Professional Documents
Culture Documents
SCHOOL OF ENGINEERING
LOOP-REACTOR SERIES
DOCTOR OF PHILOSOPHY
in Chemical Engineering
Vassileios Touloupides
Chemical Engineer
LOOP-REACTOR SERIES
‘Η έγκριση της παρούσης ∆ιδακτορικής ∆ιατριβής από το Τμήμα Χημικών Μηχανικών του
Αριστοτελείου Πανεπιστημίου Θεσσαλονίκης δεν υποδηλώνει αποδοχή των γνωµών του
συγγραφέως’ (Ν. 5343/1932, άρθρο 202, παρ. 2)
Acknowledgements
you'll never do the best you might have done with what you should have had’
Aris Rutherford*
*
Author’s note: If a job's worth doing, it's worth doing well.
*http://www.wordle.net/
i
Έκθεση Πρωτοτυπίας
Touloupides V., Kanellopoulos V., Krallis A., Pladis P. and Kiparissides C., 2010,
‘Modeling and Simulation of Particle Size Distribution in Slurry-Phase Olefin Catalytic
Polymerization Industrial Loop Reactors’, Computer Aided Chemical Engineering, 28, pp.
43-48.
Krallis A., Pladis P., Kanellopoulos V., Saliakas V., Touloupides V. and
Kiparissides C., 2010, ‘Design, Simulation and Optimization of Polymerization Processes
ii
Touloupides V., Kanellopoulos V., Pladis P., Kiparissides C., Mignon D. and Van-
Grambezen P., 2010, ‘Modeling and Simulation of an Industrial Slurry-phase Catalytic
Olefin Polymerization Reactor Series’, Chem. Eng. Sci., 65, pp. 3208-3222.
Touloupides V., Kanellopoulos V., Pladis P. and Kiparissides C., ‘Modeling and
Simulation of Particle Size Distribution in Slurry-Phase Olefin Catalytic Polymerization
Industrial Loop Reactors’, 20th European Symposium on Computer Aided Process
Engineering (ESCAPE 20), 6-9 June, 2010, Ischia, Naples, Italy.
Krallis A., Pladis P., Kanellopoulos V., Saliakas V., Touloupides V. and
Kiparissides C., ‘Design, Simulation and Optimization of Polymerization Processes Using
Advanced Open Architecture Software Tools’, 20th European Symposium on Computer
Aided Process Engineering (ESCAPE 20), 6-9 June, 2010, Ischia, Naples, Italy.
Touloupides V., Kanellopoulos V., Pladis P., Krallis A. and Kiparissides C.,
‘Modeling and Simulation of Slurry-Phase Olefin Catalytic Polymerization Industrial Loop
Reactors. Prediction of Polymer Particle Growth, Molecular, Rheological and
Morphological Polymer Properties’, AIChE 09/ Annual Meeting, November 8-13, 2009,
Nashville, USA.
iii
Touloupides V., Kanellopoulos V., Pladis P., Krallis A. and Kiparissides C., ‘Real-
time Simulation of an Industrial Scale Slurry Loop Reactor’, 8th World Congress of
Chemical Engineering (WCCE8), 23-27 August, 2009, Montreal, Canada.
Τουλουπίδης Β., Κανελλόπουλος Β., Πλαδής Π., Κράλλης Α. και Κυπαρισσίδης Κ.,
‘Δυναμική Προσομοίωση Βιομηχανικού Καταλυτικού Αντιδραστήρα Ανακύκλωσης
Πολυμερισμού’, 7ο Πανελλήνιο Επιστημονικό Συνέδριο Χημικής Μηχανικής, 3-5 Ιουνίου,
2009, Πάτρα, Ελλάδα.
Touloupides V., Kanellopoulos V., Pladis P., Krallis A. and Kiparissides C.,
‘Modeling of Industrial Catalytic Olefin Polymerization Slurry Reactors’ 7th Panhellenic
Polymer Conference, 28 Sep.-1 Oct., 2008, Ιωάννινα, Ελλάδα.
Dompazis G., Kanellopoulos V., Touloupides V., and Kiparissides C., ‘Development
of a Multi-Scale, Multi-Phase, Multi-Zone Dynamic Model for the Prediction of Particle
Segregation in Catalytic Olefin Polymerization Fbrs’, AIChE 06/ Annual Meeting, Nov.
12-17, 2006, San Francisco.
Περίληψη
Abstract
According to the proposed modelling approach, each loop reactor (i.e., consisting of
the loop reactor and the settling legs) is modelled as an ideal CSTR in series with a semi-
continuous product removal unit. Dynamic macroscopic mass and energy balances are
derived for each loop reactor in the series to predict the time variation of the concentrations
of the various molecular species as well as the reactor and jacket temperatures in the two
loop reactors. The non-continuous product withdrawal rate as well as the outflow species
concentrations from each reactor of the configuration is properly calculated via the solution
of a settling leg model. A multi-site Z-N kinetic mechanism is employed to calculate the
dynamic evolution of the average and distributed molecular properties (i.e., Mn, Mw,
MWD) of the polyolefin in each reactor of the series, in conjunction with the well-known
method of moments. In order to determine the minimum number of different catalyst active
sites for the reconstruction of the experimentally measured MWDs, as well as the kinetic
parameters values, a MWD deconvolution analysis was also employed. The Sanchez-
Lacombe equation of state is utilized to calculate the thermodynamic equilibrium
concentrations of the various molecular species (i.e., monomer(s) diluent, hydrogen, etc) in
the two phases (i.e., solids and liquid) present in the reaction mixture. All different control
loops, typically employed by an industrial slurry-phase loop reactor, are incorporated in the
reactor model. Moreover, a rheological model is employed for the prediction of the
viscoelastic behaviour of polyolefins produced in slurry-loop reactors (i.e., complex
viscosity and melt flow index). Finally, a generalized dynamic population balance model is
employed to investigate the particle-size distribution developments.
Extensive numerical simulations were carried out to assess the effects of process
operating conditions on key process variables as well as on the molecular, rheological and
morphological properties of polyolefins. The results are qualitatively consistent with the
viii
expected behavior for the process variables and polymer properties. Moreover, it has been
found that model predictions are in excellent agreement (quantitatively) with dynamic
measurements under different plant operating policies (i.e., startup, grade transition, etc.)
obtained from an industrial slurry-phase catalytic olefin polymerization loop-reactor series.
The proposed model should find wide application in the design, optimization and control
of industrial polyolefin slurry-phase loop-reactor processes.
ix
Contents
1. Introduction ................................................................................................................................................ 1
2.1. Introduction....................................................................................................................................... 25
2.2. Polyolefin Production Processes ....................................................................................................... 25
2.3. Processes Overview .......................................................................................................................... 33
2.3.1. High Pressure Processes............................................................................................................. 34
2.3.1.1. Autoclave Process............................................................................................................... 34
2.3.1.2. Tubular Process .................................................................................................................. 35
2.3.2. Solution Processes ..................................................................................................................... 36
2.3.2.1. Dowlex (Propietary: Dow).................................................................................................. 37
2.3.2.2. Sclairtech/ AST (Licensor: Nova)....................................................................................... 38
2.3.2.3. Compact (Licensor: Stamicarbon-SABIC) ......................................................................... 39
2.3.3. Slurry Processes ......................................................................................................................... 40
2.3.3.1. Phillips Slurry Loop (Licensor: Chevron Phillips) ............................................................. 41
2.3.3.2. Solvay Loop (Licensor: Solvey) ......................................................................................... 41
2.3.3.3. Hostalen (Licensor: LyondellBasell) .................................................................................. 43
2.3.3.4. Nissan (Licensor: Nissan) ................................................................................................... 43
2.3.3.5. CX Process (Licensor: Mitsui)............................................................................................ 44
2.3.4. Gas-phase Processes .................................................................................................................. 45
x
3.1. Introduction....................................................................................................................................... 57
3.2. Catalytic Systems .............................................................................................................................. 57
3.2.1. Polyolefin Catalytic Systems Historical Survey ........................................................................ 58
3.2.2. Ziegler-Natta Catalysts .............................................................................................................. 59
3.2.2.1. Stereochemical Control by Z-N Catalysts .......................................................................... 63
3.2.2.2. Mechanism of Z-N Catalyzed Polymerization.................................................................... 63
3.2.2.3. Evolution of Z-N catalysts .................................................................................................. 64
3.2.3. Metallocene Catalysts ................................................................................................................ 67
3.2.3.1. Stereoregulation .................................................................................................................. 68
3.2.3.2. Mechanism of Metallocene Catalyzed Polymerization....................................................... 69
3.2.4. Cromium Catalysts..................................................................................................................... 72
3.2.4.1. Mechanism of Chromium Catalyzed Polymerization ......................................................... 73
3.3. Generalized Polymerization Kinetic Scheme.................................................................................... 75
3.3.1. Activation of Potential Sites ...................................................................................................... 76
3.3.2. Chain Initiation .......................................................................................................................... 76
3.3.3. Chain Propagation...................................................................................................................... 77
3.3.4. Chain Transfer ........................................................................................................................... 77
3.3.5. Site Transformation ................................................................................................................... 77
3.3.6. Catalyst Deactivation ................................................................................................................. 78
3.4. Method of Moments.......................................................................................................................... 80
3.5. Calculation of Molecular Properties.................................................................................................. 84
3.6. Molecular Weight Distribution Reconstruction ................................................................................ 86
3.7. Slurry Loop Reactor Kinetic Scheme................................................................................................ 88
3.7.1. Kinetic Modelling Literature Review ........................................................................................ 89
3.7.2. Kinetic Rate Constants Sensitivity Analysis.............................................................................. 93
3.7.2.1. Definition............................................................................................................................ 93
3.7.2.2. Sensitivity Analysis Procedure ........................................................................................... 94
3.7.3. Simplified Slurry Loop Reactor Kinetic Scheme..................................................................... 101
3.7.3.1. Simplified Production/ Consumption Rates...................................................................... 101
3.7.4. Deconvolution of Molecular Weight Distribution ................................................................... 104
3.7.5. Kinetic Rate Constants Estimation Procedure ......................................................................... 110
xi
List of Tables
List of Figures
Figure 5.15: Dynamic evolution of the coolant’s inlet temperature and the temperature difference
between the coolant outlet and inlet values in the first reactor. ........................................................... 228
Figure 5.16: Dynamic evolution of the coolant’s inlet temperature and the temperature difference
between the coolant outlet and inlet values in the second reactor........................................................ 229
Figure 5.17: Dynamic evolution of polymer production rate at the exit of the first and second reactor of
the series. ............................................................................................................................................. 230
Figure 5.18: Dynamic evolution of product density at the exit of the first and second reactor of the
series. ................................................................................................................................................... 230
Figure 5.19: Dynamic evolution of average molecular weight at the exit of the first and second reactor
of the series. ......................................................................................................................................... 231
Figure 5.20: Dynamic evolution of polydispersity at the exit of the first and second reactor of the series. .. 232
Figure 5.21: Dynamic evolution of the MWD of the polymer produced in the first (a.) and in the second
reactor (b.) and overall MWD at the exit of the second reactor of the series (c.) (startup case). ......... 233
Figure 5.22: Dynamic evolution of the complex viscosity of the polymer produced in the first reactor
(startup case). ....................................................................................................................................... 235
Figure 5.23: Dynamic evolution of the complex viscosity of the polymer at the exit of the second reactor
(startup case). ....................................................................................................................................... 235
Figure 5.24: Dynamic evolution of the solids enhancement factor and valve opening cycling period in
the second reactor. ............................................................................................................................... 236
Figure 5.25: Dynamic evolution of the solids concentration and iso-butane inflow rate in the second
reactor. ................................................................................................................................................. 237
Figure 5.26: Dynamic evolution of the number average molecular weight and polydispersity index in the
second reactor. ..................................................................................................................................... 238
Figure 5.27: Dynamic evolution of the co-polymer composition in the first and second reactor of the
series. ................................................................................................................................................... 239
Figure 5.28: Dynamic evolution of the product density in the first and second reactor of the series............ 240
Figure 5.29: Dynamic evolution of the MWD and complex viscosity in the first reactor (effect of
hydrogen concentration). ..................................................................................................................... 241
Figure 5.30: Dynamic evolution of the MWD and complex viscosity in the second reactor (effect of
hydrogen concentration). ..................................................................................................................... 241
Figure 5.31: Molecular weight distribution of polymer at the exit of the first and second reactor of the
series, regarding a low-co-monomer content polymer grade. .............................................................. 243
Figure 5.32: Molecular weight distribution of polymer at the exit of the first and second reactor of the
series, regarding a medium-co-monomer content polymer grade. ....................................................... 243
Figure 5.33: Molecular weight distribution of polymer at the exit of the first and second reactor of the
series, regarding a high-co-monomer content polymer grade.............................................................. 244
Figure 5.34: Dynamic evolution of ethylene feed in the first reactor (Startup case). .................................... 245
Figure 5.35: Dynamic evolution of hexane feed in the first reactor (Startup case). ...................................... 246
xix
Figure 5.36: Dynamic evolution of hydrogen feed in the first reactor (Startup case).................................... 246
Figure 5.37: Dynamic evolution of iso-butane feed in the first reactor (Startup case). ................................. 247
Figure 5.38: Dynamic evolution of catalyst feed in the first reactor (Startup case). ..................................... 247
Figure 5.39: Dynamic evolution of ethylene off-gas composition in the first reactor (Startup case). ........... 248
Figure 5.40: Dynamic evolution of hexane off –gas composition in the first reactor (Startup case)............. 248
Figure 5.41: Dynamic evolution of hydrogen off-gas composition in the first reactor (Startup case). ......... 249
Figure 5.42: Dynamic evolution of solids concentration in the first reactor (Startup case). ......................... 249
Figure 5.43: Dynamic evolution of reactor phase density in the first reactor (Startup case)......................... 250
Figure 5.44: Dynamic evolution of temperature in the first reactor (Startup case). ...................................... 250
Figure 5.45: Dynamic evolution of coolant’s inlet temperature in the first reactor (Startup case)................ 251
Figure 5.46: Dynamic evolution of and the temperature difference between the coolant outlet and inlet
values in the first reactor (Startup case)............................................................................................... 251
Figure 5.47: Dynamic evolution of the valve opening cycling period in the first reactor (Startup case). ..... 252
Figure 5.48: Dynamic evolution of polymer production rate in the first reactor (Startup case). ................... 252
Figure 5.49: Dynamic evolution of product density in the first reactor (Startup case).................................. 253
Figure 5.50: Dynamic evolution of ethylene feed in the second reactor (Startup case). ............................... 253
Figure 5.51: Dynamic evolution of hydrogen feed in the second reactor (Startup case)............................... 254
Figure 5.52: Dynamic evolution of iso-butane feed in the second reactor (Startup case). ............................ 254
Figure 5.53: Dynamic evolution of ethylene off-gas composition in the first reactor (Startup case). ........... 255
Figure 5.54: Dynamic evolution of hexane off –gas composition in the second reactor (Startup case). ....... 255
Figure 5.55: Dynamic evolution of hydrogen off-gas composition in the second reactor (Startup case)...... 256
Figure 5.56: Dynamic evolution of solids concentration in the second reactor (Startup case)...................... 256
Figure 5.57: Dynamic evolution of reactor phase density in the second reactor (Startup case). ................... 257
Figure 5.58: Dynamic evolution of temperature in the second reactor (Startup case)................................... 257
Figure 5.59: Dynamic evolution of coolant’s inlet temperature in the second reactor (Startup case). .......... 258
Figure 5.60: Dynamic evolution of and the temperature difference between the coolant outlet and inlet
values in the second reactor (Startup case). ......................................................................................... 258
Figure 5.61: Dynamic evolution of the valve opening cycling period in the second reactor (Startup case).. 259
Figure 5.62: Dynamic evolution of polymer production rate in the second reactor (Startup case). .............. 259
Figure 5.63: Dynamic evolution of product density in the second reactor (Startup case). ............................ 260
Figure 5.64: Dynamic evolution of number average molecular weight in the first and second reactor
(Startup case). ...................................................................................................................................... 260
Figure 5.65: Dynamic evolution of polydispersity in the first and second reactor (Startup case). ................ 261
Figure 5.66: Dynamic evolution of the PSD in the first reactor (startup case).............................................. 262
Figure 5.67: Dynamic evolution of the PSD in the second reactor (startup case). ........................................ 262
Figure 5.68: Initial catalyst size distribution (Startup case). ......................................................................... 263
Figure 5.69: Dynamic evolution of ethylene feed in the first reactor (28 days case). ................................... 265
Figure 5.70: Dynamic evolution of hexane feed in the first reactor (28 days case). ..................................... 265
xx
Figure 5.71: Dynamic evolution of hydrogen feed in the first reactor (28 days case)................................... 266
Figure 5.72: Dynamic evolution of iso-butane feed in the first reactor (28 days case). ................................ 266
Figure 5.73: Dynamic evolution of catalyst feed in the first reactor (28 days case)...................................... 267
Figure 5.74: Dynamic evolution of ethylene off-gas composition in the first reactor (28 days case). .......... 267
Figure 5.75: Dynamic evolution of hexane off –gas composition in the first reactor (28 days case)............ 268
Figure 5.76: Dynamic evolution of hydrogen off-gas composition in the first reactor (28 days case).......... 268
Figure 5.77: Dynamic evolution of solids concentration in the first reactor (28 days case).......................... 269
Figure 5.78: Dynamic evolution of reactor phase density in the first reactor (28 days case). ....................... 269
Figure 5.79: Dynamic evolution of temperature in the first reactor (28 days case)....................................... 270
Figure 5.80: Dynamic evolution of coolant’s inlet temperature in the first reactor (28 days case). .............. 270
Figure 5.81: Dynamic evolution of and the temperature difference between the coolant outlet and inlet
values in the first reactor (28 days case). ............................................................................................. 271
Figure 5.82: Dynamic evolution of the valve opening cycling period in the first reactor (28 days case)...... 271
Figure 5.83: Dynamic evolution of polymer production rate in the first reactor (28 days case). .................. 272
Figure 5.84: Dynamic evolution of product density in the first reactor (28 days case). ................................ 272
Figure 5.85: Dynamic evolution of ethylene feed in the second reactor (28 days case)................................ 273
Figure 5.86: Dynamic evolution of hydrogen feed in the second reactor (28 days case). ............................. 273
Figure 5.87: Dynamic evolution of iso-butane feed in the second reactor (28 days case)............................. 274
Figure 5.88: Dynamic evolution of ethylene off-gas composition in the first reactor (28 days case). .......... 274
Figure 5.89: Dynamic evolution of hexane off –gas composition in the second reactor (28 days case). ...... 275
Figure 5.90: Dynamic evolution of hydrogen off-gas composition in the second reactor (28 days case). .... 275
Figure 5.91: Dynamic evolution of solids concentration in the second reactor (28 days case). .................... 276
Figure 5.92: Dynamic evolution of reactor phase density in the second reactor (28 days case). .................. 276
Figure 5.93: Dynamic evolution of temperature in the second reactor (28 days case). ................................. 277
Figure 5.94: Dynamic evolution of coolant’s inlet temperature in the second reactor (28 days case). ......... 277
Figure 5.95: Dynamic evolution of and the temperature difference between the coolant outlet and inlet
values in the second reactor (28 days case). ........................................................................................ 278
Figure 5.96: Dynamic evolution of the valve opening cycling period in the second reactor (28 days case). 278
Figure 5.97: Dynamic evolution of polymer production rate in the second reactor (28 days case)............... 279
Figure 5.98: Dynamic evolution of product density in the second reactor (28 days case). ........................... 279
Figure 5.99: Dynamic evolution of number average molecular weight in the first and second reactor
(28 days case)....................................................................................................................................... 280
Figure 5.100: Dynamic evolution of polydispersity in the first and second reactor (28 days case). ............. 280
Chapter 1: Introduction 1
Chapter 1
1. Introduction
1.1. Polymers
Historians frequently classify the early ages of man according to the materials that he
used for making his implements and other basic necessities. The most well known of these
periods are the Stone Age, the Iron Age and the Bronze Age. Such a system of
classification cannot be used to describe subsequent periods for with the passage of time
man learnt to use more materials. By the time of the ancient civilisations of Egypt,
Babylonia and Greece, man was employing a range of metals, stones, woods, ceramics,
glasses, skins, horns and fibres. Until the 19th century, man’s inanimate possessions, his
home, his tools, his furniture, were made from varieties of these eight classes of material.
During the last century and a half, a new class of material has been introduced which
has not only challenged the older materials for their well-established uses but has also
made possible new products which have helped to extend the range of activities of
mankind. Synthetic polymers have revolutionized our life. Today, without this class of
materials, it is difficult to conceive how such everyday features of modern life could ever
have been developed (Brydson, 1999). Plastics, fibers, elastomers, adhesives, rubber,
protein, cellulose-these all common terms in our modern vocabulary, and all a part of the
fascinating world of polymer chemistry. Innumerable examples of synthetic polymers may
be cited. Synthetic polymer materials such as nylon, polyethylene, teflon, and silicone have
formed the basis for a burgeoning polymer industry. Synthetic polymers today find
2 Chapter 1: Introduction
application in nearly every industry and area of life. Polymers are widely used as adhesives
and lubricants, as well as structural components for products ranging from children's toys
to aircraft. Moreover, they have been employed in a variety of biomedical applications
ranging from implantable devices to controlled drug delivery. Conclusively, it would not
be an exaggeration to say that we live in the Polymer age.
The word polymer is derived from the Greek words πολυ- (-poly-) meaning ‘many’;
and μέρος (-meros-) meaning ‘part’ (polymeric comes directly from the Greek ‘πολυμερές’
which means ‘having many parts’). The term was coined in 1833 by Joens Jakob Berzelius,
although his definition of a polymer was different from the modern definition. Initially,
according to Berzelius, the term polymer described organic compounds which shared
identical empirical formulas but differed in overall molecular weight. The larger of the
compounds was described as ‘polymers’ of the smallest. Thus, according to this -now
obsolete- definition, glucose, C6H12O6, would be a polymer of formaldehyde, CH2O.
way into a wide spectrum of consumer products, and its inventor had achieved the ultimate
visibility, featuring on a cover of Time. Other polymers, notably alkyd (polyester) paints
and polybutadiene rubber, were introduced about the same time. Yet despite such
commercial successes, most scientists had no clear concept of polymer structure. The
prevailing theory was that polymers were aggregates of small molecules, much like
colloids, but were held together by some ‘mysterious’ secondary force.
This aggregation or association theory eventually gave way, with no small amount of
resistance, to the theories of the german chemist Herman Staudinger, who attributed to the
remarkable properties of polymers to ordinary intermolecular forces between molecules of
very high molecular weight. Staudinger was also the first to introduce the term
‘macromolecül’ (i.e., macromolecule). In recognition of his contributions, Staudinger was
awarded the Nobel Prize in Chemistry in 1953. In the 1930’s, the brilliant work of the
American chemist Wallace Hume Carrothers placed the theories of Staudinger on a firm
experimental basis and led to the commercial development of neoprene rubber and
polyamide (nylon) fibers.
World War II led to significant advances in polymer chemistry, particularly with the
development of synthetic rubber when the natural rubber-growing regions of the Far East
became inaccessible to the Allies. Among the more significant developments of the
postwar years, was the discovery by Karl-Ziegler in Germany of new coordination
catalysts for initiating polymerization reactions and the application by Giulio Natta in Italy
of these new systems to development of polymers having controlled stereochemistry. Their
work has revolutionized the polymer industry, for these so-called stereoregular polymers
have mechanical properties superior in most instances to those of non-stereoregular
polymers. The importance of their discoveries was recognized by the award of the Nobel
Prize in Chemistry jointly to Ziegler and Natta in 1963. Equally significant was the work
of Paul Flory (Nobel Prize 1974), who established a quantitative basis for polymer
behaviour, whether it be physical properties of macromolecules in solution or in bulk or
such chemical phenomena as crosslinking and chain transfer (Stevens, 1999).
were non-existent. Polymers exhibit a very wide range of properties which allow for their
extensive use in society. Packaging remains the biggest end-use for polymers (38%)
followed by building and construction (21%), automotive (7%) and electrical and
electronic uses (6%). Other applications, which include medical and leisure, use 28%.
(Figure 1.1/ Source: PlasticsEurope Market Research Group (PEMRG)).
Packaging
38%
Others
28%
E&E
6% Building and
Construction
Automotive
7% 21%
1.2. Polyolefins
Among the broad variety of polymeric materials, polyolefins have received a special
attention during recent years. Polyolefins are among the most important commodity
plastics today due to their low production costs, reduced environmental impact, and a very
wide range of applications. They are found in products as diverse as prosthetic implants,
gas pipelines, automobile parts and accessories, synthetic fabrics, films, containers, and
toys. It may seem surprising that polymers composed only of carbon and hydrogen atoms
can be so flexible, but the versatility of polyolefins can be easily explained by the way the
monomer molecules ethylene, propylene, and higher a-olefins are connected to form the
polymer chains. It is true for any polymer, but dramatically so for polyolefins, that chain
microstructure is the key to understanding their physical properties (Meyer and Keurentjes,
2005).
Chapter 1: Introduction 5
Polyolefin is a polymer produced from a simple alkene (also called an olefin, with
the general formula CnH2n) as a monomer. This type of polymers includes large volume
materials such as homopolymers of ethylene (polyethylene, PE), and propylene
(polypropylene, PP), co-polymers of ethylene, propylene, higher olefins dienes and cyclic
olefins, as well as specialty materials such as ethylene-propylene elastomers (EPR and
EP(D)M) and polybutene-1 (PB-1).
We may state, even considering the universality of materials, that polyolefins are
unique in their ability to perfectly match the basic prerequisites for the successful growth
of a material (Galli, 1994):
Polyolefins (i.e., polyethylene and polypropylene) are the world’s mostly produced
and fastest growing polymer family. The basis of their dynamic development and still
tremendous potential lies in:
3) The energy savings during their production and use, in comparison with other
materials.
Aluminium
Copper
Steel
PS
HDPE
PVC
LDPE
PP
0 2 4 6 8 10 12 14 16
Mass Fraction of Oil Equivalents to Raw Material
Figure 1.2: Total consumption of energy and raw materials in kilogram equivalents of oil,
to obtain 1 kg of material indicated.
1.3. Polyethylene
low price,
excellent electrical insulation properties over a wide range of frequencies,
very good chemical resistance,
good processability, toughness,
flexibility, and,
transparency (in thin films of certain grades).
Chapter 1: Introduction 7
During the 1950’s three research groups working independently discovered three
different catalysts which allowed the production of essentially linear polyethylene at low
pressure and temperature. These polymers had densities in the region of 0.960 g/cm3, and
became known as high-density polyethylenes (HDPE), in contrast to the polymers
produced by the extensively commercialized high-pressure process, which were named
low-density polyethylenes (LDPE). These discoveries laid the basis for the coordination
catalysis of ethylene polymerization, which has continued to diversify. Of the three
discoveries at Standard Oil (Indiana), Phillips Petroleum, and by Karl Ziegler at the Max-
Planck-Institut für Kohlenforschung, the latter two have been extensively commercialized.
More recently the observation that traces of water can dramatically increase the
polymerization rate of certain Ziegler catalysts has led to major developments in soluble
coordination catalysts and later their supported variants.
The coordination catalysts allowed for the first time the co-polymerizaton of
ethylene with other olefins such as butene, which by introducing side branches reduces the
crystallinity and allows a low-density polyethylene to be produced at comparatively low
pressures. Although Du Pont of Canada introduced such a process in 1960, worldwide, the
products remained a small-volume specialty until 1978 when Union Carbide announced
their Unipol process and coined the name linear low-density polyethylene (LLDPE). In
addition to developing a cheaper production process, Union Carbide introduced the
8 Chapter 1: Introduction
concept of exploiting the different molecular structure of the linear product to make
tougher film. Following this lead, LLDPE processes have been introduced by many other
manufacturers. Present market needs, combined with the broad range of polyolefin
applications, have forced the polyolefin industry to operate under frequent grade transition
policies, producing more than one type of product (swing plants).
Additionally, co-polymers are made by both types of process (i.e., free radical and
catalytic polymerization). The free-radical process is used to produce co-polymers of vinyl
acetate, acrylates, methacrylates, and the corresponding acids, but chain transfer prevents
the use of higher olefins because of the drastic reduction in molecular mass of the polymer.
The coordination catalysts are able to co-polymerize olefins, but are deactivated by more
polar materials. Because of the complex interplay of the capabilities of modern plants, it is
convenient to treat separately the products, the catalysts, and the processes (Ullmann’s
Encyclopedia of Industrial Chemistry).
Most polyethylene molecules contain branches in their chains which are formed
spontaneously, in case of radical polymerization, or deliberately by copolymerization of
ethylene with α−olefins, in case of catalytic polymerization. In general, the density of PE
decreases with an increase in branch numbers; the more branching, the lower the density
(Shirayama et al., 1972, Sinclair, 1983). The crystallinity of PE decreases significantly
with an increase in branch frequency and size. Polyethylene resins are classified according
to their density which partly depends on the type of branching. The American Society for
Testing and Materials (ASTM) has classified PE into four groups:
High density polyethylene (HDPE) is essentially free of both long and short
branching and thus has stronger intermolecular forces (although very small amounts may
be deliberately incorporated to achieve specific product targets). Low degree of branching
leads to a white opaque solid material The MWD depends on the catalyst type but is
typically of medium width. It is produced mainly in slurry and gas-phase polymerization
processes (Ullmann’s Encyclopedia of Industrial Chemistry). Medium density
polyethylene (MDPE) is a type of polyethylene defined by a density range of 0.926-0.940
g/cm3. It is less dense than HDPE, which is more common. MDPE has a high degree of
resistance to chemicals and is very easy to keep clean. MDPE can be produced by
chromium/ silica catalysts, Ziegler-Natta catalysts or metallocene catalysts.
Molecular
architecture
Branching
High Little or none High
amount
Branching
Long and short Short Short
type
Density,
0.915-0.935 0.935-0.975 0.915-0.935
g/cm3
Main
Film Blow and injection molding Film
application
Production Free radical in molten Slurry, gas phase, solution Slurry, gas phase, solution
process polymer phase phase
Pressure,
69-276 2-5 2-5
Mpa
Film
52%
Other non-
extrusion
3%
Other extrusion Injection
2% Molding
12%
Sheet
2% Pipe and
Conduit
Rotational
molding Extrusion 7%
1% Wire and cable Blow Molding Coating
4%
3% 14%
1.4. Polypropylene
In the 1950’s Karl Ziegler discovered new catalysts which were to revolutionize the
plastics industry. Before then, propene could not be polymerized to high molecular mass
products with the catalysts then available, namely, free-radical, anionic and cationic
systems. Even the most favorable of these yielded only liquids consisting of many isomers
and greaselike materials unsuitable for making hard plastic products. Industrial-scale
production of polypropylene started at Ferrara, Italy, in 1957 with a 6×103 tonne/ year
plant. Nowadays, global capacity exceeds 70×103 tonne/ year (see also 1.6. Polyolefin
Market). There are several reasons for this phenomenal growth. At first propene was
readily available, almost at byproduct prices, from petrochemical cracker plants making
ethylene. The polymer itself was suitable for a wide range of existing and new applications
such as films, various fibers, large and small moldings from boat hulls to instrument parts.
In addition, new manufacturing technologies based on MgCl2 supported catalyst systems
yielded both cost savings and improved products.
which misplacements vary in frequency and in number. The tendency here is for the high
molecular mass fraction to contain fewer faults, while the short chains suffer from
excessive disruptions, which lead to reduced crystallinity.
Polymerization of
polypropylene
a. Isotactic
b. Chain imperfection
c. Syndiotactic
d. Atactic
Fiber
15%
Injection
Molding
Blow molding
37%
1%
Other Pipe and
8% profile
3%
The production of polymers has become a major item in the economy. Polymers
contribute to sustainable development and bring quality of life to citizens. It is sufficient to
say that the polymer industry within the European Union provide employment to 1.6
million people (two thirds of the number of employees in the automotive sector) and
indirectly to many times more in industries which are enabled or depend on plastics for
their products.
The diversity of applications, and the fact that present technology allows the
tailoring of different products for different applications make synthetic polymers very
Chapter 1: Introduction 15
attractive (Chiovetta, 1983). As an example, plastics make many goods in our daily life
more affordable and reduce the wastage of many valuable resources. The consumption of
polymers has increased considerably in the last tricennial and, currently, the requirement
for new products has risen. In Figure 1.7, both world and Europe production of polymers is
presented (Source: PlasticsEurope Market Research Group (PEMRG)). Nowadays, the
overall annual world polymer production may exceed 250 Mt. The data reflect a constant
increase in polymer production in spite of criticism from environmentalists. Nevertheless,
nowadays, polymer waste has become an urgent topic for industry, providing new and
challenging areas of research and development on recycling, reuse, and degradation.
Today, Europe produces approximately 60 million tonnes of plastics, representing 25% of
the global plastics production. Plastic production facilities are well placed across Europe.
Germany is the major producer, accounting for 7.5% of global production followed by
Benelux (4.5%), France (3%), Italy (2%) and the UK and Spain (1.5%) (Figure 1.8)
(Source: PlasticsEurope Market Research Group).
There are around 20 distinct groups of polymers, each with numerous grades
available to help make the best choice for each specific application. There are five high-
volume plastics families; polyethylene (including low density (LDPE), linear low density
(LLDPE) and high density (HDPE)), polypropylene (PP), polyvinylchloride (PVC),
polystyrene (solid PS and expandable EPS) and polyethylene terephthalate (PET).
Together the big five account for over 90% of all polymer demand (see Figure 1.9). It
should also be noted that polyolefins (i.e., HDPE, LDPE, LLDPE and PP) cover 60% of
the total polymer production (source: CMAI report).
16 Chapter 1: Introduction
300
World production 2008: 245
250
Annual Production, Mt Europe production
2002: 200
200
150
1989:100
100
2008: 60
1976: 50
50
1950: 1.5
0
1950 1960 1970 1980 1990 2000 2010
Year
PP
ABS 24%
4%
PVC
19% HDPE
17%
PET
7%
PC LDPE
2% 11%
PS
6% LLDPE
10%
Polyolefins are the most widely used plastics today due to their low production cost,
reduced environmental impact, and wide range of applications (e.g., packaging, building
and construction, transportation, electrics and electronics, furniture, etc.) (see Figure 1.4
and Figure 1.6). It is believed that the degree of technological and scientific sophistication
in relation to the polyolefin manufacturing has no equal among other synthetic polymer
production processes.
