and
electron holes in YSZ as a function of oxygen partial pressure.52
As can be seen in Figure 8, the ionic domain (tel<10-3) extends over more than
20 orders of magnitude. The p(0,)-range of an SOFC corresponding to a cell volt-
age of 0.8 V (air at the cathode and a H,/H,O-mixture at the anode, using the Nernst
equation E=(R~/4~ln(p(0,)""fh/po""d)), is well within the ionic domain.
In summary, the high operating temperatures of 900-1000 "C yield sufficiently
high ionic conductivity of the electrolyte and small electronic transference. On the
other hand, lower operating temperatures are desired, (i) for economic reasons, and
(ii) to minimize problems that are due to the combination of different materials in a
fuel cell. For example, problems arise from interdiffusion and chemical reactions49
between the electrodes and the electrolyte (typical electrodes in an SOFC consist of
strontium-doped lanthanum manganate, (La,Sr)MnO,_&, at the cathode and a mix-
ture of fine-grained nickel and YSZ at the anode). To obtain large enough voltages
from an SOFC, cells have to be combined to a stack. In this stack, the individual cells
are separated by the interconnector that acts as a current collector and separates the
0,- and H,-containing compartments of adjacent cells. At high temperatures, there
are severe problems (sealing, reactivity, thermal mismatch, evaporation) using
ceramic interconnectors (such as (La,Ca)CrO,), which could be avoided by going to
intermediate temperatures of about 700 "C where conventional metallic components
could be used as interconnector materials.
Temperature ("C)
900800 700 600 500 400 300
I I I
150 pm
15 Pn
1.5 pm
0.15 p
03 1 1 2 1,4 1,6 1,8 2
1000A'(K1)
Figure 9 Oxygen ion conductivities of three electrolytes as afunction of reciprocal tempera-

ture (adopted from re& 45)
However, to decrease the operating temperature of an SOFC from high tempera-

tures of 900-1000 "C to intermediate temperatures (IT) of about 700 "C, it is neces-
sary to decrease the cell resistance. This is possible by decreasing the thickness, L,
of the electrolyte and/or by new electrolyte materials with higher oxygen ion con-
ductivities. A characteristic number is the area r e ~ i s t a n c eR,=LIa,
~~, which should
be smaller than 0.15 i2 cm2. Figure 9 shows the temperature dependence of the oxy-
gen ion conductivities of YSZ and two new, recently discovered oxygen ion con-
ductors, doped lanthanum gallate (LSGM) and gadolinium-doped ceria (GDC)45,
both of which exhibit higher conductivities than YSZ.
The electrolyte thickness on the right-hand side of the diagram was calculated
from the oxygen conductivity on the left-hand side of the diagram for a desired area
resistance R,=O.15 Rcm2. To remain below this value, the thickness of aYSZ elec-
trolyte in an IT-SOFC at 700°C must be smaller than 15 pm.With such a small
thickness, a self-supported electrolyte layer is no longer possible and one must go to
anode-or cathode-supported cells. With the new materials, LSGM and GDC, thicker
electrolytes can be used even at 500-700 "C. The ionic domain of LSGM53is com-
parable with YSZ. In contrast, the ionic window of GDC is much smaller (due to
reduction of Ce4+ to Ce3+ at low oxygen partial pressures) causing problems in its
use as an SOFC e l e c t r ~ l y t eTherefore,
.~~ in the next section, we will discuss only the
new oxygen ion conductor LSGM in more detail.
3.3 Lanthanum Gallates

In the search for new oxygen ion conductors to be used as electrolytes in SOFCs,
oxides have been extensively studied in recent years, which crystallize in the
perovskite structure and in structures derived therefrom (e.g. the brownmillerite
192 Chapter 16
structure). Doped lanthanum gallate, La,-,+4xGa,-$,0,-,, has proved to be a par-
ticularly suitable candidate.5456In this material, the oxygen ion conductivity can be
clearly increased in comparison to the classic SOFC electrolyte, YSZ, by doping at
A- and B-sites. Testing comprised different compositions with partial substitution of
La by Sr, Ca or Ba and of Ga by Mg, In or Al.54Other tested gallates had e.g. the
composition (L~,&~~I)o~8Sro~2G~~,Mgo~,02~8 (Ln=Nd, Sm, Gd, Y, Yb).57The high-
est conductivities are achieved with Sr at A- and Mg at B-sites. Lanthanum gallate
of the composition L~~,Sr0~,G~,,Mg,,O2,, shows conductivities between lop2and
10-1 R-' cm-' at temperatures between 500 and 800 "C (see Figure 9).
The Gibbs energy of formation of LaGaO, has been determined using galvanic
~ e l l s ~and * -vapour
~ ~ pressure measurements combined with a defect chemical analy-
sis60,61;the enthalpy of formation was determined by calorimetry.62
The phase diagram of the binary system La,O,-Ga,O, shows only two com-
p o u n d ~ , ~LaGaO,,
, which is the basis for LSGM, and La4Ga,0,. The stability field
of Sr- orland Mg-doped lanthanum gallate was studied in several investiga-
t i o n ~However,
. ~ ~ ~ phase diagram studies are rendered difficult due to the small
cation diffusion coefficients in these oxides and the necessary long annealing times
for complete equilibration. Recently, the stability field of LSGM at 1400°C has
been determined: (i) the oxides were prepared using a modified Pechini method,68
where an aqueous solution of the metal nitrates, citric acid and ethylenglycole
yields a polyester gel, in which the cations are statistically distributed in chelat
complexes. (ii) After calcination, equilibration of the samples was ensured by long-
term annealing and by considering diffusion of the cations,69which are the slowest
moving species. Solid phase equilibria in the quasi-quaternary system
La20,-Ga,O,-MgO-SrO- near the solubility lobe of the (La,Sr)(Ga,Mg)O, -6 per-
ovskite phase were determined by XRD, SEM/EDX and electron microprobe
analysis (EPMA) of annealed and quenched samples. The resulting phase diagram
in Figure 10 shows the stability field of LSGM and the adjacent phase field^.^^,^^
The solubility of strontium in LaGaO, is only about 2 at% and much smaller than
reported in the other studies. Co-doping with strontium and magnesium increases
the Sr-solubility and yields a stability field, which also contains the composition
with the highest ionic conductivity, L%.9Sr0.,G%,9 Mg0.1O2.9.
It must be emphasized that this equilibrium phase diagram was measured at a
temperature of 1400 "C,which is much higher than the operating temperature of an
SOFC. The phase diagram and the solubilities of Sr and Mg in LSGM at lower tem-
peratures are not known. Nevertheless, LSGM prepared at high temperatures is
expected to be also stable at lower temperatures due to the very small cation mobil-
ities, by which LSGM is kinetically stabilized.
Cation self-diffusion coefficients of La, Sr and Mg in LSGM have been measured
recently for temperatures between 900 and 1400°C using stable isotopes and
Secondary Ion Mass Spectrometry (SIMS).72At 900 "C, the cation diffusion coeffi-
cients are 10 orders of magnitude smaller than the oxygen diffusion coefficients.
Despite these small values, cation diffusion is also important in an oxygen ion
conductor, since there are several processes that are determined by the slowest mov-
ing species, such as sintering, kinetic d e m i ~ i n g(due
~ ~ to different cation diffusion
“LaMgO2.5’’
0.6 A
0.0
0.0 0.2 0.4 0.6 0.8
LaGaO, x(W “SrGaO,,“
Figure 10 Section of the phase diagram of the La,O,-SrO-Ga,O,-MgO system at 1673 K.
Phase regions: ( I ) (La,Sr)(Ga,Mg)03-6perovskite
phase (P), MgGaLa,O,, MgO;
( 2 ) R MgGaLa307, MgO, SrLuGaO,; (3) R SrLaGaO, MgO; (4) P, MgO,
SrLaGaO, SrLaGa,07; and (5) R SrLaGaO, SrLaGa,O,
coefficients) or creep. The activation energy reported for creep of LSGM at temper-
atures between 1200 and 1300 “C is about 5 eV.74,75This value is in fair agreement
with the activation energy of the self-diffusion coefficients of the cations,72which
are the slowest moving species in LSGM and should determine the creep rate.
In further investigations on LSGM, the thermodynamic stability under operating
conditions76and the reactivity with electrode m a t e r i a l ~ ~were
~ . ~ *studied. Evidence of
evaporation of Ga under operating conditions at the anode side of the SOFC was
found in ref. 76 and confirmed by vapour pressure measurements and thermodynamic
calculations.70~71 However, recent studies show that gallium evaporation at the anode
can be suppressed successfully using a thin buffer layer consisting of samarium-
doped ~ e r i aWith
. ~ ~such a cell, a fairly high power density of 700 mWcm-2 could
be obtained at 800 “C, indicating that an SOFC based on an LSGM electrolyte can
operate at intermediate temperatures and would be competitive with an SOFC based
on a YSZ electrolyte.
4 Conclusion
Lanthanum gallates doped with strontium and magnesium (LSGM) are promising
new oxygen ion conductors, which can be used as electrolyte material in an SOFC
and may substitute the traditional electrolyte YSZ in IT-SOFCs. Further research is
required (i) for a better understanding of the basic physical and chemical properties
of these materials and their long-term behaviour as SOFC component, and (ii) for
the search for new and better oxygen ion conductors.
194 Chapter I6
References
1. B. Suresh, S. Schlag and K. Yokose, Chemical Economics Handbook Report “Air
Separation Gases”, SRI Consulting, Menlo Park, California, 2002.
2. N. A. Downie, Industrial gases, Kluwer Academic, Dordrecht, 1997.
3. D. M. Ruthven, S. Farooq and K. S. Knaebel, Pressure Swing Adsorption, VCH
Publishers, New York, Weinheim, Cambridge, 1993.
4. Y. Teraoka, H.-M. Zhang, S. Furukawa and N. Yamazoe, Chem. Lett., 1985, 1743.
5. Y. Teraoka, T. Nobunaga and N. Yamazoe, Chem. Lett., 1988,503.
6. P. N. Dyer, R. E. Richards, S. L. Russek and D. M. Taylor, Solid State Zonics, 2000, 134,
21.
7. H. P. Hsieh, inorganic membranes for separation and reaction, in Membrane Science and
Technology Series, 3, Elsevier, Amsterdam, 1992.
8. Recent advances in gas separation by microporous ceramic membranes, in Membrane
Science and Technology Series, 6, N. K. Kanellopoulos (ed), Elsevier, Amsterdam, 2000.
9. W. L. Worrell, Solid State Zonics, 1992, 52, 147.
10. Wang, T. Kobayashi, M. Dokiya and T. Hashimoto, J. Electrochem. SOC.,2000,147,3606.
11. H. L. Tuller, Solid State Zonics, 1992, 52, 135.
12. B. C. H. Steele, S. Carter, J. Kajda, I. Kontoulis and J. Kilner, Proceedings of the Second
International Symposium on Solid Oxide Fuel Cells, Athens, Greece, F. Grosz,
P. Zegers, S. C. Singhal, and 0.Yamamoto, (eds), 1991, 517.
13. J. H. Kuo, H. U. Anderson and D. M. Sparlin, J. Sol. Stat. Chem., 1990,87, 55.
14. I. Yasuda and T. Hikita, J. Electrochem. SOC., 1993, 140, 1699.
15. N. Sakai, T. Horia, B. A. van Hassel, T. Kawada, H. Yokokawa and M. Dokiya,
Proceedings of the 18th Symposium on Solid State Zonics, Fukuoka, Japan, 1992, 93.
16. T.Ishigaki, S. Yamauchi, K. Kishio, J. Mizusaki and K. Fueki, J. Sol. Stat. Chem., 1988,
73, 179.
17. J. Mizusaki, T. Sasamoto, W. R. Cannon and H. K. Bowen, J. Am. Ceram. SOC., 1983,66,
247.
18. J. Mizusaki, J. Tabuchi, T. Matsuura, S. Yamauchi and K. Fueki, J. Electrochem. SOC.,
1989,136,2082.
19. K . Q. Huang, M. Schroeder and J. B. Goodenough, Proceedings of the International
Symposium “Solid State Ionic Devices”, The Electrochemical Society Proceedings
Series, PV 99-13, E. D. Wachsman, J. R. Akridge, M. Liu and N. Yamazoe (eds),
Pennington, NJ, 1999, 95.
20. J. W. Stevenson, T. R. Armstrong, L. R. Pederson and J. W. Weber, A. I. Ch. E. J .
21. B. Ma and U. Balachandran, Solid State Zonics, 1997, 100 53.
22 H. U. Anderson, Solid State Zonics, 1992, 52, 33.
23 J. Mizusaki, Solid State Zonics, 1992, 52, 79.
24. ,J. Mizusaki, J. Tabuchi, T. Matsuura, S. Yamauchi and K. Fueki, K., J. Electrochem. SOC.,
1989,136,2082.
25. ,J. Mizusaki, M. Joshihiro, S. Yamauchi and K. Fueki, Solid State Zonics, 1985,543,257.
26 B. A. van Hassel, T. Kawada, N. Sakai, H. Yokokawa, M. Dokiya and H. J. M.
Bouwmeester, Solid State Zonics, 1993, 66, 29.
27 C. Wagner and W. Schottky, Z. Phys. Chem., 1930, B11,25.
28 K. Huang and J. Goodenough, J. Electrochem. SOC.,2001,148, E203.
29 H. J. M. Bouwmeester and A. J. Burggraaf, in Fundamentals of inorganic membrane sci-
ence and technology, in Membrane Science and Technology Series, 4, A. J. Burggraaf, L.
Cot (eds) Elsevier, Amsterdam, 1996.
30 S. Kim, Y, L. Yang, A. J. Jacobson, and B. Abeles, Solid State Zonics, 1999, 121, 31.
Themdynamics of New Materials 195
31. S. Kim, S. Wang, X. Chen, Y. L. Yang, N. Wu, A. Ignatiev, A. J. Jacobson, and B. Abeles,
J. Electrochem. SOC., 2000, 147, 2398.
32. P. V. Hendriksen, P. H. Larsen, M. Mogensen, F. W. Poulsen and K. Wiik, Catal. Today,
2000, 56, 283.
33. A. Atkinson, A. Selcuk, Solid State Zonics, 2000, 134, 59.
34. U. Balachandran, J. T. Dusek, P. S. Maiya, B. Ma, R. L. Mieville, M. S. Kleefisch and C.
A. Udovich, Catal. Today, 1997,36, 265.
35. T. Nakamura, G. Petzow, and L. J. Gauckler, Matel: Res. Bull., 1979, 14,649.
36. B. C. H. Steele, Matel: Sci. Eng., 1992, B13, 79.
37. A. Atkinson, T. Ramos, Solid State Zonics, 2000, 129, 259.
38. S. Kim, A. J. Jacobson, B. Abeles, Materials Research Society Symposium Proceedings,
Vol548, (Solid State Zonics V), 1999, 563.
39. M. Schroeder, K. Q. Huang, H. Y. Yin and J. B. Goodenough, Proceedings of the
Znternational Symposium “Solid State Ionic Devices ”, in The Electrochemical Society
Proceedings Series, PV 99-13, E. D. Wachsman, J. R. Akridge, M. Liu and N. Yamazoe
(eds), Pennington, NJ, 1999, 121.
40. T. L. Clites, and E. D. Wachsman, Proceedings of the International Symposium “Solid
State Ionic Devices”, in The Electrochemical Society Proceedings Series, PV 99-13, E.
D. Wachsman, J. R. Akridge, M. Liu and N. Yamazoe (eds), Pennington, NJ, 1999, 110.
41. M. Schroeder, K. Q. Huang and J. B. Goodenough,Mass and Charge Transport in
Inorganic Materials. Fundamentals to Devices, in Advances in Science and Technology,
29. Proceedings of the International Conference on “Mass and Charge Transport in
Inorganic Materials - Fundamentals to Devices ”, P. Vincencini and V. Buscaglia (eds),
Jesolo di Lido (Venice), Italy, 2000, 1507.
42. R. Doshi, C. B. Alcock, N. Gunasekaran and J. J. Carberry, J. Catal., 1993,140,557.
43. R. Bredesen and J. Sogge, SINTEF Report, S96017, 1996.
44. W. R. Grove, Phil. Mag., Ser.3, 1839,14, 127.
45. B. C. H. Steele, J. Mat. Sci., 2001,36, 1053.
46. L. J. N. Blomen and M. N. Mugerwa (eds), Fuel Cell Systems, Plenum, New York, 1993.
47 * K. Kordesch and G. Simander, Fuel Cells and their Applications ,VCH, Weinheim, 1996.
48. P. S. Manning, J. D. Sirman, R. A. De Souza and J. A. Kilner, Solid State Zonics, 1997,100,
1.
49. A. Hammou and J. Guindet, in The CRC Handbook of Solid State Electrochemistry, P. J.
Gellings and H. J. M. Bouwmeester (eds), CRC Press, Boca Raton, 1997,407.
50. W. Nernst, Z. Elektrochem., 1899, 6, 41.
51. C. Wagner, Naturwissenschafien, 1943,31,265.
52. B. C. H. Steele, J. Power Sources, 1994, 49, 1.
53. H. Yokokawa, N. Sakai, T. Horita and K. Yamaji, Fuel Cells, 2001, 2, 117.
54. T. Ishihara, H. Matsuda and Y. Takita, J. Am. Chem. SOC., 1994, 116, 3801.
55. M. Feng and J. B. Goodenough, Eul: J. Solid State Inorg. Chem., 1994, t13,663.
56. K. Huang, M. Feng and J. B. Goodenough, J. Electrochem. SOC.,1997,144,3620.
57. T. Ishihara, H. Matsuda and Y. Takita, Solid State Ionics, 1995, 79, 147.
58. A. M. Azad, R. Sudha and 0. M. Sreedharan, Mat. Res. Bull., 1991,26,97.
59. K. T. Jacob, J. Matel: Res., 2000,15,2836.
60. W. Kuncewicz-Kupczyk, D. Kobertz, M. Miller, C. Chatillon, L. Singheiser and K.
Hilpert, J. Am. Ceram. SOC.,2002, 85, 2299.
61. M. Matraszek, L. Singheiser, D. Kobertz, K. Hilpert, M. Miller and M. Martin, Phys.
Chem. Chem. Phys., 2003,53042.
62. Y. Kanke, A. Navrotsky, J. Solid State Chem., 1998, 141,424.
63. M. Mizuno, T. Yamada and T. Ohtake, Ygyo Kyokaishi, 1985,93,295.
196 Chapter 16
64.P. Huang and A. Petric, J. Electrochem. SOC.,1996,143, 1644.
65. K. Huang, R. Tichy and J. B. Goodenough, J. Am. Ceram. SOC., 1998,81,2565.
66. P. Majewski, M. Rozumek and F. Aldinger, J. A. Ceram. Soc., 2001,84, 1093.
67. P. Majewski, M. Rozumek and F. Aldinger, Znt. J. Znorg. Mat., 2001, 3, 1343.
68. 0. Schulz and M. Martin, Solid State Zonics,2000, 135, 135.
69. 0. Schulz, M. Martin and S. Flege, in Solid Oxide Fuel Cells VZZZ (SOFC-VIZZ), PV 2003-
2007, S. C. Singhal and M. Dokiya (eds), The Electrochemical Society, Pennington 2003,
304.
70. A. Matraszek, D. Kobertz, L. Singheiser, K. Hilpert, W. Kuncewicz-Kupczyk, M. Miller,
0. Schulz and M. Martin, Materialwis, Werkst., 2002,33, 355.
71. A. Matraszek, L. Singheiser, D. Kobertz, K. Hilpert, M. Miller, 0. Schulz, and M. Martin,
Solid State Zonics,2004, 166, 343.
72. 0. Schulz, M. Martin, C. Argirusis and G. Borchardt, Phys. Chem. Chem. Phys., 2003,5,
2308.
73. M. Martin, Solid Stute Zonics,2000, 136137, 331.
74. J. Wolfenstine, Solid State Zonics, 1999, 126, 293.
75. J. Wolfenstine, P. Huang and A. Petric, J. Solid State Chem., 1999, 118, 257.
76. K. Yamaji, T. Horita, M. Ishikawa, N. Sakai and H. Yokokawa, Solid Stute Zonics, 1999,
121, 217.
77. T. Horita, K. Yamaji, N. Sakai, H. Yokokawa, A. Weber and E. Ivers-Tiffee, J.
Electrochem. SOC., 200 1, 148,A456.
78. A. Zhang, S. Ohara, H. Okawa, R. Maric andT. Fukui, SolidStateIonics, 2001,139, 145.
79. K. Huang, J. Wan and J. B. Goodenough, J. Mat. Sci., 2001,36, 1093.
CHAPTER 17
Thermodynamic Prediction of the

Formation and Composition
Ranges of Metastable Coating
Structures in PVD Processes
P. J. SPENCER
1 Introduction
A large amount of systematic development work on new high-performance mixed
carbide, nitride, boride and oxide coating materials produced by physical vapour
deposition (PVD) techniques is being carried out, because the mixed coatings often
display significantly improved properties (e.g. wear-, corrosion-, oxidation-resist-
ance or certain physical property behaviour) compared to the properties of the indi-
vidual constituents.
Experiments with the deposition of metastable, multicomponent coatings using
PVD methods have shown that the required coating compositions can be achieved
relatively easily. Furthermore, variation of the substrate temperature as well as the
coating composition sometimes allows the structure and hence the properties of the
deposited coating to be changed.
By carrying out thermodynamic calculations, it is possible to predict the condi-
tions for the appearance of different metastable phases during PVD coating
processes and thereby assist in the selection of coating parameters required to pro-
duce coatings with optimum desired properties.
The general principle relating to thermodynamic calculation of the metastable
phase ranges resulting from PVD of multicomponent coatings on a low-temperature
substrate has been propounded by Saunders and Miodownik' as follows:
If an alloy, that in equilibrium contains a multiphase structure, is co-deposited

at low enough temperatures, the surface mobility is insufficient for the break-
down of the initially fully intermixed depositing atoms, and the film is therefore
198 Chapter 17
constrained to be a single-phase structure ... Nucleation and subsequent growth
processes are controlled by the substrate temperature, and therefore, if the film
is constrained to be a single phase, the structure of t h e j l m should reflect the
most energetically stable single-phase form available to it at the temperature
of the substrate. This phase can be directly found from the Gibbs energy versus
composition (G/x)diagrams of the alloy system of interest.. .
The above-defined principle has been used in a series of calculations relating to vari-
ous possible carbide and nitride coating combinations formed from among transition
metal nitrides and carbides (all with the cubic NaCl structure) on the one hand and A1N
or S i c (hexagonal wurtzite structure) or TiB, (hexagonal structure) on the ~ t h e r . ~ - ~
Further calculations have been carried out for coatings in the A120,-A1N s y ~ t e m . ~ . ~
The present paper compares the results obtained from thermodynamic prediction
of the metastable phase ranges in the AlN-TiN and A120,-A1N coating systems with
results obtained from experimental investigations of the two systems.
2 (Ti,AI)N Coatings
A1N (hexagonal wurtzite structure) and TiN (cubic NaCl (cF8) structure) show virtu-
ally no solubility in each other under equilibrium conditions at temperatures below
about 1500°C (see Figure l), but when co-deposited on a cold-temperature substrate
using PVD techniques, the metastable cubic NaCl structure can be retained to high
concentrations of A1N in the deposited coating. Such coatings are already being pro-
duced commercially for cutting tool applications. They have been shown to exhibit
superior performance as compared to TiN under conditions of wear at elevated tem-
peratures due to their better oxidation resistance and hardness. The oxidation resist-
ance of the coatings increases significantly with increasing A1 Content. However, the
deposition of (Ti,Al)N films with TUA1 ratios below about 30/70 leads to a hexago-
nal coating structure that is not suitable for tribological coatings. Until more recently,
relatively little was known about the location of the cubic to hexagonal phase transi-
tion and hence about the optimum composition of (Ti,Al)N hard coatings.
2500
/
1
1
gas + fcc
2300
\
'hexagonal
P.
2100 cubic NaCl .
G
G
1900
hexagonal+ cubic NaCl

