Академический Документы
Профессиональный Документы
Культура Документы
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
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.
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.
Topics covered, which are considered potentially important to the chemical industry
in the future, include:
ionic liquids
0
0 new materials
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.
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
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.
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.
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
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
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.
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.
X 1
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.
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:
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.
Non-Equilibrium Thermodynamics for Industry 11
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
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.
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:
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:
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:
+ Bath reactor
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
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.
A Modelling Technique for Non-equilibrium Metallurgical Processes 17
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
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
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
A Modelling Technique for Non-equilibrium Metallurgical Processes 19
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]
-
Simulation Results Components in Slag Phase
25000 1
-
4-CaO calculated
Si02 calculated
4 MnO calculated
0
5 10 15
Time [min]
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]
2600
2400
-
E
2200
-5 2000
2
b
8 1800
1600
1400
1200
0 5 10 15
Time [min]
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.
c1-
Na'
cr Ca2+
Na'
CI- CI-
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.
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.
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
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:
1
Feed 1
Catalyst
e distillation ZOI
Feed 2
Product 2
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.
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:
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
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.
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
Reactive Distillation 37
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:
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:
- 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.
-TI-
er
1 SOlVenI I
water
catalyst
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.
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:
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
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.
-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
Thermodynamic Properties from Quantum Chemistry 47
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
1000/Temperature (K-')
2
300
350 I @
I
3
c
$
.
250
dI
ez-
200
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
Thermodynamic Properties from Quantum Chemistry 49
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
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 $//’/
y’N =1.20
n
2
72
1:
\“ ,/ d//
non-polarization
/Lo
0.1 :
/ polarization
a)
9
450 A
s
2
3
c
2
Q
Q,
f 0
t- 350
25 0
0.0 0.2 0.4 0.6 0.8
Density (g ~ r n - ~ )
Figure 5 Vapor pressure and vapor-liquid phase boundary for methanol as a function of
temperature from experiment (lines) and from simulation with combinations of scal-
ing and polarization
adding up the contributions. The second is to recognize (e.g., in protein docking) that
there may be a few interaction sites that are especially important, while distant parts
of the molecule are less so. The important regions of the molecule are then treated
with a high level of theory, and the distant regions with less accurate methods, or
Thermodynamic Properties from Quantum Chemistry 51
I
0.9
0.8
0.7
0.6
0.5
1
0.4 I I I I
Figure 6 Vapor-liquid equilibrium for the methanol + acetonitrile system. The unfilled
points indicate experimental data, and the$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-')
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)
Thermodynamic Properties from Quantum Chemistry 53
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.
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
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,
Thermodynamic Properties from Quantum Chemistry 55
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
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:
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.
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:
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.
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.
~ ~~
V vapor
Lw: liquid water
H: hydrate
M: solid methane
LM:liquid
methane
I: ice
v-L,
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 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:
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
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)
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.
Thermodynamics of Natural Gas Clathrate Hydrates 73
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,
Inc., New York, 1998.
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,
Inc., New York, 1998.
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.
Thermodynamics of Natural Gas Clathrate Hydrates 75
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
International Hydrate Conference, Yokohama, Japan, May 2002,673.
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
International Hydrate Conference, Yokohama, Japan, May 2002,444.
93. C. A. Koh, Chem. Soc. Rev., 2002,31, 157.
CHAPTER 7
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.
aliphatics
7 aromatics
7
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)
Ionic Liquids in Separation Processes 79
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.
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.
x
xi *Hi,j=yiqiP (7)
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).
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
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.
16. - 1
I
:
e
e
?? 12.
'S
v)
10.
l4.I
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.
Ionic Liquids in Separation Processes 85
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
- 2
s2
1
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
Ionic Liquids in Separation Processes 87
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
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
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.
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.
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
Spectrocalorimetric Screening for Complex Process Optimization 95
Te rnperature/ *C
J
250
-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
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
Spectrocalorimetric Screening for Complex Process Optimization 97
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.
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)).
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
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.
s150 rpm --8- 280 rpm u420 rpm +560 rpm +750 rpm I
245 ,
T = I 20 "C
5 195
.?
145
E
1
(4
-S
Timeis
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
Spectrocalorimetric Screening for Complex Process Optimization 101
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.
- 400
- 390 Y
-.
-380 g
CI
- 370 $
n
-360
I-
- 350
-5 ! ' ' ' ' ' : ' ' ' ' ' : ' ' ' ' ' ; ' ' ' ' ' ~ ' ' ' ' ' ; I 340
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
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.
Spectrocalorimetric Screening for Complex Process Optimization 103
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
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
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.
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.
3.1.1 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
Thermostat
h I Calorimetric
Ce 11
To Waste
React ion#
Reaction
Vessel
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
.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.
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)
0.00025
0.00020
-i
01
0.00015
v
u
Y
0.ooo10 w
w
0.oooO5
Figure 6 Max rate vs. substrate concentrationfor the uredurease enzyme reaction
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.
Isothermal Flow-Microcalorimetry: Principles and Applications for Industry 119
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.