The following main driving forces for commercial expansion of polyolefins are:
1) The further strengthening of the already excellent cost/ property balance and the
good environmental image of polyolefins with respect to all other materials.
3) Production cost reduction. In the PE area, this means continuous market growth,
mostly in the areas of the most specialized materials, such as high-density
polyethylene (HDPE) and linear-low-density polyethylene (LLDPE).
18 Chapter 1: Introduction
This progress has been made possible by continuous polymer property developments
perfectly matching the needs of many sophisticated areas (see Figure 1.10) (Galli and
Vecellio, 2004). The history of polyolefin growth speaks for itself. The explosion in the
dynamic development of PE and PP, which is still in progress (Figure 1.11), started mainly
in the 1970’s (Figure 1.12). It appears even more impressive if it is compared with the
growth of all plastic materials. The polyolefin world market share was around 20% of the
total plastics market in the sixties, while reached almost 65% in the year 2000 (see Figure
1.13) (Galli and Vecellio, 2001, CMAI report).
Presently, the total annual world polyolefins capacity exceeds 120 Mt (Source:
CMAI report). Polyethylene and polypropylene cover 60% and 40% of the total
polyolefins production, respectively (Figure 1.14) (source: CMAI report). The annual
world-wide polyolefins market growth in the coming years is foreseen to be 4-6%, making
polyolefin manufacturing a very active research area (Mulhaupt, 2003, Galli and Vecellio,
2004, CMAI report). The major polyolefin manufacturers for 2008 are presented in Figure
1.15 and Figure 1.16 (source: Borealis Group, CMAI, LyondellBasell).
120
Total polyolefin
production
PP
80
60
40
20
0
1945 1955 1965 1975 1985 1995 2005
Year
140
PP
120
Annual Consumption, Mt
HDPE
100 LLDPE
LDPE
80
60
40
20
0
1970 1980 1990 2000 2005 2010
Year
70
60
Polyolefins Volume % 50 PP
of Total Plastics
40
30
PE
20
10
0
1951
1955
1959
1963
1967
1971
1975
1979
1983
1987
1991
1995
1999
2003
Year
PP
40%
HDPE
27%
LDPE
16%
LLDPE
17%
Lyondel Bassel
Borealis
Ineos Borealis
Ineos 13% 14%
Dow
Dow 9%
Lyondel Bassel
SABIC 19%
SABIC
9%
Total
Polimeri Europa
Total
Exxon Mobil Other
8%
8%
Repsol Polimeri Europa TVK
6% Exxon Mobil 4%
TVK Repsol
5% 5%
0 1 2 3 4 5 6
Annual Production, Mt
Lyondel Bassel
Exxon Mobil
Other CNPC
Dow 45% 3% Formosa Grp
4%
SINOPEC Borealis
4%
SABIC Total
4%
Ineos
Ineos
5%
Total
SABIC
Borealis 5%
Mathematical models for an industrial reactor are a unique tool to the study of
different research areas:
Chapter 1 Introduction
1.8. References
Chiovettta M.G., 1983, ‘Heat and Mass Transfer During the Polymerization of Alpha-
olefins from the Gas Phase’, Phd Thesis.
Galli P., 1994, ‘The Breakthrough in Catalysis and Processes for Olefin Polymerization:
Innovative Structures and a Strategy in the Materials Area for the Twenty-first
Caentury’, Prog.Polym.Sci., 19, pp.959-974.
Galli P. and Vecellio G., 2001, ‘Technology: Driving Force Behind Innovation and Growth
of Polyolefins’, Prog. Polym. Sci., 26, 8, pp. 1287-1336.
Galli P. and Vecellio G., 2004, ‘Polyolefins: The Most Promising Large-Volume Materials
for the 21st Century’, Journal of Polymer Science, Part A: Polymer Chemistry, Vol.
42, pp. 396-415.
Mattice W.L. and Stehling F.C., 1981, ‘Branch Formation in Low-density Polyethylene’,
Macromolecules, 14, pp. 1479-1484.
Meyer T.and Keurentjes J., 2005, ‘Handbook of Polymer Reaction Engineering’, Wiley-
VCH, Weinheim.
Mulhaupt R., 2003, ‘Catalytic Polymerization and Post Polymerization Catalysis Fifty
Years After the Discovery of Ziegler’s Catalysts’, Macromol. Chem. Phys., 204, 2,
pp. 289-327.
Odian G., 2004, ‘Principles of Polymerization, Fourth Edition’, Wiley Interscience, New
Jersey.
Shirayama K., Kita S. and Watabe H., 1972, ‘Effects of Branching on Some Properties of
Ethylene/a-Olefin Copolymers’, Makromol. Chem., 151, pp. 97-120.
24 Chapter 1: Introduction
Sinclair K.B., 1983, ‘Characteristics of Linear LDPE and Description of UCC Gas Phase
Process’, Process Economics Report, SRI International, Menlo Park, CA.
Stevens M.P., 1999, ‘Polymer Chemistry, an Introduction, 3rd Edition’, Oxford University
Press, New York.
Ullmann’s Encyclopedia of Industrial Chemistry, 6th edition, 2002, John Wiley & Sons.
van der Ven S., 1990, ‘Polypropylene and other Polyolefins’, Elsevier, Amsterdam.
Xie T., McAuley K.B., Hsu J.C.C. and Bacon D.W., 1994, ‘Gas Phase Ethylene
Polymerization: Production Processes, Polymer Properties, and Reactor Modeling’,
Ind. Eng. Chem. Res., 33, pp. 449-479.
PlasticsEurope Market Research Group (PEMRG), ‘The Compelling Facts About Plastics
2009’, URL: http://www.plasticseurope.org/Documents/Document/20100225141
556- Brochure_UK_FactsFigures_2009_22sept_6_Final-20090930-001-EN-v1.pdf/,
Date retrieved: May 10, 2010.
Chapter 2: Commercial Polymerization Processes 25
Chapter 2
2.1. Introduction
The purpose of this chapter is to present an overview of the processes employed for
the industrial production of polyolefins. The major industrial polymerization processes are
discussed in detail. The basic flowsheet structure of the processes is presented along with
their typical operating conditions.
large scale production commodity polyolefins require simplified processes with efficient
transitions between product grades.
Industrial polyolefin production processes can be broadly classified in high and low
pressure processes, as shown in Figure 2.1 (Ullmann’s Encyclopedia of Industrial
Chemistry). High and low pressure processes differ in both the type of reactor used and in
the physical state of the reaction media. Moreover, in high pressure processes
polymerization takes place through free-radical kinetic mechanism, while, in low pressure
processes, catalytic systems are the usually employed. High pressure processes can be
subdivided to autoclave and tubular technologies. On the other hand, low pressure
technologies include solution, slurry-phase (suspension) and gas-phase, regarding the
reaction medium. In a solution process both the catalyst and the resulting polymer remain
dissolved in a solvent that must subsequently be removed to isolate the polymer. A slurry
process is conducted in an inert diluent in which the catalyst is affixed to an inert support,
and the polymer formed during production remains suspended in the liquid medium, never
dissolving. Slurry processes can be sub-divided into conventional slurry loop processes and
stirred tank reactor processes. Gas-phase polymerization processes are divided in stirred-
bed or the fluidized-bed process. It is noted that supported catalysts are the rule in gas-
phase processes as well. The evolution of industrial processes for polyolefin production is
briefly presented in Figure 2.2.
Each type of process has limitations on the breadth of resins it can produce in terms
of polymer density and melt flow index (see Table 2.4) which directly affect the end-use
properties of the product (Figure 2.4). Production capabilities of all different process types,
regarding PE, are shown schematically in Figure 2.3. Gas-phase processes exhibit the
widest operating window, in terms of polymer density and melt flow index of the polymer
produced. However, gas-phase processes encounter difficulties in producing resins with
very low densities due to polymer stickiness and particle agglomeration. Slurry-phase
processes produce PE, covering a broad range of melt index. However, the range of
densities attainable is limited (i.e., density values greater than 0.937 g/cm3). As density
decreases, resin solubility in the dilute increases. At a density of about 0.930 g/cm3,
28 Chapter 2: Commercial Polymerization Processes
sufficient dissolution occurs to foul the reactor. Hence, slurry processes are not suitable for
LLDPE production. On the other hand, solution processes can produce PE over a broad
range of densities, but only a limited range of molecular weights. As molecular weight
increases, solution viscosity increases also. At some point, the increased viscosity limits
reactor operability and productivity. Thus, solution processes are not suitable for making
high molecular weight PE.
Liquid- Multi-zone
Process type Loop reactor Autoclave Fluidized bed
phase autoclave
Propylene, Propylene,
Co-monomer 1-butene-1 1-butene 1-hexene
1-butene 1-butene
Bis(triphenyl
silyl)
chromate/
1-3% chromic
Catalyst Ziegler-Natta Ziegler-Natta Ziegler-Natta silica-alumina
oxide/ silica
dialkyl-
aluminium
alkoxide
Ethylene feed,
Adiabatic Adiabatic
Heat removal Jacket cooling Jacket cooling adiabatic
reactor reactor
reactor
Solution Gas-phase
Residence time,
<1 5-20 2 180-300
min
Conversion per
17-20 95 95 5
pass, %
Diluent Hyper-critical
Hydrocarbon Cyclo-hexane None
ethylene
powder bed
Residence time,
45 97 - 90
hr
Conversion per
2 2 - 1-2
pass, %
Space-time yield,
94 37.5 - 100
kg/m3/hr
LDPE + + - - -
EVA co-polymers + + - - -
Acrylate co-
+ + - - -
polymers
HDPE * - + + +
HMW HDPE - - + + -
UHMPE - - * + -
LLDPE + * + * +
VLDPE + * + - +
+ = suitable
0.97
Homopolymer
0.96 Cascade
Hostalen Slurry
Mitsui Philips
Nissan Solvay
0.95 Equistar USI
Density (kg/lt)
Maruzen
Nisseki
Borstar
0.94
0.93
Solution
Gas
Dowlex
Unipol
Compact
Innovene
0.92 Sclairtech
Spheripol
Evolue
Lupotech G
0.91
Figure 2.3: Capabilities of different processes in terms of density and melt index.
Property
Property
Figure 2.4: Effect of polymer density and MFI on end-use properties (Foster, 1991).
Chapter 2: Commercial Polymerization Processes 33
One of the main concerns during the design and control of polymerization reactors is
how to remove the heat of reaction efficiently, since polymerizations are very exothermic
reactions. In polymerization reactors, it is also imperative to be able to produce polymer
that can be easily separated from unreacted monomer, catalyst residues and other
byproducts, while assuring that polymer properties remain on target for a given
polymergrade. These stringent requirements led to the design of several reactor types, each
with their own advantages and disadvantages (Asua, 2007).
Feedstock materials and additives must be purified and catalyst material must be
prepared. In case of a high pressure process, the gas must be compressed in
several stages.
Polymerization takes place either in the gas phase (fluidized bed or stirred
reactor), the liquid phase (slurry or solution), or in a high pressure environment.
Polymerization is the heart of the processes. On any one unit, only one of the
three processes is used. More details will be explained on the next pages.
Polymer particles are then separated from still existing monomers and diluents,
pelletized, dried and dispatched.
Monomers and diluents are recovered and fed again to the process.
34 Chapter 2: Commercial Polymerization Processes
The major processes for polyolefin production are summarized below (Source:
Chemical Market Resources). Polymerization processes are discussed in detail. Typical
operating conditions are reported along with their basic flowsheet structure. Extensive
analysis on the commercial production of polyolefin can be found in the work of Choi and
Ray, 1985a, Choi and Ray, 1985b, Beach and Kissin, 1986 and Malpass, 2010.
In high pressure processes, the reactor may take one of two forms: a high pressure
autoclave or a jacketed tube. The overall processes are similar except for the design of the
reactor itself. Fresh ethylene enters the reactor and is mixed with the low pressure recycle.
After further compression the mixture enters the reactor for polymerization. Oxygen or
peroxide may be used as initiators. The reaction pressure ranges from 150-350 MPa and
temperature varies from 150-300oC. Such high pressures call for very specialized
technology and many key features have remained proprietary information. The design of
thick-walled cylinders requires a different type of analysis from that for lower pressure
vessels, and fatigue is a major design consideration for pumps and compressors.
A typical autoclave design is shown in Figure 2.6. The reaction pressure ranges from
150–200 MPa. Typical overall residence time ranges from 30–60 sec, with corresponding
reactor volumes in larger plants larger than 1 m3. The autoclave functions as an adiabatic
continuous stirred-tank reactor (CSTR), with the heat of reaction being removed by the
fresh ethylene entering the reactor. The conversion of monomer to polymer is thus related
Chapter 2: Commercial Polymerization Processes 35
to the difference in temperature between the feed gas and the final reaction temperature.
Most modern reactors have two or more zones with increasing temperatures. The reaction
temperatures are maintained constant by controlling the speeds of the pumps feeding
initiators into the respective zones. The temperature is typically set equal to 180°C and
290°C for the first zone and the final zone, respectively.
ranging from 190°C for oxygen to 140°C for a peroxydicarbonate. The latter part of the
reactor functions as a product cooler similar to that of the autoclave process.
A tubular reactor works in the plug flow regime with heat transfer to the jacket. Plug
flow is achieved by the correct choice of pipe diameter relative to the flow rate so as to
give sufficient turbulence and good axial mixing. Although heat is transferred through the
reactor wall, it is not generally possible to maintain isothermal conditions, and temperature
peaks occur. Injection of initiator at various positions along the tube produces temperature
peaks, increasing the overall conversion. Thus, higher conversions than in the autoclave
reactor can be achieved, but at a higher cost in compression energy. Although conversions
of up to 35% (compared with 20% for the autoclave) have been claimed, the maximum
useful conversion depends on the product quality required, since quality deteriorates
markedly with increasing conversion.
The first commercially available technology for the production of high density,
linear polyethylene grades was the solution process. According to this technology,
polymerization is carried out in a liquid which is a solvent for the resulting polymer, which
is then flashed to remove unreacted monomer. The solvent is recovered and recycled back
in the process. The polymer is then filtered, dried and extruded in a pellet form. Solution
processes have been developed by various companies including Du Pont, Dow, DSM, and
Mitsui for the manufacture of LLDPE or HDPE/ LLDPE on a swing basis. The advantages
Chapter 2: Commercial Polymerization Processes 37
are that they readily handle a wide range of co-monomer types and product densities and,
depending on catalyst type, tend to produce narrow molecular weight distribution products
more readily. Like the high-pressure process (which is also a solution process) they are
unable to handle high-viscosity products. Polymerization occurs in a hydrocarbon solvent
(e.g., cyclohexane) or in a bulk monomer at a temperature above melting point of the
polymer. This type of process is employed for the production of LDPE (e.g., Atochem,
Sumimoto), HDPE and LLDPE (e.g., Dow, Novacor, DSM, etc.), PP (e.g., Eastman
Kodak) and polybutene-1 (e.g., Shell).
Solution processes have some unique advantages over slurry processes: MWD can
be better controlled, and, the process variables are also more easily controlled because the
polymerization occurs in a homogeneous phase. Since polymer properties such as
molecular weight distribution can be controlled by temperature, a very broad spectrum of
product grades can be obtained. The high polymerization temperature (130- 15OoC) also
leads to high reaction rates and high polymer throughputs from the reactor and low
residence times (on the order of minutes). This results in remarkable volume productivities
(up to 7500 kg/h/m3) and small reactors which are specially suited to perform grade
transitions.
Disadvantages of the process include poor monomer conversion ratio (as monomer
conversion is limited by polymer solubility; typical polymer content values in the reactor
vary from 20-35% by weight), high catalyst residuals in the polymer which must be
deashed, and high utilities consumption due to solvent recovery, purification and recycle
steps.
Solution processes are more efficient for production of low molecular weight resins
due to high operating temperatures. However, very high molecular weights cannot be
produced easily at high temperatures as a result of high viscosity in the reactor, and since
the solids content is relatively low compared with the slurry process, greater diluent
recovery may be required (Choi and Ray, 1985a).
mixture of C8 and C9 isoparaffins). Ethylene, octene-1, and catalyst are introduced into the
reactors, where they remain in solution. The reactors are run at approximately 160°C and
27 bar, and total residence time in the reactors is 30 minutes. When the solution exits the
second reactor it is flashed to remove ethylene monomer. Solvent and residual octene-1 are
removed in a 2-stage flash unit, and the remaining polymer melt is compounded with
additives, extruded, and pelletized.
The Compact solution process was originally developed by DSM in the 1970’s to
produce HDPE. It was adapted in the 1980’s to produce linear medium- and low- density
polyethylenes based on butene-1 or octene-1 co-monomers using proprietary Ziegler
catalysts. The Compact solution process is capable of producing polyethylenes across a
density range of 0.900-0.967 g/cm3 and a melt-index range of 0.8-100 using propylene,
butene-1, or octene-1 co-monomers. In the Compact solution process, depicted in Figure
2.10, ethylene and co-monomer, usually octene-1, are dissolved in hexane, cooled, and fed
to a stirred, liquid-filled reactor. The reactor is operated adiabatically, so it is necessary to
precool the reactor feed to control temperature. The enthalpy of polymerization raises the
reactor to its normal operating temperature, 200°C. Molecular weight is controlled by
40 Chapter 2: Commercial Polymerization Processes
hydrogen. Ethylene conversion of at least 95% is achieved during a short reactor residence
time. The hot solution exits the reactor into vessels where volatiles plus the major portion
of the hexane are flashed off, and catalyst residues are deactivated. The polymer
concentrate is fed to a degassing extruder where the residual hexane and octene-1 are
flashed, the appropriate additives are compounded, and the polyethylene is pelletized and
transported to silos for pack out.
The first commercially available technology for the production of high density,
linear polyethylene grades was the solution process. It was soon discovered that a more
efficient way to produce various PE grades was to carry out the polymerization in a slurry-
phase (Hottovy et. al., 2004). In slurry processes, catalyst and polymer particles are
suspended in an inert solvent, typically a light or heavy hydrocarbon. Slurry processes run
in loop reactors with the solvent circulating, stirred tank reactors with a high boiling
solvent or a ‘liquid pool’ in which polymerization takes place in a boiling light solvent. A
variety of catalysts can be used in these processes. Advanced processes combine a loop
reactor with one or two gas-phase reactors, placed in series, where the second stage of the
reaction takes place in the gas-phase reactors. For bimodal polymers, lower molecular
weights are formed in the loop reactor, while high molecular weights are formed in the
gas-phase reactor. It is noted that slurry loop reactor is further discussed and analyzed in
Chapter 4: Slurry-Loop Reactor Modelling.
Chapter 2: Commercial Polymerization Processes 41
Ethylene
To ethylene
Comonomer recovery
make-up Compressor
Fractionator
Loop reactor
To diluent
recovery
Catalyst
Purge
N2 Column
Fresh
Diluent To extruder
Catalyst, ethylene, co-monomer (if desired), and hydrogen are injected into a
vertical, loop-shaped reactor filled with isobutene diluent, as shown in Figure 2.12. Reactor
pressure is approximately 30 bar, and reactor temperature is in the range of 75-80ºC. The
42 Chapter 2: Commercial Polymerization Processes
If bimodal resin is desired, the polymer particles from the first loop reactor can be
transferred to another loop reactor by directing a portion of the slurry to settling legs,
where the polymer particles settle towards the bottom, then re-directed to another loop
reactor. Hydrogen is flashed off prior to introduction in the second loop, and additional
monomer, co-monomer and hydrogen (if desired) are added, and the mixture recirculated
in the second loop, before being harvested.
The Nissan slurry process, used to produce HDPE, is based on a stirred tank reactor
as shown in Figure 2.14 . Catalyst, ethylene, co-monomer, and hydrogen are fed to the
44 Chapter 2: Commercial Polymerization Processes
reactor which uses hexane diluent. The reactor is operated in a temperature range of 65-
85°C at a pressure of 10-14 bar. Polymer slurry exiting the reactor is sent to a flash vessel
to remove volatiles then fed to a centrifuge where the diluent is removed. The dried
powder is compounded, extruded, and pelletized. The process has been refined through the
joint efforts of Nissan, Maruzen, and Equistar.
The Mitsui process comprises two continuous, stirred-tank reactors of identical size
that are operated in series to produce film resins with bimodal MWDs (Figure 2.15). The
diluent is hexane, and the usual co-monomer is butene-1. Typical conditions are an
operating pressure of 7.8 bar, an operating temperature of 85°C, and a 45-minute residence
time per reactor. Polymer is isolated and pelletized in straightforward fashion in the Mitsui
process.
Chapter 2: Commercial Polymerization Processes 45
Gas-phase processes are by far the most common process in modern ethylene
production plants. The distinguishing characteristic of gas phase polymerization is that the
system does not involve any liquid phase in the polymerization zone. Polymerization does
occur at the interface between the solid catalyst and the polymer matrix, which is swollen
with monomers during polymerization. The gas phase plays a role in the supply of
monomers, mixing of polymer particles, and removal of reaction heat.
The reactor is operated at a temperature below the melting point of the polymer
particles. For HDPE production, the operating temperature is 90-110oC. An operating
temperature of about 90oC or lower is preferred for production of LLDPE, which contains
about 15 mol% of one or more of the C3 to C6 α-olefins (Levine and Karol, 1977).
Operating pressure ranges from 350-550 psi.
Commercial gas phase processes involve either a fluidized bed reactor, vertical
stirred reactor or a horizontal stirred reactor. In the fluidized-bed process, the monomer
flows through a perforated distribution plate at the reactor bottom and rapid gas circulation
ensures fluidization and heat removal. Unreacted polymer is separated from the polymer
particles at the top of the reactor and recycled. Fluidized-bed plants are able to produce
46 Chapter 2: Commercial Polymerization Processes
either LLDPE or HDPE and are free of constraints from viscosity (solution process) or
solubility (slurry process). A modification uses a second reactor connected in series to
perform copolymerization. The stirred-bed process uses a horizontal or vertical reactor
with compartments, in which the bed of polymer particles is agitated by mixing blades.
Compared with slurry and solution processes, however, gas phase processes do have
the following disadvantages:
1) Reactor operating temperature is limited to the resin softening point. Thus, the
productivity of catalyst is also limited.
In 1968 the Union Carbide Corporation (UCC) installed the first commercial gas
phase reactor for the production of HDPE. Later, in 1977, the process was also extended
for the production of LLDPE. Currently, it represents the largest single type of technology
for LLDPE production. In 1983, a joint venture between UCC and Shell Oil combined the
fluidized bed technology with Shell high activity supported Ti catalyst (SHACK catalyst)
to extend the process to PP production (Unipol process). Figure 2.16 depicts a schematic of
the Unipol process. Supported catalyst is fed into a reactor where it is fluidized in a stream
of ethylene, co-monomer, and hydrogen. A gas distributor plate, located near the bottom of
the reactor, supplies a uniform distribution of monomer for fluidization. Polyethylene
forms on the fluidized catalyst particles. Unreacted gases are continually withdrawn from
the reactor, compressed, cooled, and recycled. Average residence times of the particles in
the bed range from 3-5 hours during which time the particles grow to average sizes of
about 500μm in diameter. Temperatures in the range 50-70oC and pressures in the range
20-40 bar are typical. Polymer particles are removed from the reactor through a differential
valve, conveyed to an extruder, compounded with the appropriate additive formulations,
and pelletized.
Unipol II, which comprises not one but two reactors, appeared on the scene in 1992
with the intent to provide Union Carbide (and potential licensees) with the ability to
produce polyethylenes having a bimodal MWD. There are only 2 Unipol II lines in the
world, a Union Carbide plant in Taft, Louisiana, and one belonging to EQUATE
Petrochemical Company K.S.C., a Union Carbide joint venture with Petrochemical
Industries Company and Boubyan Petrochemical Company in Shuaiba, Kuwait. The
production rate of a Unipol reactor is determined by the amount of heat that can be
removed. To increase the heat-removal capacity of the fluidized bed Union Carbide
developed what it called condensed-mode operation. The recycle stream is partially
condensed, and some liquid is re-injected into the system. The enthalpy of vaporization
absorbs extra heat, enabling higher production rates. In the mid 1990’s Exxon developed
and patented what it refers to as Super Condensed Mode Technology that enables even
higher production rates. SCMT is one of the technical cornerstones of Univation
Technologies.
48 Chapter 2: Commercial Polymerization Processes
Cycle
Reactor compressor
Product
chamber
Ethylene
Comonomer
Degassing
Hydrogen Recycle
stream
To extrusion
and pack out
In 1967, BASF was the first company to develop a gas phase process for
polypropylene (Novolen process). During the eighties, the process was extended to
propylene-ethylene random and high-impact co-polymers. Figure 2.17 shows the BASF
continuous process. All the monomers, solvents and nitrogen fed into the process are first
passed through purification columns to remove compounds which act as poisons to the
catalyst. Co-catalyst blend and catalyst slurry are fed from special tanks to the reactor.
Polymerization temperatures are 50-150oC at corresponding pressures of 10-40 bar,
ensuring that the monomer phase is gaseous in the reactor. Low concentrations of
hydrogen are used to control molecular mass over wide ranges. In general, reactor
pressures higher than other gas phase processes are used in order to increase productivity.
The temperature is controlled by removing gaseous propene from the reactor head space,
condensing it with cooling water, and then recirculating it back into the reactor, where its
evaporation provides the required cooling, as well as further aeration of the stirred powder
bed. Each tonne of polymer made requires 6 tonnes of liquid propene to be evaporated as
coolant.
Chapter 2: Commercial Polymerization Processes 49
Typical reactor sizes are 25, 50, or 75 m3, equipped with proprietary helical
agitators, which give excellent agitation. Homopolymerization needs only the primary
reactor, into which the catalyst components are fed. These must be very well dispersed in
the powder bed to avoid build-up.
Powder and associated gas discharge continuously from the primary reactor dip tube
directly into a low-pressure cyclone. Propene carrier gas from this cyclone is recycled to
the reactor after compression, liquifaction, and, sometimes, distillation. The powder then
passes to a purge vessel where a deactivator quenches all residual catalyst activity, and
nitrogen strips out traces of propene from the hot powder. From here powder is conveyed
into silos for stabilization and extrusion into granules. BASF also offers a post-granulation
steam-stripping package to remove any oligomers and oxidized residues from the granules
for demanding applications.
BASF pioneered their gas-phase process with commercial production in 1969. The
products made, Novolen 1300 series, were based on high molecular mass total work up
polymers (i.e., containing atactic PP and catalyst residues) having reduced stereoregularity.
Today, such grades still find niche markets, although they are vulnerable to competition
from random co-polymers. Production is to be phased out shortly. BASF also uses their
process with a cheaper second-generation catalyst TiCl3/AlEt2Cl, which then requires an
additional dry-powder dechlorination stage.
50 Chapter 2: Commercial Polymerization Processes
To recycle
Cyclone
Condenser
Product
Monomer
Hydrogen
T
Catalyst C
Polymer
Cyclone
Condenser
Product
Monomer
Hydrogen
T
Catalyst C
2nd Reactor
This process was originally developed by Amoco in 1979 for the production of
isotactic polypropylene and then extended to polypropylene co-polymers by Chisso in
1987. The reactor consists essentially of a horizontal cylindrical pressure vessel, stirred by
paddles or an axial shaft (Figure 2.18). Its lower part is divided by weirs into four
compartments. Stirring is used as way of homogenization to avoid hot spots. Monomer, a
recycle gas stream and hydrogen for molecular weight control are fed into the bottom of
each compartment at a low enough velocity to avoid fluidizing the bed. Catalyst
components and a quench stream are sprayed into the bed from inlets in the top of the
reactor. The quench stream, added to provide evaporative cooling, consists of a light
Chapter 2: Commercial Polymerization Processes 51
Polymer flows from one compartment to another propelled by the axial stirrer till it
reaches the exit section. The powder residence time in a single vessel is equivalent to 3-5
well mixed vessels. Product is discharged from the reactor through a series of sequenced
locks. Vapors from the reactor are removed and separated into a recycle stream. The
catalytic system is based on Ti and Mg modified by a combination of esters and
chlorosulphonic acids. Catalyst poisons such as moisture, oxygen, oxygenated compounds
or sulphur must not be present at levels above one part per million.
For homopolymer production, propylene is the only monomer fed to the system. For
random co-polymer production, ethylene is additionally added. In the case of high impact
polypropylene, a second reactor is required. A gas lock system transfers powder from the
homopolymer reactor to the co-polymer reactor. After the polymerization reactors, the
powder is released periodically to a gas/ powder separation system that includes a purge
column. A countercurrent stream composed of nitrogen and steam deactivates the catalyst.
Product is sent to addition of chemical additives, extrusion and pelletization.
The main advantages of this process are its narrow solids residence time distribution,
resulting in small grade transition times, and the ability of separately controlling reaction
conditions (e.g., temperature, feed compositions etc.) in each of the compartments, leading
to products with a broad range of properties. However, achieving controlled bed uniformity
represents one of the most important operational challenges in this kind of technology.
52 Chapter 2: Commercial Polymerization Processes
Ethylene
Comonomer
Hydrogen
Cyclone
Reactor Gas Recycle
Degasser
Catalyst
Degassing
Compressor column
To extrusion
and pack out
Gas phase
reactor
Product
chamber
Flash
Diluent Product blow
tank
Recycle Recycle
stream stream
Degassing
Ethylene
Comonomer
Hydrogen
To extrusion
and pack out
2.4. References
Asua J.M., 2007, ‘Polymer Reaction Engineering’, Blackwell Publishing Ltd, UK.
Beach D.L. and Kissin Y.V., 1986, Encylcopedia of Polymer Science and Engineering;
Kroschwitz J.I. Ed., John Wiley & Sons, New York, Vol. 6, p 454.
Burdett I.D., 1992, ‘A Continuing Success: The Unipol Process’, Chemtech, Oct, pp. 616-
623.
Chandrasekhar V., Srinivasan P.R. and Sivaram S., 1988, ‘Recent Developments in
Ziegler-Natta Catalysts for Olefin Polymerization and Their Processes’, Indian
Journal of Technology, 26, pp. 53-82.
Choi K.Y. and Ray W.H., 1985a, ‘Recent Developments in Transition Metal Catalyzed
Olefin Polymerization-A Survey. I. Ethylene Polymerization’, Macromol. Chem.
Phys., C25(1), pp. 1-55.
Choi K.Y. and Ray W.H., 1985b, ‘Recent Developments in Transition Metal Catalyzed
Olefin Polymerization-A Survey. II. Popylene Polymerization’, Macromol. Chem.
Phys., C25( l), pp. 57-97.
Foster G., 1991, Polymer Reaction Engineering Course, McMaster University, Hamilton,
ON.
Levine I.J. and Karol F.J., 1977, ‘Preparation of Low and Medium Density Ethylene
Polymer in Fluid Bed Reactor’, US. Patent No. 4,011,382.
Rhee S.J.; Baker E.C., Edwards D.N., Lee K.H., Moorhouse J.H., Scarola L.S. and Karol
F.J., 1991, ‘Process for Producing Sticky Polymers’. US. Patent No. 4,994,534.
Ullmann’s Encyclopedia of Industrial Chemistry, 6th edition, 2002, John Wiley & Sons.
56 Chapter 2: Commercial Polymerization Processes
Xie T., McAuley K.B., Hsu J.C.C. and Bacon D.W., 1994, ‘Gas Phase Ethylene
Polymerization: Production Processes, Polymer Properties, and Reactor Modeling’,
Ind. Eng. Chem. Res., 33, pp. 449-479.
LyondellBasel,_‘Licenced_Technologies’,_URL:_http://www.lyondellbasell.com/Technol
ogy/LicensedTechnologies/, Date retrieved: May 10, 2010.
Chapter 3
3.1. Introduction
The purpose of this chapter is to describe the catalytic systems employed for
polymerization. Specifically, the catalytic systems presented are Ziegler-Natta,
metallocenes and chromium-based catalysts. The method of moments, which is a useful
mathematical treatment for developing rate expressions for reactor modelling and
extracting molecular developments, is described. For the definition of a simplified kinetic
scheme capable of describing olefin polymerization in slurry—phase reactors, a literature
review is presented. Moreover, a sensitivity analysis procedure is followed in order to
define the dominant reactions in terms of productivity and molecular weight. Finally, based
on the extracted simplified kinetic scheme, a deconvolution procedure, based on industrial
data, is followed for the calculation of both number of different catalyst active sites and
kinetic rate constants.
Catalysts are the key to making high performance polymeric materials. Except for
the high temperature and pressure radical initiated ethylene polymerization, all other
processes require a catalyst system for monomer activation. Typical catalytic systems
58 Chapter 3: Catalytic Polymerization Kinetics
The high pressure and high temperature processes, developed in the 1930’s, produce
branched, low density polyethylene, LDPE, with unique optical and processing yet less
impressive mechanical properties for more demanding applications. Additionally, harsh
polymerization conditions (few thousands of atmospheres of pressure and temperatures
ranging from 200–300oC) combined with the high hydrocarbon emission nature of the
older reactors still in use are considered as environmental disadvantages.
Alternative processes have been developed for ethylene polymerization and are in
use since the 1950’s. Two classes of co-ordination catalysts, the Phillips catalysts based on
silica supported CrO3 (Chromium catalysts), and Ziegler–Natta catalysts based on MgCl2
supported titanium chloride, have been employed since to produce polyethylene homo- and
co-polymer more efficiently and safely. The medium and high density polyethylene,
M:HDPE, produced with CrO3 based Phillips catalysts have both mechanical and
processing advantages but the optical properties of corresponding articles are poor and
because of relatively low catalyst activity the products contain catalyst residuals. The
linear low density polyethylene (i.e., LLDPE) produced with MgCl2 supported TiCl3 based
catalyst in a fluidized bed gas phase reactor was developed by Union Carbide researchers
in the late 1970’s. It was initially designed to replace the LDPE, it has improved
mechanical properties and its production is less environmentally stressful yet its processing
is more difficult and requires severe extrusion pressure and temperatures. Thus, the so
called LLDPE has become available as an alternative but could not replace the LDPE
resins. With respect to polypropylene until recently, only processes that employed Z-N
type catalysts existed. The catalyst system developed by Natta, initially for producing
crystalline polypropylene, had poor activity and the resulting isotactic propylene product
required several steps of washing and de-ashing with huge quantities of solvent and waste
involved. Since then, at least four new generations of catalysts have been developed each
more active and stereo-selective, eliminating more and more washing and de-ashing steps
Chapter 3: Catalytic Polymerization Kinetics 59
and creating less and less wastes and atactic fractions. However, the production of isotactic
polypropylene with the highly selective Z-N catalysts requires the addition of organic
compounds the so called internal and external electron donors, as modifiers. Additionally,
the MgCl2 supported Z-N catalysts contain substantial amounts of chlorine. The search for
versatile multipurpose clean catalysts finally led to the discovery of the metallocene
catalysts, which after years of gestation have been emerging slowly but pervasively in the
last few years. These catalysts have reached now their full industrial maturity and offer
themselves as a new class of commercially viable and environmentally clean alternative for
the production of a wide range of varieties of polyolefin homo- and co-polymers (Razavi,
2000).
