Fluidized Bed Reactor for Catalytic Olefin Polymerization · FLUIDIZED BED REACTOR FOR CATALYTIC...

i

Fluidized Bed Reactor for Catalytic Olefin Polymerization

Kinetics and Fluidization

Gerben B. MEIER

ii

CIP-gegevens Koninklijke Bibliotheek, Den Haag Meier, Gerhardus B. Fluidized bed reactor for catalytic olefin polymerization ISBN 90-36514894 Trefw.: fluidized bed reactor, olefin, polymerization, metallocene catalyst, sorption, particle mixing, segregation, draft tube, modeling Copyright 2000 by G.B. Meier, Enschede, The Netherlands No part of this book may be reproduced by any means, nor transmitted, nor translated into machine language without permission from the author Referent: Dr. ir. S.M.P. Mutsers Cover design: L.H.M. Meier and M.G.J.J. Hartgerink The research described in this thesis was performed at the Twente University, Enschede, The Netherlands. The investigations have been funded by BRITE-EURAM Project CATAPOL (BE 96-3022).

iii

FLUIDIZED BED REACTOR FOR CATALYTIC OLEFIN POLYMERIZATION

PROEFSCHRIFT

ter verkrijging van de graad van doctor aan de Universiteit Twente,

op gezag van de rector magnificus, prof. dr. F.A. van Vught,

volgens besluit van het College voor Promoties in het openbaar te verdedigen

op donderdag 30 november 2000 te 16.45 uur.

door

Gerhardus Bernardus Meier

geboren op 22 april 1972 te Stad Delden

iv

Dit proefschrift is goedgekeurd door de promotor Prof. dr. ir. W.P.M. van Swaaij

en de assistent-promotor Prof. dr. G. Weickert

v

Aan Dynah

vi

vii

Contents Summary 1

Chapter 1: General introduction 5

Outline of the thesis 7

History of the project 8

Chapter 2: Comparison of gas and liquid phase polymarization of 9 propylene with a heterogeneous metallocene catalyst

Abstract 9

Introduction 11

Experimental 11

Sorption of propylene 11

Polymer sample characterization 13

Experimental set up for liquid propylene polymerizations 13

Experimental set up for gas phase polymerizations 14

Catalyst system 15

Liquid phase polymarization procedure 16

Gas phase polymarization procedure 17

Sorption theory 17

Sorption of hydrogen 19

Kinetic model 19

Results from sorption measurements 20

Results of liquid phase experiments 23

Influence of Al/Zr ratio 23

Influence of temperature 23

Influence of hydrogen 25

Results of gas phase experiments 27


viii

Influence of pressure 27

Comparison between liquid and gas phase 29 polymarization kinetics

Conclusions 32

Notation 33

Greek 33

Sub- and superscripts 33

Abbreviations 34

Literature 34

Appendix 36

Chapter 3: Gas phase polymerization of propylene with a 37 heterogeneous metallocene catalyst. The influence of temperature, pressure and hydrogen on the reaction kinetics and molecular weight distribution

Abstract 37

Introduction 39

Experimental 39

Kinetic model 40

Molecular weight and molecular weight distribution 41

Results 43


Influence of pressure 50

Influence of hydrogen 52

Interpretation of the acceleration period 55

Molecular weight and molecular weight distribution 56

Conclusions 63

Notation 64

Greek 65

Subscripts 65

ix

Abbreviations 65

Literature 65

Chapter 4: FBR mini-plant for catalytic olefin polymerization: 67 Particle mixing and propylene polymerization

Abstract 67

Introduction 69

Sheeting, elutriation and agglomeration 69

Present work 70

Experimental 71

Fluidized bed reactor set up 71

Catalyst system 73

Catalyst injection system 74

Electrostatics 75

Experimental procedures for mixing and segregation experiments 75

Experimental procedures for semi-batch polymerizations 75

Results 76

Mixing of particles 76

Seggregation of particles 77

Batch polymerization experiments 79

Simplified reactor model 83

Discussion 86

Conclusion 87

Notation 88

Greek 88

Subscripts 88

Abbreviations 89

Literature 89

x

Chapter 5: FBR mini-plant for catalytic olefin polymerization: 91 Controlled particle mixing, propylene polymerization and reactor modeling

Abstract 91

Introduction 93

Experimental 94

FBR set up 94

Catalyst system 95

Experimental procedure for solids circulation measurements 96

Experimental procedure for semi-batch polymerizations 96

Results 97

Solids circulation 97

Semi-batch polymerizations 99

Injection of hydrogen 103

Reactor model 105

Modeling results 112

Injection of hydrogen 116

Discussion 119

Conclusion 120

Notation 121

Greek 121

Sub- or superscripts 122

Abbreviations 122

Literature 122

Samenvatting 123

Dankwoord 127

Levensloop 129

SUMMARY

1

Summary Due to the continuous improvement of catalysts and processes, polyolefins have become one of the most important plastics in the world. Polyolefins can be produced at low costs with a variety of end-use properties. Nowadays, the most important propylene polymerization processes are executed in the liquid or the gas phase or a combination of both. In contrast to propylene polymerizations in slurry phase, only a few studies have been published concerning gas or liquid phase polymerization. Especially experimental investigations of gas phase polymerizations at relevant process conditions, i.e. high pressure and temperature, are rare. The most widely established industrial gas phase technology is the fluidized bed reactor operating at 10 – 30 bar. Conversion per pass is kept low, 1-3%, to diminish concentration and temperature gradients in the reactor, which both affect the polymer properties. In 1 to 3 hours, polymer particles with a broad size distribution are obtained. There are no experimental data available in the open literature about the polymerization in such a reactor. The study reported in this thesis is concerned with a modified small-scale fluidized bed reactor to study aspects of the gas phase polymerization of propylene with a heterogeneous metallocene catalyst at conditions resembling those of industrial units as well as strongly deviating conditions. Controlled thermal gradients are provoked as they may be of interest to broaden the molecular weight distribution. To study the polymerization in such a reactor an experimentally validated kinetic model is required. The kinetic model, obtained from experiments at isothermal and isobaric conditions, is used to describe the polymerization at non-isothermal conditions in the fluid bed reactor with related molecular weight distribution of the polymer. Other important aspects that will be studied are electrostatic charging, particle mixing and elutriation of fines. Polymerization kinetics Propylene has been polymerized in both liquid and gaseous propylene in stirred tank reactors with Me2Si[Ind]2ZrCl2 / MAO / SiO2(Grace) as metallocene catalyst. Gas phase polymerizations have been executed in the temperature range of 40 to 80°C and pressures of 5 to 25 bar. Polymerizations in liquid propylene have been carried out at temperatures between 40 and 70°C and at hydrogen concentrations between 0 and 2.2% in the gas cap. To describe the reaction kinetics a simplified kinetic model has been developed, which assumes first order kinetics with respect to the number of active centers and the monomer concentration in the polymer. The monomer concentration in the amorphous part of the semi-crystalline polymer, i.e. near the active center, has been determined experimentally. At low pressures, Henry’s law can be used to describe the concentration. At higher pressures, the Flory-Huggins equation is used to fit the sorption isotherms. Decreasing values for the Flory-Huggins interaction parameter with increasing temperature have

SUMMARY

2

been found. The sorption data have been used to compare the relative reaction rates observed in gaseous and liquid propylene. Lower relative reaction rates were found in gas phase compared to the experiments in liquid phase. The activation energies found for the experiments in both phases are in the same order of magnitude. In another series of kinetic experiments, propylene have been polymerized in the gas phase at different temperatures, pressures and hydrogen concentrations using Me2Si[Ind]2ZrCl2 / MAO / SiO2(PQ) as catalyst. This catalyst is supported on another support and with a different metallocene loading, compared to the catalyst used for the experiments described above. The reaction rate curves have been described with a kinetic model, which respects both the initially increasing polymerization rate and the deactivation of the catalyst. At high temperatures, pressures and hydrogen concentrations reduced polymer yields have been found, which was interpreted as a thermal runaway on particle scale. The polymer samples were analyzed on their molecular weight by Gel Permeation Chromatography (GPC). The molecular weight and molecular weight distribution of the polymer samples could be described with a “two-site model”. At constant temperature, the chain transfer probability of both site 1 and 2 was found to depend only on the hydrogen concentration divided by the monomer concentration. Polymerization in a small-scale FBR At the High Pressure Laboratory of the Twente University, a small-scale fluidized bed reactor has been constructed for the catalytic polymerization of olefins under pressure. Also particle mixing and segregation have been studied at increased pressure. It appeared that the solids mixing is relatively fast compared to the residence time of catalyst particles in case of a polymerization process. However, the smaller particles accumulated in the upper zone. Moreover, electrostatic charging caused the forming of a layer of small particles at the reactor wall with increasing thickness in time. The particles were redispersed after injection of an anti-static agent. Semi-batch propylene polymerization experiments at different fluidization velocities showed that the vertical temperature gradients are created, which are caused by catalyst segregation as the principle factor. Specially at low gas velocities both segregation and mixing can be different under reacting conditions compared to non-reacting conditions due to different particle-particle interactions. Despite the fact that fluidized bed reactors can hardly be scaled-up from experiments in reactors with diameters below 30 cm, small-scale fluidized bed reactor experiments can contribute to address and quantify incomplete mixing and electrostatic charging effects. However, the balance between mixing and segregation cannot be compared to an industrial sized unit and vertical mixing in this un-modified fluid bed reactor should be enhanced to be more representative.

SUMMARY

3

Therefore, the small-scale fluidized bed reactor has been equipped with a draft tube and conical bottom section to control the vertical solids mixing. In this set up, particles are forced to move upwards in the draft tube section under fast fluidization conditions and enter the annulus section where they flow as a moving bed. The internal solids circulation rate is a non-linear function of the gas velocity of the unit. Strongly reduced elutriation and entrainment have been observed, compared to experiments without draft tube. Semi-batch propylene polymerizations have been executed at elevated pressures. Although the draft tube can be used to increase the solids mixing, the emphasis in the present work is put on the creation of different zones in the reactor. The temperature gradients in the reactor can be controlled from almost absent to large gradients by changing the solids circulation rate. Hydrogen injections lead to an instantaneous increased polymerization rate, probably due to the reactivation of dormant sites. The irreversible deactivation rates of dormant and active sites seem to be the same. Moreover, hydrogen appeared to be very effective to broaden the molecular weight distribution. A compartment model is developed, which is able to describe the temperature profile in the reactor and related molecular weight distribution of the polymer.

CHAPTER 1

4

CHAPTER 1

5

Chapter 1 General introduction. Polyolefins are today the major plastics materials in terms of production capacity thanks to its versatility, low costs and excellent processability. The global polypropylene market has become one of the fastest growing and the production exceeds 30 million tons in 2000. Since the discovery of the Ziegler-Natta catalysts in the early 1950’s, highly active and stereo-selective catalysts have been developed. The high catalyst activity accounts for low catalyst residues in the polymer, which can be left in the final product. Modern polymerization processes require heterogeneous catalysts to control the morphology of the polymer powder and to reduce the required amounts of cocatalyst. Thanks to the continuous development of new types of catalysts and processes, new materials have been introduced to the market thereby expanding the market share of polyolefins. Since the 1980’s chiral metallocene catalysts were synthesized, which were able (in combination with methylaluminoxane) to polymerize olefins with controlled molecular architectures. The narrow molecular weight distribution of the obtained polymer reflects the single site behavior of this type of catalyst. Despite of the great potential of the metallocene catalysts, Ziegler-Natta catalysts are still responsible for the major part of the polyolefin production. A clear trend towards gas phase polymerization processes can be observed where solvent recycling is not required and a larger variety of products can be obtained. The most widely established industrial gas phase technology is the fluidized bed reactor, see figure 1. Here, the bed is kept in the bubbling regime by recycling the reaction gases with conversion of 1-3% per pass through the reactor. The heat of reaction is removed by cooling the circulating gas and sometimes by a partial condensation in order to use the heat of evaporation of the condensates (condensed mode). The reactor is operated at constant pressure, in general between 10 and 30 bar.

CHAPTER 1

6

Polymer properties like the molecular weight distribution of the polymer are influenced by the type of catalyst, temperature, monomer concentration and hydrogen concentration. Temperature and concentration gradients in fluidized bed reactors are a strong function of the solids mixing and segregation rates inside the reactor, which are directly related to the reactor scale. At an industrial scale, bubbles rising from the distributor plate tend to accumulate in the center of the reactor, causing an upward “gulf stream” of polymer powder. Depending on the H/D a single or more mixing cells may occur. This strong mechanism of solids mixing reduces temperature and concentration gradients in fluid beds. In a small-scale unit, axial mixing is lower and “gulf streaming” is less prominent or absent. In the open literature on olefin polymerization there are no experimental data available on the fluidized bed reactor, probably because of the knowledge protection by industries. Moreover, the investment and operation costs for a pressurized experimental facility of acceptable size are high. A kinetic model based on experiments at relevant process conditions, i.e. same pressure and temperature range, is crucial for studying the polymerization in a fluidized bed reactor.

Figure 1: Several characteristic length scales in a fluid bed reactor for catalytic olefin polymerization. The reactor diameter is on the order of meters, the particles are tens to hundreds of microns and the sub fragments on the order of hundreds of nanometers.

Productremoval

Catalystinjection

Cyclone

Heatexchanger

Compressor

Catalystparticle

Heat transferto bulk

Mass transferfrom bulk

Catalyst fragmentssurrounded by

polymer

Monomer

CHAPTER 1

7

Outline of the thesis The objective of the work presented in this thesis is to study the important aspects of the polymerization of propylene with a heterogeneous metallocene catalyst in kinetic and fluid bed reactors. Some key aspects are the applicability of kinetic models derived from experimental data measured in stirred tank reactors for small-scale fluidized bed reactors, the influence of particle mixing and segregation on the temperature profile in fluidized bed reactors and how to scale-down a fluidized bed reactor for olefin polymerization. Using a modified fluidized bed reactor, the polymerization of olefins is studied at conditions resembling those of industrial fluidized beds as well as strongly deviating conditions with controlled thermal gradients. The latter may be of interest for a controlled broadening of the molecular weight distribution. At present heterogeneous metallocene catalysts, which are able to produce polyolefins with a high molecular weight at high production rates, are not commercially available. The heterogeneous metallocene catalyst used in the present work is the well-known metallocene rac-Me2Si[Ind]2ZrCl2 / MAO / SiO2. This catalyst produces polyolefins with a narrow molecular weight distribution but with a low molecular weight. Chapter 2 describes the gas and liquid phase polymerization of propylene using the heterogeneous metallocene catalyst in a 0.5 and 5 liter stirred tank reactor, respectively. The influence of temperature and pressure on the reaction rate has been investigated and is fitted with a relatively simple kinetic model. Sorption measurements have been carried out to determine the concentration of monomer in the polymer as a function of process conditions. Using the sorption data, the obtained kinetic data in gas and liquid phase have been compared. Chapter 3 presents the results of a kinetic study with the same metallocene catalyst but on a different silica support. Gas phase polymerizations at different temperatures, pressures and hydrogen concentrations have been executed and described with a kinetic model, which considers both the activation and deactivation of the catalyst. The molecular weight and molecular weight distribution of the obtained polymer samples have been measured by Gel Permeation Chromatography (GPC). Chapter 4 describes the small-scale fluidized bed reactor as built up in the High Pressure Laboratory of the Twente University. Mixing and segregation experiments have been executed, which have been used to quantify the effects on the temperature profile observed during polymerization experiments. Chapter 5 describes the controlled mixing in the small-scale fluidized bed by installation of a draft tube and conical bottom section in the reactor. Now, particles are forced to

CHAPTER 1

8

move upwards in the draft tube section under fast fluidization conditions and enter the annulus section where they move in a densified form under the action of gravity. In this way a solids circulation is set up, which can be controlled by the inlet gas velocity. Because the draft tube and annulus are generally operated at different gas velocities, both sections have different heat transport properties, temperatures and temperature gradients. Therefore, for a particle circulating through the reactor an oscillating particle temperature may occur along the particle trajectory resulting in a broadening of the molecular weight distribution. A reactor model has been developed that describes the temperature and concentration gradients inside the reactor with the related molecular weight distribution of the polymer. Each chapter has been submitted for publication and therefore has been written in such a way that it can be read independently of the other chapters. History of the project This project was started in 1996 under supervision of Prof. Dr. Ir. K.R. Westerterp and Prof. Dr. Ing. G. Weickert. After the retirement of Prof. Westerterp in 1998 the supervision was taken over in 1999 by Prof. Dr. Ir. W.P.M. van Swaaij and Prof. Weickert. The work has been done as part of a BRITE-EURAM project named Catapol (BE 3022). This project is a cooperation between 4 European universities and 4 industrial companies and is financially supported by the European Community and the participating companies. The final results of the project were presented during the first European conference on the reaction engineering of polyolefins (ECOREP), July 2000 in Lyon.

CHAPTER 2

9

Chapter 2 Comparison of gas and liquid phase polymerization of propylene with a heterogeneous metallocene catalyst. Abstract Sorption measurements are executed to study the sorption behavior of propylene in the semi-crystalline polymer. Decreasing values for the Flory-Huggins interaction parameter with increasing temperature are obtained. Large deviations are found, especially at higher temperatures, compared to data from literature. Propylene is polymerized in both liquid and gaseous propylene with rac-Me2Si[Ind]2ZrCl2 / MAO / SiO2(Grace) as metallocene catalyst. Lower relative reaction rates are found in gas phase compared to the experiments in liquid phase. The activation energies found for the experiments in both phases are in the same order of magnitude. However, the sorption data used, literature vs. experimental, have a large effect on the determined kinetic parameters.

CHAPTER 2

10

CHAPTER 2

11

Introduction In comparison with the conventional Ziegler-Natta catalyst systems, metallocene catalysts offer a higher versatility and flexibility for the synthesis and control of polyolefin polymers. Despite all the development efforts in the area of synthesis of new catalyst compounds, few experimental studies are found in the open literature concerning the kinetic behavior of these metallocene systems at relevant process conditions. Especially kinetic data for gas and liquid phase, i.e. liquid propylene, polymerizations are scarce. Moreover, only a few studies are known in which the kinetics in both phases are compared. For modern propylene polymerization processes, which contain both liquid and gas phase steps, heterogeneous catalysts are required to obtain polymer particles with a narrow particle size distribution and high bulk density. However, it is well known that the catalyst behavior depends on the carrier and the supporting technique. In general, lower activities and higher molecular weights were found using heterogeneous analogues. Propylene polymerization with a heterogeneous catalyst is a complex process, involving both chemical and physical effects. In order to reach the active centers, monomer molecules have to absorb into the amorphous part of the semi-crystalline material, then diffuse to the active centers. On the kinetic level, the reaction is based on the rates of activation, propagation and deactivation processes, all dependent on various process conditions. However, the reaction mechanism is independent of the reaction phase, i.e. the local polymerization rate depends on the local reaction conditions like temperature and monomer concentration. Hutchinson et al.1 and Samson et al.2 showed that the kinetics can be unified by calculating the monomer concentration on the basis of polymer solution thermodynamics. In this paper, results from gas and liquid phase polymerizations are reported using rac-Me2Si[Ind]2ZrCl2/MAO/TIBA/silica (Grace) as catalyst system. Sorption experiments were carried out to determine the monomer concentration near the active center. The results of the kinetic experiments are compared to check whether the kinetics are independent of the reaction phase. Experimental Sorption of propylene Sorption of propylene into polypropylene has been measured using a gravimetric method. By adding propylene gas to a vessel filled with polypropylene, the increase in weight has been used to calculate the sorption of propylene in the amorphous part of the semi-crystalline material, see figure 1.

CHAPTER 2

12

The sorption vessel (4), which has a volume of 218 ml, is filled with 75 grams of polymer and is placed on a balance (3). All tubes containing propylene gas are electrically heated to prevent condensation. The temperature inside the sorption vessel itself, also heated electrically, is controlled with an Eurotherm PID controller. To reduce any manual handling during the measurements, an actuated valve is installed to add the propylene automatically. Before starting the addition of propylene, the sorption vessel is evacuated for one hour. Prior to the addition of propylene, the weight of the vessel is measured and monitored for half an hour to ensure that the balance stabilized. After stabilization, the computer will start the propylene addition by opening the actuated valve. The valve is closed when the desired pressure has been reached. After the propylene addition the computer program waits until the pressure, temperature and sample mass are stabilized and saves the data to disc. The procedure necessary to calculate the sorption data is given in the appendix.

