Sie sind auf Seite 1von 31

Znd. Eng. Chem. Res. 1994,33, 449-479



Gas Phase Ethylene Polymerization: Production Processes, Polymer Properties, and Reactor Modeling

Tuyu Xie, Kim B. McAuley,' James C. C. HSU,and David W. Bacon

Department of Chemical Engineering, Queen's University, Kingston, Ontario, Canada K7L 3N6

A review of relevant macroscopic and microscopic processes of gas phase ethylene polymerization, both chemical and physical, is given. The commercial technology development of gas-phase ethylene polymerization processes is illustrated through a selective survey of the patent literature. Both advantages and disadvantages of gas phase polymerization processes are addressed, and the challenges of laboratory studies of gas phase polymerization are also outlined. Physicochemical phenomena of ethylene polymerization using heterogeneous catalysts are discussed, including examination of catalyst preparation, polymer morphological development, and elementary chemical reactions. Metallocene-based catalysts and their kinetic performance for olefin polymerizations are also discussed. The current state of the art for reactor modeling of polymerization rate, molecular weight development, reactor dynamics, and resin grade transition strategies is illustrated on the basis of the most recent academic studies. Finally, relationships between resin properties and polymer microstructures as well as characterization methods are described briefly. In particular, temperature-rising elution fractionation technology is emphasized for characterization of ethylene copolymers. The fundamental issues involved in gas phase ethylene polymerization and the& interrelationships are also discussed in some detail.


1. Introduction


2. Gas Phase Polymerization Processes



Commercial Gas Phase Polymerization Processes



Experimental Methods






Catalysts for Gas Phase Polymerization


3.2. Polymer Particle Morphology Developments



Chemical Reactions



Reactor Modeling



Kinetic Modeling



Dynamic Process Modeling and Control


5. Polymer Properties and Characterization



Physical and Mechanical Properties



Polyethylene Characterization





7. References


1. Introduction

Polyethylene (PE) is the largest synthetic commodity polymer in terms of annual production and is widely used throughout the world due to its versatile physical and chemical properties. The American Society for Testing and Materials (ASTM)has classified PE into four groups:

I (low density) at 0.910-0.925 g/cm3, I1 (also low density) at 0.926-0.940 g/cm3, I11 (high-density copolymers) at


g/cm3, and IV (high-density homopolymer)

at 0.960g/cm3and above (Redman,1991). In the literature,

however, PE is normally classified as low

at 0.910-0.930 g/cm3and high density (HDPE) at 0.931-

0.970g/cm3. Low-densitypolyethyleneis further classified as low-density polyethylene (LDPE) and linear low-density

polyethylene (LLDPE) based on polymer chain micro-

structure and synthesisprocesses. Accordingto the figures

reported in Mod. Plast. (19931,polyethylene production in the United States alone was over 10 million tons in

1992. The annual production of PE in Europe is about

9 million tons (Redman, 1991). The current annual worldwide capacity for PE production is over 30 million tons. Figure 1shows the US. PE production profile over the past-decade. Although the annual rate of increase slowed down slightly at the end of the 1980s, the average annual increase rate is about 894 for HDPE and about 5 % for LDPE and LLDPE for the past decade. Consumption of PE is still rising through the 1990s with development of synthesis and processing technology. The main markets and applications of LLDPE and HDPE are summarized in Figure 2 after James (1986) and Foster (1991). Polyethylene is commercially produced exclusively by continuous processes. On the basis of polymerization mechanisms and reactor operating conditions, PE pro-

duction processescan be classified into at least five process

categories as shown in


density (LDPE)

Table 1. Among them, the gas phase

polymerization process is the most recently developed and also the most versatile. Since its emergence, this process has been challenging other existing processes for market share, particularly, for production of LLDPE, due to its economic and technological advantages. Many excellent

reviews of ethylene polymerization processes have been published (Vandenbergand Perka, 1977;Short, 1981;Choi and Ray, 1985a; Nowlin, 1985; James, 1986; Beach and

However, the fundamental issues involved

Kissin, 1986).

0 1994 American Chemical Society

450 Ind. Eng. Chem. Res., Vol. 33, No. 3, 1994







zation can be traced back to the 1950s. Dye (1962),whose patent was filed in 1957,was perhaps the first to adopt a fluidized bed reactor for gas phase olefin polymerization. The original reactor consisted of three concentric super- imposed vertical sections. Polymer particles were dis- charged through an extruder, which was connected to the bottom section of the reactor. The reactor was operated at a pressure of 30 atm and a temperature of 100 "C for ethylenepolymerization. Goins(1960)carried out ethylene copolymerizationin a countercurrent fluidized bed reactor in the presence of inert diluent gas. In this process, polymer particles are passed downward in the reactor and monomer mixed with diluent gas is passed countercur- rently upward in the reactor and monomer mixed with diluent gas is passed countercurrently upward through a series of vertical fluidized bed reaction zones. The reaction zones can be controlled independently by taking off-gas from the last reaction zone, cooling it, and recycling portions of such off-gas to each of the reaction zones. Both patents (Dye, 1962;Goins, 1960)were assigned to Phillips Petroleum Company. Schmid et al. (1967) carried out ethylene polymerization in a stirred fluidized bed reactor. In this configuration, polymer particles are moved in the direction of monomer flow by stirring, and reaction heat is removed by cooling the walls of the reactor, by the gas stream, and by introduction of liquified monomers. The patent of Schmid et al. was assigned to BASF. The benefits of gas phase polymerization were recognized by these pioneer inventors. Although these inventions have not been directly applied in commercial gas phase polymer- ization processes, the fundamental ideas demonstrated by these inventors provided the foundation for the later commercial gas phase process development. The first commercial gas phase polymerization plant using a fluidized bed reactor was constructed by Union Carbide in 1968 at Seadrift, TX (Rasmussen, 1972; Batleman, 1975;Burdett, 1988). This process was devel- oped initially for HDPE production. The success of this novel technology led to the extension of the process to LLDPE, which was produced initially on a commercial reactor in 1975 (Davis, 1978;Burdett, 1988). The Union Carbide gas phase process, commonly called UNIPOL, has been licensed worldwide with more than 25 licenses operating in 14 different countries (Burdett, 1988). Pro- duction of LLDPE using gas phase processes is more difficult than production of HDPE because the difference between themelting point and polymerizationtemperature is much narrower for LLDPE. The catalyst types and equipment design developed for HDPE cannot be used to produce LLDPE because of the potential for agglomeration of polymer particles. Hence, significant engineering and chemistry research has been required to assure the success of gas phase LLDPE production. According to Karol (19831,the keys to the success of the UNIPOL technology for LLDPE production are the proprietary catalysts that operate at low pressure and low temperature and which are suitable for use in a gas phase fluidized bed reactor. Union Carbide Corporation received the 1979Kirkpatrick Chemical Engineering Achievement Award in recognition of the innovation of the UNIPOL process (Chem. Eng.,


The UNIPOL processhas been described in severalUS. Patents (Miller, 1977;Levine and Karol, 1977;Karol and Wu, 1978; Wagner et al., 1981; Jorgensen et al., 1982).A simplified flow diagram is shown in Figure 3. The fluidized bed reactor consists of a reaction zone and a disengagement zone. The reaction zone has a height to diameter ratio of




e e



6000 -


2ooo y

.""" 80











TIME, year

Figure 1. US. polyethylene production profile (data from Mod. Plast. (198C-1993)).







- 0.95








0 0.93








0.90 1





MELT INDEX, 9/10 min



Figure 2. Markets and applications for low-pressure polyethylene.

in gas phase polymerization processes have not been comprehensively discussed. With development of com- mercial gas phase polymerization processes, the impor- tance of more complete understanding of gas phase polymerization processes has been recognized by the academic community. Significant experimental and re- actor modeling work has been carried out in recent years. In the present review,the authors focus on the main issues involved in gas phase ethylene polymerization processes, includingcommercialproduction technologydevelopment, physicochemical phenomena, kinetictdynamic reactor modeling and control; polymer properties and character- ization, as well as their interrelationships.

2. Gas Phase Polymerization Processes

The distinguishing characteristic of gas phase polym- erization 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. Hence, gas phase polymerization is also called dry polymerization in some patents (Dormenval et al., 1975; Havas and Mangin, 1976). In this section, commercial gas phase polymerization processes and ex- perimental studies of gas phase polymerization are dis- cussed.

2.1. Commercial Gas Phase Polymerization Pro-

Table 1.

Polymerization Processes and Reactor Operating Conditions

Ind. Eng. Chem. Res., Vol. 33, No. 3, 1994






gas phase


bulk process


slurry polymn


reactor type

tubular or autoclave



loop or CSTR

fluidized or


stirred bed


press., atm






temp, "C






polymn mech

free radical





loci of polymn

monomer phase

monomer phase




density, g/cm3






melt index, g/10 min







Doak (1986)

Grunig and Luft


James (1986)

Choi and Ray (1985a);

James (1986)



al. (1989)

Short (1981)







an essentially constant temperature under steady-state conditions. To increase heat removal capacity and pro- ductivity, UNIPOL reactors can be operated with an inlet gas temperature at the bottom of the fluidized bed which is below the dew point temperature of that gas. Recycle gascondensesin the external cooler,andthe liquid droplets revaporize upon entry into the bed. It has been observed that droplets pass through the distributor as a mist and quickly wet the surface and pores of the polymer particles. The liquid vaporizes very quickly above the distributor.

There is a dramatic change in the gas flow pattern in the gas feed region below the distributor. However, there is no apparent effect on bulk circulation patterns of bubbles

G and resin particles above the gas distributor (Burdett,





Figure 3.

process-UNIPOL (Miller,1977;Levine and Karol,1977;Karol and Wu, 1978; Wagner et al., 1981; Jorgensen et al., 1982).

height ratio of about 1-2. To maintain a viable fluidized

bed, superficial flow through the bed is about 2-6 times the minimum flowrequired for fluidization. It is essential that the bed always contain polymer particles to prevent the formation of localized "hot spots" and to entrap and distribute the powdery catalyst. On startup, the reaction zone is usually charged with a base of polymer particles before gas flow is initiated. The catalyst is stored in a catalyst feeder under a nitrogen blanket. Catalyst is injected into the bed at a rate equal to its consumption rate at 114 to 314 of the

height of the bed.

bed is essentially equal to the catalyst concentration in the product, namely on the order of about 0.005-0.50% of bed volume. Fluidization is achieved by a high rate of gas recycle to and through the bed, typically on the order of about 50 times the rate of feed of make-up gas. The pressure drop through the bed is typically on the order of 1 psig. Make-up gas is fed to the bed at a rate equal to the rate at which polymer particles are withdrawn. A gas analyzer, position above the bed as shown in Figure 3, determines the composition of the gas being recycled, and

Catalyst concentration in the fluidized

Union Carbide gas phase ethylene polymerization

the make-up gas composition is adjusted accordingly to maintain an essentially steady-state gaseous composition within the reaction zone. The gas which has not been consumed in the bed passes through the enlarged disengagement zonewhere entrained particles drop back into the bed. Particle entrainment is further reduced by a cycloneand a filter to avoid deposition of polymer on heat-transfer surfaces and compressor


Polymerization heat is removed by a heat ex-

changer before the recyclegas is compressed and returned to the reactor. The fluidized bed can maintain itself at

The distribution plate at the bottom of a reactor plays an important role in the operation of the reactor. As the polymer particles are hot and possibly active, they must be prevented from settling to avoid agglomeration. Main- taining sufficient recycle and make-up gas flow rate through the distributor to achieve fluidization at the base of the bed is very important in fluidized bed polymerization reactor operation. The polymer particles are withdrawn close to the distributor through sequential operation of a pair of timed valves, defining a segregation zone as shown in Figure 3. The reactor is operated at a temperature below the melting point of the polymer particles. For HDPE production, the operating temperature is 90-110 "C. An operating temperature of about 90 "C or lower is preferred for production of LLDPE, which contains about 15 mol % of one or more of the Ca to c6 &-olefins (Levine and Karol, 1977). The reaction zone and disengagement zone of the reactor shown in Figure 3 are connected by a transition section having sloped walls. During polymerization, some fine particles fromthe disengagementzonefallonto the sloped walls of the transition section. These fine particles build up during reactor operation. Since the fine particles contain active catalyst, they further polymerize and form solid sheets which can grow until they block recycle gas flow or slide off the sloped walls into the polymerization zone. In the polymerization zone, the solid sheets block the flow of gas and cause fusion of the polymerization particles. Thus, large chunks of polymer can be formed which can block the entire polymerization zone unless the reactor is shut down and the sheets areremoved. To avoid this problem, Union Carbide designed a modifiedfluidized bed reactor shown in Figure 4 (Brownet al., 1981). In this process, the enlarged disengagement zone, cyclone, and filter are eliminated. Polymerization heat is removed by an internal heat exchanger. Reactor temperature is controlled by adjusting the temperature or flow rate of coolant. According to the patent (Brown et al., 1981), ethylene copolymers with densities of 0.91-0.97 gIcm3 and melt indexes of 0.1-100 or more can be produced using

452 Ind. Eng. Chem. Res., Vol. 33, No. 3, 1994










Figure 4. Gas-phaseethylene

heat exchanger (Brown et al., 1981).

polymerizationprocess with



this process. Compared with the process shown in Figure 3, this modified process has the following advantages:

1. The formation of sheets on the sloped walls of the transition zone is eliminated so that the frequency of reactor stoppages is reduced.

2. The depth of the bed in the polymerization zone can

be varied over a wide range allowing a greatly increased

rate of reactor output with good operation.

3. A smaller initial charge of powdered material is

required to start up successfully.

4. The fabrication cost of a fluidized bed reactor is


5. Gas velocity can be much lower using an internal

heat exchanger, thus reducing operation energy.

6. Since heat removal is independent of gas mass flow

rate, the reactor pressure can also be decreased to a limit defined by the polymerization kinetics.

7. The internal heat exchanger also acts as a sets of

baffles, and tends to deter the migration of bubbles to the center of the bed and to increase mixing near the reactor walls.

8. Recycle gas is at the same temperature as the entire

fluidized bed. Despite these advantages, there are no published reports on the application and performance of this type reactor. About 1964BASF first developed a gas phase ethylene polymerizationprocessusing Phillips catalysts (Wisseroth, 1969;Sailorsand Hogan, 1981),and a commercialgasphase HDPE plant was built in Germany in 1976(Choiand Ray, 1985a). BASF uses a continuous stirred-bed reactor for gas phase ethylene polymerization as shown in Figure 5 (Trieschmann et al., 1977; Choi and Ray, 1985a). The reactor is equipped with an anchor agitator and is operated at higher pressure and temperature (35atm, 100-110 OC) than those employed in UNIPOL. Polymer is discharged through a cyclone. Unreacted monomer and the heat- transfer agent, such as cyclohexane,are removed through the top of the reactor and liquefied by a condenser. The liquefied recycle monomer and cyclohexane are fed into the reactor through a valve located at the bottom of the reactor, which is controlled by a thermostat. Liquefied monomer and cyclohexane can be fed into the reactor by a metering pump or by provision of an adequate static pressure differencebetween condenser and reactor. Fresh monomer is also fed through the bottom of the reactor. The temperature in the polymerization zone can be maintained at a constant level by controlling the rate of feed of liquefied monomer and cyclohexane. Liquefied



Figure 5.