When Z-N catalysis was first discovered, it was fascinating to see ethylene, a
monomer normally difficult to activate, being polymerized under atmospheric pressure at
room temperature in a Weck-glass vessel containing gasoline as the solvent, with small
amounts of TiCl4 and Al2Et3Cl3, the latter components forming the colloidal suspended
catalyst. However, in practice, it was difficult to fabricate the resulting polymer for
practical use. It was necessary to deactivate the catalyst after polymerization, remove the
diluent and the residues of catalyst with HCl and alcohols. Washing of the polymer with
water and drying with steam was then necessary. Other complications involved purification
of the recovered diluent, feedback of the monomer after a repurification step, and, finally,
the necessity of adjusting the molecular weight by oxidative or thermal processes. The cost
of these steps diminished the advantage of low pressure processes (Sinn and Kaminsky,
1980).
transition metal compounds containing titanium (Ti), vanadium (V), chromium (Cr) and, in
special cases, molybdenum (MO), cobalt (Co), rhodium (Rh) and nickel (Ni) are primarily
used (Huang and Rempel, 1995). The most common catalyst system is titanium
tetrachloride, TiCl4, with triethyl aluminum, Al(C2H5)3, and it is used for polyethylene
production. However, various catalyst-co-catalyst combinations yield behaviour that can
be markedly different. They can vary in activity, stability, ability to polymerize specific
monomers and incorporate co-monomers, ability to control tacticity, and their effects on
the molecular weight distribution of the polymer, among other properties.
Electron donor compounds, such as amines, ethers and esters, have the potential of
complexing and reacting with the components of the catalyst or the active centers. They
have been used in controlled amounts in many Ziegler-Natta catalytic systems as a third
component to increase catalyst activity and/ or stereoselectivity.
Catalysts of this type operate at low pressures (up to 30 atm) and at temperatures up
to 120oC. The polyethylene produced is characterized by the absence of large amounts of
long- or short-chain branching, which caused variability in density, crystallinity and
melting point in the earlier discovered high-pressure, low-density polyethylene product
formed by radical-included polymerization (Table 3.1).
There are two existing theories that explain the ability for Ziegler-Natta catalysts to
produce polydisperse polymers. The first is diffusive and the second is kinetic. The former
Chapter 3: Catalytic Polymerization Kinetics 61
The kinetic theory involves the assumption that there exist different catalyst site
types, each with its own relative reactivity, due to variations in the local chemical
composition of each site type. An example is the presence of different ligands on the active
transition metal. Experimentation has provided compelling evidence that this second
theory is a more accurate description of the physical reality (Xie et al., 1994, de Carvalho
et al., 1989, Peacock, 2000). These site types tend to produce chains that differ in both
length and co-monomer composition. More specifically, the sites that produce longer
chains tend to incorporate less co-monomer. It should be noted that each different catalyst
site type can produce polymer having polydispersity number equal to 2. However, polymer
exhibiting broad molecular weight distribution can also be produced if more than one
catalyst site types are taking part in the polymerization. In this work, this theory is
followed for the extraction of the kinetic equations, rather than the mass-transfer theory.
Ziegler-Natta catalysts are used in soluble, colloidal, and heterogeneous forms. Most
Z-N catalysts systems are heterogeneous, but some homogeneous systems are also known.
It is not clear that stereoregulation and stereoselectivity are the results of heterogeneity.
Each different catalyst form has its own advantages, depending on the application.
Supported versions are used in gas-phase and slurry processes, while soluble catalysts are
used in solution processes. Although the supported types tend to have higher activity than
the non-supported ones, likely due to an increase in the number of catalyst sites that are
active, the catalytic mechanism is the same (Peacock, 2000).
Several articles have been published that give detailed information on Z-N catalysis.
These include books by Boor, 1979, Kissin, 1985 and van der Ven, 1990 as well as Choi
and Ray, 1985a,b.
62 Chapter 3: Catalytic Polymerization Kinetics
Polymethylene
Crystallinity, % at 20oC 65 75 85 95
Ziegler-Natta catalysis provided for the first time stereochemical control of the
polymerization process. By carefully selecting the combination of catalyst and co-catalyst,
one is able to produce polymers with the desired steric structure. The polyethylene
produced in Z-N polymerization is linear, which is characterized by the absence of long or
short chain branching. For a-olefin polymerization, polyolefins of isotactic or syndiotactic
structure can be obtained by using special Ziegler-Natta catalysts. There are even more
choices in steric structures for polydienes; polydienes of the 1,4-cis-, the 1,4-trans-, and the
1,2-structures, as well as the 3,4-structure in the case of substituted dienes, can be
produced with proper Z-N catalysts. Thus, under conditions easy to realize, isotactic and
syndiotactic structures are preferred to statistical atactic structure (see also Figure 1.5).
Polymer Polymer
H R
X X
C
X M CH2 CHR X M
C
X X
H H
X X
Polymer
CHR
X X
X X
X X
120
1994
100
1984
60
40
20 1975
1964 1968
1958
0
st nd rd rd th th
1 gen 2 gen 3 gen 3 gen 4 gen 4 gen
super super
Conventional catalysts were the first systems studied by K. Ziegler and G. Natta,
which were able to polymerize olefins in a stereospecific way. These catalysts consist
basically of solid TiCl3 co-crystallized with aluminium halides as a result of the reduction
of TiCl4 by organo-aluminium compounds, under controlled temperature and mixing
conditions. Commonly mentioned in the literature is the formula TiCln.xAlCl3, where
2.5<n<3.0 and 0<x<0.5. Only a small percentage of Ti atoms (1-4%) seem to be
responsible for the formation of active sites. Their activity and selectivity appears to be
related to their position in the structure of TiCl3 crystals. Those crystals aggregate,
generating porous aggregates. During polymerization, these aggregates tend to break into
smaller particles exposing new sites to the reaction medium. According to the ‘replica’
phenomenon, the final polymer particle morphology closely replicates that of the original
catalyst particle.
Chemical treatment (contact with electron donors, e.g., diisoamilic ether, di-n-
butyl ether, etc.)
Deposition on inert supports (SiO2, Al2O3. CaO, etc.) (Soga and Ikeda, 1980).
This class of catalysts appeared during the seventies as a result of research in the
area of commercial production of polyethylene and polypropylene. Its main characteristics
are the use of MgCl2 based supports as well as the introduction of Lewis bases (e.g., esters,
silanes, amines, etc.) as effective catalytic components in the preparation of the support
(‘internal electron-donor’) and during the polymerization (‘external donor’) to obtain high
isotacticity values (i.e., greater than 96%).
As in the case of TiCl3, MgCl2 particles tend to agglomerate, decreasing the support
active surface. However, electron-donor compounds adsorb the cleavage surfaces,
stabilizing the crystallites and partially regulating the stereospecific control, during the
preparation of the support. In the production of polyethylene and polypropylene, catalytic
activity values can reach 1000 tonnes/kg Ti and 1000 tonnes/kg Ti, respectively.
During the last two decades the development of industrial catalysts based on
metallocenes was a significant innovation process in polyolefin chemistry (Zechlin et al.,
2000). Metallocene compounds such as bis-cyclopentadienyl-titanium (Figure 3.4) or
zirconium complexes have been known to be mildly active for the polymerization of
olefins. In the mid 70’s Kaminski and co-workers (Anderson et al., 1976) discovered that a
significant increase in ethylene polymerization activity could be obtained by contacting
bis-cyclopentadienyl-titanium with methyaluminoxane (MAO). Methylaluminoxane is
synthesized by the addition of water in stoichiometric amounts to trimethylaluminumto
produce a cyclic oligometric structure. Apparently, the role of MAO is to form an ion pair
with an active cationic form of the metallocene. Several articles have been published that
give detailed information on metallocene catalysis (Huang and Rempel, 1995, Brintzinger
et al., 1995, Soares and Hamielec, 1995).
mechanism. This avoids the need for peroxide additives, which produce objectionable odor
and volatile by-products. A polydispersity of 2 is predicted by Schulz-Flory distribution for
polymers arising from identical catalyst centers with fixed rates of chain propagation and
chain termination. A polydispersity of 2 is thus regarded as evidence that only a single
catalyst species contributes to polymer growth in a homogeneous catalyst system
(Brintzinger at al., 1995).
3.2.3.1. Stereoregulation
The breakthrough for propylene polymerization happened with the discovery that
bridged or ansa-zirconocene catalysts could produce stereospecific configurations of
polypropylene (Huang and Rempel, 1995). The addition of bridging substituent made by
zirconocene stereorigid, resulting in different kinds of symmetry. This technique allowed
the production of polypropylene with specific chain configuration structures such as
atactic, isotactic, hemi-isotactic and syndiotactic (see also Figure 1.5).
Atactic Polymers
The metallocenes, such as Cp2ZrC12 and meso-Et(Ind)2TiCl2, have the same Cp (or
Cp’) ligands and an achiral structure. These catalysts are non-stereospecific and furnish
atactic polypropylene and other poly(α-olefins).
Isotactic Polymers
Syndioitactic Polymers
Hemi-isotactic Polypropylene
Kaminsky’s Model
Figure 3.5 shows an insertion mechanism model proposed by Kaminsky and Steiger
for ethylene polymerization with Cp2ZrC12/MA0 catalysts. Through a Zr-O-Al bond,
electron density is withdrawn from the Zr atom. If there is no ethylene present, the electron
70 Chapter 3: Catalytic Polymerization Kinetics
deficient Zr atom interacts with P-H through an agostic hydrogen bond, which can lead to a
chain transfer by β-H elimination. If ethylene is present, it can become bonded to Zr
forming a π-complex, which may be followed by ethylene insertion. The model was
explained by the existence of electron deficient compounds in penta-coordinated bimetal
complexes.
Corradini’s Model
2) the mechanism consists of two stages: the coordination of the olefin to the active
site, followed by insertion into the metal-carbon bond through a cis opening.
In Corradini’s model the active species is the Cp2MR+ complex. The cationic
character of the active species has been substantiated by the recent discovery of base-free
cationic metallocene catalysts. According to a quantum mechanical analyses for d0
complexes of the type Cp2ML (L=ligand), the best coordination of L may not be the one
(most symmetrical and presumably sterically most favorable) along the symmetry axis
which relates the two bent Cp rings. Energy minimum situations for a do complex and for
Chapter 3: Catalytic Polymerization Kinetics 71
L=H would correspond to angular deviations Δα≈65o from the symmetry axis. On the other
hand, an additional stabilization of a geometry α ≠ 0ο at the strongly unsaturated metal
center could occur because of an agostic interaction of a hydrogen in the γ-position.
Corradini predicted that α will differ from 0o whenever a monomer or solvent molecule
becomes coordinated to the equatorial belt of the bent metallocene cation, together with the
polymer chain. (Note that a geometry with α ≠ 0ο makes the metal atom chiral in cation
Cp’Cp”MR+ if the two cyclopentadienyl ligands Cp’ and Cp” are different).
As seen in Figure 3.6, at the end of each polymerization step the chain occupies the
coordination site previously occupied by the alkene monomer. Hence, most consecutive
polymerization steps correspond to a model obtained by exchanging the relative positions
of the growing chain and of the monomer. This is extremely relevant for the stereospecific
behaviour of the catalysts.
In 1953, Hogan and Banks, discovered that chromium oxide supported on silicate
and other carriers, would polymerize ethylene to high polymers (Hogan and Banks, 1958,
McDaniel, 2010). Since then, this catalyst has evolved to become the basis of the Phillips-
slurry loop process. There are three main classes of supported chromium catalysts. The
first one uses organochromium compounds, the second one is based on chromium oxide,
and the third one is a hybrid catalyst.
The first type of catalyst is prepared by calcining a high surface area carrier with
further deposition of an organochromium compound onto it anhydrously. Apparently, the
role of the calcining treatment is to partially dehydroxylate the carrier surface, leaving
mostly isolated single hydroxyls, which can then react with the organochromium
compound through a single oxide bond (Figure 3.7). One or more ligands may remain
attached to the chromium center.
In the second class of catalysts, chromium oxide binds to a fully hydrated surface.
Then, at a temperature around 200oC, a chromate species is formed through an
esterification reaction between CrO3 and two surface hydroxyls. The catalyst does not
become fully activated until more calcinations further dehydrates the surface, which
suggest that neighboring hydroxyls may interfere with polymerization if not removed. The
activity of the catalyst improves with calcining temperature up to the sintering temperature
of the support.
A variety of organo-Cr compounds are suitable for the production of active catalysts.
Examples include π-bonded compounds (e.g., bis(arene)-chromium(0)) and the
chromocene derivatives (e.g., allyl-chromium(II) and –chromium(III), alkyls of Cr(II) and
Cr(IV)). Adequate support for all three types of catalysts include silica, alumina,
aluminophosphates and silica-titanica.
Chapter 3: Catalytic Polymerization Kinetics 73
Figure 3.8: Polymerization steps for ethylene polymerization over a chromium oxide
based catalyst.
Figure 3.9: Possible co-catalyst reactions for a chromium oxide based catalyst.
Chapter 3: Catalytic Polymerization Kinetics 75
2) chain initiation,
3) chain propagation,
4) chain transfer,
6) site deactivation.
the activated vacant catalyst sites of type ‘ k ’ and ‘dead’ co-polymer chains of length ‘ n ’
produced at the ‘ k ’ catalyst active site, respectively. The symbol Pnk,i denotes the
76 Chapter 3: Catalytic Polymerization Kinetics
It should be noted that not all reactions may be applicable in every process. For
example, chain transfer to hydrogen occurs only when hydrogen is added as a chain-
transfer agent. All the different types of reactions are discussed below and are presented in
Table 3.2.
Activation is the reaction through which a potential site is converted into a reactive
vacant site. Several mechanisms are proposed in the literature (de Carvalho et al., 1989,
Kissin Y.V., 1985, Nirisen and Rytter, 1986). The present work considers the following
four ways for the catalyst to become activated (Xie et al., 1994):
by hydrogen,
by co-catalyst,
by monomer, and,
spontaneous.
If the modelled process uses a co-catalyst, and no specific information about the
susceptibility of the catalyst to activate upon reaction with either monomer or hydrogen
exists, a first approach should utilize only the activation by co-catalyst, since this is the
primary activation path. It is always important to minimize the model complexity unless
sufficient information exists to do otherwise.
k
During chain initiation, a new polymer chain is started. A vacant site of type k , P0
P1,ki . The reaction is assumed to be first order with respect to each component.
Chapter 3: Catalytic Polymerization Kinetics 77
Chain propagation is the mechanism step in which the polymer chain grows. A new
monomer molecule, M j , is inserted to the live polymer chain, Pnk,i , that is attached to the
Chain transfer is a type of reaction that terminates a live chain, producing dead
polymer and a vacant site. Generally, a chain transfer agent remains attached to the active
site, cutting off the polymer chain. Many chain transfer agents and mechanisms have been
proposed. A generalized kinetic scheme considers chain transfer:
to hydrogen,
to co-catalyst,
to solvent,
to monomer i, and,
spontaneous chain transfer.
It should be noted that most of the processes use a transfer agent for molecular-
weight control. Hydrogen is the most common species chosen (Vandenberg, 1962),
reacting with the catalyst site to disengage the polymer chain.
to hydrogen,
to co-catalyst,
to solvent,
to poison, and,
78 Chapter 3: Catalytic Polymerization Kinetics
spontaneous.
This produces an empty live site, unlike the previous chain-transfer reactions that
produce an initiated site, and a dead polymer chain.
Site deactivation is the reaction step generally accepted as the explanation for the
loss in activity experienced in some catalytic systems during the polymerization (Floyd,
1986, Nirisen and Rytter, 1986, Ray, 1972). Both occupied and vacant sites are assumed to
deactivate according to deactivation:
by hydrogen,
by co-catalyst,
by by-product,
by poison,
by monomer j, and,
spontaneous deactivation.
Chapter 3: Catalytic Polymerization Kinetics 79
K kl
Site activation to solvent: Pnk,i S
tS
P0l Dnk
k
By hydrogen: S p H 2 K
P0k K kl
P0k S P0l
aH
tS
k
by co-catalyst: S p A K
P0k B to poison: K kl
Pnk,i X P0l Dnk
aA
tX
Kk K kl
by monomer i: S p M i
aM 1
P0k M i P0k X
tX
P0l
k
S p K
P0k K kl
spontaneous: spontaneous: Pnk,i P0l Dnk
aSP
tSP
Chain initiation K kl
P0k
tSP
P0l
Kk Kk
to monomer i: Pnk,i M i
trMi
P1k,i Dnk P0k M j
dMj
C Dk
k k
spontaneous: KtrSpi
Pnk,i P0k Dnk
spontaneous: Pnk,i K
dSP
CDk Dnk
k
Site transformation P0k K
dSP
CDk
to hydrogen: K kl
Pnk,i H 2
tH
P0l Dnk
K kl
P0k H 2
tH
P0l
K kl
to co-catalyst: Pnk,i A
tA
P0l Dnk
K kl
P0k A
tA
P0l
80 Chapter 3: Catalytic Polymerization Kinetics
The moments are averages of the concentrations of polymer molecules that are
weighted by their chain lengths. The method of moments, which is a useful mathematical
treatment for developing rate expressions for reactor modelling is employed for the
calculation of molecular properties. This treatment is a statistical technique that enables us
to track various polymer chain properties without the need to include the very large
number of equations and unknowns required to account for chains of every possible length
(Hatzantonis et al., 2000, Khare, 2003, Dompazis et al., 2008).
The following ‘live’, k , ‘dead’, k , and ‘bulk’, k , moments of the corresponding
Live moments:
Nm Nm
λνk k
λν,i nν Pn,i
k
; 0, 1, 2, ... (3.1)
i 1
i 1 n 1
Dead moments:
k n Dnk ; 0, 1, 2, ... (3.2)
n 2
Bulk moments:
νk kν k ; 0, 1, 2, ... (3.3)
λν,ik and k denote the ‘ ’ moment of ‘live’ co-polymer chains ending in an ‘ i ’
( 1, 2, ..., M m ) monomer unit and the ‘ ’ moment of ‘dead’ co-polymer chains, produced
at the ‘ k ’ catalyst site, respectively. νk is the respective ‘ ’ ‘bulk’ moment. The three
leading moments, namely, the zeroth, first, and second, are sufficient for describing the
molecular weight distribution of most commercial polymers.
According to the kinetic mechanism, one can derive the net production/ consumption
rates of the various molecular species for a multi-site Z-N polymerization.
Chapter 3: Catalytic Polymerization Kinetics 81
Polymer Moments
A live polymer chain balance, applied over the kinetic scheme, gives the following
equation:
d Pnk,i N
RP k k P k s kl P k k P k
d n,i st n,i ti n,i
n ,i dt l 1 k
Nm
k k k k k
(n) ktMij 0, j ( n ) k0i 0 0 M i
P P (3.4)
; for n=1, ∞
j 1
Nm Nm
1 (n) k kpij Pnk1, j M i k kpij Pnk,i M j
j 1 j 1
1 for n 1
where (n) 0 for n 1
A dead polymer chain balance, applied over the kinetic scheme, gives:
Ns Nm Nm
RD k stk tik dk Pnk,i (3.5)
n l 1 k i 1
i 1
Polymer moments rates can de derived, multiplying all live and dead polymer chains
rates to nν and summing up for n 2, . The final moments’ rates can be calculated:
Live moments:
Ns Nm Nm v v
R k vk,i dk stkl tik k kpij M i k kpij M i qk,i
v ,i l 1 k j 1 j 1 q 0 q
(3.6)
Nm
k k k k
ktMij 0, i k0i 0 M i
P
j 1
v v!
where q q !(v q)!
82 Chapter 3: Catalytic Polymerization Kinetics
Bulk moments:
Nm Nm Nm Nm v v
k l k k k
R k ktMij 0, i M i k pij M i q,i v,i
v
i 1 j 1 i 1 j 1 q 0 q
(3.7)
Nm
k0ki P0k M i
i 1
Monomer i:
s N mN sN m s N N
RM i k0ki M i P0k k Pji k k
M i 0, j ktrMi k k
M i 0, j
k 1
j 1 k 1 j 1 k 1
(3.8)
Nm Ns Ns
k
kdMji k k k
M i 0, j kdMi M i P0
j 1 k 1 k 1
Solvent:
Ns Nm Ns Ns Nm Ns Ns
k
RS ktrS S 0,k
i
kl
( ktS S 0,k
i ) ktS kl
S P0k (3.9)
k 1 i 1
k 1 l 1 i 1 k 1 l 1
Hydrogen:
sN Ns Mm Ns Ns Mm
k
RH 2 kaH H 2 S kp k
ktrH k
H 2 0, kl k
i ktH H 2 0,i
k 1 k 1 i 1 k 1 l 1 i1
(3.10)
Ns Ns Ns Mm Ns
kl
ktH H 2 P0k ( kdH
k k
H 2 0, k
i ) kdH H 2 P0
k
k 1 l 1 k 1 i 1 k 1
Co-catalyst by-product:
Ns Ns Mm Ns
RB k
kaA A S kp kdB
k k
B 0, k k
i kdB B P0 (3.11)
k 1
k 1 i 1 k 1
Chapter 3: Catalytic Polymerization Kinetics 83
Poison:
Ns Ns Mm Ns Ns Ns Mm
kl k kl k k k
RX ktX X 0, i ktX X P0 kdX X 0,i
k 1 l 1 i 1 k 1 l 1
k 1 i 1
(3.12)
Ns
k
kdX X P0k
k 1
Co-catalyst:
Ns Ns Mm Ns Ns Mm
k
R A kaA A S kp ktrA
k
A 0,k
i ktA kl k
A 0, i
k 1
k 1 i 1
k 1 l 1
i1
(3.13)
Ns Ns Ns Mm Ns
kl
ktA A P0k (kdA
k k
A 0, k k
i ) kdA A P0
k 1 l 1
k 1 i 1 k 1
RSp k ak Sp k (3.14)
Nm
RP k ak S kp k01
k
M1 P0k k02
k
M 2 P0k ktrH
k
[ H 2 ] 0,k
i
0 i 1
Nm Nm Nm
k k k k k k
ktrA A 0, i ktrS S 0,i ktrSPi 0, i
i 1 i 1 i 1 (3.15)
Ns Nm Ns Ns
tlk 0,l i tlk P0l tkl P0k dk P0k
l 1 i 1 l 1 k 1
k
kdM k k k
1 M1 P0 kdM 2 M 2 P0
84 Chapter 3: Catalytic Polymerization Kinetics
where:
Nm
ak kaH
k k
H 2 kaA k
A kaMi k
M1 kaSP (3.16)
i 1
tkl ktH
kl kl
H 2 ktA kl
A ktS kl
S ktX kl
X ktS P (3.17)
Nm
k k k k k k
tri ktrH H 2 ktrA A ktrS S ktrMi M i ktrSPi (3.18)
i 1
Nm
dk k dH
k k
H 2 k dA k
A ktB k
B k dX k
X k dMi k
M i kdSP (3.19)
i 1
The average molecular properties of interest (e.g., number- and weight- average
molecular weights) and the polydispersity index can be calculated in terms of the overall
‘bulk’ moments, , of ‘live’ and ‘dead’ polymer chains produced over the N s catalyst
active sites.
Chapter 3: Catalytic Polymerization Kinetics 85
Co-polymer Composition
The molar polymer composition can be calculated as a ration of the polymer first
moments.
Degree of Polymerization
Ns Ns
M n 1k 0
k
MW 1 0 MW DPn MW (3.25)
k 1 k 1
86 Chapter 3: Catalytic Polymerization Kinetics
Ns Ns
M w 2k k
1 MW 2 1 MW DPw MW (3.26)
k 1 k 1
where MW is the average molecular weight of the repeating structural unit in the co-
polymer chains, that is:
Nm
MW i MWi (3.27)
i 1
Polydispersity
Mw
PD (3.28)
Mn
It should be noted that the calculation of molecular properties described by Eq. 3.20-
Eq. 3.28 refer to instantaneous values. At any time, the cumulative expression of the
calculated property can also be determined by employing a mass balance in terms of the
fresh polymer produced, described by the instantaneous property value, and the total
polymer mass in the reactor described by the cumulative property value.
polymer, one means something quite different from that which applies to small-sized
compounds. Polymers differ from the small-sized compounds in that they are polydisperse
or heterogeneous in molecular weight. The control of molecular weight and molecular
distribution is often used to improve certain desired mechanical and physical properties of
a polymer product (Kiparissides, 1996).
The reconstruction of the weight chain length distribution (i.e., WCLD) of the
polyolefin, produced at the ‘ k ’ catalyst site, can be achieved by using the two-parameter
Schultz-Flory distribution:
W k
x
y k xy k
zk
e xy
k
; k 1,2,... N s (3.29)
e ln z 1
k
where W k x denotes the mass fraction of polymer chains with a degree of polymerization
x , produced at the ‘ k ’ catalyst site. Note that the Schultz-Flory parameters can be
expressed in terms of the ‘bulk’ moments ( 0k , 1k , k2 ):
1k 0k zk 1
zk ; yk (3.30)
2k 1k 1k 0k ξ 2k ξ1k
Thus, for a multi-site catalyst, the overall MWD will be given by the weighted sum
of all polymer fractions produced over the different catalyst active sites (Figure 3.10).
Ns
Wt x W k x 1k 1 (3.31)
k 1
Ns
where 1 ( 1k ) is the total polymer mass produced over the N s catalyst active sites.
k 1
88 Chapter 3: Catalytic Polymerization Kinetics
Overall
Catalyst site type 1
Catalyst site type 2
Catalyst site type 3
Catalyst site type 4
dW/dlog(MW)
log(MW)
Figure 3.10: Overall MWD given by the weighted sum of polymer fractions produced
over the different catalyst active sites.
It should be mentioned that the number of kinetic model equations solved at any
time is a function of:
Initiation 1x107
Propagation 1x107
Deactivation 1 s-1
4. Ha et al., 2000
Initiation 10
Propagation 1
Transfer to hydrogen
Propagation 1.01x108
Deactivation 3x10-1
7. Wu et al., 2005
Propagation
Initiation 8x104
Table 3.4: Proposed kinetic mechanisms adopted by previous workers on the field.
transformation
Transfer to co-
By co-catalyst
(spontaneous)
Deactivation
Propagation
Transfer to
Transfer to
Transfer to
Reference*
Activation
monomer
hydrogen
Initiation
Transfer
catalyst
solvent
Site
1
2
3
4
5
6
7
8
3.7.2.1. Definition
To simplify models.
94 Chapter 3: Catalytic Polymerization Kinetics
A sensitivity analysis has been conducted concerning the estimation of the kinetic
parameters, so that the minimum number of elementary kinetic reactions, able to simulate
the actual process kinetics, could be defined. The effect of kinetic rate constants was
studied in terms of variability produced regarding molecular weight distribution and
productivity. After developing a single-site kinetic model, the effect of kinetic rate
constants in the output variables of interest was calculated. The kinetic scheme includes
site activation by catalyst, chain initiation and propagation reactions, chain transfer to
hydrogen and spontaneous chain transfer and spontaneous deactivation. The sensitivity
coefficient is defined as the ratio of the output variable change to the forced input variable
change. Figure 3.11-Figure 3.20 depict the effect of kinetic rate constants on the molecular
weight distribution. The results obtained from the sensitivity analysis are summarized in
Figure 3.21 and Figure 3.22. As can be seen, the elementary reactions that affect the
polymer molecular properties (e.g., average molecular weight) of the produced polymer are
site activation (Ka), chain propagation (Kp11) and chain transfer by hydrogen (KtrH) (Figure
3.21). On the other hand, the elementary reactions that affect the polymerization rate are
site activation (Ka) and chain propagation (Kp11) (Figure 3.22).
Chapter 3: Catalytic Polymerization Kinetics 95
1.5
1
Ka = 2 10 lt/mol/s
2
Ka = 2 10 lt/mol/s
2
Ka = 5 10 lt/mol/s
3
1.0 Ka = 1 10 lt/mol/s
dW/dlog(MW)
3
Ka = 2 10 lt/mol/s
0.5
Ka
0.0
2 3 4 5 6 7
log(MW)
1.5
6
K01 = 5 10 lt/mol/s
8
K01 = 5 10 lt/mol/s
10
K01 = 5 10 lt/mol/s
1.0
dW/dlog(MW)
0.5
K01
0.0
2 3 4 5 6 7
log(MW)
1.5
6
K02 = 4 10 lt/mol/s
8
K02 = 4 10 lt/mol/s
10
K02 = 4 10 lt/mol/s
0.5
0.0
2 3 4 5 6 7
log(MW)
1.5
8
Kp11 = 3 10 lt/mol/s
9
Kp11 = 3 10 lt/mol/s
10
Kp11 = 3 10 lt/mol/s
11
1.0 Kp11 = 3 10 lt/mol/s
dW/dlog(MW)
0.5
Kp11
0.0
2 3 4 5 6 7
log(MW)
1.5
9
Kp12 = 9 10 lt/mol/s
10
Kp12 = 9 10 lt/mol/s
11
Kp12 = 9 10 lt/mol/s
1.0
There is no effect of Kp12 on MWD
dW/dlog(MW)
0.5
0.0
2 3 4 5 6 7
log(MW)
Figure 3.15: Effect of Kp12 on MWD.
1.5
9
Kp21 = 3 10 lt/mol/s
10
Kp21 = 3 10 lt/mol/s
11
Kp21 = 3 10 lt/mol/s
1.0
dW/dlog(MW)
0.5
0.0
2 3 4 5 6 7
log(MW)
1.5
9
Kp22 = 9 10 lt/mol/s
10
Kp22 = 9 10 lt/mol/s
11
Kp22 = 9 10 lt/mol/s
1.0
dW/dlog(MW)
0.5
0.0
2 3 4 5 6 7
log(MW)
Figure 3.17: Effect of Kp22 on MWD.
1.5
6
KtrSP = 1 10 1/s
8
KtrSP = 1 10 1/s
9
KtrSP = 1 10 1/s
10
1.0 KtrSP = 1 10 1/s
dW/dlog(MW)
0.5
KtrSP
0.0
2 3 4 5 6 7 8
log(MW)
1.5
9
KtrH2 = 5 10 lt/mol/s
10
KtrH2 = 5 10 lt/mol/s
11
KtrH2 = 5 10 lt/mol/s
1.0
dW/dlog(MW)
0.5
KtrH2
0.0
2 3 4 5 6 7
log(MW)
Figure 3.19: Effect of KtrH2 on MWD.
1.5
1
Kd = 6 10 1/s
2
Kd = 6 10 1/s
3
Kd = 6 10 1/s
1.0
dW/dlog(MW)
0.5
Kd
0.0
2 3 4 5 6 7
log(MW)
0.13
0.12
0.00
-0.05
Figure 3.21: Effect of kinetic rate constants on number average molecular weight, Mn.