1700
1500
Figure 1 Calculated AlN-7iN section of the Al-7i-N system

Thermodynamic Prediction of the Formation 199
The significance of the metastable structural phase transition in the AlN-TiN sys-
tem on the properties of coatings produced by PVD allows the important potential
of thermodynamic calculations in predicting the ranges of composition and temper-
ature in which different metastable phases may form in a particular coating system
to be demonstrated.
The thermodynamic properties of hexagonal AlN and cubic transition metal
nitrides and carbides are known,* but for the calculation of the ranges of metastabil-
ity of the cubic and hexagonal phases in the AlN-MeN or AlN-MeC systems, the
energy differences for the transitions from the stable to the metastable structures, e.g.
MeN,,, 4 MeN,,, and AlN,,, +AIN,,, are of primary importance. In addition, the
thermodynamic properties of mixing of the metastable solutions are required. Since
virtually no experimental data are available for the energy differences referred to, a
combination of calculation and estimation techniques must be used to obtain them.
Methods available for this purpose are described below.
2.1 Metastable + Stable Structural Transformation Energies for

Nitride and Carbide Phases
A standard method of obtaining missing thermochemical values is to make plots of
relevant properties in comparison with neighbouring substances in the Periodic
Table, as shown in Figure 2 for enthalpies of transition fcc +bcc of transition metals
from the 2nd long period.
It is clear from Figure 2 that it is not a straightforward matter to predict a missing
value from such an irregular curve. However, it has been found that the reorganiza-
tion of the Periodic Table used by Pettifor’ (Figure 3), in which the elements are
listed in a single string based on what Pettifor terms the Mendeleev Number, allows
enthalpies of formation of particular types of compound with a given structure to be
plotted on a smooth curve (see ref. 10).
Figure 4 presents such a plot for the enthalpies of formation of nitrides with the
cubic-NaC1 structure.
-200001 I I I I I I I I I
0 1 2 3 4 5 6 7 8 9
Element
Figure 2 Enthalpies of the fcc +bcc transition .for elements of the 2nd Long Period of the
Periodic Table
200 Chapter 17
Figure 3 The Pettifor single string rearrangement of the Periodic Table in the sequence of the
Mendeleev Number
Mendeleev number
Figure 4 Enthalpies of formation of nitrides with the cubic-NaCl structure plotted using the
Pettifor arrangement of the Periodic Table
S t ~ l t e nhas
~ . ~developed a semi-empirical estimation method to obtain enthalpies
of formation of carbides, nitrides and oxides in metastable crystallographic struc-
tures. The calculations involved are based on the classical alloying criteria of atomic
size, valence and electronegativity, and incorporate a correction factor obtained
Thermodynamic Prediction of the Formation 20 1
using one or two reliable experimental enthalpy of formation values for the structure
concerned. When plotted as a function of the Mendeleev number, the calculated
enthalpies are in all cases essentially linear for compounds of transition metals with
Mendeleev numbers from 49 (Zr) to around 57 (Cr). At higher Mendeleev numbers,
a marked curvature is observed. Figure 5 presents the results obtained by Stolten for
the enthalpies of formation of nitrides of the transition metals in the cubic-NaC1and
hexagonal-ZnS structure.
Fernandez Guillermet and Grimvall have also written a series of papers [ 1 1-17]
devoted to deriving enthalpies of formation and standard entropies for a range of
metastable carbide and nitride phases of the 3d, 4d and 5d transition metals.
Thermodynamic values were obtained from an analysis of the cohesive properties
and vibrational entropies of the compounds concerned, whereby a quantity with the
dimensions of energy, E,, was defined as
where R is the average volume per atom and is obtained from experimental lattice
parameters.
The derived values were found to vary in a regular and related way when studied as
a function of the average number of valence electrons per atom in the compound, n,.
Figure 6 a and b, taken from Fernandez Guillermet and Grim Vall", shows the
entropy-related characteristic energy, E,(Ry), and the enthalpy of formation (per mol
of atoms) for the 3d transition metal carbides.
Tables 1 and 2 present a summary of available data, obtained both by experiment8
and by calculation for nitrides and carbides with the cubic-NaC1 and
hexagonal wurtzite structures.
Zr Hf Ti Ta Nb V WMoCrRe TcMnFeRu OsCoRh Ir Ni Pi Pd

7 0 1 ~ . ~ , ~ . ~ . ~ . ~ . ~ , , . , . , . ~ 1
m m
wurizite
m
m
0 0 0 0
m o o 0
m NaC1-cubic
0 .
0
Mendeleev Number
Figure 5 Experimental (solid squares) and c a l c ~ l a t e d(open

~ - ~ circles and stars) enthalpies
of formation of nitrides of the transition metals in the cubic-NaCl and hexagonal-
ZnS structures plotted as a function of the Mendeleev Number
202 Chapter 17
10
I YY
80 -
7 60-
s
.
I
2 40-
s
3 20-
F
4
I 0- 13
I
Figure 6(a) and (b) E,and AoH of carbides asfinctions of ne,The numbers refer to ( 1 ) ScC,
(2) Tic,(3) VC, (4) Cr3C2, (5) Cr7C3,(61 Cr2& (7) Mn7c3, (8) Mn&,
(9)Mn,C, (10)Mn,,C, ( 1 1 ) Fe3C, (12) C O ~ C(13)
, Ni3C (14) Sc2C9(15)
Sc,C3, with identical identijicationfor the E, values at the same n,. The
dashed lines refer to the (cF8) structure
Comparison of the different calculated values with each other and with experi-
mental data from Landolt-Bornstein Series8 is presented in Figures 7 and 8 for tran-
sition metal carbides and nitrides with the NaCl structure.
The calorimetric study carried out by Stolten for three compositions in the metastable
A1N-TiN system provides direct experimental data both for the enthalpy difference
AlN,, +AlN,,,, NaCl (see Table 1) as well as the mixing enthalpy of AlN-TiN
cubic NaCl ~olutions.~ These values, together with the calculated value for the Gibbs
energy difference between the NaCl and wurtzite structural forms of pure TiN,3,4
allow a thermodynamic analysis of the TiN-AlN system to be made.
Gibbs energy curves have been calculated for the hexagonal and cubic phases for
various hypothetical (Ti,Al)N deposition temperatures (Figure 9). The point of inter-
section of these curves at each temperature defines the composition at which there is a
transition from one structure to the other. It can be seen that this composition is nearly
temperature independent and has a calculated value of around 0.7 mol fraction AlN.
Experimental studies of the extent of the metastable cubic range in the section
AlN-TiN have been carried out by Knotek and Leyendecker18and more recently, with
Table 1 Enthalpies of formation of transition metal nitrides and of AlN in the cubic-
NaCl and hexagonal wurtzite structures
Nitride Af H(298 K)-cubic Re$ Af H(298 K)-hexag. Re$
Wmol-' MeN '
W mol- MeN
ZrN -365.3 8 -310 3,4
HfN -369.2 8 -340 3,4
TiN -337.6 8 -240 394
TaN -252.3 8 -40 394
NbN -236.4 8 0 394
VN -217.1 8 160 394
WN -111.8 394 420 374
MoN -38.0 394 600 334
CrN - 117.0 8 620 3,4
ReN -43.0 3,4
TcN -34.8; 90.0 3,4; 14
MnN - 185.8; -59.2 3,4; 17
FeN 2.0; 6.4 3,4; 17
RUN 137.8; 159.2 3,4; 14
OsN 121.4 5,6
CON 13.2; 37.4 3,4; 17
RhN 117.6; 108.0 3,4; 14
IrN 103.2 334
NiN 22.0; 43.2 3,4; 17
PtN
PdN 77.2 17
A1N -278.0 334 -3 18.0 8
Table 2 Enthalpies of formation of transition metal carbides in the cubic-NaCl

structure
Carbide AfH(298K)-cubic Re$ Carbide AfH(298K)-cubic Re$
Wmol-' MeC Wmol-' MeC
ZrC - 196.6 8 MnC -86.4; 12.6 3,4; 13
HfC -230.1 8 FeC 50.0; 42.0 3,4; 13
Tic -184.1 8 RuC 140.6; 74 3,4; 14
TaC - 144.1 8 osc 130.6 394
NbC - 140.6 8 COC 56.8; 51.2 3,4; 13
vc0.88 - 101.7 8 RhC 124.0; 72 3,4; 14
wc -40.0 8 Irc 114.0 394
MoC - 10.0 8 NiC 63.6; 62 3,4; 13
CrC -28.4; 1.2 3,4; 13 PtC
ReC 24.0 394 PdC 64 14
TcC 29.6; 64 3,4; 14
greater accuracy, by Cremer et aZ.19using electron probe microanalysis, X-ray photo-

electron spectroscopy and thin-film X-ray diffraction. The latter study showed that for
co-deposition temperatures between 300 and 700"C, a two-phase region exists in the
metastable AlN-TiN phase diagram. This expands with increasing temperature,
whereas at 100°C substrate temperature, no two-phase region was observed (Figure 10).
204 Chapter I 7
Zr Hf Ti Ta Nb V W MoCr Re Tc MnFe Ru 0 s C o Rh Ir N i Pt Pd
150 I ~ I ' I ' I ~ I ~ I ~ l I~ " l ' l ~
125 - O O O
100 -
h 75
- 25
8 0
$ -25
-8 -50
€ -75
2
% -100
- 8-125
3-150
5 -175
-2Ga
-225
-258
Mendeleev Number
Figure 7 A comparison of experimental and calculated enthalpies of formution for transition

metal carbides with the cubic-NaCl structure. Experimental - solid squares (Ret 8);
Calculated - open circles (refs. 3 and 4), crosses (refs. 13 and 14)
200
150
100
g
.I
-100
150
0
-250
;a
2 -300
w
-350
-400
Mendeleev Number
Figure 8 A comparison of experimental and calculated enthalpies offormation for transition

metal nitrides with the cF8 structure. Experimental - solid squares (ReJ8);
Calculated - open circles (Refs. 3 and 4), crosses (Refs. 14 and 17)
The composition of the metastable structural transition in the experimental stud-

ies lies between about 0.63 and 0.69 mol fraction AlN. The calculated transforma-
tion composition is therefore in good agreement with these observations.
Based on the experimental work, a new Ti,-,Al,N hard coating for tribological
applications with a composition of 0.62 mole fraction AlN has been developed, which
shows an increase of oxidation resistance by a factor 4 over conventional TiN coatings.
Because of the importance of the cubic to hexagonal transition in determining the
properties of coatings for particular applications, a series of calculations have been
carried out for a number of A1N-MeN systems (MeN is a transition metal nitride),
mol fraction TiN
Figure 9 Gibbs energy of formation of the cubic and hexagonal phases in the A1N-TiN sys-
tem at difierent temperatures
900 1 ~ 1 ' 1 ~ 1 '

/
cubic + hexagonal
800 -
700 -
Y 600-
1500:
(A1,Ti)N -
km: (Ti,Al)N hexagonal -
F - cubic
300 - A A I I ( A
200 -
100 -
0 1 ' 1 ' 1 ' 1 .
Figure 10 Metastable TiN-AlN phase diagram according to experimental work from Ref 19.
Experimental compositions with a large cross indicate a two-phase specimen
with a view to determining the extent of the metastable cubic and hexagonal ranges
in each case. Energies of transition between the wurtzite and NaCl structures have
been taken from Table 2, and the assumption has been made that the mixing proper-
ties are the same for the wurtzite and NaCl solutions and are equal to those in the
system AlN-TiN in all cases.
Figure 1 1 illustrates the calculated points of intersection of the Gibbs energy
curves for the cubic and hexagonal solutions at 500 K, thereby defining the ranges
of the two structures for the different, potential (Me,Al)N coatings.
It can be seen that hafnium reduces the range of the cubic phase significantly com-
pared with Ti, whereas Cr is effective in increasing the range further.
Table 3 presents a summary of the calculated metastable hexagonal and cubic
coating ranges for a number of MeN-AlN systems at 500 K.
206 Chapter 17
400
/ " " I
-
-
350
5- 300
0
c
.s 250
E
g 200
100
(3
50
0
0 01 02 03 04 05 06 07 08 09 10
rnol fraction MeN
Figure 11 Calculated ranges of the metastable cubic and hexagonal phases for a number of
MeN-AlN systems at 500 K
Table 3 Calculated composition ranges for the cubic and hexagonal metastable
coating phases in various MeN-AlN systems at 500 K
MeN NaCl-cubic range Hexagonal wurtzite range
mol fraction AlN mol fraction AlN
HfJY 0 to 0.42 0.42 to 1
ZrN 0 to 0.59 0.59 to 1
TiN 0 to 0.71 0.71 to 1
TaN 0 to 0.84 0.84 to 1
NbN 0 to 0.86 0.86 to 1
VN 0 to 0.91 0.91 to 1
CrN 0 to 0.95 0.95 to 1
3 AI,O,-AIN Coatings
Al,O,-AlN coatings are of considerable interest because they offer the possibility of
combining the advanced physical properties of electrovalent A1,03 with covalent AlN.
-
Although for temperatures below 1627 K, there is no reciprocal solubility in the
solid state, a metastable, homogeneous oxynitride phase could have potential appli-
cations in many areas because, e.g.,
the refractive index of a pure AlN film is high20, whereas the refractive index of
A1,03 is low2'
the thermal conductivity of A1N2, is high compared to that of A1,0,,,
AlN is a semiconductor, A1,0, an insulator.
If homogeneous, metastable Al-0-N phases with variable composition can be
stabilized by vapour deposition, then it may be possible that the different properties
of the binary compounds can be combined.
Metastable A1-0-N layers have been produced by reactive magnetron sputtering
ion plating using an A1 target and an Ar-0,-N, gas mixture.24For a substrate tem-
perature of 190°C and variable partial pressures of 0, and N,, a variety of coatings
with different 0 and N contents were prepared and analysed by means of X-ray pho-
toelectron spectro-scopy, Auger electron spectroscopy, high-resolution transmission
0.0 0.02 0.4 0.6 0.8 1.o

A1203 Mol fraction AlN AlN
Figure 12 Gibbs energy of formation of the spinel, amorphous and wurtzite phases in the
Al,O,-AlN system at 463 K
- 1 . 1 1 ' 1 '
300-
Hexagonal
Spinel
1
50: I I I I . I I , ! .
\ II
0.0 0.2 0.4 0.6 0.8 1.o
A1203 Mol fraction AIN AIN
Figure 13 Calculated metastable phase ranges in the Al,O,-AlN system with experimentally
observed structures at 190°C from Van Richthofen and Dimnick 24
electron microscopy and high-resolution scanning electron microscopy. Oxygen-rich

Al-0-N phases were found to be nanocrystalline with a cubic gamma-A1203(spinel)
structure. Nitrogen-rich phases displayed a hexagonal AlN (wurtzite) structure. For
O/N ratios around 1.5, the deposited phase was "EM amorphous (see Figure 13).
Using Gibbs energy values provided by a thermodynamic evaluation of the A1,0,-
A1N system,25and incorporating thermodynamic descriptions for the wurtzite and
amorphous phases for the entire composition range, Gibbs energy curves as a function
of composition have been calculated for the spinel, wurtzite and amorphous phases for
temperatures between room temperature and 600 K. The experimentally observed
transition in structure from amorphous to hexagonal at a composition close to 70 mol%
AlN was used to adjust the thermodynamic parameters of these two phases.
The calculated curves for the deposition temperature (463 K) used in the experi-
mental studies are illustrated in Figure 12.
208 Chapter I7
From the loci of the points of intersection of these curves as a function of tem-
perature, the deposition ranges of the phases have been determined (Figure 13).
4 Summary
The results presented above with respect to the AlN-TiN and Al,O,-AlN coating
systems demonstrate that thermodynamic calculations are able to provide information
of significant practical importance. In particular, the capability of calculating compo-
sition and temperature ranges of metastable coating phases produced by PVD
processes offers significant time and cost-saving predictive support in the development
of new, multi-component coatings with desired structures. Using the results of calcu-
lations to define selected experiments, much more rapid and reliable progress can be
made than that associated with widely applied trial-and-error development methods.
References
1. N.Saunders and A.P.Miodownik,J. Muter. Res. 1986, 1, 38.
2. P.J.Spencer and H.Holleck, High Temp. Sci. 1990, 27, 295.
3. H.Stolten, Ph.D. Thesis, Lehrstuhl fur Theor. Hiittenkunde, RWTH Aachen, 1991.
4. H.Stolten, P.J.Spencer and D.Neuschutz, J. Chim. Phys. 1993, 90, 209.
5 . P.J.Spencer, G.Eriksson and A.von Richthofen, Proc. CODATA’94, Chambery, France,
Sept. 1994, Springer, Berlin, 1998,
6. P.J.Spencer, in Ringberg workshop on computational thermodynamics, Group 5, new
applications of thermodynamic calculations, CALPHAD 24 (2000) 72.
7. P.J.Spencer, 2. Metullkunde, 2001, 92, 1145-1150.
8. Thermodynamic Properties of Inorganic Materials compiled by SGTE, Landolt-
Bornstein, New Series, Group lV,Vol 19, SubvolumeA - Pure Substances, Subvolume B -
Binary Alloys, Springer-Verlag, Berlin, 1999 onward.
9. D.Pettifor, New Scientist, 1986, 48.
10. P.J.Spencer, Thermochim. Actu, 1998, 314, 1.
11. A.Fernandez Guillermet and G.Grimval1, Phys. Rev., 1989, B40, 10582.
12. J.Haglund, G.Grimvul1, T.Jarlborg and A.Fernandez Guillermet, Phys.Rev., 1991, B43,
14400.
13. A.Fernandez Guillermet and G.Grimval1, J. Phys. Chem. Solids, 1992,53, 105.
14. A.Fernandez Guillermet, J.Haglund and G.Grimval1, PhyxRev., 1992, B45, 11557.
15. A.Fernandez Guillermet and K.Frisk, K., unpublished work.
16. A.Fernandez Guillermet, L.Haglund and G.Grimval1, Phys.Rev., 1993, B48, 11673.
17. A.Fernandez Guillermet and K.Frisk, J.Alloys Compds, 1994,203, 77.
18. 0.Knotek and T.Leyendecker,J.So1id State Chem., 1987, 70, 318.
19. R.Cremer, M.Witthaut and D.Neuschutz, in ‘Value Addition Metallurgy’, W.D.Cho and
H.Y.Sohn and (eds), The Minerals, Metals & Materials Society, 1998 249.
20. D.H. Wang and L. Guo, Thin Solid Films, 1987, 158, L39.
21. D.H. Wang, in Proceeding of the 172nd Meeting of the Electrochemical Society, Hawaii,
Oct. 1987; J.Electrochem.Soc., 1987, 2, 575.
22. N.K. Smit, G.A. Acker and C.J. van der Laan, Thin Solid Films, 1986, 138, 171.
23. H. Schule and W.-I. Ratzel, Technische Kerumik, lst.edn, Vulkan, Essen, 1998.
24. A. von Richthofen and R. Domnick, Thin Solid Films, 1996,283, 37.
25. M. Hillert and S. Jonsson, Z.Metullkde., 1992,83, 714.
CHAPTER 18
Thermodynamics of the
Nano-Sized Particles
TOSHIHIRO TANAKA, JOONHO LEE AND NOBUMITSU HIRAI
1 Introduction
Various thermodynamic databases have been compiled to be mainly applied to the
calculation of phase diagrams of alloys, salts and oxides.' The accumulation and
assessment of thermodynamic data and phase equilibria information to establish
those databases is sometimes called Computer Calculation of Phase Diagrams
(CALPHAD) approach.2 The CALPHAD approach has been recognized to be use-
ful in various aspects of materials science and engineering.'Y2 If it would be possible
to use the thermodynamic databases to evaluate various phy sicochemical properties
as well as phase diagrams, we could not only widen the applicability of those ther-
modynamic databases but also understand the physicochemical properties more
deeply. The authors have applied those thermodynamic databases to the evaluation
of the surface tension of liquid alloys, molten ionic mixtures and the interfacial ten-
sion between liquid steel and molten lag.^-'^ These works are aimed to comprehend
the thermodynamic properties of a material system including its surface or interface
as well as the bulk.13,14Since the effect of its surface on the total thermodynamic
properties cannot be negligible in small particles of metals and alloys, the phase rela-
tions in these metals and alloys are dependent on the size of the particle and its sur-
face property. The thermodynamics of solid-liquid-phase equilibira in a small
particle system was firstly discussed by Pawlow in 1909,15followed by Hanzen16in
1948. Since Takagi17observed in 1957 that the melting points of some pure metals
decrease with decreasing size of these metallic particles, several studies have been
carried out on the effect of the particle size on the melting phenomena of pure met-
als and As described above, the authors have evaluated the surface prop-
erties as well as phase diagrams of alloys on the basis of thermodynamic
databa~es.~-l~ Extending these techniques, a new system might be developed to esti-
mate phase equilibria of materials in small particle systems as shown in Figure 1.
The authors have tried to calculate some nano-particle binary alloy phase diagrams
210 Chapter I8
7 7
Alloys, Salts & Oxides etc.
Nano-sized Particles
(Surface Tension,
Figure 1 Concept of a new evaluation system for phase equilibria in nano-sized particle sys-
7
tems on the basis of thermodynamic databases
of Cu-Pb, Cu-Bi and Au-Si systems by using thermodynamic databases,13 and in

addition, to evaluate the effect of the surface properties on those phase diagrams on
the basis of a regular solution rn0de1.l~In the present paper, we describe an outline
of our trials13>14
to evaluate phase diagrams of binary alloys in the nano-sized parti-
cle systems.
2 Gibbs Energy in Small Particle Systems

