Polymerization Reactors

P Pladis, Chemical Process & Energy Resources Institute, Centre for Research and Technology Hellas, Thessaloniki, Greece
C Kiparissides, Chemical Process & Energy Resources Institute, Centre for Research and Technology Hellas, Thessaloniki, Greece;
Aristotle University of Thessaloniki, Thessaloniki, Greece
ã 2014 Elsevier Inc. All rights reserved.

Introduction 1
Polymerization Reactor Types 2
Polymerization Processes 5
Bulk Polymerization 6
Solution Polymerization 7
Precipitation Polymerization 7
Suspension Polymerization 7
Emulsion Polymerization 7
Solid Catalyzed Polymerizations 7
Interfacial Polymerization 8
Solid-State Polymerization 8
Reactor Design Equations 8
Thermodynamic Considerations 9
Multiscale Modeling of Polymerization Systems 10
Optimal Operation of Polymerization Reactors 11
References 12

Introduction

Plastics are major industrial goods used in the building, construction, packaging, transportation, electronic, etc., industries. The
2012 world plastics’ production was estimated at about 280 million tons.1 Plastics can be in general classified into thermoplastics,
thermosetting resins, and engineering plastics. The former yield solid materials upon cooling a polymer melt and soften upon
heating. Commodity thermoplastics are manufactured in large volumes and comprise polymers such as polyethylene (low and
high density), isotactic polypropylene, polystyrene, and polyvinyl chloride (PVC). On the other hand, thermosetting resins harden
irreversibly through chemical cross-linking reactions and, therefore, cannot be remelted and reshaped. Thermosets comprise
phenolic resins, amino resins, epoxides, alkyd resins, unsaturated polyesters, polyurethanes, etc. Finally, engineering plastics are
in general thermoplastics, possessing improved mechanical properties compared to commodity plastics. Polyethylene terephthal-
ate, polyamides, polycarbonates, some modified polystyrenes (e.g., Styrene Acrylonitrile (SAN) and Acrylonitrile Butadiene Styrene
(ABS)), various polyetherketones, polyimides, polysulfones, and some thermosetting resins belong to this class of specialty plastics.
Synthetic polymers can be synthesized from monomers by a multitude of reaction mechanisms, including addition (e.g., free-
radical, ionic, group transfer, and catalytic Ziegler–Natta polymerizations) and step-growth (e.g., polycondensation) reactions.
Their macromolecular architecture (e.g., molar mass, molecular weight distribution (MWD), copolymer composition distribution
(CCD), chain sequence length distribution (CSLD), branching distribution (BD), and stereoregularity) depends not only on the
chemical nature of the monomers, the type of polymerization mechanism, and the physical state of the reacting system (e.g.,
polymerization process) but also on the type of reactor configuration. It should be emphasized that the ‘polymer quality’ is a much
more complex issue in polymerization processes than in conventional short-chain reactions since the molecular architecture of
polymer chains strongly affects the physical, chemical, thermal, rheological, mechanical, optical, etc., polymer properties and,
therefore, its end-use applications. Thus, control of the ‘polymer quality’ in terms of process operating conditions in a polymer
reactor is key to the efficient production of high-quality, tailor-made polymers.
One of the most difficult issues in polymer reactor design, operation, optimization, and control is to determine the quantitative
relationships between the physical, thermal, mechanical, rheological, and chemical properties of a polymer and its molecular chain
architecture (e.g., MWD, CCD, CSD, and LCBD). Table 1 summarizes both types of polymer quality control measures.
Since the end-use properties of the produced polymers are directly linked with the molecular characteristics of the polymer
chains, control of the polymer chain microstructure is of profound interest to the polymer manufacturing industry. Thus, a major
objective of polymerization process modeling is to quantify the effects of kinetic mechanism, physical transport phenomena (e.g.,
mass and heat transfer), reactor type, and operating conditions on the ‘polymer quality.’ This presupposes a thorough knowledge of
the polymerization kinetics and the availability of advanced mathematical models to quantify the effects of process operating
conditions and reactor configuration on the molecular and morphological polymer properties.

Reference Module in Chemistry, Molecular Sciences and Chemical Engineering http://dx.doi.org/10.1016/B978-0-12-409547-2.10908-4 1

2 Polymerization Reactors

Table 1 Measures of polymer product quality

Molecular and morphological properties

Average molecular weights and molecular weight distribution (MWD)
Copolymer composition distribution (CCD)
Sequence length distribution (SLD)
Long-chain branching distribution (LCBD)
Stereoregularity (tacticity)
Particle size distribution (PSD)
Particle porosity, surface area, etc.
End-use properties
Physical and thermal properties (e.g., density, clarity, melting point, temperature stability, swellability, and plasticizer uptake)
Chemical (e.g., corrosion resistance)
Mechanical properties (e.g., strength, toughness, stress crack resistance, abrasion resistance, and impact resistance)
Rheological properties (e.g., flow properties, shear viscosity and elasticity, melt index, extensional viscosity, and extrudate swell)