0.11
Sensitivity Coefficient, (DC/c)/(DK/k)
0.10
0.09
0.005
0.000
Ka K01 K02 Kp11 Kp12 Kp21 Kp22 KtrH KtrSP Kd
-0.005
Literature review and sensitivity analysis results information leaded to the reduction
of the elementary reactions comprising the employed kinetic scheme. Thus, the number of
kinetic parameters that need to be estimated is reduced, while these reactions are able to
reproduce the actual molecular properties appearing in the industrial process studied. The
simplified kinetic scheme comprises of elementary reactions including site activation by
co-catalyst, chain initiation and propagation, chain transfer to hydrogen, to monomer and
spontaneous chain transfer as well as spontaneous deactivation (Table 3.5).
k
Kk
by co-catalyst: S p A K
aA
P0k B by hydrogen: Pnk,i H 2
trHi
P0k Dnk
Kk
Chain initiation by monomer j: Pnk,i M j
trMj
P1k, j Dnk
Kk Kk
by monomer i: P0k M i
0i
P1,ki spontaneous: Pnk,i
trSpi
P0k Dnk
Kk k
by monomer 1: Pnk,1 M 1
p 11
Pnk1,1 spontaneous: Pnk,i K
dSP
CDk Dnk
Pnk, 2 M 1
Kk
Pnk1,1 Kk
P0k k
p 21
dSP
CD
Kk
by monomer 2: Pnk,1 M 2
p 12
Pnk1, 2
Kk
Pnk, 2 M 2
p 22
Pnk1, 2
Polymer Moments
k
R k k01 M1 P0k k kp 21 M1 0,2
k
k kp12 M 2 0,1
k
0,1
(3.32)
k
ktrH k
H 2 ktrSP k
0,1 k
kdSP k
0,1
k
R k k02 M 2 P0k k kp12 M 2 0,1
k
k kp 21 M1 0,2
k
0,2
(3.33)
k k k k k
(ktrH H 2 ktrSP ) 0,2 kdSP 0,2
k
R k k01 M1 P0k k kp11 M1 0,1
k
k kp 21 M1 1,2
k
k kp 21 M1 0,2
k
k kp12 M 2 1,1
k
1,1
(3.34)
k k k k k
(ktrH H 2 ktrSP ) 1,1 kdSP 1,1
k
R k k02 M 2 P0k k kp12 M 2 1,1
k
k kp12 M 2 0,1
k
k kp 22 M 2 0,2
k
1,2
(3.35)
k kp 21 M1 1,2
k k
(ktrH k
H 2 ktrSP k
) 1,2 k
kdSP k
1,2
k
R k k01
2,1
M1 P0k 2 k kp11 M1 1,1
k
k kp11 M1 0,1
k
k kp 21 M1 2,2
k k
2 1,2 k
0,2
(3.36)
k kp12 M 2 2,1
k k
(ktrH k
H 2 ktrSP k
) 2,1 k
kdSP k
2,1
Chapter 3: Catalytic Polymerization Kinetics 103
k
R k k02
2,2
M 2 P0k k kp12 M 2 (2,2
k k
2 1,2 k
0,2
) k kp 22 M 2 2 1,2
k k
0,2
(3.37)
k kp 21 M1 2,2
k k
(ktrH k
H 2 ktrSP k
) 2,2 k
kdSP k
2,2
k
R k (ktrH
0
k
H 2 ktrSP k
) 0,1 k
(ktrH k
H 2 ktrSP k
) 0,2 k
kdSP k
0,1 k
0,2 (3.38)
k
R k (ktrH
1
k
H 2 ktrSP k
) 1,1 k
(ktrH k
H 2 ktrSP k
) 1,2 k
kdSP k
1,1 k
1,2 (3.39)
k
R k (ktrH
2
k
H 2 ktrSP k
) 2,1 k
(ktrH k
H 2 ktrSP k
) 2,2 k
kdSP k
2,1 k
2,2 (3.40)
Liquid-phase Components
Monomer 1 (ethylene):
s N s N s N
k
RM1 k01 M1 P0k k Pk 11 M1 0,1
k
k Pk 21 M1 0,2
k
(3.41)
k 1
k 1 k 1
Monomer 2 (1-hexene):
s N Ns Ns
k
RM 2 k02 M 2 P0k k Pk12 M 2 0,1
k
k Pk 22 M 2 0,2
k
(3.42)
k 1
k 1 k 1
Hydrogen:
sN
k k k
RH 2 ktrH H 2 0,1 0,2 (3.43)
k 1
104 Chapter 3: Catalytic Polymerization Kinetics
Co-catalyst:
sN
k
R A kaA A S kp (3.44)
k 1
k
RSp k kaA A Sp k (3.45)
k
RP k kaA
0
A S kp k01
k k
M1 k02 M 2 P0k ktrH
k
k
H 2 0,1 k
0,2 (3.46)
k k k k
ktrSP1 0,1 ktrSP 2 0,2
The most probable weight chain length distribution (i.e., WCLD) of linear homo-
and binary co-polymers may be expressed by the equation:
Chapter 3: Catalytic Polymerization Kinetics 105
w r, j 2 ( j ) r exp ( j ) r (3.47)
where w r, j is the most probable WCLD (weight fraction of polymer of chain length n
produced), r is the polymer chain length and j is the active site type.
the weight or mass fraction of polymer made by site type j and n is the number of different
catalyst active site types.
m( j ) 1
j 1
(3.49)
n 1
W r w( r,n ) m( j ) w( r, j ) w( r,n ) (3.50)
j 1
We want to minimize the difference between W r computed by Eq. 3.50 and the
Eq. 3.50 is valid for describing an instantaneous WCLD formed by the superposition
of several individual most probable WCLDs. However, in practice, one has information
106 Chapter 3: Catalytic Polymerization Kinetics
only about WCLDs for polymer accumulated over a finite polymerization time. Therefore,
to apply the proposed model to experimental WCLDs, we must assume that the WCLD for
the accumulated polymer is essentially the same as the instantaneous WCLD. This
assumption is valid when:
1) The polymerization reactor is operated at steady state conditions and the WCLD
is spatially independent.
2) The ratio of transfer to propagation rates of all active sites does not change
during the polymerization.
3) The relative amounts of polymer made by each site type do not change during
the polymerization.
4) Mass- and heat-transfer effects are negligible, since these effects could give an
instantaneous WCLD that is spatially dependent.
Finally, it should be mentioned that the values of mi and ti were estimated using a
non-linear parameter estimator and available experimental MWDs for different polymer
grades. The numerical methods used for the MWD deconvolution was the generalized
regression estimator (i.e., GREG).
According to the described analysis the deconvolution of the GPC data for a polymer
can be derived. Thus, the minimum number of catalyst active site types that accurately
describes the MWD, the weight fraction of polymer and the corresponding M n produced
at each catalyst site type and the predicted M n and M w for the entire distribution can be
determined.
In Figure 3.23, the effect of number of different catalyst active site types on %
square error deviation of the predicted MWD to the actual MWD is presented. It is evident
that as the number of catalyst active sites increases, the percentage square error deviation
decreases to a minimum value (i.e., 0.01).
In Figure 3.24, the effect of number of different catalyst active site types on the
derivative of the square error deviation of the predicted MWD to the actual MWD, is
depicted. As can be seen, the critical number of different catalyst active site types is 4, as
further increase of the number of site types would not significantly increase the derivative
of the deviation. With 4 different catalyst active site types, the expected deviation of the
predicted MWD to the actual MWD is less than 0.05%. Thus, the simplified multi-site
Chapter 3: Catalytic Polymerization Kinetics 107
catalytic kinetic scheme (see Table 3.5) can be deducted to a four-site Z-N polymerization
kinetic scheme.
In Figure 3.25, the experimentally measured MWD (marked with discrete points) is
compared with the reconstructed MWD using a two- to five-catalyst active sites model
(regarding Grade 1 data). The results indicate that at least a four-site model should be
considered to predict the shape of MWD.
Table 3.6 shows a representative set of deconvolution results for MWD provided by
TPRF plant data (regarding 6 different grades data). Table 3.6 depicts the fraction of each
site type for all different grades, based on the extracted results, for a four-site kinetic
scheme. The fraction values for each grade give us an indication on catalyst type
distribution. The average value of these calculated site type fractions for all grades can
give a ‘statistically’ correct catalyst site type distribution. Based on similar calculations, on
different grades, the site type fractions were determined equal to 0.1:0.2:0.3:0.4 for the 4
different catalyst active site types. It should be mentioned that all catalyst sites of the same
type have exactly the same kinetic rate constants. The values of number average molecular
weight produced by each catalyst site type are also presented. This information can guide
us for the determination of the kinetic rate constants through kinetic model tuning process.
108 Chapter 3: Catalytic Polymerization Kinetics
100
10
% Deviation
0.1
0.01
1E-3
1 2 3 4 5 6
Number of distributions
Figure 3.23: Effect of number of different catalyst active site types on % deviation of the
predicted MWD to the actual MWD.
0
Derivative of deviation
-4
-8
-12
-16
-20
1 2 3 4 5 6
Number of distributions
Figure 3.24: Effect of number of different catalyst active site types on the derivative of the
deviation of the predicted MWD to the actual MWD.
Chapter 3: Catalytic Polymerization Kinetics 109
a. 1.0 b. 1.0
Plant data 3 different catalyst active site types Plant data 3 different catalyst active site types
Predicted distribution Predicted distribution
Site type 1 Site type 1
0.8 0.8
Site type 2 Site type 2
Site type 3
dW/dlog(MW)
0.6 0.6
dW/dlog(MW)
0.4 0.4
0.2 0.2
0.0 0.0
2 3 4 5 6 7 8 2 4 6 8
log(MW) log(MW)
c. 1.0 d. 1.0
4 different catalyst active site types 5 different catalyst active site types
Plant data Plant data
Predicted distribution Predicted distribution
Site type 1 0.8 Site type 1
0.8
Site type 2 Site type 2
Site type 3 Site type 3
Site type 4 dW/dlog(MW) Site type 4
Site type 5
dW/dlog(MW)
0.6 0.6
0.4 0.4
0.2 0.2
0.0 0.0
2 4 6 8 2 4 6 8
log(MW)
log(MW)
active site. As the i is equal to the summation of all rates of deactivation and transfer
divided by the rate of polymerization, the rate constants for each site should be adjusted
such that the calculated i for each site becomes equal to the i obtained from
deconvolution. In the slurry-loop process, in the first reactor, the spontaneous chain
transfer constant is more important than the other transfer constants (i.e., transfer to
hydrogen. Hydrogen is mainly added in the second reactor of the process.). Therefore, by
changing the spontaneous chain transfer and propagation rate constants, i was adjusted so
that it equalled to the one obtained from deconvolution of the GPC graph related to the first
reactor of the plant. Moreover, the production rate, molecular weight, and MWD of the
product from the first reactor could be predicted. Other kinetic rate constants, such as chain
transfer to hydrogen, propagation rate constants related to the co-monomer were adjusted
in a way to achieve the desired molecular weight, production rate, MWD, and the shape of
Chapter 3: Catalytic Polymerization Kinetics 111
MWD of the produced polymer in the second reactor. The GPC graph related to the second
reactor was used just for checking the shape of the resulted MWD. If in a trial the desired
targets were not satisfied, in the next trial, the kinetic parameters of the first reactor would
be adjusted again. Considering the polymerization kinetic parameters should be the same
in the two reactors, the procedure would be repeated until the targets were achieved in each
reactor. The numerical values of all the kinetic rate constants are reported in Table 3.7
(Touloupides et al., 2010).
It should be noted that the kinetic rate constants tuning (regarding molecular
developments) is a procedure that was employed after the accomplishment of the whole
reactor model, as concentrations appearing in kinetic rate equations correspond to the local
species concentration at the active catalyst sites. Thus, only after the total reactor model
development (including kinetic, thermodynamic and reactor modelling) it was made
possible to account for the solubility effects in order to calculate the local species
concentration.
112 Chapter 3: Catalytic Polymerization Kinetics
Table 3.7: Numerical values of kinetic rate constants for a four-site Ziegler-Natta
ethylene1-hexene co-polymerization mechanism.
Activation
k
K aA 3.5 104 3.5 104 3.5 104 3.5 104 12 12 12 12
Initiation
K 0i
k 4.0 109 4.0 109 4.0 109 4.0 109 12 12 12 12
Propagation
k
K p11 5.0 1010 5.0 1010 5.0 1010 5.0 1010 12 12 12 12
k
K p21 5.0 1010 5.0 1010 5.0 1010 5.0 1010 12 12 12 12
k
K p12 1.0 1010 1.0 1010 1.0 1010 1.0 1010 12 12 12 12
k
K p22 1.0 108 1.0 108 1.0 108 1.0 108 12 12 12 12
Chain
transfer
k
K trH 2.6 1012 5.6 1011 1.8 101 6.5 1010 14 14 14 14
k
K trSpi 2.0 109 8.3 108 2.4 108 5.5 107 14 14 14 14
Deactivation
k
K dSP 6.0 103 6.0 103 6.0 103 6.0 103 12 12 12 12
Chapter 3: Catalytic Polymerization Kinetics 113
3.8. Notation
Symbols
PD : Polydispersity
t : Time, sec
A : co-catalyst
B : co-catalyst by-product
H2 : Hydrogen
Chapter 3: Catalytic Polymerization Kinetics 115
Mi : monomer ‘ i ’
S : diluent
SP : spontaneous
X : Poison
Greek letters
νth moment of the total number chain length distribution of ‘live’ co-
v :
polymer chains
νth moment of the total number chain length distribution of ‘live’ co-
v,i :
polymer chains, with terminal monomer ‘ i ’
: viscosity, kg/m/s
νth moment of the total number chain length distribution of ‘dead’ co-
v :
polymer chains
νth moment of the total number chain length distribution of ‘bulk’ co-
:
polymer chains
3.9. References
Anderson A., Cordes H.G., Herwig J., Kaminski A., Mark A., Mottweils R., Sinn J.H. and
Vollmes H.J., 1976, Anfew. Chem. Int. Ed. Engl., 15, 630.
Bhagwat M.S., Bhabwat, S.S. and Sharma M.M., 1994, ‘Mathematical Modeling of the
Slurry Polymerization of Ethylene: Gas-Liquid Mass Transfer Limitations’ Ind, Eng.
Chem. Res., 33,pp.2322-2330.
Boor J., 1979, ‘Ziegler-Natta Catalysts and Polymerizations’, Academic Press, New York.
Brintzinger H.H., Fischer D., Mulhaupt R., Rieger B. and Waymouth R.M., 1995,
‘Stereospecific Olefin Polymerization with Chiral Metallocene Catalysts’, Angew.
Chem. Int. Ed. Engl., 34, pp. 1143-1170.
Canu P. and Ray W.H., 1991, ‘Discrete Weighted Residual Methods Applied to
Polymerization Reactions’, Comp. Chem. Eng., 15, 549-564.
Choi K.Y. and Ray W.H., 1985a, ‘Recent Developments in Transition Metal Catalyzed
Olefin polymerization-A Survey. I. Ethylene Polymerization’, Polymer Reviews,
25:1, pp. 1-55.
Choi K.Y. and Ray W.H., 1985b, ‘Recent Developments in Transition Metal Catalyzed
Olefin polymerization-A Survey. II. Propylene Polymerization’, Polymer Reviews,
25:1, pp. 57-97.
Choi H.K., Kim J.H., Ko Y.S. and Woo S.I., 1997, ‘Prediction of the Molecular Weight of
Polyethylene Produced in a Semibatch Slurry Reactor by Computer Simulation’, Ind.
Eng. Chem. Res., 36, pp. 1337-1342.
Clay, P.A. and Gilbert, R.G., 1995, ‘Molecular Weight Distributions in Free-Radical
Polymerizations. 1. Model Development and Implications for Data Interpretation’,
Macromolecules, 28, 552-569.
de Carvalho A.B., Gloor P.E., Hamielec A.E., 1989, ‘A Kinetic Mathematical Model for
Heterogeneous Ziegler-Natta Copolymerization’, Polymer, 30, pp. 281-296.
Chapter 3: Catalytic Polymerization Kinetics 117
de Carvalho A.B., Gloor P.E., Hamielec A.E., 1990, ‘A Kinetic Model for Heterogeneous
Ziegler-Natta Copolymerization. Part 2: Stereochemical Sequence Length
Distributions’, Polymer, 31, pp. 1294-1311.
Dompazis G., Kanellopoulos V., Touloupides V. and Kiparissides C., 2008, ‘Development
of a Multi-scale, Multi-phase, Multi-zone Dynamic Model for the Prediction of
Particle Segregation in Catalytic Olefin Polymerization FBRs’, Chem. Eng. Sci., 63,
pp. 4735-4753.
Dusseault J.J.A. and Hsu C.C., 1993, ‘MgCl2-Supported Ziegler-Natta Catalysts for Olefin
Polymerization: Basic Structure, Mechanism, and Kinetic Behavior’, Rev.
Macromol. Chem. Phys., C32(2), pp. 103-145.
Floyd S., 1986, ‘Heat and Mass Transfer Resistances in Polymerization of Olefins over
Solid Catalysts’, PhD thesis, University of Wisconsin-Madison.
Fontes C.H. and Mendes M.J., 2005, ‘Analysis of an Industrial Continuous Slurry Reactor
for Ethylene–Butane Copolymerization’, Polymer, 46, pp. 2922-2932.
Galli P. and Vecellio G., 2001, ‘Technology: Driving Force Behind Innovation and Growth
of Polyolefins’, Prog. Polym. Sci., 26, 8, pp. 1287-1336.
Ha K-S., Yoo K-Y. and Rhee H-K., 2001, ‘Modeling and Analysis of a Slurry Reactor
System for Heterogeneous Olefin Polymerization: The Effects of Hydrogen
Concentration and Initial Catalyst Size’, Journal of Applied Polymer Science, Vol.
79, pp. 2480-2493.
Hakim S. and Moballegh L., 2006, ‘Simulation of a Series of Industrial Slurry Reactors for
HDPE Polymerization Process Using Deconvolution of the GPC Graph of Only the
First Reactor’, Iranian Polymer Journal, 15, pp. 655-666.
Hatzantonis H., Yiannoulakis H., Yiagopoulos A. and Kiparissides, C., 2000, ‘Recent
Developments in Modeling Gas-Phase Catalyzed olefin Polymerization Fluidized-
Bed Reactors: The Effect of Bubble Size Variation on the Reactor’s Performance’,
Chem. Eng. Sci., 55, pp. 3237-3259.
Hogan J.P. and Banks R.L., 1958, ‘Polymers and Production Thereof’, U.S. Patent No.
2,825,721.
118 Chapter 3: Catalytic Polymerization Kinetics
Huang J. and Rempel G.L., 1995, ‘Ziegler-Natta Catalysts for Olefin Polymerization:
Mechanistic Insights from Metallocene Systems’, Prog. Polym. Sci., 20, pp. 459-
526.
Lorenzini P. and Bertrand P. and Vilermaux J., 1991, ‘Modeling Ethylene and a-Olefin
Copolymerization Using Ziegler-Natta Catalyst’, Can. J. Chem. Eng., 69, 682.
McAuley K.B., MacGregor J.F. and Hamielec A.E., 1990, ‘A Kinetic Model for Industrial
Gas-Phase Ethylene Polymerization’, AIChE Journal, 36, pp. 837-850.
McDaniel M.P., 2010, ‘A Review of the Phillips Supported Chromium Catalyst and Its
Commercial Use for Ethylene Polymerization’, Advances in Catalysis, 53, pp. 123-
606.
McKenna T.F. and Soares J.B.P., 2001, ‘Single Particle Modelling for Olefin
Polymerization on Supported Catalysts: A Review and Proposals for Future
Developments’, Chem Eng Science, 56, pp. 3931–3949.
Neto A.G.M., Freitas M.F., Nele M. and Pinto J. C., 2005, ‘Modeling Ethylene/1-Butene
Copolymerizations in Industrial Slurry Reactors’, Ind. Eng. Chem. Res., 44, pp.
2697-2715.
Nirisen O. and Rytter E., 1986, ‘First Order Activation Second Order Deactivation
Kinetics Applied to the Polymerization with Supported Ziegler-Natta Catalysts’,
Macromol. Chem., 7, pp. 103-108.
Peacock A.J., 2000, ‘Handbook of Polyethylene’, Marcel Dekker, New York, pp. 1-10, 43-
66.
Razavi A., 2000, ‘Metallocene Catalysts Technology and Enviroment’, C. R. Acad. Sci.
Paris, Serie Iic, Chimie/Chemistry, 3, pp. 615-625.
Saltelli A., Tarantola S., Campolongo F. and M. Ratto, 2004, ‘Sensitivity Analysis in
Practice: A Guide to Assessing Scientific Models’, John Wiley & Sons, Ltd.
Saltelli A., Ratto M., Andres T., Campolongo F., Cariboni J., Gatelli D., Saisana, M., and
Tarantola S., 2008, ‘Global Sensitivity Analysis. The Primer’, John Wiley & Sons.
Soares J.B.P. and Hamielec A.E., 1995, ‘Metallocene/aluminoxane catalysts for olefin
polymerization. A review’, Polymer Reaction Engineering, 3, pp. 131-200.
Soga K.T. and Ikeda S., 1980, ‘Polymerization of Propylene over Metal Oxides-Supported
TiCl3 Catalysts’, Polymer Bulletin, 1, pp. 849-856.
Touloupides V., Kanellopoulos V., Pladis P., Kiparissides C., Mignon D. and Van-
Grambezen P., 2010, ‘Modeling and Simulation of an Industrial Slurry-phase
Catalytic Olefin Polymerization Reactor Series’, Chem. Eng. Sci., 65, pp. 3208-
3222.
120 Chapter 3: Catalytic Polymerization Kinetics
Vandenberg E.J., 1962, ‘Process for Polymerizing Olefins Wherein Hydrogen Is Utilized
as a Molecular Weight Control Agent’, U.S. Patent No. 3,051,690.
van der Ven S., 1990, ‘Polypropylene and Other Polyolefins: Polymerization and
Characterization’, Elsevier Science Publishers B. V., Netherlands.
Wells G.J., Ray W.H. and Kosek J., 2001, ‘Effects on Catalyst Activity Profiles on
Polyethylene Reactor Dynamics’, AIChE Journal, Vol. 47, No. 12, pp. 2768-2780.
Xie T., McAuley K.B., Hsu J.C.C., Bacon, D.W., 1994, ‘Gas Phase Ethylene
Polymerization: Production Processes, Polymer Properties, and Reactor Modeling’,
Industrial and Engineering Chemistry Research, 33, pp. 449-479.
Yiannoulakis H., Yiagopoulos and A., Kiparissides C., 2001, ‘Recent developments in the
particle size distribution modeling of fluidized-bed olefin polymerization reactors’,
Chem. Eng. Sci., 56, pp. 917-925.
Zacca J.J. and Ray H.W., 1993, ‘Modelling of the liquid phase polymerization of olefins in
loop reactors’, Chem. Eng. Sci., Vol. 48, No. 22, pp. 3743-3765.
Zechlin J., Hauschild K. and Fink G., 2000, ‘Silica Supported Metallocene/MAO-Systems:
Comparison of the Polypropylene Growth During Bulk Phase polymerization with
Slurry phase experiments’, Macromol. Chem. Phys., 201, pp. 597-603.
Chapter 4: Slurry-loop Reactor Modelling 121
Chapter 4
4.1. Introduction
In the present chapter, the modelling approach followed for the dynamic simulation
of a typical industrial-scale slurry-loop reactor series is presented. All the steps followed
for the simulation of the industrial slurry-loop reactor, including kinetic modelling, mass
and energy balances, thermodynamic considerations, physical and transport properties
calculation, reactor control system modelling, rheological properties calculation, particle
size distribution calculation, etc., are thoroughly analyzed.
The first commercially available technology for the production of high density,
linear polyethylene grades was the solution process. It was soon discovered that a more
efficient way to produce various PE grades was to carry out the polymerization under
slurry conditions (Franklin et al., 2004, Hottovy et. al., 2004). In the sixties, Phillips
Petroleum Co. developed the first loop reactor technology for the production of polyolefins
in a slurry-phase. According to this technology, the polymerization is carried out in one or
more continuous loop reactors operating in series. The product is withdrawn from the
bottom of the loop reactor via the semi-continuous operation of a product withdrawal
system (e.g., settling legs). The slurry-phase loop reactor technology has gained
122 Chapter 4: Slurry-loop Reactor Modelling
The main reasons for the wide use of slurry-phase loop reactors are:
Figure 4.1: Schematic representation of an industrial slurry-phase olefin catalytic polymerization cascade-loop reactor series.
124 Chapter 4: Slurry-loop Reactor Modelling
The first reactor of the series is continuously fed with monomer, co-monomer,
hydrogen, diluent and catalyst. The olefin feed used for the polymerization is at least one
olefin selected from 1-olefins having a maximum of 8 carbon atoms per molecule and no
branching nearer the double bond than the 4-position. Examples of olefins which can be
polymerized in a slurry-loop reactor include ethylene, propylene, 1-butene, 1-pentene and
1,1-butadiene. Particle form polymer can be prepared from ethylene and mixtures of
ethylene with minor amounts of other unsaturated hydrocarbons as co-monomers, such as
propylene, 1-butene, 1-pentene, 1-hexene (Scoggin, 1966). It should be noted that the
monomers are introduced into the reactor as a gas. The monomers dissolve in the liquid
phase for the reaction to take place on the solid catalyst. Typical ethylene content in
reactors’ feed varies from 70-250 wt. % based on the weight of the diluents (Marechal,
2006). Suitable diluents include parafins such as those having from 3-12 (preferably 3-8)
carbon atoms per molecule (e.g., n-butane, n-pentane, iso-pentane, n-hexane, n-decane,
etc.) and saturated cyclic hydrocarbons (e.g., cyclohexane, cyclopentane and
methylcyclopentane, methylcyclohexane, etc.) (Scoggin, 1966).
Catalyst is introduced via a catalyst introduction port, before the centrifugal pump
(Kufeld et al., 2006). Typical catalyst concentration ranges from 0.001-5% by weight
based on the diluent (Scoggin, 1966). It should be noted that the catalyst type greatly
affects the polymer particles morphology that, in turn, can affect the particle circulation
Chapter 4: Slurry-loop Reactor Modelling 125
behaviour and particle agglomeration that can cause the appearance of temperature hot
spots in the loop reactor (Marechal, 2006). Moreover, the physical properties of a
polyethylene product vary depending on the catalytic system employed for the polymer
production. This is because different catalytic systems tend to yield different MWDs. A
number of catalytic systems have been employed for the production of polyethylene in a
slurry-loop reactor (Lonfils at al., 2000). The slurry-loop process originally employed
chromium-based catalysts (i.e., Phillips catalyst). Chromium-based catalytic systems for
the loop reactor are discussed by the Patent of Lonfils at al., 2000. Today, industrial scale
olefin polymerization reactors are commonly operated in the presence of third and fourth-
generation Z-N catalysts (Marechal, 2006, Reginato et al., 2003). The system employed for
this project work is ethylene-1-hexene Z-N polymerization with iso-butane as a diluent.
During polymerization, the polymer solids are gradually collected in the settling legs
placed at the lower part of the loop reactor. The settling legs periodically open to remove
the highly concentrated slurry (i.e., consisting of polymer solids and a fraction of the liquid
phase). The product stream leaving the first loop reactor is fed to the second reactor of the
series together with fresh monomer(s), diluent and hydrogen. The highly concentrated
slurry product leaving the second reactor is then flashed from the operating pressure (e.g.,
41 bar) to 2 bar (Salmon, 2003). The vaporized unreacted monomer(s) and diluent exit the
flash chamber via a conduit for further processing which includes condensation by simple
heat exchange using a recycle condenser, and return to the system without the necessity for
compression via a recycle diluents line. Thus, the diluent is completely recovered due to
the high monomer(s) conversion (i.e., 95%-98%) while there is no need for monomer(s)
recovery. Polymer particles are withdrawn from the high pressure flash chamber for further
processing. They are passed to the low pressure flash chamber and they are recovered as
the polymer product (Kufeld et al., 2006), which finally is dried and pelletized.
mixing conditions (Zacca and Ray, 1993). The velocity should be maintained high enough
to avoid saltation or deposition of solids from the slurry. The saltation velocity is defined
as the minimum flow velocity needed to avoid saltation and varies with process conditions.
The value of the saltation velocity is proportional to the reactor diameter, as well as, to the
percentage of solids in the slurry. Since high solids are desirable, one way to maintain flow
above the saltation velocity at high solids content is to use relatively small-diameter reactor
(Kufeld et al., 2006). Moreover, the highly turbulent flowrate ensures a high heat transfer
rate from the reaction mixture to the water coolant, flowing inside the tube’s jacket and, at
the same time, prevents reactor fouling caused by particle deposition on the internal
surface of the tube’s wall. The high flow circulation rates make possible the high-solids
operation of the loop reactor (e.g., larger than 30 % w/w). This represents a clear
advantage of the loop reactors over the operation of conventional stirred tanks where
serious mixing and heat transfer limitations can arise at high-solids concentrations (Drusco
and Rinaldi, 1984, Ferrero and Chiovetta, 1990).
The reactor is run ‘liquid full’ (Kufeld et al., 2006). In general, in a slurry-phase loop
reactor two phases, namely, a liquid phase (i.e., consisting of diluent, monomer, co-
monomer and hydrogen) and a polymer phase (i.e., consisting of polymer and sorbed
quantities of diluent, monomer, co-monomer and hydrogen) are present. Typical reactor
mass fractions for polymer, diluent and monomer would be 30%, 68% and 2%,
respectively, although these fraction values may vary considerably (Ayres, 1986).
Various polyolefin grades with broad or/ and bimodal MWDs can be produced using
two slurry-loop reactors in series. It is well known that in order to achieve good
processability of polyethylene resins, it is desired that the flow properties and the shear
response of the polyethylene are improved by broadening the MWD. Moreover, in order to
improve the mechanical properties of the resin, it is desired to separate as far as possible
the polymerization reactions occurring in the two reactors so that the difference in density
and molecular weight between the two resin fractions is enhanced for any given target
density of the resin.
One of the reactors of the series produces a high-molecular weight fraction, and the
other of the reactors produces a low- molecular weight fraction. The resultant bimodal
resin, comprising a chemical blend of the two fractions has particular applications (e.g., for
use as a pipe resin which exhibits good mechanical properties such as environmental stress
crack resistance and slow crack growth resistance) (Marechal, 2006).
Chapter 4: Slurry-loop Reactor Modelling 127
A typical proportion between the high and low molecular weight fractions is 50/ 50
wt. % blend between the two fractions. For a 50/ 50 wt. % blend between the two fractions,
it is clear to achieve a target density of the combined blend forming the pipe resin, which
target density is required commercially, for any decrease in density of high-density, low
molecular weight fraction there must be a corresponding increase in the density of the low
density high molecular weight fraction. However, the proportion between the high and low
molecular weight fractions may vary depending on the final product properties desired.
Typically, for pipe resins, the blend comprises from 40-57 wt. % high molecular weight
fraction. For film resins, the range is broader depending on the key resin properties that are
required (e.g., 30-75 wt. %). Reducing the high molecular weight fraction, decreases the
density of the high molecular weight resin at constant final resin density. Increased reactor
independence enables the achievement of enlarged density differences or enlarges
differences in molecular weight between the two fractions, leading to improved mechanical
properties (Marechal, 2006).
Typically, polyolefins of high molecular weight and low density are produced in the
first reactor of the series, commonly operated at low hydrogen concentrations and high co-
monomer concentrations (e.g., 2-15 wt. % based on the weight of the diluent). On the other
hand, in the second reactor of the series, high hydrogen (e.g., 0-5mole % based on the
weight of the diluents in the second reactor) and low co-monomer concentrations result in
the production of low molecular weight and high density polyethylene (Marechal, 2006).
The polymerization temperature in the first loop reactor typically ranges from 70-
100oC (preferably about 80oC) when the high molecular weight fraction is produced in the
first reactor and for the second reactor from 80-120oC (preferably around 95oC) when the
low molecular weight fraction is produced in the second reactor. It should be noted that the
temperature affects the solids content as an increase in temperature tends to lower the
viscosity of the diluent (Marechal, 2006). On the other hand, there is a practical limit on
increasing temperature because the point is quickly reached where the polymer goes into
solution and fouls the reactor. In general, reactor fouling occurs when polymer particles
dissolve into reactors’ liquid phase. Specifically, the polymer particles when dissolving
increase the volume that they occupy by several orders of magnitude. This increase in
volume causes the viscosity to increase in the reactor. If the polymer particles continue to
dissolve, thereby increasing the viscosity, eventually a gel-like substance is formed which
128 Chapter 4: Slurry-loop Reactor Modelling
plugs the reactor. As it is well known, unplugging a fouled reactor is a time-intensive and
costly undertaking (Benham et al., 1991).
Solids content is at least 30 wt. % based on the weight of the diluent, for a Z-N
catalyst (Marechal, 2006), while solids concentrations values equal to 45 wt. % are typical.
The solids concentration can be increased by using a circulating pump having a diameter
that is larger than the diameter of the reactor tube. This is achieved by providing a
localized enlargement in the reactor tube at the location of the propeller of the pump.
Furthermore, it is known that the replacement of the conventional settling legs in a slurry-
loop reactor, which are provided for periodic and sequential take-off of the polyethylene
fluff by a so called ‘continuous product take-off’ can also lead to higher solids
concentration in the reactor (Marechal, 2006).
Finally, the polymerization pressure ranges from 30-90, preferably about 41 bars
(Marechal, 2006). For the case of loop reactors in series, second reactor’s pressure is kept
in lower values in order the transport of first reactor’s outflow to the second reactor to be
feasible by means of pressure difference.
solids concentration of the slurry finally recovered as product slurry. They are attached to
the lower horizontal section of the loop reactor to gravitationally draw off a concentrated
fraction of the solid polymer (Marwill, 1966).
Polymer slurry is periodically removed in small batches through lines and valves
which serve to isolate the severe conditions in the reactor from the downstream equipment.
Two valves are incorporated within the receiving zone permitting the entrapment of a
portion of the settled polymer product. Periodically, the bottom valve closes and the upper
valve opens to permit the introduction of polymer product into the lock. Then, the top
valve closes and the bottom opens to prevent the escape of the polymer with a small
amount of accompanying diluent and reactant (Scoggin, 1966) (see Figure 4.2).
of the pressure in the loop reactor, affecting reactant concentrations and product quality
and consistency (Salmon, 2003).
Note that, during a given reactor period (e.g., startup, steady-state, etc.), a certain
number of settling legs can be in operation (i.e., 1-6). Thus, each settling leg will have its
next firing after a full firing-cycle (i.e., firing period) of all other settling legs in operation
is completed. This permits polymer settling in the leg, enhancing solids concentration.
During steady state operation of polymerization process, firing period values range from
20-50 sec (depending on reactor pressure and number of settling legs in operation)
(Marwil, 1966). The use of settling legs in a loop reactor for manufacturing an ethylene
butane co-polymer is well described by the patent of Marwil, 1966.
Settling legs do present two problems. First, they represent the imposition of a
‘batch’ technique onto a basic continuous process. Each time a settling leg reaches the
stage where it ‘fires’, accumulated polymer it causes an interference with the flow of slurry
in the loop reactor upstream and the recovery system downstream. Also, the valve
mechanism essential to periodically seal off the settling legs from the reactor upstream and
the recovery system downstream requires frequent maintenance due to the difficulty in
maintaining a tight seal with the large diameter valves needed for sealing the legs.
Secondly, as reactors have gotten larger, logistic problems are present by the settling legs.
If a pipe diameter is doubled, the volume of the reactor goes up four-fold. However,
because of the valve mechanism involved, the size of the settling legs cannot easily be
increased further. Hence, the number of legs required begins to exceed the physical space
available.
Despite the problems that may arise from the semi-continuous operation of the
settling legs, they are still employed in slurry-phase olefin polymerization processes
(Franklin et al., 2004), particularly for the production of high density polyethylene (Noll,
2005). This is because, unlike bulk slurry polymerizations (i.e., where the polymer is the
diluent) where solids concentrations are higher than 60%, olefin polymer slurries in a
diluent are generally limited to no more than 40% weight. Hence, settling legs have been
believed to be necessary to give a final slurry product at the exit of the settling legs of
greater than 40% (Franklin et al., 2004, Hottovy at al., 2004). The enhancement factor (i.e.,
the ration of solids concentration at the exit of settling legs to solids concentration in the
reactor) in steady-state operation for settling legs operation is around 1.25. It must be
Chapter 4: Slurry-loop Reactor Modelling 131
emphasized that in a commercial operation as little as one percent point increase in solids
concentration is of major significance (Franklin et al., 2004).
The importance and benefits of reactor modelling have been widely recognized by
both industrial and academic researchers. Process models play a significant role in
improving product quality, plant efficiency, and safety. A mathematical model that can
reliably predict the behaviour of a specific unit or process, becomes a valuable tool that can
be applied to all tasks of process operation. The use of these models in process design,
simulation, optimization, and control, promises to have a profound commercial impact on
the chemical industry is highly desirable. In the past 30 years, the development of
computer simulation tools for the chemical industry has dramatically advanced.