TIPI

TI

PI

VACUUM

VENT

NC

PI

ARGON

2

3

41

5

PI

NC

Valve

Control Valve

Pressure Reducer

Relief Valve

Check Valve

1 Propylene vessel

2 Water bath

3 Balance

4 Sorption vessel

5 Capillary

Figure 1: Experimental set up for sorption measurements.

CHAPTER 2

13

Polymer sample characterization The polymer used for the sorption experiments is a product of a liquid pool experiment at 60°C, using the same catalyst system as has been used for the present kinetic measurements, see below. The density (pores excluded), crystallinity and porosity have been determined, see table 1. The porosity is determined by mercury intrusion, considering a pore size between 0.1 and 10 µm. To verify that these characteristics do not change during the experiments, the product has been analyzed prior and after the experiments. From DSC curves, it could be concluded that the crystallinity does not change in the whole range in which measurements have been executed.

Table 1: Some characteristics of the polymer used for the sorption experiments.

Polymer Density [g/ml]

Crystallinity [%]

Porosity [%]

Prior to measurements

0.91

43.5

22

After experiments 0.91 43.7 20

Experimental set up for liquid propylene polymerizations Liquid phase polymerizations have been executed in the same set up as was used by Samson et al.3. The set up comprises a 5 liter jacketed reactor, catalyst injection system and purification systems for propylene, hydrogen, nitrogen and hexane. Additionally, the reactor is equipped with a helical stirrer and a sampling system, see also Shimizu et al.4. Batch experiments have been carried out at isothermal conditions. The reaction rate has been determined by a calorimetric method assuming a constant heat transfer coefficient during the experiment.

CHAPTER 2

14

Experimental set up for gas phase polymerizations The experimental set up for gas phase polymerizations is schematically shown in figure 2.

The set up is based on the set up described by Samson et al.2. Some adjustments, see below, have been carried out in order to operate at higher pressures and to handle a different catalyst. The set up consists of a stainless steel 0.5 liter Büchi reactor for pressure up to 40 bar, a catalyst injection system, a small vessel to inject a certain amount of hydrogen, an evaporation vessel and a temperature control system.

PI

85 C

PI

NC

Valve

Control Valve, NC

Pressure Reducer

Relief Valve

Rupture Disk

1 Reactor

2 Catalyst injection

3 Hydrogen injection

4 Evaporation vessel

5 Traced mass flow

controller

PI

NC

VACUUM

NCN

O

PC

TI

PURGE

PI

TI

NOControl Valve, NO

1

2

3

4

Hydrogen

5

Figure 2: Experimental setup for gas phase polymerizations.

CHAPTER 2

15

By keeping the temperature and pressure constant during the experiment, the reaction rate can be calculated from the feed rate required to keep the reactor pressure constant. Propylene gas, evaporated in the evaporation vessel, is fed into the reactor via a traced mass flow controller. The liquid propylene in the evaporator is kept at a temperature of 70°C, to be able to execute experiments up to 25 bar. To prevent condensation of propylene, all tubes of the monomer feed system are traced at 85°C. A special helical stirrer has been used to enforce good mixing inside the reactor. Moreover, 50 grams of inert sodium chloride have been used for every experiment to prevent catalyst particles from sticking to each other and to the reactor wall. The sodium chloride also improves the heat transfer from the reacting particles to the reactor wall. The stirrer forces the powder mixture to move upwards along the reactor wall, and downwards along the stirrer shaft under the influence of gravity. A lower propeller stirrer has been mounted to whirl up the powder on the bottom of the reactor. The temperature inside the reactor, used to control the temperature within 0.1°C, is measured just above the stirrer, but in direct contact with the powder flow. To inject a dry catalyst powder, a new catalyst injection system has been developed. The catalyst is prepared under nitrogen atmosphere in a glove box and mixed with 50 grams of sodium chloride. Local high concentrations of catalyst in the reactor are avoided this way, which may lead to local hot spots. The catalyst mixture is brought into an injection vessel, which is connected to the reactor set up. Shortly before start-up, the catalyst mixture is injected via a valve into the evacuated reactor. After the experiment, the injection vessel is checked for catalyst losses, but they were never found. Catalyst system The metallocene catalyst used for the gas and liquid phase polymerizations is rac-Me2Si[Ind]2ZrCl2, see figure 3. Several groups5,6 studied the polymerization behavior of this well-known system. Spaleck et al.5 reported a rather low molecular weight (Mw = 36000 g/mol) of the polymer obtained after polymerization in liquid propylene at 70°C in the absence of hydrogen. Bonini et al.7 studied the heterogeneous analogue in slurry at low pressure (2 bar) and low temperature (40°C). The system used for this study, was kindly supplied by Witco Co. Bergkamen (Germany). It is supported on Grace silica with a concentration of 1 wt%. The MAO/SiO2 – support used for immobilization of the metallocene contained 25 wt% of alumina, giving a [Al]/[Zr] ratio of 386. The average particle size of the silica used (SD3216-30, 10 – 110 µm) is 51 µm. A SEM photo of the morphological structure of one catalyst particle is given in figure 4. It is well known that the polymerization rate can substantially be increased by adding small amounts of aluminum alkyls7, especially triisobuthylaluminium (TIBA). To be able

CHAPTER 2

16

to compare the results of gas and liquid phase experiments, the catalyst was precontacted with TIBA (Akzo Nobel) for 30 minutes prior to injection. The amount of TIBA used for precontacting has been kept constant resulting in a total [Al]/[Zr] ration of 750.

Liquid phase polymerization procedure Before each experiment, the evacuated reactor is filled with 1 liter of liquid propylene and heated to a temperature of 60°C. Next, 200 mg of TIBA is injected to scavenge all impurities. After one hour of stirring at 1000 rpm, the reactor content is purged through the drain. Then the prescribed amount of hydrogen and 2.6 liters (at 20°C) of liquid propylene are added. The hydrogen concentration in the gas cap above the liquid at reaction conditions has been calculated by calculating the volume of this gas cap and compensating for the amount sorbed into the liquid propylene. Mizan et al.8 reported about the solubility of hydrogen in liquid propylene and found a Henry type of sorption, see equation 1.

22 HHLH PkC = (1)

Here kH is the Henry coefficient and 2HP the partial pressure of hydrogen.

In table 2 some relevant data for calculation of the hydrogen concentration in the gas cap of the liquid phase reactor is given. Data about the density of liquid propylene and the vapor density have been taken from the VDI-Wärmeatlas9.

Me2Si ZrCl2

Figure 3: Catalyst used for the kinetic experiments in liquid and gas phase.

Figure 4: SEM photo of the catalyst morphology.

CHAPTER 2

17

Table 2: Some relevant data required to calculate the hydrogen concentration in the gas cap of the liquid phase reactor.

T [°C]

kH [mol/l.bar]

ρG [kg/m3]

LmC

[kg/m3] VG

[l] VL [l]

40

0.0205

33.96

475.50

2.39

2.61

50 0.0245 43.18 454.58 2.32 2.68 60 0.0289 55.27 431.00 2.23 2.77 70 0.0334 72.98 403.17 2.13 2.87

Before catalyst injection, the reactor content is brought to the desired temperature. During the experiment, the temperature of the reactor and the coolant at the in- and outlet of the reactor jacket are measured every 20 seconds. The calorimetric method introduced by Samson et al.3 has been used to calculate the reaction rate in time. After 75 minutes, the reaction is terminated by fast purging of the reactor. Due to evaporation of the liquid propylene, the temperature drops rapidly reducing the amount of polypropylene produced during the non-isothermal phase. The obtained polymer product is removed from the reactor and dried under vacuum overnight. Gas phase polymerization procedure Before each experiment, the reactor is heated to 90°C under vacuum for 30 minutes followed by flushing with nitrogen. Next, the reactor is brought to the desired temperature. The hydrogen injection vessel is pressurized with the desired amount of hydrogen. To prevent initial direct contact of the catalyst with pure hydrogen, the hydrogen injection vessel is pressurized with propylene up to 9 bar. Then the catalyst/salt mixture is injected into the evacuated reactor. A computer program written in HP VEE starts filling of the reactor to the desired pressure. Due to the automation of the set up, operation is simplified to a large extend. During the reaction, the operator can control the set up from outside a concrete bunker. The reaction rate is obtained by measuring the monomer mass flow required to keep the pressure constant during the experiment. After 60 minutes the reactor content is purged to stop the reaction. The polymer product is washed with water to separate the sodium chloride and dried under vacuum overnight. Sorption theory In case the amount of sorbed monomer is sufficiently small, i.e. low monomer-polymer interaction, Henry’s law can be used to describe the sorption in the amorphous part of a semi-crystalline polymer.

CHAPTER 2

18

Pkc Hm = (2 )

The volume fraction of penetrant inside the amorphous part can be calculated with:

Lm

mH

C

PMk=φ (3 )

At higher monomer concentrations, the interaction between monomer and polymer increases. For this situation the Flory-Huggins equation is often used, see equation 4.

20

)1()1(lnln φχφφ −+−+=P

P (4 )

Here P and P0 are the partial pressure and saturation pressure of the monomer, respectively, and χ is the Flory-Huggins interaction parameter. Samson et al.3 estimated the Flory-Huggins interaction parameter χ with the Laar-Hildebrand equation10, see equation 5.

( ) spmm

RTχδδ

νχ +−= 2

(5)

Here νm is the molar volume of the monomer, δm and δp are the solubility parameters of the monomer and polymer, respectively, and χs is the correction for entropic interaction. The solubility parameter δ is a function of temperature, but νm/R(δm-δp)

2 and χs are often nearly independent of temperature11. Therefore, the Flory-Huggins parameter should decrease with increasing temperature, according to equation 6, which is in contrast to Samson et al.3.

BT

A+=χ (6)

Moreover, the Flory-Huggins parameter may depend on the concentration, depending on the interaction between the solvent, amorphous and crystalline part of the polymer. The Flory-Huggins equation appears to be very useful when used as a correlative method, but cannot be used in a predictive way if reasonable accuracy is required. From various experiments, e.g. Favre et al.12, it has been concluded that the Flory-Huggins interaction parameter remains an empirical fitting parameter. The monomer concentration inside the polymer used for kinetic modeling has been estimated by the following equation:

Lmm CC ⋅φ= (7)

CHAPTER 2

19

Here LmC is the concentration of liquid monomer for a given temperature9.

Sorption of hydrogen It is well known that hydrogen not only influences the molecular weight of the polymer product but also affects polymerization kinetics. Knowledge of the sorption behavior of hydrogen as function of process conditions is therefore as important as monomer sorption behavior. However, for sparingly soluble penetrants sorption can be described by Henry’s law. For our system the situation is more complicated, because large amounts of propylene are sorbed into the amorphous part of the semi crystalline material. If Henry’s law is still valid for the sorption of hydrogen in a polymer matrix swollen with propylene, polymerization experiments in liquid phase should be executed with the same amount (mol%) of hydrogen in the gas cap as used for the gas phase experiments when the results are to be compared later on. Kinetic model The propagation rate is dependent on monomer concentration and the number of active sites. For propylene polymerizations with zirconocenes at low monomer concentration, several authors13,14 reported the reaction order with respect to monomer concentration to be higher than one, see equation 8. Fait et al.15 postulated a kinetic model based on the presence of a single-center, two-state catalyst system, in which both states differ in propagation rate. According to the proposed mechanism, interconversion between the two states will lead to a lower concentration of the slow state at higher monomer concentration leading to an overall order of reaction in the monomer larger than one.

nm

*pp CCkR = (1<n<2) (8)

In the simple power law equation 8 kp=kp,0exp(-Eact,p/RT), where kp is the propagation rate constant and Eact,p the activation energy for propagation. Further, C* is the number of active centers per gram of catalyst and Cm the monomer concentration based on the sorption equations presented before. The decay in polymerization rate generally observed after the built-up period has been discussed in literature intensively. It is generally believed16 that this loss of activity in time cannot be explained by intra-particle monomer diffusion limitation through the growing polymer layer. Most of the experimental results can be analyzed according to a decreasing number of active centers in time due to chemical deactivation. We have chosen to describe the deactivation by a simple first order relation:

CHAPTER 2

20

**

Ckdt

dCd−= (9)

Here kd=kd,0exp(-Eact,d/RT), where kd is the deactivation constant and Eact,d the activation energy for deactivation. Integration of equation 9 will lead to the number of active sites as a function of time.

tk*max

* deCC −= (10)

with *maxC the maximum number of active centers per gram of catalyst.

Substitution into equation 8 leads to:

tkmax,p

tk*maxmpp

dd eReCCkR −− == (11)

with Rp,max the maximum reaction rate. Note that equation 11 is valid after the built-up period. Results from sorption measurements Propylene sorption experiments have been done at different temperatures (41, 52, 62 and 73°C) and pressures (5 – 25 bar). In figure 5, the results of the sorption experiments are given, together with a best fit according to Henry’s law. The results are summarized in table 3. These fits are only based on the sorption measurements at low volume fractions of.propylene. The data at low volume fractions could be fitted with Henry’s law, but fails for higher values.

Figure 5: Fitting the sorption measurements with Henry’s law.

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0 5 10 15 20 25 30

P [bar]

φφ [-

]

41°C

52°C

62°C

73°C

CHAPTER 2

21

Table 3: Henry constants at different temperatures used to fit the experimental data at low volume fractions.

Temperature [°C]

kH [mol/l⋅bar]

Temperature [°C]

kH [mol/l⋅bar]

41

0.141

62

0.095

52 0.115 73 0.075

For every data point, the Flory-Huggins interaction parameter χ has been determined. The error in the determined interaction parameter is about 5% for the measurements at 41°C and increases to about 9% for the measurements at 73°C. In figure 6, each temperature series have been fitted with an average value for the interaction parameter. In figure 7, the experimental data are fitted using interaction parameters given by Samson et al.3. The experimentally determined values for the Flory-Huggins interaction parameter together with the literature values are summarized in table 4. The measured value for the Flory-Huggins interaction parameter decreases with increasing temperature, whereas the values given by Samson et al.3 are increasing dramatically with temperature.

Figure 6: Fitting measurements with the Flory-Huggins equation.

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0 5 10 15 20 25 30

P [bar]

φ [−

]φ

[−]

41°C

52°C

62°C

73°C

CHAPTER 2

22

Figure 7: Comparison of sorption measurements with literature data.

Figure 8: Measured Flory-Huggins parameters vs. the reciprocal temperature.

0.5

0.55

0.6

0.65

0.7

0.75

0.8

0.85

0.00285 0.0029 0.00295 0.003 0.00305 0.0031 0.00315 0.0032

1/T [K-1]

χχ [-

]

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0 5 10 15 20 25 30

P [bar]

φφ [-

]

41°C

52°C

62°C

73°C

Literature

41°C 52°C

62°C

73°C

CHAPTER 2

23

Table 4: Flory-Huggins interaction parameter as function of temperature; literature vs. experimental data.

Temperature [°C]

Interaction parameter given by Samson et al.3

Average measured value

41

0.86

0.82

52 0.99 0.75 62 1.16 0.69 73 1.45 0.61

Based on equation 6, a linear relation between the measured interaction parameter and the reciprocal temperature is expected, see figure 8. Equation 12 represents the best fit.

501.1T

43.730−=χ (12)

All experiments have been executed using the same polymer made in liquid propylene at 60°C with a heterogeneous metallocene catalyst. Care should be taken when using the absolute values, since the Flory-Huggins interaction parameter may e.g. depend on the degree of crystallinity of the polymer. Results of liquid phase experiments Influence of Al/Zr ratio The sensitivity towards the amount of TIBA have been tested during three different polymerization experiments at different Al/Zr ratios, see figure 9. All experiments have been carried out at 60°C and 2% hydrogen in the gas cap. The polymerization rate has been calculated as the amount of polypropylene per gram of metallocene on the silica support per hour, kg/gmet.hr. This is to compare the results later on with catalysts with different metallocene loading. The large initial fluctuations in polymerization activity do not depict real fluctuating reaction rates, they are caused by the PID controller. Although the temperature can be controlled within 0.2°C, the cooling water temperature does show some small oscillation in the beginning of an experiment, which directly influences the apparent polymerization rate. The experiments at different Al/Zr ratios show only minor deviations with each other. This indicates that in this range, a variation in the ratio does not influence the polymerization rate too much. Influence of temperature The influence of temperature on the reaction rate has been studied by varying the temperature from 40 to 70°C, see figure 10. Table 5 gives an overall summary of the experiments, presenting the reaction conditions and some fitted kinetic parameters. Note

CHAPTER 2

24

the large differences between the values for the monomer concentration based on theoretical literature sorption data and the values based on the measured sorption data, especially at higher temperatures. At 70°C, the measured monomer concentration is about 4 times higher than the concentration calculated by Samson et al.3.

0

200

400

600

800

1000

0 10 20 30 40 50 60 70

Time [min]

Rp

[kg

PP

/gr m

et h

r]

Al/Zr = 2091

Al/Zr = 700

Al/Zr = 1070

0

200

400

600

800

1000

0 10 20 30 40 50 60 70

Time [min]

Rp

[kg

PP

/gr m

et h

r]

70 °C 60 °C 50 °C 40 °C

Figure 9: Influence of Al/Zr ratio on the polymerization rate during liquid phase polymerization.

Figure 10: Influence of temperature on the polymerization rate during liquid phase polymerizations.

CHAPTER 2

25

Table 5: Summary of the liquid phase experiments at different temperatures.

T [°C]

H2 gas cap

[mol%]

H2 liquid

[mol%]

Cm [kg/m3]

This work

Cm [kg/m3]

Samson et al.3

Rp,max

[kg PP/gmet.hr]

kd

[min-1]

40

2.45

0.028

208.8

200.3

125.8

0.0088

50 2.38 0.043 233.0 156.5 232.5 0.0096 60 2.22 0.064 257.9 111.0 377.5 0.0179 70 2.05 0.086 276.1 70.4 578.0 0.0266

The experiment at 70°C showed the presence of a critical polymer concentration after about 20 minutes, i.e. a propylene conversion above 40 to 50% causing a sharp decrease of the heat transfer coefficient manifest from the temperature recordings, see also Samson et al.3. The observed transition in heat transfer is ascribed to a changing flow behavior of the reactor content, see Samson et al.3. Influence of hydrogen The influence of hydrogen on the polymerization kinetics has been investigated at 60°C in liquid phase. Experiments have been executed between 0 and 2.2% hydrogen in the gas cap. This relatively small range has been chosen because the sensitivity of metallocene catalysts towards hydrogen, related to the molecular weight of the polymer, is large compared to classical Ziegler-Natta catalysts17. The maximum reaction rate Rp,max and deactivation constant kd have been determined by fitting the reaction rate curves to equation 11. In figure 11 and 12, the maximum reaction rate and deactivation constant are given as a function of the hydrogen concentration. A linear relation is obtained between the maximum reaction rate and the hydrogen concentration in the gas cap. No saturation effect, as shown by Samson et al.18, was found for the range in which the experiments have been executed. However, the increasing trend for the deactivation constant with the hydrogen concentration, observed at higher hydrogen concentrations, was also found by Samson et al.18. In general, a higher deactivation rate is observed at higher reaction rates, either caused by higher temperatures or hydrogen. No deactivation phenomena were found for the experiment without hydrogen associated with a low reaction rate. Several possible explanations for the increased catalyst activity in the presence of hydrogen are given in literature18. One possibility may be the shortcutting of slow propagation steps, which occur after isolated secondary insertions (2-1 insertions).

CHAPTER 2

26

Figure 11: Influence of hydrogen on the maximum polymerization rate.

0

50

100

150

200

250

300

350

400

0 0.5 1 1.5 2 2.5

H2 [%]

Rp,

max

[kg

PP

/gm

et h

r]

Figure 12: Influence of hydrogen on the deactivation constant kd.

0

0.004

0.008

0.012

0.016

0.02

0 0.5 1 1.5 2 2.5

H2 [%]

k d [

min

-1]

CHAPTER 2

27

Results of gas phase experiments Influence of temperature The influence of temperature on the polymerization rate has been studied at temperatures between 40 and 70°C at 10 bar pressure and 2 mol% of hydrogen. As can be seen in figure 13, the reaction rate as well as the deactivation rate increases with increasing temperature. In contrast to the experiments in liquid phase, the catalyst does not show activity directly after injection. Especially at lower temperatures some time is required to reach maximum polymerization activity. The model presented earlier does not take any induction period or activation processes into account.

Influence of pressure The influence of pressure has been studied between 5 and 25 bar at 70°C. Figure 14 shows that the maximum reaction rate increases more or less linear with the monomer concentration in the amorphous part of the polymer, reflecting the first order reaction kinetics. Cm has been calculated according to the experimental sorption data presented earlier in this paper. Fitting the data at low concentration, below 15 bar, a trend line with an intercept not equal to zero is obtained, line b in figure 14. This supports the theory of a reaction order of higher than one discussed before. Another possible explanation may be a less accurate estimate of the monomer concentration at low pressures.