Polymerization Reactor Types

A variety of polymerization reactors, including continuous, semibatch, and batch reactors, are used in industrial polymer
manufacturing. In general, batch reactors are usually employed for small-to-intermediate production of polymers or specialty
grades. On the other hand, continuous reactor systems (e.g., continuous stirred tanks, train of CSTRs, loop-type reactors, and
fluidized bed reactors (FBRs)) are suitable for large-volume production of commodity polymers such as polyolefins, high-impact
polystyrene, and poly(methyl methacrylate).2 Semibatch reactors are usually single stirred tanks similar to the batch reactor units.
Their operation usually follows the following process step: initial charge of the reactants (i.e., monomer(s), initiator(s), and
solvent(s)) into the reactor that is followed by continuous/discrete feeding of some reactant (e.g., comonomer, chain transfer agent
(CTA), and initiator) to control the polymerization rate and/or the polymer chain properties.
The choice of the reactor type or reactor configuration (e.g., number of CSTRs in series) for a given polymerization depends on
many factors, including the polymerization process (e.g., solution, suspension, and emulsion), the polymerization mechanism
(e.g., free radical and catalytic), the number of phases present in the polymerization system and thermodynamic partitioning of the
various reactive species (e.g., monomer, solvent, and initiator in the different phases, gas, liquid, and solid, present in the
polymerization), reaction conditions (e.g., temperature, pressure, and viscosity), heat removal rate, polymer product properties,
investment and operating cost, operability, and controllability.2
In the continuous production of polymers by free-radical polymerization, different types of reactors, including continuous-flow
stirred tank reactors, tower reactors, horizontal linear flow reactors, tubular reactors, and screw-type reactors, are commonly used.3
The stirred tank reactor is typically an agitated, jacketed pressure vessel, often with a provision for an overhead condenser for
solvent or monomer reflux; see Figure 1. Reactors other than stirred tanks may be functionally equivalent to stirred tanks. For
example, loop reactors are widely used in the polymer industry, especially for solution and slurry-phase catalytic olefin polymer-
izations. In a loop reactor, the agitator in the stirred tank is replaced by a circulation pump. The loop reactor usually consists of a
jacketed pipe and may include several ‘legs or tube sections’ connected in series; see Figure 2. Due to the very high circulation rate of
the reaction mixture, the operation (design equation) of a loop reactor can be approximated by that of a well-stirred tank vessel.
In general, a continuous reactor will behave like a CSTR when there are not any spatial inhomogeneities in temperature and
composition and its residence time distribution (RTD) follows that of a classical CSTR.3,4
Continuous multizone stirred tank reactors have been used in the high-pressure free-radical polymerization of ethylene (LDPE
autoclaves) and in the polymerization of styrene. The reactor is either a vertical or a horizontal vessel with a large length-
to-diameter ratio. The reactor consists of several equal or nonequal volume compartments (i.e., reaction zones) in series, separated
by disks. Radial and axial mixing of the polymerization mixture is effected by a stirrer shaft, equipped with several impellers (e.g.,
pitched blades and turbine), that is running across the multizone reactor.2,5 Usually, in a multizone polymerization reactor, the
flow-mixing behavior of the reaction mixture significantly deviates from that observed in ideal flow reactors (e.g., plug flow and
perfectly mixed reactors). Thus, the flow behavior in a multizone reactor needs to be approximated by a more complex RTD (e.g.,
multizone models with recycle) to account for the effect of macromixing on the polymerization kinetics.5 The main motivation for
using a multizone reactor is that the polymerization conditions (e.g., temperature and monomer and initiator concentrations) can
widely vary along the length of the multizone reactor, which greatly improves our ability to control the final polymer properties
and polymer productivity. For example, in the high-pressure, free-radical polymerization of ethylene in autoclaves, the reaction is
practically carried out in an adiabatic mode. Thus, cooling of the polymerization mixture is effected by the introduction of cold
monomer at several side-feed points along the multizone reactor. Moreover, the polymerization temperature in a zone is controlled
by manipulating the corresponding initiator feed rate into a specific reaction zone; see Figure 3.
Fluidized bed catalytic polymerization reactors (Figure 4) for solid catalyzed polymerization have long been a common
configuration for producing high-density polyethylene grade and linear low-density polyethylene grade (HDPE and LLDPE).
They are also used in a cascade-reactor configuration in the production of high-impact polypropylene. In a catalytic FBR, catalyst
particles, in the size range of 20–80 mm, are continuously fed into the reactor at a point above the gas distributor and react with the
 Polymerization Reactors 3

Figure 1 Schematic representation of a continuous stirred tank reactor (CSTR).

Figure 2 The slurry-phase catalytic loop polyethylene process.

4 Polymerization Reactors

Figure 3 Schematic representation of a high-pressure LDPE multizone autoclave.