(Trieschmann et al., 1977; Choi and Ray, 1985a).

BASF gas phase ethylene polymerization process










- I!





Figure 6. AMOCO gas phase polymerization process (Shepard et al.,1976;Jezl et al.,1976;Peters et al.,1976;Jezl and Peters, 1978a,b).

monomer and cyclohexane vaporize in the reaction zone and rise through the bed at a very slow velocity of about 2 cm/s so that the polymer particles are not fluidized. The bed is mixed by agitation with a mean circulation time of 30-100 s. Fresh monomer is added at a rate such that the pressure in the system remains constant. The pressure and temperature in the polymerization zone are main- tained atlevelssuchthat the vaporizedmonomer and heat- transfer agent remain in the gaseous state. Magnesium- supported titanium halides and aluminum alkyl com- pounds and silica-supported modified chromium oxide catalysts are used for HDPE production. The AMOCO (Standard Oil Company) gas phase polymerization process was developed in the mid-1970s (Shepard et al., 1976;Jezl et al., 1976;Peters et al., 1976; Jezl and Peters 1978a,b). A compartmented horizontal stirred bed reactor is used in this process as shown in Figure 6. The reactor is divided into several stirred polymerizationzoneswhich are separated by weirs. Make- up and recycle monomer and hydrogen are fed into each reaction zone along the bottom of the reactor. Catalyst and quench liquid areintroduced into each polymerization zone through spray nozzles at the top of the reactor. Unreacted monomer, hydrogen, and quench vapor, nor- mally isobutane or isopentane, are removed through the top of the reactor to the recycle and separation sections. The weirs extend upward to somewhat above the middle of the reactor so that the polymer bed fills about half the volume of the reactor. As the polymer powder exceeds the weir height, it falls into the adjacent reaction zone in the direction of the polymer discharge. The polymer

discharge vessel contains a postpolymerization zone, wherein a controllable amount of adiabatic polymerization takes place,producingheat which,together with externally added heat, melts the polymer to form easily transferable liquid polymer. Following treatment with water and suitable additives, the melt polymer is finally converted to pellets.

Each polymerization zone is equipped with flat paddles extending transversely from the shaft and making close clearance with the inside wall of reactor to ensure adequate bed mixing. The paddles are so constructed to minimize any foward or backward movement of the bed. The agitation speed is about 5-30 rpm to provide the desired heat and mass transfer, to avoid polymer particle being thrown up into the space above the bed, and to provide

a slow and regular turnover of the entire polymer bed

contained in the reactor. The distinguishing feature of

the AMOCO process, compared with previous processes,

is that it can be operated like a set of CSTRs in series. The

residence time distribution (RTD) of this process almost approaches that of plug flow. Tracer studies on the commercial reactor indicate that the RTD in the AMOCO reactor is roughly equivalent to that of three to five well- stirred tanks in series. This unique feature of the

horizontal reactor permits production of ethylene-pro- pylene copolymers with high impact strength and good stiffness (Brockmeier, 1991). AMOCO and Chisso have been collaborating for commercialization of ethylene- propylene gas phase polymerization processes (Krieger,


Soon after the Union Carbide process was introduced, Naphtachimie developed a gas phase polymerization process using a fluidized bed reactor (Dormenval et al., 1975; Havas and Magin, 1976). A commercial-scale fluidized bed plant with an initial capacity of 25 000 t/y was commissioned in 1975 (Delgrange, 1987). The gas phase polymerization technology developed by Naph- tachimiewas later transferred to BP Chemicals(Delgrange, 1987; Brockmeier, 1987). After the amalgamation of Napthachimie intoBP Chemicals, high priority was given to further development of the novel gas phase polymer- ization process, as can be confirmed by recent patents

assigned to BP Chemicals (Dumain and Raufast, 1985; Collomb-Ceccarini and Crouzet, 1986;Bailly and Speak- man, 1986;Raufast, 1987;Bailly and Collomb,1988;Chinh and Dumain, 1990;Speakman, 1991;Matens and Moterol, 1992;Havas and Lalanne-Magne, 1992). The BP Chem- icals gas phase process now has a capacity of 170 kt/y and can produce both LLDPE and HDPE with a density range from 0.91 to 0.96 (Delgrange,1987;Redman, 1991;Speak- man, 1991). The major distinction of the BP Chemicals process from the UNIPOL process is that prepolymer particles rather than catalyst particles are fed into the fluidizedbed reactor. The BP Chemicalsgasphase process

is a continuoustwo-stagepolymerization process,asshown

in Figure 7 (Chinh and Dumain, 1990; Redman, 1991; Matens and Moterol, 1992; Havas and Lalanne-Magne, 1992). This process uses a combination of a stirred tank reactor and a fluidized bed reactor in series. During conventional gas phase reactor operation, the catalyst and the cocatalyst may be brought into contact either prior to their introduction into the fluidized bed, or in the interior of the reactor. Whichever method is employed, the polymerization reaction always starts up very abruptly and attains a maximum rate soon after the catalyst system is introduced into the fluidized bed. It is therefore in the initial phase of polymerization that the risks of hot spots forming and grains bursting into fine

Ind. Eng. Chem. Res., Vol. 33, No. 3, 1994









Figure 7. BPChemicalstwo-stagegasphase ethylenepolymerization process (Chinh and Dumain, 1990;Redman, 1991;Maters and Moterol, 1992;Havas and Lalanne-Magne,1992).

particles are greatest. Hot spots may lead to formation

of agglomerates and to settling of polymer inside the fluidized bed (Collomb-Ceccarini and Crouzet, 1986). Furthermore, during polymerization small variations in the feed rates of catalyst, monomer, and comonomer or in the withdrawal rate of polymer will also cause an unex- pected increase in the quantity of heat evolved by the


these small variations can cause hot spots in the reaction

bed and formation of agglomerates by melting polymer. Such variations can therefore make it difficult to obtain

a polymer of consistent quality, in particular, of constant

molecular weight and particle size (Chinh and Dumain, 1990; Matens and Moterol, 1992; Havas and Lalanne- Magne, 1992). These problems can be eliminated by adopting a prepolymerization stage, as shown in Figure 7.

Prepolymerization gives advantages in polymer particle size control and control of the catalyst activity in the fluidized bed reactor. Prepolymerization can be carried out in a liquid hydrocarbon medium or in a gas phase stirred reactor at temperatures from 40 to 115"C. Catalyst

is introduced into the prereactor in the form of dry powder

or in suspension in a liquid hydrocarbon. Prepolymer- ization is carried out to a conversion wherein the pre- polymer contains 0.002-10 millimol of transition metal/g of polymer. The diameter of the prepolymer is in the range from 200 to 250 pm (Bailly and Speakman, 1986; Chinh and Dumain, 1990; Martens and Moterol, 1992; Havas and Lalanne-Magne, 1992). Prepolymer is fed into the fluidized bed reactor through a metering feed device. The fluidized bed reactor operates at a superficialvelocity of approximately 0.5 m/s, 2-8 times the minimum fluid- ization velocity (Dumain and Raufast, 1985). To avoid induction time at startup, the prepolymer is treated with triethylaluminum for polymerizationwith chromiumoxide as catalyst (Durand and Morterol, 1986a,b; Bailly and Speakman, 1986). Tocreate more porosity,the prepolymer is also treated with n-hexane to remove low molecular weight polymer (wax) (Bailly and Speakman, 1986). Monomer, comonomer, hydrogen, and inert gas are fed through the bottom of the fluidized bed. Some comonomer and inert volatile hydrocarbon, such as isopentane, are introduced into the inlet line of the heat exchanger to avoid fine particles depositing on heat exchanger surfaces and compressor blades. The ratio of comonomer to monomer partial pressure is kept constant (normally 0.1- 0.2)) in the reactor (Chinh and Dumain, 1990). Toavoidpressure fluctuations during polymer discharge, BP Chemicals developed a continuous discharge device

If the heat cannot be removed efficiently,


Ind. Eng. Chem. Res., Vol. 33, No.3,1994

Table 2. Ethylene Gas Phaw Polymerization Prowsws and Reactor Operating Conditions


Union Carbide



BP Chemic&


fluidized bed

stirred bsd

horizontal stirred bsd

stirred reactor and fluidd bed



supportedTi. V and CrCOa catalyat


supported CrCOS


~~pportsdTi and


supported Ti and


-60 for prepolymn

press., atm

temp. OC










1-buteneor 1-hexene


propylene or 1-butene

I-butene or I-hexene




H,and temperature


MWD. Mw/Mm


narrow tu broad







palymer particle size. pm




Rasmuassn (1972):

Trieaehmann et al. (1977):

Shepard et al. (1976);


Nidetti et al. (1988)

Choi and Ray (1985a)

Jezl and Petern (1978a,b)

and SpsaLman (1986)

consisting of continuously rotating plug valves shown as

A and B in Figure 7.

each other in such a way that the two valves do not open

at the same time, rotating at a speed of about 0.5-1.0

The volume of the vessel between the valves is between

These valves are connected with





0.2% and 1% ofthevolumeofthefluidizedsolidcontained



the reactor (Raufast, 1987).




One of the disadvantages of gas phase polymerization isthat the reactor operating temperature must be lower than the meltingpoint ofthepolymerproduced. Recently, Baillyand Collomb(1988;1990)found that agglomeration ofthepolymerparticlescommencesat higher temperatures

whenpulverulent inorganicparticles areemployed. Thus, higher polymerization temperaturescan be used to produce LLDPE. Thepulverulent inorganicparticles, suchas silica

oralumina,arechemicallyinerttocatalystsandmonomers. The inert particles have porosity 0.2-2.5 mL/g, surface

area 20-90 mz,and diameter 0.5-20fim. The particle size is 5(t500 times smaller than the mean diameter of prepolymer particles. Accordingto the patent (Baillyand Collomb, 1988,LLDPE with density 0.860-0.910g/cm3 can be produced with 0.005-0.2 wt % inert particles based

on polymer in the reaction zone. The inert particles are

mixed with prepolymer particles and then fed into the reactor. Rhee et al. (1991)also used inert particles, e.g., carbon black, silica, and clays, in the production of ethylene-propylene rubbers and found that the reactor can be operated at temperatures higher than softening points of copolymers when amounts of about 0.3-50% by weight of inert material are employed. These inventions suggest that the soft polymer particles can he stabilized using inert particles. However, the morphology devel- opment of polymer polymerized at or near melting point has not been reported. Typical reactor operating conditions and production capabilities of the gas phase processes discussed above are summarized in Table 2. Comparedwith slurry and solution processes,gasphase processes have many distinct advantages. Slurry-based processesproducePEcoveringa broad rangeof melt index; however, the range of densities attainable is limited. As density decreases, resin solubility in the dilute increases. At a density of about 0.930 g/cm3 sufficient dissolution occursto 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 somepoint, the increased viscosity limits reactor operabilityand productivity. Thus, solution processesare not suitable for making high molecular weight PE. Production capabilities of slurry and solution processes are shown schematically in Figure 8, after James (1986).


5 0.93









MELT INDEX, 9/10 min

Figure 8.

Production capabilities of aolution and slurry polymer-

ization processes.

0910 L

" "1

" I"

I U"


MELT INDEX, g/10 min


FiKure 9.

Production capability of gas phase polymerization

processes (James. 1986;Speakman, 1991).

Gasphase processes,which donot involveanyliquid phase in the reaction zone, are not constrained hy solubility and viscosity. Hence, a gas phase process can produce a complete range of PE with densities of 0.91-0.97 g/cm$ and melt indexes of <0.01 g/10min up to 200 g/10 min, asshowninFigure9(James, 1986,Speakman,1991).Other major advantages of gas phase processes over liquid phase processes include production of high comonomer content polyethylenes, such as high impact strength ethylene propylenecopolymers,ethylene-propylenediene rubbers. Comonomercontent is limited in the liquid phase processes because of problems with viscosity effects from dissolved copolymer and comonomer (Brockmeier, 1991;Rhee et al.,1991;Burdett. 1992). Furthermore,gasphaseprocesaes can reduce construction costs up to 30% and operating costs up to 35% over conventional liquid phase processes (Burdett, 1992;Krieger, 1992).

Table 3.

Process and Low-Pressure Gas Phase Process

Comparison of High-pressure Free-Radical

high-pressure process

tubular or autoclave reactor

high operating pressure, -3400 atm high reaction temperature, -300 OC high energy requirement high capital cost limited to low-densityPE

no comonomerrequired (homopolymer)for LDPE production catalyst is less sensitive to impurities in the feedstock low raw material cost can produce ethylene and vinyl acetate copolymer polymer with long-chain branching


fluidized bed or stirred bed reactor low operating pressure,



low reaction temperature, 80-110 OC low energy requirement low capital cost can produce both high- and low-densityPE up to l0-15% comonomer required (copolymer)

catalyst is very sensitive to impurities high raw material cost currently a limited range of comonomers mostly linear polymer chains with short-chain branching

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


2. The poor heat-transfer efficiency of the gas phase is

a disadvantage. Hence,additional inert heat transfer agent

is required to maintain stable reactor operating conditions

at high production rates.

3. There is a possibility of sintering and agglomeration

of the polymer particles due to formation of local hot spots

when a high-activity catalyst is used. 4. As polymerization progresses, fine particles will deposit on heat-transfer surfaces, compressor blades, and sloped walls of the reactor.

5. Sensitivity to variation in operating conditions, such

as flow rates of catalyst and monomer feed and polymer discharge, is a disadvantage. Gas phase processes offer many advantages over con- ventional high-pressure processes. The conventionalhigh- pressure free-radical process can produce only low-density polyethylene due to the nature of its polymerization mechanisms. Table 3 compares some features of the high- pressure LDPE process and the low-pressure LLDPE gas phase process (Imhausen et al., 1981; Karol, 1983).

Significant progress has been made in overcoming the disadgantages of gas phase polymerization and in im- provingheat removal capacity,as discussedabove. Overall, gasphase polymerizationprocesseshave had a great impact on polyethylene production technologydevelopment.