When a pure solid phase is selected as the reference state of Gibbs energy, the total
Gibbs energies in liquid and solid phases, Adrotdl Phase (Phase=Liq or Sol), of an A-B
binary alloy system in a small particle with its radius r are described in the following
equations (1)-(6):l39l4
AGTotal, Phase = A G B u l k , Phase + AGSurface, Phase

(1)
The Gibbs energy of the bulk of A-B binary alloy in liquid and solid phases,
A G B u l k ? Liq and A G B u l k , in Equation (l),which corresponds to AGTotal*(P=Liq
or Sol) with r=-, are expressed in Equations ( 2 ) and (3).
where AGkS and AGf are Gibbs energies of pure liquid phases relative to those of
pure solid phases, in other words, the Gibbs energy of the melting. GExcess, L1q and
GExcess,Sol are the excess Gibbs energies of liquid and solid phases in the A-B alloy.
N A and N B are mole fractions of components A and B. Although detailed functions
of compositions and temperature for excess Gibbs energies are generally stored in
thermodynamic databases, the regular solution model is applied here to evaluate the
excess Gibbs energy GExcess, Phase because various interaction energies can be selected
easily as follows:
GExcess, Phase = N
A
.N~ . w,PPase
(4)
Thermodynamics of the Nano-Sized Particles 21 1
where WEae (Phase=Liq or Sol) is the interaction energies in liquid and solid phases.
In addition, the Gibbs energy of the melting AGF is also compiled for various ele-
ments in thermodynamic databases, but here we use the following simple equation (5)
as Pelton and Thompson32assumed this equation in their evaluation of phase diagrams:
where Tx,mp and Sx,mp are the melting point and the entropy of fusion for pure sub-
stance X(X=A or B). S , mp is roughly assumed to be 10JK-' mol-' according to
Richard's rule.33In addition, we select here TA: mp= 1200K and TB,,,=600K as an
example to evaluate a phase diagram of A-B binary system.
AGsu~ace9Phase in Equation (l), the effect of the surface on AGTota'*
Phase, is assumed
as f01lows:'~J~
2 Phase v Phase - 2 ( N A o r 1 V r 1 + N o V
AGsUrfaCe, Phase =
r
" ") (Phase = Liq or Sol) (6)
where oLiqand oSo1 are surface tensions of liquid alloy and solid solution, VLiqand p'
the molar volumes of liquid alloy and solid solution, r the radius of a particle, orland
or1the surface tensions of pure solid A and B and Vrl and VSp' the molar volumes of
pure solid A and B. The second term on the right-hand side of Equation (6) comes from
the fact that a pure solid phase is selected as the reference state of Gibbs energy.
On the surface tension of pure liquid metals at their melting points we use oFmp,
the following approximation:l4
TXm
o!&,~= (Nm-') (7)
1000
In addition, the temperature coefficient of the surface tension of pure liquid has been
reported to be about 0.0001 Nm-' K-1.34Thus, we assumed the following equation
for okiq and o,L'4:
= 0.6 - 0.0001 (T - TB,,,)/Nm-'(T,, mp = 600K) (9)
As shown in Equation (6), we need the value of the surface tension oF1of pure solid
X, but the precise information on the value of OF'and its temperature dependence
are i n ~ u f f i c i e n t . From ~ data reported in some reference^,^^-^^ the value of oT1
~ ~ - ~ the
of pure solid at the melting point is found to be 25% larger than the surface tension
of pure liquid on the average. Equation (10) is, therefore, assumed to express the sur-
face tension OF'of pure solid X in the present work:
aopq
a?' = +
1.250~~qm, 7(T - Tx, mp) ( X = A or B ) (10)
where oXLfgmisp the surface tension of pure liquid X at its melting point Tx,mp. The
temperature dependence of o?' in Equation (10) is assumed to be the same as that
of opq. In addition, the effects of crystal faces on o;c"lare ignored here. When
212 Chapter 18
Equation (10) is applied for the surface tension opl of pure solid X , the following
relations are obtained from Equations (8) and (9):
or1= 1.25 X 1.2-0.0001 (T - TA, mp)/Nm-l (1 1)

= 1.25 X 0.6 - 0.0001 (T - TB, mp)/Nm-l (12)
On the molar volumes of pure components, we used the following values in Equation
(13) by assuming that the temperature dependence of the molar volume and the vol-
ume change due to the melting are neglected because the effect of the excess Gibbs
energy and the surface tension on the phase equilibria is focused.13*14
VzhaSe= VPhase
B = 10 X lO-'/m3rnol-' (Phase = Liq or Sol) (13)
It is assumed that the molar volumes of liquid alloy VLiq and solid alloy VSo1in
Equation (6) are obtained from the following simple additivities:
VPhase = N VPhase +N VPhase
(14)
A A B B
The authors have evaluated the surface tension oPhase in Equation (6) of A-B binary
alloys on the basis of Butler's equation [38] as follow^:^-^^
In Equations (15a) and (15b), gurf and Ni& are mole fractions of element B in the sur-
face and the bulk, respectively. A~ase=LN~b(VxM)2b (No Avogadro number: X=A or
B, L= 1.091) is the molar surface area of pure X , and this is obtained from the molar
volume Vxm q mPhase, Bulk (T, Ni"lk)and G
s 9 F Bulk (T, @") are partial excess
Gibbs energies of component X in the bulk and the surface, respectively, as functions
of T and Ni& or @". Since the partial molar excess Gibbs energy of component X
(X= A or B) in the bulk GFEss*phase, Bullc (T, Ni&) in Equations (15a) and (15b) can be
obtained directly from GExcesst Phase in Equation (4) by using Equations (16a) and (16b),
a GExcess, Phase
GF
Phase, Bulk (T, N,Bulk) = G Excess, Phase -
( W
NB aNB
a G Excess, Phase
G?, Phase, Bulk (T, ~BBulk) = G Excess, Phase - - NB) aNB ( 16b)
For the excess Gibbs energy in the surface, we derived the following equation( 17)3-'4
based on the model proposed by Yeum et ~ 1 . ~ ~
Thermodynamics of the Nano-Sized Particles 213
where PMrX*
Liq = 0.83:liquid alloys & pMIX*
Liq = = 0.75:solid solutions.
12
Equation (17) indicates that G F s"f(T,iVid),which has the same formula
as GF is obtained by replacing iVFkby NiUrfin G F Phase, s"f
Phase, Bulk (T, iVi"Lk),
(T, N i d ) (X=A or B) and then multiplying P M I X y Phase with G F (T, N p - 9 .
pMIX* Phase is a parameter corresponding to the ratio of the coordination number in the
surface to that in the bulk considering the surface rela~ation.~-'~ In particular, for the
solid solutions, here we use the value 0.75 as p MIX* by assuming that the coordi-
nation number in the bulk is 12 and that in the surface is 9 in a closed-packed struc-
ture. The surface tension crPhasein Equation (15) of liquid alloys or solid solutions
can be calculated from Equations (15)-( 17) as follows:
(1) Setting temperature T and composition iVpkof a solution.

(2) Inserting the values for surface tension cryase
and molar volume V!hase of pure
substances at the above temperature in Equations (15a) and (15b).
(3) Determining excess Gibbs energies in the bulk phase at the above temperature
and composition, and substituting them in to Equations (15a) and (15b).
(4) Then, Equations (15a) and (15b) become the simultaneous equations with
unknown Nid and cr Phase. These equations are solved for those unknown
and CT Phase numerically.
3 Calculation of Nano-Sized Phase Diagrams of Binary

Alloys
From Equations (l), (2), ( 6 )and (lo), the Gibbs energy of pure component X includ-
ing the surface is obtained as follows:
The temperature T, which yields A g o t a l = O in Equation (18), is the melting point of

pure X at a given radius r of a particle. Figure2 shows the change in the melting point
of pure Au with the radius of the particle calculated from Equation (18) with exper-
imental data.20*21The data34740necessary in this calculation are given in Table 1.
Liquid-solid-phase equilibria in an alloy are obtained from the following thermo-
dynamic conditions:
P p l , Liq =
P P l , (20)
a AG TO^, Phase
PpalPhase
, = & T o t a l . Phase -
NB aN, (Phase=Liq or Sol) (21)
214 Chapter 18
li
J
5a 1250
hp, Mdtlny polnt of pure Au
115B 10 20 30 40 50
Radius of particle / nm
f1
Figure 2 Change in melting point of pure Au with particle size. Expe.: Sambles,20Coombes.21
Table 1 Parameters used in the calculation of phase diagrams

Melting point Entropy of Sur$ace Tension Molar volume
IK fusion lNm- lm3mol-'
IJK- 'mol-
Substance A: 1200 10 1,2-0.0001( T- 12OO):Liq OX IO-Wq.
1.2X 2.5 -0.000 1 (T- 1200):Sol. 10X 1O-? Sol.
Substance B: 600 10 0.6-0.0001(T- 600):Liq . OX 10-6:Liq.
0.6X2.5-0.0001(T-600):S01. lox 10-6:S~l.
30 15 0 - 15
w,;q/k.J 20 Figure 4(a)
15 Figure 4(b)
0 Figure 4(c) Figure 3(b) Figure 3(a)
and Figure 3(c)
-20 Figure 3(d) Figure 3(d)
aAGT0ta'.Phase
ppl,Phase = AGTota1. Phase + (1 - N ~ ) (Phase = Liq or Sol) (22)
JNB
Figures 3 and 4l3,I4show the calculated results of the phase diagrams using the param-
eters in Table 2. In these figures, the solid curves indicate the phase equilibria in the
bulk (r=m). On the other hand, the chain and the dotted curves are the calculated
results for r= 10 and 5nm, respectively. Figure 3(a) indicates the phase diagrams for
the ideal solutions in both solid and liquid phases, in other words, Wk2=WZ1=O.
When W&' increases from 0 to 30 kJ with Wgq=O, the phase diagrams change as
I I
Figure 3 Calculated results of the phase diagrams with systematic change of W;? and WAS':
on the basis of the regular solution model
shown in Figures 3(b) and (c). Figure 3(d) shows a phase diagram for Whq= -20 kj
and Wz1=-15kj as an example for Wzq<O and W E k O . When W g l is fixed to be
30 kJ and W&q changes from -20 to +2 kJ, the phase diagrams are obtained as shown
in Figures 4(a)-(d). As can be seen in Figure 4(a), when the interaction parameters are
Wzq%O and W&l%O, the effect of the particle size on the phase diagrams is remark-
able. As described in the authors' previous the liquid or solid alloys indicate
the large downward curvature of the composition dependence of the surface tension in
the conditions of Wkq>>Oand W&'%O. These composition dependences of the surface
tension affect the contribution of AGSurface*
Phase to AGOd, Phase largely. In addition, when
Wzl>O, the solid solutions do not appear in some of the phase diagrams of the bulk
as shown in Figures 3(c), (d) and 4(d). However, when the particle size decreases, the
contribution of AGsurfacev to AQod, in the solid phase cannot be ignored.
Consequently, as shown in Figures 3 and 4, the solid solution appears in the small par-
ticle systems even when the bulk phase diagrams do not show the solid solutions.
As metioned above, when an alloy has a large positive interaction among the com-
ponents ( W P > O ) ,the phase diagram of the alloy changes remarkably with the size
of a particle. Figure 5 shows one example of the calculated result of Cu-Pb alloy,
which has a large positive excess Gibbs energy.13 Since this phase diagram consists
of a liquid phase and pure solid phases, the solid solutions are not considered in the
calculation. The calculation procedures and the data are given in detail in the
authors' previous work. l 3
216 Chapter I8
, I . . -
..------....--. -,..-........
-.....-....-. :;*.I*...
:I;:;:
Figure 4 Calculated results of the phase diagrams with systematic change of WkLq in the liq-
uid phase and W;;'= 30 kJ on the basis of the regular solution model
As can be seen in Figures 2-5, the phase diagrams of binary alloys in the nano-
sized particle systems can be estimated from the information stored in thermody-
namic databases with some fundamental physical properties, although the rough
assumptions have been used here.I3J4As Howe pointed the atomic bonding
near the surface of the bulk materials is weaker than that in the bulk, which causes
the surface melting. In this phenomenon, the thin layers near the surface of metals
are melted below the melting points of the bulk. For example, Kojima and S U S ~ ~ ~
investigated the surface melting of pure Cu by using the MD method. They reported
that the surface melting occurs at 900 K in the face [ 1111, although the bulk Cu melts
at 1356K. Thus, in the nano-sized particles, whose surface area per unit volume is
Table 2 Data necessary for calculation of the melting point of pure Au

Surj4ace tension o;/Nm-' of pure liquid Au: Ref: 34 adz
-/Nm-'T-'
aT
0ku= 1.169-0.00025 (T- 1336.15) -0.00025
ok,,dNrn-': Ref. 33 Tx, mp/K:Ref. 34
1.169 1336.15
Molar volume/ m3mol-' of pure liquid Au: Ref. 34
Vku=11.3X10-6{ 1.0+0.000069 (T-1336.15))
Gibbs energy change of fusion of Pure Au/ J mol-': Ref. 40
AGk:= 12552.0-9.385866 T
Figure 5 Phase diagram of Cu-Pb alloy
larger than that of the bulk systems, the surface phenomenon may occur more
remarkably than the bulk system. We should, therefore, consider the combination of
those microscopic approaches with the above macroscopic thermodynamics for fur-
ther development of the evaluation of the nano-sized phase diagrams.
4 Concluding Remarks
The authors have described some fundamental procedures to evaluate the solid-
liquid phase relations of nano-sized particles in binary alloys from the calculation of
the surface properties as well as the phase equilibira on the basis of thermodynamic
databases, which are usually used for the calculation of phase diagrams of the bulk
materials. In order to obtain quantitatively precise values of the melting points and
liquidus temperatures in alloys, we should carry out further investigation as follows:
More detail discussion on the thermodynamic criteria on the melting of nano-

sized particles.
The reliability of the data used for the evaluation of the melting phenomena of
the nano-sized particles, especially the surface energies of liquid and solid alloys.
The change in the bonding of atoms near the surface from that of the bulk in
relation to the surface melting phenomena.
We can, however, evaluate the phase equilibria of small particles qualitatively from
the macroscopic thermodynamic point of view as described in the present paper.
Thus, the authors wish that enormous attention be paid to the application of thermo-
dynamic databases to the calculation of the phase diagrams of the nano-sized parti-
cles, whose information will be in demand increasingly in the near future.
218 Chapter 18
References
1. C. W. Bale and G. Eriksson, Canad. Metall. Quart. 1990,29, 105.
2. T. Nishizawa, Matel: Trans. JIM, 1992,33,713.
3. T. Tanaka and T. Iida, Steel Research, 1994, 65, 21.
4. T. Tanaka, K. Hack, T. Iida and S. Hara, 2.Metallkd., 1996,87, 380.
5 . T. Tanaka, S. Hara, M. Ogawa and T. Ueda, 2. Metallkd., 1998,89, 368.
6. T. Tanaka, S. Hara, M. Ogawa and T. Ueda, Molten Salt Forum, 1998,5-6,213.
7. T. Tanaka, S. Hara and T. Ueda, in Proceedings of the 11th International Symposium on
Molten Salts, San Diego, ECS, USA, P.C. Trulove, H.C. De Long, G.R. Stafford, S.Deki
(eds), 1998,645.
8. T. Tanaka, K. Hack and S . Hara, MRS Bulletin, 1999, 24,45.
9. T. Tanaka and S. Hara, Electrochemistry, 1999,67, 573.
10. T. Tanaka and S. Hara, 2.Metallkd., 1999, 90, 348.
11. T. Ueda, T. Tanaka and S. Hara, 2.Metallkd., 1999, 90, 342.
12. T. Tanaka, K. Hack and S. Hara, Calphud, 2001,24,463.
13. T. Tanaka and S. Hara, 2.Metallkd., 2001, 92,467.
14. T, Tanaka and S. Hara, 2.Metallkd., 2001, 92, 1236.
15. P. Pawlow, 2.Phys. Chem., 1909,65, 1.
16. H. Reiss and I.B. Wilson, J. Colloid Sci., 1948, 3, 551.
17. M. Takagi, J. Phys. SOC.Jpn., 1954,9, 359.
18. K.J. Hanszen, 2.Phys., 1960, 157, 523.
19. C.R.M. Wronski, Brit. J. Appl. Phys., 1967,18, 1731.
20. J.R. Sambles, Proc. Roy. SOC.Lond. A., 1971,324, 339.
21. C.J. Coombes, J, Phys. F., 1972,2,441.
22. Ph. Buffat and J-P. Borel, Phys. Rev., 1976,13, 2287.
23, P.R. Couchman and W.A. Jesser, Nature, 1977,269,481.
24. P.R. Couchman and C.L. Ryan, Phil. Mag., 1978, A37,369.
25. G.L. Allen, W.W. Gile and W.A. Jesser, Acta Metall., 1980, 28, 1695.
26. J. Ross and R.P. Andres, Su@ Sci., 1981,106, 11.
27. J.P. Borel, Surf.Sci., 1981, 106, 1.
28. G.L. Allen, R.A Bayles, W.W. Gile and W.A. Jesser, Thin Solid Films, 1986,144,297.
29. H. Saka, Y. Nishikawa and T. Imura, Phil. Mag. A, 1988, 57, 895.
30. K. Sasaki and H. Saka,, Phil. Mag. A, 1991,63, 1207.
31. J.G. Lee, H. Mori and H. Yasuda, Phys. Rev. B, 2002,66,012105-1.
32. A.D. Pelton and W.T. Thompson, Prog. Solid State Chem., 1975, 10, 119.
33. D.R. Gaskell, Zntroduction to the Thermodynamics of Materials, 3rd edn, Taylor &
Francis, Washington DC, USA, 1995
34. T. Iida and R.I.L. Guthrie, The Physical Properties of Liquid Metals, Clarendon Press,
Oxford, 1988.
35. B.C. Allen and W.D. Kingery, Trans. Metall. Sco. AIME, 1959,215, 30.
36. L.E. Murr, Interfacjal Phenomena in Metals and Alloys, Addison-Wesley,London, 1975.
37. V.K. Kumikov and Kh.B. Khokonov, J. Appl. Phys., 1983,54, 1346.
38, J.A.V. Butler, Proc. Roy, Soc., 1932, A135, 348.
39. K.S. Yeum, R. Speiser and D.R. Poirier, Metall. Trans., 1989, 20B, 693.
40. A.T. Dinsdale, CALPHAD, 1991,15, 317.
41. J.M. Howe, Interfaces in Materials, John Wiley & Sons, Inc., New York, 1997.
42. R. Kojima and M. Susa, High Temp. High Press., 2002,34.
CHAPTER 19
Thermodynamics of Electrolyte
Systems of Industry
KAJ THOMSEN
1 Standard States and Chemical Potentials

The chemical potential pi of a species i is the partial molar derivative of the Gibbs
energy G, enthalpy H , Helmholtz energy A, or internal energy U of substance i1
Pi = (1)
In Equation (l),niis the amount of component i, T is the temperature, P is the pres-

sure, S is the entropy, and V is the volume. Matter flows spontaneously from a region
of high chemical potential to a region of low chemical potential just like electric cur-
rent flows from a region of high electric potential to a region of low electric poten-
tial and mass flows from a position of high gravitational potential to a position of
low gravitational potential. The chemical potential can therefore be used to deter-
mine whether or not a system is in equilibrium. When the system is in equilibrium,
the chemical potential of each substance will be the same in all phases of the system.
In an ideal solution, the chemical potential pjdeal
of species i is calculated from the
standard chemical potential PO, the gas constant R, the Kelvin temperature T, and the
mole fraction xi as2
The chemical potential of component i will be equal to the standard chemical poten-
tial when the mole fraction of component i equals 1. The standard state is thus pure
component i. If species i is an ion, the standard state is the hypothetical state of pure
ion i. In a real solution, an extra term, RT In is added to the chemical potential in
order to account for the deviation from ideality:
220 Chapter 19
x
where is the activity coefficient of component i. In the standard state, the activity
coefficient equals 1. There is no way of directly measuring the standard chemical
potential of pure ions. It has therefore been found practical to introduce a normal-
ized activity coefficient equal to the ratio of the activity coefficient and the activity
coefficient at infinite dilution:
‘
y* = - + 1
[ I T
for xi+ o (4)
By introducing this unsymmetrical activity coefficient in Equation (3), the expres-

sion for the chemical potential of component i can be written as:
The term p;= p; + RT In YF does not refer to the chemical potential in a specific
physical state, but has the advantage that it can be determined by experiment in the
limit of infinite dilution.
A molal activity coefficient f can be defined as the product of the water mole
fraction and the unsymmetric activity coefficient:
By insertion of this expression into Equation (5), the following expression for the
chemical potential of component i is obtained as a function of the molality mi(mol per
kg water) of component i, and the molecular mass M , of water in kg per mol water:
In this equation, the following definition of the chemical potential of ion i in “the
hypothetical state of the pure ions dissolved in pure water to give an ideal solution
at unit mean molality, under the standard state pre~sure”~has been used:
Like p*,p: can be measured and is the value that is most often tabulated for the
standard chemical potential of ion^^-^. mo= 1.0 mol per kg water is inserted in order
to make the expressions dimensionless.
The molal activity coefficient f is the activity coefficient usually measured
experimentally and reported, rather than the unsymmetric or symmetric mole frac-
tion activity coefficients. Most activity coefficient models, however, yield values of
the symmetric or the unsymmetric mole fraction activity coefficient. It is therefore
often necessary to convert back and forth between activity coefficients and chemical
potentials in the various systems.
1.1 Ionic Mean Properties

Electroneutrality is required in all solutions; thus, it is not possible to measure the
properties of a single ion without influence from an ion of opposite charge. By
convention, the standard enthalpy of formation and the standard entropy of the
Thermodynamics of Electrolyte Systems of Industry 22 1
hydrogen ion in aqueous solution are given the value zero at 298.15 K6These values
refer to the molal standard state. A value of zero is also assigned to the (molal) stan-
dard Gibbs energy of formation, heat capacity, and volume of the hydrogen ion. The
corresponding values for all other ions can be calculated relative to the value for the
hydrogen ion. In some tables, other conventions are used for the standard state heat
capacity and volume of the hydrogen ion.5
Like the standard state properties, the activity coefficient of a single ion is
immeasurable by any thermodynamically valid m e t h ~ dThe . ~ measured and calcu-
lated activity coefficients are therefore usually presented as the geometric mean of
the cationic and the anionic activity coefficients weighted according to the amounts
of each ion:
A similar expression can be obtained for the symmetric and the unsymmetric mole
fraction activity coefficient. In a completely dissociated solution of n mol Na2S0,,
the mean molal activity coefficient is
or
The mean molality is similarly defined by