incoming fluidizing monomer(s) to form a broad distribution of polymer particles (e.g., 100–5000 mm). The particulate polyolefin
product is continuously withdrawn from the reactor at a point, preferably, close to the bottom of the bed. The polymerization heat
is removed via the unreacted gas mixture exiting the bed, which is subsequently cooled and recycled back to the reactor. Industrial
FBRs typically operate at temperatures of 70–110
C and pressures of 20–40 bar. The superficial gas velocity can vary from 3 to
7 times the minimum fluidization velocity. The catalyst feed rate can vary from 0.01 to 0.5 g s1, depending on the catalyst activity
and reactor capacity. The single-pass monomer conversion ranges from 2% to 5%, whereas the overall monomer conversion can be
as high as 98%. The catalyst/polymer particles in the reactor can grow due to polymerization, can undergo particle breakage and/or
particle agglomeration, or can be entrained by the fluidizing gas.6 Since polymer particles are formed in the gas phase in the absence
of a diluent, there is no need for separation of diluent from the particulate polymer product except for the removal of unreacted
monomer. This is a clear advantage of the gas-phase process over the slurry and solution processes.
Tubular reactors have certain advantages over stirred tank reactors: design simplicity, good heat transfer capability, and narrow
MWD of the polymer due to minimal back mixing. Other advantages of tubular reactors relative to stirred tanks include suitability
for use at higher pressures and temperatures and the fact that severe energy transfer constraints may be surmounted readily using
this configuration.7 The large surface area/volume ratio is particularly advantageous for the dissipation of heat generated by
exothermic polymerizations. However, one major problem in the operation of tubular polymerization reactors is that the viscosity
of the polymerizing mixture increases significantly as the monomer conversion increases. Consequently, large radial and axial
temperature gradients may exist and the velocity profile can be significantly distorted.2 In addition, during reactor operation, the
inner tube wall can be fouled due to polymer deposition.
A high-pressure LDPE tubular reactor typically consists of a spiral-wrapped metallic pipe having a large length-to-diameter
ratio.8 The total length of the tube ranges from 500 to 1500 m, while its internal diameter does not exceed 70–80 mm. With respect
to the process heat requirements, the tubular reactor is divided into a number of zones (i.e., preheating, reaction, and coolant
zones). In the preheating zone, the ethylene feed stream is heated (by steam or water) up to a temperature at which the
polymerization can be initiated via the thermal decomposition of a mixture of initiators. The tubular reactor also includes a
number of reaction and cooling zones, arranged in a certain sequence, in accordance with the reactor design and product
specifications. A commercial tubular reactor usually consists of 2–6 reactor zones and 6–12 coolant zones. Ethylene polymerization
is a highly exothermic reaction. The heat of reaction is partially removed through the reactor wall by the heat transfer fluid (usually
water). Approximately half of the reaction heat is removed by the coolant streams flowing into the various reactor jackets, either
cocurrently or countercurrently to the reaction fluid. This results in a no isothermal reactor operation. The temperatures and flow
rates of the coolant streams are used to control the heat removal rate and, thus, the temperature profile along the tube. Fouling of a
reactor zone is partially overcome by a periodic increase of the reactor pressure by means of the operation of a control valve located
 Polymerization Reactors 5

Figure 4 Schematic representation of a gas-phase catalytic olefin polymerization in an FBR unit.

near the reactor exit. The periodic operation of the valve results in an increase of the velocity of the reaction mixture that tears away
the polymeric deposits.
The tubular reactor usually includes a number of monomer, initiator, and CTA side injection points to control the polymer
production rate and polymer quality. The various initiator feed streams usually contain a mixture of 2–4 initiators, having different
decomposition temperatures. In general, a multitude of polyethylene grades can be produced from a single reactor line. Thus, the
monomer conversion can vary in the range of 20–35%, while the LDPE density ranges from 0.915 to 0.935 g cm3.8 In Figure 5,
a schematic representation of a high-pressure tubular LDPE plant is shown.

Polymerization Processes

Polymers can be classified according to one or more of the following criteria: (i) chemical nature of monomers, (ii) molecular
structure of polymers, (iii) polymer chain growth mechanism, and (iv) type of polymerization process. Notice that the previously
mentioned criteria are not unique and several other alternative classifications have been proposed in the literature.
In general, the various industrial polymerization processes can be broadly classified into homogeneous and heterogeneous
ones. The first class comprises polymerization processes that are carried out in a single phase. In heterogeneous polymerizations,
either the polymer is insoluble in the monomer phase (e.g., bulk precipitation polymerization of acrylonitrile), or the polymer-
ization involves the presence of different phases (e.g., suspension polymerization, emulsion polymerization, and solid–gas
catalytic polymerization). More specifically, the following polymerization processes are commonly employed for the manufactur-
ing of various polymers: (i) bulk, (ii) solution, (iii) precipitation, (iv) suspension, (v) emulsion, (vi) solid catalyzed polymeriza-
tion, (vii) interfacial polycondensation, and (viii) solid state.9
Table 2 presents the main characteristics of some typical addition polymerization processes.10
6 Polymerization Reactors

Figure 5 Schematic representation of a high-pressure LDPE tubular reactor process.

Table 2 Types of addition polymerization processes

Process Bulk and solution Precipitation Liquid slurry Gas dispersion Emulsion Suspension

Particle size (mm) – 102–104 10–104 10–104 0.1–1.0 100–500

Commercial reactor types Batch, CSTR, tube, Batch, CSTR Batch, CSTR, Fluidized bed, Batch, CSTR Batch
extruder extruder loop stirred bed
Kinetic mechanism All All ZN ZN FR, I FR, I
Examples (including PS, LDPE, PMMA, PVC, PTFE, PF HDPE, PP, EP HDPE, LLDPE, PP, PS, PMMA, PVA, PMMA, PS,
copolymers) PVAC EP PTFE PVC

Mechanisms: FR, free radical; I, ionic; ZN, Ziegler–Natta.

Monomers: PS, polystyrene; PMMA, polymethacrylate; PVA, polyvinyl acetate; LDPE, low-density polyethylene; HDPE, high-density polyethylene; LLDPE, linear low-density
polyethylene; PTFE, polytetrafluorethylene; PF, polyformaldehyde; EP, ethylene–propylene copolymers; PP, polypropylene.