2.2. Experimental Methods. Despite the commercial

success of gas phase ethylene polymerization technology, the public literature contains no accounts of fundamental scale-up studies of gas phase processes. There is a need, therefore, for a comprehensive understanding of detailed polymerization behavior in gas phase polymerization. A challenge for academic researchers studying gas phase polymerization of ethylene is how to scale down commercial processes for experimental laboratory studies. Kinetic studies of ethylene gas phase polymerization were first carried out by Edgecombe (1963) and Lipman and Norrish (1963) using conventional Ziegler-Natta catalysts. Wisseroth (1969) reported on gas phase po- lymerization of ethylene using Phillips catalysts. With the development of commercial gas phase polymerization processes, more university-based researchers have at-

Ind. Eng. Chem. Res., Vol. 33, No. 3, 1994








Figure 10. Laboratory reactors for gas phase ethylene polymeri- zation. (1) Choi et al. (1983, 1985b); (2) Lynch et al. (1991); (3) Dusseault (1991).

tempted to extend experimental studies of gas phase polymerization in recent years. Choi et al. (1983,1985b) designed a 1-L gas phase stirred bed reactor equipped with an anchor-type agitator. Mabilon and Spitz (1985) and Spitz et al. (1988)used a 675-cm3stainless steel reactor stirred by vertical oscillationsfor gasphase polymerization. However, the detailed reactor configuration used in this investigation was not described. More recently, Jejelowo et al. (1990) and Lynch et al. (1991) constructed a 1-L stainless steel reactor with a specially designed paddle agitator. The reactor temperature is controlled using a 26-Loilbath. Toimprovereactor mixingand heat transfer, Dusseault (1991) designed an 800-cm3horizontal stirred reactor with both internal coolingcoiland external cooling/ heating jacket. The agitator is designed in such a way so that the entire polymer bed can be turned over. Weist et al. (1989)used a 20-mm-diameter fluidized bed reactor to study polymer morphology development at the initial polymerization stage. However,the reactor configuration and performance were not described. Figure 10 shows typical reactors used for ethylene gasphase polymerization studies reported in the literature. Because the reactivity of the catalyst is very high, only

a very small amount of catalyst, normally in the order of milligrams,is required for laboratory-scale polymerization using a 0.5-1.0-L reactor. Therefore, inert particles are usually added into the reactor to assist the initial mixing of the catalyst. Glass beads with diameters of 500-5000 pm are often used (Keii, 1972;Doi et al., 1982;Choi et al.,

1983,1985b;Dusseault, 1991). Mabilon and Spitz (1985) and Spitz et al. (1988) used salt (KC1) or polyethylene powder to assist initial mixing. Lynch and Wanke (1991) used 450-pm Teflon powder. Glass beads may further grind the catalyst during the initial stage of polymerization. Saltissensitiveto moisture; hence carefuldrying procedure

is required, but the separation of polymer from salt is very


for initial mixing, it is not good for polymer property analysis because newly produced polymer can not be separated from the initial polymer powder. However, it may be useful for polymerization rate measurements without requiring polymer separation. The laboratory-scale reactors shown in Figure 10 can providesatisfactory temperature controlwithin a restricted operating temperature range. With high-activitycatalysts,

Although polyethylene powder is effective

456 Ind. Eng. Chem. Res., Vol. 33, No. 3, 1994

Table 4.

Ethylene Gas Phase Polymerization

Main Requirements of Industrial Catalysts for

polymerization process requirements

polymer property control

high productivity molecular weight

proper kinetic behavior

catalyst morphology density

control of polymer morphology good comonomer incorporation

easy feed to reactor low cost and reproducible catalyst preparation

molecular weight distribution

chain-branching distribution particle size and bulk density polymer chain unsaturation

the initial polymerization rate is very high, and so the initial heat is difficult to remove with an oil or water bath cooling system. Therefore, reactor temperature control systems shown in Figure 10 are still not adequate for a broad range of reactor operating conditions. To carry out gas phase ethylene polymerization under commercial reactor operating conditions, the reactor must be able to provide fast heat removal to avoid hot spots in the polymerization zone and adequate mixing conditions to disperse a small amount of catalyst powder.

3. Physicochemical Phenomena

In light of the preceding observations, it is evident that

gas phase polymerization involves a complicated physi- cochemical transition from gaseous monomer to solid polymer, although the synthesis process itself is relatively simplecompared with liquid phase processes. In gasphase polymerization, catalyst preparation and polymer particle morphology development play important roles in reactor operation. Hence, it is crucial to understand the relevant microscopic processes of polymerization and their rela- tionships.

3.1. Catalysts for Gas-Phase Polymerization. Ca-

talysis is the most active area of research in ethylene

polymerization with worldwide participation in both industrial and academic laboratories. In recent years, hundreds of publications on polyethylene synthesis and

manufacture have been recorded annually in Chemical Abstracts. Most are related to catalysis studies. The extent and variety of research work dealing with the same problem reflects not only the great interest and extensive commercial applications in this area, but also the com-

plexity of catalysts for ethylene polymerizations.

progress concerning catalysts for ethylene polymerization has been a focus of several recent international conferences (Seymour and Cheng, 1987; Kaminsky and Sinn, 1988; Quirk et al., 1988;and Keii and Soga, 1990;Vandenberg and Salamone, 1992),and excellent reviewshave appeared in the literature (Zakaharov and Yermakov, 1979; Karol, 1984; Hsieh, 1984; McDaniel, 1985; Nowlin, 1985; Hsieh et al., 1987;Barbe et al., 1987;Tait, 1989;Kryzhanovskii and Pranchev, 1990;Dusseault and Hsu, 1993). Catalyst development plays a vital role in the polymerization of ethylene, particularly in successful gas phase polymeri- zation processes (Karol, 1983). For gas phase polymer- ization, although high productivity of the catalyst is a necessary condition, other features, such as copolymeri- zation kineticbehavior and catalyst particle sizeand shape also play important roles in the production of high-quality LLDPE and HDPE. The main requirements of industrial catalysts are summarized in Table 4 (Karol et al., 1987; Dall'Occo et al., 1988; Burdett, 1988). Two major catalyst families have been commonly used in commercial ethylene production: titanium1 vanadium based catalysts (Ziegler-Natta catalysts) and chromium oxidebased catalysts (Phillips catalysts). In the past four


Table 5.

Polymerization (Joraensen et

Properties of Silica Support Used for Gas Phase


1982: Nicoletti et al




preferable range

particle size, pm



surface area, m2/g

at least



pore size, A

at least



concn of catalyst metal, mmol/g of support



decades, research has been focused mainly on titanium and chromium oxide based catalysts (Karol et al., 1988). A typical industrial catalyst consists of active metal, modifiers,and inert support. Porous silicaand magnesium halide are common supports. Table 5 shows typical properties of silica support used for commercial gas phase polymerization (Jorgensen et al., 1982; Nicoletti et al., 1988). Internal pore size distribution is in the range of 80-1000 A, and with the distribution peak in the range of 100-200 A, depending on preparation (Weist et al., 1987). A comparison of the performance of titanium-, vanadium-, and chromium-based catalysts is summarized in Table 6 (McDaniel, 1985;Karol, 1988). The significant difference between Ti- and Cr-based catalysts is that the molecular weight distribution (MWD) of polymers made using Ti catalysts in narrower than that using Cr-based catalysts. High activity or productivity of catalyst means that the catalyst leaves no more than 10 ppm catalyst metal and 30 ppm chloride in the polymer formed. Titanium levels higher than this level lead to polymer color problems, and higher chloride contents result in excessive corrosion problems in processing equipment (Hsieh, 1980, 1984). Accordingto recent patents, the metal content in polymer is very low for the catalyst used in gas phase polymer- izations, e.g., <5 ppm of Ti (Jorgensen et al., 1982), <0.5 ppm of Ti (Collomb-Ceccarini and Crouzet, 1986), and about 2 ppm of Cr (Speakman, 1991). Several methods have been developed for Ti-based catalyst preparation (Jorgensen et al., 1982;Karol, 1984; Collomb-Ceccariniet al. 1986). Three steps are normally involved: (a) formation of a precursor containing Ti, (b) incorporation of the precursor onto the support, and (c) activation of the catalyst. The different methods can be distinguished by the mode of formation of a precursor containing titanium, as summarized in Table 7 together with some examples (Karol, 1984). Method 1 in Table 7 is often used for chromium oxide based catalyst prepa- ration. Methods 2 and 3 are commonly used for Ti-based industrial catalyst preparation (Jorgensen et al., 1982; Collom-Ceccariniet al., 1986;Nicoletti et al., 1988). The catalyst precursor prepared using method 2 in Table 7 is then impregnated onto a porous support such as silica, as shown in Table 5. Impregnation of the support with the precursor is accomplished by dissolving the precursor in the electron donbr compound and admixing the support with dissolved precursors. The solvent is then removed by drying at a temperature of up to 85 "C (Jorgenson et al., 1982; Nicoletti et al., 1988). The catalyst precursor prepared by ball-milling (method 3 in Table 7) has a relatively small diameter, generally less than 50 pm. It is too small to be used directly for gas phase polymerization in a fluidized bed reactor. For this reason, the precursor and cocatalyst are converted to prepolymer particles in a prepolymerization stage as mentioned for BP Chemicals processes. The concentration of titanium in the prepoly- mer is controlled in the range of 4-30 ppm. Prepolymer particle size is about 70-250 pm before it is fed into the fluidized bed reactor (Collomb-Ceccarini et al., 1986). Hence, one of the functions of prepolymerization is to form prepolymer in place of silica as a catalyst support for gas phase polymerization.

Table 6.

Ind. Eng. Chem. Res., Vol. 33, No. 3, 1994

Performance of Titanium, Vanadium, and Chromium Oxide Based Catalysts (McDaniel, 1986: Karol, 1988)









R&l +halocarbon promoter

thermal, 300-1000 OC





Hz response





narrow, 3-6


intermediate-broad, 6-30

a-olefin incorporation




activity decay rate



very low

polymer chain unsaturation


very low

one C=C per polymer molecule

long-chain branching



low concn of long-chain branching

ED is an electron donor.

Table 7.

Catalysts (Karol, 1984)

Operational Methods to Produce High-Activity

Table 8.

Catalyst (Chien et al

Approximate Dimensions of MgCll Supported

1983: Chien. 1987: 1988)

1.chemical anchoring to surface of support



composition MgC12 (HCl treated) MgCldEB MgClz/EB/PC/AlEts

(c)(CKHdoCr/SiO, 2. formation of bimetallic complexes






(a)MgClz + 2TiC4 + 8Poc13- [TiCl~~lZ[Mg(POC13)sl~+~2POC13 (b)2MgC1~+Tic4+ 7THF -[TiCls(THF)l-[Mg2Cl~(THF)61+


MgC12 +Tic4+4CH3COzCzHs-TiMgCl&H&OzCzHs)4
























3. insertion into defects of support (a)MgClz + Tic14+ethylp-toluate -ball-milling (b)MgClz +TiQedioxane -ball-milling 4. formation of high surface area sponge

(a) Tic4 +EtzAlCl-,9-TiC1,.xEtAlC12



bwmyl ether


MgClz with an electron donor (internal Lewis base), such as ethyl benzoate, can produce a disordered crystal

structure and reduce crystalline size to 20-40 A. The

surface area of the support increases to a maximum value of ca. 6.5 m2/g. The surface area of the support further increases through each step of the modification. Table 8 shows approximate dimensions of a MgClz supported catalyst obtained by ball-milling (Chienet al., 1983;Chien, 1987, 1988). Although the composition of commercial catalysts may not be the same as that shown in Table 8, the main features of MgC12 supported catalysts are the same. From Table 8 it can be seen that MgClz supported catalysts have large surface areas and internal pore volumes. Both surface area and pore volume change with modifier composition. Incorporation of Tic4 with such a support results in dispersion or isolation of the titanium centers so that the population of active centers which are accessiblefor polymerization increases dramatically com- pared with that form unsupported conventional Ti cat- alysts. This is the main reason why supported Ti catalysts have productivities so much higher than those of con- ventional catalysts. In addition, magnesium ions can stabilize activetitanium from deactivation, enhance chain- transfer processes (& of PE decreases when Mg/Ti ratio increases), and lead to narrow MWD for polyethylene production (Karol, 1984; Nowlin, 1985). According to Chien (19871, both internal Lewis base and internal alcoholincrease surface areaand pore volume, but they do not compete for the same surface sites of MgC12 crystal because the content of EB on the support is the same after addition of internal alcohol. Therefore, they may adsorb on different planes of MgClz crystal. Alu- minum alkylssuchas triethylaluminum (AlEt3) are usually used as cocatalysts to activate the catalyst. The activating action of AlEt3 is believed to reduce Ti4+ to Ti3+ and alkylation. In addition to the activation requirement, AlEt3 also acts as a scavenger of impurities of the polymerization system. The optimum ratio of Al/Ti is in the range of 10-30 (Dusseault and Hsu, 1993). AlEt3 can alsoreduce MgClzsupportparticle sizeand increasesurface area 2-3-fold if AlEt3 is used as an internal modifier during support preparation (Chien, 19871, as shown in Table 8. Compared with Ti-based catalyst preparation, chro- mium-based catalyst preparation seems to be simpler. Chromium-based catalyst (Phillips catalyst) is usually prepared by impregnating a chromium compound onto a wide-poresilicaand then calciningan oxygenenvironment to activate the catalyst. This leaves the chromium in the

treated solid - [TiCl~~(EtA1C12)o,03(ether)o,oJ Tic4- [MgClz.Mg(OEt)z.Mg(TiC&)l

(c)Mg(0Et)Z+Ti(O-nC,H& +EtAlC12-trimetallic sponge 5. formation of solid solutions by cocrystallization (coprecipitation)


EtMgCl +Tic4-TiCls-MgClz+organic fragments

The patent literature usually gives only the state of the art for preparation of catalysts. The reasons for the choices of different compounds are not explained. To better understand the effect of each component used on the properties of a catalyst, Chien and his co-workers have fully characterized a typical MgC12 supported catalyst:

MgC12/EB/PC/A1Et3/TiC14-A1Et3/MPT,where EB, PC,

and MPT are abbreviations for ethyl benzoate, p-cresol, and methyl p-toluate, respectively (Chien, 1987, 1988, 1992). The major findings of academic studies on MgC12 supported catalysts have been comprehensively reviewed by Dusseault and Hsu (1993). The main reasons that MgCl2 is the most widely used as a catalyst support can be summarized as follows:

1. MgClz has crystalline forms similar to those of Tic13

(conventionalZiegler-Natta catalyst) (Chien,1987;Barbe, 1987; Dusseault and Hsu, 1993). This distinct feature

suggests that MgClz should have the ability to mimic the structure of active Tic13 and the ability to efficiently incorporate TiC14.

2. MgCl2 possesses desirable morphology as a support.

Mercury porosimetry showed that MgClz has a large number of pores with radii > 200 nm (Chien et al., 1983).

The structure of MgClz is sufficiently resistant to fracture by operational manipulation, but weak enough so that its internal structure will be broken down during polymer- ization.

3. MgC12 has a lower electronegativity compared with

those of other metal halides. A metal halide with lower electronegativity than Tic13 will increase productivity of

ethylene polymerization (Soga et al., 1987). MgClz also enhances chain-transfer reactions because Mndecreases with an increase in Mg/Ti ratio (Karol, 1984).