1
lnm, = - x n j l n m j
ions
A corresponding definition is used for the mean mole fraction. The mean ionic activ-
ity is the product of the mean concentration and the mean activity coefficient.
2 Activity Coefficient Models

Activity coefficient models are equations representing the Gibbs or the Helmholtz
energy of solutions. Activity coefficients and related properties are derived form
these energy functions by proper differentiation (Equation ( 1)).
In addition to the short-range interactions between species in all solutions, long-
range electrostatic interactions are found in electrolyte solutions. The deviation from
ideal solution behavior caused by these electrostatic forces is usually calculated by
some variation of the Debye-Huckel theorys or the mean spherical approximation
(MSA).9 These theories do not include terms for the short-range attractive and repul-
sive forces in the mixtures and are therefore usually combined with activity coefficient
models or equations of state in order to describe the properties of electrolyte solutions.
The Electrolyte NRTL modellO-lland the Extended UNIQUAC are
examples of activity coefficient models derived by combining a deb ye-Huckel term
with a local composition model. Equation of state models with electrostatic terms for
222 Chapter 19
electrolyte solutions have not yet been developed to a level where they can be of
practical use in industry.Ig The Pitzer mode120 was developed by adding virial
expansion terms to the Debye-Huckel equation. The Pitzer model is based on
another definition of the ideal solution than the one given in Equation (2). A UNI-
QUAC model combined with a Debye-Huckel term and a virial expansion term was
introduced by Wang et al. in 2002.,l
3 Equilibrium Calculations
Equilibrium calculations for electrolyte solutions include speciation equilibrium,
vapor-liquid equilibrium, solid-liquid equilibrium, and liquid-liquid equilibrium.
As an example of the first three types of equilibria, we will consider the ternary
H,O-NH,-CO, system.
Chemical and phase equilibria are attained when the chemical potential of the
reactants equals that of the products. This criterion for equilibrium can be expressed
by the mass action law as
AGO is the change in standard state Gibbs energy by the process, calculated from the
standard state chemical potentials that can be found t a b ~ l a t e d .R~ *is~the gas con-
stant, T is the temperature in Kelvin, aiis the activity of species i, and vi is the stoi-
chiometric coefficient of component i in the process under consideration.
Components on the left-hand side in the process equation are given negative stoi-
chiometric coefficients. Usually, only the values of the chemical potentials at
298.15K are found in tables. It is then necessary to calculate the value of AGO at the
temperature of interest. This can be done by using the Gibbs-Helmholtz equation:
a(AGo/R7') -- --AHO
aT RT2 at constant pressure (14)
The variation of AHO, the change in standard state enthalpy by the process, with tem-
perature can be calculated from the heat capacity of the species involved in the
process.22
3.1 Speciation Equilibrium

The term speciation is used to describe the reactions that take place when an electrolyte
is dissolved in water. Water dissociates, sour gases hydrolyze, some ions dissociate,
and other ions associate until thermodynamic equilibrium is attained. The liquid phase
of the ternary H20-NH,-C0, system contains at least the following nine species: H20,
NH,(aq), CO,(aq), H', OH-, N H:, HCO;, CO:-, and NH,COO-. (aq) indicates that
the species is in aqueous solution to avoid ambiguity. In order to adequately model this
system, interaction parameters for the interaction between each pair of species need to
be determined; thus, speciation calculations are performed simultaneously with the
parameter estimation, and the calculated amount of each species is compared with
experimental data. Some models20also require ternary parameters and consequently an
additional amount of data to determine these parameters.
Thermodynamics of Electrolyte Systems of Industry 223
One of the speciation processes taking place in this solution is the formation of
carbamate ions, NH2COO-, from CO, and NH,:
H 2 0 + C02(aq) + NH,(aq) H30f + NH2COO- (15)
The AGO for this process to be used in Equation (13) is calculated by
In Equation (16), it has been assumed that the pure component standard state is used
for all components and that the activity coefficients of the solutes are normalized
according to Equation (4). The right-hand side of Equation (13) becomes
~ H , o~NH,COO-
+ ) =In( xHIO+ 1/*H30+XNH,C00- lfNH2C00-
'H20 'C02 'NH,(aq) XH,OYH,O x c o 2 ~ oXNH,&H,
,
When all independent speciation equilibria in this way are written in the form (13), the
equations can be solved for the mole fractions. The equation set must be solved by an
iterative procedure, as the activity coefficients are dependent on the mole fractions.
3.2 Vapor-Liquid Equilibrium

The gas phase of the system will mainly consist of H,O(g), NH,(g), and C02(g). In
order to perform vapor-liquid equilibrium calculations with an activity coefficient
model at pressures higher than ambient pressure, the activity coefficient model can be
combined with an equation of state for the gases.', Usually, there is no need for binary
interaction parameters in the equation of state as gas phase fugacities are only slightly
dependent on these binary interaction parameters. For the equilibrium between
ammonia in the liquid phase and ammonia in the vapor phase, Equation (13) gives:
The mole fraction and the fugacity coefficient are y and #, and the total pressure is
P . This is the so-called gammdphi formulation of vapor-liquid equilibrium.,
3.3 Solid-Liquid Equilibrium

A number of different solid phases can precipitate from a solution of ammonia and
carbon dioxide in water. As an example, the equilibrium of ammonium bicarbonate
with an aqueous solution is written in the form (13):
NH,HCO, E NH: + HCO,
As a standard state for the solid salt, the state of pure, solid salt is used. For solid-liq-
uid equilibrium calculations in mixed solvent systems, it is important to choose the
proper standard states for the ions.15
224 Chapter 19
3 -4 Liquid-Liquid Equilibrium
For liquid-liquid equilibrium to occur, it is required that the activity of each com-
ponent is the same in both phases. For a system of NaCl, water, and isopropanol (IP)
three equations of the form (13) can be written for the equilibrium between liquid
phase I and liquid phase 11:
The left-hand side of each of these equations is zero when the same standard state
for each species is used in both phases.23The mean ionic activity is being used for
the salt. For solutions containing more salts, it will be necessary to add one equation
for the mean activity of each independent salt.
4 Phase Diagrams
Solid-liquid equilibrium phase diagrams play an important role in the design of
industrial crystallization processes.24-26The calculation of phase diagrams can be
used to validate the activity coefficient model used for process simulation.
An example of a binary phase diagram is shown in Figure 1. Experimental data
from the open literature are plotted in the diagram along with the calculated phase
diagram. In the temperature range from 49 to 89 "C, magnesium nitrate has three dif-
ferent solubilities at each temperature corresponding to two different solubilities of
the hexahydrate, and the solubility of the dihydrate. Figure 2 shows the 50°C
solubility isotherm for the ternary H20-Na2S04-MgSO, system. This isotherm
consists of three different branches. The first branch represents the solubility of
sodium sulfate, the second branch represents the solubility of astrakanite
(Na2S04*MgS04.4H20),and the third branch represents the solubility of hexahy-
drite, (MgS04.6H20). In Figure 3, the liquid-liquid phase diagram for the ternary
H20-Na,$04-t-butanol system is shown. The units in this phase diagram are mass
percent. Polytherms for ternary systems can also be presented in two-dimensional
phase diagrams.17
Phase diagrams for reciprocal salt pairs can be presented by the method of
J a n e ~ k eAccording
.~~ to this method, a projection of the three-dimensional diagram
is made, and the water content is indicated by contour lines. In the Janecke diagram
in Figure 4, the water content is marked in mass percent at each grid intersection.
The concentrations of the four ions, K', NH4+, C1-, and NO,-, are indicated as
charge fractions. The lines in Figure 4 indicate compositions of liquids in equilib-
rium with two salts. Each field in the diagram represents a range of compositions
that are saturated with one salt.
The phase diagrams shown in Figures 1 4 were calculated with the Extended
UNIQUAC model,12 and the experimental data marked with circles in the diagrams
come from various sources.
0 20 40 60 80 100
Mass percent Mg(NO,),
Figure 1 The solubility of Mg(NO,), in water as afunction of temperature
T = 50.0"C
20 80
0
0 10 20 30 40 50 60 70 80 90 100
MgS04
Figure 2 The 50 "C solubility isotherm in the H,O-Na,SOcMgSO, system. Tie lines indicate
the three solid phases. Concentrations are in mass 96
For systems with more than four independent components, two-dimensional

phase diagrams can be made by fixing the amount of each additional component. An
isotherm for such a system will then consist of a series of phase diagrams with vary-
ing amounts of these components.
226 Chapter 19
Figure 3 Liquid-liquid equilibrium in the H,O-Na$Oct-butanol system. Calculated and

experimental tie lines
I I I I I I
I l l 1
I I I I I
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
T = 20.0"C CI- Anion Charge fradon NO;
Mass percent water given at grid intersections
Figure 4 Janecke diagram for the reciprocal salt pair KCl-NH,,NO, in water. Calculated
two-salt saturation lines and experimental two-salt saturation points
5 Density
The apparent molar volume Vp of a salt is
vp = nV- 'H20 '&O

ns
43.0
42.5
42.0
41.5
m 41.0
%
i
:::
39.5
39.0
38.5
38.0
0 10 20 30 40 50 60
sqrt (c, m0Ym3)
Figure 5 Apparent molar volume of potassium nitrate solutions calculated with Masson's
rule (22), compared with experimental data
where V is the molar volume of the solution, n is the total amount in mol, nH,ois the
amount of water, Viz0is the molar volume of pure water, and ns is the amount of salt.
The apparent molar volumes of salts are often linear functions of the square root
of the molar concentration c in the concentration range from infinite dilution to sat-
uration of the salt:28
Equation (22) is known as Masson's rule. The apparent molar volume of potas-
sium nitrate is shown in Figure 5 as a function of the square root of the concen-
tration c in mol m-3 up to saturation at 25 "C.Experimental data are plotted,
together with the line calculated from Equation (22) with a=38.13X m3 mol-'
and b=7.45x (m3 m ~ l - ' ) ~ ' ~ .
A similar relation, which can be derived theoretically from the deb ye-Hiickel the-
ory, exists for the partial molar volume of electrolyte^.^^
Apparent molar volumes of ions are often considered additive. The a and b param-
eters from Equation (22) cannot be measured for single ions, but the parameter val-
ues for salts can be split appropriately between the anions and cations of the salt if
the parameter values for one ion are fixed by convention. Unfortunately, several con-
ventions are in use.5 As data based on different conventions are incompatible, it is
important to check some reference values when using this type of data.
A good approximation for the density of multicomponent electrolyte solutions can
be obtained by adding the individual ionic contributions to the apparent molar vol-
ume:
228 Chapter 19
The density p of a multicomponent salt solution can then be calculated from
Temperature-dependent parameters for calculating the densities of binary and ter-

nary aqueous solutions of salts have been compiled by Sohnel and N o v ~ t n L . ~ ~
6 Viscosity
The viscosity of aqueous electrolyte solutions is usually described with an equation
of the following form, as a function of the concentration c in m ~ l / m ~ : ~ l
'
-= 1
'0
+ Ac'" + Bc (25)
The A parameter in this equation can be calculated t h e ~ r e t i c a l l y .The

~ ~ ,B~ ~parame-
ter is determined from experimental data. 77 is the viscosity of the solution and qois
the viscosity of the pure solvent. The B parameter is usually considered additive and
an approximation of the viscosity of a multicomponent salt solution can be obtained
by adding the contributions from all ions.
References
1. J. M. Prausnitz, R. N. Lichtentaler and E. G. Azevedo, Molecular Thermodynamics of
Fluid-Phase Equilibria, 3rd edn, Prentice-Hall PTR, Upper Saddle River NJ 07458, 1999.
2. J. M. Smith, H. C. Van Ness and M. M. Abbott, Introduction to Chemical Engineering
Thermodynamics, 6th edn, McGraw-Hill, New York, 2001.
3. D. D. Wagman, W. H. Evans, V. B. Parker, R. H. Schumm, I. Halow, S. M. Bailey, K. L.
Churney and R. L. Nuttall, J. Phys. Chem. Re$ Data, 1982, 11, Suppl. 2.
4. NIST Chemical Thermodynamics Database Version 1.1, U.S. Department of Commerce,
National Institute of Standards and Technology, Gaithesburg MD, 1990,
5. Y. Marcus, "Ion Properties", Marcel Dekker Inc. New York, 1997.
6. J. D. Cox, D. D. Wagman, V. A. Medvedev, CODATA Key Valuesfor Thermodynamics,
Hemisphere Publishing Corporation, New York, 1989.
7. R. P. Buck, S. Rondinini, A. K. Covington, F. G. K. Baucke, C. M. A. Brett, M. F.
Camdes, M. J. T. Milton, T. Mussini, R. Naumann, K. W. Pratt, P. Spitzer and G. S .
Wilson, Pure Appl. Chem., 2002, 74 (1 l), 2169.
8. P. Debye and E. Hiickel, Phys. Z., 1923,24, 185.
9. A. H. Harvey, T. W. Copeman and J. M. Prausnitz, J. Phys. Chem., 1988,92,6432.
10. C. C. Chen, H. I. Britt, J. F. Boston and L. B. Evans, AIChE J., 1982,24,588.
11. C. C. Chen and L. B. Evans, A.Z.Ch.E. J., 1986,32,444.
12. K. Thomsen, P. Rasmussen and R. Gani, Chem. Eng. Sci., 1996,51,3675.
13. K. Thomsen and P. Rasmussen, Chem. Eng. Sci., 1999,54, 1787.
14. K. Thomsen and P. Rasmussen in Steam Water and Hydrothermal Systems: Physics and
Chemistry Meeting the Needs ofhdustry, P. G. Hill, P. Tremaine, D. Irish and P. V.
Balakrishnan (eds), NRC Press, Ottawa, 2000, 118.
15. M. Iliuta, K. Thomsen and P. Rasmussen, Chem. Eng. Sci., 2000, 55,2673.
16. S . Pereda, K. Thomsen and P. Rasmussen, Chem. Eng. Sci., 2000, 55,2663.
17. M. C. Iliuta, K. Thomsen and P. Rasmussen, A.Z. Ch.E J., 2002,48,2664.
18. S. G. Christensen and K. Thomsen, Ind. Eng. Chem. Res., 2003,42,4260.
19. J. A. Myers, S. I. Sandler, R. H. Wood, Znd. Eng. Chem. Res., 2002, 41,3282.
20. K. S. Pitzer, Activity Coeficients in Electrolyte Solutions, CRC Press, Boca Raton, FL,
1991,33431.
21. P. Wang, A. Anderko and R. D. Young, Fluid Phase Equilibr. 2002,203, 141.
22. K. Thomsen, Aqueous electrolytes: model parameters and process simulation, Ph.D.
Thesis, Department of Chemical Engineering, Technical University of Denmark, 1997.
23. H. Zerres and J. M. Prausnitz, A.Z.Ch.E. J., 1994, 40,676.
24. B. Fitch, Ind. Eng. Chem., 1970, 62,6.
25. K. Thomsen, P. Rasmussen and R. Gani, Chem. Eng. Sci., 1998,53, 1551.
26. L. A. Cisternas and R. E. Swaney, Znd. Eng. Chem. Res., 1998,37, 2761.
27. Janecke, E., 2. Anorg. Chem., 1906,51, 132.
28. D. 0. Masson, Philos. Mag. and J. Sci., 1929, 8,218.
29. 0. Redlich and P. Rosenfeld, 2. physikal. Chem. Abt. A., 1931, 155,65.
30. 0. Sohnel and P, Novotny, in Densities of Aqueous Solutions of Inorganic Substances,
Physical Sciences data, vol22, Elsevier, Amsterdam, 1985.
3 1. H. D. B. Jenkins and Y. Marcus, Chem. Rev., 1995,95,2695.
32. H . Falkenhagen and M. Dole, Phys. Z., 1929, 30,611.
CHAPTER 20
Thermodynamics of
Crystallization
AMYN S. TEJA AND RONALD W. ROUSSEAU
1 Introduction
Crystallization generally involves the separation of a solute from a solution (mother
liquor) via the creation of a second (solid) phase. The solid phase may be a pure
compound, a nonstoichiometric compound, a “mixed crystal” phase, or a solid solu-
tion. The solid may be thermodynamically stable, or metastable due to slow rates of
change to more stable forms. Metastable solid phases are quite common, and can
result from surface heterogeneities as well as slow kinetics. The driving force for
crystallization is supersaturation, which is defined as the difference between the
chemical potential of the solute in the solution under the prevailing conditions, and
that in the solution at equilibrium. Although a state of equilibrium may not be
achieved in solid-liquid systems because of kinetic considerations, thermodynamics
plays a key role in determining the driving force for the crystallization process via
the determination of supersaturation. Phase diagrams provide information about the
equilibrium characteristics of a particular system and are essential to an understand-
ing of the crystallization process, The focus of what follows is therefore on phase
diagrams of solid-liquid systems, and the representation of the solid-liquid equilib-
rium curves by means of thermodynamic relationships.
2 Pure Component Phase Behavior

Phase diagrams of mixtures terminate at pure-component freezing points where the
pure solid crystallizes. A proper understanding of the solid-liquid behavior of multi-
component systems therefore requires a description of the freezing points of the pure
components. A typical pressure-temperature phase diagram for a single-component
system is shown in Figure l(a). There are three single-phase regions on this diagram,
denoted by S (solid), L (liquid), and G (gas). Equilibrium between any two phases is
represented by curves such as AD, BD, and CD. AD represents the locus of
Thermodynamics of Crystallization 23 1
Figure 1 Phase diagrams of single-component systems
solid-vapor equilibria and is the sublimation curve; BD represents the locus of

solid-liquid equilibria and is the melting/freezing curve; and finally, CD represents the
locus of vapor-liquid equilibria and is the vapor pressure curve of the substance. The
three loci meet at the triple point D, where the S, L, and G phases simultaneously coex-
ist at equilibrium. Quite often, the sublimation curve is discontinuous, as shown in
Figure l(b) where a system that exhibits two crystalline polymorphs (aand p ) is
shown. AD' and D'D in the diagram represent the sublimation curves of the two solids,
and the two curves intersect at a transition point D'. This is also a triple point because
two solids and a vapor coexist simultaneously at this point. If the liquid is cooled
slowly, the solid Sp crystallizes first and then changes to the solid S, on further cool-
ing. Several such transitions are known to exist in some substances. It is also possible
for one of the solids (Ss in Figure l(c)) to exist as a metastable polymorph. In
Figure 1(c), A'D' represents the sublimation curve of the metastable solid S p . A hypo-
thetical transition point between the two solids exists at D". However, D" is at a higher
temperature than the melting point of the solid and, therefore, cannot exist in practice.
The coexistence curves in Figure 1 can be represented by the Clapeyron equation
where (dP/dncoexrepresents the change in pressure with temperature along the coex-
istence curve, and AH, AV are the changes in enthalpy and volume due to the phase
transition. Thus, for example, the change in sublimation pressure with temperature
is related to the enthalpy of sublimation, and to the difference in volume between the
vapor and solid phases. The Clapeyron equation can be used to determine the con-
ditions for crystallization of pure substances either from the melt or via sublimation,
using appropriate values of AH and AV.
3 Binary Solid-Liquid Equilibria

Only the solid and liquid phases are of interest in most crystallization processes;
therefore, the vapor phase will generally be ignored in the following discussion.
Furthermore, pressure has little effect on the properties of condensed phases, and
will also be ignored in most of what follows.
232 Chapter 20
3.1 Systems that Form Solid Solutions

A system of two components can form solid solutions in which the atoms or molecules
of one of the species occupy substitutional or interstitial sites in the (crystal) lattice of
the other species, without substantially affecting the crystal structure of that species.
Such a system is called isomorphous because of complete miscibility in both the liquid
and solid phases. Figures 2(a) to 2(c) show the temperature-composition phase dia-
grams of such systems. Three different regions can be found in these diagrams: S, L,
and S +L. The curve ACB separating the L and S +L regions is the liquidus or freezing
point curve of the solution, and the curve ADB separating the S and S +L regions is the
solidus or melting point curve. A homogeneous liquid phase is present at all tempera-
tures and compositions above the liquidus, whereas a homogeneous solid phase is pres-
ent at all temperatures and compositions below the solidus. When subjected to cooling,
the two components are crystallized simultaneously and a homogeneous solid phase is
obtained at equilibrium conditions below the melting point curve. Any point M in the
region bounded by the liquidus and solidus curves represents a mixture of liquid C and
solid solution D. Compositions and relative amounts of the phases may be determined
by drawing horizontal tie-lines (since the temperature must be the same in both phases)
and using the lever rule. A minimum can be exhibited by the liquidus and solidus curves
as shown in Figure 2b or, in rare cases, a maximum as shown in Figure 2(c). However,
the thermodynamic description is the same for all three types of phase behavior.
The thermodynamic relationships for equilibrium between a binary solid solution
and a binary liquid solution are given by
where zi,xi are the compositions, and ys, yk are the activity coefficients of com-
ponent i in the solid and liquid phases, respectively. The fugacities of the pure solid
x or z
Figure 2 Phase diagrams of binary systems
Thermodynamics of Crystallization 233
fis and of the pure subcooled liquid fiL at the temperature and pressure of the sys-
tem may be obtained using a thermodynamic path described by Prausnitz et aL1:
where Ahp is the enthalpy change upon melting of component i at its triple point
temperature Tti,and ACpi is the difference between the heat capacities of the com-
ponent i in the liquid and solid phases. The differences in the ACpi terms are often
small in comparison with the Ahys term, and can generally be neglected.' The equa-
tion therefore simplifies to
If the solid undergoes polymorphic transitions, then one or more terms involving the
transition enthalpies and temperatures must be added. Thus
where Ahqmsis the enthalpy change of transition of component i at the transition

temperature Ti. Finally, if the pressure is large and its effects must be considered,
then a term involving the volume change must also be included:
In Equation (6), AV is the difference in volume between the liquid and solid phases, and
the integration is carried out from the triple point pressure to the pressure of the system.
Equation (2) and any one of Equations ( 3 ) to (6) may be written for each of the
two components and solved to obtain the freezing and melting curves shown in
Figures 2(a) to 2(c), provided the pure-component properties (Mys, Ah,ms, ACpi,AV,
Tti, and Ti) are available, and provided the activity coefficients f , xL can be esti-
mated using methods such as UNIFAC.2
3.2 Systems that Exhibit Eutectic Behavior