0

10

20

30

40

50

0 10 20 30 40 50 60 70

Time [min]

Rp

[kg

PP

/gm

et h

r]

70 °C

60 °C

50 °C

40 °C

Figure 13: Influence of temperature on the polymerization rate during gas phase polymerization at 10 bar pressure.

CHAPTER 2

28

In figure 15 the deactivation rate constant kd is given as function of the monomer concentration. Again a linear relation is obtained. According to the model the deactivation rate should not depend on the monomer concentration, however this dependency has been found before. Kohara et al.19 also obtained a linear relation between the deactivation constant and the monomer concentration using a Ziegler-Natta catalyst.

Figure 14: Influence of monomer concentration on maximum polymerization rate during gas phase polymerization rate at 70°C. a: Fit of all data, b: Fit at low concentrations; below 15 bar.

0

50

100

150

200

250

300

350

0 20 40 60 80 100 120 140

Cm [kg/m3]

Rp,

max

[kg

PP

/gr m

ethr

]

a

b

Figure 15: Influence of monomer concentration on the deactivation constant during gas phase polymerization at 70°C.

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0 20 40 60 80 100 120 140

Cm [kg/m3]

k d [

min

-1]

CHAPTER 2

29

Comparison between liquid and gas phase polymerization kinetics The main difference between the gas and liquid phase experiments seems to be the monomer concentration near the active center. One would expect that the same reaction mechanism holds for both phases and the same relative polymerization rate (Rp/Cm) should be obtained. In figure 16 and 17, the Arrhenius plots are given of the gas and liquid phase experiments based on the measured and (theoretical) literature3 sorption data, respectively. The constants for the kinetic model, based on both sorption data sets, are summarized in table 6.

Table 6: The constants for the kinetic model for the gas and liquid phase polymerizations.

Liquid phase Gas phase Sorption data set This work Literature3

sorption data

This work Literature3 sorption data

*max0p Ck [m3/hr.gmet] 8.67· 105 3.17· 1012 4.69· 109 1.94· 1011

Eact,p [kJ/mol] 36.8 76.2 48.0 69.7 kd0 [min-1] 5.09· 103 7.95· 103

Eact,d [kJ/mol] 35 31

Figure 16: Arrhenius plots (propagation) of gas and liquid phase experiments using the measured sorption data.

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

2.9 2.95 3 3.05 3.1 3.15 3.2 3.25

1000/T [K-1]

ln(R

p,m

ax/C

m)

Liquid phase

Gas phase

CHAPTER 2

30

-1

-0.5

0

0.5

1

1.5

2

2.5

2.9 2.95 3 3.05 3.1 3.15 3.2 3.25

1000/T [K-1]

ln(R

p,m

ax/C

m)

Liquid phase

Gas phase

Figure 17: Arrhenius plot (propagation) of the gas and liquid phase experiments using the sorption data given by Samson et al.3.

Figure 18: Arrhenius plot (deactivation) for the gas and liquid phase experiments.

-5.5

-5

-4.5

-4

-3.5

-3

-2.5

-2

-1.5

2.9 2.95 3 3.05 3.1 3.15 3.2 3.25

1000/T [K-1]

ln k

d

Liquid phase

Gas phase

CHAPTER 2

31

In general lower relative reaction rates, Rp,max/Cm, have been found in gas phase, which is in agreement with the results of Samson et al.2. Our simplified kinetic model does not take activation of the catalyst into account, although a clear initial increase in polymerization activity has been observed during all gas phase experiments. The time required to reach the maximum polymerization activity is for a gas phase experiment about 3-5 minutes. In this time, deactivation processes are already influencing the polymerization rate, causing a lower maximum polymerization rate. This may explain the lower relative activity observed for the gas phase experiments. An alternative method could be to use the initial reaction rate instead of the maximum polymerization rate by extrapolation to time zero. However, such extrapolations would lead to unrealistic values of the initial reaction rates. This effect is enhanced by the higher deactivation rate observed for the gas phase experiments, see figure 18. Samson et al.2 also found higher deactivation rates during gas phase polymerizations. An arrhenius plot for the gas and liquid phase experiments based on the initial reaction rate is given in figure 19.

-1

-0.5

0

0.5

1

1.5

2.9 2.95 3 3.05 3.1 3.15 3.2 3.25

1000/T [K-1]

ln(R

p0/C

m)

Liquid phase

Gas phase

Figure 19: Arrhenius plot (propagation) of the gas and liquid phase experiments using the measured sorption data. Rp0 is determined by extrapolation to time zero.

CHAPTER 2

32

The activation energies found for the liquid and gas phase experiments are in the same order of magnitude, see table 6. However, the sorption data used, literature vs. experimental, have a large effect on the determined kinetic parameters. The temperature series in liquid phase was unfortunately not executed at completely constant hydrogen concentration in the gas cap. The concentration dropped from 2.45% at the lowest temperature to 2.05% at the highest temperature, see table 5. The experiments in liquid phase, in which the hydrogen concentration in the gas cap was varied, show that this variation in hydrogen concentration will affect the maximum reaction rate. So, the determined activation energy for propagation for the liquid phase experiments is in fact too low, about 10%. This may explain the slightly lower activation energy (based on experimental sorption data) found for the liquid phase experiments. However, the sorption behavior of hydrogen in a polymer may depend on the amount of propylene sorbed in the polymer. Such information may help to explain the differences between the gas and liquid phase kinetic data. Conclusions Sorption data of propylene in polypropylene have been measured using a gravimetric method. At low pressures, sorption could be described with Henry’s law, at higher concentrations the Flory-Huggins equation appeared to be a useful fitting equation. Decreasing values of the Flory-Huggins parameter were measured with increasing temperature, which is in contrast to data given by Samson et al.3 but in line with theoretical expectations. Polymerizations have been executed in both gaseous and liquid propylene at different temperatures, pressures and hydrogen concentrations. Hydrogen appeared to have a large influence on the reaction and deactivation rate. In general, a higher deactivation rate is observed at higher reaction rates, either caused by higher temperatures or hydrogen. The influence of the monomer concentration was studied at 70°C in gas phase. At low concentrations, below 15 bar, a reaction order above one is obtained. This supports the theory of a single-center, two-state catalyst postulated by Fait et al.15. The sorption data of propylene have been used to compare the results of the gas and liquid phase polymerizations. Lower relative reaction rates have been found for gas phase polymerizations. The activation energies found for liquid and gas phase experiments are comparable. The sorption data used, either data from literature or experimental data, have a large effect on the estimated kinetic parameters. For a correct comparison between gas and liquid phase kinetic data, not only adequate monomer sorption data are required, but also hydrogen sorption data. Unfortunately, the

CHAPTER 2

33

sorption behavior of hydrogen in polypropylene swollen with propylene is unknown and difficult to determine exactly. Acknowledgement - This work has been funded by BRITE-EURAM Project CATAPOL (BE 96-3022). We

greatly acknowledge K. van Bree and F. ter Borg for the construction of the experimental set ups and

technical assistance. G.H. Banis is acknowledged for his technical support and W.R. Smit, B.G.C.J. Wijers

and S.N. Kuper for their assistance in the experimental part.

Notation C* Number of active sites mol/gmet

cm Monomer concentration mol/l Cm Monomer concentration kg/m3

2HC Hydrogen concentration mol/l Eact Activation energy J/mol kd Deactivation rate constant min-1

kH Henry constant mol/l.bar kp Propagation rate constant m3/mol.hr m Mass kg Mm Molecular weight kg/mol P Pressure bar P0 Saturation pressure bar R Gas constant J/mol.K Rp Reaction rate kg/gmet.hr t Time min T Temperature K V Volume m3 X Crystallinity - Greek χ Flory-Huggins interaction parameter - δ Solubility parameter (J/cm3)0.5

φ Volume fraction - ν Molar volume cm3/mol ρ Density kg/m3

Sub- and superscripts G Gas phase H2 Hydrogen L Liquid phase

CHAPTER 2

34

m Monomer Max Maximum condition n Reaction order p Polymer p Polymerization, propagation PP Polypropylene 0 Initial condition Abbreviations Al Aluminium MAO Methylaluminoxane TIBA Tri-isobuthylaluminium Zr Zirconium Literature 1. Hutchinson, R.A., Ray, W.H., J. Appl. Polym. Sci., 41, 51-81, (1990) 2. Samson, J.J.C., van Middelkoop, B., Weickert, G., Westerterp, K.R., AIChE J.,

45(7), 1548-1558, (1999) 3. Samson, J.J.C., Weickert, G., Heerze, A.E., Westerterp, K.R., AIChE J., 44(6), 1424-

1437, (1998) 4. Shimizu, F., Pater, J.T.M., Weickert, G., submitted to J. Appl. Pol. Sci., (2000) 5. Spaleck, W., Aulbach, M., Bachmann, B., Küber, F., Winter, A., Macromol. Symp.,

89, 237-247, (1995) 6. Carvill, A., Tritto, I., Locatelli, P., Sacchi, M.C., Macromolecules, 30, 7056-7062,

(1995) 7. Bonini, F., Fraaije, V., Fink, G. J., Polym. Sci. Part A: Polym. Chem., 33, 2393-

2402, (1995) 8. Mizan, T.I., Li, J., Morsi, B.I., Chang, M.Y., Maier, E., Singh, C.P.P., Chem. Eng.

Sci., 49(6), 821-830, (1994) 9. VDI-Wärmeatlas; VDI-Gesellschaft Verfahrenstechnik und Chemieingenieurwesen

(GVC), 6th edition, (1991) 10. Barton, A.F.M., CRC Handbook of polymer-liquid interaction parameters and

solubility parameters, CRC Press, Boca Raton, 284-292, (1990) 11. Prausnitz, J.M., Lichtenthaler, R.N., de Azevedo, E.G., Molecular thermodynamics

of fluid-phase equilibria, Englewood Cliffs, N.J: Prentice-Hall, 2nd edition, (1986) 12. Favre, E., Nguyen, Q.T., Schaetzel, P., Clément, R., Néel, J. J., Chem. Soc. Faraday

Trans., 89(24), 4339-4346, (1993) 13. Fink, G., Herfert, N., Montag, P., Ziegler Catalysts, Springer-Verlag, 159-179,

(1995)

CHAPTER 2

35

14. Jüngling, S., Mülhaupt, R., Stehling, U., Brintzinger, H.H., Fischer, D., Langhauser, F. J., Polym. Sci., Part A: Polym. Chem., 33, 1305-1318, (1995)

15. Fait, A., Resconi, L., Guerra, G., Corradini, P., Macromolecules, 32, 2104-2109, (1999)

16. Kim, I., Woo, S.I., Korean J. of Chem. Eng., 7(2), 95-99, (1990) 17. Blom, R., Dahl, I.M., Macromol. Chem. Phys., 200, 442-449, (1999) 18. Samson, J.J.C., Bosman, P.J., Weickert, G., Westerterp, K.R., J. Pol. Sci., Part A:

Polym. Chem., 37(2), 219-232, (1999) 19. Kohara, T., Shinoyama, M., Doi, Y., Keii, T., Makromol. Chem., 180, 2139-2151,

(1979)

CHAPTER 2

36

Appendix Calculation procedure for sorption measurements The weight increment from a sorption experiment has been calculated from the measured mass before and after the addition of propylene gas.

sorbedG12 mmmmm +=−=∆ (13)

The volume of the vessel is used to calculate the volume of the gas not sorbed in the amorphous part of the polymer, i.e. gas between polymer particles and gas in the pores of the polymer.

p

pvesselpvesselG

mVVVV

ρ−=−= (14)

Multiplying this volume by the gas density results in the mass of gas, see equation 15.

GGG Vm ρ= (15)

Using the density of liquid propylene the volume of sorbed propylene can be calculated.

Lm

Gsorbed C

mmV

−∆= (16)

The volume fraction of the monomer in the amorphous parts of the polymer, φm, can be calculated with:

( ) 0,psorbed

sorbed

0,amorph,psorbed

sorbed

VX1V

V

VV

V

−+=

+=φ (17)

Swelling of the polymer results in an increase of the polymer volume:

sorbed0,pp VVV += (18)

The calculated polymer volume Vp in turn changes the volume of the gas not sorbed into the polymer, which influences the volume fraction of monomer sorbed into the polymer, etc. This step has to be repeated until the volume fraction φ has converged.

CHAPTER 3

37

Chapter 3 Gas phase polymerization of propylene with a heterogeneous metallocene catalyst. The influence of temperature, pressure and hydrogen on the reaction kinetics and molecular weight distribution. Abstract Gas phase polymerizations have been executed at different temperatures, pressures and hydrogen concentrations using rac-Me2Si[Ind]2ZrCl2 / MAO / SiO2(PQ) as catalyst. The reaction rate curves have been described with a kinetic model, which takes the initially increasing polymerization rate into account. The monomer concentration in the polymer has been calculated with the Flory-Huggins equation. The kinetic parameters have been determined by fitting the reaction rate curves with the model. At high temperatures, pressures and hydrogen concentrations, a runaway on particle scale may occur leading to reduced polymer yields. The molecular weight and molecular weight distribution of the polymer samples could be described with a “two-site model”. At constant temperature, the chain transfer probability of site 1 and 2 depends only on the hydrogen concentration divided by the monomer concentration.

CHAPTER 3

38

CHAPTER 3

39

Introduction Despite the rapid developments in the field of metallocene catalysts, only a few papers provide kinetic data based on experiments in gaseous monomer, although most commercial polypropylene processes are carried out in gas phase. These processes are operated at elevated pressure, e.g. 25 bar, using heterogeneous catalysts. Resconi et al.1 showed that both molecular weight and degree of isotacticity are strongly influenced by the monomer concentration. In this paper, the gas phase polymerization of propylene with a heterogeneous metallocene catalyst is described together with the related molecular weight and molecular weight distribution of the product. The catalyst system used in this study is rac-Me2Si[Ind]2ZrCl2 with methylaluminoxane (MAO) as cocatalyst. Both components are supported on Pennsylvania Quarts (PQ) silica. Experimental The experimental set up consists of a 0.5 liter jacketed Büchi reactor, a helical stirrer, injection system for dry catalyst powder and an evaporation vessel for monomer supply. The set up has been described in detail in a previous paper2. Polymerization reactions have been executed at constant temperature, pressure and hydrogen concentration. The reaction rate is calculated from the feed rate required to keep the pressure constant, and measured by a mass flow controller. The catalyst used, is the well-known metallocene rac-Me2Si[Ind]2ZrCl2. The catalyst, kindly supplied by Witco Co. Bergkamen (Germany), is supported on PQ silica with a concentration of 0.79 wt%. The MAO/SiO2 – support used for immobilization of the metallocene contained 19.1 wt% of alumina. Tri-isobuthylaluminium (TIBA) has been used to increase the polymerization activity. The amount of TIBA and the precontact time has been kept constant for all experiments. Some characteristics of the silica carrier (PQ MS3040) are summarized in table 1. Note the quite large average particle diameter (108µm). The exact polymerization procedure for the gas phase polymerizations has been described in a previous paper2. Table 1: Pore size and particle size distribution of the PQ silica carrier.

Minimum

[µm]

Average

[µm]

Maximum

[µm]

Pore diameter

0.004

0.2

5

Particle size 27 108 170

The polypropylene samples were characterized on their molecular weight and molecular weight distribution by Gel Permeation Chromatography (GPC). These GPC

CHAPTER 3

40

measurements were carried out at DSM Research (The Netherlands) on a Waters M150C apparatus with TSK columns at 140°C using 1,2,4-trichlorobenzene as solvent. Kinetic model The polymerization rate in gas phase obtained with the type of catalyst used in this work, clearly shows an acceleration and deactivation stage in time2,3. The nature of both the acceleration and deactivation stage has been discussed by many authors before. Kim et al.4 explained the acceleration stage for a highly active Ziegler-Natta catalyst with the kinetics of absorption of the cocatalyst. Bonini et al.3 excluded chemical activation processes, because their metallocene catalyst was preactivated by MAO during the supporting procedure. They explained the acceleration by diffusion limitations of the monomer inside the catalyst particle. Their postulated particle growth model accounts for the shell by shell fragmentation from outside and a final multigrain structure of the particle.

To formulate a model to describe the complex physical and chemical effects, we have chosen to describe the reaction rate by empirical relations. Scheme 1 gives a schematic representation of the activation, propagation and deactivation processes. The non-activated catalyst, C, is activated by a first order relation to an activated catalyst system C*, see equation 1.

Ckdt

dCi−= (1)

Here ki=ki,0exp(-Eact,i/RT), where ki is the activation rate constant and Eact,i the activation energy for activation. Deactivation is described by a decreasing number of active sites, again using a first order relation, see equation 2.

*di

*

CkCkdt

dC−= (2)

C* stands for the overall number of active sites including the C*-Pj species from scheme 1. Furthermore, it is assumed that all active sites deactivate with the same rate, independently whether a chain is attached to it or not.

C C* C*-Pj

ki kp

kd kd

Deactivated catalyst

Scheme 1: Activation, propagation and deactivation of the catalyst.

CHAPTER 3

41

In equation 2, kd=kd,0exp(-Eact,p/RT) where kd is the deactivation rate constant and Eact,d the activation energy for deactivation. In a previous paper2 it was observed that at relatively high monomer concentrations, above 15 bar, the polymerization reaction is first order with respect to the number of active sites and the monomer concentration in the amorphous part of the polymer.

mpp CCkR *= (3)

Here kp=kp,0exp(-Eact,p/RT), where kp is the propagation rate constant and Eact,p the activation energy for propagation. The monomer concentration in the polymer has been calculated with the Flory-Huggins equation. The Flory-Huggins interaction parameter is based on the experimental data published before2. Integration and substitution, under isothermal conditions, of equations 1 and 2 leads to:

( )tktk

id

i0

* di eekk

kCC −− −

−= (4)

Here C0 is the initial amount of non-activated catalyst. Substitution into equation 3 leads to:

( )tktk

id

i0mpp

di eekk

kCCkR −− −

−= (5)

Molecular weight and molecular weight distribution The molecular weight and molecular weight distribution are important parameters for the physical, mechanical and rheological behavior of polymers. As a result of the uniform catalytically active sites of a metallocene catalyst, it is possible to control the molecular weight and molecular weight distribution. At constant temperature, pressure and gas composition, the instantaneous molecular weight distribution of a polymer synthesized with a single-site catalyst can be described with a Schulz-Flory distribution, see equation 6.

jq2dj ejqy −= (6)

Here, yjd is the density function of the instantaneously formed molecular weight

distribution. This equation depends only on the chain termination probability q. This parameter is dependent on the type of catalyst and the process conditions, like temperature, pressure and the hydrogen concentration. Different types of chain transfer reactions have been observed, e.g. β-hydrogen transfer to the metal and monomer, chain transfer to aluminium, chain transfer with hydrogen and chain transfer with the monomer.

CHAPTER 3

42

Chain transfer with hydrogen is in general the most effective one, and is often used to control the molecular weight of polypropylenes under industrial conditions. The molecular weight of a polymer sample depends on the ratio between the overall propagation rate and the total chain transfer rates and is in general independent of the polymerization activity. Combining chain transfer and chain growth reactions, one can derive:

xHHmmmp

xHHmm

kCkCkCk

kCkCkq

22

22

+++++

= (7)

Here, km, 2Hk and kx are the rate constants for chain transfer to monomer, hydrogen and

all other chain transfer mechanisms, respectively. During polymerization, high molecular weight polymers are formed; therefore it can be assumed that:

xHHmmmp kCkCkCk22

++>> (8)

Furthermore, we expect chain transfer reactions with hydrogen and monomer to be much more important than other chain transfer mechanisms. This leads to equation 9, which corresponds to the well-known Mayo equation.

m

H

p

H

p

m

C

C

k

k

kk

q 22+= (9)

Note that 2HC , Cm and the temperature are constant during the experiment, which means

that the instantaneous molecular weight distribution corresponds to the integral distribution obtained after the experiment. In this case we neglect the possible change in concentration of other components, e.g. MAO or TIBA. It may be possible that these components are diluted because of the production of polymer. Data from the GPC analyses can help to calculate the chain transfer probability, see equation 10.

w

m

n

m

MM

2MM

q == (10)

Here, Mm is the molecular weight of the monomer, Mn is the number average molecular weight and Mw is the weight average molecular weight of the polymers, see Weickert5. The equations presented above are in general only valid for a single site catalyst, where Mw/Mn = 2, in case of dominating transfer reactions for chain growth termination.