4. MgC12 is inert to the chemicals used for polymeri-

zation. It can be left in the final polymer product without

deashing at the levels used in commercial production. Because of the above features of MgClz, Tic14 can be incorporated with MgClz in different ways. Co-milling of

458 Ind. Eng. Chem. Res., Vol. 33, No. 3, 1994

hexavalent state, monodispersed on the silica surface. Chromium trioxide (CrOa) has been impregnated most commonly (McDaniel, 1985). Cr03chemically anchors to the silica surface to form a new surface composition of silica (Karol, 1984;Hsieh, 1984;McDaniel, 1985;Nowlin, 1985). However,whether the chromium atoms are isolated as chromate, or exist in pairs as dichromates, is still debated in chromiumcatalyst studies. It is believed that chromium trioxide probably binds to the silica as chromate initially, and during calcining some chromate may convert to dichromate (McDaniel, 1985). At least six methods have been described in preparation of Cr03 catalysts for ethylene gas phase polymerization (Baillyand Speakman, 1986). The main distinction among them is due to incorporation of modifiers. After CrOs is impregnated with porous silica, the catalyst is further impregnated with a titanium compound such as titanium tetraisopropoxide. Then, the catalyst is activated by a heat treatment at 350-1000 "C under air for 30 min-5 h. The final catalyst contains 0.05-10% chromium and 0.50- 20% titanium (Bailly and Speakman, 1986; Speakman,


Catalyst prepared following the above procedure is not suitable for direct use in gas phase polymerization because of its long induction time. Cr(VI)/silica catalyst develops polymerizationactivitygradually whenexposedto ethylene monomer. It is believed that Cr(V1)is reduced by ethylene to the Cr(I1) active site. Below 100 "C the induction time becomes longer until at 60 "C there is almost no activity. However, at 150 "C, the catalyst exhibits an immediate and constant activityin solution polymerization(McDaniel, 1985). For gasphase polymerization,the normal operating temperature is between 80 and 100 "C. The Cr-based catalyst has a significant induction period which causes difficulty not only in starting polymerization, but also in maintaining constant polymerization conditions in the reaction zone. When the polymerization is started, the catalyst activity rapidly accelerates, which also makes polymerization conditions difficult to control. Therefore, the catalyst has to be prereduced to eliminate the induction time. CO and alkyls of aluminum, boron, and zinc can be used as reducing agents to eliminate induction time (McDaniel,1985). However,for gas phase polymerization, catalyst with fully activated chromium could lead to the formation of hot spots and cause agglomeration due to poor heat transfer in the reaction zone. Accordingto Bailly and Speakman (1986)and Speakman (1991),the best way to overcome these problems is to convert the catalyst to prepolymer particles through a prepolymerization stage as described above. It should be pointed out that the activationtemperature is a very important parameter in controlling catalyst activity and polymer molecular weight for ethylene polymerization using Cr-based catalyst. Figure 11 shows the effects of activation temperature on the reactivity and molecular weight of polymer (McDaniel, 1985,1988).Here, activity is defined as the inverse of the time needed to produce 5000 g of polymer/g of catalyst. The activity increases with an increase in activation temperature up to maximum at around 925"C, and then declines as sintering destroys the surface area and porosity of the silica base. The weight-average molecular weight decreases with an increase in calcining temperature up to 925 "C; then the trend reverses as the silica begins to sinter. The melt index shows the same type of relationship. The relation- ships shown in Figure 11suggest that the chain termination reaction is very sensitive to the activation temperature for Cr-based catalyst systems.











Figure 11.

polymer properties (McDaniel, 1985, 1988). Reprinted with per- mission from McDaniel (1985). Copyright 1985 Academic Press.

Effecta of activation temperature on reactivity and

The silicasupported catalystsdescribedabovehavebeen successfullyused for PE production. Deashing processes have been eliminated for many years. However,silicadoes remain in the final polymer product as an impurity. This impurity causes the polyethylene to have a poor film appearance. Hence, it is desirable to eliminate silica by use of polymer particles as a support. Research in this area has been reported in recent patent publications (Cana et al., 1992). Polypropylene and polyethylene or their copolymers have been used as supports. The catalyst precursor is bonded on the polymer particles. However, detailed information about how the catalyst is bonded on the polymer particles is not given. The results using polymer supported catalyst are sufficiently promisingthat polymer supported catalysts may replace nonpolymer supported catalysts for the manufacture of polyethylene film product in the near future. The term "supported catalyst" is used in a very wide sense and includes not only systems in which the transition metal compound is linked to the substrate by means of a chemical bond, but also systems in which the transition metal atom may occupy a position in a lattice structure, or where complexation, adsorption, or even occlusionmay take place (Tait, 1989). In commercialapplications, MgC12 supported catalyst is further supported on silica or polymer through prepolymerization. All of these supports play a role in polymerization and in polymer morphology de- velopment. Besides the heterogeneous catalysts discussed above, another catalyst family that may soon be commercialized in gas phase olefin polymerizations is metallocene-based catalysts. The metallocene-based catalysts are also called homogeneousZiegler-Natta catalystsin the literature. The first homogeneousZiegler-Natta catalyst,CpzTiClz/AlEt~- C1 (Cp = cyclopentadienyl) was independently discovered by Natta and Pino (1957) and by Breslow and Newburg (1957). The kinetic behavior of this catalyst was studied by Breslow and Newburg (19591,Chien (19591,and others (Reichert, 1983). The productivity of this catalyst is very low, less than 300 g of PE/(g of Tiehaatm) (Chien, 1959; Chien et al., 1990). However, if cocatalysts (AlEt, or AlEtC12) are treated with water, both productivity and molecular weight (MW) of PE increase dramatically (Reichert and Meyer, 1973;Andresen et al., 1976;Chihlar

et al., 1978). It was believed that aluminoxane plays an important role in the increase in productivity (Andresen et al., 1976;Chihlar et al., 1978;Sinnand Kaminsky, 1980). Significant breakthroughs however occurred in the early 1980swhen Kaminsky and his co-workersfound very high productivity for ethylene polymerization using CpzZrCl2 as catalyst and methylaluminoxane (MAO) as cocatalyst (Sinnetal., 1980;Kaminsky et al., 1983;Kaminsky, 1983; Herwigand Kaminsky, 1983;Kaminsky and Luker, 1984). Furthermore, Ewen (1984) and Kaminsky et al. (1985) demonstrated that chiral rac-Et[IndH412TiClz and rac- Et[Ind]zZrClz/MAO (Ind = indenyl) catalysts are capable of stereospecific polymerization of a-olefins. Since the mid-l980s, metallocene-based catalysts have become of academic and commercial interest. The level of industrial research activities in metallocene catalysis has been tremendous. More than 50 companies worldwide have been involved in metallocene catalyst studies. Among them, Mitsui Petrochemical, Indemitsu, Exxon, Mitsui Toatsu, Hoechst,Asahi Chemical,and Dow Chemical are the leading companies in terms of the number of patents issued. To date, over 50 metallocene catalysts have been described based on Ti, Zr, and Hf metallocenes

(SRI, 1993). Exxon Chemical was the first company to apply metallocene catalysis to polyolefin production on a commercial scale. The company built a 15000 t/y high- pressure autoclave reactor unit at the Baton Rouge site. This plant was started up on June 1991and has produced specialty linear ethylene-based polymers called EXACT resins: about 20 grades with a range of densities from 0.865 to 0.935 g/cm3 (Mod. Plast., 1991; SRI, 1993; Montagna and Floyd, 1993; Rogers, 1993a). To further increase product range, Exxon and Mitsui Petrochemical have been collaborating forcommercializationof gas phase processtechnologyusing metallocene catalysts (SRI,1993; Rogers, 1993a). To further increase product range, Exxon and Mitsui Petrochemical have been collaborating for commercialization of gas phase process technology using metallocene catalysts (SRI, 1993; Rogers, 1993b). The gas phase process is expected on stream in the mid-1990s and will be operated at temperatures of 80-90 "C and will produce 100 000 t/y of ethylene copolymers with densities ranging from 0.86 to 0.975 g/cm3. Dow Chemical, in competition with Exxon, has developed a new family of polyolefins using Constrained Geometry Catalyst Tech- nology, This technology was commercialized under the trade name INSITEin mid-1993(Rotmanand Wood, 1993; Schwank, 1993; Lindsay, 1993;Mod. Plast.,1993; Story and Knight, 1993). The distinct characteristic of INSITE polyethylenes is improvement in physical properties without sacrificingprocessability as a result of a significant amount of long-chain branching formed in polymer chains (SRI, 1993; Stevens, 1993). Thus, for the first time, polyethylenes made through coordination polymerization have the melt rheologicaladvantages currently obtainable only in free-radical-polymerized high-pressure LDPE. INSITE resins are now produced in solution processes (Rotmanand Wood, 1993). However,Dow claims to have successfully carried out polymerizations using its metal-

locene catalysts in other

pressure, gasphase, and slurryprocesses (SRI,1993).Mobil Chemical has also carried out ethylene and l-hexene copolymerization using metallocene catalysts in a pilot- scale fluidized bed reactor. Commercialization of a gas phase process using a metallocene catalyst is on the way (Furtek, 1993). In addition to the companies above, there are others in Japan, Germany, and Belgium that appear

processes as well, including high-

Ind. Eng. Chem. Res., Vol. 33, No. 3, 1994


to be well on the way to commercialization of metallocene technologiesfor polyolefins(SRI,1993;Kuber, 1993;Payn,


While polyolefincompanies worldwidejostle in the race to commercialize metallocene-based technologies, the fundamental studies of metallocene catalysts carried out in companies are highly confidential. In academic insti- tutes the kinetic studies of metallocene catalyzed olefin polymerization were mainly carried out in solution pro- cesses. The following metallocene-based catalysts have been studied extensively: CpZMClZ/MAO (M = Ti, Zr, Hf) (Herwig and Kaminsky, 1983;Kaminsky et al., 1983; Kaminsky and Luker, 1984;Kaminsky and Hahnsen, 1987; Chien and Wang, 1988; Mallin et al., 1988; Ahlers and Kaminsky, 1988;Tsutaui and Kashiwa, 1988;Tait, 1988; Chien and Wang, 1989;Chien and Wang, 1990;Kaminaka and Soga, 1992; Uozumi and Soga, 1992); [IndlzZrCld MA0 (Ahlers and Kaminsky, 1988;Chien and He, 1991a- d); chiral metallocenes, e.g., Et[IndIzZrClz (ethylene bis(indeny1)zirconium dichloride), Et[IndIzZrMez, and Et[IndH4]zZrCl2(Ewen, 1984;Kaminsky et al.,1985,1988; Kaminsky, 1986;Ahlersand Kaminsky, 1988;Drogemuller et al., 1988; Ewen et al., 1988;Tait, 1988;Antberg et al., 1990; Chien and He, 1991a; Chien and Sugimoto, 1991; Kioka et al., 1990; 1992; Heiland and Kaminsky, 1992; Soga and Kaminaka, 1992;Uozumi and Soga, 1992;Soga

et al., 1993; Chien and Nozaki, 1993; Chien and Gong, 1993); and other metallocene derivatives (Kaminsky et al., 1986,1991; Ewen et al., 1988;Tait et al., 1992;Chien et al., 1992). Compared with heterogeneous Ziegler-Natta catalysts, metallocene-basedcatalystshave distinct kinetic characteristics for olefin polymerizations, which are sum- marized as follows. Superhigh productivity and good polymerization rate profile. Polymerization rate reaches a maximum within 5 min, decays slightly, and then remains almost

constant. For the Cp2MRz

catalyst family, zirconium catalysts are more active than titanium or hafnium systems, at temperatures over 50 OC. The cocatalyst methylaluminoxane is better suited than ethylaluminoxane or isobutylaluminoxane (Sinn and Ka- minsky, 1980; Kaminsky et al., 1983, 1986; Herwig and Kaminsky, 1983;Kaminsky and Hahnsen, 1987;Mallin et al., 1988;Tsutsui and Kashiwa, 1988;Chien and Razavi, 1988;Chien and Wang, 1989,1990; Chien and Sugimoto, 1991;Tait et al., 1992). For Et[Indl2MC12 (M = Zr,Hf), Hf catalysts showhigher activities (Heiland and Kaminsky,

1992). Productivity of metallocene catalysts can be up to

3 X lo7 g of PE/(g of Zr*h) (Kaminsky et al., 1986). Very high initiation efficiency. Active center con- centration is 75-100% of Zr for ethylene polymerization using zirconocene/MAO, increases monotonically with temperature, and is sensitive to Al/Zr ratio (Chien and Razavi, 1988; Tait, 1988; Chien and Wang, 1989; Chien and He, 1991b; Tait et al., 1992). Narrow molecular weight distribution. In general, polydispersity of PE or its copolymers produced using metallocene catalysts is between 2 and 5; in most cases, polydispersity is about 2 (Tsutsui and Kashiwa, 1988; Chien and Wang, 1990;Chien and He, 1991a;Uozumi and Soga, 1992). Polydispersity can be controlled by mixing metallocene catalysts or by mixing metallocene with heterogeneous Ziegler-Natta catalysts (Ahlers and Ka- minsky, 1988;Kioka et al., 1990;Heiland and Kaminsky, 1992; Fries and Bowles, 1993). Molecular weight sensitivity to hydrogen and temperature. Chain transfer to hydrogen and @-hydride elimination rate constants are 2-3 orders of magnitude greater than the corresponding values found for MgClz

(M= Ti, Zr, Hf; R = C1, CH3)

460 Ind. Eng. Chem. Res., Vol. 33, No. 3, 1994

supported heterogeneous Ziegler-Natta catalysts (Ka- minsky and Luker, 1984; Chien and Wang, 1990). Hy- drogen also reduces polymerization rate (Kaminsky and Luker, 1984; Kioka et al., 1992). Uniform incorporation of comonomers. Composi- tion of copolymer is almost the same as the monomer feed composition, and comonomer is uniformly distributed in copolymer chains regardless of chain length (Chien and He, 1991a,d; Uozumi and Soga, 1992). The effect of comonomer on ethylene polymerization rate depends on polymerization temperature. 1-Hexene or propylene enhances ethylene consumption rate at 30 and 50 "C (Tsutsui and Kashiwa, 1988; Koivumaki and Seppala, 1993). However, 1-hexene reduces ethylene polymeriza- tion rate at 70 and 95 "C (Chien and Nozaki, 1993; Koivumaki and Seppala, 1993). Stereochemicalcontrol in microstructures of olefin polymers. Metallocene catalysts by manipulating the structures of organocompoundscan polymerize olefins with very high stereoregularity to give either isotactic or syndiotatic polymers (Ewen, 1984;Kaminsky et al., 1985, 1986; Kaminsky, 1986; Ewen et al., 1988;Antberg et al., 1990; Chien and Sugimoto, 1991; Antberg et al., 1991; Uozumi and Soga, 1992). Very high cocatalyst ratio. To obtain high polym- erization rate and high active center concentration, the Al/Zr ratio needsto be 1OL1O4orhigher (Chienand Razavi, 1988; Sinn et al., 1988; Chien and Wang, 1989). If the AVZrratio issmall,the polymerizationrate increasesslowly and has an induction period (Chien and Sugimoto, 1991). The reason for high cocatalyst ratio is still unclear. The cost of cocatalyst could be more than 200-300 times the cost of the catalyst in commercial production (SRI, 1993). Hence, high cocatalyst ratio is one of the major barriers to commercialization of metallocene technologies. Metallocene catalysts must be supported on a carrier for use in gasphase processes. Chien and He (1991~)found that Et[Ind12ZrClz supported on Si02 does not change polymerization behavior. Interestingly, a smaller amount of MA0 is required in the case of the supported catalyst than for the homogeneous system to achieve the same catalytic activity. Et[IndHrlzZrClz supported on A1203 or MgC12 can be activated by common trialkylaluminums; however, the catalyst cannot be activated by common alkylaluminums if it is supported on Si02 (Soga and Kaminaka, 1992;Sogaet al., 1993). The MgClz supported metallocene catalysts produced a broader MWD (Soga et al., 1993). CpzZrClz supported on A1203 or MgClz produces atactic polypropylene, but is inactive to propylene using Si02 as a carrier (Kaminaka and Soga, 1992). The effect of a carrier on the performance of metallocene catalysts is unclear at present. For six decades, the polyolefinindustry has experienced several great revolutions. Despite the fact that there are some technological difficulties, commercialization of met- allocene-based catalyst technologies may be the greatest revolution of all. The impact of metallocene-based catalysts on polyolefin industry is perhaps beyond the current imagination. 3.2. Polymer Particle Morphology Developments. One of the most important features of ethylene polym- erization using heterogeneous catalysts is replication of polymer particles (Berger and Grieveson, 1965;Mackie et al., 1967). This suggeststhat the morphologydevelopment of polymer particles depends on the original morphology of the catalysts. Polymer growth on titanium crystal surfaces has been carefully examined using electron microscopy(Carradine and Rase, 1971;Baker et al., 1973),