Mathematical models can be helpful in process analysis and control in the following ways:
In recent years, interest has focussed on the prediction of the end-use polymer
properties (i.e., number- and weight average molecular weight, molecular weight
distribution, co-polymer composition, polymer density, complex viscosity, particle size
distribution, etc.). To develop a predictive model, account must be taken of the chemistry
and physics of all of the relevant microscopic processes which occur in the polymerization
process. Detailed physical property and thermodynamic data on the partitioning of species
among phases is required to quantitatively calculate the concentrations of reactants at the
loci of polymerization. Valid kinetic rate constants (frequency factors and activation
energies) are also required (Ulmann’s Encyclopedia of Industrial Chemistry). The fact that
each polymerization process involves a number of unique physical and chemical
phenomena (e.g., reaction kinetics, physical and transport phenomena, thermodynamics,
reactor configuration, etc.) increases the scope for the development of custom-made
computer aided design software tools for specific polymerization processes (Krallis et al.,
2010).
Polymer chain reactions (e.g. chain growth, chain branching, chain termination, etc.)
related with the kinetic mechanism occur at the microscale. If the elementary reaction steps
of a polymerization mechanism are known, several mathematical techniques are available
for calculating the molecular property distributions, in terms of the kinetic rate constants
and the concentration of the reactants. The most powerful approach for modelling
polymerization kinetics is the detailed species balance method. Based on the classical
physical law of conservation of mass, one can derive an infinite set of differential or
algebraic difference equations, depending on the reactor type, for the different molecular
species (e.g., monomer(s), growing polymer chains, dead polymer chains) present in the
reaction mixture. The resulting system of species balance equations must be solved to
obtain information on the desired molecular property distributions.
Finally, at the macroscale, one has to deal with the development of models
describing the macromixing phenomena in the reactor, the overall mass and energy
balances, particle population balances, the heat and mass transfer from the reactor as well
as the reactor dynamics and control.
The whole process model integrates information coming from the solution of
different sub-models at all different space/ time scales (e.g., kinetic, particle, PSD and
reactor). It is apparent that a comprehensive mathematical model capable of predicting the
molecular and morphological developments in a polymer reactor in terms of the process
operating conditions should include appropriate model representations of all chemical and
physical phenomena occurring at the three scales. Guidelines for the development of a
mathematical model are thoroughly analysed in the work of Kiparissides, 1996.
The slurry-phase catalytic olefin polymerization has been extensively studied in the
last two decades. Ha et al., 2001 investigated the operation of a semi-batch slurry ethylene
polymerization reactor by employing a multi-scale, multi-phase model. Their model,
describing the gas phase, the liquid phase (including gas bubbles), and the solid polymer
particles inside the reactor was developed to assess the effect of the hydrogen
concentration and the initial size of the catalyst particles on the polymerization reaction. It
was reported that the average molecular weight of the polymer material can be influenced
by the initial size of the catalyst particles because of the existence of mass transfer
limitations, an effect that may become very important when catalyst activities are high.
Khare et al., 2002 developed a generalized polymerization process simulator to describe
the steady-state operation of a slurry-phase industrial plant. They studied the effects of
operating conditions on the polyolefin production rate and molecular properties (i.e.,
average molecular weights and polydispersity index). The model was capable of simulating
the transient operation of the polymerization plant, and, thus, it was possible to calculate
the optimal operating conditions to increase the polyolefin production rate. Fontes and
Mendes, 2005 developed a model for a continuous slurry ethylene-butene co-
polymerization reactor that was able to predict the production rates and the number and
Chapter 4: Slurry-loop Reactor Modelling 135
weight molecular weight averages of the final polymer. The kinetic scheme employed
comprises of initiation, propagation, first order deactivation, hydrogen transfer, ethylene
transfer, transfer to co-catalyst and β-hydride elimination, assuming two different catalyst
site types taking part in polymerization. The method of moments is employed for the
calculation of molecular developments. A non-uniform solid phase is considered. The
multi-grain model (i.e., single particle model) is employed for the calculation of the radial
distribution of chemical species such as the living and dead polymer chains in o a single
polymer particle. Neto et al., 2005 developed a mathematical model to simulate the
dynamic behaviour of ethylene/1-butene industrial slurry-phase polymerization reactor.
Their model is able to describe the dynamic evolution of the molecular weight averages,
co-monomer content, particle size averages, melt index, and density of the final polymer
resin.
All previous workers (i.e., Uvarov and Tsevetkova, 1974, Lepski and Inkov, 1977,
Ferrero and Chiovetta, 1990) have employed the continuous stirred-tank reactor (CSTR)
136 Chapter 4: Slurry-loop Reactor Modelling
model to describe the loop reactor. On the other hand, Weimin et al., 1991 described a
steady state mathematical model for polypropylene production in a loop reactor, employing
plug flow tubular equations. The influence of different operational parameters was
analyzed through simulation. Since only high ratios of recycle ratios were considered, no
significant concentration and temperature gradients along the length of the reactor were
found. Zacca and Ray, 1993 assumed that the flow behaviour in the loop reactor can be
approximated by a series of tubular sections interconnected by perfectly mixed inlet and
outlet zones, allowing the investigation of recycle rate, axial dispersion and heat transfer.
Reginato et al., 2003 simulated the operation of a loop reactor by assuming a non-ideal
CSTR behaviour. In their work, they introduced a constant discharge parameter accounting
for the difference in the concentrations of polymer solids in the reactor and the product
withdrawal stream.
calculate the temperature along the tube. A schematic representation of the proposed model
is depicted in Figure 4.3. The polymer molecular properties (i.e., number- and weight-
average molecular weights and molecular weight distribution) are determined by
employing a generalized multi-site, Ziegler-Natta (i.e., Z-N) kinetic scheme in conjunction
with the well-known method of moments. The Sanchez-Lacombe Equation of State (i.e.,
S-L EOS) is applied to calculate the thermodynamic equilibrium concentrations of the
various molecular species (i.e., monomer(s) diluent, hydrogen, etc) in the various phases
(i.e., solids, liquid and gas) present in the reaction mixture. Thermal effect related to
monomers adsorption in the diluents as well as Heat transfer resistances between the
phases are assumed negligible. Finally, a rheological model based on the reptation and
Rouse relaxation theories is employed to calculate the rheological behaviour (i.e., melt
viscosity versus frequency) of polyolefins produced in the industrial slurry-phase loop
reactor series in terms of the calculated MWDs. Numerical simulations are carried out to
investigate the effects of the process operating conditions (i.e., reactor temperature and
pressure, inflow rates of catalyst and monomers, feed composition, etc.) on the dynamic
behaviour of the cascade-loop reactor series (i.e., startup and grade transition) in terms of
plant operability and molecular and rheological properties of polyolefins. It is shown that
the proposed comprehensive model is capable of simulating the dynamic operation of an
industrial slurry-phase cascade-loop reactor series under different plant operating policies
(i.e., startup, grade transition, etc.).
Figure 4.3: Schematic representation of loop reactor modelling approach.
Chapter 4: Slurry-loop Reactor Modelling 139
The modelling approach followed is depicted in a logic diagram form in Figure 4.4.
The simulation model incorporates all different models presented in the precious sections
under FORTRAN environment. It should be noted that the differential equations solved at
every time, corresponding to the ones coming from mass balances for species of interest
and from energy balances, regarding reactor temperature, are solved using the ordinary
differential equation solver routine, divpag, from the numerical library of IMSL.
The actual flow sheet of the process modelled is shown in Figure 4.3. According to
our modelling approach, each loop reactor of the whole process is simulated as a CSTR. A
loop reactor can be modelled as a recycle reactor (Figure 4.5a.). The recycle ratio,
Rrecycle , of a loop reactor, defined as the ration of volumetric flow rates between the
Qrec
Rrecycle
Qout
can be made to vary from zero to infinity. Reflection suggests that as the recycle ratio is
raised, the behaviour shifts from plug flow ( Rrecycle 0 , Plug Flow Reactor, PFR) to
mixed flow ( Rrecycle , Continuous Stirred Tank Reactor, CSTR) (Figure 4.5b.). Thus,
recycling provides a means for obtaining various degrees of backmixing with a PFR
(Levenspiel, 1972). A loop reactor must converge to the stirred tank reactor (CSTR) when
the recycle ration tends to extremely large values. In practice, there is no general way to
determine this threshold recycle ratio where the loop reactor begins to behave as a CSTR.
The recycle ratio of a loop reactor can be defined as the ratio of volumetric flow rates
between the recycle stream and the outlet stream:
According to Zacca and Ray, 1993, at very high recycle ratios (i.e., Rrecycle 150 )
the reactor is fairly uniform presenting almost no concentration and temperature gradients
along the axial distance. They propose, as a ‘rule of thumb’ that for recycle ratios above
30, the loop reactor behaves as a CSTR.
For industrial loop reactor plants operating conditions, a typical recycle ratio can be
calculated:
Chapter 4: Slurry-loop Reactor Modelling 141
As can be easily understood, the assumption that industrial loop reactors behave as
CSTRs holds true, since the recycle ratio remains at large values. Thus, the whole process
(2 loop reactors in series) can be safely simulated as a two CSTR train.
spontaneous chain transfer as well as spontaneous deactivation (see Table 3.5). The
methodology followed in order to derive a simplified kinetic scheme able to simulate the
molecular developments in a slurry-loop reactor is thoroughly analyzed in section 3.7
Slurry-loop reactor kinetic scheme.
According to the proposed modelling approach, each loop (i.e., consisting of the
reactor tube and the settling legs) is modelled as an ideal CSTR (in terms of perfect mixing
conditions) in series with a semi-continuous process. In other words, the whole process
consists of a semi-continuous (in terms of reactor outflow) CSTR, where reactor outflow
may have different concentration (enhanced in solids) compared to the one in the reactor.
th
According to the modelling approach, j reactor of volume VR , j , in series may
have a continuous inflow, regarding all species of interest, i , Fin,i , j . Moreover, species of
interest may also have a consumption/ production rate term, Ri, j . This term corresponding
to the consumption/ production rates, related to the kinetic scheme employed, are
calculated in 3.5 Slurry-loop reactor kinetic scheme section (see Eq. 3.32- Eq. 3.46). It
should be noted that, the molecular species of the reactor’s bulk phase that are taken into
account are ethylene, 1-hexene, iso-butane, hydrogen, co-catalyst, vacant catalyst sites,
potential catalyst sites, zero, first and second ‘live’ and ‘dead’ moments for each catalyst
site and for each monomer. For reactor pressure values, PR, j , not exceeding the maximum
dCi , j Fin,i, j
Ri , j PR, j PR, j ,max (4.1)
dt VR, j ,
As can be seen in Eq. 4.1, no outflow term is taking part in the mass balances
equation. The outflow depends on the pressure of each reactor and the maximum allowable
pressure value. After accomplishing the appropriate thermodynamic calculations, the
outflow stream of the two reactors, as well as the corresponding inflow from first to second
Chapter 4: Slurry-loop Reactor Modelling 143
reactor was directly linked with reactor pressure. Thus, every time in which reactor
pressure, PR, j , exceeds its maximum allowable pressure, PR, j ,max , a product volume
proportional to the settling legs volume is automatically discharged from the loop reactor
(i.e., settling leg firing). The reduced values of compositions of all species of interest (in
terms of mol/ m3), as reactor volume remains the same, also reduces reactor pressure value.
Thus, the model follows a ‘zig zag’ pattern regarding the overall reactor pressure, as fresh
monomers and diluent fed into the reactor causing a new pressure increase. The
composition of the outflow volume can be either the same as in the reactor either enhanced
in solids, due to the settling legs operation. Moreover, first reactor’s outflow is fed to the
second reactor, increasing second reactor’s species of interest concentrations. Reactor
concentrations are then corrected as can be seen in Eqs. 4.2, 4.2 and 4.4.
Ci',1
Ci,1 VR,1 VL,1 Cout ,i,1 VL,1
(4.2)
, PR,1 PR,1,max
VR,1 VL,1
and,
Ci',2
Ci,2 VR,2 VL,2 Cout ,i,1 VL,1
(4.3)
VR,2 VL,2
Ci',2
Ci,2 VR,2 VL,2 Cout ,i,2 VL,2
, PR,2 PR,2,max (4.4)
VR,2 VL,2
144 Chapter 4: Slurry-loop Reactor Modelling
VL, j represents the settling leg volume (i.e., the bulk volume that is discharged every time
th
there is a leg firing). Cout ,i, j represents the outflow concentration of the i species of
th
interest from the j reactor. In order to calculate this term, all species of interest,
regarding this part of calculations, are separated into ‘liquid’ or ‘solid’ phase species. In
this way, a partial, reduced concentration, regarding liquid or solid phase can be calculated
for all species, assuming that the two phases are perfectly separated. Thus, monomers,
diluent, hydrogen and co-catalyst are treated as ‘liquid’ species, while catalyst active sites
and polymer moments are treated as ‘solid’ species. Thus, the reduced concentrations of
both ‘liquid’ and ‘solid’ species can be calculated in terms of solids volume fraction in the
reactor, s, j .
‘Liquid’ species
Cred , j Ci, j 1 s, j (4.5)
‘Solid’ species
Ci, j
Cred , j (4.6)
s, j
According to all previous calculations, assuming that solids volume fraction in the
settling leg (i.e., in the reactor outflow), out , j , can be calculated, outflow concentration
th th
of the i species of interest from the j reactor, Cout ,i , j , can be therefore calculated.
The final equation forms for both reactors and for both ‘liquid’ and ‘solid’ species, are
presented in Eq. 4.7-4.12.
It is noted that the solids volume fraction in the settling leg, out , j , is calculated
through settling leg modelling.
Chapter 4: Slurry-loop Reactor Modelling 145
Ci',1
Ci,1 VR,1 VL,1 Cred ,i,1 VL,1 1 out ,1
, PR,1 PR,1,max (4.7)
VR,1 VL,1
and,
Ci',2
Ci,2 VR,2 VL,2 Cred ,i,1 VL,1 1 out ,1 (4.8)
VR,2 VL,2
Ci',1
Ci,1 VR,1 VL,1 Cred ,i,1 VL,1 out ,1
, PR,1 PR,1,max (4.9)
VR,1 VL,1
and,
Ci',2
Ci,2 VR,2 VL,2 Cred ,i,1 VL,1 out ,1
(4.10)
VR,2 VL,2
Ci',2
Ci,2 VR,2 VL,2 Cred ,i,2 VL,2 1 out ,2
, PR,2 PR,2,max (4.11)
VR,2 VL,2
Ci',2
Ci,2 VR,2 VL,2 Cred ,i,2 VL,2 out ,2
, PR,1 PR,1,max (4.12)
VR,2 VL,2
146 Chapter 4: Slurry-loop Reactor Modelling
Product withdrawal is achieved through the settling legs operation. Settling legs play
an important role in the whole loop reactor process. Their operation is crucial as they affect
reactor concentration, which in turn, affect reactor pressure, which again affects settling
leg operation. The quantification of the difference between the concentration inside the
reactor and the one at the reactor output is achieved through the settling leg modelling. To
calculate the solids concentration in the settling legs of a loop reactor as well as their
periodic operation (i.e., discrete times of valve openings), a comprehensive model,
describing the dynamic operation of the settling legs, was developed (Touloupides et al.,
2010a). According to the proposed approach, each settling leg in a loop reactor is assumed
to comprise two zones (see Figure 4.6), namely:
1) A well packed polymer zone at the bottom of the settling leg. The void volume
fraction in the packed polymer zone is assumed to be filled with a liquid phase
of the same composition as that in the loop reactor).
2) An upper zone at the top of the settling legs containing a slurry-phase of similar
composition with that in the loop reactor.
Operating Status
Figure 4.6: Schematic representation of the settling legs employed for product withdrawal.
Chapter 4: Slurry-loop Reactor Modelling 147
Thus, when the pressure in the loop reactor reaches a specified set-point value (i.e.,
maximum allowable operating pressure), the discharge valve of a specific settling leg
opens automatically (i.e., settling leg firing) so that the product content of the leg is
discharged. This also results in a decrease of the reactor pressure. Note that, during a given
reactor period (e.g., startup, steady-state, etc.), a certain number of settling legs can be in
operation (e.g., 1-6).
The time period between the valve openings of two consecutive legs is called firing
period, t f , and corresponds to two consecutive maxima in the reactor pressure.
Assuming that N legs are in operation, the total time available for polymer solids to be
settled in a specific leg, will be equal to the summation of the last N firing periods. The
valve opening cycling period, ts , is defined as the time period between two consecutive
openings of the same valve:
ts t f (4.13)
N
In this research study, it is assumed that the volume fraction of solids in the packed
zone depend on the solids concentration in the reactor-phase as well as on the valve
opening cycling period,. Accordingly, the solids volume fraction, f p , in the packed
f p Sr s ts (4.14)
where Sr is a packing coefficient related with the solids settling rate in a leg and s is the
solids volume fraction in the reactor slurry-phase. Packing coefficient, Sr , depends on the
viscocity and the geometric characteristics of the reactor studied. The value of this
parameter was defined based on industrial plant data, provided by our industrial partner. It
should be noted that the value of this parameter practically remains stable as recycling
velocity also remains stable. Thus, for every loop reactor operating settling legs, a different
but stable value can be defined. It should be noted that the number of settling legs in
operation strongly affects the solids concentration in the product withdrawal stream. As the
number of settling legs in operation increases, the valve opening cycling period increases,
148 Chapter 4: Slurry-loop Reactor Modelling
giving more time for polymer settling and, leading to higher values of solids volume
fraction in the product outflow stream (or higher values of enhancement factor). This in
turn results in an increase of solids concentration in the product outflow stream (as more
time available for solids’ settling is given).
Similarly, the solids volume fraction in the product outflow stream, out , will be
given by:
out f p p 1 f p s (4.15)
where p is the volume fraction of the packed polymer particles in a settling leg. For this
research study, a face-centered cubic packing arrangement of spherical polymer particles
of equal size was assumed (see Figure 4.7). Thus, after simple geometric calculations one
can define that 0.7405 of the space is occupied by spheres and 0.2595 is unoccupied space.
To account for the difference in the solids concentrations in the reactor and the
product stream, a solids enhancement factor, ef , was introduced.
e f out s (4.16)
Contrary to previous studies (Reginato et al., 2003), it was found that the value of
the enhancement factor did not remain constant during reactor startup, exhibiting an
Chapter 4: Slurry-loop Reactor Modelling 149
asymptotic decrease to its final steady-state value. Comparative results, regarding settling
legs’ operation are given in Chapter 5: Results and Discussion.
Heat removal from the reactor is achieved via heat exchangers. Because of the good
mixing conditions in the loop reactor, it can be assumed that reactor temperature is, at any
time, the same in the reactor (but not constant). In other words, we simulate the loop
reactor as a CSTR.
All possible terms are taken into account for the calculation of energy balance,
regarding reactor mixture, including heat input, heat outpout, heat from reaction, heat
transferred to reactor jacket, heat loss to environment and heat provided by the pump
power (Figure 4.8).
150 Chapter 4: Slurry-loop Reactor Modelling
dTr , j N
VR , j R , j C p , R , j qin ,i , j i , j C p.i , j Tin ,i , j Tref
dt i 1
H r , j R p , j
U e , j Ae , j TR , j Te
Pp
q in,i, j , q out, j and q c, j represent the inlet and outlet volumetric flowrates in the ‘ j ’ reactor
of the series and the coolant flowrate in the ‘ j ’ reactor jacket, respectively. Tref and
Tc,out, j denote the reference temperature and the jacket outlet temperature, respectively.
Finally, the term U e, jA e, j Tr, j Te accounts for the heat losses from the free reactor
It should be noted that for the establishment of the energy balances a similar to the
mass balance approach has been followed. Thus, heat input and heat output terms take part
in the equations only when volume exits the reactor (i.e., only when pressure reaches
maximum pressure limit ant settling leg volume exits on/ and enters reactor, as explained
for mass balances).
Chapter 4: Slurry-loop Reactor Modelling 151
Similarly, one can derive the following pseudo-steady-state energy balances for the
coolant flowing in the reactor jackets. The temperature in the coolant cannot be assumed
constant through the coolant tube length. Thus, contrary to reactor mixture energy balance
calculation, the coolant jacket is simulated as a PFR.
dTc, j
dz
d r, jU j / q c, jc, jCpc, j Tr, j Tc, j
(4.19)
0 z Lc, j , Tc, j Tc,in, j at z0
d r, j and U j are the inside tube diameter and the overall heat transfer coefficient from the
reaction mixture to the coolant in the ‘ j ’ reactor, respectively. By integrating Eq. 4.19, one
can easily show that the axial variation of the coolant temperature in the jacket will be
dTc dr U r dT dr U r
dVc ' mc Cpc Ac '
Tr Tc c
dz ' m c Cpc
Tr Tc
dTc T T
a dz ' ln r c az
Tr Tc Tr ,0 Tc,0
(4.20)
Tr Tc
Tr ,0 Tc,0
e a z Tr Tc Tr ,0 Tc,0 e a z
Tc Tr Tr ,0 Tc,0 e a z
where:
dr U r
a (4.21)
m c Cpc
152 Chapter 4: Slurry-loop Reactor Modelling
Tr ,0 Tr at z=0 (4.22)
where Tc ,in , j (t , 0) and Tc, j (t , z ) denote the jacket inlet temperature and the jacket
The overall heat transfer coefficient is calculated by the following equation (Brodkey
and Hersey, 1988).
1
U Rf (4.24)
RR Rc Rw
R f denotes heat transfer resistances because of reactor fouling. The resistances accounting
for the reaction mixture, or the coolant jacket and the reactor wall are calculated as:
dc
RR (4.25)
d R hR
1
Rc (4.26)
hc
d
d w log w
Rw dR (4.27)
2 w
d
d w log w
1 d
c dR 1 R (4.28)
f
U d c hR 2k w hc
where hR and hc denote the inside and outside heat transfer coefficients (i.e., reactor and
coolant jacket heat transfer coefficients), respectively. For high Reynolds Number values
(i.e., turbulent flow and Re > 10,000), the inside and outside partial heat transfer
coefficient is calculated using the equations based on the work of Sieder and Tate (Coulson
and Richarson, 1999, Geankoplis, 1993):
0.14
hd
Nu 0.027 Re0.8 Pr1/ 3 b (4.29)
k w
where b and w are the reaction mixture viscosities calculated at the bulk reaction (or
coolant) temperature and near the reactor wall temperature, respectively.
It should be mentioned that Eq. 4.29 holds for Re 10000 , 0.7 Pr 16000 and
length to diameter ratio, Leq / Deq 60 . It should also be noted that Eq. 4.29 provides very
good results when the term Re Pr Deq / Leq takes values higher than 10. If this number
take numbers lower than 10, the outside partial heat transfer coefficient is underestimated.
For low values of Reynolds number (i.e., laminar flow and Re 2100 ) the outside
partial heat transfer coefficient is calculated using a different equation (Bird et. al., 2002):
0.33 0.14
hd d b
Nu 1.86 Re Pr (4.30)
l w
The dimensionless Reynolds and Prandlt numbers appearing in Eq. 4.29 and Eq.
4.30 are defined as follows:
d V
Re (4.31)
Cp
Pr (4.32)
154 Chapter 4: Slurry-loop Reactor Modelling
For the stable as well as dynamic operation of the simulated industrial slurry-phase
cascade-loop reactor series, the simulation of the various PID feedback controllers, shown
in Figure 4.9, was included. In particular, the dynamic process model included the
simulation of the two temperature controllers, one ethylene mass fraction controller, two
polymer solids concentration controllers and two reactor pressure controllers.
5 5
TC TC
CC
CC CC
2 2
Iso-butane Iso-butane
Ethylene Ethylene
1-Hexene
Hydrogen
Hydrogen
Settling legs Settling legs
PC PC
3 3
Catalyst
Co-Catalyst
Diluent
Polyethylene
The three basic feedback control modes that are employed are proportional (P),
integral (I) and derivative control (D). The controller compares the measured value to the
setpoint and takes the appropriate corrective action by sending an output signal to the
control valve. The ideal continuous (analog) proportional-integral-derivative (i.e., PID)
controller operation can be described by the following equation (Seborg, 1989):
1 t de(t )
p (t ) p K c e(t ) e(t ') dt ' D (4.33)
I 0 dt
To convert the ideal continuous (analog) PID controller expression (Eq. 4.33) to its
digital equivalent, the following finite difference approximations can be followed.
t n
e(t ')dt ' ek t (4.34)
0 k 1
de(t ) en en 1
(4.35)
dt t
The digital PID controller equation can be written in two ways, the position form and
velocity form. By substituting Eqs. 4.34 and Eq. 4.35 into Eq. 4.33, the position form of
the digital PID control algorithm can be obtained:
t n
pn p K c en ek D (en en 1) (4.36)
I k 1 t
Chapter 4: Slurry-loop Reactor Modelling 157
or,
t n
pn 1 p K c en 1 ek D (en 1 en 2 ) (4.37)
I k 1 t
Subtracting Eq. 4.37 from Eq. 4.36 we obtain the velocity form of the PID
controller:
t
Pn pn pn 1 K c (en en 1 ) en D (en 2en 1 en 2 ) (4.38)
I t
Trapezoidal approximation
Another variation of the digital PID controller is based on the more accurate
trapezoidal approximation for the integral:
t n e e
k k 1 t
e(t ')dt ' 2
(4.39)
0 k 1
After this expression is substituted into Eq. 4.38 the velocity form of the control law
takes the form employed in this research project.
t ek ek 1 D
Pn pn pn 1 K c (en en 1 ) (en 2en 1 en 2 ) (4.40)
I 2 t
After the control system is installed, the controller settlings must be adjusted until
the control system performance is considered to be satisfactory. This activity is referred to
as controller tuning or field tuning of the controller. Controller tuning is usually done by a
trial and error procedure. A typical approach can be summarized as follows:
158 Chapter 4: Slurry-loop Reactor Modelling
1) Eliminate integral and derivative action by setting D at its minimum value and
I at its maximum value.
In the reactor, two phases co-exist, namely a liquid phase (including isobutene,
ethylene, 1-hexene and hydrogen) and a solid-polymer phase (including polymer and
sorbed quantities of iso-butane, ethylene, 1-hexene and hydrogen). To calculate the species
consumption/ production rates, the concentration of the sorbed species in the amorphous
polymer phase must be known. Moreover, reactor pressure must be accurately calculated
because this is what defines when settling leg volume exits the reactor, decreasing the
pressure. For the thermodynamic calculations (i.e., both solubility and pressure
calculations) the Sanchez-Lacombe Equation of State is employed.
Chapter 4: Slurry-loop Reactor Modelling 159
In the reactor, two phases co-exist, namely a liquid phase (including isobutene,
ethylene, 1-hexene and hydrogen) and a solid-polymer phase (including polymer and
sorbed quantities of iso-butane, ethylene, 1-hexene and hydrogen). It should be noted that
the gas phase may be neglected due to its low fraction (tends to be equal to zero) in the
reactor phase. Moreover, the gas phase formation in the reactor is undesired due to
operating problems that the gas phase causes (i.e, trapped gas phase in the reactor top,
cavitation problems, etc.). For the calculation of the monomer(s) consumption rate(s) and
molecular weight properties of polyolefins in the slurry-phase loop reactor, the
concentrations of the sorbed monomers and other species in the amorphous polymer phase
should be known. In the heterogeneous catalytic olefin polymerization, the initially formed
polymer chains are amorphous. However, as the ‘dead’ polymer chains move away from
the polymerization loci (i.e., active catalyst sites) due to chain propagation and termination
reactions, crystallization of the amorphous polymer chains takes place. Thus, it can be
assumed that the produced polyolefin will consist of two phases, namely, an amorphous
polymer phase swollen with the sorbed reaction species and a crystalline one that acts as a
barrier to the reaction species transfer to the active catalyst sites (Hutchinson et al., 1992,
Kanellopoulos et al., 2004). Based on the mass balances, bulk concentrations in the reactor
are calculated. However, only the species which are sorbed in the amorphous polymer
phase take part in the polymerization reactions. Thus, their ‘active’ concentration is not the
same as in the bulk phase. Thus, sorbed concentrations have to be calculated.
pi k H C (4.41)
160 Chapter 4: Slurry-loop Reactor Modelling
pi is the partial pressure of the solute in the gas above the solution, C is the
concentration of the solute and k H is a constant with the dimensions of pressure divided
by concentration. The constant, known as the Henry's law constant, depends on the solute,
the solvent and the temperature.
2
log k H 2.38 1.08 c
T
(4.42)
T
The Sanchez-Lacombe equation of state (i.e., S-L EOS) is employed to calculate the
solubilities of the various reaction species (i.e., ethylene, 1-hexene, isobutene) in semi-
crystalline polyolefins over a wide range of pressures and temperatures (Sanchez and
Lacombe, 1978, Orbey et al, 1998, Rodgers, 1993).
Sanchez and Lacombe derived a dimensionless equation of state for classical fluids
based on the Ising (lattice) fluid model (see Eq. 4.43). According to the lattice theory,
fluids are mixtures of molecules and holes, confined to sites in a lattice. The S-L EOS can
be applicable for polymer species as well as conventional components. It is noted that S-L
EOS can accurately predict molar volume, fugacity coefficients, heat capacities, and
departures for enthalpy, entropy, and Gibbs free energy.
2 P T[ln(1 ) (1 1/ f ) ] 0 (4.43)
Chapter 4: Slurry-loop Reactor Modelling 161
P P / P* (4.44)
/ * (4.45)
T T / T * (4.46)
* * *
where P , and are characteristic parameters for pure substances.
The number of sites (mers) occupied in the lattice, f , can be related to the
molecular weight of the pure component, MW, according to the following equation:
f P* MW /( R T * * ) (4.47)
Ethylene, 1-hexene, iso-butane and polyethylene are taken into consideration for the
thermodynamic calculations in the reactor. It should be noted that, the reactor bulk phase
almost consists of those four species. S-L EOS can calculate the existence of two different
phases, namely, the liquid and the solid phase, their weight fraction, their density as well as
the weight fraction of each component in each phase.
By employing S-L EOS we can calculate the mass fraction of monomers in the solid
phase. The concentration in the solid phase is calculated as:
frsolid ,i solid
Csorbed , solid ,i (4.48)
MWi
162 Chapter 4: Slurry-loop Reactor Modelling
It is important to point out that in all theoretical calculations, the values of the
*
characteristic parameters of pure components P* , T * and , appearing in the S-L EOS
were directly obtained from molecular dynamics simulations (Kanellopoulos et al, 2006).
*
The values P* , T * and parameters used for ethylene, 1-hexene, iso-butane and LDPE
are depicted in Table 4.1.
In all cases, the binary interaction parameters, kij and ij , accounting for the
interactions between the sorbed molecules and the polymer chains were introduced to fit
the EOS-based predictions to available solubility measurements (Moore and Wanke 2001,
Novak et al., 2006, Parrish, 1981) for solubility and pressure calculations. The binary
interaction parameters, kij and ij , were estimated by fitting the S-L EOS to binary
equilibrium data and they are reported in Table 4.2. The binary interaction parameters for
the systems ethylene-iso-butane and hexene-iso-butane for which there is not experimental
data, are set equal to zero.
Figure 4.10. As can been seen, the binary interaction parameter exhibits an exponential
decrease with temperature. In fact, at higher temperatures (e.g., T > 65 oC), the value of k ij
asymptotically approaches zero (i.e., the interactions between the polymer segments and
the sorbed molecules become negligible).
164 Chapter 4: Slurry-loop Reactor Modelling
In Figure 4.11 and Figure 4.12, experimental and predicted solubilities of 1-hexene
and iso-butane in LLDPE (ethylene-1-hexene) co-polymer grades are depicted. Contrary to
the ethylene-LLDPE binary system, the solubilities of 1-hexene and iso-butane in LLDPE
exhibit a strongly nonlinear behaviour with pressure. Clearly, the theoretical predictions
obtained by the S-L EOS are in excellent agreement with the experimentally measured
solubilities. It is important to point out that the numerical value of the binary interaction
parameter, k ij , for the (iso-butane-LLDPE) binary system is independent of temperature.
On the other hand, the binary interaction parameter for the (1-hexene-LLDPE) binary
system follows a linear decrease with temperature (see inserted scheme in Figure 4.11).
This means that, at higher polymerization temperatures, the molecular interactions between
the unlike molecules (i.e., polymer segments and 1-hexene) decrease.
-2
6.0x10
0.05
Predicted by S-L EOS
Binary Interaction Parameter, kij
Polynomial Approximation
Solubility, greth/gram LLDPE-1-Hexene
0.04
0.03
0.02
-2 0.01
4.0x10
0.00
-5 2
kij = 0.075 - 0.019*T + 1.125*10 T
-0.01
25 30 35 40 45 50 55 60 65 70
o
Temperature, C
-2
2.0x10 o
Moore and Wanke, 2001, T = 27.7 C
S-L EOS kij = 0.030
o
Moore and Wanke, 2001, T = 47.7 C
S-L EOS kij = 0.010
o
Moore and Wanke, 2001, T = 67.7 C
S-L EOS kij = 0.0
0.0
0 10 20 30 40
Pressure, bar
1.0
0.4
Predicted by S-L EOS)
0.4
0.2
0.1
0.2
0.0
-3
kij = 0.48 - 3.5*10 T
-0.1
60 80 100 120 140 160
Temperature, K
0.0
0 2 4 6 8 10 12
Pressure, bar
0.3
Solubility, grisobutane/gram LLDPE-1-hexene
o
Parrish, 1981, T=74 C
S-L EOS kij=0.02
o
Parrish, 1981, T=82 C
S-L EOS kij=0.02
0.2
o
Parrish, 1981, T=93 C
S-L EOS kij=0.02
0.1
0.0
0 5 10 15
Pressure, bar
Csorbed , solid ,i
Csorbed , am,i (4.49)
1 Wcrys
In Figure 4.14 the linear fit as well as the corresponding experimental data is shown.
Moreover, the effect of hexene incorporation on the crystallinity of the produced
polyethylene according to other workers on the field is depicted. It is clear that crystallinity
data exhibit a degree of dispersion. According to plant crystallinity measurements
(provided by our industrial partner), it was observed that number average molecular weight
of the polymer produced may also affect crystallinity. Based on the plant data, the
following correlation can be extracted, as a function of hexene incorporation as well as
number average molecular weight, M n .