Frequently, the solidus curve in Figure 2(b) is horizontal and gives rise to two
regions in the phase diagram where a liquid solution is in equilibrium with a solid
phase (Figure 2(d)). Note that in the system shown in Figure 2(d), there is negligi-
ble solubility of solid A (S,) in solid B (S,) and of S, in S,. Horizontal line CED in
the figure represents the temperature at which solid mixtures of S, and S, melt or
crystallize. AE and BE are liquidus curves, which bound regions AEC and BED,
respectively. Solid is produced when processing (cooling or concentrating) moves a
solution within one of these regions. In region SA+L, pure solid A is in equilibrium
with liquid solutions whose compositions fall on the liquidus curve AE. Similarly, in
region S,+L, pure solid B exists in equilibrium with liquid solutions on curve BE.
Horizontal tie-lines again give the composition of the liquid in equilibrium with
234 Chapter 20
either S, or S,. Although both AE and BE are liquidus curves, AE is often called the
“solubility curve”, and BE is called the “freezing point curve”, when A is considered
to be the “solute” and B is considered to be the “solvent”. However, it should be
clear from this discussion that the terms “solute” and “solvent” depend on which
component is in excess in a solution. The liquidus curves meet at the eutectic point
E, which represents the lowest temperature at which a liquid may coexist in equilib-
rium with one or more solid phases in the system. At point E, a liquid of composi-
tion xE exists in equilibrium with two pure solids S , and S,.
The thermodynamic relationships along the liquidus curves are given by
Equation (7) is obtained from Equation ( 2 ) by noting that the solid phase is pure, and
therefore the mole fraction and activity coefficient in the solid phase are both unity.
The ratio of pure-component fugacities can be obtained from any one of Equations
x,
((3) to (6)), and the activity coefficient in the liquid, must be estimated. The com-
position-temperature behavior along the liquidus curves may then be calculated. The
eutectic point is found from the intersection of the two liquidus curves.
If the solution is dilute (i.e. the liquidus curve exists over a narrow concentration
range near the pure solvent axis), then the activity coefficient of the solute in the liq-
uid solution can be replaced by its value at infinite dilution. Furthermore, a simple
relationship may be employed for the temperature-dependence of the infinite-dilu-
tion activity coefficient as follows:
Combining Equations (3), (7), and (8) leads to expressions of the type
lnx = a + b/T (9)
that are often quite satisfactory for correlating the solubility behavior of sparingly
soluble solute^,^ as shown in Figure 3 for several amino acids.
If the two components react with one another to form one or more compounds, then
phase diagrams of the type shown in Figure 2(e) are obtained. Again, there is negli-
gible solubility of S, in S,, or of S, in S,. However, the system incorporates a reac-
tion A + B + AB. AB may be a solvate or a hydrate, for instance. In the range of
compositions between C and E in Figure 2(e), pure solid A (S,) exists in equilibrium
with a liquid solution and AE is the liquidus or solubility curve of AB in A. In the
range of compositions between E and D’, solid AB (S,) exists in equilibrium with
liquid solutions along the liquidus curve FE. FE is then the liquidus or freezing point
curve for AB in the presence of A. AB can exist by itself only at the composition rep-
resented by the vertical line FF’. This vertical line is similar, in principle, to the ver-
tical axes used to represent the two pure components. Pure AB melts at F, the
congruent melting point of S,. Similarly, curves FE’ and BE’ are the solubility and
freezing curves of AB and B, respectively. When system compositions are to the right
of F’, C’E’D is the temperature below which solids B and AB crystallize as a eutectic
-5.00 -f-
d
‘
c -5.20
L-isoleucine
-5.40
-5.60
-5.80
0.0030 0.0031 0.0032 0.0033 0.0034 0.0035

1IT(K)
Figure 3 Solubilities of a series of amino acids
mixture. The complete phase diagram may therefore be considered to be two simple
eutectic diagrams (one for the A-AB system and one for the AB-B system) joined at
the vertical line FF’.
The liquidus curves can be obtained by solving the thermodynamic relationship
already presented above, since pure solid A, pure solid B, or pure solid AB is in equi-
librium with a liquid phase in each of the instances described. The fugacity of pure
subcooled liquid AB may be obtained by considering the reaction equilibrium in the
liquid phase, so that
AG; =
fib
-RT In -
JAJB
The standard Gibbs energy may be obtained from the standard Gibbs energies of
fusion and dissolution of solid AB.
Partial miscibility in the solid phase is also possible, as shown in Figure 2 ( 0 . In
this case, there are three single-phase regions (S,, Sp, and L), and three two-phase
regions (S, + L, So+ L, and S, + S,) in the diagram. The S, phase is a solid solu-
tion rich in A, whereas the S, phase is a solid solution rich in B. The “solubility
limit” of S, corresponds to the curve ACC’. It increases with temperature to a max-
imum at C (where all Sp has melted) and then decreases until the melting point of
pure A is attained. Similarly, the curve BDD’ represents the “solubility limit” of Sp.
S, and Sp coexist at all temperatures and compositions in the region S,+SB. AE and
BE are the liquidus curves of S, and Sp and they meet at the eutectic point E. Phase
equilibrium relationships in all regions except S, + Sphave already been presented
in the previous discussion. In the region S, + S,, we have
xsazrf,as = f p ~ , B f , . ’ ~ (i = A, B) ( 1 1)
or
236 Chapter 20
These relationships may be solved at a given temperature for the phase compositions
:z and z r , provided the activity coefficients can be estimated.
4 Ternary Solid-Liquid Equilibria

4.1 Systems that Form Solid Solutions
Ternary systems of two solutes in a solvent can form solid solutions. In this case,
phase diagrams of the type shown in Figure 4(a) are obtained. Points a and b on the
diagram represent the solubilities of solute A and solute B in solvent C at a fixed
/
0
0
0
0
XC
mole fraction
Figure 4 (a) Ternary systems that form solid solutions. (b) Ternary systems in which the three
constituent binaries exhibit eutectic behavior. (c) Projection of liquidus curves in
(b). (d) Ternary system in which two of the constituent binaries form eutectics. ( e )
Ternary system in which A and B combine to form AB
C
A B
Figure 4 (continued)
238 Chapter 20
temperature (and pressure), and curve adb is an isotherm representing the solubility
at a specific temperature. At point m,a liquid of composition d is in equilibrium with
crystals of composition e. The broken lines are tie-lines at the given temperature (and
pressure).
The thermodynamics of ternary mixtures that form solid solutions may be
obtained by writing Equation (2) for the two solutes A and B. Thus, for solute A:
The pure-component fugacities may again be obtained from enthalpies of fusion and
triple point temperatures as described above. If Equation (3) is used for the fugacity
ratio, then
ZA - r,” M p
-- - -(T TtA - 1)
xA r,” RTtA
Alternatively, in the absence of enthalpies of fusion and triple point temperatures, the
solubility of pure A in solvent C may be used to obtain the fugacity ratio as described
by Teja et d4Equation (14) then becomes
Similar expressions may be obtained for solute B. A relative-distribution ratio or

“selectivity” may be defined as follows:
Note that the relative distribution of the solutes depends on nonidealities in both the
liquid and solid phases, in addition to pure-component properties. Also, if one of the
solutes is an impurity, then crystallization of the other solute from a solution may
result in crystals containing significant levels of the impurity.
4.2 Eutectic Systems

The phase diagram of a ternary system in which the three species do not form solid
solutions with each other and the constituent binary systems form eutectics, is shown
in Figure 4(b). The temperatures TA, TB, and Tc correspond to the melting points of
A, B, and C, respectively. The vertical faces of the prism represent the tempera-
ture-concentration behavior of the three binaries. Note that the behavior of each
binary system is that shown in Figure 2d. The solidus lines are not shown for the
sake of clarity. Points EAB, EAC, and EB, are the eutectic points of the three binary
systems, and point D is the ternary eutectic point. At D, a homogeneous liquid phase
coexists in equilibrium with the three pure solids. The dashed lines from the binary
eutectic points to D represent the lowering of the binary eutectic temperatures upon
the addition of the third component. Projections of the dashed lines onto the trian-
gular base are shown in Figure 4c, where the apexes of the triangle represent the pure
solids at their melting points. The binary eutectic points are indicated on the three
Thermodynumics of Crystallization 239
sides of the triangle. The three liquidus curves DE,, DE,, and DEB, meet at the
ternary eutectic point D. The temperature falls from the apexes and from the sides of
the triangle toward the eutectic point D. A few isotherms in one region of the dia-
gram are shown in Figure 4c (T,>T,>T,) in order to follow the cooling of a ternary
solution. A ternary liquid solution of composition rn starts to deposit crystals of pure
B at temperature T I .Upon further cooling, more B is deposited and the composition
changes along the line BmD' until the liquidus curve is reached at D'. A second solid
(pure C) now begins to crystallize along with B and the composition of the mother
liquor moves along D'D toward D. At the ternary eutectic point D, the third solid (A)
crystallizes along with B and C, and the system solidifies without further change in
the composition of the residual liquid.
The thermodynamic relationships for equilibrium between a pure solid and a (ter-
nary) liquid solution have already been presented in equation (7). However, the
activity coefficient is now a function of the mole fractions of the three components,
as well as the temperature. Along the liquidus curve EBCD, pure B and pure C crys-
tallize as the solution is cooled. We may therefore write
and
At any fixed temperature along the liquidus curve, we may simultaneously solve
equations (17) and (18) for the unknown compositions xB,xc, and xA (= 1-xB-xC).
The location of the ternary eutectic point (TD,xB,xc) requires the simultaneous solu-
tion of three equations such as equation (17) written for each of the three components.
Many different types of phase diagrams are possible in a ternary system. Figure
4d shows a system in which there are two eutectics in systems A+C and B+C (a
system of two salts and water, for example). The two solids (salts) are completely
immiscible, and there are therefore four regions on the phase diagram, labeled L,
L+SA, L+SB, and L+SA+SB. The ternary eutectic and the liquid curves are labeled
as in previous figures. Figure 4e shows a system in which A and B react to form a
new compound F, which is comprised of A and B in a specific stoichiometric ratio.
Such a species is referred to as a double salt. This leads to six regions on the phase
diagram labeled L, L+SA, L+SB, L+Sp L+SA+SF, and L+SF+S,. The fugacity of
the compound F is obtained by considering its reaction equilibrium using relation-
ships such as Equation (10). The other equilibrium equations are similar to those
described above. Thus, many of the phase diagrams for ternary and higher mixtures
can be calculated using the equations presented in this work.
5 Solid-Fluid Equilibria
Crystallization from the gas phase, and in particular from a gas phase above the crit-
ical temperature and pressure of the solution, has attracted much interest in the lit-
erature for making particles of uniform size via a process known as supercritical
240 Chapter 20
P
Figure 5 The solubility of a nonvolatile solid in a supercritical fluid, as a function of pres-
sure at two temperatures close to the critical temperature of the system
cry~tallization.~ Generally, a change in pressure is employed to bring about a large

change in solubility, and hence in supersaturation, of a nonvolatile solute. Since
pressure plays a significant role in these crystallizations, it is instructive to outline
the thermodynamic relationships employed in the calculation of solid-supercritical
fluid equilibria.
Figure 5 shows a typical solubility vs. pressure curve for a nonvolatile solute such
as naphthalene in a solvent such as carbon dioxide. As the pressure increases from a
low value, the solubility first decreases and then dramatically increases before it
attains an approximately constant value. The dramatic increase in solubility occurs
near the critical point of carbon dioxide, so that a small change in pressure near the
critical point can lead to large changes in solubility. Large supersaturations can there-
fore be generated by small changes in pressure near the critical point of the solvent.
The thermodynamic relationships for a pure solid solute A (assuming that the gas
does not dissolve in the solid) in equilibrium with a supercritical solution is given by
where $A is the fugacity coefficient of the solute in the supercritical solution, yA is

its solubility, and P is the system pressure; fiSis the fugacity of the pure solid solute
at the temperature and pressure of the system and may often be replaced by a prod-
uct of its sublimation pressure and a Poynting correction for pressure. Thus,
The fugacity coefficient qAis generally calculated from an equation of state. However,
many equations of state require a knowledge of the critical parameters of the solute,
which may not always be available. Nevertheless, solubilities can be correlated, and
sometimes extrapolated, using this approach. The addition of more solutes poses few
problems from a thermodynamic viewpoint, as long as the appropriate solid-state
fugacity is used in the calculations. This type of approach may also be used to study
Thermodynamics of Crystallization 24 1
the effect of co-solutes and co-solvent on the solubility behavior. Alternatively, an
approach based on the dilute solution theory6 may be employed for this purpose.
6 Kinetics of Dissolution and Crystal Growth

Crystal nucleation and growth are key factors in determining characteristics of crys-
talline products that may determine their usefulness or even the ability to separate
them from their mother liquor. The kinetics of these phenomena and of dissolution
are related to the difference between the values of the system variables and the Val-
ues of those variables at equilibrium. In crystallization, this difference is referred to
as supersaturation and it is given by the expression
where p is the chemical potential of the crystallizing species at the prevailing sys-
tem conditions, and p* is the equilibrium value at the system temperature and pres-
sure. Common practice has resulted in the use of more easily expressed driving
forces for the kinetic phenomena, often without a sound basis; among these are the
following:
concentration difference: Ac = c - c* (21a)
temperature difference: AT = T* - T (21b)
saturation or supersaturation ratio: S = c/c* (2 1c)
relative supersaturation: CT = ( c - c*)/c*= S - 1 (21d)
where the asterisk (*) designates the value at equilibrium.

The relative supersaturation has special utility because it approximates the rigor-
ous definition of supersaturation in Equation (21), which is related to solute con-
centrations by the expression
AP
RT =In-
YC
y*c*
For ideal solutions, where both yand y* are approximately unity, or for conditions
that lead to the approximation y= f ,
AP =In-;-c = l n S
-
RT c (23)
When the supersaturation ratio, S, is small (say less than l.l), it follows that
For this reason, then, kinetics of nucleation, growth, and dissolution are often
expressed in terms of relative supersaturation. The actual relationships for these
242 Chapter 20
kinetic phenomena depend on the controlling mechanism and are beyond the scope
of the present d i s c u s ~ i o nHowever,
.~~~ the importance of the solubility is clear from
the definition of supersaturation.
7 Summary
This chapter outlines thermodynamic relationships and phase diagrams relevant to
the crystallization of relatively simple molecules. In particular, electrolytes and
mixed solvents have not been treated, although many features of their phase dia-
grams are similar to those described here. Specialized books and monographs should
be consulted for further information on all t~pics.~-l*
References
1. J. M. Prausnitz, R. N. Lichtenthaler and E. G. Azavedo, Molecular Thermodynamics of
Fluid Phase Equilibria, 3rd edn, Prentice-Hall, Upper Sadle River, NJ, 1999.
2. A. Fredenslund, J. Gmehling and P. Rasmussen, Vapour-Liquid Equilibria Using UNZ-
FAC, Elsevier, Amsterdam, 1977.
3. J. C. Givand, A. S. Teja and R. W. Rousseau, A.Z.Ch.E. J., 2001,47,2705.
4. A. S. Teja, J. C. Givand and R. W. Rousseau, A.1.Ch.E. J., 2002,48, 2629.
5. A. S. Teja and T. Furuya, Encyclopedia of Separation Science, Vol 111, Academic Press,
London, 2000.
6. J. Mendez-Santiago and A. S. Teja, Z & E C Res., 2000,39,4767.
7. J. W. Mullin. Crystallization, 4th edn, Butterworth-Heinemann,Oxford, 2001.
8. R. W. Rousseau, Kirk-Othmer Encyclopedia of Chemical Technology, 4th edn, Vol 7,
Wiley, NJ, 1993.
9. W. C. Blasdale. Equilibria in Saturated Salt Solutions. ACS Monograph Series, Chemical
Catalog Co., New York, 1927.
10. R. Haase and H. Schonert (translated by E. S. Halberstadt), Solid-liquid Equilibrium,
Pergamon Press, London, 1969.
11. K. S . Pitzer (ed). Activity Coeficients in Electrolyte Solutions, 2nd edn, CRC Press, Boca
Raton, FL, 1991.
12. G. T. Hefter and R. P. T. Tompkins (eds), The Experimental Determination of Solubilities,
Wiley, Chichester, UK, 2003.
CHAPTER 21
Thermodynamics of Adsorption
ALAN L. MYERS
1 Introduction
The attachment of molecules to the surface of a solid by adsorption is a broad subject.
This chapter is focused on the adsorption of gases in high-capacity solid adsorbents
such as active carbon or zeolites.2These commercial adsorbents owe their enormous
capacity to an extensive network of nanopores of various shapes (cylinders, slits) with
specific volumes in the range from 100 to loo0 cm3 kg-'. Applications of adsorption
exploit the ability of nanoporous materials to adsorb one component of a gas preferen-
tially. For example, the preferential adsorption of nitrogen from air passed through an
adsorption column packed with zeolite creates a product stream of nearly pure oxygen.
Thermodynamics has the remarkable ability to connect seemingly unrelated prop-
erties. For example, the temperature coefficient of adsorption is directly proportional
to the heat of immersion of the solid adsorbent in the gas. The most important appli-
cation of thermodynamics to adsorption is the calculation of phase equilibrium
between a gaseous mixture and a solid adsorbent.
The basis for thermodynamic calculations is the adsorption isotherm, which gives
the amount of gas adsorbed in the nanopores as a function of the external pressure.
Adsorption isotherms are measured experimentally or calculated from theory using
molecular simulation^.^ Potential functions are used to construct a detailed molecu-
lar model for atom-atom interactions and a distribution of point charges is used to
reproduce the polarity of the solid material and the adsorbing molecules. Recently,
ab initio quantum chemistry has been applied to the theoretical determination of
these potentials, as discussed in another chapter of this book.
Thermodynamics applies only to equilibrium adsorption isotherms. Equilibrium
means that any point can be reached from either direction by raising (adsorption) or
lowering (desorption) the pressure at constant temperature. If the desorption
isotherm does not coincide with the adsorption isotherm, then equilibrium has not
been achieved and the usual thermodynamic equations do not apply. The mismatch
of adsorption and desorption, which is called hysteresis, does not occur in pores
smaller than 2 nm but is observed' when the pores are large enough for the adsorb-
ing molecules to condense to a liquid. For adsorption of supercritical gases or for
244 Chapter 21
adsorption of subcritical vapors in nanopores, most experiments and simulations
yield equilibrium isotherms with no evidence of hysteresis.
Molecular simulations yield absolute adsorption or the actual number of mole-
cules in the nanopores. Experiments measure excess adsorption, which is the num-
ber of molecules in the nanopores in excess of the amount that would be present in
the pore volume at the equilibrium density of the bulk gas. The difference between
absolute and excess adsorption is negligible at the sub-atmospheric pressures of
greatest interest. For supercritical gases adsorbed at high pressure (e.g. 100 bar), the
difference between absolute and excess adsorption is too large to i g n ~ r e . ~
2 Adsorption Isotherm and Equation of State

Whether the adsorption isotherm has been determined experimentally or theoreti-
cally from molecular simulation, the data points must be fitted with analytical equa-
tions for interpolation, extrapolation, and for the calculation of thermodynamic
properties by numerical integration or differentiation. The adsorption isotherm for a
pure gas is the relation between the specific amount adsorbed n (moles of gas per
kilogram of solid) and P, the external pressure in the gas phase. For now, the dis-
cussion is restricted to adsorption of a pure gas; mixtures will be discussed later. A
typical set of adsorption isotherms is shown in Figure 1. Most supercritical
isotherms, including these, may be fit accurately by a modified virial e q ~ a t i o n . ~
r
2
,-
h
c
(
g
1
0
0 40 80 120
PkPa
Figure 1 Adsorption isotherms of C,H, in NaX (zeolite structure FAU), where n is the-amount
absorbed and P is the pressure. Points indicate experimental data.6Solid lines indi-
cate Equation ( 1 )
Thermodynamics of Adsorption 245
where K is the Henry's constant, or the slope of adsorption isotherm dn/dP at the
limit of zero pressure, rn is the saturation capacity (mol kg-I), and the Ciare virial
coefficients, three of which usually suffice to fit the data within experimental error.
If the virial coefficients are all zero, Equation (1) reduces to the well-known
Langmuir equation.' Equation (1) has the form P(n) so that the inverse function n(P)
is implicit. This slight inconvenience is offset by the fact that the implicit form can
be integrated analytically for the thermodynamic functions (see below).
The determination of an accurate value for Henry's constant ( K ) is essential for
the calculation of thermodynamic properties and for mixture calculations. According
to Equation (l), a plot of ln(P/n) as a function of n intersects the y-axis at 1/K. If the
scatter of the data at low pressure is so large that an accurate value of K is impossi-
ble to determine, then Henry's constant should be measured at a higher temperature.
It is difficult to obtain reliable values of Henry's constants from gravimetric meas-
urements because the amount of gas adsorbed at low pressure is given by the differ-
ence of two weight measurements that differ by an infinitesimal amount. Volumetric
measurements are preferred for measuring Henry's constants because the amount of
gas adsorbed is determined by the large difference between the amount of gas dosed
to the system and the amount of gas left in the system after adsorption.
Interpolation of adsorption isotherms with respect to temperature is based on the
thermodynamic e q ~ a t i o n : ~
where h is the differential enthalpy of adsorption, a negative quantity because adsorp-

tion is exothermic. The absolute value of h is called "isosteric heat". The partial differ-
entiation is performed at constant n. In the rigorous equation, the pressure P is replaced
by the fugacity of the gas. The differential enthalpy may be expressed as a polynomial:
-
h(n) = D o + D,n -k D,n2 + D3n3+ (3)
The constants Diare assumed to be independent of temperature. For wide variations