and it was found that the polymer first accumulates on boundaries of catalyst fragments and at existing cracks. Carradine and Rase (1971) observed that the polymer appearing on lateral surfaces is nodular in character while that growing on basal surfaces is fibrous. Baker et al. (1973) observed nodular polymer growth around small catalyst particles (-0.1 pm in diameter) during gas phase polymerization. The shape of the polymer growth was usually governed by that of the active catalyst particle. Wristers (1973a,b) examined the morphology of classical Ziegler-Natta catalyst usingelectron scanningmicroscopy,

and found that the gross catalyst particles are composed of primary particles whose size and shape depend on the preparation conditions of the catalyst. According to Hock (1966)and Graff et al. (1970),the size of a primary particle is in the range 10-1000 A in diameter. The shape and reactivity of primary catalyst crystals are responsible for the fiberlike structures that make up the final polymer particles (Wristers, 1973a,b). On the basis of electron microscopy and adsorption data, Bukatov et al. (1982) suggested that the 10-30-pm catalyst granules consist of three levels of particles: 200-1000-A, 1000-5000-A, and <5-pm particles. The size of catalyst particle in the final

polymer matrix is in the range of

et al. (1986)used microspherical mesoporous copolymers of styrene and divinylbenzene as a catalyst support, and found that the microspheres of support polymer also dispersed in the polyethylene matrix. The morphology replication nature of ethylene polym- erization provides a guideline for catalyst development (Galli et al., 1981, 1984). Polymer morphology control can be considered in the catalyst preparation stage. Although the reactivity and chemicalcompositionof MgClz or silica supported catalyst have been changed compared with those of a conventional Ziegler-Natta catalyst, the morphological replication nature of ethylene polymeri- zation using the supported catalyst is the same. The morphology of a final polymer particle is governed by the morphology of the supported catalyst. However, for supported catalysts, the primary particle is the support material, e.g., MgC12 crystal or silica. It has been exper- imentally observed that polymer grows on the edges of MgC12 crystals treated with Tic4 and activated by AlEt3 (Barbeet al., 1983;Galliet al., 1984). Polymer morphology development using supported catalystshas been examined in many laboratories. All the evidence shows that the polymerparticle sizeand shape replicate itsparent catalyst size and shape (Galli et al., 1981, 1984;Hosemann et al., 1982;Munoz-Escalona et al., 1987,1988;Karol et al., 1987; Kakugo et al., 1988;Dall'Occo et al., 1988). This suggests that the MgClz particle and/or silica surface favors the most isotropic distribution of Tic13 active centers. Figure 12 shows the replication relationship between silica supported catalyst and polymer particles made with the catalyst (Munoz-Escalona et al., 1988). The morphology of a supported catalyst depends on the preparation conditionsand variousmodifiers. No detailed morphology study on a supported catalyst has yet been reported to date. The particle size shown in Table 8, estimated by Chien (1987), is very small. The results in Table 8 suggest that MgC12 supported catalysts are composed of minute crystals with large surface area and pore volume. As discussed in the previous section, the catalyst particle size depends on preparation conditions. Commercial catalysts have a particle size distribution in the range of 10-30 pm (Galli et al., 1981, 1984; Bohm et al., 1988;Kakugo et al., 1988;Kim and Woo, 1989). Kim and Woo (1989)examinedTiCUMgCldSiOZcatalyst using

200 to 500 A. Mkrtchyan

100 1























Figure 12. Particle size replication of ethylene polymerizationwith


scanning electron microscopy and found that the catalyst consists of small particles with a rough surface. However, the internal structure of the catalyst was not shown. If the catalyst is impregnated on silica support, the particle size is in the range of 30-250 pm (Bailly and Speakman, 1986; Munoz-Escalona, 1988). The average polymer particle size is about 20 times larger in diameter than the catalyst particle size (Galli et al., 1981, 1984;Karol et al.,


Kakugo et al. (l988,1989a,b) recently studied internal morphology development of polypropylene particles using transmission electron microscopy. Their results indicate that each polymer primary particle has a nucleus 50-170 Ain diameter, which is believed to be the catalyst primary particle. It is interestingto noticethat the catalyst primary

particle sizeidentified in the polymerparticle is in excellent agreement with catalyst crystal particle size estimated by Chien (19871, as shown in Table 8. This suggests that the minicrystal size of a catalyst achieved by ball-milling cannot be reduced further during polymerization, or that the Ti catalyst active center is fixed on the surface of this mini MgClz crystal and not inside the minicrystal. The diameter of a polymer primary particle is about 0.2 pm. The fine polymer globules, about 1 pm in diameter, are formed by fusion of several polymer primary particles. The gross polymer particles are composed of these primary particle aggregates. Weist et al. (1989) found that the pore distribution of silica supported catalyst does not change at the initial stage of polymerization. However, the pore size of a catalyst increases with an increase in polymer yield, indicating fragmentation during polym- erization. The SEM photographs of polymer particles show that the catalyst particles are completely encapsu- lated by polymer in the early stage (<5 min) of polym-


occurs as early as 5 min after the initiation of polymer- ization. Both spherical and fiberlike particles are formed during ethylene polymerization (Kim and Woo, 1989).

On the basis of the available evidence in the literature, polymer morphological development during polymeriza- tion with supported heterogeneous catalysts can be summarized as follows:

1. An activated porous catalyst particle is composed of primary particles of size in the range of 50-200 A. These catalyst primary particles are held together by van der

Rupture of the outer layers of polymers particles

Ind. Eng. Chem. Res., Vol. 33, No. 3, 1994 461

Waals force. There are interstices or pores between the primary particles, providing a large surface area.

2. At the beginningof polymerization,monomer diffuses

through the interstices and adsorbs onto the surface of catalyst primary particles, where the active centers are located (Bohm, 1978a).

3. The propagation reaction occurs on the surface of

the catalyst primary particles. When monomer is con- verted to polymer, polymer precipitates and accumulates around the catalyst primary particles. As polymerization proceeds,the interstices are gradually filledwith polymers. Hydraulic pressure or stress between catalyst primary particles increases due to accumulation of polymer chains in the pores leading to separation of catalyst primary particles. Polymer around the catalyst primary particles cements these primary particles. The distance between

the catalyst primary particles increases with an increase

in polymer yield. Thus polymer particles grow isotropi- cally during polymerization. When the polymer particles with high polymer yield are examined using electron microscopy, catalyst primary particles are found to be uniformly dispersed in the polymer matrix as though the catalyst were broken up during polymerization. In fact, this should be regarded as a separation process. Catalyst primary particles should be broken in the preparation stage. Interstices or pores must exist between catalyst primary particles, where active centers are located, so that polymerization occurs on the surfaces of the interstices, leading to separation of catalyst primary particles.

4. The polymer around the catalyst primary particles is swollen with monomers (ethylene and a-olefin). Since olefin polymerization is a highly exothermic process, the temperature within the polymer primary particle could increase rapidly, and may exceed the softening or melting point of the resin (Karol et al., 1987). Due to the rapid growth of primary particles, the polymer chains between primary particles fuse together to form polymer primary particle aggregates of diameters in the range of 0.5-1.0 pm.

5. The final polymer particles are composed of aggre-

gates of primary particles with average diameters in the range of 200-500 pm (Galli et al., 1981, 1984; Kakugo, 1988; Bohm et al., 1988). The morphology of the final particles depends on polymer yield and the balance between annealing and distorting forces during particle growth. The primary particles grow isotropically(Kakugo, 1988);therefore, in general, the polymer particles replicate their parent catalyst particles in shape and size.

The mechanism described above can be expressed

schematicallyas in Figure 13. To avoid confusionbetween catalyst primary particle and polymer primary particle, the catalyst primary particle is called a domain in Figure

13. Polymer primary particles are formed with a catalyst

domain as a core, which is the smallest resulting catalyst particle during polymerization. 3.3. Chemical Reactions. From the preceding ob- servations, the chemical reactions of ethylene polymeri- zation can be envisioned as occurring at the interface between the solid catalyst and the polymer matrix, where the active centers are located. From gas-state monomer to solid-state polymer, ethylene experiences a dramatic physicochemical transition within a very shorttime. The polymerizationenvironment changeswith the composition of catalyst, polymerization process, reactant composition, reactor operating conditions, and extent of polymerization. Although intensive research activity has been focused on Ziegler-Natta catalyst systems since their discovery in the early 1950s, no definitive, unequivocal chemical

462 Ind. Eng. Chem. Res., Vol. 33, No. 3, 1994







10-30 micron in diameter (30-100 micron for silica supported catalyst) Composed of minute crystal domains 20-200 A"in diameter

Formation of polymer primary particles, -0 2 micron in diameter Separation of catalyst domain

Growth of primary particles Formetionof primary particle aggregates, -1 micron in diameter

Figure 13. Schemeof polyethylenemorphology development during gas phase polymerization.

reaction mechanismshave been developedto fullydescribe the kinetic behavior of ethylene homo/copolymerization, due to the complexity of the systems employed. Never- theless, the key elementary reactions have been estab- lished, which include formation of active centers, insertion of monomer into the growing polymer chains, chain- transfer reactions, and catalyst deactivation. Most of the proposed mechanisms are based on information about polymerization rate, molecular weight and its distribution, polymer chain microstructure, and active center concen- trations. Detailed mechanisms have been discussed in monographs (Boor, 1979; Kissin, 1985) and in recent reviews(Taitand Watkins, 1989;Dusseault and Hsu, 1993). Among the proposed propagation mechanisms, the one proposed by Cossee (Cossee, 1964; Arlman and Cossee, 1964)together with its subsequent modifications has been widelyadopted. With Cossee's mechanism in mind, Bohm (1978a) proposed perhaps the most comprehensive ele- mentary reactions involved in the ethylene polymerization process. Since commercial production of LLDPE and HDPE consists of a copolymerization process, obviously, copo- lymerization mechanisms are required to understand kinetic behavior and polymer properties. Therefore, Bohm's reaction model has been modified and extended to the ethylene copolymerization processes in recent modeling studies (Villermaux et al., 1989;de Carvalho et al., 1989,1990;McAuleyet al., 1990;Lorenzini et al., 1991; Hutchinson et al., 1992). Furthermore, to illustrate multimodal distribution of polymer chain composition and broad moleclar weight distribution, the multiple active center concept has been adopted in recent modeling studies. Therefore, a comprehensive chemical reaction mechanism should include not only reactions proposed by Bohm (1978) but also reactions involving multicom- ponents at different active centers. Whether each type of active center has the same or different reaction mecha- nisms is still uncertain (multiple center formation itself has not been proved fundamentally). If one assumes that all of the active centers perform the same reaction mechanisms, but with different reaction rates for each elementary reaction, then elementary reactions which are commonlyadopted in modelingstudies can be summarized as in Table 9, where C, is the catalyst potential active

Table 9.

and a-Olefin Copolymerization

Summary of Elementary Reactions for Ethylene




spontaneous activation

activation by aluminum alkyl (A) activation by electron donor (E) activation by hydrogen


activation by monomer

1 (Mi)

activation by monomer

2 (Mz)



[MI3 - Pm+l,n,I K~LZ


+ [M21

- p*m,nt1,2


Pm,n,z + [MI3 -*



Pm,n,z + [Mz]- Pm,n+l,z

Initiation initiation of M1 by normal active center initiation of Mz by normal active center initiation of MI by active center with H initiation of Mz by active center with H initiation of MI by active center with A initiation of Mz by active center with A initiation of M1 by active center with E initiation of Mz by active center with E Propagation propagation of chain type 1with M1 propagation of chain type 1with M2 propagation of chain type 2 with MI propagation of chain type 2 with Mz

Chain Transfer spontaneous chain transfer or &elimination chain transfer to hydrogen



- PO+ 4'm.n



Pm n L + [HJ - PH.O+ Q'm.n



Pm,n,L + [AI - P*A,O + Q'm,n


Pm,n,i + [El -+ P*E,o+ P'm,n

Pm,ni + [MI]


- P*I,o,I+ Q'm,n


Pm,n,i + [Mz] - *o,I,,

+ q'm.n

chaintransfer to aluminum alkyl (A)

chain transfer to electron

donor (E) chain transfer to M1

chain transfer to Mz










' qm,n

Deactivation spontaneous deactivation


+ lZ1 -*




- d'












- d'

+ qm,n

+ qm,n

' 4m.n

+ qm,n

+ 4m,n

deactivation by

poison (Z) deactivation by aluminum alkyl (A) deactivation by electron donor (E)

deactivation by

deactivation by monomers

impurities or

hydrogen (H2)

Other Possible Reactions

PI,OJ + q'ra


- pr+1+,1


pO,l,z + q'ra

- Pr++1.2


P*m,n,l + qfra -+


pm,n,2 + Q'ra


- Pm+r,nta,z

formation of short-chain branches formation of short-chain

branches formation of long-chain branches (rare) formation of long-chain branches (rare)

center; P*o is the active center without polymer chain; P*m,n,iis the active center with m units of monomer 1 and n units of monomer 2, with the third subscript idenoting the chain terminal monomer type bonded to the active

center; q’m,nis a dead polymer chain with a terminal double

is a dead polymer chain without terminal

double bond. For simplification of notation, no subscript is shown corresponding to the type of active center. The reactions shown in Table 9 should be considered to occur at each type of active center. The terminal model is assumed to be valid for binary copolymerizations. How- ever, if the penultimate model is valid for some systems, the mechanisms in Table 9 can be modified accordingly. Hutchinson et al. (1992) considered transformation re- actions,whichare assumed to form different active centers. If an active center formed by transformation is indepen- dent of the original polymer chain, then the effect of a transformation reaction is on the initiation reactions which have been shown in Table 9. If an active center formed by transformation depends on the original polymer chain produced by this center, then the formation of each active center must be treated individually. The detailed mech- anisms of transformation reactions have not been given (Hutchinson, 1992). Formation of short-chain branches was proposed by Bohm (1978). It was assumed that a polymerchain with a terminal double bond will participate in propagation at an active center with one monomer unit. Formation of long-chain branching is unlikely for Ziegler- Natta polymerization systems. However, for Phillips’

bond; and q,,,,

catalyst systems,polymerchainswith long-chainbranching are formed, although branching frequency is low (McDaniel, 1985). For ethylene polymerization with a Ziegler-Natta catalyst at high temperature and high pressure, polymer chains dissolve in the monomer phase. Hence, polymer chains with a terminal double bond are able to diffuse to active centers to participate in propa- gation and to form long-chain branching. However, detailed information regarding this reaction has not been

published to date. The mechanisms in Table 9 are summarized based on Ti-based catalysts. Reaction mech- anisms for chromium oxide based catalysts are a subset of those in Table 9. It should be pointed out that second- order deactivation of active centers is not included in Table


4. Reactor Modeling

Reactor modeling is the determination of a quantitative relationship between reactor performance and reactor operating conditions. It requires comprehensive under- standing of polymerization processes, physical phenomena, and chemical reaction mechanisms. The importance and benefits of reactor modeling have been widely recognized

The ap-

plication capability of a model depends on scope of the modeling effort. Ray (1986, 1991) defined a modeling hierarchy as microscale, mesoscale, and macroscale, ac- cording to the characteristics of the polymerization reactor systems. The relationships between modeling scale and polymerization systems are outlined in Figure 14 after Ray (1986,1991). The emphasis of each modeling level, in particular for ethylene polymerization systems, can be summarized as follows:

by both industrial and academic researchers.