M n
%Wcrys (% Hexene mol fraction in PE ) exp( ) (4.52)
It should be noted that, Eq. 4.52 can be employed for the calculation of both
instantaneous (i.e., for fresh polymer produced) or cumulative crystallinity value (i.e, the
weighted sum of polymer crystallinity regarding all polymer mass in the reactor),
depended on the %Hexene mol fraction in PE which may refer to the instantaneous or
cumulative value of the co-monomer composition, respectively.
80
Quijada et al., 1997, DSC measurements
Quijada et al.,1997, Density Measurements
Ko et al., 1998, DSC measurements
quijada et al., 1995, DSC measurements
60 Quijada et al., 1995, DSC measurements
Villar et al., 2001, DSC measurements
Linear fit
%Crystallinity
40
20
0 5 10 15
Hexene fraction (% mol)
Figure 4.14: Literature review on the effect of 1-hexene incorporation on the crystallinity
of the produced polyethylene.
168 Chapter 4: Slurry-loop Reactor Modelling
%Wcrys %Wcrys
polymer polymer ,cryst (1 ) polymer , am (4.53)
100 100
For polyethylene, the corresponding values for polymer ,cryst and polymer , am at
25oC, are 0.997 kg/m3 and 0.854 kg/m3, respectively (Yiagopoulos et al., 2001).
where polymer ,25o C is the polymer density at 25oC as calculated from Eq. 4.53.
Chapter 4: Slurry-loop Reactor Modelling 169
1.45
0 MPa
1.40 20 MPa
1.35
Specific Volume, cm /g
3
1.30
1.25
1.20
1.15
1.10
1.05
1.00
0 50 100 150 200 250 300
o
Temperature, C
0.96
Experimental values
Linear fit
0.95
3
Density, g/cm
0.94
0.93
0.92
0.91
20 30 40 50 60 70 80 90 100 110
o
Temperature, C
The apparent viscosity of slurry is higher than the one of the diluent alone. The solid
particles behave as volumes where no velocity gradient occurs, increasing the velocity
gradient and the shear rate to the remaining distance to the reactor wall, and thus the
apparent viscosity. Different equations have been proposed to estimate to slurry viscosity
as a function of the solids content.
The viscosity of the reaction mixture can be calculated by the following correlations.
Thomas, 1965 suggested the following expression for the calculation of viscosity of
suspensions of uniform spherical particles, mix , in terms of the viscosity of the diluent,
Barnea and Mizrahi, 1976, have also suggested a correlation for the calculation of
viscosity of suspensions of spherical solid particles (Eq. 4.56).
5 3 s
mix liquid exp (4.56)
1 s
Gay, Nelson, Armstrong, 1969, have introduced the extra term, max , describing the
maximum obtainable volume fraction solids for the system. max is a tuning parameter,
depending on the suspension system studied (Eq. 4.57). Practically, max expresses the
limit-value of solids volume fraction, s , where viscosity tends to infinity values. For the
system studied (i.e., polyethylene particles suspension in i-butane), the value of max may
range between 0.45-0.70.
Chapter 4: Slurry-loop Reactor Modelling 171
s s
0.48
mix
liquid exp 2.5 (4.57)
max s
max
For the viscosity of the liquid phase it is assumed that equals to i-butane viscosity.
According to Noll, 2005, i-butane viscosity can be calculated in terms of temperature, T .
7.3891 2582.6
liquid exp (4.58)
172.23 T
In Figure 4.17, all the results of all the proposed correlations are compared, for a
typical reactor temperature, equal to 90oC. As can be seen, all the correlations are in good
agreement between them for the values of volume fraction from 0-0.3. For values of
volume fraction higher then 0.3, Gay, Nelson and Armstrong and Thomas correlation,
which are in agreement between them, predict higher values of viscosity comparing to the
Barnea and Mizrahi correlation. In the present research project, Thomas correlation was
employed for the results presented in Chapter 5: Results and Discussion.
3.0
Thomas, 1965
Gay, Nelson and Armstrong, 1969
2.5 (max. volume fraction=0.65 )
Barnea and Mizrahi, 1973
o
2.0 Temperature=90 C
Viscosity, cP
1.5
1.0
0.5
0.0
0.0 0.2 0.4 0.6
Volume fraction
Roy and Thodos estimation technique is employed for the calculation of thermal
conductivity of gases (i.e., ethylene and 1-hexene). According to the methodology
employed, a reduced thermal conductivity may be expressed as:
r (4.59)
1
T MW 3 6
210 c (4.60)
P4
c
pressure, respectively. The reduced thermal conductivity, r , can be separated into two
parts. The first, attributed only to translational energy, was obtained from a curve fit of the
data of rare gases (Roy, 1967), varies only with the reduced temperature, Tr . The second
term corresponds to the contribution of rotational and vibrational interchange, relating to
the reduced temperature and a specific constant estimated from group contributions. The
equations may be written as:
T
Tr (4.64)
Tc
Chapter 4: Slurry-loop Reactor Modelling 173
where C coefficient is a function of the number of carbons and their positions in the
carbon chain. For ethylene C 2.73 and for 1-hexene C 10.62 .
Due to the low mass fraction of ethylene and 1-hexene in the reaction mixture, it is
assumed that pressure has no effect on the value of thermal conductivity for those gases.
However, if necessary, Stiel and Thodos modification may estimate the effect of pressure
on thermal conductivity of gases. Roy and Thodos method for the calculation of thermal
conductivity of gases as well as Stiel and Thodos modification are presented in detail by
Ried et al., 1986.
Latini et al. method is employed for the calculation of thermal conductivity of pure
liquids, L (i.e., i-butane).
0.38
A 1 Tr
L (4.66)
Tr1/ 6
A* Tb a g g
A for 50 mol MW 250 mol (4.67)
MW Tc
where L is the thermal conductivity of the liquid in W/(m.K) and Tb is the normal boiling
point temperature, K. The terms A* , and depend on the compound studied. For
For simple organic liquids, the thermal conductivities values are 10 to 100 times
higher than the corresponding thermal conductivities of the low pressure gases at the same
temperature. There is little effect of pressure, and by increasing the temperature usually the
174 Chapter 4: Slurry-loop Reactor Modelling
thermal conductivity values decrease. These characteristics are similar to those noted for
liquid viscosities, although the temperature dependence of the latter is nearly exponential.
mix i mi (4.68)
Table 4.3 reports values of ethylene heat capacity as can be found in literature.. In
Table 4.4 the values of saturated ethylene for the operating range of temperature are also
reported.
Cp, J/mol/K,
Temperature, K P, bar
(Perry and Green, 1984)
275 50.97 -
6000
Saturated ethylene, Perry's Handbook
Chao, et al., 1983
5500 Egan and Kemp, 1937
Neto et al, 2005
5000
Heat Capacity, J/Kg/K
4500
4000
3500
3000
2500
2000
160 180 200 220 240 260 280 300 320 340 360
Temperature, K
Table 4.5 reports the values of 1-hexene heat capacity, as can be found in literature.
Table 4.6 reports the values heat capacity of liquid 1-hexene for the operating range of
temperature. The following correlation is proposed by McCullough, et al., 1957 for
hydrocarbons with 6 atoms of carbon (temperature range: T=150-310K):
In Figure 4.19, the measured values of 1-hexene heat capacity are compared to the
calculated values (according to McCullough, et al., 1957).
Chapter 4: Slurry-loop Reactor Modelling 177
Table 4.6: Heat capacity of liquid 1-hexene (Kalinowska and Woycicki, 1985).
243.956 1999
178 Chapter 4: Slurry-loop Reactor Modelling
2600
2400
2300
2200
2100
2000
1900
1800
160 180 200 220 240 260 280 300 320 340 360 380
Temperature, K
Table 4.7 reports constant values of i-butane heat capacity, as suggested in the
literature. Table 4.8 reports the values of saturated i-butane for the operating range of
temperature.
Chapter 4: Slurry-loop Reactor Modelling 179
Table 4.8: Heat capacity of liquid i-butane (Perry and Green, 1984, 3-199).
4000
Aston et al, 1940
3000
2500
2000
1500
1000
500
0
0 50 100 150 200 250 300 350 400
Temperature, K
Table 4.9 reports constant values of solid polyethylene constant pressure heat
capacity as can be found in literature.
883.6 or 1,147.8 5-350 K. Value per monomer unit. Chang, Westrum, et al., 1975
C p,cryst 4,185.8 0.1205 1.810 103T , J/kg/K, 310K to melting, Crystalline PE (4.70)
C p , am 4,185.8 0.3985 0.538 103 T , J/kg/K, 250-470K, Amorphous PE (4.71)
where Wcrys is the crystallinity coefficient. In Figure 4.21 the PE heat capacity range
between fully amorphous and fully crystalline morphology is displayed.
2600
Specific Heat at Constant Pressure, J/(Kg K)
2400
2200
2000
Wunderlich, B., 1962
100% Amorphous polymer
100% Crystalline polymer
1800
320 340 360 380 400
Temperature, K
Table 4.10 reports the heat of polymerization for polyethylene as can be found n
literature. As can be seen, the value of heat of polymerization may vary from 83-136
kJ/mol depending on the characteristics of the process where polymerization takes place. It
is noted that the heat of polymerization value selected for this research project was the one
proposed by Zacca and Ray, 1993.
The physical properties of coolant water are given by the following equations.
Cpc , j 4.02 exp 1.99 104 Tc , j (4.75)
Temperature, K Fit
Specific heat, Cpc , j 4.215 2.31 103 Tc , j 273.15 3.851 105 Tc , j 273.15 1.555 107 Tc , j 273.15
2 3
Viscosity, Pa s 17.9 10-4 10 10-4 6.53 10-4 4.66 10-4 3.55 10-4 2.82 10-4 c , j 0.00153 exp (Tc , j 273.15) / 29.187 0.00025
Thermal
kc , j 5.645 10 4 1.887 10 6 (Tc , j 273.15) 7.267 10 9 (Tc , j 273.15) 2
conductivity, 5.65 10-4 5.98 10-4 6.27 10-4 6.52 10-4 6.69 10-4 6.80 10-4
kJ/m2/s/K
Chapter 4: Slurry-loop Reactor Modelling 185
1025
Experimental measurement (deMan, 1999)
Fit on experimental data
Correlation of Xie et. al., 1991
1000
3
Density, Kg/m
975
950
0 20 40 60 80 100
o
Temperature, C
4.35
Experimental data (deMan, 1999)
Fit of experimental data
Correlation of Daupert and Danner, 1985
Specific heat capacity, KJ/Kg/K
4.30
4.25
4.20
4.15
0 20 40 60 80 100
o
Temperature, C
0.002
Experimental data (deMan, 1999)
Fit of experimental data
Correlation of Daupert and Danner, 1985
Viscosity, Pas
0.001
0.000
0 20 40 60 80 100
o
Temperature, C
7.0
Experimental data (deMan, 1999)
Fit of experimental data
6.8 Correlation of Daupert and Danner, 1985
Thermal conductivity, 10 KJ/(msK)
6.6
-4
6.4
6.2
6.0
5.8
5.6
5.4
0 20 40 60 80 100
o
Temperature, C
In the present research project, Pladis et al., 2008 rheological model is employed for
the prediction of the viscoelastic behaviour of polyolefins produced in slurry-loop reactors
(i.e., complex viscosity and melt flow index). In what follows, an advanced rheological
model for the calculation of the viscoelastic properties of linear polyolefins in terms of
their respective molecular weight properties is presented. The methodology for the
calculation of melt index (i.e., MI) (i.e., referring to the number of grams of polymer melt
per ten minutes pushed out of a die of prescribed dimensions under the action of a
specified load) is presented in detail in the work of Pladis et al., 2008.
The rheological model employed is based on the reptation and Rouse relaxation
theories, to calculate the rheological curve (i.e., melt viscosity versus frequency) of linear
polyolefins in terms of their respective MWDs. The dependence of the longest relaxation
time, , the polymer chain diffusion coefficient, D , and the polymer melt viscosity, ,
on the polymer chain molecular weight, MW , are given by the following correlations:
MW 2 ; D MW 1 ; η MW for MW MWe
MW 3 ; D MW 2 ; η MW 3 for MW MWe
According to Tsenoglou, 1991 and des Cloizeaux, 1990 the dynamic relaxation
modulus, Grept (t ) , of a polymer melt can be related to the molecular weight distribution
0 1/ (4.78)
Grept (t ) G N Fm ( t , MW ) w ( MW ) d log( MW )
log( MWe )
0
where G N is the plateau modulus and Fm (t , MW ) is a monodisperse relaxation function
for polymer chains of molecular weight, MW , at time t . The parameter is usually set
equal to two. However, slightly higher values for b are also often used. In this study, the
value of b was set equal to 2.25 (Van Ruymbeke et al., 2002a, Van Ruymbeke et al.,
2002b).
8 exp( i 2U (t , MW ))
Fm (t , MW )1/ (4.79)
2 iodd i2
The function U (t , MW ) relates the polymer molar mass to the relaxation time:
t MW * Mt
U (t , MW ) g (4.80)
rept ( MW ) MW M * rept ( MW )
The reptation time, rept ( MW ) , exhibits a power law dependence on the polymer
molar mass, MW :
Finally, the function g , appearing in Eq. 4.80 has the following functional
dependence on x :
2
1 e i x
0.5
g ( x) x x ( x )0.5 (4.82)
i 1 i2
To account for the contribution of the polymer chains below the entanglement
molecular weight to the dynamic relaxation modulus, the following equation was
employed (Pattamaprom and Larson, 2001, Van Ruymbeke et al.,2002a):
0
G R (t ) G N FR (t , MW ) w( MW ) d log( MW ) (4.83)
log( MWe )
i 2 t i 2 t
1 ( MW ) 1 N R ( MW )
FR (t , MW ) e R e (4.84)
N iN 3 i 1
weight, MW :
R ( MW ) K R MW 2 (4.85)
than MWe , while the second term accounts for the slower relaxation modes, but only in the
longitudinal dimension. Following the original developments of Van Ruymbeke et al.,
2002a the overall dynamic relaxation modulus will be given by the sum of the reptation
and Rouse contributions (see Eq. 4.78 and Eq.4.83):
190 Chapter 4: Slurry-loop Reactor Modelling
G (t ) Grept (t ) GR (t ) (4.86)
The numerical values of the model parameters were taken from Van Ruymbeke et
al., 2002a and Doelder, 2006 and are reported in Table 4.12. It should be pointed out that
the numerical values of the rheological model parameters can be obtained either from the
literature or from combined experimental measurements of the MWD and dynamic storage
and loss moduli of selected industrial PE grades. The estimated values of the model
parameters can then be employed to predict the rheological behaviour of other polymer
grades independently of the MWD type (e.g., unimodal, bimodal, monodisperse, broad).
Table 4.12: Numerical values of the rheological model parameters for polyethylene.
MWe (1) 0
GN (1)
MW * (1) K rept (1) KR (2)
The well known Schwarzl approximations are employed for the calculation of the
' ''
storage and loss muduli, G and G (Schwarzl, 1971):
' ''
It should be noted that the above approximations for G ( ) and G ( ) provide
excellent results in the frequency range of commonly-available experimental
*
measurements. From Eq. 4.87 and Eq. 4.88, the complex viscosity, | ( ) | , can easily be
obtained, using the following expression:
1 1/ 2
| * ( ) | (G ' ( )) 2 (G '' ( )) 2 (4.89)
G ' ( )
tan (4.90)
G '' ( )
Moreover, the zero-shear viscosity, 0 , and the steady state recoverable compliance,
0 0 G (t ) dt ; J e002 0 tG (t ) dt (4.91)
4.7.1. Introduction
skewness of the distribution, unimodal and/ or bimodal character, etc.). For reactive
particulate processes, the quantitative calculation of the evolution of the PSD presupposes
a good knowledge of the particle nucleation, growth, breakage, and aggregation
mechanisms. These mechanisms are usually coupled to the reaction kinetics,
thermodynamics (e.g., solubility of a reactant in the particulate phase), and other
microscale phenomena including mass and heat transfer between the different phases
present in the system (Kiparissides et al., 2004).
Particle nucleation often results in the formation of a large number of small particles
within a short period of time. Particle growth due to chemical reaction results in an
increase of the mean particle size and can affect the form of the PSD, particularly in size-
dependent particle growth processes. Finally, particle aggregation and breakage can result
in significant changes in the form of the PSD. The time evolution of the PSD is commonly
obtained from the solution of the general population balance equation (i.e., PBE)
governing the dynamic behaviour of a particulate process (Hulburt and Katz, 1964,
Ramkrishna, 2000).
In general, the numerical solution of the dynamic PBE for a particulate process,
especially for a reactive one, is a notably difficult problem because of both numerical
complexities and model uncertainties regarding the particle nucleation, growth,
aggregation, and breakage mechanisms that are often poorly understood. Usually, the
numerical solution of the PBE requires the discretization of the particle volume domain
into a number of discrete elements that results in a system of stiff, nonlinear differential or
algebraic/ differential equations (i.e., DAEs) that are solved numerically.
In the open literature, several numerical methods have been developed for solving
the steady-state or dynamic PBE. These include the full discrete method (Hidy, 1965) the
method of classes (Marchal et al., 1988, Chatzi and Kiparissides, 1992) the discretized
PBE (i.e., DPBE) (Batterham et al., 1981, Hounslow et al., 1988) the fixed and moving
pivot DPBE methods (Kumar and Ramkrishna, 1996a, Kumar and Ramkrishna, 1996b),
the high-order DPBE methods (Sastry and Gaschignard, 1981, Landgrebe and Pratsinis,
1990) the orthogonal collocation on finite elements (i.e., OCFE) (Gelbard and Seinfeld,
1979) and the Galerkin method (Nicmanis and Hounslow, 1998). Kiparissides et al., 2004
presents a detailed description of all the numerical methods employed for the solution of
the system of nonlinear DAEs.
Chapter 4: Slurry-loop Reactor Modelling 193
Zacca et al., 1996 developed a population balance model using the catalyst residence
time as the main coordinate, to model particle-size developments in multistage olefin
polymerization reactors. A two-site copolymerization kinetic scheme was employed to
account for the polymerization kinetics and some functional, obtained by fitting simulated
data from a detailed particle growth model, was used to imitate the effect of mass transfer
resistances (effectiveness factor) on particle growth. Choi et al., 1994 followed the
population balance approach of Kunii and Levenspiel, 1991, to investigate the effect of
feed catalyst size distribution on the average molecular properties and the particle size
distribution of the polyolefin produced in a fluidized-bed reactor for both non-deactivated
and deactivated catalysts. A two-site copolymerization kinetic scheme and a multigrain
solid core model were used to describe the molecular weight developments and monomer
diffusion in the growing polymer particle. Soares and Hamielec, 1995 developed a
mathematical model to analyse the effects of ideal and non-ideal reactor residence time
distribution on the polymer particle-size distribution. For reaction rates relevant to typical
Ziegler-Natta and metallocene catalysts, the effects of non-ideal flow and catalyst
deactivation were analysed. In Touloupides et al., 2010b, a comprehensive multi-scale
mathematical model was developed to predict the dynamic evolution of morphological
(i.e., particle size distribution (PSD)) and molecular (i.e., molecular weight distribution
(MWD)) distributed polymer properties in catalytic Ziegler-Natta (Z-N) gas and slurry
phase olefin polymerization reactors. According to the multi-scale description of the
continuous olefin polymerization reactors, models at three different levels, namely, a
kinetic model, a single particle model and a population balance model were employed.
Moreover, the effect of reactor media (e.g., slurry versus gas phase) on the MWD and PSD
of the produced polymer particles was thoroughly analyzed.
According to the developments of Randolph and Larson, 1971, Litster et al., 1995,
Dompazis, et al., 2008 and Touloupides et al., 2010b, the following dynamic population
balance equation, accounting for both particle growth and particle agglomeration can be
derived to describe the evolution of PSD in a continuous stirred tank reactor:
n p D,t G D n p D,t 1
Bt D,t Dt D,t Fc nc D,t Fp n p D,t (4.92)
t D W
where:
Bt D,t
D 2
(D3 Dmin
3
)1 / 3
K ag D D
3 3 13
D D ,t n D,t dD
,D n p 3 3 13
p
(4.93)
2 Dmin
D D
3
3 23
and,
Dmax
Dt D,t n p D,t K ag D,D n p D,t dD (4.94)
Dmin
Fc and Fp (g/s) denote the catalyst particles feed into the reactor and the product
withdrawal rate respectively. W t (g) is the mass of the solids in the reactor that can vary
with time. n p D,t and nc D,t , expressed in (#/g/cm), are the corresponding number
density function of the particles in the reactor and in the feed stream, respectively. The
term n p D,t dD denotes the number of particles in the size range D, D dD .
G D 2 R p D ρ p πD 2 (4.95)
Chapter 4: Slurry-loop Reactor Modelling 195
The product withdrawal rate, Fp (g/s), can be easily calculated via a mass balance
equation:
Dmax
ρ p πD 3
Fp Fc W G D n p D,t d
6 (4.96)
Dmin
Notice that the second term on the right-hand side of Eq.4.96 accounts for the total
polymer production rate in the reactor. From Eq. 4.96, one can calculate the product
withdrawal rate, Fp , provided that the solids weight, W t , the particle growth rate,
For the special case of a uniform size catalyst feed, the total number of catalyst
particles in the feed stream per unit mass will be given by:
N c Dc 6 p Dc3 (4.97)
where Tav denotes the average temperature of the two colliding particles of sizes Di and
D j , and Tsf is the average polymer softening temperature of the two particles, calculated
m j ) are the respective temperatures and masses of the two colliding particles.
K0 and Eg are proportionality constants that can be estimated from experimental data on
particle agglomeration.
hydrodynamic behaviour of the polymer particles in the reactor (e.g., the mechanism of
dual particle collision, the diameters of the colliding particles Di and D j , etc.) and,
In this study, a simple agglomeration kernel1 that favours the collisions of small
with small particles was employed.
K ag K0 f Di ,D j K0 Di2 D 2j Di D j
4
(4.100)
Finally, one can define in terms of n p D,t , the probability density function
Pp D,t dD , which represents the mass fraction of particles in the size range D, D dD :
p D 3
Pp D,t dD n p D,t dD (4.101)
6
Pp D,t dD 1 (4.101)
Dmin
Chapter 4: Slurry-loop Reactor Modelling 197
The continuous form of the PBE (Eq. 4.92) can be solved using the orthogonal
collocation on finite elements (i.e., OCFE) method. Gelbard and Seinfeld, 1978 first
employed the OCFE method to solve the dynamic PBE. The method was successfully
applied to the solution of a class of pure aggregation problems and to the general PBE (Eq.
4.92) for short aggregation times. Recently, Nicmanis and Hounslow, 1998 employed a
collocation method to determine the steady state PSD in a continuous particulate process
undergoing combined particle growth and aggregation.
In the OCFE method, the particle volume domain is first divided into ‘ ne ’ elements
nc 1
nV V ,t nej t j t (4.102)
j 0
where nej denotes the value of nV V ,t at the ‘ j ’ internal or boundary collocation point.
Following the general developments of Finlayson, 1980, Eq. 4.92 is recast into a
system of ( ne nc ) residual equations corresponding to all the internal points of the ‘ ne ’
volume elements.
198 Chapter 4: Slurry-loop Reactor Modelling
nc 1
d j e dGV e
nie t n j t J GV Vi ni t
e
j 0 d i dV i
ne nc
nie t wkG J Vi e ,Vk f nkf t
f
f 1 k 1
g 1 nc (4.103)
wGj J Vi e V j f ,V j f nie nV Vi e V j f
f
f 1 j 1
Vie / 2
V V ,V nV V nV Vi e V dV
e
i
V1g
According to the standard finite element formulation, the global volume domain of
each element ‘ e ’ is linearly transformed to a local domain [−1, 1]. The index ‘ g ’ denotes
e f
the element containing Vi / 2 . J is the Jacobian of the volume transformation and wkG
are the integration weights of the Gauss–Legendre quadrature rule. At the boundary points
between the various elements, the number density function and its first derivative are
forced to be continuous. Thus, the following ne 1 continuity conditions between all the
ne nc
n01 t n01 t w j J V01 ,V j f nkf t
f
(4.104)
f 1 k 1
Since the total number of unknown nodal values of nej (i.e., ne nc 1 1 ), is less
than the total number of residual equations and continuity conditions (i.e.,
n n
e c 1 1 ), two additional equations are needed to produce a closed system of DAEs.
These equations correspond to the values of nej at the minimum Vmin ( e 1, j 0 ) and
residual equation similar to Eq. 4.103 can be written. On the other hand, at V Vmin , the
residual equation for the free boundary condition takes the following form:
Chapter 4: Slurry-loop Reactor Modelling 199
nc
d j nc
d j e 1
j n j t J
e e 1
n e
t J (4.105)
j 1 d j 1 d
nc1 0
In the present research study, the orthogonal collocation on finite elements method
(i.e., OCFE) was employed for solving the dynamic PBE (i.e., Eq. 4.92) (Alexopoulos et
al., 2004). A third order Lagrange polynomial is chosen to approximate the actual solution
within the integration domain. Note that the selected polynomial approximation satisfies
the numerical solution criteria on convergence and computational efficiency. The resulting
system of stiff, non-linear algebraic differential Eqs. 4.103–4.105 was solved using the
double precision Petzold-Gear BDF method (IMSL, routine DASPG).
200 Chapter 4: Slurry-loop Reactor Modelling
4.8. Notation
Symbols
Csorbed , am,i : Concentration of the sorbed molecular species ‘ i ’ in the amorphous part of
the solid polymer phase, mol/l
Csorbed , solid ,i : Concentration of the sorbed molecular species ‘ i ’ in the solid polymer
phase, mol/l
d : Diameter, m
e : Error signal
ef : Enhancement factor
fp : Volume fraction of the leg volume that consists of packed/ settled polymer
Chapter 4: Slurry-loop Reactor Modelling 201
frsolid ,i : Mass fraction of the molecular species ‘ i ’ in the solid polymer phase
l : Length, m
Nu : Nusselt number
P : Pressure, bar
p (t ) : Controller output
: Dimensionless pressure
P
Pr : Prandtl number
qout , j : Volumetric flowrate of molecular species ‘ i ’ stream leaving the reactor, l/s
202 Chapter 4: Slurry-loop Reactor Modelling
Re : Reynolds number
t : Time, sec
T : Temperature, K
: Dimensionless temperature
T
V : Volume, l
b : bulk
Chapter 4: Slurry-loop Reactor Modelling 203
c : Coolant
e : Environment
j : Referring to ‘ j th ’ reactor
L : Settling leg
mix : Mixture
R : Reactor
w : Wall
Greek letters
t f : Firing period, s
Kc : Controller gain
: Viscosity, Pa s
: Density, kg/l
204 Chapter 4: Slurry-loop Reactor Modelling
: Dimensionless density
D : Derivative time
I : Integral time
4.9. References
Alexopoulos A. H., Roussos A. I. and Kiparissides C., 2004, ‘Part I: Dynamic evolution of
the particle size distribution in particulate processes undergoing combined particle
growth and aggregation’, Chem. Eng. Sci., 24, 5751-5769.
Anderson D.R., 1966, ‘Thermal conductivity of polymers’, Chem. Rev, 66, pp. 677–690.
Arastoopour H., Huang C.H. and Weil S.A., 1988, ‘Fluidization Behaviour of Particles
Under Agglomerating Conditions’, Chem. Eng. Sci., 43, pp. 3063-3075.
Aston J.G., Kennedy R.M. and Schumann, S.C., 1940, ‘The heat capacity and entropy,
heats of fusion and vaporization and the vapor pressure of isobutane’, J. Am. Chem.
Soc., 62, pp. 2059-2063.
Ayres C.A., Scott Jr. J.N. and Shrek F.T., 1986, Sep. 23, ‘Loop Reactor Settling Leg
System for Separation of Solid Polymers and Liquid Diluent’, US Patent No.
4,613,484.
Barnea E. and Mizrahi J., 1976, ‘On the ‘Effective’ Viscosity of Liquid-Liquid
Dispersions’, Ind. Eng. Chem., 15, No. 2, pp. 120-125.
Batterham R.J., Hall J.S. and Barton G., 1981, ‘Pelletizing Kinetics and Simulation for
Full-Scale Balling Circuits. In Proceedings of the 3rd International Symposium on
Aggregation’, Nurnberg, West Germany, p A136.
Benham E.A., McDaniel M.P., McElvain R. and Schneider R.O., 1991, Dec. 10, ‘High-
temperature Slurry Polymerization of Ethylene’, US Patent No. 5,071,927.
Bird R.B., Stewart W.E. and Lightfoot E.N., 2002, ‘Transport Phenomena’, Second
Edition, John Wiley and Sons.
Brandrup J. and Immergut E.H., 1989, ‘Polymer Handbook’, third edition, John Wiley and
Sons, Inc.
Brodkey R.S. and Hersey H.C., 1988, ‘Transport Phenomena, a Unified Approach’,
McGraw Hill International Editions, Chemical Engineering Science.
Chang S-S., 1976, ‘Heat capacities of polyethylene IV. High molecular weight linear
polyethylene’, J. Res., NBS A80, pp. 51-57.
206 Chapter 4: Slurry-loop Reactor Modelling
Chang S-S., Westrum E.F., Jr. and Carlson, H.G., 1975, ‘Heat capacities of polyethylene
III. One linear and one branched sample from 5 to 350 K’, J. Res., NBS 79A, pp.
437-441.
Chao J., Hall K.R. and Yao J.M., 1983, ‘Thermodynamic properties of simple alkenes’,
Thermochim. Acta, 64(3), pp. 285-303.
Chatzi E.G. and Kiparissides C., 1992, ‘Dynamic Simulations of Bimodal Drop Size
Distributions in Low-coalescence Batch Dispersion Systems’, Chem. Eng. Sci., 47
(2), 445-456.
Chen C.H., Vermeychuk J.G., Howell J.A. and Ehrlich, P., 1976, ‘Computer Model for
Tubular High-Pressure Polyethylene Reactors’, AIChE Journal, 22,3, pp. 463-471.
Chen C.M., 1993, ‘Gas Phase Olefin Polymerization with Ziegler-Natta Catalysts’, PhD
Thesis, University of Wisconsin.
Choi K.Y., Zhao X. and Tang S., 1994, ‘Population Balance Modelling for a Continuous
Gas Phase Olefin Polymerization Reactor’, J. Appl. Polym., Sci., 53, pp. 1589-1594.
Coulson J.M. and Richardson J.F., 1999, ‘Chemical Engineering’, The Bath Press, Great
Britain.
Daupert T.E. and Danner R.P., 1985, ‘Data Compilation Tables of Properties of Pure
Compounds’, Design Institute for Physical Properties Data, AIChE: New York.
deMan, J.M., 1999, ‘Principles of Food Chemistry’, 3rd Edition, Springer, Maryland.
Doelder J.D., 2006, ‘Viscosity and Compliance from Molar Mass Distributions Using
Double Reptation Models’, Rheol. Acta, 46, 195.
Egan C.J. and Kemp J.D., 1937, ‘Ethylene. The heat capacity from 15°K to the boiling
point. The heats of fusion and vaporization. The vapor pressure of the liquid. The
entropy from thermal measurements compared with the entropy from spectroscopic
data’,J. Am. Chem. Soc., 59, pp. 1264-1268.
Ferrero M.A. and Chiovetta, M.G., 1990, ‘Preliminary Design of a Loop Reactor for Bulk
Propylene Polymerization’, Polym. Plast. Technol. Eng., 29 (3), pp. 263-287.
Finlayson B.A., 1992, ‘Numerical Methods for Problems with Moving Fronts’, Ravenna
Park Publ. Inc, Seattle.
Floyd S., Heiskanen T., Taylor T.W., Mann G.E. and Ray W.H., 1987, ‘Polymerization of
Olefins Through Heterogeneous Catalysis VI. Effect of Particle Heat and Mass
Transfer on Polymerization Behavior and Polymer Properties’, J. Appl. Polym. Sci.,
33, pp. 1021-1065.
Floyd S., Heiskanen T. and Ray W.H., 1988, ‘Solid Catalyzed Olefin Polymerization’,
Chem. Eng. Prog., 84, 11, pp. 56-62.
Fontes C.H. and Mendes M.J., 2005. ‘Analysis of an Industrial Continuous Slurry Reactor
for Ethylene–Butane Co-polymerization’, Polymer, 46, pp. 2922-2932.
Franklin ΙΙΙ, R.K., Hottovy J.D., Hensley H.D., Przelomski D.J., Cymbaluk T.H. and Perez
E.P., 2004, Jun. 1, ‘High Polymer Solids Slurry Polymerization Employing 1-Olefin
Comonomer’, US Patent No. 6,743,869 B2.
Gay E.C., Nelson P.A. and Armstrong, W.P., 1969, ‘Flow Properties of Suspensions with
High Solids Concentration’, AIChe Journal, 15, 6, pp. 815-822.
Geankoplis C.J., 1993, ‘Transport Processes and Unit Operations’, 3rd edition, Prentice
Hall.
Gelbard F. and Seinfeld J.H., 1979, ‘Exact Solution of the General Dynamic Equation for
Aerosol Growth by Condensation’, J. Colloid Interface Sci., 68 (1), pp. 173-183.
Ha K., Yoo K. and Rhee H., 2001, ‘Modeling and Analysis of a Slurry Reactor System for
Heterogeneous Olefin Polymerization: The Effects of Hydrogen Concentration and
Initial Catalyst Size’, J. Appl. Polym. Sci., 79, pp. 2480-2493.
208 Chapter 4: Slurry-loop Reactor Modelling
Hartel R.W. and Randolph A.D., 1986, ‘Mechanisms and Kinetic Modeling of Calcium
Oxalate Crystal Aggregation in a Urinelike Liquor. Part I: Mechanisms’, AIChE J.,
32, pp. 1176-1185.
Hatzantonis H., Goulas A. and Kiparissides C., 1998, ‘A comprehensive model for the
prediction of particle-size distribution in catalyzed olefin polymerization fluidized-
bed reactors’, Chem. Eng. Sci., 53, 18, pp. 3251-3267.
Hidy G.M., 1965, ‘On the Theory of the Coagulation of Noninteracting Particles in
Brownian Motion’, J. Colloid Sci., 20, 123-144.