Bulk Polymerization
In bulk polymerization, the reaction mixture consists essentially of pure monomer with small amounts of dissolved catalyst and
molecular weight modifiers. Bulk polymerization may be of either homogeneous or heterogeneous type depending on the mutual
solubility of monomer and polymer. Styrene, vinyl acetate, and methyl methacrylate are all miscible with their respective polymers
and hence polymerize homogeneously in bulk. Vinyl chloride and acrylonitrile are typical examples of monomers where the
polymerization is carried out heterogeneously. In these cases, the reaction mixture forms a slurry almost from the beginning of the
reaction. Typical features of homogeneous bulk polymerization are the high viscosity of reaction mixture and the poor heat transfer
characteristics. This is particularly true with homogeneous free-radical bulk polymerization where the viscosity increases dramat-
ically. The heat transfer coefficient is further reduced because polymer can deposit on the reactor wall. At intermediate conversions,
the rate of heat generation often accelerates with conversion (due to the gel effect). Thus, it becomes impossible to carry the reaction
much further than 55–65% conversion in homogeneous bulk, while with heterogeneous reactions, a somewhat higher conversion
is possible because of the milder increase in viscosity. The polymer product will generally contain residual monomer, which is
usually removed by flash evaporation and/or steam heating.
 Polymerization Reactors 7

Solution Polymerization
Solution processes are widely used in production of rubbers such as medium cis-polybutadiene, which is manufactured with
anionic catalysts and hexane as solvent, and also butyl rubber, which is obtained by copolymerization of isobutylene and isoprene
in methyl chloride with a cationic polymerization catalyst. Polymerization kinetics can be either homogeneous or heterogeneous
(precipitation). Solution polymerization has the advantage that the viscosity can be reduced; thus, the removal of the heat of
reaction becomes easier. However, the separation of the polymer from the solution is often a costly operation.

Precipitation Polymerization
When the polymer is insoluble in its monomer or in the monomer–solvent solution, a polymerization that begins in a
homogeneous phase becomes quickly heterogeneous as the polymer precipitates as a second phase. Some well-known polymer-
ization processes, which follow precipitation kinetics, include the manufacture of PVC, polyacrylonitrile, and low-density poly-
ethylenes under certain operating conditions. In precipitation polymerization, the overall kinetics is assumed as the sum of
independent reactions in the two phases, namely, the monomer-rich phase and the polymer-rich phase. In the former phase,
normal solution kinetics is followed. In the latter, which is swollen with monomer, all reactions (e.g., termination and propaga-
tion) may become diffusion-controlled.

Suspension Polymerization
In suspension polymerization, monomer droplets with dissolved initiator are commonly dispersed in water. As the polymerization
progresses, the droplets are transformed into sticky, viscous monomer–polymer particles that finally become rigid, spherical
polymer particles of size 50–500 mm. The concentration of polymer is typically 30–50% in the fully converted suspension. The
viscosity of the suspension is determined by the dispersion medium and, therefore, remains essentially constant throughout the
reaction. Thus, temperature control is not difficult. To maintain the dispersion of monomer droplets/polymer particles from
coalescing, surface-active stabilizing agents are added into the dispersion and proper agitation is applied. The control of particle size
distribution (PSD) is in general a complex issue and so is the scaling-up of the process. Suspension polymerization is used on a
large scale (e.g., 150 m3 batch reactors) for the manufacture of PVC and for specific grades of polystyrene.

Emulsion Polymerization
In emulsion polymerization, an aqueous dispersion of monomer(s) is converted by free-radical polymerization into a stable
dispersion of polymer particles of 0.1–3 mm in diameter. A typical emulsion recipe comprises the dispersing medium (e.g., water),
monomer(s), a water-soluble initiator, and an emulsifying agent (e.g., sodium and potassium salts of saturated long-chain acids).
The excess of surfactant forms micelles (c. 5–10 mm in diameter), which are clusters of surfactant molecules. A very small part of the
monomer can be found in solution, some is solubilized by the micelles, but most stays in the monomer droplets (c.10 mm in
diameter). Usually, one has 1018 micelles/cm3 and 1010–1011 monomer droplets/cm3. Polymerization does not occur in the
monomer droplets but either in interior of the micelles (micellar nucleation) or in the continuous phase (homogeneous
nucleation) when the monomer is slightly soluble.
The main advantages of the emulsion polymerization are the following: low viscosity of the reaction mixture, easy thermal
control of the reactor, virtually 100% conversion that may be achieved, high polymerization rates, and final latex that may be
directly usable. Its disadvantages are the following: emulsifier and coagulants are hard to remove from the final polymer (e.g., high
residual impurity), and the production cost is higher than that of suspension systems. Typical emulsion processes include
copolymerization of styrene and butadiene (SBR rubber) and polymerization of chloroprene (neoprene rubber). Also, vinyl acetate
and several acrylic monomers are polymerized in emulsion in the manufacture of latex paints.