1. Microscale modeling: polymerization rate devel-

opment, or polymer yield; reactant species distribution; molecular weight development and ita distribution; chem- ical composition of polymer chains and its distribution; microstructures of polymer chains, including chain branch- ing, unsaturated groups, and sequence distribution.

Ind. Eng. Chem. Res., Vol. 33, No. 3,1994










Figure 14. Levels of polymerization reactor modeling.

2. Mesoscale modeling: interphase heat and mass

transfer; intraphase heat and mass transfer; fluid me-

chanics and micromixing; polymer particle morphology development; polymer particle size and distribution.

3. Macroscale modeling: macromixing and residence

time distribution; overall material and energy balance; heat and mass transfer from the reactor; reactor dynamics and control; polymer grade transition and control.

There are no definitive boundaries between these modeling levels. In fact, they often overlap during modeling studies. For instance, knowledge of some microscale modeling is required for mesoscale modeling studies, and macroscale modeling depends upon under- standing of microscale and mesoscale phenomena. A

complete model for ethylene polymerization, including all three levels, has not been developed in the literature. However, significant modeling efforts with emphasis on specific levels have been published. Microscale and mesoscale modeling studies are normally reported as kinetic modeling in the literature. Macroscale modeling is often referred to as dynamic modeling. Major contri- butions to ethylene polymerization reactor modeling are now discussed.

Kinetic modeling has focused

mainly on polymerization rate and polymer molecular weight development, which can be expressed as follows:

4.1. Kinetic Modeling.



i=l j-1



where binary copolymerizationis assumed. With the long- chain approximation, polymerization rate is a function of the propagation rate constants, monomer concentrations, and active center concentration. Complexity can arise through the expression for active center concentration. Molecular weight development is a function of all the components which affect polymer chain length, as shown in Table 9. One of the important features of ethylene polymerization is the broad molecular weight distribution of the polymer. Two theories have been developed to explain broad MWD of polyethylene, namely, diffusion theory and multiple active center theory. The former emphasizes the effect of monomer concentration on molecular weight development. The latter is more con- cerned with the influence of active centers and associated kinetic parameters. Some of the experimental results in support of these theories are as follows.

Diffusional limitations:

1. Polymerization rate depends on stirring speed when

agitation speed is below a critical level (Berger and Grieveson, 1965; Bohm, 1978b).

464 Ind. Eng. Chem. Res., Vol. 33, No. 3, 1994

2. Molecular weight of PE formed in the initial period

of polymerization is much higher than that of polymer

formed in later stages (Crabtree et al., 1973).

3. The polymerization rate of ethylene increases sig-

nificantly in the presence of small amounts of comonomers, such as propylene and l-hexene (Finogenova et al., 1980; Soga et al., 1989). Soga et al. (1989) found that the crystallinity of the copolymer decreases with an increase in comonomer concentration. The reactivity ratio of propylene to ethylene is lessthan 1. Therefore, an increase in reaction rate in the presence of a small amount of propylene is due to a decrease in the monomer diffusion resistance. This suggests that diffusion resistance is very significant in highly crystalline polymers, such as HDPE (Soga et al., 1989). The results of Finogenova et al. (1980) and Soga et al. (1989) are very convincing. In fact, the effect of crystal- linity on monomer concentration in the polymer matrix has not been comprehensivelystudied. However,the effect of a-olefins on the ethylene polymerization rate depends on the a-olefin types. It is not certain whether physical or chemical effects dominate. Crabtree’s result is in contradiction with data reported by Bohm (1978~)and others (Kissin, 1985). Bohm (1978~)showedthat number- average molecular weight increases significantly with reaction time during the early period of polymerization and then levels off with an increase in reaction time.

Multiple active centers: The following phenomena

cannot be explained by diffusion limitation theory (Zuc- chini and Cecchin, 1983; Karol, 1984).

1. Different transition metal catalysts can provide large

changes in polymer MWD.

of catalyst changes the

breadth of the MWD.


Chemical modification

3. Heterogeneous catalyst can produce broad MWD,

even when the polymer is in solution.

Homogeneous soluble catalysts provide narrow

MWD, even when the polymer is insoluble in the reaction medium.

5. High-activity catalyst does not necessarily provide

broad MWD.

6. Ti-based catalyst can provide polymer with narrow

MWD, and vanadium-based catalyst produces resins of broad MWD. A combination of these two catalyst systems, however, gives MWD of intermediate breadth with the same reactor operating conditions (Nicoletti et al., 1988).


7. Compositionof LLDPE separated using temperature-

rising elution fractionation had two or three distinct peaks

(Wildet al., 1982a;Usami et al., 1986;Kelusky et al., 1987; Mirabella et al., 1987;Hosoda, 1988). The reactivityratios, as measured by 13C NMR, were different for each peak (Usami et al., 1986). Although the above phenomena cannot provide con- clusive proof of multiple active center formation mech- anisms,they dosuggestthat molecular weightdevelopment of PE depends on the nature of the catalyst and on the polymerization mechanisms. Many mathematical models have been developed to quantitatively describe the kinetic phenomena of ethylene polymerization. Most of the early modelsfocusedon polymerization rate development (Keii, 1972). However,the emphasis has been placed on polymer quality development in more recent modelingstudies. The progress of kinetic model development is now discussed. Bohm (1978a) developed a model based on a compre- hensive chemical reaction scheme, which is applicable to both homogeneous and heterogeneous catalyst systems. The model development is based on assumptions of steady state for all active species, kinetic rate constants inde-

pendent of chain length, and slow deactivation of the catalyst. However,the model is limited to polymerization rate and number-average molecular weight (M,)of ho- mopolymerization. Higher order molecular weight and MWD were not considered. Dussealt (1991)and Dusseault and Hsu (1993)developed areaction rate modelfor gasphase ethylene polymerization based on the multiple active sites hypothesis. This model is able to incorporate different distribution functions of active centers to simulate polymerization rate behavior. Both first-order and second-order deactivation mecha- nisms are considered in the model calculations (Dussealt, 1991). However, the model treatment is limited to polymerization rate. The effect of multiple active site distribution on polymer molecular weight development was not considered. de Carvalho et al. (1989) developed a comprehensive ethylene and a-olefin copolymerization model accounting €or the formation, initiation, and deactivation of active centers, and for the spontaneous and chain transfer to hydrogen, monomer, and organometallics. The state of the art to model copolymerization on multiple active sites was described in detail. Thisgeneral model wassimplified by assuming a stationary state of growing chains and neglecting deactivation of active sites. The molecular weight distribution was developed on the basis of these assumptions. This model shows that the instantaneous polydispersity for each active site is equal to 2, and the instantaneous polydispersity for the entire amount of polymer produced would deviate from 2 by the ratio of the variance of the propagation rate constants to the mean propagation rate constant. The effect of the propagation rate constant distribution on molecular weight develop- ment was demonstrated by assuming unimodal and skewed bimodal distributions of propagation rate constants. The pseudo-kineticrate constant method was used to evaluate the kinetic rate constants for each site. However, chain- transfer reactions were not considered in the active center fraction calculations. Effects of chain-transfer reactions on active center fraction calculations can be very important if chain-transfer reactions are significant (Xie and Hamielec, 1993a-c). de Carvalho et al. (1990) recently extended their model to calculate both the stereo and chemical sequence distributions of copolymer. The models developed by de Carvalho et al. (1989,1990)have not been evaluated using experimental data. Villermauax et al. (1989) and Lorenzini et al. (1991) developed a model for ethylene and a-olefin copolymer- ization with Ziegler-Natta catalyst at high temperature and high pressure. The model development is based on a single active site, which is then extended to two or three active sites using a mixing law. This model can fit molecular weight averages and distribution data over the range of the experimental conditions. The kinetic pa- rameters were estimated by fitting the model to experi- mental data (Lorenzini et al., 1991). However,the effects of monomer type on the reactivity of active sites are neglected in the parameter estimation. In other words, ethylene and butene-1 copolymerization were treated as r1 = r2 = 1. The polydispersity of polymer obtained for their system at higher temperature and high pressure is between 3 and 5. It is worth noting that the polymerization occurs in a supercritical phase, the polymer being soluble in the monomer mixture. Therefore, there is no monomer diffusion limitation in this system. However, the poly- dispersity of polymer is similar to that obtained in a low- temperature system. This suggests that multiple active sites play a central role in determining molecular weight



0 7ooOo




+ +

Laboratory Data



,%J 4oooo-

' 3oooo b





io 4b









Time in Hours Figure 15. Comparison of model predictions and industrial labo- ratory data for gas phase ethylene copolymerization in a fluidized bed reactor. Reprinted with permission from McAuley et al. (1990). Copyright 1990 AIChE.

development. On the basis of this model analysis, the realkylation, spontaneous transfer and deactivation re- actions play only minor roles during high-temperature and high-pressure polymerization. McAuleyet al. (1990)developed a comprehensive model for gas phase ethylene copolymerization in a fluidized bed reactor, such as the UNIPOL system shown in Figure 3. Besides elementary reactions described by de Carvalho et al. (19891, the model also accounts for the effect of impurities on catalyst deactivation. This model can be used to predict average chain lengths for multiple copo- lymerization. The kinetic rate constants in each active site are expressed as pseudo-kinetic rate constants. Ap- plying Stockmayer'sbivariate distribution function (Stock- mayer, 1945)to each active site, McAuleyet al. (1990),for the first time, were able to calculate molecular weight and composition distributions for ethylene copolymers. The model was evaluated assuming two active sites of catalyst. The model predictions are in excellent agreement with experimental data as shown in Figure 15. The models discussed above focused on the effects of chemical reactions and catalyst on polymer property development using the hypothesis of multiple active center distribution. Physical phenomena were simplifiedin these model developments. Assuming the polymer particles to be isothermal and spherical, Galvan and Tirrell (1986) developed a model for molecular weight calculation of olefinhomopolymerization which accounted for two active sites and the limitation of monomer diffusion. The simulation results showed that the presence of different types of active sites is responsible for the broad MWD of olefin polymerization. Even in the case in which there are apparent monomer diffusion limitations, multiple active sites still play an important role in determining molecular weight development. Significant contributions have been made by Ray and his co-workers over the past decade toward a better understanding of olefin polymerizations using Ziegler- Natta catalysts (Ray,19881,includingexperimental studies of olefin polymerization behavior (Taylor et al., 1981;Yuan et al., 1982; Choi et al., 1983;Choi and Ray, 1985b;Yoon and Ray, 1987);polymer molecular weights and polydis- persity modeling (Nagel et al., 1980; Floyd et al., 1987; 1988); microparticle, macroparticle, and external film mass- and heat-transfer phenomena (Floyd et al., 1986a- c); particle ignition, extinction, monomer sorption phe- nomena, and particle morphology development (Hutch- inson and Ray, 1987, 1990, 1991a,b; Hutchinson et al., 1992); and reactor dynamics and control studies (Choi and Ray, 1985c, 1988). On the basis of the polymer morphology development






0 )


Ind. Eng. Chem. Res., Vol. 33, No. 3, 1994



deactivating cataIyst(I

Actlvlty :-5500








10-~ cdis


Mb - 4




Mn2 - 30,000





WElaHT FRACTION Figure 17. Polydispersityresulting from combinationof two polymer

fractions, each with polydispersity of

from Floyd et al. (1987). Copyright 1987 John Wiley & Sons.

discussed above,as shownin Figure 13,a multigrain model

of particle growth was used to study heat and mass transport phenomena. According to Floyd et al. (1986a), intraparticle temperature gradients arenegligiblefor both the microparticles and the macroparticles in gas or slurry polymerization reactors. Exceptions would be for large highly active catalyst particles early in the lifetime of the polymer particles in gas phase reactors. This conclusion is in agreement with commercial observations (Bailly and Speakman, 1986). Intraparticle monomer concentration gradients in the microparticles can be significant for high- activity-catalyst systems having large primary crystals of catalyst in a gas phase polymerization system. However, concentration gradientsin the macroparticles will normally be negligible for gas phase polymerization.

2. Reprinted with permission

the effects of

intraparticle and external boundary layer transport re- sistance on kinetic behavior and polymer property de- velopment based on the multigrain model. It was found that diffusion resistance affects the polymerization rate more strongly than the molecular weights. Their simu- lation results show that even in a highly diffusion-limited situation, the polydispersity drops below 4 within 30 min after polymerization begins. A typical simulation result is shown in Figure 16 (Floyd et al., 1987). Therefore, monomer concentration gradients in the macroparticle

Floyd et al. (1987, 1988) further studied

466 Ind. Eng. Chem. Res., Vol. 33, No. 3, 1994





u * 2 cm/s












hslonloneous brrohuv









2 5

TIME (sec)




Figure 18. External film heat- and mass-transfer resistances in gas phase polymerization of polyethylene under stirred bed conditions with high activity catalyst. Reprinted with permission from Floyd et al. (1986~).Copyright 1986 John Wiley & Sons.

alone cannot explain the breadth of the MWD under


of active sites affect polydispersity significantly. Figure 17 shows the effects of molecular weight and weight fraction

normalreactor operating conditions.

produced in each site on polydispersity assumingtwo active sites (Floyd et al., 1987). This figure shows that the molecular weights produced by the two sites must differ by at least an order of magnitude to achieve broad polydispersities (PD > 5). If polydispersity is the same for each site, the overall polydispersity is always at a maximum when each site contributes 50% of the total weight of polymer. A bimodal MWD will result if the polymer with polydispersity greater than 6.0 is produced from two types of active site. To produce unimodal MWD with high polydispersity, a large number of different types of active site (>4) is required. However, an increase in the number of active site types may not necessarilyincrease the polydispersity. Polydispersity and MWD are not sensitive to catalyst particle size distribution. Floyd et al. (1986b,c) calculated external film heat- transfer resistances in gas phase ethylene polymerization in a stirred bed reactor using the Ranz-Marshall corre- lation. They found that operating a gas phase reactor with either a low- or high-activity catalyst would lead to a significant particle temperature rise that wouldinfluence observed kinetics early in the polymerization, and would lead to agglomeration and sticking of polymer particles for high-activity catalyst above 30-60 pm in diameter. Figure 18 shows simulations of external film temperatures and monomer concentration gradients of polymer particles in an early stage of polymerization (Floyd et al., 1986~). It can be seen that the concentration gradient is less than 1%; however, temperature rise on the particle surface is very significant for large catalyst particles. It should be pointed out that this simulation was based on a stirred bed reactor with a very slowgas flowvelocity (2 cm/s). For fluidizedbed reactors, thetemperature riseshould be much less than that shown in Figure 18. The main results of heat and mass transport for gas phase polymerization are summarized in Table 10(Floydet al., 1986a,b,1987,1988). Hutchinson and Ray (1991a)recently studied the effect of prepolymerization on polymer property development and found that prepolymerizationdoesnot affectmolecular weight development, but increases polymer bulk density. Bulk density increaseswith an increasein reactor operating temperature. If prepolymerization is carried out at high

Table 10.