CHAPTER 3

43

Results The influence of temperature, pressure and hydrogen on the polymerization kinetics have been studied by using the kinetic model described earlier in this paper. An example of an experiment together with a fit of the model is shown in figure 1. Fitting is done by minimizing the deviations between the experimental and model curve. In this figure, also a fit with the classical model without activation is given. It is obvious that the model without activation is not able to describe the reaction rate during the first couple of minutes of the experiment. After about 10 minutes the two different models, both using a first order deactivation, coincide. Note that an induction period, in general about 1-2 minutes, is not taken into account. Therefore, t = 0 is defined when Rp increases from zero.

For each experiment, the measured reaction rates have been fitted to the model with activation. The values obtained for the activation constant ki, the deactivation constant kd and the kinetic parameter kpC0 have been used to analyze the various process parameters in a qualitative and quantitative manner. Table 2, 3 and 4 give an overall summary of the experiments, presenting the process conditions and determined kinetic parameters.

0

20

40

60

80

100

120

140

160

0 10 20 30 40 50 60

Time [min]

Rp

[kg

PP

/ g m

et. h

r] Model with activation

Model without activation

Figure 1: Fitting an experiment with the kinetic model with and without activation.

CHAPTER 3

44

Table 2: Summary of the experiments; temperature and pressure series.

T [°C]

p [bar]

Cm [kg/m3]

2HC [kg/m3]

kpC0 [m3/gmet.hr]

kd [min-1]

ki [min-1]

Rp,max [kg/gmet.hr]

50 15 92.4 0.011 1.031 0.033 0.191 65.9

60 15 67.1 0.011 1.481 0.038 0.320 74.6

60 15 67.1 0.011 1.393 0.040 0.349 70.6

70 15 49.8 0.011 2.563 0.048 0.440 98.8

70 15 49.8 0.011 2.630 0.049 0.424 100.5

75 15 44.0 0.011 2.348 0.046 0.290 73.1

75 15 44.0 0.011 2.201 0.044 0.315 70.4

80 15 38.2 0.011 2.149 0.050 0.311 57.9

70 5.1 14.2 0.011 1.211 0.072 0.256 9.3

70 10.1 30.7 0.011 2.374 0.070 0.187 35.5

70 14.9 49.7 0.011 2.541 0.077 0.290 71.3

70 17.9 64.9 0.011 2.766 0.079 0.300 100.8

70 19.5 73.8 0.011 3.022 0.087 0.349 128.5

70 22.6 95.5 0.011 2.509 0.099 0.719 164.6

70 24.6 113.4 0.011 1.801 0.144 1.147 145.8

CHAPTER 3

45

Table 3: Summary of the experiments; hydrogen series at 70°C and 15 bar.

Cm [kg/m3]

2HC [kg/m3]

kpC0 [m3/gmet.hr]

kd [min-1]

ki [min-1]

Rp,max [kg/gmet.hr]

50.7 0 0.520 0.040 0.146 16.2

50.7 0 0.600 0.028 0.210 22.3

50.5 0.0034 1.204 0.056 0.362 43.3

50.4 0.0037 1.131 0.060 0.348 48.2

50.3 0.0053 1.384 0.053 0.305 46.9

50.3 0.0054 1.371 0.067 0.355 65.1

50.0 0.0109 1.760 0.061 0.477 60.4

50.0 0.0107 1.752 0.061 0.345 69.7

49.9 0.0116 1.908 0.067 0.495 88.0

49.3 0.0214 2.276 0.069 0.744 103.3

49.2 0.0322 2.551 0.080 1.218 87.7

49.0 0.0324 2.136 0.069 1.167 104.4

48.6 0.0443 2.534 0.064 1.200 120.1

46.2 0.0684 3.175 0.074 1.544 126.1

CHAPTER 3

46

Table 4: Summary of the experiments; temperature series without hydrogen at 15 bar.

T [°C]

Cm [kg/m3]

kpC0 [m3/gmet.hr]

kd [min-1]

ki [min-1]

50 91.8 0.165 0.050 0.428

50 91.8 0.152 0.042 0.405

60 66.9 0.311 0.035 0.440

60 66.9 0.366 0.043 0.588

70 50.6 0.600 0.028 0.210

70 50.6 0.520 0.040 0.146

80 38.2 1.305 0.069 0.288

80 38.2 1.077 0.065 0.325

Influence of temperature The influence of temperature has been investigated at 15 bar at temperatures between 50 and 80°C. The hydrogen concentration has been kept constant at 0.011 kg/m3, which is about 1 vol%. In figure 2 the Arrhenius plot for propagation is given, showing a linear behavior for temperatures between 50 and 70°C. The experiments at 75 and 80°C showed a considerable lower activity. In figure 3 and 4 the Arrhenius plots for deactivation and activation are given respectively, showing again a more or less linear relation between 50 and 70°C, and a deviating behavior above 70°C. The determined kinetic parameters are summarized in table 5. Table 5: Summary of the determined kinetic parameters.

Model parameters

Eact,p

43.9

kJ/mol

Eact,d 17.4 kJ/mol Eact,I 35.6 kJ/mol kp0C0 1.20⋅107 m3/gmet.hr kd0 21.3 min-1

ki0 1.19⋅105 min-1 Validity range 50 – 70 °C

CHAPTER 3

47

-0.2

0

0.2

0.4

0.6

0.8

1

1.2

2.8 2.9 2.9 3.0 3.0 3.1 3.1 3.2

1000/T [K-1

]

ln(k

pC0)

70 50607580

T [°C]

7 min prepolymerization



Standard experiment

Figure 2: Arrhenius plot for propagation. The open dots represent the prepolymerization experiments.

-3.5

-3.4

-3.3

-3.2

-3.1

-3

-2.9

2.8 2.9 2.9 3.0 3.0 3.1 3.1 3.2

1000/T [K-1

]

ln(k

d)

Figure 3: Arrhenius plot for deactivation.

CHAPTER 3

48

The activation energy for propagation, 43.9 kJ/mol, is quite close to the value found previously2, 48.0 kJ/mol, where the same metallocene on a different support was used for gas phase polymerizations at 10 bar pressure. Note that in our previous work the activation energy was determined based on a model without activation.

The possibility of particle overheating at the start of the reaction, where the highly concentrated catalyst particles still have a small external surface, has been the subject of many discussions in literature6-8. The elevated particle temperature may lead to chemical reactions between the various catalyst components, especially the stability of MAO at higher temperatures is questionable. The particle temperature is dependent on the heat transfer coefficient, the heat transfer area and the difference between bulk and particle temperature. The internal temperature gradients are in general small. Samson et al.9 found for both gas and liquid phase polymerization with a Ziegler-Natta catalyst, rapid catalyst deactivation above a certain temperature. The absolute polymerization rate, per gram of catalyst (including support), are in the same order of magnitude as observed in this work; 1-3 kg PP/(gcat.hr). Prepolymerizations were executed by Samson et al.9 to prevent a thermal runaway on particle scale. Prepolymerization may help to prevent a thermal runaway on particle scale. As the reaction rate is much lower in prepolymerization step compared to the main polymerization, particles have time to grow in a controlled way to a size where the external surface is large enough to remove all the reaction heat. The observed reaction rates by Samson were significantly higher than without prepolymerization. The prepolymerizations were carried out at low temperatures in either liquid propylene or pentane slurry. In order to test the same procedure for a

-1.8

-1.6

-1.4

-1.2

-1

-0.8

-0.6

2.8 2.9 2.9 3.0 3.0 3.1 3.1 3.2

1000/T [K-1]

ln(k

i)

Figure 4: Arrhenius plot for activation.

CHAPTER 3

49

heterogeneous metallocene catalyst, prepolymerizations have been executed in gaseous propylene at 40°C and 5 bar monomer pressure. The prepolymerization time has been varied between 3 and 7 minutes, after this the temperature and pressure have been raised to 80°C and 15 bar, respectively. As can be seen in figure 2, prepolymerization of the catalyst results in a higher reaction rate for the main polymerization reaction. However, a large deviation with the Arrhenius plot still exists, even for prepolymerization time of 7 minutes. Obviously, there are more effects than only initial particle overheating to be taken into account. In a next series, the influence of temperature in the absence of hydrogen has been investigated at 15 bar and temperatures between 50 and 80°C. In figure 5 the Arrhenius plot is given, showing a linear behavior in contrast to the temperature series with hydrogen. No discrepancy was found at temperatures above 70°C. Because of the low reaction rate in absence of hydrogen, the chance of a runaway is small, even at 80°C. The activation energy for propagation derived for this case is 58.9 kJ/mol. This value is higher compared to the value obtained in the presence of hydrogen (43.9 kJ/mol). A possible explanation for this may be that the catalyst applied is not completely single site, and certain sites have different sensitivity towards hydrogen. For further discussion on a possible non-single site behavior of the heterogeneous metallocene catalyst, see below.

-2

-1.5

-1

-0.5

0

0.5

2.8 2.85 2.9 2.95 3 3.05 3.1 3.15

1000/T [K-1]

ln(k

pC0)

Figure 5: Arrhenius plot for propagation in the absence of hydrogen.

CHAPTER 3

50

Influence of pressure The influence of pressure has been investigated in the range between 5 and 25 bar at 70°C. The hydrogen concentration has been kept constant at 0.011 kg/m3, which is the same value as used for the temperature series. In figure 6, the influence of monomer concentration on the kinetic parameter kpC0Cm is shown. Here, the monomer concentration Cm is based on the monomer concentration in the amorphous part of the polymer. For a first order reaction, a linear relation is expected with an intercept equal to zero. As can be seen from figure 6, a linear relation is obtained for pressures between 5.1 and 19.5 bar, while deviations occur at higher pressures (22.6 and 24.6 bar). Moreover, the intercept is not equal to zero indicating a higher reaction order at low monomer concentrations. This is in agreement with work published before2,10.

The deviations observed at high pressures, can be explained in a similar way as was done for the temperature series. At high polymerization rates, i.e. high monomer concentrations or bulk temperatures, the particle temperature may be substantial higher than the bulk temperature, which may lead to chemical deactivation reactions. The maximum polymerization activity observed for the case of particle excess temperature behavior, should be in the same order of magnitude for the pressure and temperature series as in those cases the actual particle temperature can be expected to be similar. The behavior described above is confirmed by the determined Rp,max values presented in table 2.

0

50

100

150

200

250

300

0 20 40 60 80 100 120Cm [kg/m3]

k pC

0Cm

[kg

PP

/gm

et.h

r]

Figure 6: Influence of monomer concentration on kpC0Cm.

CHAPTER 3

51

The deviations observed at high pressures, have not been found in the work published before2, where the same catalyst was used but on a different silica support. The average particle size of the support used in this work is about a factor of two (108 vs. 51 µm) larger than used in this previous work. Therefore, the volume to surface area ratio of the catalyst used in this work is about a factor of two higher, while the activity per gram of catalyst is about the same. This makes heat removal from the larger particle more difficult. In figure 7 and 8, the influence of the monomer concentration on the deactivation and activation constant is given. As can be seen, the two constants are hardly affected by the monomer concentration, except for the two experiments at 22.6 and 24.6 bar. The values for kd and ki for these two experiments are higher, compared to the values at low concentration. This is in contradiction to the behavior observed in the temperature series.

0.00

0.04

0.08

0.12

0.16

0 20 40 60 80 100 120

Cm [kg/m3]

k d [

min

-1]

Figure 7: Influence of monomer concentration on kd.

CHAPTER 3

52

Influence of hydrogen The influence of hydrogen has been investigated at 15 bar and 70°C between 0 and 0.068 kg/m3 (0 – 5.5 mol%). The hydrogen concentration is calculated as the concentration in the gas phase, because sorption data of hydrogen in a polymer swollen with monomer is not available. In figure 9 the influence of hydrogen on the kinetic constant kpC0 is given, showing a strong increase at low hydrogen concentration. At high hydrogen concentrations the slope decreases. Samson et al.11 also found a decreasing slope with increasing hydrogen concentration. Several possible explanations can be given for increasing polymerization rate at higher hydrogen concentrations12. Firstly, hydrogen may react with the surface of the heterogeneous catalyst to increase the number of active sites. Secondly, hydrogen may reactivate deactivated sites, which have been deactivated by components like metalallylics. Thirdly, hydrogen may prevent the forming of unsaturated chain-end groups, which may act as a poison. However, the most plausible explanation is that hydrogen prevents the formation of dormant sites, which occur after a secondary insertion11. The explanation for the decreasing slope at higher hydrogen concentration can also be explained by several theories. Firstly, at higher hydrogen concentration the concentration of dormant sites will decrease causing a reduced reactivation effect of hydrogen. Secondly, the temperature and pressure series presented above support the theory of a runaway on particle scale above a critical polymerization rate. In table 3 values for the

0

0.2

0.4

0.6

0.8

1

1.2

1.4

0 20 40 60 80 100 120

Cm [kg/m3]

k i [

min

-1]

Figure 8: Influence of monomer concentration on ki.

CHAPTER 3

53

maximum polymerization activity are presented. These values are in the same order of magnitude, in case of high hydrogen concentrations, compared to the values obtained in the temperature and pressure series. The given possible explanations do not exclude each other, they might occur simultaneously.

In figure 10 and 11 the influence of hydrogen on the deactivation and activation constant is given. The deactivation rate in the absence of hydrogen is substantially lower compared to higher hydrogen concentrations. The activation constant seems to be linearly dependent on the hydrogen concentration.

Figure 9: Influence of hydrogen on kpC0.

0

0.5

1

1.5

2

2.5

3

3.5

0.0000 0.0100 0.0200 0.0300 0.0400 0.0500 0.0600 0.0700 0.0800

[H2] [kg/m3]

k pC

0 [m

3 /gm

et.h

r]

CHAPTER 3

54

0.000

0.020

0.040

0.060

0.080

0.100

0.0000 0.0100 0.0200 0.0300 0.0400 0.0500 0.0600 0.0700 0.0800

[H2] [kg/m3]

k d [

min

-1]

Figure 10: Influence of hydrogen on kd.

0.00

0.40

0.80

1.20

1.60

2.00

0.0000 0.0100 0.0200 0.0300 0.0400 0.0500 0.0600 0.0700 0.0800

[H2] [kg/m3]

k i [

min

-1]

Figure 11: Influence of hydrogen on ki.

CHAPTER 3

55

Interpretation of the acceleration period As discussed before, several explanations for the initial acceleration period have been proposed in literature. The particle growth model proposed by Bonini et al.3, assuming a shell by shell fragmentation from outside, is based on a diffusion limitation process. According to this theory, the yield obtained at the point of maximum polymerization activity should be the same for every experiment, assuming little chemical deactivation processes in the acceleration period. Following our model, the yield at the moment of maximum polymerization activity can be determined from the surface below the rate-time curves, see equation 11.

( )

( ) ( )

−−−

−

=−−

=

−−

−−∫

tk

d

tk

iid

im0p

t

0

tktk

id

im0p

di

di

e1k1

e1k1

kkk

CCk

dteekk

kCCkYield

(11)

In figure 12 the calculated yield at the moment of maximum polymerization activity is given for the pressure series. As can be seen, the yield at maximum polymerization activity is certainly not the same for every experiment. Therefore, we interpret the increasing polymerization rate during the initial period as a physical and chemical process where several components, e.g. metallocene, MAO, TIBA, react to form the active species. The physical process is highly related to the particle morphology, which in turn is predefined in the early (fragmentation) stage of polymerization. Catalyst

Figure 12: Influence of the relative hydrogen concentration on the chain termination probability; samples from the hydrogen series.

0

1

2

3

4

5

6

7

8

0 20 40 60 80 100 120

Cm [kg/m3]

Yie

ld a

t R

p,m

ax [

kg P

P/g

met

]

CHAPTER 3

56

fragmentation depends mainly on two factors: hydraulic pressure generated by the polymer produced and rigidity of the support, see Weickert et al.13. A fast increase of the hydraulic pressure, i.e. high initial polymerization rates, leads to a large number of fragments, whereas low initial rates corresponds to a small number of fragments. We have chosen to describe the initial acceleration period with an empirical model because the mechanism seems to be very complex. Molecular weight and molecular weight distribution The results from the GPC measurements (Mw, Mn and Mw/Mn) are presented in table 6 for some of the samples produced during the kinetic studies at different process conditions. The low molecular weight of the polymer is in agreement with results from literature3,12,14. In general, a narrow molecular weight distribution is obtained during polymerization with homogeneous metallocene catalysts at constant temperature and gas composition (Mw/Mn ≈ 2). Higher polydispersities are observed after polymerization with the heterogeneous analogues3,15. The samples produced during the kinetic measurements described before show polydispersities of around 2.5. One can calculate the chain termination probability q in two different ways, either based on Mn or based on Mw resulting in the same value if Mw/Mn = 2, see equation 10. In figure 13, the chain termination probabilities based on Mn and Mw of the hydrogen series have been plotted against the relative hydrogen concentration (

2HC /Cm), theoretically giving a straight line (see equation 9). Because Mw/Mn > 2, the two lines in figure 13 do not coincide. Moreover, at high relative hydrogen concentrations both kinds of chain termination probabilities are smaller than expected. Several possible explanations can be given for this behavior. Firstly, the error made during the GPC analyses may increase with decreasing molecular weight. Secondly, a possible runaway on particle scale, as argued before, results in overheating of particles during the start of the reaction. However, a relative low amount of lower molecular weight polymer will be produced in this period. Degradation reactions of MAO or other cocatalyst compounds at higher temperatures may also lead to the forming of chain termination agents.

CHAPTER 3

57

Table 6: GPC data.

T [°C]

p [bar]

2HC /Cm [-]

Mn [kg/mol]

Mw [kg/mol]

Mw/Mn [-]

q(Mn) [-]

q(Mw) [-]

50

15

1.19⋅10-4

11

25

2.3

3.82⋅10-3

3.36⋅10-3

60 15 1.64⋅10-4 8.6 23 2.7 4.88⋅10-3 3.65⋅10-3

70 15 2.21⋅10-4 8.1 19 2.3 5.19⋅10-3 4.42⋅10-3

75 15 2.50⋅10-4 6.9 17 2.5 6.09⋅10-3 4.94⋅10-3

80 15 2.88⋅10-4 7 17 2.4 6.00⋅10-3 4.94⋅10-3

50 15 0 13 34 2.6 3.23⋅10-3 2.47⋅10-3

60 15 0 14 31 2.2 3.00⋅10-3 2.71⋅10-3

80 15 0 9.8 24 2.4 4.29⋅10-3 3.50⋅10-3

70 5.1 7.75⋅10-4 3.4 7.3 2.2 1.24⋅10-2 1.15⋅10-2

70 14.8 2.21⋅10-4 8.7 17 2.0 4.83⋅10-3 4.94⋅10-3

70 17.9 1.69⋅10-4 10 20 2.0 4.20⋅10-3 4.20⋅10-3

70 19.5 1.49⋅10-4 10 21 2.1 4.20⋅10-3 4.00⋅10-3

70 22.6 1.15⋅10-4 11 24 2.2 3.82⋅10-3 3.50⋅10-3

70 15 0 12 27 2.3 3.50⋅10-3 3.11⋅10-3

70 15 7.34⋅10-5 9.6 22 2.3 4.38⋅10-3 3.82⋅10-3

70 15 1.05⋅10-4 8.3 20 2.4 5.06⋅10-3 4.20⋅10-3

70 15 2.18⋅10-4 7.3 17 2.3 5.75⋅10-3 4.94⋅10-3

70 15 4.34⋅10-4 5.4 13 2.4 7.78⋅10-3 6.46⋅10-3

70 15 9.05⋅10-4 3.9 10 2.6 1.08⋅10-2 8.40⋅10-3

70 15 1.48⋅10-3 3.3 8.1 2.5 1.27⋅10-2 1.04⋅10-2

Variation of the reaction temperature and pressure will also lead to a variation of the relative hydrogen concentration, because of the changing monomer concentration in the polymer. In figure 14 the chain termination probability based on Mw of all kinetic series (temperature, pressure and hydrogen) are given as function of the relative hydrogen concentration. It can be seen that the dependency on temperature is not as strong as on the relative hydrogen concentration. This phenomenon is characteristic for the used

CHAPTER 3

58

catalyst and not generally valid. Phillips catalysts, CrO/SiO2, are for instance hardly affected by hydrogen but are strongly dependent on temperature15.

Figure 13: Influence of the relative hydrogen concentration on the chain termination probability; hydrogen series.