in temperature over several hundred degrees Kelvin, this approximation can be cor-
rected by introducing heat capacities. The integrated form of Equation (2) is
which provides the temperature dependence P(T) given a reference point P*(T")
measured at the same value of n. Combination of Equations (1) and (4) yield an
adsorption equation-of-state, which includes the temperature variable:
x exp[C,n + C2n2+ C3n3+ - - * J (5)
where the constants K , m, and the Cjrefer to the reference isotherm at T * .The constants
of Equation (5) for the adsorption isotherms in Figure 1 are: T*= 298.15 K, K= 1.9155
246 Chapter 21
0.1 1 10 100
PMa
Figure 2 Adsorption isotherms of C,H, in NaX (zeolite structure FAU), where n is the amount
adsorbed and P is the pressure. Points indicate experimental data.6 Solid lines cal-
culated from Equation (5)
mol kg-' kpa-', rn=2.9997 mol kg-', C,=0.841 kg mol-', C2= -0.0631 1 kg2 mol-2,
C3= -0.009415 kg3m ~ l - Do= ~ , -39.5 W mol-', andD1=2.25 kJ kg molP2.The exper-
imental data are compared with Eq. (5) in Figure 2, which includes interpolated and
extrapolated isotherms. Logarithmic plots are useful for examining the accuracy of the
equation-of-stateat low pressure. The calculation of enthalpy, free energy, and entropy
from these constants is explained in the next section.
Usually, the differential enthalpy is determined from Equation (2) using two or
more adsorption isotherms. Alternatively, the differential enthalpy can be measured
directly using a calorimeter.8In either case, a reference isotherm should be measured
for the lowest temperature at which an accurate value of the Henry constant can be
extracted. In the example shown in Figure 1, the reference isotherm is at 25 "C. For
a particular gas and solid, the combination of a reference isotherm with the differ-
ential enthalpy provides complete thermodynamic information about the system.
3 Thermodynamic Functions
The grand potential plays a central role in adsorption thermodynamics. The grand
potential is defined by
where F is the Helmholtz free energy. The independent variables of the grand potential
are temperature, volume, and chemical potential. These variables are precisely the ones
needed to describe the amount adsorbed from a bulk gas at specified values of temper-
ature and chemical potential in a solid adsorbent of fixed volume. For the same reason,
molecule simulations of adsorption are conveniently performed in the grand canonical
ensemble for which a=-kT In E,where Z is the grand canonical partition fun~tion.~
For adsorption of a pure gas, the grand potential is obtained from an isothermal
integrati~n:~
R is expressed in J kg-' of solid adsorbent. Physically, the grand potential is the free
energy change associated with isothermal immersion of fresh adsorbent in the bulk
fluid. The absolute value of the grand potential is the minimum isothermal work nec-
essary to clean the adsorbent. Since adsorption occurs spontaneously, the cleaning or
regeneration of the adsorbent after it equilibrates with the feed stream is the main
operating cost of an adsorptive separation process.
Any extensive thermodynamic property of the system (free energy, enthalpy,
entropy, or heat capacity) may be written as the sum of three terms for:
1. the value of the property for the adsorbate molecules at the state of the equili-
brated bulk gas mixture at { T, P, yi};
2. the value of the property for the clean solid adsorbent in vacuo at T; and
3. the change in the property associated with immersion of the clean adsorbent in
the bulk gas at constant { T, P , y i } .
The thermodynamic functions for items 1 and 2 are calculated using the standard
equations for bulk gases and solids, respectively,1° so that the focus for adsorption
thermodynamics is on item 3. It follows from Equations (5) and (7) that the grand
potential (free energy of immersion) for each pure component is
2 3
+ -C2n3
3
+ -C3n4 + ...
4
2 3
+ -D,n3 + -D3n4 +
+-
[f
- - -:](kDln2
3 4
- - a
The constants rn and Cirefer to the values for the reference isotherm at T";the con-
stants Di refer to the polynomial for the differential enthalpy in Equation ( 3 ) . Note
that the free energy is independent of the limiting value of the enthalpy at zero pres-
sure, Do in Equation (3).
The enthalpy of immersion (H) is the integral of the differential enthalpy (h):
H = /:%dn (9)
The enthalpy of immersion, like R, has units of J kg-'. From Equations ( 3 ) and (9):
1 1 1
H(n) = Don +-
2
Dln2 + -DD,n3 + -DD,n4
3
4 + ... (10)
It is convenient to report the enthalpy of immersion as an integral molar enthalpy (J

mol- ') using h= Hln:
1 1 1
h(n) = Do + -Dln
2
+ -DD,n2
3
+ -D3n3
4
+
248 Chapter 21
Given the free energy of immersion ( R ) and the enthalpy of immersion (H),the
entropy of immersion is
H-R
s= -
T
4 Mixtures
The grand potential provides the standard state for the formation of adsorbed solu-
tions from the pure components. Given the pressure ( P ) , temperature (0,and mole
fraction of component 1 in the gas phase (yl) for a binary mixture, three equations
are solved simultaneously7 to establish the amounts adsorbed (ny,ni) at the standard
state:
Thus, the partial pressures on the left-hand side of Equations (13) and (14) are
known and the three unknowns are ny, n;, and xl, where x2= 1 -xl. For mixtures con-
taining more than two components, each additional component adds two equations
and two unknowns (np and xi). In the rigorous form of Equations (3, (7), and
(13)-( 1 9 , the pressure or partial pressure is replaced by the fugacity.
Given the adsorbed-phase composition x 1 from the solution of Equations
(13)-( 15), the selectivity of the adsorbent for component i relative to componentj is
The larger the selectivity, the easier the separation of component i from component
j by adsorption. Zeolites with a selectivity as high as 10 for nitrogen relative to oxy-
gen are used in pressure-swing adsorption processes” to produce oxygen from air.
The specific amount of each component adsorbed for an ideal solution is given by
ni=n,xi (17)
where the total specific amount adsorbed from a mixture of gases is

1
--
n, -?$ (18)
In summary, the procedure for predicting the thermodynamic properties of an

adsorbed mixture begins with the determination of the thermodynamic properties of
each individual component as expressed by its equation of state, Equation (5). After
fixing the independent variables { T, P , yi} for a system containing Nc components,
the set of (2Nc- 1) Equations (13)-(15) is solved for the adsorbed-phase mole frac-
tions xi and standard-state amounts adsorbed (np),with the constraint that x i x i = 1.
Knowledge of the standard states and the adsorbed-phase composition allows the
calculation of the selectivity by Equations (16) and the amount of each species
adsorbed by Equations (17) and (1 8). Finally, the entropy and enthalpy of immersion
are given by the equations:
H; and S; are evaluated at the standard-state amount adsorbed (n;). It may seem at
first glance that an entropy of mixing term is missing from Eq. (20), but S refers to
the entropy of immersion of the solid in the gas mixture. The total entropy of the
adsorbate mixture relative to its pure, perfect-gas reference state includes a separate
term for mixing and compressing the adsorbate gas to its equilibrium state { T, P , y i }.
The integral enthalpy H of the mixture divided by the total amount adsorbed is the
integral molar enthalpy h, as in Equation (1 1) for adsorption of a single component.
5 Example
The application of Equations (13)-(20) is illustrated for binary mixtures of ethylene
(1) and ethane (2) adsorbed on NaX zeolite (faujasite). The constants for the single-
gas adsorption equations of state5are given in Tables 1 and 2. The selectivity of NaX
for ethylene relative to ethane is a function of temperature, pressure, and the
composition of the gas. The selectivity at constant temperature (20°C) is shown in
Figure 3. The selectivity at the limit of zero pressure is the ratio of Henry's constants
(K,/K2=33.7). At constant mole fraction of ethylene in the gas, the selectivity
decreases rapidly with increasing pressure. At constant pressure, the selectivity
decreases with increasing mole fraction of ethylene in the gas. The selectivity at con-
stant pressure and gas composition decreases with temperature, as shown in Figure
4. Decrease of the selectivity with temperature, pressure, and the mole fraction of the
preferentially adsorbed species is typical behavior for binary adsorption.
Table 1 Constants of Eq. ( 1 ) for reference adsorption isotherms of gases in NaX

zeolites at 293.15 K. Virial coeficients Ci in units of kg' mol-'
Gas K (mol kg-' kPa-') m (mol kg-') C, c
2 c.3 c
4
C2H, 5.2039 4.5341 0.385 0.0075 0.0012 0.0012

C2H6 0.1545 3.8937 -0.267 -0.0499 0.0192 0.0
Table 2 Constants of Eq. (3)for diflerential enthalpy (isosteric heat) of adsorption

of gases in NaX zeolite5 at 298.15 K. Mrial coeficients Di in units of W kgi
mol- ( i + 1)
Gas Do 4 D2 D3 D4
C2H4 -41.836 0.3215 - 1.2203 0.9452 -0.1576
C2H6 -26.893 -1.1719 0.0328 -0.1195 0.0
250 Chapter 21
0 0.2 0.4 0.6 0.8 1

Y1
Figure 3 Selectivity (xIy2)/(x2y1)

for adsorption of ethylene (1) relative to ethane (2) in NaX
(zeolite structure FAU) at 20"C,plotted against y,, the mole fraction of C,H, in the
gas. Isobars calculated from Equations (13)-(16)using the constants for pure gases
in Table 1
0 20 40 60 80 100
TIT
Figure 4 Selectivity (xly2)/(xg,)for adsorption of ethylene (1)relative to ethane (2)in NaX (zeo-
lite structure FAU) at 100 kPa and y,=O.l, plotted against the temperature. Calculated
from Equations (13)-(16) using the constantsfor pure gases in Tables 1 and 2
The selectivities in Figures 3 and 4 were calculated from the single-gas isotherms
using Equations (1 3) and (14), which are written for ideal adsorbed solutions (IAS)
with activity coefficients ~ = 1 These
. equations are rigorous at the limit of
low pressure. At high pressure, mixtures adsorbed in nanopores display negative
Thermodynamics of Adsorption 25 1
7 3
3
4
2
;*
1
0
4
I
4
7
x
M
24
3 2
E
\
I: P=lOkPa
1
C2H6
0
2
-I
w
24
“01
E.
-
I: P=lkPa
a I
I 2 0.4 0.6 0.8 0
Y1
Figure 5 Individual and total isotherms at 20 “C for isobaric adsorption of mixtures of
ethylene (1) and ethane (2) in NaX (zeolite structure FAU), where n is the amount
adsorbed and y l is the mole fraction of ethylene in the gas. Dashed lines calculated
from Equations (13)-(15), (17) and (18) using the constants for pure gases in
Table 1. Solid lines indicate experimental datas
252 Chapter 21
Figure 6 Enthalpy for isobaric adsorption of mixtures of ethylene ( I ) and ethane (2) in NaX,
where h is the integral enthalpy and x, is the mole fraction of ethylene in the
nanopores. Dashed lines calculated from Equations (13)-(15) and (19) using con-
stants for the pure gases in Tables 1 and 2. Solid lines indicate experimental data.’
At 1 kPa, the dashed and solid line coincide
deviations from Raoult’s law (x< 1). These deviations are dominated by hetero-
geneity of the gas-solid energy and therefore cannot be estimated from the activity
coefficients of the bulk fluids. The strongest deviations from ideality are observed
for mixtures in zeolites such as NaX (faujasite), which has strong electric fields and
electric field gradients in its nanopores that interact differently with quadrupolar
(C,H,) and nonpolar molecules (C,H,). Mixtures adsorbed in materials with weak
electric field gradients such as silicalite (MFI structure) or active carbon are more
nearly ideal ( p l ) than zeolites like NaX, which contain exchangeable nonframe-
work cations.
Activity coefficients for nonideal mixtures have been r e p ~ r t e dThe
. ~ error associ-
ated with the use of IAS theory is shown in Figure 5. The solid lines are the experi-
mental data and the dashed lines were calculated from Equations (13)-( 18). The
comparison of the IAS prediction with experimental data in Figure 5 raises the fol-
lowing question: is it possible to predict activity coefficients? Correlations of activ-
ity coefficients with single-gas adsorptive properties5 suggest that such predictions
are possible, and reliable methods may be discovered in the future.
The estimate of the integral enthalpy ( h )by Equation (19) is shown by the dashed
lines in Figure 6. The solid lines are the experimental data determined by calorime-
try.5 The error in the estimated enthalpy (dashed lines) increases with pressure but
the largest error is 1.6%.The values at the two end points (yl=O and y,= 1) are the
integral enthalpies for pure ethylene and ethane given by Equation (1 1).
6 Summary
Equation (5) is an equation-of-state for the adsorption of a pure gas as a function of
temperature and pressure. The constants of this equation are the Henry constant, the
saturation capacity, and the virial coefficients at a reference temperature. The
temperature variable is incorporated in Equation (5) by the virial coefficients for the
differential enthalpy. This equation-of-state for adsorption of single gases provides
an accurate basis for predicting the thermodynamic properties and phase equilibria
for adsorption from gaseous mixtures.
References
1. D. M. Ruthven, Principles of Adsorption and Adsorption Processes, John Wiley & Sons,
New York, 1984,7, 50, 56.
2. http://www.iza-structure.org/databases
3. D. Nicholson and N. G. Parsonage, Computer Simulation of the Statistical Mechanics of
Adsorption, Academic Press, London, 1982
4. A. L. Myers, P. A. Monson, Adsorption in porous materials at high pressure: theory and
experiment, Lungmuir, 2002,18, 10261-10273.
5. F. R. Siperstein and A. L. Myers, Mixed-gas adsorption, A.Z.Ch.E.J., 2001, 47,
1141-1 159.
6 . S. H. Hyun and R. P. Danner, J. Chem. Eng. Data, 1982,27, 196.
7. A. L. Myers, Thermodynamics of adsorption in porous materials, A.I.Ch.E. J., 2002, 48,
145-1 60.
8. J. A. Dunne, R. Mariwala, M. Rao, S. Sircar, R. J. Gorte and A. L. Myers, Calorimetric
heats of adsorption and adsorption isotherms. 1. O,, N,, Ar,CO,, CH,, C&, and SF6 on
silicalite, Lungmuir, 1996, 12, 5888-5895.
9. D. A. McQuarrie, Statistical Mechanics, Harper & Row, New York, 1976, p. 51.
10. J. M. Prausnitz, R. N. Lichtenthaler and E. G. de Azevedo, Molecular Thermodynamics
of Fluid-Phase Equilibria, 3rd edn, Chapter 3, Prentice-Hall, Upper Saddle River, New
Jersey, 1999.
11. D. M. Ruthven, S. Farooq and K. S. Knaebel, Pressure Swing Adsorption, John Wiley &
Sons, New York, 1993.
CHAPTER 22
Mesoscopic Non-equilibrium
Thermodynamics of Polymer
Crystallization
DAVID REGUERA AND J.M. RUB1
1 Non-Equilibrium Thermodynamics and the

Mesoscopic Level of Description
In 193 1, Onsager published two papers on simultaneous irreversible processes' in
which he formulated the well-known reciprocal relations between phenomenologi-
cal (Onsager) coefficients establishing in this way the basis for the formulation of a
thermodynamical theory of irreversible processes, referred nowadays to as non-equi-
librium thermodynamics.* The success of the theory is mainly due to the simplicity
of its method, which runs parallel to that of equilibrium thermodynamics, to the
many situations pertaining to the macroscopic world to which it applies, and to its
solid connection with statistical mechanics and kinetic theory. Irreversible processes
as for example heat conduction, mass diffusion and electrical phenomena, which
constitute fundamental mechanisms of transport in bulk phases and at interfaces3 and
are the basis of many industrial processes, can convincingly be explained under the
framework of the theory. Non-equilibrium thermodynamics nowadays constitutes a
powerful tool with many applications in physical, physicochemical and biological
systems as well as in engineering processing.
1.1 Meso-Structures
The scope of non-equilibrium thermodynamics is restricted to irreversible processes
taking place in macroscopic systems, i.e. at length scales much larger than any
molecular characteristic length. The continuum hypothesis, for which matter is
assimilated to a continuum medium, invoked to set up a local description of the sys-
tem, ignores its molecular nature. Although this approach may work out extremely
well at long length scales, when reducing the observation scales focusing on regions
Mesoscopic Non-equilibrium Thermodynamics of Polymer Crystallization 25 5
whose energy is not drastically much larger than thermal energy, the molecular
nature of the system becomes unavoidable manifest.
The more experimental accessibility to shorter and shorter scales and the interest
in elucidating the intimate nature of materials and the elementary mechanisms gov-
erning irreversible processes make the formulation of a non-equilibrium thermody-
namics valid at the mesoscopic level a task of primary importance. Many interesting
irreversible processes in nature as for example nucleation, growth and agglomeration
phenomena, which constitute basic mechanisms in industrial applications in metal-
lurgy, food industry and material science, occur at the mesoscale. In these processes,
the starring entities (such as the emerging crystals, droplets or aggregates) are meso-
structures, which are neither particles nor macroscopic objects. They undergo gen-
uine meso-scale processes as assembling and impingement and they are affected by
the fluctuations of the thermal bath in which they are embedded and by an external
driving as is the case in industrial processing.
1.2 Mesoscopic Non-Equilibrium Thermodynamics (MNET)

It has been shown that the scheme of non-equilibrium thermodynamics can be
extended to describe irreversible processes occurring at the mes~-scale.~*~ This gen-
eralization entails the formulation of the local equilibrium hypothesis, the comer-
stone of non-equilibrium thermodynamics, in a more general context. The local state
of the system has to incorporate the variables that have not yet relaxed toward equi-
librium at mesoscopic scales. Starting from the corresponding Gibbs equation that
incorporates those degrees of freedom, one then obtains the entropy production and
from it the expression for the currents in terms of the thermodynamic forces. As a
result, in general, one obtains expressions for the currents in the space of the meso-
scopic degrees of freedom of the system and the kinetic equations describing the
evolution of the probability density in that space.
In this paper, we will illustrate the application of mesoscopic non-equilibrium
thermodynamics to the dynamics of activated processes. Together with transport
process, they constitute the two basic mechanisms for which a system can be
removed from an equilibrium state. Although both processes act in the sense of
yielding non-equilibrium situations, their natures are markedly different. Whereas
transport processes always exhibit a linear regime that manifests at small values of
the driving force, activated processes are genuinely non-linear since a certain
amount of energy, not necessarily small, has to be given to the system in order that
it may overcome the free energy barrier to change its state. Non-equilibrium ther-
modynamics cannot account for the kinetic of such non-linear processes.
For a chemical reaction, for example, non-equilibrium thermodynamics formu-
lates a linear relationship between the reaction rate and the affinity, which constitute
only the first term in the development of the law of mass action. To obtain the full
law, one has to take into account not only the initial and final states of the kinetics
but all intermediate configurations, i.e. one has to introduce a mesoscopic degree of
freedom accounting for the different molecular configurations. When this is done, in
the framework of mesoscopic non-equilibrium thermodynamics, one arrives at the
law of mass action governing the kinetics for arbitrary values of the thermodynamic
256 Chapter 22
force. To illustrate the use of the theory, we will focus, in particular, on polymer
crystallization, a basic mechanism in many industrial applications.
2 Polymer Crystallization
Crystallization of polymers is a process for which structural reorganization of a
melted polymer leads to the emergence of an ordered phase. The transition from
melted to crystalline polymer first occurs at the melting temperature. The crystalline
phase manifests in the temperature range comprised between the melting tempera-
ture and the glass transition temperature. For temperatures higher than the melting
temperature, thermal agitation impedes the formation of an ordered phase, whereas
for temperatures smaller than the glass transition temperature, diffusion of the
monomers to form a crystal is prevented.
We can schematically distinguish between three stages in the entire crystallization
process: nucleation, growth and secondary crystallization. At the earliest stages,
crystals are formed from the melted phase via a nucleation process. The presence of
density fluctuations may eventually create small aggregates of polymers having the
same properties as the crystalline phase. The small crystals are continuously being
created and destroyed by fluctuations. The appearance of a small crystal involves
competition between two effects. One is the decrease of energy due to the fact that,
below the melting temperature, the chemical potential of the crystal phase is lower
than that of the melt, which competes with the energetic cost due to the surface ten-
sion associated with the creation of an interface. For clusters of small size, surface
effects are dominant. As a consequence, their growth is energetically unfavorable
and the small crystals tend to spontaneously dissolve. There exist, however, a criti-
cal size beyond which the volume effects become dominant, and the growth of the
cluster is favored by a global reduction of the energy. In the second stage, or growth,
once formed, crystals grow freely until they progressively begin to compete to fill
the whole space. They may eventually hit each other, and therefore the growth is
stopped at the contact surface. This phenomenon, called impingement, is relevant at
later stages of the process and determines the final morphology of the system. In the
final stage of crystallization, also known as secondary crystallization, amorphous
non-crystallized matter that remains trapped among clusters can eventually self-
organize, joining to the main crystal structure and therefore increasing the degree of
crystallinity of the sample.
In quiescent homogeneous melts, crystallization proceeds by the formation of
isotropic quasi-spherical clusters known as spherulite. Figure 1 illustrates the final
structure of a crystallized polymer (PHB), showing clearly the different spherulites
originating from nucleation, whose growth competing to fill the space was stopped
by impingement.
The usual theoretical framework to describe this process is the Avrami- Kolmogoroff
based on purely stochastic grounds, and aimed at describing the space filling
in the crystallization kinetics. The underlying physical ingredients of this model are the
nucleation and growth rates. This highlights the importance of providing an accurate
theory to obtain nucleation and growth rates that ideally may incorporate the influences
of the host melted phase and also the peculiarities of the processing of the melt.
Mesoscopic Non-equilibrium Thermodynamics of Polymer Crystallization 257
Figure 1 Crystal structure of PHB grown at 5.5" Cfrom Ref: 10)
3 The Mesoscopic Non-EquilibriumThermodynamics

Approach to Polymer Crystallization
Nucleation and growth are the most important processes in determining the final
morphology of the crystallized polymer and consequently the properties of the mate-
rial. Both processes can be quite conveniently described within the framework of
mesoscopic non-equilibrium thermodynamics.
Crystallization of polymers is an activated process that may be influenced by the
presence of mechanical stresses and external gradients. Clusters are formed at the
expense of the metastable phase due to the presence of thermal fluctuations. The
process can be modeled by a particle crossing a free energy barrier, whose height
depends on the interactions and the external drivings, upon the intervention of
Figure 2 Schematic representation of crystallization as a difSusion process over a free

energy barrier
258 Chapter 22
fluctuations. The states at both sides of the barrier correspond to the metastable melt-
ing phase and to the crystallized phase (see Figure 2).
3.1 Mesoscopic Non-Equilibrium Thermodynamics of Activated