Phase Polymerization (Flosed et al., 1986a.b. 1987. 1988)

Heat- and Mass-Transfer Resistance in Gas





negligible,<1 OC



can be significant for high-activity catalyst and for those yieldinglarge primary crystallites after breakup



negligible except for large particles


of high-activity catalyst early



in particle lifetime negligible except for large particles of high-cavity catalyst early in particle lifetime



often very significant during


initial stage of particle growth




temperature and at low pressure, reactor fouling is more serious. An attempt to model morphology development of polymer particles has also been made in a recent publication (Hutchinson et al., 1992). Significant modeling studies of heat and mass transfer during olefin polymerization from the gas phase have also been carried out by Laurence and Chiovetta (1983). Their model development was based on the following assumptions: microparticle and macroparticle are spher- ical; diffusivity is independent of temperature; and mi- croparticle sizesare equal. Their simulationsshowedthat there is no significant monomer diffusion effect during the progress of the reaction. The monomer concentration profile becomes almost uniform inside the microparticles within 3min. The temperature gradient is less than 1"C. However, the monomer gradient in the initial stage is significant. The resultsof Floyd et al. (1986a,b,1987,1988) discussed above are in general agreement with model predictions by Laurence and Chiovetta (1983). Webb et al. (1989a) determined sorption of ethylene onto the catalyst and silicasupport usingsolid-gas chromatography and gravimetry and found that ethylene sorbs on both the silica support and the catalyst. However, sorption on polymerization catalyst is about 20% higher than on its support. The sorption of ethylene in the high yield polymer particles is 50 times less than the sorption of ethylene by the bare catalyst. Webb et al. (1989b)measured internal porosity and intraparticle diffusivity using solid-gas chromatography. It was observed that internal void space and transport rate decrease significantly at a yield of 0.1 g of polymer/g of catalyst. Polymerization to such low yields fills the macropores of the catalyst nearest the externalparticle surfacewith polymer. However,the rate of adsorption of monomer is very fast and is not rate limiting. Webb et al. (1991) further discussed morpho- logical influences in the initial stage of gas phase polym- erization and proposed a kinetic model accounting for nonideal morphological influences. On the basis of the above studies, it is safe to state that multiple active sites play an important role in determining PE molecular weight development, whereas the effect of diffusion limitations on molecular weight development plays only a minor role. However, diffusion limitation may be significant in controlling the polymerization rate for a system with a high-activity catalyst and highly crystalline polymermatrix. Even in a diffusionallylimited situation, MWD is still controlled by the nature of the catalyst system and the ethylene polymerization mech- anisms. Therefore, in the future, polymer property modeling should focus on multiple active site effects and exploit the mechanisms of formation of different active centers. The current state of the modeling of molecular weight development published in the literature is sum- marized in Table 11. Accounting for multiple active sites, detailed elementary chemical reactions, and mole fraction

Table 11.

Approaches to Polymer Property Modeling

model structure

basic assumption

P,, ,,P polydispersity

R,, P,, ,,P chain length

distribution, polymer compositiondistribution, stereo and chemicalsequence

length distribution

R,,M,,a,, Ma,MWD, short chain branches, double bonds

R,, P,, p,, bivariate molecular weight and composition distribution of copolymer

M,,a,, polydispersity

isothermal, spherical particles, two active sites, effect of diffusion control


multiple active sites, concentration of reactant is independent of active sites, deactivation depends on site type, steady state for growing chains two or three active sites, steady state for living chains, first-order deactivation multiple active sites, monomer concentration in the active sites is proportional to gas phase pressure and independent of sites, steady state for growing chains, uniform gas compositionand temperature multiple active sites, consider effect of diffusion limitation, steady state for growing chains, multiple grain particles

of each active site, Xie et al. (1993) recently developed a comprehensive model for molecular weight calculations of ethylene and a-olefin copolymers. The model can predict molecular weightwith broad polydispersityranging from 2 to 40 or higher depending on polymerization conditions. The model can be used for copolymer mo- lecular weight calculations of not only heterogeneous Ziegler-Natta catalyst but alsometallocene-based catalyst systems.

4.2. Dynamic Process Modeling and Reactor Con-

trol. Dynamic process modeling considers the behavior of the entire reactor system, as shown in Figure 14. Therefore, dynamic modeling should incorporate kinetics of polymerization, heat and mass transfer, and reactor characteristics in arealistic manner. Choiand Ray (1985~

1988) were the first to model the dynamic behavior of gas phase polymerization in a fluidized bed reactor and in a stirred bed reactor. Their model development for a fluidizedbed reactor isbased on the followingassumptions:

1. The emulsion phase is perfectly backmixed and is at incipient fluidizing conditions with constant voidage.

Ind. Eng. Chem. Res., Vol. 33, No. 3, 1994


reactor and process

batch, slurry system, homopolymerization

batch or CSTR, slurry system, copolymerization


Galvan and Tirrell, (1986)

de Carvalho et al. (1989,1990)

CSTR, high-temperature

Villermauxet al. (1989);

and high-pressure Lorenzini et al. (1991)


fluidized bed reactor or CSTR, multicomponent gas phase polymerization

McAuley et al.(1990)

batch, slurry or gas phase homo- or copolymerization

Nagel et al. (1980);

Floyd et al.(1987); Hutchinson et al. (1992)

stabilized, provided that the recycle heat exchanger has sufficient cooling capacity. However, even under closed- loop control, temperature runaway can occur in some situations when the rate of heat removal by recirculating gas becomes lower than the rate of heat generation in the reactor. Hence, additional coolant is required to maintain stable reactor operation conditions for high reaction rate systems. Assuming that the stirred bed reactor is adiabatic and heat removal depends on evaporation of liquid monomer, Choi and Ray (1988) found that a stirred bed reactor was open-loop-unstable for most operating conditions with or without a bed level controller. The reactor instability was due to the limited heat removal capability of the system. Although the reactor is open-loop-unstable,the dynamics are so sluggish that application of feedback controllers on bed level, temperature, and pressure can easily stabilize the system. Under the assumptions listed above, Choi and Ray (1985~)did not consider the existence of a maximum

bubble size, nor the effects

of a varying bed height or inert


The flow of gas in excess of minimum fluidization

gas on the fluid dynamics of the reactor. Talbot (1990)

passes through the bed in the form of bubbles. The bubbles are spherical, uniform, and in plug flow. The bubble phase is always at a quasi-steady state.

and McAuley et al. (1993)relaxed those assumptions and considered the bubbles to grow only to their maximum stable size under certain reactor operating conditions,and

3. Polymerization occurs only in the emulsion phase.

assumed further than the bubbles are of constant size at

4. Mass and heat transfer occur at uniform rates over

all levels within the bed and travel through the bed in

the bed height.

plug flow. The simulation results show that bubble size


There is negligible mass- and heat-transfer resistance

critically influences the rate of heat and mass transfer

between the solid polymer particles and the emulsion phase.

within the bed, and when the bubbles are as small as those predicted by the maximum stable size correlations, there


No elutriation of solids occurs.

is little or no resistance to the transfer of monomer and


There is no catalyst deactivation.

heat between the phases. McAuley et al. (1993) further


The fluidization bed height (reaction zone height) is

simplified the two-phase model to the well-mixed model

constant. The model simulations show both steady-state multi- plicity and bifurcation phenomena in a fluidized bed reactor. Temperature runaway in the reactor can occur in open-loop operation if parameters are set beyond the Holf bifurcation point or if disturbances carry the reactor beyond the domain of attraction of the lower steady state. Under feedback temperature control the reactor can be

by assuming unrestricted mass and energy transfer be- tween the bubble and emulsion phases. The simulations show that only minor deviations exist between the predictions of the two models for superficial velocities in the range of 3-6 times the minimum fluidization velocity and for the small bubble size predicted by the maximum stable size correlations. Therefore, the well-mixed gas assumption isvalid for modelinggas phase polymerization.


30u '

Ind. Eng. Chem. Res., Vol. 33, No. 3, 1994






Partlcle Slze (cm)



Figure 19. Particle size distribution of catalyst (Talbot, 1990).

.- in


b ,,,,,,,,, ,,,,,,,,,

o. aa

a. as

a. io

0. IS

a. 20

Partlcle Size (cm)

Figure 20. Effect of catalyst particle size distribution on polymer particle size distribution (Talbot, 1990).

The dynamic model developed by Talbot (1990) also accounts for the effects of a varying bed height and inert gas on the fluid dynamics. These modifications permit the model to simulate reactor startup behavior. It was found that when the bed height is allowed to increase to

a desired level before product removal begins, the tem-

perature rise in the bed is more rapid than when the bed

height is kept constant at a desired level. Hence, a temperature overshoot may occur when a varying bed

height startupprocedure is employed.

balance approach, Talbot (1990) also developed a model for polymer particle size calculations. This model can be used to predict the effects of catalyst particle size and distribution on polymer particle size development. For instance, a distribution of catalyst particle size can produce particle size distribution of polymer product which is broad and skewed, as shown in Figures 19 and 20 (Talbot, 1990). Understanding the dynamic behavior of a reactor is important not only for reactor design and process control under normal operating conditions, but also for product quality control, particularly during the resin grade tran- sition period. Although gas phase polymerization is a versatile process, as shown in Figure 9, specific reactor operating conditions produce a specific grade of polyeth- ylene. Tomeet the market demand,a polyethyleneprocess

is required to produce a number of different product grades

in the same reactor. For example, the Phillips plant produces as many as 50 different grades of polymer (McDaniel, 1988). Therefore, all polyethylene manufac- turers regularly face grade changeover operations, and

Usinga population




TIME, hr





Figure 21. Grade changeover scheme for ethylene polymerization (Sinclair, 1987).

detailed procedures for grade changeover are important knowledge for an individual manufacturer. Very little information on this topic is available in the open literature. Sinclair (1987) has outlined the considerable economic importance and difficulties of grade changeoveroperation. Figure 21 shows an example of lost production time and off-specificationproduct during changeoverfromonegrade to another (Sinclair, 1987). In the case shown in Figure

to 88% of the

nameplate capacity at the beginning of the grade change point a to maintain safe reactor conditions. The polym- erization rate is held at the reduced rate until point b while the reactor conditions change toward those required for the new target grade. At point c, the polymerization rate is gradually increased within the reactor safety limitations. At point d, the polymerization rate reaches the lower specification limit for the product, and steady- state conditions for the new grade are finally reached at point e. Off-specification product is produced between the beginning of the grade change (point a) and the attainment of specifications for the new grade (point d). The amount of this product is the area under the polymerization rate curve from point a to point d. The lost production time can be estimated as the difference between what the reactor could have produced during the grade changeover and the actual production during the change. Therefore, knowledge of how to minimize off- specificationproduct and maximizeproduction rate during grade changeovers is crucial in plant operation. Ramanathan and Ray (1991) have simulated grade transition strategies for two fluidized bed polyolefin reactors in series and found that catalyst feed rate, hydrogen feed rate, and gas bleed rate are important variables for accomplishing rapid grade transitions while maintaining the polymer production rate. More recently, optimal open-loop policies for accomplishing grade changeovers of polyethylene production at commercial operating conditions were studied by McAuley and MacGregor (1992)using dynamic model-based optimiza- tion. McAuleyand MacGregor(1991)related the multiple active center model with appropriate simplifications to the melt index of polyethylene so that the model can be used to predict melt index which can be measured in the industriallaboratory environment. Adjustableparameters in the model are updated on-line by a recursive prediction error method as measurement data become available. Therefore, the model can be used for on-line prediction based on laboratory measurements and for grade transition policy studies. A desirable grade transition policy must satisfy the following conditions:

21, the polymerization rate must be reduced

Ind. Eng. Chem. Res., Vol. 33, No. 3, 1994 469 Oo4 0 0.4 0.8 1.2
Ind. Eng. Chem. Res., Vol. 33, No. 3, 1994
0 0.4
0 0.4
Figure 22. Steady-state copolymercompositiondistributionsof two
grades of polyethylene (McAuley and MacGregor, 1992).
Figure 23. Compositiondistribution of copolymerat end of grade
transition for polyethylene (McAuleyand MacGregor, 1992).
a. Safety and stability of reactor performance during
the grade changeover must be ensured. Excursions above
the sticking temperature of the polymer and disruption
of fluidization, heat transfer, and mixing in the bed cannot
be tolerated.
b. Minimum quantity of off-specification polymer and
short transition time are desired.
c. All reactor operating variables should settle at their
prespecifiedfinalvalueswhen grade transition is complete.
Production of consistent on-specification product under
new reactor operating conditions is required.
With these conditions as criteria, McAuley and MacGre-
gor (1992) used the following integral quadratic objective
functions to obtain optimal transition policies.
a,(ln MI, - In
+ a,(pi - P,,)~I dt + J~[w5(VP-
- Rpde~)~]dt + wS(Tf8p
wT(Bd8p - Bwde8)z
where Fl(u)and Fz(u)are objective functions, MI and p
are the melt index and density of polymer, respectively,
Tfsp and Tdes are the final steady-state temperature and
desired temperature, Vp and Vpdes are the gas bleed valve
position and desired valve settling position, R, is the
polymerization rate, B, is the bed level, to and tf are the
grade transition beginning time and the time at which the
reactor reaches steady state beyond the end of the
changeover, and wi (i = 1-8) and ai (j = 1-4) are weighting
factors related to the relative costs of deviations of each
variable. Subscripts c and sp represent the accumulated
value and the set point, respectively. Three policies of
grade changeover were demonstrated using eqs 3 and 4:
produced before the start of the grade changeover remains
in the reactor after the changeover, having been mixed
with the new polymer produced during the changeover.
Figures 22 and 23 suggest that a longer time is required
for a complete grade changeover of the composition
distribution, compared with the times for average prop-
Except the bed level, all of the operating variables
discussed above only affect the properties of polymer
produced after the beginning of the grade changeover,
because the properties of dead polymer are determined
by the kinetics of polymerization and reactor operating
conditions at the time at whichthe polymer was produced.
Therefore, reactor operating variables only affect newly
born polymer in the reactor. Thus, to achieve fast grade
changeover, properties of the new polymer should be
brought as close as possible to their targets. On the basis
of this philosophy, McAuley and MacGregor (1993)
designed a nonlinear model-inverse based feedback con-
troller, choosing instantaneous MIi and pi as controlled
variables, instead of accumulated values. An on-line
polymer property inferencescheme (McAuleyand MacGre-
gor, 1991) and open-loop optimal policies (McAuley and
MacGregor, 1992) are implemented in the designed
controller. The control action alsoaccounts for impurities
and catalyst composition disturbance during grade
changeovers. Simulation results show that the controller
is able to control the properties of many grades of
polyethylene andto accomplishnearly optimal changeover
between these grades.
5. Polymer Properties and Characterization
The objective of process development and kinetic/
dynamic modeling studies is to produce a quality polymer
with high productivity. Hence, it is very important to
understand the relationships among polymer end-use
properties, polymer characteristics, and reactor operating
5.1. Physical and Mechanical Properties. Physical
McAuley and MacGregor (1992) pointed out that the
molecular weight and composition distributions may not
reach the steady-state specification even though the
average properties MI, and p, of the polymer in the reactor
are within their specification limits, as shown in Figures
and 23 for a changeover from grade B to grade A. Figure
shows the steady-state composition distributions for
both grades A and B. At the end of the changeover, the
composition of grade A is as shown in figure 23, which is
very different from the steady-state grade A shown in
Figure 22. This is because some of the old polymer
and mechanical properties of polyethylene material depend
on density, crystallinity, molecular weight and its distri-
bution, chemical composition, and other characteristics.
These variables, in turn, are controlled by polymerization
conditions and are not mutually independent. The
essential variable, whichaffects allthe physical properties,
is the chain microstructure of polyethylene. Backbone
linearity contributes to improvetensile and tear properties,
while branching increases toughness of the material. The
combination of the two in LLDPE resins gives physical
properties that are generally superior to those of conven-