0

0.002

0.004

0.006

0.008

0.01

0.012

0.014

0.016

0.018

0.02

0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012 0.0014 0.0016 0.0018

[H2]/Cm

q [-

]

q based on Mn

q based on Mw

Figure 14: Influence of the relative hydrogen concentration on the chain termination probability; samples from all series.

0

0.002

0.004

0.006

0.008

0.01

0.012

0.014

0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012 0.0014 0.0016

[H2]/Cm

q(M

w)

[-]

Hydrogen series

Pressure series

Temperature series

Temperature series wihout hydrogen

CHAPTER 3

59

In figure 15 an example is given of the result from a GPC measurement together with two curves based on a Schulz-Flory distribution from q’s determined from Mn and Mw. The GPC curve is broader, which was expected because of the higher polydispersity. Fait et al.10 postulated a kinetic model for a single-center, two-state catalyst. The two states differ in either their monomer insertion or coordination rate, thus having two different propagation rate constants. Different chemical possibilities for the nature of the two stated were discussed. The ratio between the two states depends, among others, on the monomer concentration. We observed first order reaction kinetics at pressures above 15 bar, but nevertheless observed non-single site behavior. This may be related to the fact that our catalyst is supported allowing for interaction of active sites and the support.

0

0.2

0.4

0.6

0.8

1

1.2

1.4

2 2.5 3 3.5 4 4.5 5 5.5 6

Log Mw

Y

b

a

c

Figure 15: Simulation of the molecular weight distribution with Schulz-Flory distributions. a) Result from GPC analyses. b) Based on chain termination probability calculated with Mn. c) Based on chain termination probability calculated with Mw.

CHAPTER 3

60

Table 7: Chain transfer probabilities for site 1 and 2 assuming an equal productivity for both sites (50/50).

2HC /Cm

[-]

q1

[-]

q2

[-] ( )∑

=

−N

1i

2

ielmodGPC YYN

1

0

2.05⋅10-3

4.95⋅10-3

0.28⋅10-3

7.34⋅10-5 2.49⋅10-3 5.91⋅10-3 0.38⋅10-3

1.05⋅10-4 2.64⋅10-3 6.90⋅10-3 0.49⋅10-3

2.18⋅10-4 3.28⋅10-3 8.07⋅10-3 0.82⋅10-3

4.34⋅10-4 4.02⋅10-3 1.18⋅10-2 1.00⋅10-3

9.05⋅10-4 4.86⋅10-3 1.79⋅10-2 1.90⋅10-3

1.48⋅10-3 7.12⋅10-3 2.61⋅10-2 2.42⋅10-3

0

0.2

0.4

0.6

0.8

1

1.2

2 2.5 3 3.5 4 4.5 5 5.5 6

Log Mw

Y

GPC result

Two-site model

Figure 16: Simulation of the molecular weight distribution with the two-site model.

CHAPTER 3

61

Table 8: Chain transfer probabilities for site 1 and 2 assuming a relative productivity for site 1 and 2 of 10 and 90%.

2HC /Cm

[-]

q1

[-]

q2

[-] ( )∑

=

−N

1i

2

ielmodGPC YYN

1

0

1.18⋅10-3

3.77⋅10-3

4.18⋅10-3

7.34⋅10-5 1.40⋅10-3 4.71⋅10-3 6.55⋅10-3

1.05⋅10-4 1.32⋅10-3 5.50⋅10-3 10.0⋅10-3

2.18⋅10-4 1.73⋅10-3 6.22⋅10-3 7.10⋅10-3

4.34⋅10-4 2.07⋅10-3 8.45⋅10-2 8.35⋅10-3

9.05⋅10-4 2.36⋅10-3 1.18⋅10-2 12.8⋅10-3

1.48⋅10-3 3.20⋅10-3 1.56⋅10-2 11.3⋅10-3

Table 9: Chain transfer probabilities for site 1 and 2 assuming a relative productivity for site 1 and 2 of 90 and 10%.

2HC /Cm

[-]

q1

[-]

q2

[-] ( )∑

=

−N

1i

2

ielmodGPC YYN

1

0

2.90⋅10-3

9.00⋅10-3

2.52⋅10-3

7.34⋅10-5 3.54⋅10-3 1.20⋅10-2 2.86⋅10-3

1.05⋅10-4 3.87⋅10-3 1.47⋅10-2 4.11⋅10-3

2.18⋅10-4 4.58⋅10-3 1.65⋅10-2 4.45⋅10-3

4.34⋅10-4 5.97⋅10-3 2.45⋅10-2 8.00⋅10-3

9.05⋅10-4 7.73⋅10-3 3.91⋅10-2 13.5⋅10-3

1.48⋅10-3 9.57⋅10-3 5.23⋅10-2 18.1⋅10-3

In order to obtain a better description of the molecular weight distribution, a “two-site model” has been used. The distributions obtained from the hydrogen series have been used to fit each distribution with two Schulz-Flory distributions, assuming an equal

CHAPTER 3

62

polymerization rate of each site (50/50). In figure 16 an example of such a fit is given presenting a much better fit as was obtained by using a single site model. The determined values for the chain termination probability of both sites together with the errors of the model predictions are given in table 7. To check the assumption of equal productivity of each site, fitting has been repeated assuming a productivity ratio of site 1 and 2 of 10/90 and 90/10, respectively. The obtained values for the chain termination probability of site 1 and 2 for these cases are given in table 8 and 9, respectively. It can be seen that the errors found for these cases are larger compared to the errors found for the case in which equal site productivity was assumed. However, the productivity ratio between the two sites has not been optimized. The obtained values for the chain termination probability of site 1 and 2 for all 3 cases are plotted versus the relative hydrogen concentration, see figure 17. It appears that both sites show a linear relation in all 3 cases, although the errors for the 10/90 and 90/10 model are much larger. Obviously, the linearity found is not a sufficient criterion to select the correct productivity ratio of the two-site model. The relations for the 50/50 model are given in equations 12 and 13.

0023.0C

C195.3q

m

H1

2 += (12)

0051.0C

C20.14q

m

H2

2 += (13)

0

0.01

0.02

0.03

0.04

0.05

0.06

0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012 0.0014 0.0016

[H2]/Cm

q [-

]

q1 based on 50/50 productivity






Figure 17: Determined chain termination probabilities for site 1 and 2 versus the relative hydrogen concentration.

CHAPTER 3

63

Conclusions Gas phase polymerizations of propylene with a heterogeneous metallocene catalyst have been executed at different temperatures, pressures and hydrogen concentrations. The experimentally measured reaction rate curves have been simulated with a model, which considers both the activation as well as the deactivation of the catalyst. It is argued that the clearly observed initial acceleration stage is due to chemical activation processes rather than a diffusion limitation process. At relative high polymerization rates deviations have been found from the model, which is attributed to a thermal runaway on particle scale. At these reaction conditions, the particle temperature may well be over the bulk temperature resulting in decomposition of vital catalyst components. Prepolymerization at low pressure and temperature increased the reaction rates. However, at higher temperatures large deviations with the Arrhenius plot still exist even for prepolymerization time of 7 minutes. Obviously, there are more effects than only particle overheating to be taken into account. For temperatures between 50 and 70°C, an activation energy for propagation was found to be 43.9 kJ/mol. This value is close to the value found by Meier et al.2, 48.0 kJ/mol where the same metallocene was used on a similar support with a smaller average particle size for gas phase polymerizations at 10 bar pressure. Important to mention is that also the same monomer sorption data has been used in the present paper to allow comparison of these kinetic data. A higher activation energy has been found for polymerizations in the absence of hydrogen at 15 bar pressure; 58.9 kJ/mol. The influence of monomer concentration has been investigated at pressures up to 25 bar. Again, deviations at high polymerization rates have been found, supporting the theory of a runaway at particle scale. The deactivation and activation constant are hardly affected by the monomer concentration for the range in which experiments have been done. The influence of hydrogen on the polymerization kinetics has been investigated at concentrations between 0 and 5.5 mol%. At low concentrations, the reaction rate strongly increases whereas at high concentrations this behavior flattens off. The most plausible explanation for the increased activity with increasing hydrogen concentration is the reactivation of dormant sites by hydrogen. The molecular weight and molecular weight distribution of the polymer samples produced during the kinetic study have been analyzed by GPC. The low molecular weight of the polymer is characteristic for the applied catalyst system. The molecular weight distribution is broader than expected for a single site catalyst, but in agreement with results reported for heterogeneous metallocene catalysts. The results from the

CHAPTER 3

64

temperature series are reasonably consistent with the results from the hydrogen and pressure series when the chain termination probability is plotted versus the relative hydrogen concentration. An empirical “two-site model” has been used in order to give a better description of the molecular weight distribution. Two Schulz-Flory distributions have been used to fit the observed distribution assuming an equal productivity of each site. Variation of the productivity ratio leads to larger errors between the model and GPC curves. The determined chain termination probabilities for each site are found to be linearly dependent on the relative hydrogen concentration, independently of the productivity ratio. The linearity itself is obviously not a sufficient criterion to select the correct productivity ratio of the two-site model. Acknowledgement - This work has been funded by BRITE-EURAM Project CATAPOL (BE 96-3022). We


technical assistance. G.H. Banis is acknowledged for his technical support and W.R. Smit, O. Slotboom,

M.F. Bergstra and M.J.M. Hattink for their contribution in the experimental part. The authors further

wishes to thank DSM Research for the GPC measurements.

Notation C Number of potential active sites mol/gmet

C* Number of active sites mol/gmet

Cm Monomer concentration kg/m3

2HC Hydrogen concentration kg/m3

Eact Activation energy J/mol j Chain length - kd Deactivation rate constant min-1

2Hk Chain transfer constant for hydrogen m3/mol.hr ki Activation rate constant min-1 km Chain transfer constant for monomer m3/mol.hr kp Propagation rate constant m3/mol.hr kx Chain transfer constant (other than H2 and m) m3/mol.hr Mm Molecular weight of the monomer kg/mol Mn Number averaged molecular weight kg/mol Mw Weight averaged molecular weight kg/mol p Pressure bar q Chain termination probability - R Gas constant J/mol.K Rp Reaction rate kg/gmet.hr t Time min

CHAPTER 3

65

T Temperature K djy Density function of the instantaneous MWD -

Y Standard format for GPC curve; jMy mdj kg/mol

Greek χ Flory-Huggins interaction parameter - Subscripts 0 Initial value 1 Site 1 2 Site 2 d deactivation i activation p propagation Abbreviations GPC Gel Permeation Chromatography MAO Methylaluminoxane TIBA Tri-isobuthylaluminium Literature 1. Resconi, L., Fait, A., Piemontesi, F., Colonnesi, M., Rychlicki, H., Zeigler, R.,

Macromolecules, 28, 6667-6676, (1995) 2. Meier, G.B., Weickert, G., van Swaaij, W.P.M., submitted to J. Appl. Pol. Sci.,

(2000) 3. Bonini, F., Fraaije, V., Fink, G., J. Polym. Sci. Polym. Chem. Ed., 33, 2393-2402,

(1995) 4. Kim, I., Choi, H.K., Kim, J.H., Woo, S.I., J. Pol. Sci. Pol. Chem. Ed., 32, 971-977,

(1994) 5. Weickert, G., Modellierung von polymerisationsreaktoren; Springer-Verlag Berlin

Heidelberg, Chapter 3, p 37, (1997) 6. Floyd, S., Choi, K.Y., Taylor, T.W., Ray, W.H., J. Appl. Polym. Sci., 31, 2231-2265,

(1986) 7 Floyd, S., Choi, K.Y., Taylor, T.W., Ray, W.H., J. Appl. Polym. Sci., 32, 2935-2960,

(1986) 8 McKenna, T.F., Spitz, R., Cokljat, D., AIChE J., 45(11), 2392-2410, (1999) 9. Samson, J.J.C., van Middelkoop, B., Weickert, G., Westerterp, K.R., AIChE J.,

45(7), 1548-1558, (1999)

CHAPTER 3

66

10. Fait, A., Resconi, L., Guerra, G., Corradini, P., Macromolecules, 32, 2104-2109, (1999)

11. Samson, J.J.C., Bosman, P.J., Weickert, G., Westerterp, K.R., J. Pol. Sci. Polym. Chem. Ed., 37(2), 219-232, (1999)

12. Carvill, A., Tritto, I., Locatelli, P., Sacchi, M.C., Macromolecules, 30, 7056-7062, (1997)

13. Weickert, G., Meier, G.B., Pater, J.T.M., Westerterp, K.R., Chem. Eng. Sci., 54, 3291-3296, (1999)

14. Spaleck, W., Aulbach, M., Bachmann, B., Küber, F., Winter, A., Macromol. Symp., 89, 237-247, (1995)

15. Blom, R., Dahl, I.M., Macromol. Chem. Phys., 200, 442-449, (1999)

CHAPTER 4

67

Chapter 4 FBR mini-plant for catalytic olefin polymerization: Particle mixing and propylene polymerization. Abstract Particle mixing and segregation have been studied in a small-scale fluidized bed reactor under pressure. It appeared that the solids mixing is relatively fast compared to the residence time of catalyst particles in case of a polymerization process. However, the smaller particles accumulate in the upper zone. Moreover, electrostatic charging causes the forming of a layer of small particles at the reactor wall with increasing thickness in time. The particles are redispersed after injection of an anti-static agent. Semi-batch propylene polymerization experiments at different fluidization velocities showed that vertical temperature gradients are caused by catalyst segregation as the principle factor. Specially at low gas velocities both segregation and mixing can be different under reacting conditions compared to non-reacting conditions due to different particle-particle interactions. Catalyst concentration gradients caused by incomplete mixing are expressed remarkably by the exothermic reaction even at low polymerization rates. Despite the fact that fluidized beds can hardly be scaled-up from experiments in reactors with diameters below 30 cm, small-scale fluidized bed reactor experiments can contribute to address and quantify incomplete mixing and electrostatic charging effects.

CHAPTER 4

68

CHAPTER 4

69

Introduction Nowadays, gas phase polymerization of olefins is one of the most important polymerization processes. Compared to other processes like the slurry and solution polymerization, gas phase processes have many distinct advantages. Gas phase processes can reduce capital and operating costs over conventional processes. Moreover, gas phase polymerizations offer a large variety of products, which could not be produced by other processes. The fluidized bed reactor is one of the main types of reactors for producing polyolefins in the gas phase. In the open literature on polyolefin polymerization there are no experimental data available obtained with this type of reactor because of the high investment costs for creating a pressurized experimental facility of this type, the high activity of the catalyst and its sensitivity to small traces of impurities. Moreover, phenomena like sheeting, electrostatic charging and entrainment of small particles may occur. Industrial fluidized bed reactors for propylene polymerization operate in a pressure range of 20 to 30 bar. Gas phase conversion per pass is kept low, 1-3%, to facilitate cooling of the reactor and to diminish concentration and temperature gradients in the reactor. At the typical residence time of 1 to 3 hours, polymer particles with a broad size distribution are obtained. Segregation phenomena can play an important role as the bigger particles, referred as the jetsam, tend to accumulate at the bottom of the bed. The smaller particles, referred as flotsam, show the tendency to accumulate at the top section. Mixing and segregation are competitive processes, both affected by the fluidization velocity. Specially at gas velocities much higher than the minimum fluidization velocity of the big particles, mixing is strongly enhanced. Because the catalyst/polymer particles grow in size during polymerization, it is possible that the upper section is rich in small and very active particles due to segregation. This may even introduce a vertical temperature gradient, which will affect the subsequent polymer properties. As the product in such a reactor is removed from the bottom section of the reactor, the bigger particles are preferentially extracted. This segregation effect results in a narrowing of the particle size distribution of the product. Sheeting, elutriation and agglomeration Sheeting and agglomeration of polymer particles are two serious problems, which can appear in modern gas phase polymerization processes. Sheeting of the reactor wall influences the performance of a fluidized bed reactor negatively and can plug the reactor in a later stadium. Moreover, blow out of highly active small particles elutriated from the bed can cause sheeting and plugging in other parts of the system. Agglomeration of particles affects the particle size distribution, fluidization behavior and polymer

CHAPTER 4

70

properties. Moreover, agglomeration of polymer particles reduces the internal heat and mass transfer thereby limiting the overall yield per gram of catalyst. Sheeting of the reactor wall of a fluidized bed occurs when the shear forces near the reactor wall are smaller than the adhesion forces between the particles and the reactor wall. There are in general three main causes for sheeting of the reactor wall. First, overheating of particles may occur due to very high reaction rates. The temperature of a particle can rise in this way above the softening temperature. Secondly, sticking may be due to polymer properties itself. Some (co-)polymers with low glass temperatures can easily form agglomerates. Thirdly, sheeting may develop due to the presence of excess negative or positive static charges pressing charged particles against the wall. The critical static voltage level for sheet formation is a complex function of polymer sintering temperature, operating temperature, pressure, gas composition, particle size distribution, slip velocity between gas and polymer particle, catalyst activity, etc. When the static charge reaches the level where the particles begin to stick to the wall, a layer of catalyst-containing polymerizing particles forms a non-fluidized layer at the reactor wall. The temperature in this layer can rise due to limited heat removal until the particles melt and fuse. At this point other particles will stick to the layer and it will grow in size. In some cases the layer will plug the system otherwise it will disturb operation by becoming loose from the wall. The phenomena mentioned above are rarely described in the open literature. Some authors have mentioned the effects but in most cases these are only studied qualitatively. The fluidization hydrodynamics have a dominant influence on agglomeration and sheeting since they both directly influence the physical parameters of the fluidizing bed. In US patent 5,283,2781 the use of an anti-static agent is described. The anti-static agent can be introduced into the reactor in the form of a solution, but dispersing a liquid uniformly is difficult. Another way to introduce the anti-static agent is to provide it in a prepolymerization step. It is known that prepolymerization itself can also prevent agglomeration of particles as described in US patent 5,241,0232. When the catalyst is injected without prepolymerization, reaction commences very intensively, creating a local speeding up of the reaction in the fluidized bed which can give rise to runaway reactions and also breaking up of solid particles of catalyst and loss of control of particle size may occur. Present work This paper reports on an experimental set up for the polymerization of propylene in a small-scale fluidized bed reactor under pressure. The set up is used to quantify the effects of particle mixing and segregation on the temperature profile in the reactor. It is obvious that the small-scale chosen, 10 cm in diameter and 1 m in height, is not representative for

CHAPTER 4

71

an industrial system. Fluidization in reactors with a small diameter may result in a slugging type of fluidization because the bubble size may approach the reactor diameter. This type of fluidization has a smaller vertical mixing rate, compared to the bubbling regime generally applied in industrial sized reactors. To compare well-mixed stirred semi-batch reactors, often used for kinetic measurements, with fluidized bed reactors, particle mixing and segregation in the particular fluid bed will generally be taken into account. Specially for small-scale fluid beds these effects may be crucial. Moreover, because of the high surface to volume ratio of small-scale reactors, electrostatic charging is enhanced, agglomeration of particles occurs more easily due to the lower vertical mixing rate. In comparison with industrial reactors, the experiments in such small-scale reactors may be interpreted as a worst case magnifying the different reactor problems mentioned. For the present work this was further strengthened by using a catalyst that produces a low molecular weight sticky polymer. In future work3 the reactor will be modified to allow also comparison with expected full-scale reactor conditions. In the current paper mixing and segregation have been studied at pressures up to 26 bar. Isobaric semi-batch polymerization experiments with a heterogeneous metallocene catalyst have been executed. Especially the influence of the superficial gas velocity has been investigated. The heterogeneous metallocene catalyst used, was the subject of an extensive kinetic study published before4. Experimental Fluidized bed reactor set up The experimental set up is schematically shown in figure 1. The reactor consists of a fluid bed zone and a disengagement zone. The internal diameter and height of the fluid bed zone is 10 and 100 cm, respectively. The diameter is limited by the capacity of the compressor and the height of the reactor results in a gas residence times comparable to industrial units. The disengagement zone has a diameter of 38.5 cm and a height of 25 cm. Observing glasses are available at four different vertical positions (12.5, 37.5, 62.5 and 87.5 cm above the distributor plate). At a distance of 15 cm above the distributor plate, tracer particles can be injected. Polymer samples can be withdrawn at 7 heights from 20 to 80 cm above the distributor at 10 cm interdistance. The catalyst is injected 15 cm above the distributor plate, see below. Product can be removed from the bottom section of the reactor, 5 cm above the distributor plate, semi-continuously by opening a valve to a container. The distributor plate consists of sintered stainless steel spheres with an average pore diameter of 0.3 mm. Steel nuts are packed under the distributor in order to increase mixing and break-up of the inlet velocity profile. Temperatures in the reactor are measured 50 mm below the distributor plate and at 6 different vertical positions with 10 cm interdistance, starting at 25 cm above the distributor.