Processes
Mesoscopic non-equilibrium thermodynamics provides a description of activated
processes. In the case considered here, when crystallizationproceeds by the formation
x
of spherical clusters, the process can be characterized by a coordinate which may
represent for instance the number of monomers in a cluster, its radius or even a global-
order parameter indicating the degree of crystallinity. Polymer crystallization can be
viewed as a diffusion process through the free energy barrier that separates the melted
phase from the crystalline phase. From mesoscopic non-equilibrium thermodynamics
we can analyze the kinetic of the process. Before proceeding to discuss this point, we
will illustrate how the theory applies to the study of general activated processes.
Since activation kinetics is affected by the presence of fluctuations in the
metastable phase, we will provide a description of the system in terms of the proba-
bility distribution function P(xt). Our task will be then to derive the expression of
the activation current, a quantity accessible to experimentsin many instances. To this
end, our starting point is the expression for the Gibbs equation accounting for
entropy variations due to the underlying diffusion process of the probability density
(1)
where P ( x t ) is the probability density, p(xt) is a non-equilibrium chemical potential

defined in yspace and Tis the temperature of the system assumed to be constant. In the
absence of impingement effects, the chemical potential is given through the expression
Here, H(33is the free energy barrier and k is Boltzmann’s constant. The presence of
impingement effects can be taken into account by introducing an activity coefficient
in the chemical potential as done for non-ideal mixtures.
The evolution of the probability density is governed by the continuity equation
where J is the activation current defined in yspace, which has to be determined. Its
value can be obtained from the entropy production CJ = dS/dt, which follows from
the Gibss equation, by taking the time derivative and then using the continuity equa-
tion. Due to the fact that the definition domain of yis bounded, the current has to dis-
appear at the initial and final states of the process. After performing a partial
integration, one then arrives at
Mesoscopic Non-equilibrium Thermodynamicsof Polymer Crystallization 259
The integral consists of contributions of flux-force pairs and then has the habitual
form of the entropy production in non-equilibrium thermodynamics. Also assuming
linear dependence between fluxes and forces, one establishes the relations
where L is the corresponding Onsager coefficient, which may in general depend on

the internal coordinate. To derive this equation, locality in the internal space has also
been taken into account, for which only currents and forces at the same value of the
coordinate become coupled. To obtain the expression of the current, we need to elab-
orate eq (5) more. Using Equation (2), it can be expressed as
where b=kUP is a diffusion coefficient assumed to be constant in first approxima-

tion. Activation processes are usually fast enough to favor the setup of a quasi-
stationary state with a nearly constant current. Integration in yspace then leads to
which expresses the constant current J in terms of forward and backward contribu-
tions corresponding to the two exponentials and 1 is a kinetic coefficient that can be
identified from the integration procedure. We have then shown that application of the
scheme of non-equilibrium thermodynamics to a mesoscopic description of the sys-
tem enables us to obtain the non-linear kinetic laws of activated processes. The der-
ivation we have presented is general and therefore valid for polymer crystallization.
In the next subsection, we will discuss the way in which MNET can also account for
the peculiar characteristics of these processes.
3.2 Polymer Crystallization

Crystallization of polymers takes place usually under inhomogeneous conditions as
those related to the presence of temperature gradients or mechanical processing of the
melt, such as extrusion, shearing or injection, which largely influence the kinetics.
The formalism introduced in the previous subsections is able to incorporate the
effect of these influences in the crystallization kinetics, thus providing a more real-
istic modeling of the process, which is mandatory for practical and industrial pur-
poses. Due to the strong foundations of our mesoscopic formalism in the roots of
standard non-equilibrium thermodynamics, it is easy to incorporate the influence of
other transport processes (like heat conduction or diffusion) into the description of
crystallization. In addition, our framework naturally accounts for the couplings
between all these different influences.
Contrary to the case of solidification of simple substances, crystallization of poly-
mers occurs at a wide range of temperatures. Moreover, the low thermal conductivites
and large heat capacities, characteristic of many polymer substances, imply that, even
the latent heat naturally released in the phase transformation sets up important tem-
perature gradients in the melt, able to alter the crystallization process significantly.
260 Chapter 22
Within the framework outlined in the previous section, the coupling between ther-
mal effects, diffusion and crystallization can be easily worked out in a simple and
unified framework, as described in ref. 8.
Crystallization in a sheared melt8 can also be described by the formalism intro-
duced in the previous subsection in which the internal coordinate represents the
coordinates of the center of mass of the crystallites and their number of monomers.
The main influence of the flow on crystallization kinetics is to modify the free
energy barrier, which is given by
1
Q(n,u) = +(n) + -rn(n)
2 (u-vo)2 (8)
+
where is the free energy of formation of a cluster with n monomers at rest, rn its
mass, u its velocity and vo the shear flow velocity. Proceeding along the lines indi-
cated previously, one obtains two significant results. First, the diffusion coefficient
of the crystallites in the direction of the velocity gradient is modified by the presence
of the shear.9 Its new expression is given by
where Do is the diffusion coefficient in the absence of flow. This expression clearly
shows that the correction depends on the shear rate p, the pressure p and the viscos-
ity 77 through the dimensionless factor @/p. In practical applications, crystallization
occurs at very large values of the shear rate, which originates through the injection
process of the melt; moreover, the viscosity of the supercooled melt is also very
large. This implies a significant correction to the diffusion mechanism of the crys-
tallites due to the presence of an external flow that originates as a result of the pres-
ence of driving forces in the processing of the melt.
The second important result is related to the change in the crystallization rate due
to the shear flow. Since the rate of addition of polymer to a spherulite is roughly pro-
portional to the diffusivity of the molecules, variations of the diffusion coefficient
Figure 3 Polarizing optical micrographs of poly(ethy1ene terephthalate) (PET) crystallized

at 240°C in the absence (A) and in the presence (B) of a shearing. As a conse-
quence of the shearing, nucleation becomes increasingly profuse, and the shape of
the spherulites becomes elliptical. (From re$ 1 I )
Mesoscopic Non-equilibrium Thermodynamicsof Polymer Crystallization 26 1
affect the value of the forward current in Equation (7). Moreover, the presence of the
flow destroys the isotropy of the system and leads to a distinction between growth
rates (and diffusion) in different directions. The growth then proceeds by the forma-
tion of non-spherical crystallites, as shown in Figure 3.
4 Conclusions
We have shown that mesoscopic non-equilibrium thermodynamics satisfactorily
describes the dynamics of activated processes in general and that of polymer crys-
tallization in particular. Identification of the different mesoscopic configurations of
the system, when it irreversibly proceeds from the initial to the final phases, through
a set of internal coordinates, and application of the scheme of non-equilibrium ther-
modynamics enable us to derive the non-linear kinetic laws governing the behavior
of the system.
This description has to be compared with that proposed by non-equilibrium ther-
modynamics in terms of only two states, corresponding to the melted and crystallized
phases in the example we are discussing, from which only one may account for the
linear domain, when the chemical potentials at the wells are not very different. This
feature imposes serious limitations in the application of NET to activation processes
since that condition is rarely encountered in experimental situations and has therefore
restricted its use to only transport processes. The mesoscopic version of non-equilib-
rium thermodynamics, on the contrary, circumvents the difficulty offering a promis-
ing general scenario useful in the characterization of the wide class of activated
processes, which appear frequently in systems outside equilibrium of different nature.
We have applied the MNET theory to a particular activated process: polymer crys-
tallization, in which the initial and final states of the system correspond to the melted
and crystallized phases, respectively. The processing of the melt is carried out under
out-of-equilibrium conditions due to the presence of driving forces and gradients
usually of significant strength. Our conclusion is that MNET can also account for the
kinetics in these more extreme situations.
References
1. L. Onsager. Phys. Rev., 1931,37,405;Phys. Rev., 1931,38,2265.
2. S . R. de Groot and P. Mazur, Non-equilibrium Thermodynamics,Dover, New York, 1984.
3. B. Hafskjold and S . Kjelstrup, J. Stat. Phys., 1995, 78, 463 ;A. Rosjorde, D.W. Fossmo,
S. Kjelstrup, D. Bedeaux and B. Hafskjold, J. Colloid Znter$ Sci., 240, 355.
4. J. M. G. Vilar and J. M. Rubi, Proc. Nut. Acad. Sci.,2001, 98, 11081.
5. D. Reguera and J.M. Rubi, Phys. Rev. E 2001,64,061106.
6. M. Avrami, J. Chem. Phys. 1939,7, 1103; ibid 1940,8, 212; ibid 1941,9, 177.
7. A. N. Kolmogorov, Bull. Acad. Sci. USSR Mat. Ser., 1937, 1, 355.
8. D. Reguera, J.M. Rubi and L.L. Bonilla, Chapter 3 in Mathematical Modelling for
Polymer Processing, V. Capasso (ed), Springer-Verlag, Berlin, 2003.
9. I. Santamm’a-Holek, D. Reguera and J.M. Rubi, Phys. Rev. E, 2001,63, 05 1106.
10. J. K. Hobbs, D. R. Binger, A. Keller and P. J. Barham, J. Polymer Sci., 2000, 38, 1575.
11. H. S. Myung, W. J. Yoon, E. S. Yoo, B. C. Kim and S. S. Im, J. Appl. Poly. Sci., 2001,80,
2640.
CHAPTER 23
Applied Thermodynamics for

Petroleum Fluids in the Refining
Industry
DERESH RAMJUGERNATH AND RAJ SHARMA
1 Introduction
Crude petroleum has been explored, extracted and processed for thousands of years.’
The modern petroleum industry as we know it, however, is just over a hundred years
old. The processing of crude oil into various products, by-products, charge-stocks
for petroleum-based finished products and petrochemicals has been determined by
the ever-changing demand of the product slate and new products or processes.
Historically, the petroleum industry is, perhaps, the single industry having had the
greatest impact on the development of Chemical Engineering as a discipline. It is
definitely a case where technology has preceded engineering science!
The use and application of ‘thermodynamics of petroleum fluids’ encompasses all
facets of the petroleum industry, ranging from production and refining of crude oil
to processing of petrochemicals. This chapter is limited to a review of the most
widely used thermodynamic methods in the petroleum refining industry, together
with a discussion on a few recent developments. The chapter provides a single
source of easily accessible relevant information.
Perhaps the most important thermophysical properties required in petroleum
refining for rating as well as process and equipment design are enthalpy and
vapor-liquid equilibria (fugacities). Enthalpies and fugacities often provide suffi-
cient information to calculate mass and energy balances across most unit operations
in a refinery. Since crude oil is an undefined, wide-boiling mixture of many differ-
ent types of hydrocarbons ranging in carbon numbers extending to 70 or beyond,
conventional thermodynamic methods used for property prediction of pure compo-
nents or well-defined mixtures are no longer applicable. The several direct products
of crude petroleum such as gasoline, kerosene, diesel and others are also defined by
their boiling point range cuts and the conventional methods cannot be used. Special
Applied Thermodynamics for Petroleum Fluids in the Re$ning Industry 263
methods and correlations have been developed over the past several decades to pre-
dict these, and other, properties of petroleum crude and its fractions. Most of these
correlations or methods rely on bulk property characterization of crude oil and its
cuts, such as specific gravity (SG or "API), various average boiling points, etc.
2 Composition of Petroleum
In spite of different sources and origins of crude oils across the world, surprisingly
the elementary chemical composition is relatively uniform (Table 1), even though
the physical characteristics (specific gravity, boiling ranges, pour points and the like)
vary ~ i d e l y . ~The
9 ~ physical characteristics are, however, generally affected by the
small differences in composition, which also, consequently, have an impact on the
value and processing of the crude oil.
Crude petroleum is an undefined mixture of several hydrocarbons (with carbon
numbers up to 70 or more) consisting of mainly paraffinic (including their isomers),
naphthenic and aromatic-type molecules. The small quantities of non-hydrocarbon ele-
ments such as sulfur, nitrogen, vanadium, nickel, chromium and oxygen present in the
crude oil are usually present as part of complex structures involving hydrocarbon^.^?^
While olefinic hydrocarbons are generally not present in the whole crude, various frac-
tions of petroleum may contain a small percentage as a result of some dehydrogena-
tion of paraffins and naphthenes that could occur during downstream processing.
3 Characterization of Petroleum and its Fractions

3.1 Measured Data
Historically, and to date, perhaps the two most important bulk physical properties
measured in the refining industry to characterize petroleum and its fractions are the
distillation datdcurve [True Boiling Point (TBP) - US Bureau of Mines m e t h ~ dor,
;~
ASTM D 866] and the mid-percent gravity curve in terms of "API. The distillation
datdcurve gives the boiling point profile and the mid-percent gravity curve as a
function of liquid volume percent distilled (Figure 1). The functional relationship
between "API and specific gravity (SG) is given as
141.5
OAP1 = SG6%00~ - 131.5
The distillation data are generally reported to be 572°F (-573 K) at 40 mm Hg

(-5.33 kPa) U.S. Bureau of Mines method5), which corresponds to about 790°F
Table 1 Typical Elementary Composition of Petroleum

Element Average percent
Carbon - 85
Hydrogen - 12
Nitrogen - 0.1
Sulfur up to 5%
264 Chapter 23
1000 -
-- 80
3 800-
-- 60
i$ 600- E
c
i 400- -- 40
200 - -- 20
0 I I I I 0
Figure 1
:
1111
1088
901
8.8
A 711
T
u Mfi
R
E 50°
400
300
OM 0.05 1.2 0.5 1 2 5 10 21 30 40 50 60 70 811 90 95 91 99 99.899.9 99.99

LIQUID VOLUME PERGNT DISTILLED
Figure 2 Illustration of the use of probability graph paper - crude distillation data (ternper-
ature vs liquid volume percent distilled)
(-694 K) at atmospheric pressure (760 mmHg; -101.32 kPa). However, the end
points of the whole crude ordinarily extend beyond 1000"F (-81 1 K). Hence, often,
the distillation data (temperature as a function of percentage distilled) are plotted on
a probability graph paper and a 'best-estimate' straight line drawn extrapolating to
about 1200 OF (-922 K) (Figure 2). The data points are then picked off from this line
to obtain a smooth TBP curve over the entire range as in Figure 1.
These two bulk properties (TBP data/ "API) are then used to calculate other 'con-
stants' such as molecular weight (MW), the pseudo-critical temperature (T,) and
pressure (P,),respectively, and the pseudo-acentric factor (a). The other properties
generally measured are the kinematic viscosities at 100°F (-311 K) and 200°F
(-366 K), respectively, and the Reid Vapor Pressure (RVP) (mainly for the gasoline
range cut, defined as the vapor pressure exerted by the cut at 100 O F (-31 1 K) ). All
of the above-measured properties and the calculated constants are generally
Applied Thermodynamics for Petroleum Fluids in the Refining Industry 265
Distillation
OAPT
Estimate
Other Properties
Mass and Energy Balances
Figure!3 Hierarchy of property prediction for 'design'
adequate to completely characterize 'petroleum' and enable estimation of various

thermo-physical properties. Borrowing from the concept of the onion diagram, the
hierarchy of thermodynamic property prediction in the refining industry for design
is perhaps best illustrated by Figure 3.
3.2 Calculated Parameters

From the specific gravity (SG/"API) and the distillation curve/data, the mean aver-
age boiling point (MeABP) and the Watson characterization factor (K,) are calcu-
lated. The MeABP is the arithmetic average of the molal and cubic average boiling
points and is mainly used for correlating hydrogen content, molecular weight (MW),
pseudo-critical properties (T,, P,), pseudo-acentric factor (m), K,, and the like. For
a petroleum fluid, MeABP may be estimated directly from the volume average boil-
ing point (VABP) by subtracting a differential expressed as a function of VABP and
the slope of the TBP curve as shown in Figure 4. Detailed correlations/curves are
presented by M a ~ w e l l . ~
The Watson characterization factor (K,) is calculated as follows:
The Kw for crude oil and its fractions generally range between 10 (highly naphthenic
crude) and 13 (highly paraffinic crude). For highly aromatic compounds to highly
paraffinic compounds, the value of Kw usually ranges between 10 and 15, respec-
tively, and therefore, can be said to be a qualitative measure of either the aromaticity
or the paraffinicity of a crude oil.
In addition to MeABP and Kw,the pseudo-critical properties (T,, P,), MW and the
pseudo-acentric factor (m) are also required for estimating various thermo-physical
properties of crude and its fractions. Several correlations/methods exist for calculating
266 Chapter 23
-- I MeAB= VABPa'DIFFERENTIAL'
VABP-096 + +90% (DISTILLEDTEMPS)
0 2 4 6 8 10 12
Figure 4 Schematic of correlation between mean average boiling point and volumetric aver-
age boiling point (Maxwell)
Tc,Pc, w and MW, in terms of MeABP and SG (or K,.,).Some of the most widely used
methods for estimating these parameters are listed in Table 2. Most of these methods
use a simple 2-parameter correlation of the form
8 = f (MeABP,SG) (3)
where 8is the property (Tc,Pc, w, MW) to be predicted.
Table 2 Some Widely Used Parameter Estimation Methods for Characterization of

Petroleum and its Fractions
Parameter Methods
Molecular weight, MW Hariu-Sage8
Kesler-Lee9
Riazi-Daubert
Riazi-Daubert
Critical temperature, T, Kesler-Lee9
Riazi-Daubert lo
l-bVU12
Riazi-Daubert
Critical pressure, P, Kesler-Lee9
Riazi-Daubert l o
Twu'*
Riazi-Daubed
Critical volume, V, Riedel'
%U'2
Acentric factor, o Kesler-Lee9
Applied Thermodynamicsfor Petroleum Fluids in the Refining Industry 267
i-
f
Liquid Vol % Distilled
Figure 5 Illustration of pseudo-component ‘characterization’ of petroleum crude into ‘nar-

row’ boiling range ‘compounds’.A similar approach is used for estimating spec@
gravities of ‘pseudo-components’
3.3 Pseudo-Components Characterization

Petroleum crude, in recent times, is also on occasion defined in terms of pseudo-
components by breaking up the overall distillation curve (Figure 1) into several com-
ponents as in Figure 5. These pseudo-components are not single components, but
mixtures boiling over narrower temperature ranges as compared to the whole petro-
leum or its products. These pseudo-components are again characterized in terms of
an average boiling point and SG for estimating other properties using the methods
of Table 2 as outlined in Calculated Parameters above. Since the crude oil is now
represented in terms of a defined mixture, the well-known multicomponent methods
for rating and design of process equipment become available. It is generally accepted
that the pseudo-multicomponent approach improves accuracy in thermodynamic
property predictions over the conventional petroleum correlations leading to better
designs. However, it has been the author’s experience that perhaps the pseudo-com-
ponent characterization of crude oil leads to more elegant methods becoming avail-
able for calculations, but, with relatively no improvements in accuracy even at the
cost of a significant increase in computation time and effort. There are several under-
lying reasons for this, starting with the assumptions about mixing rules to predict
petroleum mixture properties.
4 Thermo-Physical Property Prediction Methods

4.1 Enthalpy
Enthalpy is one of the most important thermo-physical properties required for cal-
culating heat loads in process design. The most common models used for estimating
enthalpies of petroleum and its fractions in the refining industry are based on the cor-
responding-states approach. Methods for calculating the necessary critical properties
268 Chapter 23
and an acentric factor have been discussed in Calculated Parameters above. The
American Petroleum Institute Technical Data Book - Petroleum Refining14 is the
most widely used reference manual in petroleum refining and presents enthalpy cor-
relations in terms of the two standard bulk properties measured in refineries - the
MeABP and SG (and, in turn, K.J.
It is important to mention here that since it is not possible to estimate the absolute
enthalpy of a compound, a reference temperature is generally chosen. Ideally, this
reference temperature is such that all the enthalpies under consideration are zero at
this temperature. In the absolute sense, unfortunately, different compounds have dif-
ferent molecular motions (and hence different enthalpies) at a given temperature.
Hence, the solution lies in choosing a temperature that is sufficiently low so that dif-
ferent compounds may be assumed to have the same enthalpy at the reference
point.15 Kesler-Lee9 defined the reference state for enthalpies of petroleum fluids as
saturated liquid at - 200 O F (- 144 K), and it is reasonable to assume that enthalpies
of all the petroleum fractions at this low temperature will be zero.
In general, the estimation of enthalpy (departure) can be represented in the fol-
lowing manner:
AH = I
Tb
=ref
T
Cp,dT + AHv + ICp,dT
Tb
+ AHpc (4)
where Cpland Cpvare the liquid and vapour heat capacities, respectively, AH,, is the
latent heat of vaporization and AHx is the pressure correction term to compensate
the effect of system pressure on enthalpy. Often, the effect of pressure on a liquid
that exists below 1000 psia ( -6.9 MPa) pressure is neglected.
The Kesler-Lee9 correlations for liquid and vapour phase heat capacities of petro-
leum fluids are used for estimating the respective enthalpies at temperatures of inter-
est. The Lee-Kesler16 corresponding-states method is used for obtaining estimates of
the heats of vaporization and for developing the saturation envelope enthalpies. This
method uses the Curl and Pitzer17 approach and calculates various thermodynamic
properties by representing the compressibility factor of any fluid in terms of a sim-
ple fluid and a reference fluid as follows:
where Z ( O ) and Z(') are given by
The enthalpy departure (and other thermodynamic properties) is then calculated

using equations obtained from the compressibility factor via the usual thermody-
namic relationships.
In the case of reacting systems in petroleum refining, the required estimates of
heats of reaction are obtained, as usual, by the difference between the enthalpy
departures of the products and reactants. As an example, the endothermic heat of
Applied Thermodynamics for Petroleum Fluids in the Refining Industry 269
cracking reactions in a fluidized-catalytic-cracker (FCC) may be estimated as fol-
lows with excellent results.'*
The heat of cracking reaction may be defined as
is the enthalpy of reac-

where m p r o d u c t s is the enthalpy of products, and AHrcactants
tants. As discussed above, in general, the enthalpy departure is given by Equation
(4). Substituting Equation (4) into Equation (8) gives
where Cpl and Cpgare calculated using the Kesler-Lee9 correlation and AHvapis
calculated using the generalized thermodynamic correlation reported by Lee and
Kesler16 as before.
4.2 P-V-T Relationship

Density or the pressure-volume-temperature (P-V-7') relationship is considered
along with enthalpy and vapour-liquid equilibria as the three most essential thermo-
physical properties in the petroleum refining industry. The compressibility or den-
sity is commonly used in the petroleum industry to determine the volumetric
properties of gases and liquids, information that is vital for transportation, safety and
sale of petroleum fluids.
The pressure-volume-temperature relationship is concisely represented by the
compressibility factor,
PV
Z=--
RT
Depending on the conditions (temperature and pressure) and the complexity and
accuracy required for the computations, generalized correlations for gases or analyt-
ical equations of state can be used to compute the gas compressibilities and the sub-
sequent molar volumes.
Generalized correlations for the compressibility factor, 2, as well as analytical
expressions, based on the second virial coefficients, have been developed by Pitzer
et aZ.19The correlation for 2 takes the form:
where Zo and 2' are functions of both reduced temperature (T,) and reduced pressure
(P,). Of the Pitzer correlations, the one developed by Lee and Kesler16 has found
greatest favour.
The analytical equations of state used to compute the pressure-volume-tempera-
ture relationships are generally one of the following three classes, viz. virial equa-
tions of state, cubic equations of state and the Benedict-Webb-Rubin equation or
modifications thereof.
270 Chapter 23
For the liquid phases, the compressibilities are most commonly computed using
the generalized correlations for liquids and analytical equations of state. Reid et aL20
describe in detail the various correlations available for the computation of liquid
phase densities for pure components, with the modified Rackett equation being the
most popular. For mixtures, the use of an analytical equation of state is most prefer-
able with cubic equations of state being the favourite type of equation of state.
4.3 Vapour-Liquid Equilibria (Including Vapour Pressure)