Hottovy J.D., Hensley H.D., Przelomski D.J., Cymbaluk T.H., Franklin III R.K. and Perez
E.P., 2004, Oct. 19, ‘High Solids Slurry Polymerization Using Heat Exchange to
Condense the Flashed Diluent’, US Patent No. 6,806,324 B2.
Hounslow M.J., Ryall R.L. and Marshall V.R., 1988, ‘A Discretized Population Balance
for Nucleation, Growth, and Aggregation’, AIChE J., 34 (11), 1821-1832.
Hulburt H.M. and Katz S., 1964, ‘Some Problems in Particle technology. A Statistical
Mechanical Formulation.’, Chem. Eng. Sci., 19, pp. 555-574.
Hutchinson R.A., Chen C.M. and Ray W.H., 1992, ‘Polymerization of Olefins through
Heterogeneous Catalysis X: Modelling of Particle Growth and Morphology’, J. Appl.
Polym. Sci., 44, 1389-1414.
Kalinowska B. and Woycicki W., 1985, ‘Heat capacities and excess heat capacities of (an
alcohol + an unsaturated hydrocarbon). 1. (Propan-1-ol + n-hex-1-ene)’, J. Chem.
Thermodynam., 17, pp. 829-834.
Kanellopoulos V., Dompazis G., Gustafsson B. and Kiparissides C., 2004, ‘Compregensive
Analysis of Single-Particle Growth in Heterogeneous Olefin Polymerization: The
Random-Pore Polymeric Flow Model’, Ind. Eng. Chem. Res., 43, pp. 5166-5180.
Kanellopoulos V., Mouratides D., Pladis P. and Kiparissides C., 2006, ‘Prediction of
Solubility of α-olefins in Polyolefins Using a Combined Equation of State-Molecular
Dynamics Approach’, Ind. Eng. Chem. Res., 45 (17), pp. 5870–5878.
Chapter 4: Slurry-loop Reactor Modelling 209
Khare N.P., Seavey K.C., Liu Y.A., Ramanathan S. and Lingard, S., Chen, C., 2002,
‘Steady-State and Dynamic Mdeling of Commercial Slurry High-Density
Polyethylene (HDPE) Process’, Ind. Eng. Chem. Res., 41, pp. 5601-5618.
Kiparissides C., Alexopoulos A., Roussos A., Dompazis G. and Kotoulas C., 2004,
‘Population Balance Modeling of Particulate Polymerization Processes’, Ind. Eng.
Chem. Res., 43, pp. 7290-7302.
Kiparissides C., Baltsas A., Papadopoulos S., Congalidis J.P., Richads J.R., Kelly M.B.
and Ye Y., 2005, ‘Mathematical Modeling of Free-Radical Ethylene
Copolymerization in High Pressure Tubular Reactors’, Ind. Eng. Chem. Sci., 44, pp.
2592-2605.
Ko Y.S., Han T.K., Sadatoshi H. and Woo S.I., 1998, ‘Analysis of Microstructure of
Ethylene–1-Hexene Copolymer Prepared over Thermally Pretreated
MgCl2/THF/TiCl4 Bimetallic Catalyst’, Inc. J. Polym Sci A: Polym Chem, 36, pp.
291-300.
Krallis A., Pladis P., Kanellopoulos V. and Kiparissides C., 2010, ‘Development of
Advanced Software Tools for Computer-Aided Design, Simulation, and
Optimization of Polymerization Processes’, Macromolecular Reaction Engineering,
4, pp. 303-318.
Kufeld S.E, Reid T.A., Burns D.H., Verser D.W., Hensley H.D., Przelomski D.J.,
Cymbaluk T.H., Franklin R.K.,Perez E.P. and Hottovy J.D., 2006, Apr. 25, ‘Slurry
Polymerization Reactor Having Large Length/ Diameter Ratio’, US Patent No.
7,033,545 B2.
Kumar S. and Ramkrishna D., 1996a, ‘On the Solution of Population Balance Equations
by Discretizations I. A Fixed Pivot Technique’, Chem. Eng. Sci., 51 (8), 1311-1332.
Kumar S. and Ramkrishna D., 1996b, ‘On the Solution of Population Balance Equations
by Discretizations II. A Moving Pivot Technique’, Chem. Eng. Sci., 51 (8), 1333-
1342.
210 Chapter 4: Slurry-loop Reactor Modelling
Landgrebe J.D., Pratsinis S.E., 1990, ‘A Discrete-Sectional Model for Powder Production
by Gas-Phase Chemical reaction and Aerosol Coagulation in the Free-Molecular
Regime’, J. Colloid Interface Sci., 139 (1), pp. 63-86.
Li J., Zhang J., Ge W. and Liu X., 2004, ‘Multi-scale methodology for complex systems’,
Chem. Eng. Sci., 59, pp. 1687-1700.
Litster J.D., Smit D.J. and Hounslow M.J.,1995, ‘Adjustable Discretized Population
Balance for Growth and Aggregation’, AIChE J., 41, pp. 591-603.
Lonfils N., Bodart P. and Debras G., 2000, Aug. 1, ‘Catalysts for Polyethylene Production
and Use Thereof’, US Patent No. 6,096,679.
Marchal P., David R., Klein J.P. and Villermaux J., 1988, ‘Crystallization and Precipitation
Engineerings-I. An Efficient Method for Solving Population Balance in
Crystallization with Agglomeration’, Chem. Eng. Sci., 43 (1), 59-67.
Marechal P., 2006, Apr. 25, ‘Process for Producing Bimodal Polyethylene Resins’, US
Patent No. 7,034,092 B2.
Marwil S.J., 1966, Dec. 20,‘Withdrawal of Solids from a Flowing Stream Comprising a
Slurry of Same’, US Patent No. 3,293,000.
McCullough J.P., Finke H.L., Gross M.E., Messerly J.F. and Waddington G., 1957, ‘Low
temperature calorimetric studies of seven 1-olefins: effect of orientational disorder in
the solid state’, J. Phys. Chem., 61, pp. 289-301.
Moore S.J. and Wanke S.E., 2001, ‘Solubility of Ethylene, 1-butene and 1-hexene in
Polyethylenes’, Chem. Eng. Sci., 56, pp. 4121-4129.
Neto A.G. M., Freitas M.F., Nele M. and Pinto J.C., 2005, ‘Modeling Ethylene/1-Butene
Copolymerizations in Industrial Slurry Reactors’, Ind. Eng. Chem. Res., 44, pp.
2697-2715.
Chapter 4: Slurry-loop Reactor Modelling 211
Noll P., 2005, Jan. 27, ‘Process for Preparing Polyethylene’, US Patent No 2005/ 0020784
A1.
Novak A., Bobak M., KoseK J., Banaszak B.J., Lo D.,Widya T., Ray, W.H and de Pablo
J.J., 2006, ‘Ethylene and 1-Hexene Sorption in LLDPE under Typical Gas-Phase
Reactor Conditions’, Journal of Applied Polymer Science, 100, pp. 1124-1136.
Orbey H., Bokis C.P. and Chen, C.C., 1998, ‘Equation of State Modeling of Phase
Equilibrium in the Low Density Polyethylene Process: The Sanchez-Lacombe,
Statistical Associating Fluid Theory and Polymer-Soave-Redlich-Kwong Equations
of State’, Ind. Eng. Chem. Res., 37, 4481-4491.
Park G.S., Shomate C.H., Kennedy W.D. and Crawford B.L. Jr., 1937, ‘The entropies of n-
butane and isobutane, with some heat capacity data for isobutane’, J. Chem. Phys., 5,
pp. 359-363.
Passaglia E. and Kevorkian, H.K., 1963, ‘The heat capacity of linear and branched
polyethylene’, J. Appl. Poly. Sci., 7, pp. 119-132.
Pattamaprom C., Larson R.G., 2001, ‘Predicting the Linear Viscoelastic Properties of
Monodisperse and Polydisperse Polystyrenes and Polyethylenes’, Rheol. Acta, 40,
pp. 516-532.
Perry R. H. and Green, D., 1984, ‘Perry’s Chemical Engineer’s Handbook’, sixth edition,
McGraw-Hill International Editions, Chemical Engineering Series.
Pladis P., Kanellopoulos V., Chatzidoukas C. and Kiparissides C., 2008, ‘Effect of
Reaction Conditions and Catalyst Design on the Rheological Properties of
Polyolefins Produced in Gas-Phase Olefin Polymerization Reactors’, Macromol.
Theory Simul., 17, pp. 478-487.
Quijada R., Scipioni R.B., Mauler R.S., Galland G.B. and Miranda M.S.L., 1995,
‘Synthesis and Characterization of Ethylene-1-hexene Copolymers Using
Homogeneous Ziegler-Natta Catalysts’, Polymer Bulletin, 35, pp. 299-306.
212 Chapter 4: Slurry-loop Reactor Modelling
Quijada R., Rojas R., Mauler R.S. Galland G.B. and Scipioni R.B., 1997, ‘Study of the
Effect of the Monomer Pressure on the Copolymerization of Ethylene with 1-
Hexene’, Inc. J Appl Polym Sci, 64, pp. 2567-2574.
Randolph A. D., Larson M. A., 1971, ‘Theory of Particulate Processes: Analysis and
Techniques of Continuous Crystallization’, Academic Press, N.Y.
Reginato A.S., Zacca J.J. and Secchi A.R., 2003, ‘Modeling and Simulation of Propylene
Polymerization in Nonideal Loop Reactors’, AIChe Journal, 49, No. 10, pp. 2642-
2654.
Reid R.C., Prausnitz J.M. and Poling B.E., 1986, ‘The Properties of Gases and Liquids’,
Fourth edition, McGraw-Hill international Editions.
Salmon E.J., 2003, May 20, ‘Continuous Recovery of Polymer from a Slurry Loop
Reactor’, US Patent No. 6,566,460 B1.
Sastry K.V.S. and Gaschignard P., 1981, ‘Discretization Procedure for the Coalescence
Equation of Particulate Processes’, Ind. Eng. Chem. Fundam., 20, pp. 355-361.
Scoggin J.S., 1966, Mar. 22, ‘Method and Apparatus for the Recovery of Solid Olefin
Polymer From a continuous Path Reaction Zone’, US Patent No. 3,242,150.
Chapter 4: Slurry-loop Reactor Modelling 213
Seborg D.E., Edgar T.F. and Mellichamp D.A., 1989, ‘Process Dynamics and Control’,
Wiley Series in Chemical Engineering, John Wiley and Sons, USA, Chapter 8, pp.
183-198, Chapter 13, pp. 294-309, Chapter 26, pp. 614-616.
Schwarzl F.R., 1971, ‘Numerical Calculation of Storage and Loss Modulus from Stress
Relaxation Data for Linear Viscoelastic Materials’, Rheol. Acta, 10, pp. 165-173.
Soares J.B.P. and Hamielec A.E., 1995 ‘Effect of Residence Time Distribution on the Size
Distribution of Polymer Particles Made With Heterogeneous Ziegler-Natta and
Supported Metallocene Catalysts. A Generic Mathematical Model’, Macromol.
Theory Simul., 4, pp. 1085-1104.
Touloupides V., Kanellopoulos V., Pladis P., Kiparissides C., Mignon D. and Van-
Grambezen P., 2010a, ‘Modeling and Simulation of an Industrial Slurry-phase
Catalytic Olefin Polymerization Reactor Series’, Chem. Eng. Sci., 65, pp. 3208-3222.
Touloupides V., Kanellopoulos V., Pladis P. and Kiparissides C., 2010b, ‘Modeling and
Simulation of Particle Size Distribution in Slurry-Phase Olefin Catalytic
Polymerization Industrial Loop Reactors’, Computer Aided Chemical Engineering,
28, pp. 43-48.
Ullmann’s Encyclopedia of Industrial Chemistry, 6th edition, 2002, John Wiley & Sons.
Uvarov, B.A. and Tsevetkova V.I., 1974, ‘Development of a Mathematical Model for
Controlling the Yield of Propylene Polymerization in Loop Reactors’, Polim.
Protsessy Appar. 165.
Van Ruymbeke E., Keunings R., Bailly C., 2002a, ‘Determination of the Molecular
Weight of Entagled Linear Polymers from Linear Viscoelasticity Data’, J. Non-
Newtonian Fluid Mech., 105, pp. 153-175.
Van Ruymbeke E., Keunings R., Stephenne V., Hagenaars A. and Bailly C., 2002b,
‘Evaluation of Reptation Models for Predicting the Linear Viscoelastic Properties of
Entangled Linear Polymers’, Macromolecules, 35, pp. 2689-2699.
214 Chapter 4: Slurry-loop Reactor Modelling
Villar M.A., Failla M.D., Quijada R., Mauler R.S., Valles E.M., Galland G.B. and
Quinzani L.M., 2001, ‘Rheological Characterization of Molten Ethylene-a-olefin
copolymers synthesized with Et[Ind]2ZrCl2/MAO catalyst’, Polymer, 42, pp. 9269-
9279.
Wang Z., Tham M.T. and Morris A.J., 1992, ‘Multilayer feedforward neural networks:
Canonical form approximation of nonlinearity’, Int. J. Control, 56, pp. 655-672.
Weimin Z., Weiduan S. and Hongsi W., 1991, ‘Modeling and Analyis of Propylene Loop
Reactor-Steady State Behavior and Operable Region’, Chemical Reaction
Engineering and Technology, 7, pp. 68-76.
Wunderlich B., 1962,b, ‘Motion in polyethylene. III. The amorphous Polymer’, J. Chem.
Phys., 37, pp. 2429-2432.
Xie T.Y., Hamielec A.E., Wood P.E. and Woods, D.R., 1991, ‘Experimental Investigation
of Vinyl Chloride Polymerization at High Conversion: Mechanism, Kinetics and
Modeling’, Polymer, 32(3), pp. 537-557.
Yiannoulakis H., Yiagopoulos A. and Kiparissides C., 2001, ‘Recent developments in the
particle size distribution modeling of fluidized-bed olefin polymerization reactors’,
Chem. Eng. Sci., 56, pp. 917-925.
Yiagopoulos A., Yiannoulakis H., Dimos V. and Kiparissides C., 2001, ‘Heat and Mass
Transfer Phenomena During the Early Growth of a Catalyst Particle in Gas-phase
Olefin Polymerization: the Effect of Prepolymerization Temperature and Time’,
Chem. Eng. Sci., 56, pp. 3979-3995.
Zacca, J.J. and Ray, H. W., 1993, ‘Modelling of the liquid phase polymerization of olefins
in loop reactors’, Chem. Eng. Sci., Vol. 48, No. 22, pp. 3743-3765.
Zacca J.J., Debling J.A. and Ray W.H., 1996, ‘Reactor Residence Time Distribution
Effects on the Multistage Polymerization of Olefins-I. Basic Principles and
Illustrative Examples, Polypropylene’, Chem. Engng Sci., 51, pp. 4859-4886.
Zoller P. and Walsh D., 1995, ‘Standard pressure-volume-temperature data for polymers',
Technomic publishing, Lancaster, Basel, pp. 36-37.
Chapter 5: Results and Discussion 215
Chapter 5
5.1. Introduction
Model tuning has been conducted in terms of plant data, provided by our industrial
partner. Model input variables comprises ethylene, 1-hexene and hydrogen inflow rates
policies, reactor temperature, pressure, solids concentration and ethylene off-gas
composition (regarding first reactor operation) setpoint policies as well as the policy on the
number of settling legs that will be in operation. It should be pointed out that in all
different simulation runs, the same set of kinetic, transport and other model parameters
were utilized.
216 Chapter 5: Results and Discussion
The volume of each reactor in the cascade-loop reactor configuration was assumed
to be equal to 45 m3 (i.e., tube diameter and tube length were assumed equal to 0.48 m and
248 m, respectively). The volume of each settling leg was set equal to 0.07 m3. Finally, for
the stable as well as dynamic operation of the simulated industrial slurry-phase cascade-
loop reactor series, the simulation of the various PID feedback controllers, shown in Figure
4.9, was included. In particular, the dynamic process model included the simulation of the
two temperature controllers, one ethylene mass fraction controller, two polymer solids
concentration controllers and two reactor pressure controllers.
To simulate a typical startup policy for the cascade-loop reactor series, the ethylene
inflow rates as well as the number of settling legs in operation in the first and second
reactor were assumed to follow the discrete dynamic profiles shown in Figure 5.1. That is,
at time t 0 , the two loop reactors were assumed to be filled with iso-butane and known
concentrations of ethylene and hydrogen. In fact, the initial mass fractions of ethylene and
hydrogen in the first and second loop reactor were assumed to be equal to 0.6 and 0.14 and
0.0 and 0.0, respectively. The initial pressure in each reactor was assumed to be equal the
respective set-point value of the two pressure controllers. The mass flowrate ratios of (1-
hexene to ethylene) and (hydrogen to ethylene) in the feed stream were set equal to 0.07
and 1.15 10-5 in the first reactor and 0.0 and 1.5 10-3 in the second reactor, respectively.
Chapter 5: Results and Discussion 217
It should be pointed out that during the process startup, the setpoint values of the
temperature controllers in the first and the second reactor of the series were set equal to
T 85o C and T 95o C , respectively. In addition, the setpoint values of the ethylene mass
fraction, solids concentration and reactor pressure controllers were kept constant and equal
to typical values employed in industrial slurry-phase cascade-loop reactors (Marechal,
2006, Noll, 2005).
218 Chapter 5: Results and Discussion
a.
12 tn/hr
10.5 tn/hr
Ethylene Inflow Rate
1.5 tn/hr
6
5 Settling Legs in Operation
4
Startup
3
First reactor
2
b.
12 tn/hr
10.5 tn/hr
Etylene Inflow Rate
0 tn/hr
6
Settling Legs in Operation
4 Startup
Second reactor
2
0 10 20 30 40
Time, hr
Figure 5.1: Dynamic evolution of the ethylene inflow rate and number of settling legs in
the first (a.) and second (b.) loop reactor of the series during process startup.
Chapter 5: Results and Discussion 219
In Figure 5.2 and Figure 5.3, the dynamic evolution of the pressure in the first and
second reactor is depicted. According to the semi-continuous operation of the settling legs,
each time that the operating pressure in the loop reactor exceeds an upper limit (i.e., see
dashed line in Figure 5.2 and Figure 5.3), the discharge valve of a settling leg opens so its
polymer-solids content is emptied from the leg. This results in a decrease of the pressure in
the loop reactor (i.e., lower limit in Figure 5.2 and Figure 5.3) that is followed by the
closing of the discharge valve of the settling leg. Subsequently, as fresh monomer(s) and
diluent are continuously fed into the reactor and new polymer is formed, the reactor
pressure gradually reaches again its maximum allowable value. Thus, the discharge valve
of the next settling leg opens. This on-off valve opening of the sequence of the settling legs
results in a periodic variation (i.e., ‘zig-zag’) of the pressure in the loop reactor (see Figure
5.2 and Figure 5.3).
45
45
First reactor
44
Reactor Pressure, bar
43
44 Upper limit
42
Set point
41
Lower limit
40
Reactor Pressure, bar
43 39
46900 46920 46940 46960 46980 47000
Time, sec
42
41
40
39
0 10 20 30 40
Time, hr
Figure 5.2: Dynamic evolution of the operating pressure in the first reactor.
220 Chapter 5: Results and Discussion
44 44
Second reactor 43
43 Upper limit 41
Set point
Dimensionless Reactor Pressure
40
Lower limit 39
42 38
46800 46840 46880 46920 46960 47000
Time, sec
41
40
39
38
0 10 20 30 40
Time, hr
Figure 5.3: Dynamic evolution of the operating pressure in the second reactor.
Figure 5.4 illustrates the dynamic evolution of the ethylene off-gas composition (i.e.,
% w/w ethylene mass fraction in the reactor-phase, except for the polymer) and the catalyst
inflow rate into the first reactor of the series. It should be pointed out that the ethylene
mass fraction in the reactor-phase in the first reactor is controlled by manipulating the
catalyst inflow rate. As can be seen, after approximately 30 hours of dynamic operation,
the reactor reaches its final steady-state value with respect to the ethylene concentration.
Note that the total ethylene concentration in the reactor-phase (i.e., liquid phase and
swollen polymer particles) is a very important process variable since high ethylene
concentrations can lead to the formation of a separate gas-phase in the reactor with
undesired effects regarding the reactor operability (i.e., appearance of a stagnant ethylene
gas phase, insufficient recirculation rate of the reaction mixture, appearance of cavitation
phenomena, etc.). On the other hand, there is no controller for the ethylene off-gas
composition in the second reactor (as there is no fresh catalyst feed in the second reactor).
Figure 5.5 illustrates the dynamic evolution of the ethylene off-gas composition in the
second reactor. As can be seen, ethylene off-gas composition reaches higher values
compared to the ones in the first reactor. Moreover, a delay in the polymerization rate,
leading to an increase of the ethylene off-gas composition is observed due to the fact that
while the reactor is fed with fresh ethylene and un-reacted ethylene, coming from the first
reactor, there is not enough catalyst in the reactor to propagate the polymerization.
Chapter 5: Results and Discussion 221
1.5 150
First reactor
0.5 50
Manipulated Variable
0.0 0
0 10 20 30 40
Time, hr
Figure 5.4: Dynamic evolution of the ethylene mass fraction and catalyst inflow rate in the
first reactor.
4
Second reactor
Ethylene Off-gas Composition, % w/w
0
0 10 20 30 40
Time, hr
Figure 5.5: Dynamic evolution of the ethylene mass fraction and catalyst inflow rate in the
second reactor.
222 Chapter 5: Results and Discussion
The dynamic evolution of the 1-hexene and hydrogen off-gas composition in the first
and second reactor of the series are illustrated in Figure 5.6 and Figure 5.7, respectively. It
should be pointed out that the 1-hexene and hydrogen mass fractions reach their final
steady-state values after approximately 15 hours of operation.
The dynamic evolution of the solids concentration and iso-butane inflow rate (i.e.,
manipulated variable) in the first and second reactor are depicted in Figure 5.8 and Figure
5.9, respectively. As can be seen, the solids concentration controller is inactive during the
first 6 hours of the startup because of the very low concentration of solids in the reactor. It
should be pointed out that, based on the selected reactor startup policy, after about 30
hours of operation the solids concentration in the first reactor reaches its final steady-state
value.
As has already been mentioned, in the reactor, two phases co-exist, namely a liquid
phase (including isobutene, ethylene, 1-hexene and hydrogen) and a solid-polymer phase
(including polymer and sorbed quantities of iso-butane, ethylene, 1-hexene and hydrogen).
The time evolution of the reactor, polymer and liquid phase density in the first and second
reactor are depicted in Figure 5.10 and Figure 5.11, respectively. As can be seen, regarding
the first reactor, the reactor-phase density at t 0 is equal to the density of the pure liquid
phase (i.e., liquid 440 kg / m3 ). However, as the solids concentration increases, the
density of the reaction mixture (i.e., liquid plus solids) in the first reactor reaches its final
value (i.e., b 640 kg / m3 ).
Chapter 5: Results and Discussion 223
5 0.04
First reactor
0.02
0.01
1
0 0.00
0 10 20 30 40
Time, hr
Figure 5.6: Dynamic evolution of the 1-hexene and hydrogen off-gas composition in the
first reactor.
4.0 3
Second reactor
3.5
3.0
2
2.5
2.0
1.5
1
1.0
0.5
0.0 0
0 10 20 30 40
Time, hr
Figure 5.7: Dynamic evolution of the hydrogen mass fraction in the first reactor.
224 Chapter 5: Results and Discussion
60 20
First reactor
40
Controlled Variable
30 10
20
Manipulated Variable
5
10
-3
0 0
0 10 20 30 40
Time, hr
Figure 5.8: Dynamic evolution of solids concentration and iso-butane inflow rate in the
first reactor.
60 15
Second reactor
50
Controlled Variable
40 10
30
20 5
Manipulated Variable
10
-3
0 0
0 10 20 30 40
Time, hr
Figure 5.9: Dynamic evolution of the solids concentration and iso-butane inflow rate in
the second reactor.
Chapter 5: Results and Discussion 225
900
First reactor
800
Reactor phase
3
Density, kg/m
600
500
400
0 10 20 30 40
Time, hr
900
800
Second reactor
Slury phase
3
Density, kg/m
600
500
400
0 10 20 30 40
Time, hr
Figure 5.12 depicts the dynamic evolution of the solids enhancement factor as well
as the valve opening cycling period of the settling legs in the first reactor during process
startup. It is apparent that the value of the enhancement factor (i.e., ef out s )
decreases exponentially to its final steady-state value (i.e., e f 1.25 ). This means that the
solids volume fraction in the product stream is about 25% higher than that in the reactor-
phase. Notice that during the process startup (i.e., the startup period) the number of settling
legs in operation changes from 2 to 6 (see Figure 5.1). At the same time, the valve opening
cycling period changes from about 100 seconds at the beginning of the startup to about 48
seconds at the final reactor steady-state. In Figure 5.13 the dynamic evolution of the solids
enhancement factor as well as the valve opening cycling period of the settling legs in the
second reactor during process startup is depicted. As can be seen, at t 10hours , there is
an abrupt increase in both enhancement factor and valve opening cycling period due to the
corresponding increase f the number of settling legs in operation (i.e., from 4 to 6, see
Figure 5.1).
3.0 100
First reactor
60
2.0
40
1.5
20
1.0 0
0 10 20 30 40
Time, hr
Figure 5.12: Dynamic evolution of the solids enhancement factor and valve opening
cycling period in the first reactor.
Chapter 5: Results and Discussion 227
6.0 120
Second reactor
5.5
4.5
Enhancement Factor
80
4.0
3.5 60
3.0
40
2.5
2.0
20
1.5
1.0 0
0 10 20 30 40
Time, hr
Figure 5.13: Dynamic evolution of the solids enhancement factor and valve opening
cycling period in the second reactor.
100
80
75
0 10 20 30 40
Time, hr
Figure 5.14: Dynamic evolution of temperature in the first and second reactor.
80 8
First reactor
Coolant Inlet Temperature, C
o
70
ΔT Coolant, C
4
o
60
Manipulated Variable 2
50 0
0 10 20 30 40
Time, hr
Figure 5.15: Dynamic evolution of the coolant’s inlet temperature and the temperature
difference between the coolant outlet and inlet values in the first reactor.
Chapter 5: Results and Discussion 229
100 8
Second reactor
o
Coolant Inlet Temperature, C 90
6
DT Coolant, C
80
70
o
Manipulated Variable 2
60
50 0
0 10 20 30 40
Time, hr
Figure 5.16: Dynamic evolution of the coolant’s inlet temperature and the temperature
difference between the coolant outlet and inlet values in the second reactor.
In Figure 5.17 the dynamic evolution of polymer production rate at the exit of the
first and second reactor of the series is depicted. As can be seen, polymer production rate
follows ethylene feed rate, exhibiting high conversion values (i.e., higher than 95%).
In Figure 5.18, the variation of the polyolefin density at the exit of the first and
second reactor (i.e., final product density, free of any sorbed solvent) is shown. As
expected, polymer produced in the first reactor exhibits lower values of density due to the
higher co-monomer content in the reactor’s feed. In the inserted figure, the polymer
crystallinity is presented.
230 Chapter 5: Results and Discussion
30
First reactor
Second reactor
25
Polymer Production Rate, tn/hr
20
15
10
0
0 10 20 30 40
Time, hr
Figure 5.17: Dynamic evolution of polymer production rate at the exit of the first and
second reactor of the series.
0.96
First reactor
Second reactor
Product Density, kg/lt
64
62
60
58
56
54
Crystallinity
0.94
52
First reactor
50
Second reactor
48
46
44
42
40
38
36
0 10 20 30 40
Time, hr
0.92
0 10 20 30 40
Time, hr
Figure 5.18: Dynamic evolution of product density at the exit of the first and second
reactor of the series.
Chapter 5: Results and Discussion 231
In Figure 5.19 and Figure 5.20 the dynamic evolution of the average molecular
weight and polydispersity at the exit of the first and second reactor is presented. Moreover,
in Figure 5.21 the dynamic evolution of the MWD of the polymer produced in the first
(Figure 5.21a.) and in the second reactor (Figure 5.21b.) and overall MWD at the exit of
the second reactor of the series (Figure 5.21c.) is depicted. As can be seen, the MWDs
produced in the first and the second reactor are relatively narrow. On the other hand, the
overall MWD at the exit of the second reactor is broader reflecting the formation of a low
molecular weight polymer fraction due to the high hydrogen concentration in the second
reactor. The MWD of the polymer produced by all different catalyst site types in the first
and the second reactor, regarding steady-state (i.e., at t 14hours ), is presented, in the
inserted figures of Figure 5.21a. and Figure 5.21b., respectively.
1.0
-5
First reactor
Number Average Molecular Weight, Mn, x10
Second reactor
0.8
0.6
0.4
0.2
0.0
0 10 20 30 40
Time, hr
Figure 5.19: Dynamic evolution of average molecular weight at the exit of the first and
second reactor of the series.
232 Chapter 5: Results and Discussion
35
30
25
Polydispersity
20
First reactor
Second reactor
15
10
0 10 20 30 40
Time, hr
Figure 5.20: Dynamic evolution of polydispersity at the exit of the first and second reactor
of the series.
Chapter 5: Results and Discussion 233
a. 0.8 0.8
2 hours Overall
Catalyst site type 1
Catalyst site type 2
dW/dlog(MW)
14 hours 0.4
0.6 0.2
Polymer produced
in the first reactor 0.0
dW/dlog(MW)
2 3 4 5 6 7 8
log(MW)
0.4
Time
0.2
0.0
b. 0.8 0.8
Overall
2 hours Catalyst site type 1
Catalyst site type 2
dW/dlog(MW)
14 hours 0.4
0.6
0.2
dW/dlog(MW)
0.0
2 3 4 5 6 7 8
log(MW)
0.4
Time
0.2
Polymer produced
in the second reactor
0.0
c. 0.8
2 hours Polymer at the exit
8 hours of the second reactor of the series
14 hours
0.6
dW/dlog(MW)
0.4 Time
0.2
0.0
2 3 4 5 6 7 8
log(MW)
Figure 5.21: Dynamic evolution of the MWD of the polymer produced in the first (a.) and
in the second reactor (b.) and overall MWD at the exit of the second reactor of the series
(c.) (startup case).
234 Chapter 5: Results and Discussion
In the present study, the complex viscosity of polyolefins produced in the slurry-
phase cascade-loop reactor series was calculated by employing a rheological model based
on the reptation and Rouse relaxation theories (Pladis et. al., 2008). The application of this
model enables the prediction of the rheological behaviour of polymer melts (i.e., complex
viscosity versus frequency) for linear polymer chains in terms of their respective MWD.
Figure 5.22 and Figure 5.23 depict the calculated rheological curves of the polymer
produced in the first reactor and polymer at the exit of the second reactor of the series,
respectively. It is evident that, the complex viscosity in the first reactor is higher than that
in the second reactor reflecting the higher molecular weight values of the polymer
produced in the first reactor. Moreover, the transition from the Newtonian plateau to the
shear thinning behaviour is more abrupt for narrow MWDs (i.e., see discrete lines in the
inserted figure in Figure 5.22).
In the following sections of the research project, the effect of the total number of
settling legs in operation, the effect of the co-monomer inflow rate and the effect of the
hydrogen inflow rate on the slurry-loop reactor series process is studied. It should be noted
that the simulations regarding all those effects investigation, begin at t 0hours ,
corresponding to the end of steady-state simulation, studied at 5.2.1. Process startup.
Chapter 5: Results and Discussion 235
6
10
2 hours
6 hours
14 hours
Complex Viscosity, η, Pa.s 5
10
4
10
3
10
2
10
1E-3 0.01 0.1 1 10 100 1000
Frequency, 1/sec
Figure 5.22: Dynamic evolution of the complex viscosity of the polymer produced in the
first reactor (startup case).
6
10
2 hours
8 hours
14 hours
5
Complex Viscosity, η, Pa.s
10
4
10
3
10
2
10
1E-3 0.01 0.1 1 10 100 1000
Frequency, 1/sec
Figure 5.23: Dynamic evolution of the complex viscosity of the polymer at the exit of the
second reactor (startup case).
236 Chapter 5: Results and Discussion
It is well known that the selection of the number of settling legs in operation is
directly related with the overall process performance and product quality. For example, a
decrease in the number of settling legs in operation may occur due to either problematic
operation of settling legs (e.g., settling leg fouling due to extensive particle agglomeration,
etc.) or settling legs maintenance. In such a case, the non-operating leg is directly set out of
order and, thus, the reactor operates with a reduced number of settling legs.
In Figure 5.24, the dynamic evolution of the enhancement factor as well as of the
reactor valve opening cycling period (i.e., time period between two consecutive valve
openings of the same leg) are depicted. In this simulation, the number of settling legs in
operation in the second loop reactor of the series changes from 6 to 5 at time t 10hours
and finally to 4 at t 20hours . It can be seen that as the number of settling legs in
operation decreases the valve opening cycling period decreases. This results in a decrease
of the solids enhancement factor since the time period available for solids settling is
reduced. This in turn leads to a significant increase in the iso-butane inflow rate (see Figure
5.25) caused by the higher iso-butane removal rate (i.e., low values of e f ).
1.4 100
60
1.2
40
1.1
6
20
5
4
Settling Legs in Operation
1.0 0
0 10 20 30
Time, hr
Figure 5.24: Dynamic evolution of the solids enhancement factor and valve opening
cycling period in the second reactor.
Chapter 5: Results and Discussion 237
48 12
42
8
Manipulated Variable
40
-3
38 6
0 10 20 30
Time, hr
Figure 5.25: Dynamic evolution of the solids concentration and iso-butane inflow rate in
the second reactor.
In Figure 5.26, the dynamic evolution of the number average molecular weight and
polydispersity index of the polymer produced in the second reactor of the series are
depicted. It is apparent that as the number of settling legs in operation decreases, the
number average molecular weight increases while the polydispersity index decreases due
to the increase of the ethylene concentration in the reactor (i.e., in both liquid and polymer
phase, see inserted figure in Figure 5.26). This can be attributed to the decrease of that the
ethylene conversion as a result of the decrease of the reaction mixture residence time (due
to the increase of the iso-butane inflow rate, see Figure 5.25).
In summary, the number of settling legs in operation affects both the reactor
performance (i.e., the iso-butane inflow rate in the second reactor changes from 7.5 x 103
kg/hr to 10.0 x 103 kg/hr when the number of settling legs in operation changes from 6 to
4, see Figure 5.25) as well as the molecular properties of the polyolefin (see Figure 5.26).