Solid Catalyzed Polymerizations

HDPE, LLDPE, and isotactic polypropylene are produced in low-pressure reactors in the presence of Ziegler–Natta catalysts.
Although some catalysts are used in liquid form, most are heterogeneous, solid catalysts that can be supported or unsupported.
The monomer in contact with the catalyst could be gaseous (the gas-phase process), a pure liquid (the liquid solution process), or
dissolved in a diluent (the slurry process). In all cases, the porous catalyst fractures and fragments of the catalyst are dispersed in the
polymer and become the locus of polymerization.
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 polymers
particles, and removal of reaction heat. Hence, gas-phase polymerization is also called dry polymerization. Solution processes have
some unique advantages including better molecular weight control, process variables that can be more easily controlled because
polymerization occurs in a homogeneous phase, and high productivity due to higher temperatures. However, polymers of high
molecular weight cannot be produced easily. Furthermore, lower solid content results in greater diluent recovery.
8 Polymerization Reactors

The liquid slurry polymerization process encompasses by far the largest group of solid catalyzed olefin polymerization
technologies. Excellent temperature control is a major attraction of the liquid slurry process. On the other hand, swelling of
polymer in the slurry medium can be a major problem since it lowers the polymer production rate. An example of slurry
polymerization is the polymerization of propylene in n-heptane in the presence of a solid Ziegler–Natta catalyst. The polymer
chains grow from the solid surface and remain attached to it. The polymer becomes rapidly insoluble and precipitates out to form a
separate solid phase swollen with the monomer/solvent. In some processes, each catalyst particle generates one particle (e.g.,
limited particle agglomeration). Recently, several types of homogeneous catalyst systems have been developed (e.g., metallocenes)
for precipitation polymerization. In this case, polymerization starts in solution, but as the polymer chains grow, they become
insoluble, agglomerate, and finally precipitate to form a separate solid phase.

Interfacial Polymerization
This type of process requires two phases, for example, an organic phase containing an acid chloride and an aqueous phase
containing a diamine. Polymerization is typically carried out at the interface between the two immiscible liquid phases following a
step-growth polymerization mechanism. The polymerization rate is quite fast, and contrary to the usual step polymerization,
high-molecular-weight polymer is produced at room temperature. Notice that the polymerization rate is controlled by the diffusion
rates of monomers to the polymerization interface.

Solid-State Polymerization
Polymerization in the solid state can produce macroscopic polymer single crystals and, in some cases, crystals of high optical
quality. The propagation reaction is controlled by the crystal structure and symmetry of the monomer. Thus, in certain cases, the
propagation reaction may lead to a crystal-to-crystal transformation. Many novel polymers of perfect stereoregularity, including
polydiacetylenes, crystalline 1,4-polybutadiene, ultrathin oriented layers of vinyl polymers, and various optically active polymers
in crystalline organization, have been produced by solid-state polymerizations.

Reactor Design Equations

To describe the dynamic operation of a complex polymerization process, one commonly needs to derive a multiphase, multiscale
mathematical model, accounting for the physical and chemical phenomena taking place in the polymerization reactor. In general,
a population balance approach is presented to follow the molecular weight (e.g., MWD and LCB) and morphological (e.g., PSD)
developments in a particulate polymerization process.
The development of the general population balance equation (PBE) follows the works of Hulburt and Katz11 and Kiparissides.12
Let us assume that N(x, y, z, p1, p2, . . . pk, t) represents a distribution of countable entities (e.g., polymer chains, monomer droplets,
and polymer particles) where x, y, and z denote the spatial coordinates, pi represents the ith property of the entity, and t is the time.
For example, pi could denote the degree of polymerization or the comonomer content in a copolymer chain or the volume of a
polymer particle. Thus, the term N dx dy dz dp1 dp2 . . . dpk represents the fraction of entities in the geometric volume element
dV ¼ dx dy dz, with respective property values in the range dp1 dp2, . . ., dpk. Note that the distribution function will satisfy the
following condition:
ð
Nðx, y, z, p1 , p2 , .. ., pk Þdxdydz dp1 dp2 ... dpk ¼ 1 [1]

Following the general developments of Himmelblau and Bischoff,13 the differential form of the PBE, governing the conserva-
tion of the population of countable entities, will be

@N @ @  @ Xk
@
+ ðux NÞ + uy N + ðuz N Þ + ð ui N Þ + D  B ¼ 0 [2]
@t @x @y @z i¼1
@p i

where ux ¼ dx/dt, uy ¼ dy/dt, and uz ¼ dz/dt denote the usual geometric velocities in the spatial coordinates x, y, and z, respectively.
ui ¼ dpi/dt is the rate of change of the property pi. B and D represent the corresponding rates of ‘birth’ and ‘death’ of entities per unit
time, unit geometric volume, and unit property change. Very often, the average values of the population properties in the entire
reactor volume, V, are only required. Accordingly, one can define the following geometrically averaged distribution function,
ð
1
n¼ N dV [3]
V V

By integrating eqn [3] over the geometric volume, V, and using Gauss’ divergence theorem, the following macroscopic PBE is
obtained:
 Polymerization Reactors 9

1 @ Xk
@ 1
ðVnÞ + ðui nÞ + D  B ¼ ðqin nin  qout nout Þ [4]
V @x i¼1
@p i V

where and qin and qout are the inlet and outlet volumetric flow rates, respectively.
In general, the time-spatial variation of the distributed molecular properties of interest (i.e., MWD, CCD, and LCBD) of the
polymer chains in a polymerization reactor will be described by the general PBE [4].
Thus, for a well-mixed free-radical batch reactor, one can derive the following macroscopic PBEs for the ‘live,’ R(n,t), and ‘dead,’
D(n,t), linear polymer chains of length n:

1 d½VR P ðn, t Þ
¼ rP ðn, t Þ [5]
VR dt
1 d½VR Dðn, t Þ
¼ rD ðn, t Þ [6]
VR dt
where rP(n,t) and rD(n,t) are the corresponding net formation rates for the ‘live’ and ‘dead’ polymer chains of length n. These rates
can be easily calculated from the established polymerization kinetic mechanism.12
Accordingly, for a high-pressure LDPE tubular polymerization reactor, one can derive the following steady-state PBEs for the
‘live’ and ‘dead’ nonlinear polymer chains:

@ ½uP ðb, nÞ

¼ rP ðb, nÞ; b ¼ 0,1, . .., Nb ; n ¼ 1,2, ... , Nn [7]
@z
@ ½uDðb, nÞ
¼ rD ðb, nÞ; b ¼ 0,1, . .. ,Nb ; n ¼ 1, 2, .. ., Nn [8]
@z
where P(b,n) and D(b,n) are the concentration of ‘live’ and ‘dead’ polymer chains of length n with b long-chain branches and rP(b,n)
and rD(b,n) are the corresponding net formation rates.8 u denotes the velocity of the reaction mixture in the tube and z is the axial
coordinate. Note that the total number of molar balance equations for the ‘live’ and ‘dead’ nonlinear polymer chains (see eqns [7]
and [8]) will depend on the maximum number of long-chain branches, Nb, and the maximum degree of polymerization, Nn, that
is, it will be equal to 3  Nb  Nn. Consequently, the computational effort required for the solution of the complete set of molar
species differential balance equations is prohibitively high. Thus, to calculate the joint (MWD–LCBD) distribution along the
reactor tube, one needs to solve eqns [7] and [8] using a suitable numerical method (e.g., method of double moments, ‘numerical
fractionation,’ global orthogonal collocation, orthogonal collocation on finite elements, fixed-pivot method, and stochastic Monte
Carlo method).8,12

Thermodynamic Considerations

In heterogeneous polymerization processes, the polymerization rates in the different phases present in the reactor will depend on
the concentrations of radicals (or active catalyst sites), monomer(s), etc., in the different phases. These variables largely control the
reactor productivity as well as the molecular properties of the polymer. Thus, knowledge of the solubilities of monomers and other
reactive species (e.g., CTA(s) and inert species) in a phase (e.g., polymer particles) is of profound importance in polymer reactor
design and simulation. Since it is very difficult from an experimental point of view to measure experimentally the solubilities of all
sorbed species of interest in different types of polymers, thermodynamic models, in the form of an equations of state (EOS), are
often employed to calculate the partitioning of the reactive species in a binary/ternary polymerization system.
Several thermodynamic models (i.e., activity coefficient and EOS models) have been used to describe the thermodynamic
equilibrium behavior of multiphase polymerization systems. Note that, at high pressures, activity coefficient models exhibit a
number of problems (e.g., unknown values of reference states for supercritical fluids, temperature-dependent parameters, and use
of different models for vapor and liquid phases). Thus, they cannot predict important critical phenomena and many thermody-
namic properties cannot be calculated. On the other hand, the use of EOS models (i.e., the perturbed-hard-chain theory of
Donohue and Prausnitz,14 the statistical associating fluid theory of Chapman et al.,15 and their modifications) is significantly more
accurate for complex fluids (e.g., hydrogen bonding fluids, supercritical fluids, and polymers16). Another approach for determining
the phase equilibrium of multicomponent polymer systems is based on the Sanchez–Lacombe equation of state (SL EOS).17 The SL
EOS model is based on lattice fluid theory and statistical mechanics. It can predict the thermodynamic properties and the phase
behavior of both polymeric and nonpolymeric systems. A large number of evaluations of the SL EOS have been reported in the
literature, mainly regarding liquid–liquid equilibrium behavior.
In general, the predictive capabilities of an EOS model strongly depend on a number of parameters that need to be determined
experimentally. In fact, to predict the solubility of a compound in an amorphous polymer, using the SL EOS, the characteristic
parameters (i.e., P , T , and r ) of both small molecules (monomers, solvents, etc.) and polymer need to be known. In addition,
the value of a binary interaction parameter, kij, accounting for the interaction energy between the polymer segments and the sorbed
molecules, should be known a priori. The estimation of the parameters appearing in an EOS model is usually carried out by fitting
10 Polymerization Reactors

the selected EOS model to a series of experimental PVT and phase equilibrium data of the pure components as well as of their
mixtures using nonlinear parameter estimation methods. To avoid the previously mentioned tedious and costly experimental
procedure, a molecular dynamics approach can be employed to calculate the characteristic parameters (e.g., P , T , and r )
appearing in the SL EOS.18

Multiscale Modeling of Polymerization Systems

To quantify the effects of process operating conditions (i.e., temperature, monomer/comonomer molar ratio, and initiator
concentration) on the molecular polymer properties (e.g., MWD, CCD, and LCBD), detailed mathematical models need to be
derived based on the known polymerization kinetics and an accurate description of all physical and transport phenomena (e.g.,
mass and heat transfer) taking place in a polymerization reactor. In the past, a great number of modeling studies have been reported
dealing with the prediction of polymer molecular properties for both linear and nonlinear polymer chains. Furthermore, in recent
years, new theoretical tools have been developed to predict the polymer end-use properties in terms of the molecular chain
architecture.
Thus, for a multiphase polymerization process, one needs to develop various mathematical models, at different time and length
scales, accounting for the physical and chemical phenomena taking place in a polymerization process. The present challenge in the
polymerization process modeling is to bridge the gap between molecular scale, nanoscale, microscale, mesoscale, and macroscale
phenomena in industrial-scale polymerization processes. In addition to the modeling of the reactor, other upstream and
downstream unit operations including separators, condensers, compressors, and heat exchangers need to be simulated to describe
the complete dynamic operation of a polymer production plant. The plant-scale model can then be used to conduct process
simulation studies, to calculate the optimal operating conditions to maximize plant productivity for a desired polymer grade, or to
minimize the off-spec polymer production during a grade transition. Needless to say, the predictive capabilities of a dynamic
polymerization model are strongly related to the availability of accurate information on the thermodynamic and transport
properties of the polymerization system.
In Figure 6, the different multiscale models are schematically shown for the gas-phase catalytic olefin polymerization in an FBR.
The multiscale description incorporates models at different timescales and size scales. In particular, at the molecular level, a
multisite, Ziegler–Natta, kinetic mechanism is derived to predict the evolution of the polymer molecular properties. To calculate
the particle growth, and the spatial monomer and temperature profiles in a polymer particle, the random pore, polymeric flow
model (RPPFM) is utilized. The RPPFM is solved together with a dynamic particle population balance model, to predict the PSD in
the bed.19