The emphasis on vapour-liquid equilibria (including vapour pressure) is inherant in
the petroleum industry due to the importance of distillation in separations. If separa-
tions by extraction are to be undertaken, then liquid-liquid equilibrium is equally
important. Fugacities for thermodynamic equilibrium (flash calculations) are proba-
bly one of the most sought-after properties. This is because fugacities and enthalpies
often provide sufficient information to calculate a mass and energy balance.
The description of vapour-liquid equilibrium behaviour can be obtained from
analytical equations and generalized correlations. The generalized correlations are
generally for the equilibrium ratio, K, and the fugacity coefficients.
For the analytical equations, there are two methods to compute the vapour-liquid
equilibrium for systems. The equation of state method (also known as the direct or
phi-phi method) uses an equation of state to describe both the liquid and vapour
phase properties, whereas the activity coefficient method (also known as the
gamma-phi approach) describes the liquid phase via an activity coefficient model
and the vapour phase via an equation of state. Recently, there have also been modi-
fied equation of state methods that have an activity coefficient model built into the
mixing rUles.21*22These methods can be both correlative and predictive. The predic-
tive methods rely on the use of group contribution methods for the activity coeffi-
cient models such as UNIFAC23and ASOG. Recently, there have also been attempts
to develop group contribution methods for the equation of state method, e.g.
PRSK.24For a detailed history on the development of equations of state and their
applications, as well as activity coefficient models, refer to Wei and S a d u ~ , ~ ~
Sandler26and W a l a ~ . ~ ~
DePreister 28 produced nomographs for K-values of light hydrocarbons. The
K-values represent the lightness of a compound or fraction, and are computed from
the vapour and liquid equilibrium phase compositions as follows:
Yi
K. = -
' Xi
The K-values are correlated as functions of temperature and pressure. The DePreister
correlations have approximate validity for mixtures of light hydrocarbons and other
simple molecules.
Vapour pressure is a key property in VLE calculations and is thus an important
petroleum property. The most common method for prediction of vapour pressures is
the corresponding states method. The method requires knowledge of the critical
properties and the acentric factor. For petroleum fractions, the Ma~well-Bonnell~~
method is standard.
Applied Thermodynamics for Petroleum Fluids in the Rejhing Industry 27 1
The generation of pseudo-components for petroleum fluids has been discussed in

Pseudo-Components Characterization. Once the various pseudo-components have
been generated and classified, the conventional methods for multicomponent VLE
apply and the methods discussed above for VLE computation can be used.
5 Codes and Data Sources

The most widely used codes and data sources today are perhaps the following three:
1. API Technical Data Book - Petroleum Refining.14

2. Aspen Physical Property System - Physical Property Methods and Models.30
3. SUPERTRAPP - National Institute of Standards and Technology (NIST)
Thermophysical Properties of Hydrocarbon Mixtures Databa~e.~'
All the above three provide extensive data sources and simulators for estimating
various thermodynamic properties of pure compounds, well-defined mixtures and
petroleum fluids. The respective manuals describe in detail various methods used for
estimating the properties and their limitations while at the same time providing the
user with choices. The Aspen Physical Property System and SUPERTRAPP can also
be used to estimate thermodynamic properties using the pseudo-component charac-
terization of petroleum fluids. The reader is referred to the respective manualshim-
ulators for details.
6 Future Trends
Even though thermo-physical property prediction of petroleum fluids has come a
long way since the early years of crude oil refining, the approach used by industry
still mainly remains rooted in methods developed several decades ago. Most of these
methods, as discussed earlier, rely on bulk property characterization deriving from
them other parameters needed for property predictions. The current analyses of
petroleum fluids are typically limited to identifying the front-end through to C,/C,
(or C,) and then characterizing the C l / C i (or C;) fraction in terms of boiling points
and specific gravities. In view of the complexity of composition, a detailed compo-
nent-wise characterization of petroleum fluids is very difficult and is neither practi-
cal nor f e a ~ i b l e . ~However,
~ 3 ~ ~ with growing emphasis on improved designs, better
characterization and property prediction methods are required. The pressure-vol-
ume-temperature (PVT) and density predictions are particularly sensitive to accu-
rate characterizations of the heavier hydrocarbons. Recently, attempts have been
made to better characterize the heavier hydrocarbons in petroleum fluids using a
continuous model for C,+ fractions. The continuous model is based on the use of dis-
tribution functions for characterizing c,+ fraction^^^,^^, and very good agreement
between predicted values and experimental data has been claimed.34The character-
ization parameters, including the pseudo-critical properties/acentric factors,
obtained from the distribution function approach, are then used for predicting the
phase-behaviour of petroleum f l ~ i d s .The ~ ~methods
, ~ ~ developed are mostly applied
272 Chapter 23
to simple petroleum fluids with a high concentration of light hydrocarbons and a low
concentration of C,’ fractions. Most recently, Fazlali et aZ.33have developed a new
phase behaviour prediction method for complex petroleum fluids (consisting of high
concentrations of the C,’ fraction), using a combination of the Peng-R~binson~~ and
the Riazi-Mansoori3* equations of state, which requires a minimum of input charac-
terization data such as an average value of molecular mass and the specific gravity
of the C,+ fraction. Prediction of phase-behaviour of complex petroleum fluids using
‘continuous thermodynamics’ is a new area of research and will perhaps lead to
important developments in the field.
Further, molecular simulation and computational chemistry have evolved, and are
evolving, into important tools for developing better characterization techniques
where it is not possible to measure all data.32Even so, it is precisely the molecular
complexity of petroleum fluids that seems to be an inhibiting factor in the use of
these methods for developing better characterization methods. However, identifica-
tion of important functional groups in petroleum fluids applying molecular simula-
tion and/or computational chemistry for use with group contribution methods to
predict thermo-physical properties may be an area for further research.
References
1. J. G. Speight, The Chemistry and Technology of Petroleum, 3rd edn, Marcel Dekker, New
York, 1999.
2. J. H. Gary and G.E. Handwerk, Petroleum Refining - Technology and Economics, Marcel-
Dekker, New York, 1975.
3. W. L. Nelson, Petroleum Refinery Engineering, 4th edn, McGraw-Hill, New York, 1958.
4. M. C. K. Jones and Hardy, R. L., Znd. Eng. Chem., 1952,44,2615.
5. N. A. C. Smith, H. M. Smith, 0. C. Blade and E. L. Garton, ‘The Bureau of Mines
Routine Method for the Analysis of Crude Petroleum’, Bureau of Mines Bull., 1951,
Number 490.
6. .ASTM Designation D 86, 1934.
7 J. B. Maxwell, Data Book on Hydrocarbons, D. Van Nostrand, New York, 1950.
. I
8 I0. H. Hariu and R. C. Sage, Hydrocarbon Process., 1969,48(4), 143-148.

9. .M. G. Kesler and B. I. Lee, Hydrocarbon Process., 1976,55(3), 153-158.
10 M. R. Riazi and T. E. Daubert, Hydrocarbon Process., 1980,59(3), 115.
I. .
11. . M. R. Riazi and T. E. Daubert, Ind.Eng.Chem.Res.,1987,26,755-759.

12. ’C. H. Twu, Fluid Phase Equilibl:, 1984, 16, 137.
13. .Reidel, L. API Procedure 4A 1.15.
14. .American Petroleum Institute Technical Data Book - Petroleum Refining, 5th edn, T. E.
Daubert and R. P. Danner (eds), American Petroleum Institute, Washington, DC, 1988.
15 V. K. Pareek, A. A. Adesina, A. Srivastava and R. Sharma, Chem. Eng. J., 2003,92(1-3),
101 - 109.
16. B. I. Lee and M. G. Kesler, A.I.Ch.E.J., 1975, 21,3, 510.
17. R. F. Curl and K.S. Pitzer, Ind. Eng. Chem., 1958, 50, 265.
18. V. K. Pareek, A.A. Adesina, A. Srivastava, and R. Sharma, J Mol. Cat, A: Chem., 2002,
181(1-2), 263-274.
19. K. S. Pitzer , Thermodynamics, 3rd edn, McGraw-Hill, New York, 1995.
20. C. R. Reid, J. M. Prausnitz and B. E. Poling, The Propeties of Gases and Liquids, 4th edn,
McGraw-Hill, Singapore, 1988.
Applied Thermodynamicsfor Petroleum Fluids in the Refining Industry 273
21. M. J. Huron and J. Vidal, Fluid Phase Equilibl:, 1979, 3,255-271.
22. D. S. H. Wong and S. I. Sandler, A.I.Ch.E. J., 1992,38(5), 671-680.
23. A. Fredenslund, J. Gmehling and P. Rasmussen, Vapor-Liquid Equilibria using UNIFAC.
A Group Contribution Method, Elsevier, Amsterdam, 1977.
24. T. Holderbaum and J. Gmehling, Fluid Phase Equilibl:, 1991, 70, 25 1.
25. Y. S. Wei and R. J. Sadus, A.I.Ch.E. J., 2000, 46, 169-196.
26. S. I. Sandler, Models for Thermodynamics and Phase Equilibria Calculations, Marcel
Dekker, New York, 1994.
27. S. M. Walas, Phase Equilibria in Chemical Engineering, Butterworths, Stoneham, 1985.
28. C. L. DePriester, Chem. Eng. Progl: Symp. Sel:, 1953,49(7), 1-43.
29. Maxwell-Bonnell, API Procedure 5A1.15, 1955.
30. Aspen Physical Property System 11 . l , Aspen Tech Inc, Cambridge, MA, 2001.
3 1. NIST Thermophysical Properties of Hydrocarbon Mixtures Database (SUPERTRAPP),
National Institute of Standards and Technology (US Department of Commerce),
Gaithersburg, MD, 2003.
32. C. Tsonopolous and J. H. Heidman, Thermophysical property needs into the 21st century:
petroleum refining and petrochemical industry view, 13th Symposium on Thermophysical
Properties, Boulder, CO, USA, June 22027, 1997.
33. A. Fazlali, H. Modarress and G.A. Monsoori, Fluid Phase Equilibl:, 2001,179,297-317.
34. M. R. Riazi, Ind. Eng. Chem. Res., 1997,36,4299-4307.
35. L-S Wang and J. Gmehling, A.I.Ch.E. J., 1999,45, 5 , 1 125-1 134.
36. H. Manafi, G.A. Mansoori and S. Ghotbi, J.Petroleum Sci. Eng., 1999, 22, 67-93.
37. D. Y. Peng and D.B. Robinson, Ind. Eng. Chem. Fund., 1976,15,59-64.
38. M. R. Riazi and G. A. Mansoori, Oil Gas J., 1993,12, 108-1 11.
Subject Zndex
Ab initio quantum Mechanical Complex Process Optimization, 88
Calculations, 43 Composition of Petrol, 263
Activity, 37 Continuous Injection, 97
Activity Coefficients, 35, 37 COSMO-RS Model, 54
Activity Coefficient Model, 174 Cost of Recycling, 160
Activated Processes, 258 Cost-Benefit Analysis, 171
Adsorption, 79,243 Critical Point, 151
Adsorption Isotherm, 244 Crystal Growth, 241
Alloys, 213 Crystallization, 230
Analysis Tools, 169 Cubic Equation of State, 173
Anionic Polymerization, 91, 108
Applied Thermodynamics, 166, 167 Data Bases, 129
Arrenhenius Plot, 107 Data Packages, 175
Aspen, 166 Data Sources, 271
AspenTech Clients, 172 Decarburisation rate, 18
Density, 3, 226
Binary Alloys, 213 Deployment Tools, 169
Binary Solid-Liquid Equilibria, 23 1 Differential Scanning Calorimetry,
Bulk Thermodynamic Properties, 149 154
Diffusion Coefficients, 127
Calculation of Phase Diagrams, 213 Dissolution of Crystal Growth, 241
Calculation Engine, 168 Distillation, 33
Calorimetric Data, 118 Distillation Columns, 5, 33
Calorimetric Design, 111 Donnan Equilibrium, 25
Calorimetric Measurement, 144 Drug Receptor Studies, 108
Catalyst Nature, 98 Economic Impact, 166
Catalytic Partial Oxidation, 180 Economic Value, 166
Catalytic Reactor, 123 Electrolytes, 174, 189
Catalytic Species, 92 Electrolytes for Fuel Cells, 189
Ceramic Membranes, 180 Electrolyte Systems, 174
Chemical Equilibrium, 36 Energetic Considerations, 165
Chemical Potentials, 29,219 Energies of Transformation, 199
Chemical Reactors, 8 Enthalpy Predictions, 267
Chemical Recycling, 159 Entropy Production, 5, 41
ChemApp, 16,30 Equation of State, 173, 244
ChemSage, 30 Equilibrium Calculations, 176, 222
ChemSheet Thermodynamic Software, Equipment, 40
30 Estimation Methods, 172
Coatings (A1,0, - AlN), 206 Eutectic Behaviour, 233
Coatings (Ti, Al)N, 198 Eutectic Systems, 238
Codes, 271 Extractive Distillation, Process, 77
Computational Quantum Mechanics, 43 Extraction Process, 78
Subject Index 275
Fibres, 23, 24 Lennard Jones Potential, 45
Flow Calorimeter, 111 Liquid-Liquid Equilibrium, 80,223
Flow Calorimetric Equations, 113 Liquid-Liquid Extraction, 78
Flow Microcalorimeter, 110, 119
Flow-Through Microcalorimeters, 115 Mass Transfer, 134
Force Fields, 44 Material Recycling, 159
Fuel Cells, 187 Membranes, 180, 186
Fugacity Coefficients, 35 Mesoscope Non-Equilibrium
Fusion Crystallization, 152 Thermodynamics,255,259
Meso-Structures,254
Gaussian Comutational Package, 5 1 Metallurgical Processes, 12
Gas Clathrate Hydrates, 57 Metastable Coatings, 197
Gas-Liquid Equilibrium, 80 Microcalorimetry, 104
Gas-Phase Flow Calorimetry, 113 Micro-Particle Production, 132
Gas Solubility, 86 Mixtures, 248
GEMC NVT simulations, 48 Model Deployment, 177
Gibbs Energy in Small Particle Systems, Modelling, 16
210 Modelling Electrolyte Systems, 174
Gibbs Energy Minimization, 16, 28, 71 Modelling Polymer Systems, 175
Gibbs Energy Model, 27 Molecular Simulations, 44
Glass Transition Temperature, 154 Monte Car10 Simulations, 45
Multi-Phase Thermodynamics, 23
Heat Capacities of Gases, 145 Multiple Miniature Calorimeter, 108
Heat Capacities of Liquids, 147
Heat of Transfer, 4 Nano-Particle Production, 132
Henry Coefficients, 85 Nano-Sized Particles, 132, 209
Heterogeneous Catalysis, 4 1 Nano-Sized Phase Diagrams, 213
High Pressure VLE, 135 Natural Gas, 57
Hydrate Phase Diagram, 65 Natural Gas Hydrates, 57
Hydrate Thermodynamics, 59 New Materials, 180
Non-Equilibrium Thermodynamics, 1, 254
Infinite Dilution Activity Coefficients, 83 Non-Equilibrium Thermodynamics of
Industrial Applications, 170 Polymer Crystallization, 254
International Dimension, 129 Non-Isothermal Reaction Calorimetry, 101
Interfacial Properties, 177
Ionic Liquids, 76, 82 Oxide Fuel Cells, 187, 189
Ionic Mean Properties, 220 Oxygen Permeation, 184
Irreversible Thermodynamics, 72 Oxygen Separation, 180
Isothermal Conditions, 92
Isothermal Flow-Microcalorimetry, 110 Particles Morphology, 135
Isothermal Microcalorimetry, 110 Petrol Composition, 263
Petroleum Fluids in the Refining Industry,
Kinetics of Dissolution of Crystal Growth, 262
24 1 pH, 31
Kinetic Parameters, 118 Pharmaceutical Industry, 104
Kinetic Stability, 106 Phase Diagrams, 66, 224
Phase Equilibrium, 34, 170
Lactams, 91 Phase Equilibrium Modelling, 170
Lanthanum Gallates, 191 Phase Transitions, 2
L-D Convertor, 12 Phase Transition Thermal Properties, 152
276
Plastic Recycling, 159 Solvent Screening, 170
Polymer Crystallization, 256,259 Speciation Equilibrium, 222
Polymer Production, 160 Spectro-CalorimetricScreening, 88
Polymer Systems, 175 Stability Determinations, 105
Prediction Methods, 267 Stability of Membranes, 186
Prediction Methods for PVT Standard States, 219
Relationships, 269 Statistical Thermodynamic Model, 67
Prediction Methods for VLE, 270 Steel Melting, 15
Process Design, 39 Stirring Level, 100
Process Model, 12 Successive Injection Technique, 96
Process Synthesis, 38 SupercriticalAnti-solvent Precipitation,
Programmable Interfaces, 169 132
Pseudo-Binary Diagrams, 139 Supercritical Fluids, 132
Pseudo-Component Characterization, 267 Surface Tension, 4
Pulp Fibres, 23
Pulp-Ion Interaction, 25 Ternary Solid-Liquid Equilibrium, 236
Pulp Suspension, 23 Theoretical Predictions, 127
Pure Component Behaviour, 230 Thermal Conductivity, 126
PVD Processes, 197 Thermal Recycling, 159
PVT Predictions, 269 Thermal Separation Processes, 77
Thermodynamic Considerations, 165
Quantitative Cost-Benefit Assessments, Thermodynamic Functions, 246
171 Thermodynamics of Adsorption, 243
Quantum Chemistry, 23 Thermodynamics of Crystallization, 230
Quantum Mechanical Calculations, 43, 53 Thermodynamics of Electrolyte Systems,
Quick Screen Solvents, 170 219
Thermodynamics of Nano-Sized Particles,
Raman Spectroscopy, 65 209
Reaction Calorimetry, 88 Thermodynamics of New Materials, 180
Reaction Kinetics, 37 Thermodynamic Prediction, 197
Reactive Distillation, 33 Thermo-MechanicalProperties, 149
Recycling, 160 Thermo-Physical Properties, 144, 267
Refining Industry, 262 Titration Calorimetry, 108
Reversible Thermodynamics, 72 Transport and Interfacial Properties, 177
Transport Coeficients, 14
SAS Processing, 133 Transport Properties, 122
Scale-Up Problem, 39 Transport Thermodynamics, 181
Second Law, 6 , 9
Selective Solvent, 77, 80, 81 Unmet Needs of AspenTech Clients, 166,
Selectivity at Infinite Dilution, 83 172
Separating Agent, 77 UNIFAC Model, 5 5 , 171
Separating Processes, 76
Simulation, 17 Van der Waals and Platleeuw Model, 69, 71
SimuSage, 16 Vapour-Liquid Equilibrium, 79, 85, 223
Slag, 14 Virial Coefficient Prediction, 46
Solid-Fluid Equilibria, 239 Viscosity, 125, 228
Solid-Liquid Equilibrium, 223, 23 1, 236 VLE Data, 85
Solid Oxide Fuel Cells, 187 VLE Measurements, 36
Solid Solutions, 232 VLE Predictions, 270
Solution-Phase Flow Calorimeters, 112 VLE Simulation, 48, 49, 54

Chemical Thermodynamics For Industry

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Chemical Thermodynamics For Industry

Загружено:

Авторское право:

Доступные форматы

Chemical Thermodynamics for Industry

0 The Royal Society of Chemistry 2004

All rights reserved

Published by The Royal Society of Chemistry,

Registered Charity Number 207890

For further information see our web site at www.rsc.org

Typeset by Macmillan India Ltd, London, UK

The newly established areas include:

ab initia quantum chemistry

I wish to record my thanks to authors and to the publishers in producing a book of

Chapter 2 A Modelling Technique for Non-equilibrium Metallurgical

Chapter 3 Multiphase Thermodynamicsof Pulp Suspensions 23

Chapter 4 Reactive Distillation 33

Chapter 5 Thermodynamic Properties from Quantum Chemistry 43

Chapter 6 Thermodynamics of Natural Gas Clathrate Hydrates 57

Chapter 10 Isothermal Flow-Microcalorimetry:Principles and

Chapter 12 Micro- and Nano-particlesProduction

Chapter 13 CalorimetricMeasurements of Thermophysical

Chapter 14 Plastic Recycling 159

4 Industrial Applications 170

Chapter 22 Mesoscopic Non-equilibrium Thermodynamicsof

Chapter 23 Applied Thermodynamicsfor Petroleum Fluids in the

Subject Index 274

1 What Can Non-Equilibrium Thermodynamics Offer?

2 Developments and Status of NET

3 NET Applied to Phase Transitions

OCP 0 NEMD simulations

4 NET Equations for Distillation Columns

5 The Second Law Optimal Path of Operation

The Hamiltonian for the problem was constructed as

6 Second Law Optimal Distillation Columns

Figure 6 Examples of operating lines of an adiabatic and a diabatic column

7 Second Law Optimal Chemical Reactors

Entropy production is important for a proper definition of fluxes and forces in

A Modelling Technique for

2 Process Model Development

(FeO) +[C]+ [Fe]+ { CO} (2)

(FeO) + { CO} + [Fel+ { CO,} (3)

nx=Kx, elTA(cx, metal-'^, phase boundary) (5)

However, there is no reliable information on transport coefficients, reaction rate con-

Figure 2 Cell model for the LD converter process

define the conditions for a “local” equilibrium calculation,

Figure 3 Integration of ChemApp into SimuSage

C-content of hot metal

Figure 5 Decarburisation rate

Figure 6 Components in hot metal

Figure 7 Components in slag phase

normally Si is oxidised preferentially to carbon. The limitation of transport to the

1600 I I Hot spot

Figure 8 Gas production

Figure 9 Temperature evolution

2 Fibres in Aqueous Solution

Fiber wall External suspension solution

Figure 1 Schematic presentation of a pulp fibre in an aqueous electrolyte solution

3.1 Donnan Equilibrium

p;+RT In a[+z,Fy' = p;+RT In a;.'+z,Fy" (1)

3.2 Specific Interactions

4 Solid and Gaseous Phases

5 General Gibbs Energy Model

Figure 3 Calcium concentrations in the suspension of eucalyptus sulfate (ESA) and t h e m -

Table 2 Data input for the ESA and TMP models

a reduction of the investment,

A typical reactive distillation set-up is depicted in Figure 1.