CHAPTER 4

72

Two cyclones and a filter have been installed to remove fines entrained from the reactor. Gas leaving the reactor is cooled by a vertically positioned heat exchanger. A Eurotherm 900 EPC PID controller, controls the temperature of the gas entering the compressor. The control circuit comprises cold and hot water streams and an injection pipe of cold water directly into the inlet pipe of the heat exchanger. A second heat exchanger is used to

NO

Valve

Control Valve

Rupture Disk

Relief Valve

Check Valve

Flow meter

Filter

Heat exchanger

Compressor

IR

H2

NO

Inlet propylene andhydrogen

Product removal

CatalystinjectionV

ent

Ven

t

NO

Buffer vessel

CyclonesT6

T5

T4

T3

T2

T1

T0

Figure 1: Small-scale fluidized bed reactor for propylene polymerization.

CHAPTER 4

73

control the gas inlet temperature of the reactor within 0.2 °C. The reactor is placed in an air thermostat in order to decrease heat losses from the reactor wall to the surroundings. A blower circulates heated air in this cage with a temperature of about 50°C. The heat losses through the wall are thus reduced to less than 5% of the amount of convective cooling by the inlet gas. The compressor consists of two parallel heads each containing a double membrane. A buffer vessel is installed in order to compensate for pressure fluctuations. The flow circulating through the reactor system is regulated by two control valves (Orion 9000, USA) and is measured by a Brooks mass flow meter (5865 Ex). One control valve is located just before the reactor, a second valve is located in the bypass from the compressor loop. To measure and control the gas composition in the reactor, a sample flow from the reactor is lead continuously through a cascaded IR propylene analyzer (Servomex Xendos 2500) and hydrogen analyzer (Maihak, Thermor 615 thermal conductivity meter). Catalytic olefin polymerizations require extremely pure raw materials to avoid poisoning of the catalyst. Propylene, hydrogen and nitrogen are cleaned in separate purification systems to remove traces of O2, H2O, CO, etc. The purification system consists of BTS and molsieves columns, see Samson et al.5. Propylene is fed to the system as a liquid just before the first heat exchanger. A PID controller controls the partial pressures of propylene and hydrogen. The system is designed to operate at a maximum pressure of 30 bar. The installation is placed in a well ventilated concrete bunker and fully computer controlled. Thermocouples and pressure indicators in various places are used to determine the temperatures and pressures in the system every 10 seconds. If a measured value indicates a potentially dangerous situation, catalyst injection is stopped. A relief valve is placed on top of the reactor, which discharges at 30 bar. A rupture disk protects the installation for pressures exceeding 32 bar. Combustible gas detectors are installed, to monitor leakage of combustible gas. If a leak is observed, the experiment is stopped and the installation is flushed with nitrogen. Catalyst system The metallocene catalyst used for the polymerizations is rac-Me2Si[Ind]2ZrCl2, which produces a low molecular weight polymer. The sticky polymer is useful for studying agglomeration and sheeting in the fluidized bed reactor. The catalyst system used for this study, was kindly supplied by Witco Co. Bergkamen (Germany). It is supported on Grace silica with a metallocene concentration of 1 wt%. The MAO/SiO2 – support used for immobilization of the metallocene contained 25 wt% of alumina, giving a [Al]/[Zr] ratio

CHAPTER 4

74

of 386. The average particle size of the silica used (SD3216-30, 10 – 110 µm) is 51 µm. Tri-isobuthylaluminium (TIBA) has been used as a cocatalyst to increase the polymerization activity. The amount of TIBA and the precontact time has been kept constant for all experiments. The preparation procedure has been described before4. Catalyst injection system Injection of the catalyst is crucial in the procedure and needs special precaution, as only a few milligrams of these components have to be injected. A semi-continuous system has been developed to inject a reproducible amount of catalyst, see figure 2. The injection proceeds as follows. Catalyst is prepared under a nitrogen atmosphere in a glove box. The catalyst is diluted with polypropylene powder to prevent the catalyst particles from sticking to each other and to the reactor walls, which would possibly lead to local hot spots in the reactor. The catalyst container is connected to the catalyst injection system and all connection sections are thoroughly flushed with nitrogen. A three-way ball valve, located under the catalyst storage vessel, is used as a dosing device by injecting the contents of the ball valve by flushing with nitrogen. The valve thus provides a reproducible amount of polymer/catalyst mixture. Shortly before injection a plunger is displaced, which seals the reactor. To prevent monomer from entering the injection system, a small continuous flow of nitrogen blows into the reactor when the plunger is pulled back. A small nitrogen overpressure is always kept in the whole injection system to prevent monomer leakage into the injection system. The injection procedure itself is automated but the operator can adjust the time space between the successive injections from outside the box. For the semi-batch experiments described in this paper, the total amount of catalyst is injected in 12 injections at the beginning of the experiment with 20 seconds time space in between.

Catalyst storage

Plunger

Dosage valveN2

N2

N2

Rea

ctor

wal

l

Figure 2: Injection system for catalyst.

CHAPTER 4

75

Electrostatics To prevent electrostatic charging in the fluidized bed reactor, the whole set up is connected to the electrical earth. Moreover, in all regular tests Larostat 519 (dimethylethyl ammonium ethosulfate supported on silica, kindly supplied by PPG/Mazer chemicals) as anti-static agent has been used. The bulk density of the powder is 540 kg/m3 and the mean particle size is 9 µm. No effect of the anti-static agent on the polymerization rate has been found during separate tests in a 0.5-Liter stirred reactor. Addition of a small amount (~200 mg) prior to the catalyst injections assures a sufficient amount of anti-static agent in the reaction zone. This amount of anti-static agent corresponds to about 0.01 wt% of the total reactor contents. Due to the small average particle size of the powder, some anti-static agent leaves the fluidized bed reactor zone at high superficial velocities but the anti-static effect remains. Experimental procedures for mixing and segregation experiments For the mixing and segregation experiments a LLDPE powder has been used, kindly supplied by BASF AG (Germany), with a particle size distribution from 100 to 1400 ìm. The average particle size is 900 ìm. The powder is of type B, according to the Geldart classification system. For the mixing experiments, black tracer particles have been prepared from a sieved fraction (420 - 720 ìm) by coloring with black ink. For a mixing experiment, 15-18 grams of tracer particles have been injected into the fluidized bed reactor. At a specified time interval, samples are taken from the reactor at 7 vertical locations. The concentration of tracer is determined using a picture of the sample and counting the colored particles with imaging software. The vertical segregation profile has been determined by analyzing the particle size distribution of samples taken from the bed by laser diffraction (Sympatec Helos LF). The mixing and segregation experiments have been executed, with and without anti-static agent, using nitrogen as fluidization medium at 25 °C. The minimum fluidization velocities of the fluid bed at different conditions have been determined in a classical way from the pressure drop curve by changing the gas velocity. Experimental procedures for semi-batch polymerizations On start-up, the fluidized bed reactor is charged with 1.9 kg of a polypropylene powder, kindly supplied by DSM (The Netherlands). This powder consists of almost ideally spherical particles with a narrow particle size distribution. The average particle size is 570 µm. Before polymerization can take place the system must be freed from contamination. This can be done by thoroughly flushing the system with nitrogen and propylene followed by evacuation for 1 hour. Propylene is much more effective in cleaning the system than nitrogen. The reason for this is probably that propylene absorbs in the polypropylene under pressure, removing most of the contamination after venting.

CHAPTER 4

76

Before polymerizations are started, 30 grams polypropylene precontacted with 0.5 gram of TIBA is injected to remove the last traces of impurities. When the system is at the desired conditions, i.e. constant temperature, pressure, fluidization velocity and gas composition, the system is ready to start the catalyst injections. In this paper, the results from semi-batch experiments are reported. During these experiments, a certain amount of catalyst is injected at the beginning of the experiment. No product is removed from the reaction zone. The amount of polymer produced in such a batch experiment is small (less than 300 g compared to 1.9 kg start-up powder), which implies a limited change in bed height. During the experiments the inlet temperature, fluidization velocity, the total pressure and partial pressures are kept constant. The reaction is followed by monitoring the temperature at 6 vertical positions inside the reactor. Results Mixing of particles The minimum fluidization velocity has been determined experimentally at 15 and 26 bar nitrogen pressure and 25°C using the LLDPE powder. The determined minimum fluidization velocity at 15 bar is 7.5 cm/s, at 26 bar a value of 6.0 cm/s has been found. At 15 bar, slugs have been observed in the upper part of the bed starting at 14 cm/s. At 26 bar, slugs became visible at a slightly higher velocity: 16 cm/s. Particle mixing has been studied at 11 cm/s and 26 bar nitrogen pressure corresponding to 1.8*Umf. In figure 3, the concentration of tracer versus the vertical position in the reactor at different time intervals is presented. The concentration of tracer has been made dimensionless by division with the mean tracer concentration after complete mixing. After 15 seconds, the tracer is only detected at the second sampling point, probably due to upward transport of the tracer during injection with nitrogen. The tracer is mixed-up completely after 300 seconds, which is fast compared to the average residence time for a catalyst particle in a continuous polymerizing system.

CHAPTER 4

77

Segregation of particles Vertical segregation has been studied by measuring the flotsam concentration of the LLDPE powder at different vertical positions. Initially these experiments were carried out in absence of the anti-static agent. The flotsam has been defined as particles smaller than 500 ìm. In figure 4 the effect of the fluidization time on segregatio n at 11 cm/s and 26 bar is shown. It can be seen that the concentration of flotsam in the bottom zone of the bed is much lower compared to the concentration in the upper zone. After 1 hour, a fully segregated pattern can be distinguished. Moreover, the concentration of flotsam in the upper zone decreases with fluidization time. However, the particles are not leaving the reactor because no particles were found in the collectors of the cyclones or filter. A layer of fines with increasing thickness over time could be observed visually over the length of the reactor wall obviously caused by electrostatic charging. After addition of the anti-static agent the wall deposits disappeared after a few minutes. In the absence of electrostatic charges the particles can move more freely through the bed causing better mixing. In figure 5, the concentration of particles smaller than 420 ìm are compared of samples taken before and after the injection of the anti-static agent. The average concentration of particles smaller than 420 ìm observed before the addition of the anti -static agent is smaller than the original value, presented by the horizontal line in figure 5.

0

0.5

1

1.5

2

2.5

3

3.5

4

0 100 200 300 400 500 600 700 800 900

Z [mm]

C/C

mea

n [-

]

t = 15 s

t = 30 s

t = 45 s

t = 75 s

t = 120 s

t = 300 s

Figure 3: Tracer concentration at different time intervals and vertical position. Tracer particles injected at Z = 150 mm. U0 = 11 cm/s, 26 bar N2 and 25°C.

CHAPTER 4

78

After addition of the anti-static agent, the observed average concentration equals the average of the original powder indicating that no layer is build-up anymore. The results of the mixing and segregation experiments, together with additional measurements, will be the subject of a more detailed paper6.

0

5

10

15

20

25

30

35

0 200 400 600 800 1000

Z [mm]

Cfl

otsa

m [

vol%

]

1 hour

2 hours

3 hours

4 hours

Figure 4: Segregation profile of flotsam as measured at different time intervals illustrating the build up of a stagnant layer of fines. U0 = 11 cm/s, 26 bar N2 and 25 °C.

Figure 5: Effect of anti-static agent on the observed concentration profile of particles < 420 ìm. U 0 = 11 cm/s, 26 bar N2 and 25 °C.

0

2

4

6

8

10

12

0 100 200 300 400 500 600 700 800 900

Z [mm]

C [

vol%

]

With anti-static agent

Without anti-static agent

Mixed value of original powder

CHAPTER 4

79

Batch polymerization experiments For the batch polymerization experiments a polypropylene powder with almost no fines has been used for each experiment, see the experimental section, to minimize the risk of fines sticking at the reactor wall. At 25 bar and 68.5°C, Umf has been derived by measuring the pressure drop over the bed at different velocities, see figure 6. Using the same gas composition as used for the polymerization experiments, see table 1, a value of 4.2 cm/s has been found. Using air at atmospheric pressure and room temperature, a minimum fluidization velocity has been found of 7.4 cm/s. The bed height at minimum fluidization conditions could be estimated by visual observation. The results are presented in table 2 together with some physical constants and parameters of the reactor model for the batch polymerization experiments. Table 1: Experimental operating conditions.

Inlet temperature 68.5 °C Pressure 25 bar Amount of catalyst 0.5 gram Fluidization velocity Variable Gas composition Propylene 65 % Nitrogen 33 % Hydrogen 2 %

Three polymerization experiments were executed at 3 different superficial gas velocities. 0.5 gram of catalyst was used for each experiment. The process conditions are summarized in table 1. The development of the vertical temperature profile in time, both experimental result and model prediction, are given in figure 7, 8 and 9. The model prediction is based on a simple dynamic model using the well-mixed assumption and will be explained below. The superficial velocities were 8, 10 and 13.5 cm/s, respectively. During these experiments the thermocouple at the highest position, T6, is located above the surface of the bed. T0 represents the thermal couple below the distributor plate. T1 till T6 are positioned at 6 different vertical positions, starting at 25 cm above the distributor.

CHAPTER 4

80

Table 2: Physical constants and parameters for modeling of the fluidized bed reactor.

Parameter Value Parameter Value Ar (m

2) 7.854⋅10-3

Hmf (m)

0.54*

Cp,g (J/kg⋅K) 1473** ∆Hr (J/kg) 2.47⋅106 Cp,p (J/kg⋅K) 1926 kd0 (s

-1) 1.325⋅102 Cp,w (J/kg⋅K) 450 kp0 (m

3/kg.s) 1.303⋅109 Dr (m) 0.1 Umf (m/s) 0.042*

dw (m) 3⋅10-3 εmf 0.44* dp (m) 570⋅10-6 ρg (kg/m3) 35.04**

Eact,d (J/mol) 31.0⋅103 ρp (kg/m3) 910

Eact,p (J/mol) 48.0⋅103 ρw (kg/m3) 7860 * Experimentally determined at 25 bar and 68.5°C, see experimental results. ** Value based on 65% propylene, 33% nitrogen, 2% hydrogen at 341.5 K and 25 bar. As a finite amount of catalyst is introduced, which is subject to deactivation, the responding temperature behavior of the semi-isolated fluid bed will be of the shape of a broad pulse. Only if the bed would be well mixed all thermocouples would indicate the same temperature. The experiment at 8 cm/s, 1.9*Umf, shows a very diffuse and complex temperature profile, see figure 7. Due to the low value of U0-Umf, mixing in the bed is limited and

Figure 6: Pressure drop versus the superficial gas velocity at 25 bar and 68.5 °C.

0

5

10

15

20

25

0 5 10 15 20 25

U0 [cm/s]

pres

sure

dro

p [m

bar]

CHAPTER 4

81

local hot spots are observed. It is remarkably that even with a low activity catalyst as used in the present work, high local temperatures can be observed. In the beginning of the experiment the highest temperature is observed in the bottom section of the reactor. It seems that a hot front, slowly moves upwards to the top of the bed. This can be explained in two different ways. Firstly, by a slowly segregating bed where a zone with high catalyst concentration is moving to the top of the bed. Secondly, by a propagating thermal wave through a packed bed where all reaction heat is generated in bottom part of the bed at the beginning of the experiment. The speed of such a propagating wave can be calculated, see equation 1.

( )( ) ( )

ppgp

gp

0front CC

CUU

ρ+ρ

ρ≈ (1)

The speed of the front observed during the experiment is about 0.03 cm/s, whereas 0.23 cm/s is expected based on equation 1. Thus, the slowly segregating bed is the most probable explanation for the experimentally observed behavior.

The mixing experiments presented in this paper, carried out by tracer injection at 11 cm/s (about 1.9*Umf), showed a relative fast mixing behavior. This seems to be in contradiction with the results of the batch polymerization experiment at 8 cm/s, about 1.8*Umf. There are several reasons for the reduced mixing under reaction conditions. Firstly, it has to be mentioned that the mixing experiments were carried out with a different polymer, LLDPE vs. iPP, with a different particle size distribution. Moreover,

60

70

80

90

100

110

0 10 20 30 40 50 60 70 80

Time [min]

T [

°C]

T0

T1

T2

T3

T4

T5

T6

Model

Figure 7: Temperature profile, experiment and model prediction, at 8 cm/s; 1.9*Umf.

CHAPTER 4

82

the mixing experiments were carried using nitrogen as fluidization gas. During the polymerization experiments, propylene is solved into the amorphous phase of the polypropylene, which causes swelling of the polymer. The molecular weight of the polypropylene produced with the applied metallocene catalyst system is very low; about 20 kg/mol. This polymer is somewhat sticky, which is probably the major source of the limited mobility in the bed. The experiment at 10 cm/s, 2.4*Umf, shows a more regular temperature pattern with a lower maximum temperature and some temperature gradients inside the bed. The lowest temperature was found at the bottom section of the bed, as expected. The two thermocouples T5 and T6 measure a significant higher temperature than the thermocouples in the bed indicating a relative high polymerization activity in the top section of the bed and the lean phase above the bed.

The experiment at 13.5 cm/s, 3.2*Umf, shows the lowest maximum temperature and lowest temperature gradients in the bed. The two thermocouples T5 and T6 measure again a significant higher temperature than the thermocouples in the bed. It appears that the upper zone of the bed has a high catalyst concentration due to segregation of the catalyst. Due to the enlarged disengagement zone the elutriation of fine particles did not result in entrainment of particles from the reactor because no material, except from some anti-static agent, was found in the cyclones.

65

70

75

80

85

90

0 20 40 60 80 100

Time [min]

T [

°C]

T0

T1

T2

T3

T4

T5

T6

Model

Figure 8: Temperature profile, experiment and model prediction, at 10 cm/s; 2.4*Umf.

CHAPTER 4

83

Simplified reactor model The modeling of fluidized bed reactors for olefin polymerization has been the subject of several papers. No experimental data were used for verification in these publications. Choi et al.7 developed a model that takes temperature and concentration gradients within the bubble phase into account. McAuley et al.8 compared the two-phase model with a simple well-mixed model and concluded that the well-mixed model is appropriate for predicting the temperature and concentration in the gas phase of industrial fluidized bed polyethylene reactors if the bubble size is small as predicted by several maximum stable bubble size correlations. If the well-mixed assumption is valid, both for concentration as well as temperature profiles, experimental results and correlations developed from well-stirred gas phase laboratory reactors could be used immediately to model industrial-scale fluidized bed reactors. None of the models mentioned above takes the segregation of particles into account. In order to compare the measured temperature profiles, a simple dynamic model is given here assuming the well-mixed assumption.

67

68

69

70

71

72

73

74

75

0 10 20 30 40 50 60 70

Time [min]

T [

°C]

T0

T1

T2

T3

T4

T5

T6

Model

Figure 9: Temperature profile, experiment and model prediction, at13.5 cm/s; 3.2*Umf.

CHAPTER 4

84

Model assumptions: 1. During the reaction polymer is formed causing an increase in the bed level and

possibly a change in average bed properties since the polymer formed has different polymer and morphological properties. These aspects are not taken into account. The model calculates with an average density and heat capacity of the reaction zone.

2. The heat transfer from the bed to the reactor wall combined with the heat transfer through the wall is relatively fast compared to the heat transfer from the reactor wall to the reactor surroundings. It is assumed that the temperature of the reactor wall equals the temperature of the reactor contents. Heat losses from the wall are not taken into account.

3. The kinetic constants are based on an experimental kinetic study (see Meier et al.4) with the same catalyst system.

Under these assumptions, the modeling equations are as follows: The reaction rate is first order with respect to the number of active sites, C*, and the monomer concentration in the polymer, Cm.

m*

pp CCkR = (2)

Here kp=kp,0exp(-Eact,p/RT), where kp is the propagation rate constant and Eact,p the activation energy for propagation. The monomer concentration in the polymer has been calculated with the Flory-Huggins equation. The Flory-Huggins interaction parameter, is based on the experimental data published before4. The deactivation of the catalyst is described by a first order relation, see equation 3.