Figure 6 A multiscale model description of the catalytic olefin polymerization in a gas-phase FBR.
 Polymerization Reactors 11

From the numerical solution of this multiscale model, the morphological and molecular polymer properties in a gas-phase,
catalytic olefin polymerization reactor can be determined. It should be pointed out that the distributed polymer molecular
properties greatly affect the final mechanical and rheological properties of polyolefins and, thus, their end-use applications.

Optimal Operation of Polymerization Reactors

Process optimization and control can have a significant strategic impact on polymer plant operability and economics. The
operating objectives in polymerization processes must satisfy complex property requirements for the final polymer and simulta-
neously achieve the greatest economic potential during the batch operation. Meeting the specified targets during operation requires
an increased effort in many different levels of the production procedure, such as the quality of the measurements, the accuracy of
the model predictions, the performance of the control system, and the ability to determine and implement the optimal operating
conditions that would ensure the satisfaction of certain economic and product quality criteria. Commonly employed polymer
product quality indicators (e.g., mechanical strength, tear strength, and rheological properties) are directly or indirectly linked with
the molecular structural properties of polymer chains (e.g., MWD, CCD, and CSLD), which are difficult (sometimes impossible) to
measure online. Very often, average polymer molecular weight properties (e.g., number- and weight-average molecular weights),
which can be indirectly inferred from the online measurement of the solution viscosity or melt index of the polymer, are selected as
the major controlled variables that need to be maintained within well-determined limits so that desired product quality criteria can
be satisfied.
In theory, time optimal control policies that ensure the satisfaction of the polymer product property requirements and the
operational constraints can be calculated off-line and implemented as set-point changes of the regulatory control system. However,
the operation of a polymerization reactor is heavily affected by disturbances resulting from a variety of process-related sources and
an inherent model – process mismatch. Process interaction with upstream units or through recycle streams, modeling mismatch
due to simplifying assumptions and unmodeled phenomena, variation of the model parameters due to changing operating
conditions, and unmeasured process disturbances cause the plant to drift away from a predefined trajectory path. Thus, feedback
of the error between the actual values of the controlled variables and the model predictions is required so that the control system
can adjust the operating conditions (e.g., polymerization temperature and initiator concentration) accordingly. The calculation of
the optimal operating conditions depends on the adequacy of the process model, the values of the model parameters, and a
knowledge of the unmeasured stochastic disturbances that affect the process.
Note that the end-use polymer properties (e.g., density, melt index, impact strength, rigidity, tensile strength, chemical
resistance, thermal stability, and plasticizer uptake) can be related, mainly, through empirical relationships, to the molecular
and morphological properties of polymers. On the other hand, for a given reactor configuration, the molecular properties will be
strongly dependent on the reactor operating conditions. Therefore, the production of polymers with desired end-use properties can
be translated to a corresponding molecular property optimization problem.
Basically, two types of optimization problems are encountered in polymerization reactors. The first, the static optimization
problem, deals with the selection of the best (optimum) time-invariant controls so that without disturbances, the molecular
properties attain some desired values. The second, the time optimal control problem, refers to the determination of the optimal
control trajectories, in order to translate a polymer reactor from its initial state to a desired final one. The former problem is usually
related with the steady-state operation of continuous polymer reactors, while the latter is concerned with the dynamic operation of
batch and semibatch reactors and the start-up policies for continuous vessel reactors. The general solution to these optimization
problems can be obtained in terms of the following elements: (i) an accurate model of the process, (ii) a selected number of control
variables, (iii) an objective function, and (iv) a suitable numerical method for solving the specific optimization problem.20
Many optimization problems in polymerization reactors have been studied by defining a single scalar objective function that
combines all identifiable performance measures with weighting factors chosen a priori. However, the combination of several terms
in a single objective function is not always easy in the sense that some controlled variables may react in opposite directions to
manipulations of a control variable. In general, these objectives will fall into the following categories: molecular property
specifications, safety, reactor and environmental constraints, and economic objectives.
Let us assume that we desire to operate a continuous polymer reactor at a maximum production rate to obtain a polymer with
desired molecular properties (e.g., Mn, PD ). These two optimization goals can be combined into a single objective function of the
following form:
 2  2
Mn PD
J ¼ w1 XM + w2  1 + w3 1 [9]
Mn  PD

where w1, w2, and w3 represent the values of the weighting factors. It is obvious that there are practically no limitations on the exact
form of the objective function. The previously mentioned performance index is an implicit function of the control variables.
Therefore, the general optimization problem becomes one of maximizing (minimizing) the objective function [9] subject to some
model and constraint equations describing the quantitative relationships between the state and the control variables:

x_ ¼ f ðxðt Þ, uðt Þ, t Þ [10]