Isothermal Flow-Microcalorimetry: Principles and Applications for Industry 121
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
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)
PRODUCT Water
Water
Preheater
Catalyst
bed
lnterstage
Unit
Catalyst
bed
610 K,30
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
Transport Properties and Industry 125
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.
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
Transport Properties and Industry 129
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.
Transport Properties and Industry 131
References
1. R. B. Bird, W. E. Stewart and E. N. Lightfoot Transport Phenomena, John Wiley & Sons
Inc., New York, 1960.
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
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.
Figure 1 Large needle-like crystals of Prednisolone precipitated from Methanol at 1.50 bal;
40 "C,20 mg/mL
Micro- and Nano-particles Production Using Supercritical Fluids 137
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
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.
Micro- and Nano-particles Production Using Supercritical Fluids 139
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
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
Micro- and Nano-particles Production Using Supercritical Fluids 141
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
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.
Micro- and Nano-particles Production Using Supercritical Fluids 143
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
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.
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
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
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
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
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 -
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
Calorimetric Measurements of Thermophysical Properties for Industry 153
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
(81)310
I .
31I
.
310 315 330 335
I
YO
T0IWW-l-K
200
180
160 .
60
40
20
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.
-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.
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;
Plastic Recycling 163
(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
T=20 "C
unsoluble +
compounds (111) filtration T-140 "C
T-100 "C
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.
Plastic Recycling 165
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
Calculation engine
Analysis tools
Programmable interfaces
Deployment tools
I t t
Aspen Engineering Suite@
t t f
Yodels and Data far Physlcat Properties, Solution Chemistries, and industriai Applicatlons
T t t
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
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.
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.
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.
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.
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.
Industry Perspective on the Economic Value of Applied Thermodynamics 179
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)
Carrier gas
0 2 P" ( 0 2 )
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
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
re" Fe2+ 1
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.
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:
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
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,
a b
Figure 7 Schematic set-up of a fuel cell. ( a ) proton-conducting electrolyte, and (b) oxygen
ion-conducting electrolyte
Thermodynamicsof New Materials 189
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)
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)
Assuming equilibrium for both reactions and considering that the concentration of
oxygen vacancies is fixed by the high Y203fraction, cvo= cy203,we obtain:
150 pm
15 Pn
1.5 pm
0.15 p
1000A'(K1)
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
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:
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
1500
-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.
m
0 0 0 0
m o o 0
m NaC1-cubic
0 .
0
Mendeleev Number
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
Thermodynamic Prediction of the Formation 203
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
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
200
150
100
g
.I
-100
150
0
-250
;a
2 -300
w
-350
-400
Mendeleev Number
Figure 9 Gibbs energy of formation of the cubic and hexagonal phases in the A1N-TiN sys-
tem at difierent temperatures
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
Thermodynamic Prediction of the Formation 207
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
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
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:
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):
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:
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
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
Thermodynamics of the Nano-Sized Particles 215
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
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:
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
Pi = (1)
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)
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.
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
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.
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
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.
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.,
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.
Thermodynamics of Electrolyte Systems of Industry 225
0 20 40 60 80 100
Mass percent Mg(NO,),
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
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
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
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)
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.
Thermodynamics of Electrolyte Systems of Industry 229
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.
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.
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
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
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
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
Thermodynamics of Crystallization 235
-5.00 -f-
d
‘
c -5.20
L-isoleucine
-5.40
-5.60
-5.80
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.
/
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
Thermodynamics of Crystallization 237
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
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.
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
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.
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:
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
,-
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 : ~
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:
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.~
Thermodynamics of Adsorption 247
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)
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)
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.
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).
Thermodynamics of Adsorption 253
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
1.1 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.
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
(1)
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
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.
+
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
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
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$ning Industry 263
methods and correlations have been developed over the past several decades to pre-
dict these, and other, properties of petroleum crude and its fractions. Most of these
correlations or methods rely on bulk property characterization of crude oil and its
cuts, such as specific gravity (SG or "API), various average boiling points, etc.
2 Composition of Petroleum
In spite of different sources and origins of crude oils across the world, surprisingly
the elementary chemical composition is relatively uniform (Table 1), even though
the physical characteristics (specific gravity, boiling ranges, pour points and the like)
vary ~ i d e l y . ~The
9 ~ physical characteristics are, however, generally affected by the
small differences in composition, which also, consequently, have an impact on the
value and processing of the crude oil.
Crude petroleum is an undefined mixture of several hydrocarbons (with carbon
numbers up to 70 or more) consisting of mainly paraffinic (including their isomers),
naphthenic and aromatic-type molecules. The small quantities of non-hydrocarbon ele-
ments such as sulfur, nitrogen, vanadium, nickel, chromium and oxygen present in the
crude oil are usually present as part of complex structures involving hydrocarbon^.^?^
While olefinic hydrocarbons are generally not present in the whole crude, various frac-
tions of petroleum may contain a small percentage as a result of some dehydrogena-
tion of paraffins and naphthenes that could occur during downstream processing.
1000 -
-- 80
3 800-
-- 60
i$ 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
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
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)
i-
f
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 Cpl and Cpgare calculated using the Kesler-Lee9 correlation and AHvapis
calculated using the generalized thermodynamic correlation reported by Lee and
Kesler16 as before.
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.
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