*d

*

Ckdt

dC−= (3)

Here kd=kd,0exp(-Eact,d/RT), where kd is the deactivation rate constant and Eact,d the activation energy for deactivation. The mass balance for monomer:

( ) ( ) pm0,mgr0mgr RXXAU

dt

XVd−−ρ=

ερ (4)

The energy balance:

( ) ( ) rp0g,pgr0av,pavwr HRTTCAU

dt

TCVd∆+−ρ=

ρ+ (5)

CHAPTER 4

85

The initial conditions are as follows: at t = 0, C* = C0 Xm = Xm,0 T = T0

(6)

Vr+w is based on the volume of the reaction zone plus the volume of the reactor wall:

( ) r2r

2wrrwr HDD

4

1VV −π+= ++ (7)

with Dr+w = Dr + 2dw. The average density and heat capacity of the volume heated-up:

w

w

g

g

p

pav mmm

1

ρ+

ρ+

ρ

=ρ (8)

w,pwg,pgp,ppav,p CmCmCmC ++= (9)

Here mp, mg and mw are the mass fractions of the polymer, monomer and wall respectively. They are defined as:

avwrgrp

pp VVM

Mm

ρ+ερ+=

+

(10)

avwrgrp

grg VVM

Vm

ρ+ερ+

ερ=

+

(11)

gpw mm1m −−= (12)

The mass fractions are influenced by the superficial gas velocity U0 since the volume of the reaction zone is increasing with increasing U0. Correlations given by Kunii and Levenspiel9 have been used to calculate the fraction of bubbles in the bed, which is required to calculate the bed volume. Many fluidization correlations published in literature are based on experimental work executed at normal pressure. It is known that the gas density applied for the polymerization of olefins can cause severe deviation. Properties like bubble size and regime of fluidization may change with increasing pressure. Moreover, the absorption of propylene in the amorphous part of the polymer may effect the interaction between polymer particles due to swelling of the polymer. This will affect hydrodynamic

CHAPTER 4

86

properties like minimum fluidization velocity Umf. Therefore, Umf, Hmf and εmf have been determined experimentally. The results of the model prediction are presented together with the experimental results in figure 7 – 9. The simple model seems to be able to describe the general shape of the temperature profiles inside the bed if U0 is not too low. Since the model is based on an ideal mixed system of both the gas and solid phase, i.e. segregation and mixing phenomena have not been taken into account, the model is not able to predict the temperature gradients observed. The increase in temperature during the first couple of minutes seems to be overestimated. This is because the kinetic model does not take the activation of the catalyst into account. The over prediction of the temperature rise for the experiment at 13.5 cm/s is due to the elutriation above the highest thermocouple in the reactor. It is clear from these results that the mixing and segregation processes of the solids should be included in the model. Such a model should also consider scaling-up relations for the actual solids mixing and segregation under reaction conditions if the model is to be used for industrial applications. Discussion It has to be noted that only batch polymerizations have been executed where at the beginning of the experiment a certain amount of catalyst is injected. In case of a continuous system, fresh catalyst is added continuously resulting in a particle size distribution, which is directly related to the residence time distribution of catalyst and polymerization kinetics. This of course will affect the mixing and segregation behavior of the system. A more complex model based on CFD and discrete particle description, see e.g. Hoomans et al.10, should be able to cope with these phenomena. Vertical solids mixing in a fluidized bed reactor is strongly related to the reactor scale. The solids mixing rate is in general expressed in engineering models in the form of an overall diffusion or dispersion coefficient, although this concept does not explain the complex physical phenomena involved. Matsen11 reported that the dispersion is directly proportional to bed diameters up to 10 m. In our small-scale fluidized bed reactor the average bubble size is about half of the reactor diameter. In case of an industrial sized reactor, this ratio is quite different. Here the reactor diameter and height are about 4 and 12 meters, respectively. Moreover, the mixing of solids is determined by the gross circulation pattern of solids generally observed in reactors of this scale. The bubbles tend to accumulate in the center of the reactor causing an upward “gulf stream” of solids in the center and down flow near the reactor wall. These phenomena are not observed in reactors at a small-scale as used in the present study. Matsen11 concluded that the limited

CHAPTER 4

87

success of bubbling bed models for the description of large units can be explained by the fact that bubble velocities and bed densities cannot be calculated from knowledge of superficial gas velocity and bubble size caused by the “gulf stream” present in the reactor. In our next paper3 we will report upon the controlled mixing in a small-scale fluidized bed reactor by installation of a draft tube in the center of the reactor. The small-scale fluidized bed reactor as described in the present paper cannot be used for scale-up. However, phenomena like electrostatic charging, sheeting and agglomeration, catalyst screening, slugging and the elutriation of fines are subjects that can be studied at actual reaction conditions in the present set up. Conclusion Vertical particle mixing and segregation have been studied in a small-scale fluidized bed under pressure. At higher gas velocities relative fast mixing of tracer particles has been found compared to the overall residence time of catalyst particles normally encountered in a polymerizing system. Particle size analysis of samples taken at different vertical positions showed that the smaller particles accumulate in the upper zone of the bed. The influence of electrostatic charging is important. A layer of small particles with increasing thickness in time is formed over the total length of the reaction zone but is redispersed after injection of an anti-static agent. The simple back-mixed model is not able to describe the vertical temperature gradients observed during the batch polymerization experiments, because no segregation and no mixing phenomena have been taken into account. Mixing and segregation phenomena of the solid phase (valid for reaction conditions) have to be incorporated. Three semi-batch polymerization experiments have been executed at different fluidization velocities. The phenomena observed during the mixing and segregation experiments were also observed during the polymerization experiments. It turned out that the catalyst particles accumulate in the upper zone of the bed creating a vertical temperature gradient. The speed of segregation is directly related to the mixing rate and therefore to the fluidization velocity. The balance between segregation and mixing cannot be compared to an industrial sized unit and vertical mixing should be enhanced to be more representative. Acknowledgement - This work has been funded by BRITE-EURAM Project CATAPOL (BE 96-3022). We


technical assistance. G.H. Banis is acknowledged for his technical support and E. Vellenga, M.J.

Mollenhorst, N.F. Geijsen and A.A. van Klaveren for their contribution in the experimental part. The

author further wishes to thank DSM Research and BASF AG for the materials and help they provided.

CHAPTER 4

88

Notation Ar Cross-sectional area of fluidized bed m2 C* Amount of active sites kg

Cm Monomer concentration in polymer kg/m3 Cp Heat capacity J/kg⋅K dp Mean particle size of polymer m dw Wall thickness of reactor m Dr Diameter of fluidized bed reactor m Eact,p Activation energy for propagation J/mol Eact,d Activation energy for deactivation J/mol Hr Bed height m Hmf Bed height at Umf m ∆Hr Heat of reaction J/kg kd Reaction rate constant for deactivation sec-1 kp Reaction rate constant for propagation m3/kg.s m Mass fraction - Mp Mass of start-up powder kg R Gas constant J/mol.K Rp Reaction rate kg/s t Time s T Temperature K T0 Temperature of inlet gas K Umf Minimum fluidization velocity m/s U0 Inlet gas velocity m/s V Volume m3 X Mass fraction of monomer in gas phase - Z Vertical position in reactor m Greek ε Void fraction of gas in bed - ρ Density kg/m3 Subscripts av Average value g Gas m Monomer p Polymer r Reactor w Wall

CHAPTER 4

89

0 Initial value Abbreviations Al Aluminium MAO Methylaluminoxane TIBA Tri-isobuthylaluminium Zr Zirconium Literature 1. US 5,283,278: BP Chemicals Ltd., (1994) 2. US 5,241,023: BP Chemicals Ltd., (1993) 3. Meier, G.B., Weickert, G., van Swaaij, W.P.M., submitted to AIChE J., (2000) 4. Meier, G.B., Weickert, G., van Swaaij, W.P.M., submitted to J. Appl. Pol. Sci.,

(2000) 5. Samson, J.J.C., Weickert, G., Heerze, A.E., Westerterp, K.R., AIChE J., 44, 1424,

(1998) 6. Roos, P., Westerterp, K.R. “Segregation and mixing at elevated pressures in a

fluidized bed polymerization reactor”, in preparation. 7. Choi, K.Y., Ray, W.H., Chem. Eng. Sci., 40(11), 2261 – 2279, (1985) 8. McAuley, K.B., Talbot, J.P., Harris, T.J., Chem. Eng. Sci., 49(13), 2035 – 2045,

(1994) 9. Kunii, D., Levenspiel, O., “Fluidization Engineering”, 2nd edition, Butterworth-

Heinemann, USA, (1991) 10. Hoomans, B.P.B., Kuipers, J.A.M., van Swaaij, W.P.M., in Fluidization IX, L.-S.

Fan, T.M. Knowlton (eds), 485, (1998) 11. Matsen, J.M., Powder Techn., 88, 237-244, (1996)

CHAPTER 4

90

CHAPTER 5

91

Chapter 5 FBR mini-plant for catalytic olefin polymerization: Controlled particle mixing, propylene polymerization and reactor modeling. Abstract A small-scale fluidized bed reactor has been equipped with a draft tube and cone to control the vertical solids mixing. The internal solids circulation rate is a non-linear function of the gas velocities. Strongly reduced elutriation and entrainment have been observed, compared to experiments without draft tube. Semi-batch propylene polymerizations have been executed at elevated pressures. The temperature profiles can be controlled by the solids circulation rate. Hydrogen injections led to an instantaneous increased polymerization rate, probably due to the reactivation of dormant sites. The irreversible deactivation rates of dormant and active sites seem to be the same. Moreover, hydrogen appeared to be very effective to broaden the molecular weight distribution. A compartment model is developed to describe the temperature profile in the reactor and related molecular weight distribution of the polymer.

CHAPTER 5

92

CHAPTER 5

93

Introduction The most widely established industrial gas phase technology for polyolefin production is the fluidized bed reactor operating at elevated pressures between 20 and 30 bar. In these large-scale units, bubbles rising from the distributor plate tend to accumulate in the center of the reactor, causing an upward “gulf stream” of polymer powder. Particles closer to the wall are generally moving in downward direction thus creating an overall circulation. Depending on the H/D a single or more mixing cells may occur. Intensive exchange between up and down flows may create a diffusion type of mixing. This strong mechanism of solids mixing reduces temperature and concentration gradients in fluid beds. In polymerization temperature and concentration gradients would affect directly the molecular weight distribution of the polymer. However, the rate of mixing in fluid beds is strongly dependent on the scale of operation, see e.g. Matsen et al.1, and on a small scale the axial mixing is lower and “gulf streaming” less prominent or absent. In a previous paper2, we have reported upon the solids mixing and propylene polymerization in a small-scale fluidized bed reactor. Severe particle segregation and elutriation was found. At polymerization conditions, where the small particles are most active, high local temperatures were found due to segregation insufficiently counteracted by vertical mixing. In the present paper, a modified small-scale fluidized bed reactor is used to study aspects of the polymerization of olefins at conditions resembling those of industrial fluidized bed reactors as well as strongly deviating conditions with controlled thermal gradients. The latter may be of interest for a controlled broadening of the molecular weight distribution of the polymer. To realize this, the vertical solids mixing, elutriation and entrainment should be controlled. A vertical draft tube installed in the center of the fluid bed is a well-known instrument of chemical engineers to control the axial solids mixing but has never been used for a fluidized bed polymerization reactor before according to the scientific and patent literature. Using the draft tube, the internal solids circulation rate can be controlled by the gas velocity in the draft tube and in the annulus space. In this way it is possible to simulate the solids circulation pattern observed at industrial scale in a small-scale unit. Because the draft tube and annulus are generally operated at different gas velocities, both sections have different heat transport properties, temperatures and temperature gradients. Therefore, for a particle circulating through the reactor an oscillating particle temperature may occur along the particle trajectory. The temperature profiles in the reactor will broaden the molecular weight distribution of the polymer. Moreover, it should be possible to control the molecular weight distribution of the polymer by controlling the circulation rate of the solids in the reactor.

CHAPTER 5

94

Note that a recently developed new multi-zone process for the polymerization of olefins3 is based on the same principles. However, here the solids circulation is external and based on a riser-downer combination. Although the draft tube can be used to increase the solids mixing in the present investigation, the emphasis is placed on the elimination of segregation and the creation of different zones in one reactor. This was already realized with one single gas phase reactant injection point although the draft tube would also allow injection of different gases in different zones. The latter is not part of the present work. Experimental FBR set up The experimental set up, see figure 1, comprises a fluidized bed reactor, 10 cm inner diameter and 1 m in height, a special catalyst injection system, two cyclones, a membrane compressor, two heat exchangers, flow control valves, mass flow meter and has been described in detail in a previous paper2. For the experiments presented in this paper, a draft tube and conical bottom section have been installed to force the solids circulation with a single gas injection point. In this set up particles are forced to move upwards in the draft tube section and enter the annulus section where they move in a densified form with no bubbles present under the action of gravity (“moving bed”). Polymer particles leaving the annulus section are entering the well-mixed cone section and are reintroduced to the draft tube section. In this way a solids circulation is set up, which can be controlled by the inlet gas velocity of the unit. The length, inner diameter and wall thickness of the draft tube are 770, 50 and 5, respectively. The draft tube, made of glass, is placed in the center of the reactor 130 mm above the distributor plate and is fixed in place with three horizontal rods at the bottom and the top. The inlet diameter and height of the cone are 50 and 150 mm, respectively. Thermocouples in the annulus section are available at 6 different vertical positions (25, 35, 45, 55, 65 and 90 cm above the distributor plate). 2 thermocouples are installed in the draft tube section (20 and 60 cm above the distributor plate). The device to measure the pressure over the annulus is connected to the reactor at 15 and 90 cm above the distributor plate. The installation is placed in a concrete bunker and fully computer controlled.

CHAPTER 5

95

Catalyst system The catalyst used, is the well-known metallocene rac-Me2Si[Ind]2ZrCl2. The catalyst, kindly supplied by Witco Co. Bergkamen (Germany), is supported on PQ silica with a concentration of 0.79 wt% of metallocene. The MAO/SiO2 – support used for immobilization of the metallocene contained 19.1 wt% of alumina. Tri-isobuthylaluminium (TIBA) has been used as cocatalyst to increase the polymerization

NO

Valve

Control Valve

Rupture Disk

Relief Valve

Check Valve

Flow meter

Filter

Heat exchanger

Compressor

IR

H2NO

Inlet propylene andhydrogen

Product removal

Ven

t

Ven

t

NO

Buffer vessel

Cyclones

Te

Catalystinjection

Ta(1)

Ta(6)

Ta(4)

Ta(5)

Ta(3)

Ta(2)

Td(2)

Td(1)

dP

Figure 1: Schematic representation of the fluidized bed reactor equipped with draft tube.

CHAPTER 5

96

activity. The amount of TIBA and the precontact time has been kept constant for all experiments. The exact preparation procedure has been described in a previous paper4. Experimental procedure for solids circulation measurements The solids circulation rate as function of the inlet gas velocity has been determined by measuring the residence time of tracer particles in the annulus section. On start-up, the reactor is charged with 2 kg of polypropylene powder. This powder consists of almost ideally spherical particles with a narrow particle size distribution and may be regarded as a type B powder according to the Geldart classification system. The average particle size is 570 µm. With this amount of polymer in the system, the bed level in the annulus is about 10 cm below the top of the draft tube. To prevent electrostatic charging, about 200 mg Larostat 519 (dimethylethyl ammonium ethosulfate supported on silica, PPG/Mazer chemicals) as anti-static agent has been used for every experiment. Before the tracer injection, the system is brought to the desired temperature, pressure, gas composition and gas velocity. About 25 grams of colored tracer particles are injected from above at the top of the annulus bed. Because the particles are moving under nearly plug flow conditions in the annulus, the tracer particles can be followed through the observing glasses located at 4 different vertical positions (12.5, 37.5, 62.5 and 87.5 cm above the distributor plate). The time required to flow through the annulus section has been measured in this way and is used to calculate the circulation rate. Experimental procedure for semi-batch polymerizations On start-up, the reactor is charged with 1.9 kg of a polypropylene powder, the same as was used for the circulation rate measurements. Before polymerization can take place the system must be freed from contamination. The exact cleaning procedure has been described in a previous paper2. When the system is at the desired conditions, i.e. constant temperature, pressure, anti-static agent concentration, gas velocity and gas composition, the system is ready to start the catalyst injections. In this paper, the results of semi-batch experiments are reported. During these experiments, a small amount of catalyst is injected at the beginning of the experiment. No product is removed from the reaction zone. The amount of polymer produced in such a batch experiment is relatively small (less than 300 g) compared to the amount of start-up powder, which implies a limited change in bed height and circulation rate. However, the bed level in the annulus never exceeds the height of the draft tube during the experiment. During the semi-batch experiments the inlet temperature, gas velocity, the total pressure and partial pressures are kept constant. As the reactor is to a large degree isolated from the environment, the reaction rate in semi-batch operation can be followed by the temperature response, interpreted by the in-stationary heat balance. The temperature is therefore monitored continuously at the several vertical positions in the annulus and draft tube section. The polypropylene samples obtained at the end of the experiment were characterized with

CHAPTER 5

97

respect to their molecular weight and molecular weight distribution by Gel Permeation Chromatography (GPC). These GPC measurements were carried out on a Waters M150C apparatus with TSK columns at 140°C using 1,2,4-trichlorobenzene as solvent. Results Solids circulation Experiments at different pressures and gas compositions have been executed, see table 1, to study the influence on the circulation rate. The residence time of solids in the annulus section has been measured as function of the gas velocity, see figure 2. Note that the gas velocity is based on the velocity at the cone inlet, i.e. 50 mm in diameter. This velocity does not represent the fluidization velocity in the draft tube or annulus section, as the gas is divided over these two sections. As can be seen from figure 2, the residence time of solids in the annulus decreases with increasing gas velocity. The gas density hardly influences the circulation rate, although some differences have been observed between the experiments at 15 and 25 bar with 65% propylene. The measured solids velocity in the annulus stays for all experiments (between 1.7 and 3.5 cm/s) below the minimum fluidization velocity, i.e. 4.2 cm/s. Table 1: Process conditions for circulation rate measurements.

T [°C] p [bar] Propylene [%] Nitrogen [%] ρg [kg/m3] 338

25

65

35

35.5

338 25 0 100 25.0 338 15 65 35 20.4 293 1 Air 1.0

A certain minimum gas velocity (Ue,min) is required to start the circulation of solids in the system. This minimum gas velocity decreases with increasing density of the gas phase. The minimum gas velocity required at atmospheric pressure is about 42 cm/s, whereas this velocity is about 24 cm/s at 25 bar with 65% propylene. Further visual observation of the solids circulation experiments showed that the tracer particles were mixed through the reactor contents after 3 to 4 circulations. Most of the mixing is achieved in the cone section. The flow behavior of solids in the draft tube is of the slugging type. At very high velocities, 45 cm/s, a transition towards turbulent fluidization is observed. Only minor dispersion of solids in the annulus is observed. No bubbles in the annulus section have been detected.

CHAPTER 5

98

The pressure drop over the annulus has been measured at different gas velocities and bed weights using nitrogen at 25 bar and 60 °C as fluidization medium, see figure 3. As expected, larger pressure drops have been observed with increasing bed weight. Increasing the gas velocity results in a decreasing pressure drop over the annulus, indicating a lower solids hold up in the draft tube with increasing gas velocity because of the coupled pressure balance for the annulus and draft tube. The bed height in the annulus is almost constant with increasing gas velocity. The decreasing pressure drop over the system with increasing gas velocity also indicates a decreasing upward gas velocity in the annulus section with increasing gas velocity. The lower gas bypassing to the annulus section with increasing gas velocity is in agreement with the results reported by Ji et al.5. The Ergun equation have been used to calculate the minimum upward gas velocity in the annulus, i.e. at the highest solids circulation rate. Assuming a bed porosity of 0.42 and a shape factor of 1, a slip velocity of 4.6 cm/s have been found. The downward solids velocity at these conditions is 3.5 cm/s. The minimum upward gas velocity in the annulus is therefore 1.1 cm/s.

0

10

20

30

40

50

60

70

80

90

0 20 40 60 80 100Ue [cm/s]

Res

iden

ce t

ime

[s]

25 bar, 100% Nitrogen

25 bar, 65% Propylene

15 bar, 65% Propylene

1 bar, air

Ue,min

Figure 2: Solids residence time in the annulus with a length of 77 cm as function of gas velocity. Ue,min represents the minimum gas velocity to start solids circulation.