470 Ind. Eng. Chem. Res., Vol. 33, No. 3, 1994





















0 97

Figure 24. Relationships between polymer properties and resin density (Foster, 1991).

tional LDPE (Sinclair, 1983). Commercial resins are mainly specified by density and molecular weight (melt index) and distribution (melt flow ratio). In other words, the processingand end-useproperties are mainly governed by these variables. Density of PE depends on the following factors: the number of branches; branch chain length; branch fre- quency distribution; chemical composition distribution; molecular weight and distribution. 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. Hence, any physical properties related to crystallinity, such as stiffness and yield stress, will be affected by branching or chemical composition. Density decreases with an increase in molecular weight. This is because of the inhibition of crystallization by longer molecular chains. In commercial copolymersmade using heterogeneous Ziegler-Natta cat- alysts, high molecular weight fractions generally contain less comonomer units and fewer branch chains, while low molecular fractions contain 2-4 times the comonomer content of the high molecular weight fractions. This suggeststhat comonomerunits are more sensitive to chain transfer and other termination reactions during polym- erization. This is also one of the reasons why the density of polymer depends on MWD. The basic relationships between polymer properties and density can be summa- rized in Figure 24 (Foster, 1991). The effects of molecular weight and distribution on mechanical properties of polymers are very complex (Martin et al., 1972; Nunes et al., 1982). Stiffness and hardness of PE decrease with an increase in molecular weight for molecularweights < 106. For molecular weights

is independent of molecular

weight (Margolies, 1970). The stiffness of a resin is a function of crystallinity. As molecular weight of PE increases, the percent crystallinity decreases so that the stiffness decreases (Shirayama et al., 1972; Nunes et al.,

> lo6, the flexural stiffness


As the molecular weight of PE increases, the tensile impact and the Izod impact increase and reach a maximum at a molecular weight of about 106, and then decrease gradually with an increase in molecular weight (Margolies,


Figure 25 summarizes general relationships between polymer properties and molecular weight (Foster, 1991). The significance of these relationships also depends on

\ \ L

; v)

















Figure 25.

molecular weight (Foster, 1991).

Relationships between polymer properties and resin

h 0.93-0.97HDPE




0.91-0.93 giml


0.91-0 93 glml

Figure 26.

Schematic of molecular structures for commercial


MWD. Resins with a broad MWD show enhanced melt shear thinning compared to that of narrow MWD resins, and therefore tend to flow more easily at the high shear rates prevailing in resin processing. Hence, broad MWD resins are widelyused where extrusion cross-sectionalareas are relatively large and where a broad MWD results in easier processing. In blow molding of large containers, the combination of high molecular weight and broad MWD gives high melt strength, low parison sag and easy processing; the same applies for pipe extrusion (Sinclair,



Polyethylene Characterization. Commercial

polyethylenes are essentially branched polymers, LDPE with long-chain branching, and both LLDPE and HDPE with short-chain branching, as illustrated Figure 26. Hence, polyethylene characterization requires analysis of not onlymolecularsizebut alsobranching and composition. In commercial practice, polyethylene resin grade speci- fication is generally quoted in terms of melt index and density (ASTM, 1990)rather than molecular weight and composition. Polydispersity is indicated using a melt flow ratio (Spitz, 1987) rather than the ratio of M,/i@,,. The melt index and melt flowratio give onlyrelative properties of a polymer and cannot provide detailed information about the polymer structure. It is difficult to relate melt index to polymerization conditions. However,melt index measurement is the most convenient way to determine relative molecularsizein an industrial environment, which is why melt index is generally used for PE quality control.

Ind. Eng. Chem. Res., Vol. 33, No. 3, 1994
















y 16 W $ 13 E 10

W 2


g 6.3






1 - - -

2.0 0 01






, "1





m 5 12



a 10






MELT INDEX, 9/10min

Figure 27. Relationship between weight-averagemolecular weight and melt index of polyethylene. (m) Sinclair (1983);(+) Spitz (1987); (0)LLDPE (1-octene as comonomer); (*) LLDPE (1-butene as comonomer); (A)HDPE (Bremner and Rudin, 1990).

One way to relate the melt index to polymerization conditions is to correlate melt index with molecular weights. Unfortunately, only limited data are available in the literature for correlation of melt index and molecular weight. Figure 27 shows the relationship between weight- average molecular weight and melt index (Sinclair, 1983; Spitz, 1987; Bremner et al., 1990). It can be seen that molecular weight increases with a decrease in melt index; however, there is no simple relationship between them over a broad range of molecular weights. According to measurements by Bremner and Rudin (1990),Mwand the melt index relationship depend on the density or polymer composition. The data for LLDPE with 1-butene as a comonomerare very scattered. The data for LLDPE with 1-octene comonomer are in agreement with the data reported by Sinclair (1983)and Spitz (1987). These data

can be correlated with the solid line shown in Figure

A linear relationship between MWand melt index can only be used over a limited range of molecular weights. The results of Bremner and Rudin show that, at the same melt index,HDPE has a higher molecular weight than LLDPE. Poor correlation between molecular weight and melt index can be attributed to both measurement error and the nature of PE structure. Commercial low-pressure PES are copolymers. Polymers with the same molecularweight may have different compositions, i.e., shortbranches. The melt flowproperty depends on both molecular weight and branching. Hence, polymers with the same molecular weight may have different melt index values. In other words, melt index is a function of the whole distribution of molecular weights and compositions and not just a single molecular weight average value (Bemner et al.,1990;Pang and Rudin, 1992). Furthermore, melt indexes and mo- lecular weights published in the literature are often measured using commercial polymers, for which the synthesis histories of the samples axe often unknown. Therefore, caution must be taken when using such data. Figure 28 showsthe relationship between polydispersity and melt flow ratio, which is used to indicate broadness of molecular weight distribution (Spitz, 1987). Although the data are scattered, polydispersity seems to be linearly related to flow ratio. Although melt index and melt flow ratio can provide relevant information about molecular weight and distri- bution, the true molecular weight and its distribution are required to better understand polymerproperties andtheir










Figure 28. Relationship between polydispersity and melt flowratio of polyethylene (Spitz, 1987).

relationships with polymerization mechanisms and reactor operating conditions. Size exclusion chromatography (SEC and GPC) can, in principle, provide both molecular weight averages and distribution. The Waters Associates 150C ALC/GPC (high-temperature GPC) is commonly used in both academic and industrial research for PE molecular weight measurement. o-Dichlorobenzene or 1,2,44richlorobenzene is used as solvent (Kissin, 1985). Virtually all polyethylenes are insoluble at room temper- ature; hence, the GPC is normally operated at 135-150 OC. A GPC consists of two essential sections: a separation sectionand a detector section. A GPC columnfractionates according to the hydrodynamic volume of polymer solute in the mobile phase. Since the hydrodynamic volume depends on both molecular weight and composition (branch frequency), the polymer eluted in the detector cell will not be uniform either in molecular weight or in composition (Vela Estrada and Hamielec, 1992-93). Hence, strictly speaking, GPC can only be valid for molecular weight measurement of homopolyethylene. Current commercially available detectors for GPC measurements include (i) a differential refractive index detector (DRI) to monitor polymer concentration in the eluant as a function of retention time; (ii) continuous measurement of the weight-average molecular weight of the eluting species using a low-anglelaser light scattering photometer (LALLS); (iii) continuous measurement of the molecular weight using a continuous viscometer (CV). The LALLS signal is proportional to CMand CV responds according to CMa,while DRI is proportional to C, where C is concentration and M is molecular weight (Pang and Rudin, 1992). DRI and CV are insensitive to high molecular weight and low concentration fraction. On the other hand, LALLS is very sensitive to high molecular weight fractions but not to low molecular weight. There- fore, DRI and CV measurements may lose information on high molecular weight fractions and LALLS may lose the information on low molecular weight fractions, so that three different molecular weights and distributions may be obtained for the same PE sample using three different detectors. Pang and Rudin (1992) concluded that no current GPC technique provides a picture of the true molecular weight distribution of broad MWD polyethyl- ene. For copolymer, such as LLDPE, GPC can provide only approximate values of molecular weight and its distribution

472 Ind. Eng. Chem. Res., Vol. 33, No. 3, 1994

if the entire polymer sample is used. Since copolymer has

a bivariate distribution of molecular weight and chemical

composition, it is desirable to measure both molecular weight and chemical composition of the polymer. This can be achieved by a combination of temperature-rising elution fractionation and GPC techniques. The fractionation of polyethylene copolymer on the basis of molecular weight alone can occur only if a solvent gradient elution technique is performed at a temperature abovethe melting point, while the fractionation takes place mainly on the basis of the degree of short-chain branching if the temperature-rising elution technique is employed. The term “temperature-rising elution fractionation” (TREF) was first coined by Shirayama et al. (1965) to describe a method used to fractionate low-density poly- ethylene according to the degree of short-chain branching. The actual technique had been described earlier by Desreux and Spiegels, who first recognized the potential of using elution at different temperatures to achieve a crystallization separation (Hawkins and Smith, 1958). Although the principle of TREF has been recognized for over four decades, significant development of TREF analytical techniques occurred only in the past decade. It is interesting to note that the development of TREF technology and application has taken place almost ex- clusivelywithin industrial research laboratories (Hawkins and Smith, 1958;Shirayama et al., 1965,1972; Wild and Ryle, 1977;Bergstrom and Avela, 1979;Nakano and Goto, 1981; Wild et al., 1982a,b; Knobeloch and Wild, 1984; Kimura et al., 1984; Usami et al., 1986; Kelusky et al., 1987; Cady, 1987; Hosoda, 1988; Mirabella et al., 1987, 1992;Wild, 1989; Wilfong and Knight, 1990). Only very recently has this technology captured the attention of the academic community (Schouterden et al., 1987;Kulin et al., 1988; Hamielec, 1989; Glockner, 1990; Vela Estrada and Hamielec, 1992-93). No TREF instrument has been commercialized to date. Hence, TREF instruments are currently constructed by individual laboratories. Few university-based polymer institutes have built TREF equipment in recent years,e.g., McMaster University (Vela Estrada and Hamielec, 1992-93), University of Waterloo, and the University of Alberta. The development of TREF is clearly driven by the need to understand the nature of polyethylenecopolymerswhichexhibit behaviorindicative of considerable structural heterogeneity and to estimate the kinetic parameters associated with detailed kinetic models. The fundamental basis of TREF technology is the relationship between crystallizability and branching of polymer. Crystallizability of a polymer decreases with an increase in branch frequency in polymer chains. In other words, polymer chains with different branch concentra- tions crystallize at different temperatures. The higher the degree of crystallinity, the higher is the dissolution temperature of polymer chains; hence, polymer chains with fewer branches dissolve at higher temperatures. To separate polymer chains with different branch concen- trations effectivelyin a TREF process, the polymer is first coated on a spherical supportmaterial, normally a column packing, by slowly cooling a polymer solution within the support material. Thus, polymer chains with fewer branches deposit on the supportfirst, and polymer chains with more branches deposit later as the temperature decreases. To separate the polymer fractions,the polymer precipitated on the support is stripped off by extraction with a solvent at increasing temperatures. Therefore, polymer eluted at lower temperature has a higher fre- quency of branches and polymer eluted at higher tem-




Figure 29. Schematic representation of temperature rising elution fractionation.

perature has a lower branch concentration. One may envisage polymer chains with different branch concen- trations depositing on the support in different layers, as shown in Figure 29. In a real TREF process,the thickness of polymer on a support may be very thin, depending on the surface area of support and amount of polymer used. Detailed information regarding polymer deposition on the support has not been reported. Two types of TREF mode have been developed, namely, preparative TREF and analytical TREF. In the prepar- ative mode, the polymer sample is fractionated and subsequently each fraction is analyzed by additional methods such as GPC and NMR. Thus, relatively large samples (1-log) are separated (Wild,1990,1993;Glockner, 1990; Vela Estrader and Hamielec, 1992-93). Typical preparative TREFs have been designed by Bergstrom and Avela (1979)and Usami et al. (1986). The main elements include an oven or a bath with temperature control, a packed column, and solvent and solution flow systems. A TREF column is normally packed with glass beads (Nakanoet al.,1981),ChromosorbP (Bergstromand Avela, 1979; Wild et al., 1982; Kulin et al., 1988), or silanated silica gel (Kelusky et al., 1987). Coolingand heating rates are in the range 0.5-6.5 “C/h. Analytical TREF refers to fractionation systems in which the separated polymer eluate is continuously monitored as a function of elution temperature. This type of operation usually fractionates only small quantities (<0.5g) of polymer (Wild,1990,1993). Typical analytical TREF systems have been developed by Wild et al. (1982b),Usami et al. (1986),Kelusky et al. (19871, and Hazlitt (1990). Perhaps the most advanced analytical TREF developed to date is that by Hazlitt (19901, shown in Figure 30. The instrumentation used in analytical TREF is derived directly from the already well- developed GPC with necessary modifications of a tem- perature rise control system (Wild, 1990). Analytical TREF is a quick method to obtain the short-chain- branching distribution, but the amount of sample used is too small to measure the MWD of separate fractions by GPC. The main results for polyethylene homo and copolymers using TREF measurement are summarized as follows. The number of branches in fractions (methyl groups/ 100 C) decreases with an increase in elution temperature (Shirayama et al., 1972; Wild et al., 1982; Usami et al., 1986; Schouterden et al., 1987; Kelusky et al., 1987;

Ind. Eng. Chem. Res., Vol. 33, No.