238 Chapter 5: Results and Discussion
1.0 32
-4
1.05
0.95
0.80 30
0.75
0.70
Polydispersity
0 10 Time, hr 20 30
0.9 28
26
0.8 24
0 10 20 30
Time, hr
Figure 5.26: Dynamic evolution of the number average molecular weight and
polydispersity index in the second reactor.
In order to change the density of the final polymer product, a step change in the 1-
hexene inflow rate in the first reactor of the series was introduced. Figure 5.27 presents the
dynamic evolution of the co-polymer composition (i.e., in 1-hexene) in the first and second
reactor of the series. In this simulation, a transition from a polymer grade having a co-
polymer content equal to 0.5 % mol, at the exit of the second reactor of the series, to a
polymer grade with a co-polymer content equal to 0.95 % mol is depicted. Moreover, the
dynamic evolution of the co-polymer composition at the exit of the first reactor of the
series is illustrated (see dotted line in Figure 5.27). In the inserted figure, the time variation
of the mass fractions of 1-hexene in the liquid and polymer phase in the first reactor are
depicted, during the above grade transition.
Chapter 5: Results and Discussion 239
2
First reactor
Second reactor
12
First reactor
0
0 10 20 30
Time, hr
0
0 10 20 30
Time, hr
Figure 5.27: Dynamic evolution of the co-polymer composition in the first and second
reactor of the series.
In Figure 5.28, the variation of the polyolefin density in the first and second reactor
(i.e., final product density, free of any sorbed solvent) as well as the applied step-change in
the 1-hexene inflow rate (i.e., from 800 kg/hr to 1600 kg/hr) are depicted with respect to
the polymerization time. It is apparent that the product density at the exit of the second
reactor of the series is higher than that in the first reactor due to the lower co-monomer
incorporation rate in the second reactor. It should also be noted that the product density
reaches its new steady-state value in about 5 hours after the introduction of the step change
in the 1-hexene inflow rate. Note that, in the first reactor of the present slurry-phase
cascade-loop reactor series, low crystalline polymer (e.g., 35-40%) is produced due to the
higher co-monomer concentration, while, in the second reactor of the series, high
crystalline polymer (e.g., 55-60%) is produced due to the lower co-monomer incorporation
rate.
240 Chapter 5: Results and Discussion
0.96 5000
4000
3000
0.92 2000
1000
0.90
0 10 20 30
Time, hr
Figure 5.28: Dynamic evolution of the product density in the first and second reactor of
the series.
In this case, a step change in the hydrogen inflow rate was introduced to change the
final MWD of the polyolefin. More specifically, the hydrogen to ethylene mass flowrate
ratios in the first and second reactor of the series changed from 1.0 10-5 to 1.0 10-4 and
from 6.0 10-4 to 4.0 10-3, respectively. It is well known that, in Z-N catalytic olefin
polymerization, the hydrogen concentration is employed to control the molecular weight of
the polyolefin. In Figure 5.29 and Figure 5.30, the effect of the hydrogen concentration on
the dynamic evolution of the MWDs of polyolefin produced in the first and the second
reactor of the series are depicted. It is evident that as the hydrogen to ethylene mass
flowrate ratio increases (i.e., the hydrogen concentration increases), the MWD shifts to
lower molecular weights. The inserted figures in Figure 5.29 and Figure 5.30 show the
corresponding rheological curves during the imposed transition in the hydrogen inflow
rate. It is apparent that higher molecular weight polyolefins exhibit higher complex
viscosities and vice versa.
Chapter 5: Results and Discussion 241
0.8 10
6
Initial
Initial 2 hours
dW/dlog(MW) 0.6 10
3
2
10
1E-3 0.01 0.1 1 10 100 1000
Frequncy, 1/sec
0.4 Time
0.2
0.0
2 3 4 5 6 7 8
log(MW)
Figure 5.29: Dynamic evolution of the MWD and complex viscosity in the first reactor
(effect of hydrogen concentration).
0.8 6
10
Initial Initial
2 hours
Complex Viscocity, η, Pa.s
4 hours
2 hours
5
10
4 hours 10
4
0.6 10
3
2
10
dW/dlog(MW)
0.4
Time
0.2
0.0
2 3 4 5 6 7 8
log(MW)
Figure 5.30: Dynamic evolution of the MWD and complex viscosity in the second reactor
(effect of hydrogen concentration).
242 Chapter 5: Results and Discussion
The model developed under this research project was employed for the modelling of
an industrial slurry-loop reactor. The predicting capability of the model has been tested by
simulating the actual reactor operation and comparing model’s predictions to plant
measurements. Model predictions were compared with actual plant data as provided by our
industrial partner. The following representative results correspond to a typical reactor start
up and a whole month operation which includes a grade transition. It should be noted that
the model has been tested over different operating conditions for the production of
different polymer grades, exhibiting great accuracy in all cases. Finally, it should be
mentioned that all results are presented scaled due to confidentiality issues.
0.8
First reactor, model prediction Grade 1
First reactor, plant data
Second reactor
Second reactor, plant data
dW/dlog(MW) 0.6
0.4
0.2
0.0
1 2 3 4 5 6 7 8
log(MW)
Figure 5.31: Molecular weight distribution of polymer at the exit of the first and second
reactor of the series, regarding a low-co-monomer content polymer grade.
0.8
First reactor, model prediction Grade 2
First reactor, plant data
Second reactor
Second reactor, plant data
0.6
dW/dlog(MW)
0.4
0.2
0.0
1 2 3 4 5 6 7 8
log(MW)
Figure 5.32: Molecular weight distribution of polymer at the exit of the first and second
reactor of the series, regarding a medium-co-monomer content polymer grade.
244 Chapter 5: Results and Discussion
0.8
First reactor, model prediction Grade 3
First reactor, plant data
Second reactor
Second reactor, plant data
0.6
dW/dlog(MW)
0.4
0.2
0.0
1 2 3 4 5 6 7 8
log(MW)
Figure 5.33: Molecular weight distribution of polymer at the exit of the first and second
reactor of the series, regarding a high-co-monomer content polymer grade.
Model’s capability to predict the dynamic response of the two loop-reactor series
process is tested over a typical startup of the reactor series. In Figure 5.34-Figure 5.36 the
reactor first reactor’s feed policies regarding ethylene, hexane and hydrogen are presented.
As can be seen, reactor feed is gradually increased to reach steady state operation (in terms
of polymer production rate) after 40 hours of operation. In Figure 5.37 the iso-butane feed
is depicted. It is noticed that iso-butane feed is manipulated in order to control reactor
solids concentration (see Figure 5.42). Manipulated catalyst feed is presented in Figure
5.38. As can be seen in Figure 5.39-Figure 5.42, the model is capable of accurately
predicting reactor composition, regarding ethylene, hexane, hydrogen and solids
concentration, leading to an accurate calculation of reactor phase density (Figure 5.43).
The dynamic evolution of temperature, coolant’s inlet temperature into the reactor’s jacket
and the time variation of the temperature difference between the coolant outlet and inlet
values can be predicted by the model (see Figure 5.44-Figure 5.46). The settling leg model
Chapter 5: Results and Discussion 245
presented in this research project is able to calculate the dynamic evolution of the valve
opening cycling period presented in Figure 5.47. Finally, the dynamic evolution of
polymer production rate and polymer density (i.e., final product density, free of any sorbed
solvent) is presented in Figure 5.48 and Figure 5.49, respectively. In Figure 5.50-Figure
5.63, simulation results regarding the second reactor of the series are presented.
The dynamic evolution of polymer density for the first and second reactor is
presented in Figure 5.49 and Figure 5.63, respectively. Moreover, the dynamic evolution of
number average molecular weight and polydispersity for the first and second reactor is
presented in Figure 5.64 and Figure 5.65, respectively. As can be seen in Figure 5.64, for
the first 4 hours of reactor series operation, the polymer produced in both reactors exhibit
the same number average molecular weight. This is attributed to the fact that the second
reactor is not fed with fresh catalyst but with catalyst coming from the first reactor outflow
and ‘live’ polymer which continues to polymerize. It is noted that the transition time for
polymer production with the defined end-use properties in terms of number average
molecular weight, polydispersity and density does not exceed 15 hours.
1.2
Startup
1.0
Scaled Ethylene Feed, Kg/hr
0.8
0.6
0.4
First reactor
0.2
Plant Data
Model Input
0.0
0 10 20 30 40 50
Time, hr
Figure 5.34: Dynamic evolution of ethylene feed in the first reactor (Startup case).
246 Chapter 5: Results and Discussion
1.2
Startup
1.0
Scaled Hexene Feed, Kg/hr
0.8
0.6
0.4
First reactor
0.2
Plant Data
Model Input
0.0
0 10 20 30 40 50
Time, hr
Figure 5.35: Dynamic evolution of hexane feed in the first reactor (Startup case).
1.2
Startup
Scaled Hydrogen Feed, 10 x Kg/hr
1.0
-3
0.8
0.6
0.4
First reactor
0.2
Plant Data
Model Input
0.0
0 10 20 30 40 50
Time, hr
Figure 5.36: Dynamic evolution of hydrogen feed in the first reactor (Startup case).
Chapter 5: Results and Discussion 247
1.4
Startup
1.2
0.8
Manipulated Variable
0.6
0.4
First reactor
0.2
Plant Data
Model Prediction
0.0
0 10 20 30 40 50
Time, hr
Figure 5.37: Dynamic evolution of iso-butane feed in the first reactor (Startup case).
1.6
Startup
1.4
Scaled Catalyst Feed, Kg/hr
1.2
Manipulated variable
1.0
0.8
0.6
Figure 5.38: Dynamic evolution of catalyst feed in the first reactor (Startup case).
248 Chapter 5: Results and Discussion
Plant Data
Model Prediction
Controlled Variable
Setpoint
1
0
0 10 20 30 40 50
Time, hr
Figure 5.39: Dynamic evolution of ethylene off-gas composition in the first reactor
(Startup case).
1.4
Startup
Scaled Hexene Off-gas Composition, % w/w
1.2
1.0
0.8
0.6
0.4
First reactor
Figure 5.40: Dynamic evolution of hexane off –gas composition in the first reactor
(Startup case).
Chapter 5: Results and Discussion 249
16
Plant Data
Model Prediction
12
0
0 10 20 30 40 50
Time, hr
Figure 5.41: Dynamic evolution of hydrogen off-gas composition in the first reactor
(Startup case).
1.4
Startup
Scaled Solids Concentration, % w/w
1.2
Setpoint
1.0
Controlled Variable
0.8
0.6
0.4
First reactor
Figure 5.42: Dynamic evolution of solids concentration in the first reactor (Startup case).
250 Chapter 5: Results and Discussion
1.2
Startup First reactor
3
Scaled Reactor Phase Density, Kg/m
1.1
1.0
0.9
0.8
Plant Data
Model Prediction
0.7
0 10 20 30 40 50
Time, hr
Figure 5.43: Dynamic evolution of reactor phase density in the first reactor (Startup case).
1.1
Startup
Scaled Temperature, C
o
Setpoint
1.0
Controlled Variable
First reactor
Plant Data
Model Prediction
0.9
0 10 20 30 40 50
Time, hr
Figure 5.44: Dynamic evolution of temperature in the first reactor (Startup case).
Chapter 5: Results and Discussion 251
1.6
Startup First reactor
1.2
1.0
Manipulated Variable
0.8
0 10 20 30 40 50
Time, hr
Figure 5.45: Dynamic evolution of coolant’s inlet temperature in the first reactor (Startup
case).
1.4
Startup
1.2
1.0
Scaled ΔT coolant, C
o
0.8
0.6
0.4
First reactor
Figure 5.46: Dynamic evolution of and the temperature difference between the coolant
outlet and inlet values in the first reactor (Startup case).
252 Chapter 5: Results and Discussion
2.0
Startup
1.0
Plant Data
Model Prediction
0.0
0 10 20 30 40 50
Time, hr
Figure 5.47: Dynamic evolution of the valve opening cycling period in the first reactor
(Startup case).
1.2
Startup
Scaled Polymer Production Rate, Kg/hr
0.8
0.4
First reactor
Plant Data
Model Prediction
0.0
0 10 20 30 40 50
Time, hr
Figure 5.48: Dynamic evolution of polymer production rate in the first reactor (Startup
case).
Chapter 5: Results and Discussion 253
1.10
Startup
1.6
Scaled Crystallinity, %
1.4
First reactor
Scaled Product Density, kg/lt 1.2
1.0
1.05
0.8
0 10 20 30 40 50
Time, hr
1.00
Plant Data
Model Prediction
0.95
0 10 20 30 40 50
Time, hr
Figure 5.49: Dynamic evolution of product density in the first reactor (Startup case).
1.2
Startup
Scaled Ethylene Feed, kg/hr
0.8
0.4
Second reactor
Plant Data
Model Input
0.0
0 10 20 30 40 50
Time, hr
Figure 5.50: Dynamic evolution of ethylene feed in the second reactor (Startup case).
254 Chapter 5: Results and Discussion
0.18
Startup
0.16
0.14
Scaled Hydrogen Feed, Kg/hr
0.12
0.10
0.08
0.06
2.4
Startup Second reactor
2.0
Iso-butane Feed, kg/hr
1.6
1.2
Manipulated Variable
0.8
0.4
Plant Data
Model Prediction
0.0
0 10 20 30 40 50
Time, hr
Figure 5.52: Dynamic evolution of iso-butane feed in the second reactor (Startup case).
Chapter 5: Results and Discussion 255
Plant Data
Model Prediction
0
0 10 20 30 40 50
Time, hr
Figure 5.53: Dynamic evolution of ethylene off-gas composition in the first reactor
(Startup case).
0.8
Startup Second reactor
Hexene Off-gas Composition, % w/w
0.6
0.4
0.2
Plant Data
Model Prediction
0.0
0 10 20 30 40 50
Time, hr
Figure 5.54: Dynamic evolution of hexane off –gas composition in the second reactor
(Startup case).
256 Chapter 5: Results and Discussion
300
250
200
150
100
50
Plant Data
Model Prediction
0
0 10 20 30 40 50
Time, hr
Figure 5.55: Dynamic evolution of hydrogen off-gas composition in the second reactor
(Startup case).
1.4
Startup Second reactor
1.2
Setpoint
%wt Solids Concentration
1.0
0.8
Controlled Variable
0.6
0.4
Figure 5.56: Dynamic evolution of solids concentration in the second reactor (Startup
case).
Chapter 5: Results and Discussion 257
1.1
Startup Second reactor
3
Scaled Reactor Phase Density, Kg/m
1.0
0.9
0.8
0.7
Plant Data
Model Prediction
0.6
0 10 20 30 40 50
Time, hr
Figure 5.57: Dynamic evolution of reactor phase density in the second reactor (Startup
case).
1.20
Startup Second reactor
1.15
Scaled Temperature, C
o
Setpoint
1.10
Controlled Variable
1.05
Plant Data
Model Prediction
1.00
0 10 20 30 40 50
Time, hr
Figure 5.58: Dynamic evolution of temperature in the second reactor (Startup case).
258 Chapter 5: Results and Discussion
1.8
Startup Second reactor
1.4
1.2
Manipulated Variable
1.0
Plant Data
Model Prediction
0.8
0 10 20 30 40 50
Time, hr
Figure 5.59: Dynamic evolution of coolant’s inlet temperature in the second reactor
(Startup case).
1.4
Startup Second reactor
1.2
Scaled ΔT Coolant, C
1.0
o
0.8
0.6
0.4
Figure 5.60: Dynamic evolution of and the temperature difference between the coolant
outlet and inlet values in the second reactor (Startup case).
Chapter 5: Results and Discussion 259
2.4
Startup
1.6
1.2
0.8
Second reactor
0.4
Plant Data
Model Prediction
0.0
0 10 20 30 40 50
Time, hr
Figure 5.61: Dynamic evolution of the valve opening cycling period in the second reactor
(Startup case).
1.2
Startup
Scaled Polymer Production Rate, Kg/hr
0.8
0.4
Second reactor
Plant Data
Model Prediction
0.0
0 10 20 30 40 50
Time, hr
Figure 5.62: Dynamic evolution of polymer production rate in the second reactor (Startup
case).
260 Chapter 5: Results and Discussion
1.08
1.7
Startup
1.6
Scaled Crystallinity, %
1.5
Scaled Product Density, kg/lt 1.06
1.4
1.3
1.04 1.2
0 10 20 30 40 50
Time, hr
1.02
Second reactor
1.00
Plant Data
Model Prediction
0.98
0 10 20 30 40 50
Time, hr
Figure 5.63: Dynamic evolution of product density in the second reactor (Startup case).
1.4
Scaled Nunber Average Molecular Weight, Mn
Startup
1.2
1.0
0.8
First reactor
Second reactor
0.6
0.4
0.2
0.0
0 10 20 30 40 50
Time, hr
Figure 5.64: Dynamic evolution of number average molecular weight in the first and
second reactor (Startup case).
Chapter 5: Results and Discussion 261
5.0
Startup
4.5
3.5
2.0
1.5
1.0
0.5
0 10 20 30 40 50
Time, hr
Figure 5.65: Dynamic evolution of polydispersity in the first and second reactor (Startup
case).
The dynamic evolution of the particle size distribution in the first and second reactor
of the series, is illustrated in Figure 5.66 and Figure 5.67, respectively. It is important to
point out that during the process startup, the PSD moves to higher diameters as the system
approaches its final steady state operation. It can be seen that a steady state PSD is
achieved after approximately 14 hours of operation. As has already been mentioned,
polymer particles produced in the first reactor, are fed into the second reactor of the series
where polymerization continues without fresh catalyst feed. Thus, the PSD of the polymer
particles at the exit of the second reactor shifts to higher diameters, compared to the one in
the first reactor, as the same polymer particles, coming from the first reactor continue
growing. In Figure 5.68, the initial catalyst size distribution is depicted. It should be
mentioned that the initial catalyst size distribution is important for the PSD of the polymer
produced in the reactors.
262 Chapter 5: Results and Discussion
60
Startup First reactor
50 4 hours
6 hours
10 hours
40 14 hours
Pp, 1/cm
30
Time
20
10
0
200 400 600 800 1000
Diameter, μm
Figure 5.66: Dynamic evolution of the PSD in the first reactor (startup case).
70
Startup Second reactor
60 6 hours
10 hours
50
12 hours
14 hours
Pp, 1/cm
40
30
Time
20
10
0
200 400 600 800 1000
Diameter, μm
Figure 5.67: Dynamic evolution of the PSD in the second reactor (startup case).
Chapter 5: Results and Discussion 263
0.15
Catalyst Size Distribution
0.10
Pp, 1/cm
0.05
0.00
10 20 30 40 50
Diameter, μm
In this simulation run, the reactor model simulates a whole month reactor series
operation, containing a grade transition. More specifically, the 1-hexene feed rate in the
first reactor is increases by over 6 times, for the production of less crystalline polymer
grade. At the same time, hydrogen feed in the first reactor is decreased in order to shift
MWD of polymer produced in the first reactor to higher molecular weights.
In Figure 5.69-Figure 5.71 the reactor first reactor’s feed policies regarding ethylene,
hexane and hydrogen are presented. As can be seen at t=11 days, 1-hexene feed rate shifts
to a higher value while hydrogen feed rate is lowered. In Figure 5.72, the dynamic
evolution of iso-butane feed rate is depicted. Iso-butane feed rate changes in order to
control the solids concentration to follow the desired set point policy (see Figure 5.77). The
dynamic evolution of reactor composition, regarding ethylene, 1-hexene and hydrogen is
presented in Figure 5.74-Figure 5.76. As expected, 1-hexene composition increases and
hydrogen composition decreases after the corresponding changes in the reactor’s feed rates
increase of 1-hexene feed in the first reactor at t=11days. The dynamic evolution of reactor
phase density is presented in Figure 5.78. The dynamic evolution of temperature, coolant
264 Chapter 5: Results and Discussion
inlet temperature as well as the temperature difference between the coolant outlet and inlet
values are depicted in Figure 5.79, Figure 5.80 and Figure 5.81, respectively. The dynamic
evolution of the valve opening cycling period, regarding the first reactor is shown in Figure
5.82. It is clear that the model is able to follow the variability exhibited by the valve
opening cycling period.
In Figure 5.83 the dynamic evolution of polymer production rate is depicted. The
dynamic evolution of polymer density for polymer produced in the first reactor is
presented in Figure 5.84. As can be seen, at t=11 days, the polymer density increases due
to the corresponding increase of the 1-hexene inflow rate. It should be noticed that the
polymer density reaches its new steady state 6 hours after the 1-hexene inflow rate step
change. It should be noticed that the density of the polymer produced in the first reactor
also affects the polymer at the exit of the second reactor (see Figure 5.98).
In Figure 5.85-Figure 5.98, simulation results regarding the second reactor of the
series are presented. The dynamic evolution of number average molecular weight and
polydispersity for the first and second reactor is presented in Figure 5.99 and Figure 5.100,
respectively. As can be seen, the step change in hydrogen inflow rate in the first clearly
affects the number average molecular weight of the polymer produced in the first reactor,
shifting it to higher values.
Chapter 5: Results and Discussion 265
2.0
28 days First reactor
1.6
Scaled Ethylene Feed, kg/hr
1.2
0.8
0.4
Plant Data
Model Input
0.0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.69: Dynamic evolution of ethylene feed in the first reactor (28 days case).
2.4
28 days
2.0
Scaled Hexene Feed, kg/hr
1.6
1.2
0.8
First reactor
0.4
Plant Data
Model Input
0.0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.70: Dynamic evolution of hexane feed in the first reactor (28 days case).
266 Chapter 5: Results and Discussion
4.0
28 days First reactor
3.5
2.5
2.0
1.5
1.0
0.5
0.0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.71: Dynamic evolution of hydrogen feed in the first reactor (28 days case).
2.4
28 days First reactor
2.0
Scaled Iso-butane Feed, kg/hr
1.6
1.2
Manipulated Variable
0.8
0.4
Plant Data
Model Prediction
0.0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.72: Dynamic evolution of iso-butane feed in the first reactor (28 days case).
Chapter 5: Results and Discussion 267
2.4
28 days First reactor
2.0
Scaled Catalyst Feed, kg/hr
1.6
1.2
0.4
Plant Data
Model Prediction
0.0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.73: Dynamic evolution of catalyst feed in the first reactor (28 days case).
2.0
Sclaed Ethylene Off-gas Composition, % w/w
1.5
1.0
Controlled Variable
0.5
Plant Data
Model Prediction
0.0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.74: Dynamic evolution of ethylene off-gas composition in the first reactor (28
days case).
268 Chapter 5: Results and Discussion
2.0
28 days
1.2
0.8
First reactor
0.4
Plant Data
Model Prediction
0.0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.75: Dynamic evolution of hexane off –gas composition in the first reactor (28
days case).
6
Scaled Hydrogen Off-gas Composition, % w/w
Plant Data
Model Prediction
0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.76: Dynamic evolution of hydrogen off-gas composition in the first reactor (28
days case).
Chapter 5: Results and Discussion 269
1.1
28 days First reactor
1.0
0.9
Controlled Variable
0.8
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.77: Dynamic evolution of solids concentration in the first reactor (28 days case).
1.1
28 days First reactor
3
Scaled Reactor Phase Density, Kg/m
1.0
Plant Data
Model Prediction
0.9
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.78: Dynamic evolution of reactor phase density in the first reactor (28 days case).
270 Chapter 5: Results and Discussion
1.05
28 days First reactor
Scaled Temperature, C
o
Controlled Variable
1.00
Plant Data
Model Prediction
0.95
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.79: Dynamic evolution of temperature in the first reactor (28 days case).
1.4
28 days First reactor
Scaled Coolant Inlet Temperature, C
o
1.3
Plant Data
Model Prediction
1.2
1.1
1.0
0.9
0.8
0.7
0.6
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.80: Dynamic evolution of coolant’s inlet temperature in the first reactor (28 days
case).
Chapter 5: Results and Discussion 271
2.0
28 days First reactor
1.2
0.8
Plant Data
Model Prediction
0.4
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.81: Dynamic evolution of and the temperature difference between the coolant
outlet and inlet values in the first reactor (28 days case).
2.0
28 days First reactor
Scaled Valve Openinfg Cycling Period, sec
1.6
1.2
0.8
0.4
Plant Data
Model Prediction
0.0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.82: Dynamic evolution of the valve opening cycling period in the first reactor (28
days case).
272 Chapter 5: Results and Discussion
2.0
28 days First reactor
1.2
0.8
Plant Data
Model Prediction
0.4
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.83: Dynamic evolution of polymer production rate in the first reactor (28 days
case).
1.05
1.4
28 days
Scaled Crystallinity, %
1.2
Scaled Product Density, kg/lt
1.0
0.8
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
1.00
First reactor
Plant Data
Model Prediction
0.95
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.84: Dynamic evolution of product density in the first reactor (28 days case).
Chapter 5: Results and Discussion 273
2.0
28 days
1.2
Plant Data
Model Input
0.4
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.85: Dynamic evolution of ethylene feed in the second reactor (28 days case).
0.25
28 days
0.20
Scaled Hydrogen Feed, kg/hr
0.15
0.10
Second reactor
0.05
Plant Data
Model Input
0.00
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.86: Dynamic evolution of hydrogen feed in the second reactor (28 days case).
274 Chapter 5: Results and Discussion
2.0
28 days Second reactor
1.0
0.5
Plant Data
Model Prediction
0.0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.87: Dynamic evolution of iso-butane feed in the second reactor (28 days case).
5
28 days Second reactor
Scaled Ethylene Off-gas Compositon, % w/w
1
Plant Data
Model Prediction
0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.88: Dynamic evolution of ethylene off-gas composition in the first reactor (28
days case).
Chapter 5: Results and Discussion 275
1.0
0.8
0.6
0.4
Second reactor
0.2
Plant Data
Model Prediction
0.0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.89: Dynamic evolution of hexane off –gas composition in the second reactor (28
days case).
300
Scaled Hydrogen Off-gas Composition, % w/w
250
200
150
100
50
Plant Data
Model Prediction
0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.90: Dynamic evolution of hydrogen off-gas composition in the second reactor
(28 days case).
276 Chapter 5: Results and Discussion
1.05
28 days Second reactor
1.00
0.95
Plant Data
Model Prediction
0.90
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.91: Dynamic evolution of solids concentration in the second reactor (28 days
case).
1.10
28 days Second reactor
3
Scaled Reactor Phase Density, kg/m
1.05
1.00
0.95
0.90
0.85
Plant Data
Model Prediction
0.80
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.92: Dynamic evolution of reactor phase density in the second reactor (28 days
case).
Chapter 5: Results and Discussion 277
1.15
28 days Second reactor
1.13
Controlled Variable
1.12
1.11
Plant Data
Model Prediction
1.10
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, days
Figure 5.93: Dynamic evolution of temperature in the second reactor (28 days case).
1.4
28 days Second reactor
Scaled Coolant Inlet Temperature, C
o
1.2
Manipulated Variable
1.0
0.8
Plant Data
Model Prediction
0.6
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.94: Dynamic evolution of coolant’s inlet temperature in the second reactor (28
days case).
278 Chapter 5: Results and Discussion
2.0
28 days
1.5
Scaled ΔΤ Coolant, C
o
Plant Data
Model Prediction
0.0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.95: Dynamic evolution of and the temperature difference between the coolant
outlet and inlet values in the second reactor (28 days case).
2.0
28 days Second reactor
Scaled Valve Opening Cycling Period, sec
1.6
1.2
0.8
Plant Data
Model Prediction
0.4
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.96: Dynamic evolution of the valve opening cycling period in the second reactor
(28 days case).
Chapter 5: Results and Discussion 279
2.0
28 days Second reactor
1.2
0.8
Plant Data
Model Prediction
0.4
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.97: Dynamic evolution of polymer production rate in the second reactor (28 days
case).
1.05
2.0
28 days
1.8
Scaled Crystallinity, %
1.04
Scaled Product Density, kg/lt
1.6
1.4
1.03 1.2
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
1.02
Second reactor
1.01
Plant Data
Model Prediction
1.00
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.98: Dynamic evolution of product density in the second reactor (28 days case).
280 Chapter 5: Results and Discussion
1.4
1.0
0.8
First reactor
0.6
Second reactor
0.4
0.2
0.0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.99: Dynamic evolution of number average molecular weight in the first and
second reactor (28 days case).
6
28 days
5
Scaled Polydisperity
3 First reactor
Second reactor
0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Time, Days
Figure 5.100: Dynamic evolution of polydispersity in the first and second reactor (28 days
case).
Chapter 5: Results and Discussion 281
5.3. References
Marechal P., 2006, Apr. 25, ‘Process for Producing Bimodal Polyethylene Resins’, US
Patent No. 7,034,092 B2.
Noll P., 2005, Jan. 27, ‘Process for Preparing Polyethylene’, US Patent No 2005/ 0020784
A1.
Pladis P., Kanellopoulos V., Chatzidoukas C. and Kiparissides C., 2008, ‘Effect of
Reaction Conditions and Catalyst Design on the Rheological Properties of
Polyolefins Produced in Gas-Phase Olefin Polymerization Reactors’, Macromol.
Theory Simul., 17, pp. 478-487.
282
Chapter 6: Concluding Remarks and Future Work 283
Chapter 6
According to the proposed modelling approach, each loop reactor (i.e., consisting of
the loop reactor and the settling legs) is modelled as an ideal CSTR in series with a semi-
continuous product removal unit. Dynamic macroscopic mass and energy balances are
derived for each loop reactor in the series to predict the time variation of the concentrations
of the various molecular species as well as the reactor and jacket temperatures in the two
loop reactors. The non-continuous product withdrawal rate as well as the outflow species
concentrations from each reactor of the configuration is properly calculated via the solution
of a settling leg model.
284 Chapter 6: Concluding Remarks and Future Work
Extensive numerical simulations were carried out to assess the effects of process
operating conditions on key process variables as well as on the molecular, rheological and
morphological properties of polyolefins. The results were qualitatively consistent with the
expected behavior for the process variables and polymer properties. Moreover, it was
found that model predictions were in excellent agreement (quantitatively) with dynamic
measurements under different plant operating policies (i.e., startup, grade transition, etc.)
obtained from an industrial slurry-phase catalytic olefin polymerization loop-reactor series.
It was demonstrated that the present model can simulate bimodal MWDs produced in
multistage reactor configurations. The proposed model should find wide application in the
design, optimization and control of industrial polyolefin slurry-phase loop-reactor
processes.
(e.g., catalyst feed rate, etc.) and mixing conditions on the molecular and
morphological properties as well as on the molecular species dispersion
characteristics, energy requirements (i.e., recirculation pump power
consumption, pressure losses along the reactor length, etc). Moreover, the
employed multi-zone modelling approach will be capable of evaluating various
reactor instabilities and problematic operations in the slurry loop reactor
process.
Particle agglomeration phenomena as well as the operability of the slurry loop
reactors largely depend on the comprehension of the growth of single catalyst/
polymer particles. Particle agglomeration and/ or attrition phenomena in the
slurry loop reactors can be further studied. To achieve this, a single particle
model should be employed to investigate if possible hot spots formation (i.e.,
particle agglomerates) in the reactor zones may take place (i.e., axial reactor
locations) taking into account the local reactor operating and mixing conditions
along with the particle size distribution model.
Application of model predictive control (i.e., MPC) on the developed reactor
model to define the optimum reactor policy (regarding reactor feed rates,
controllers’ setpoints, etc.) in order to minimize grade transition times or reach a
preferred steady state under the most advantageous reactor policy route.
CV of V. Touloupides i
PERSONAL INFORMATION
Name Vassileios Touloupides
Address Auxendiou Str. 3, Triandria 55337, Thessaloniki, Greece
Phone Number +30-2310-922469 (Home) +30-6972-268369 (Mobile)
Email v.touloupides@gmail.com, vast@cperi.certh.gr
Nationality Greek
Date of Birth 16.05.1982
Place of Birth Thessaloniki/ Greece
Marital Status Single
EDUCATION
2009- Aristotle University of Thessaloniki (Α.U.TH.)
Medical School
2005-10 Ph.D in Chemical Engineering
Aristotle University of Thessaloniki (Α.U.TH.)
Department of Chemical Engineering, Faculty of Engineering.
Ph.D Thesis: ‘Mathematical Modelling and Simulation of an
Industrial α-Olefins Catalytic Slurry Phase Loop-reactor Series’
Corresponding Professor: Costas Κiparissides, Professor of Chemical
Engineering, Aristotle University of Thessaloniki.
2000-05 Diploma in Chemical Engineering
Aristotle University of Thessaloniki (Α.U.TH.)
Department of Chemical Engineering, Faculty of Engineering.
Graduation Degree: 8.5/10.0 Ranking: First within 110 Graduated
Students.
Diploma Thesis: ‘MATLAB Applications in Chemical
Engineering’
Corresponding Professor: Costas Κiparissides, Professor of Chemical
Engineering, Aristotle University of Thessaloniki.
EXPERIENCE
2004-05 Technical & Economical Study on Ethylene-oxide Production Plant,
Department of Chemical Engineering, Faculty of Engineering
(A.U.TH.).
July-August 2004 Training on process dynamics, Centre for Research and Technology
Hellas.
AWARDS & GRANTS
2011 ‘Borealis Student Innovation Award’ for the best doctoral graduate
thesis.
2005-10 Scholarship from Centre for Research and Technology Hellas.
2005 Award for Excellent Academic Achievements from the Technical
Chamber of Greece.
2005 Award for Excellent Academic Achievements from the Pan-Hellenic
Association of Chemical Engineers.
2004 Award for Excellent Academic Achievements from the Department
of Chemical Engineering, Aristotle University of Thessaloniki.
2004-05
2003-04 Scholarship from the National Institute of Scholarships.
2000-01
ii CV of V. Touloupides
COMPUTER PROGRAMS
Excellent Knowledge of:
WINDOWS 2000/NT/XP/VISTA/7 Operation Systems and Internet
Applications,
MICROSOFT OFFICE,
FORTRAN 77/90, MICROSOFT DEVELOPER STUDIO,
MATLAB,
ORIGIN, PHOTOSHOP and VISIO Software Packages.
FURTHER INFORMATION
Member of the Technical Chamber of Greece (2005).
Member of the Panhellenic Association of Chemical Engineers
(2007).
Driving License.
Personal Interests:
Playing Guitar, Music, Medicine, Gardening, Reading, Travel.