12 Polymerization Reactors

cðxðt Þ, uðt Þ, t Þ  0 [11]

For the finite time optimal open-loop control problem, the objective function can take the following general form:
ð tf
   
J ¼ h x tf , tf + gðx ðt Þ,uðt Þ, t Þdt [12]
0

where h is a nonlinear scalar function of the final state variables and g is a time-varying function of the state, x(t), and control, u(t),
variables. Accordingly, given a process model [10] and the constraints [11], one can calculate the open-loop optimal trajectory
policies for batch and semibatch polymer reactors that minimize (maximize) the objective function [12].
Despite the rich literature on the open-loop dynamic optimization of polymerization processes, there are a limited number of
successful applications on the closed-loop control of industrial polymerization reactors. There are several valid reasons for the lack
of closed-loop ‘quality’ control applications, more specifically the following:

• Polymerization processes are highly nonlinear, containing a large number of time-varying kinetic and transport parameters.
• Modeling of polymerization systems typically leads to nonlinear integrodifferential equations. Their infinite-dimensional
nature does not allow their direct use for the design of nonlinear controllers that can be easily implemented in real time.
• Online measurement of MWD, CCD, LCB, PSD, etc., is very difficult, and the available secondary measurements are often
insufficient for the accurate inference of the entire molecular weight and/or PSD.
• Formulation of a meaningful objective function in terms of distributed molecular and morphological properties (e.g., MWD
and PSD) is not easy, especially when the end-use properties of the product need to be controlled. Furthermore, the controlled
variables react in opposite directions to changes in the manipulated variables.
However, recent advances in online measurements and control theory have made possible the nonlinear optimization and control
of molecular weight and PSDs in particulate polymerization processes. In particular, for semibatch processes, different control
strategies, including batch-to-batch, online model-based control, and midcourse correction, have been applied to control the MWD
and PSD in a polymerization reactor. On the other hand, for continuous particulate polymerization processes, online soft sensors
for ‘polymer quality’ monitoring, optimal grade change transition, and scheduling are some of the proposed tools for maximizing
polymer productivity and improving product quality while maintaining safe reactor operation.

References
1. Plastics Europe. Plastics-The Facts: An Analysis of European Latest Plastics Production, Demand and Waste Data. http://www.plasticseurope.org.
2. Choi, K.-Y. In Handbook of Vinyl Polymers: Radical Polymerization, Process, and Technology; Mishra, M.; Yagci, Y., Eds.; 2nd ed.; CRC Press: Boca Raton, 2009; pp 347–368,
Chapter 12.
3. Nauman, E. B. In Handbook of Polymer Reaction Engineering; Meyer, T.; Keurentjes, J., Eds.; Wiley-VCH: Weinheim, 2005; pp 533–552, Chapter 10.
4. Touloupides, V.; Kanellopoulos, V.; Pladis, P.; Kiparissides, C.; Mignon, D.; Van-Grambezen, P. Chem. Eng. Sci. 2010, 65, 3208–3222.
5. Pladis, P.; Kiparissides, C. J. Appl. Polym. Sci. 1999, 73, 2327–2348.
6. Dompazis, G.; Kanellopoulos, V.; Touloupides, V.; Kiparissides, C. Chem. Eng. Sci. 2008, 63, 4735–4753.
7. Hill, C. G.; Root, T. W. Introduction to Chemical Engineering Kinetics and Reactor Design, 2nd ed.; Wiley, 2014.
8. Kiparissides, C.; Krallis, A.; Meimaroglou, D.; Pladis, P.; Baltsas, A. Chemical Engineering Technology 2010, 33, 1754–1766.
9. Kiparissides, C. Chem. Eng. Sci. 1996, 51, 1637–1659.
10. Ray, W. H. Can. J. Chem. Eng. 1991, 69, 626–629.
11. Hulburt, H. M.; Katz, S. Chem. Eng. Sci. 1964, 19, 555–574.
12. Kiparissides, C. J. Process Control 2006, 16, 205–224.
13. Himmelblau, D. M.; Bischoff, K. B. Process Analysis and Simulation: Deterministic Systems; John Wiley & Sons, Inc.: New York, 1968
14. Donohue, M. D.; Prausnitz, J. M. AICHE J. 1978, 24, 849–860.
15. Chapman, W. G.; Gubbins, K. E.; Jackson, G.; Radosz, M. Ind. Eng. Chem. Res. 1990, 29, 1709–1721.
16. Wei, Y. S.; Sadus, R. J. AICHE J. 2000, 46, 169–196.
17. Sanchez, I. C.; Lacombe, R. H. J. Phys. Chem. 1976, 80, 2352–2362.
18. Kanellopoulos, V.; Mouratides, D.; Pladis, P.; Kiparissides, C. Ind. Eng. Chem. Res. 2006, 45, 5870–5878.
19. Dompazis, G.; Kanellopoulos, V.; Kiparissides, C. Macromol. Mater. Eng. 2005, 290, 525–553.
20. Kiparissides, C.; Papadopoulos, E.; Morris, J. In NATO-ASI on Methods of Model Based Process Control; Berber, R., Ed.; Martnus Nijhoff Publishers: Dordrecht (Netherlands),
1995; pp 495–530.