CHAPTER 5

99

Semi-batch polymerizations Semi-batch propylene polymerizations have been executed at 25 bar and different gas velocities, i.e. different solids circulation rates. The first experiment has been executed with a gas inlet temperature of 68°C. Before the catalyst is injected the system is stabilized, i.e. constant temperature, pressure, gas velocity, etc. Then the total amount of catalyst, 1 gram, is injected in about 4 minutes. The responding temperature behavior of the semi-isolated fluid bed will be of the shape of a broad pulse, because the catalyst is subject to deactivation. Only at a very high solids circulation rate all thermocouples, both in the draft tube and annulus, would indicate the same temperature. At lower circulation rates temperature gradients will appear in both sections as follows. Particles coming from the annulus will be cooled down in the cone by the relative cold gas thereby absorbing monomer. On the way up in the draft tube, the temperature will increase a little due to the polymerization reaction although this remains a small effect due to the short residence time in this section. Along the way down in the annulus much more polymerization takes place at these longer residence times and less convective cooling capacity is available resulting in a relative large temperature gradient. The temperature profile of the annulus section of the first experiment is presented in figure 4. The initial decrease in temperature is due to the catalyst injections with cold nitrogen. After the catalyst injections, the temperature starts rising rapidly and reaches a

Figure 3: Influence of total polymer hold up on the pressure drop over the annulus as function of the gas velocity using 25 bar N2 at 60°C.

13

14

15

16

17

18

19

20

18 22 26 30 34 38 42 46 50 54

Ue [cm/s]

pres

sure

dro

p [m

bar]

1.8 kg

1.9 kg

2.0 kg

2.1 kg

2.2 kg

2.3 kg

2.4 kg

CHAPTER 5

100

maximum after about 15 minutes. At this point, the largest temperature gradients inside the reactor have been observed, about 5°C. Then the catalyst starts to deactivate and the temperature inside the reactor returns to its initial value after about 1.5 hours. The maximum temperature observed in this experiment was about 83°C. This would be above the critical temperature for which a thermal runaway on particle scale may occur4.

Table 2: Process conditions for batch polymerization experiments with Te = 58°C.

Ue [cm/s] ua,s [cm/s] [C3H6] [%] [N2] [%] [H2] [%] 25.1

1.7

69.6

29.4

1.0

30.4 2.3 69.0 30.0 1.0 36.2 3.0 66.6 32.4 1.0 42.7 3.5 67.1 31.9 1.0

There are several ways to prevent the thermal runaway, for instance by prepolymerization or by lowering the gas inlet temperature. The next series of experiments were executed with a gas inlet temperature of 58°C to minimize the risk of a thermal runaway, see table 2. The temperature profiles of the annulus section of these experiments are given in

62

65

68

71

74

77

80

83

86

0 10 20 30 40 50 60 70 80 90 100

t [min]

T [

°C]

Ta(1)

Ta(2)

Ta(3)

Ta(4)

Ta(5)

Ta(6)

Catalyst injections

Figure 4: Temperature profile of the annulus as a function of time during a semi-batch polymerization. Ta(1) and Ta(6) indicate the temperature in the bottom and top of the annulus, respectively.

CHAPTER 5

101

figure 5 – 8. The experiment at the lowest gas velocity, 25.1 cm/s, shows the highest internal temperature gradient and absolute temperature. The experiment at 42.7 cm/s shows only minor internal temperature gradients, about 1°C, and may be interpreted as an experiment in a CSTR. The scattering during the initial phase of the experiment at 30.4 cm/s and twice during the experiment at 36.2 cm/s is due to some solids circulation disturbances, probably because of some temporary sticking of particles at thermocouples. The temperature gradients measured inside the draft tube section are much less due to the high gas velocity. The gas velocity in the draft tube is for the experiments at 25.1 and 42.7 cm/s about 3.3 and 7.6 times Umf, respectively. The temperature gradients in the draft tube observed for the experiment at the highest and lowest circulation rate were 0.2 and 1°C, respectively.

55

58

61

64

67

70

73

76

0 10 20 30 40 50 60 70 80

t [min]

T [

°C]

Ta(1)

Ta(2)

Ta(3)

Ta(4)

Ta(5)

Ta(6)

Figure 5: Temperature profile of the annulus as a function of time during a semi-batch polymerization. Ue = 25.1 cm/s, ua,s = 1.7 cm/s.

CHAPTER 5

102

55

58

61

64

67

70

73

76

0 10 20 30 40 50 60 70 80

t [min]

T [

°C]

Ta(1)

Ta(2)

Ta(3)

Ta(4)

Ta(5)

Ta(6)


55

57

59

61

63

65

67

69

0 10 20 30 40 50 60 70 80 90

t [min]

T [

°C]

Ta(1)

Ta(2)

Ta(3)

Ta(4)

Ta(5)

Ta(6)


CHAPTER 5

103

Injection of hydrogen From kinetic experiments published before4, it was concluded that the current catalyst system is sensitive towards hydrogen. The molecular weight of the produced polymer is decreasing with increasing hydrogen concentration, whereas the polymerization activity is increasing with increasing hydrogen concentration. The main reason for the increased polymerization rate is that hydrogen prevents the formation of dormant sites, which occur after a secondary insertion of propylene. In a next series of experiments, see figure 9, hydrogen has been injected in the gas recycle at different moments during semi-batch experiments. Thus a certain fraction of the end product is produced at a low hydrogen concentration giving a high molecular weight polymer, the other part of the product is produced at a high hydrogen concentration giving a low molecular weight polymer. During the experiment the propylene concentration is kept at its initial value, 60 vol%, i.e. the pressure increases about 1 bar after the injection of hydrogen. The hydrogen concentration before injection is 0.0217 kg/m3 (∼1 vol%) , after the injection of hydrogen the concentration is raised to 0.123 kg/m3 (7 vol%). All experiments have been executed with a gas inlet temperature of 59.5 °C and a gas velocity of 36 cm/s, i.e. about 3 cm/s solids velocity in the annulus.

55

57

59

61

63

65

0 10 20 30 40 50 60 70 80 90 100 110

t [min]

T [

°C]

Ta(1)

Ta(2)

Ta(3)

Ta(4)

Ta(5)

Ta(6)


CHAPTER 5

104

Due to some problems with the catalyst injection system, the experiments showed lower polymerization activities. The results will be interpreted in a qualitative way, based on the temperature measurements of the annulus, by normalizing the curves. In figure 9, the results are given of the hydrogen pulse experiments showing the difference between the gas inlet temperature and the temperature at the bottom of the annulus, Ta(1), during the experiments. The negative value of Ta(1)-Te during the first couple of minutes is the result of the catalyst injections with cold nitrogen. As can be seen from figure 9, the temperature immediately increases after injection of hydrogen, indicating an almost instantaneous increase in polymerization rate. The reaction rate to reactivate the dormant sites is probably fast.

A second feature observed from figure 9 is that the different temperature curves follow more or less the same master curve before and after the injection of hydrogen. The active sites probably deactivate with about the same rate at the applied process conditions. Moreover, the relative increase in temperature after the injection is more or less constant. Because the increase is probably caused by the reactivation of dormant sites, one may conclude that the relative concentration of dormant sites, relative towards the amount of active sites, is constant during the experiment. Dormant sites deactivate probably with the same rate as the active sites.

Figure 9: Normalized temperature profiles in the bottom of the annulus during different semi-batch polymerizations with hydrogen injections at different moments. Ue = 36 cm/s, ua,s = 3 cm/s.

-2

-1

0

1

2

3

4

5

6

7

0 10 20 30 40 50 60 70

t [min]

∆∆T [

°C]

Injection after 12.5 min

Injection after 16 min



CHAPTER 5

105

Reactor model A reactor model has been developed, which should be able to describe the temperature and concentration gradients inside the reactor and the related molecular weight distribution of the polymer product. The reactor model uses the measured solids circulation rate measurements as input. The results of the circulation rate measurements are highly related to the geometry of the set up, e.g. diameter and length of the draft tube, the distance between draft tube and cone and the angle of the cone. Several authors have reported that the geometry of the reactor has a large influence on the solids circulation rate6-9. No data are available on pressurized systems. In the present paper at the given geometry, just the influence of the gas velocity on the solids circulation rate has been investigated and is used as input for the model. The assumptions for the reactor model are summarized below.

1. The reactor is divided in three sections; the annulus, draft tube and cone section. The draft tube and annulus sections are divided in N CSTR’s in series to simulate plug flow in both sections. The cone is considered as one CSTR. The reactor model is schematically presented in figure 10.

2. The resistance towards mass and heat transfer between the gas and solids in each compartment is negligible, which is acceptable10 in case the catalyst activity is not extremely high and the particles are relatively small. Particles circulating in the system are continuously absorbing and

Solids

Gas

Ann

ulus

Dra

ft tu

be

2

1

Nd

3

Na

3

2

1

Cone

Gas out

Figure 10: Schematic presentation of the reactor model.

CHAPTER 5

106

desorbing propylene due to the temperature gradients in the system. Equilibrium is assumed in each compartment.

3. The energy effects due to absorption and desorption of propylene have been neglected.

4. The gas and solids velocities in the reactor are assumed to be constant, because the monomer conversion per pass through the reactor is low.

5. Minimum fluidization conditions for the annulus have been assumed. The voidage of the cone and draft tube section are equal to the experimentally determined value.

6. No heat transfer from the annulus to the draft tube section takes place. 7. The heat transfer from the annulus bed to the reactor wall combined with the heat

transfer through the wall is relatively fast compared to the heat transfer from the reactor wall to the reactor surroundings. This, because the reactor is placed in an air thermostat at 50 °C, which reduces the heat losses to a large extend. It is assumed that the wall temperature of the adjacent annulus compartment equals the temperature of the reactor contents. Heat losses are described using a Nusselt correlation for heat transfer from a flat plate to laminar flowing air.

The kinetic model and kinetic constants are based on an experimental study4 with the same catalyst system. The model does respect the initiation and deactivation of the catalyst, see scheme 1. The reaction rate is described by first order kinetics with respect to the number of active sites, C*, and the monomer concentration in the polymer, Cm. The monomer concentration is calculated using the Flory-Huggins equation. The Flory-Huggins interaction parameter, is based on the experimental data published before11. The density and heat capacity of the gas phase are functions of the local temperature and gas composition and have been described according to Peng-Robinson and Daubert et al. (1985), respectively.

To obtain a clear understanding of the reactor model, the material and energy balances will be described separately for the cone compartment, an arbitrary draft tube compartment and an arbitrary annulus compartment.

C C* C*-Pj

ki kp

kd kd

Deactivated catalyst

Scheme 1: Activation, propagation and deactivation of the catalyst.

CHAPTER 5

107

Cone compartment Mass balance for non-activated catalyst in the cone:

( ) ciCc

C1,as,m

c CkXXdt

dC−−φ=

(1)

Here CcX is the mass fraction of non-activated catalyst in the cone, defined as:

sccc

cCc )1(VC

CX

ρε−+=

(2)

The first term in the right side of equation 1 concerns the convective flow of non-activated catalyst. The second term concerns the initiation of non-activated catalyst with initiation rate constant ki. Mass balance for activated catalyst in the cone:

( ) *cdci

Cc

Ca,1sm,

*c CkCkXX

dtdC **

−+−φ= (3)

Here *C

cX is the mass fraction of activated catalyst in the cone, defined as:

scc*c

*cC

c )1(VCC

X*

ρε−+=

(4)

The first term in the right side of equation 3 concerns the flow of activated catalyst, the second term the initiation towards active catalyst and the third term the deactivation of the catalyst. Mass balance for monomer in the cone:

( ) ( ) c,pmc

megee

mcgcc RmmAU

dt

mVd−−ρ=

ρε

(5)

The first term in the right side of equation 5 concerns the flow of monomer through the cone. The second term represents the consumption of monomer in the cone due to polymerization. The reaction rate in the cone compartment is described with:

c,m*cpc,p CCkR = (6)

Here *cC stands for the amount of activated catalyst in the cone compartment.

Energy balance for the cone:

CHAPTER 5

108

( ) ( ) ( ) rc,pc1,as,ps,mceg,pgee

cc,av,pc,avc HRTTCTTCAUdt

TCVd∆+−φ+−ρ=

ρ

(7)

The first two terms in the right side of equation 7 concern the enthalpy associated with the gas and solid flow through the cone. The third term stands for the amount of heat produced due to polymerization. Starting conditions for the cone at t = 0:

0,cc C)0(C =

0)0(C*c =

0c T)0(T =

m0

mc m)0(m =

(8)

Draft tube compartment i The mass and energy balances for a draft tube compartment are analogous: Mass balance for non-activated catalyst:

( ) i,diC

i,dC

1i,ds,mi,d CkXX

dt

dC−−φ= −

(9)

Here Ci,dX is the mass fraction of non-activated catalyst, defined as:

sdi,di,d

i,dCi,d )1(VC

CX

ρε−+=

(10)

Mass balance for activated catalyst:

( ) *i,ddi,di

Ci,d

C1i,ds,m

*i,d CkCkXX

dt

dC **

−+−φ= − (11)

Here *Ci,dX is the mass fraction of activated catalyst, defined as:

sdi,d*

i,d

*i,dC

i,d )1(VC

CX

*

ρε−+=

(12)

Mass balance for monomer:

CHAPTER 5

109

( ) ( ) i,d,pm

i,dm

1i,dgdg,d

mi,dgdi,d RmmAu

dt

mVd−−ρ=

ρε−

(13)

The equation for the polymerization rate:

i,d,m*

i,dpdi,p CCkR = (14)

Energy balance:

( ) ( ) ( )

ri,d,p

i,d1i,ds,ps,mi,d1i,dg,pgdg,di,d

avi,d,p

avi,di,d

HR

TTCTTCAudt

TCVd

∆+

−φ+−ρ=ρ

−−

(15)

Starting conditions for t = 0:

0,di,d C)0(C =

0)0(C*i,d =

0id, T)0(T =

m0

mi,d m)0(m =

(16)

Coupling between cone and draft tube:

c0,d TT =

Cc

C0,d XX =

** Cc

C0,d XX =

mc

m0,d mm =

(17)

Annulus compartment i Mass balance for non-activated catalyst:

( ) i,aiC

i,aC

1i,as,mi,a CkXX

dt

dC−−φ= +

(18)

Here Ci,aX is the mass fraction of non-activated catalyst, defined as:

CHAPTER 5

110

sai,ai,a

i,aCi,a )1(VC

CX

ρε−+=

(19)

Mass balance for activated catalyst:

( ) *i,adi,ai

Ci,a

C1i,as,m

*i,a CkCkXX

dt

dC **

−+−φ= + (20)

Here *Ci,aX is the mass fraction of activated catalyst, defined as:

sai,a*

i,a

*i,aC

i,a )1(VC

CX

*

ρε−+=

(21)

Mass balance for monomer:

( ) ( ) i,a,pm

i,am

1i,agag,a

mi,agai,a RmmAu

dt

mVd−−ρ=

ρε−

(22)

The equation for the polymerization rate:

i,a,m*

i,apai,p CCkR = (23)

Energy balance:

( ) ( ) ( )

( )∞

+−

−πα−∆+

−φ+−ρ=ρ

TTD4

1HR

TTCTTCAudt

TCVd

i,a2ri,ari,a,p

i,a1i,as,ps,mi,a1i,ag,pgag,ai,a

avi,a,p

avi,ai,w,a

(24)

Because the heating and cooling of the reactor wall is respected in the accumulation term, the density and heat capacity of the annulus compartment is based on the weight-averaged value of the gas, solids and the stainless steel wall. The last term in equation 24 concerns the heat transfer from the reactor wall to the air flowing through the isolated metal cage. Starting conditions for t = 0:

0,ai,a C)0(C =

0)0(C*i,a =

0ia, T)0(T =

(25)

CHAPTER 5

111

m0

mi,a m)0(m =

Coupling between cone and draft tube and annulus:

c0,a TT =

da N,d1N,a TT =+

CN,d

C1N,a da

XX =+

*

d

*

a

CN,d

C1N,a XX =+

mc

m0,a mm =

(26)

Sensitivity towards hydrogen Both the propagation rate constant as well as the initiation rate constant depend on the concentration of hydrogen in the gas phase, see Meier et al.4. The deactivation rate constant was found to be not sensitive towards hydrogen above hydrogen concentrations of 0.01 kg/m3. The sensitivity towards hydrogen of both kp and ki are described as empirical relations4, see equations 28 and 29.

( ) RT

E

pc

2pp

p,act

eb]H[ak−

+= (27)

( ) RT

E

i2ii

i,act

eb]H[ak−

+= (28)

RT

E

o,dd

d,act

ekk−

= (29)

Molecular weight distribution The molecular weight distribution of the product from a single site catalyst can in general be described with a Schulz-Flory distribution. It has been shown in a previous paper4 that the molecular weight distribution of the polymer produced with the current heterogeneous metallocene catalyst can best be described with a “two-site model”. The chain transfer probability of site 1 and 2 depends only on the hydrogen concentration divided by the monomer concentration (in the polymer). At constant temperature, pressure and gas composition, the instantaneous molecular weight distribution of a polymer synthesized with a single-site catalyst can be described with a Schulz-Flory distribution, see equation 30.

CHAPTER 5

112

jq2dj ejqy −= (30)

Here, djy is the density function of the instantaneously formed molecular weight

distribution. The “two-site model” presented in a previous paper4 uses two Schulz-Flory distributions, equal polymerization rate of each site but with a different chain termination probability q. The chain transfer probability of site 1 and 2 depends only on the relative hydrogen concentration4, see equations 31 and 32. Cm, the monomer concentration in the polymer, depends on the temperature.

0023.0C

C027.3q

m

H1

2 += (31)

0051.0C

C31.14q

m

H2

2 += (32)

The molecular weight distribution of the polymer product obtained after a polymerization experiment in the internally circulating fluidized bed reactor can be calculated by integration over time and place in the reactor. For our reactor model, we can summarize the molecular weight distributions produced during an integration interval for every compartment, based on the amount of polymer produced in that time interval. The related weight and number averaged molecular weight of the polymer can be calculated by the moments of the distribution:

dj

1j

1nj,wn yMN ∑

∞

=

−= (33)

1

2w N

NM = ,

0

1n N

NM =

(34)

Modeling results The number of mixing cells in the annulus and draft tube (Na and Nd) remain to be determined or fixed. It was found from preliminary calculations that 20 compartments in the annulus and draft tube were sufficient (Na=Nd=20) to simulate the plug flow behavior in these two sections. Increasing the number of compartments in these two sections does not affect the predicted results. Other input data are summarized in table 3.

CHAPTER 5

113

Table 3: Input data for reactor model.

Constant Value Unit Constant Value Unit ap

1.8· 107

m3/(kg.s.(kg/m3)c)

Eact,p

43.9

kJ/mol

ai 9.72· 104 m3/kg.s ∆Hr 2471 kJ/kg Ae 1.96· 10-3 m2 kd0 0.36 1/s Aa 5.03· 10-3 m2 Na 20 - Ad 1.96· 10-3 m2 Nd 20 - bp 2.27· 106 m3/kg.s T∞ 50 °C bi 1.07· 103 1/s εa 0.44 - c 0.3 - εc 0.75 - dw 4· 10-2 m εd 0.75 - Eact,d 17.4 kJ/mol ρs 910 kg/m3

Eact,i 35.6 kJ/mol

The system of differential equations (164 in total) is numerically solved with a Runge-Kutta method. In figure 11 and 12 the model prediction of the temperature profile in the annulus together with the results of the experiments at 25.1 and 42.7 cm/s are presented. The shapes of the temperature curves are comparable with the experimental results. It seems that the rate of deactivation of the catalyst is overpredicted by the model, leading to an under prediction of the temperature at the end of the experiment. The temperature gradient over the length of the annulus is somewhat underpredicted. This may be explained as follows. The model uses the solids circulation measurements described earlier in this paper as input data. These measurements were carried out under non-polymerization conditions with the same polymer powder as has been used as start up powder for the polymerization experiments. However, during the polymerization experiments a low molecular weight polymer is formed which behaves somewhat sticky. This may cause a reduced mobility of the polymer powder inside the reactor leading to reduced circulation rates and higher internal temperature gradients than expected. Furthermore, the model assumes minimum fluidization conditions for the annulus. The solids circulation experiments showed that the annulus is probably below the minimum fluidization velocity. Therefore, the model calculates with a too large upward gas velocity in the annulus, which implies an under prediction of the temperature gradient. Model simulations showed however that the sensitivity of the temperature gradient in the annulus towards the gas velocity in the annulus is small. The input for the solids circulation rate is much more important.

CHAPTER 5

114

In figure 13 Mw, Mn and Mw/Mn of the formed polymer are presented as function of time for the experiment at 25.1 cm/s. This experiment shows the highest temperature gradients (both in time and over the annulus), leading to the highest polydispersity of the final product compared to the experiments at higher gas velocity. As can be seen from figure 13, both Mw and Mn are decreasing in time and Mw/Mn is increasing in time. The polydispersity at the beginning of the experiment is around 2.56, reflecting the two-site model. However, the increase of Mw/Mn is very small indicating that the temperature gradients are too small to see any broadening of the molecular weight distribution.

55

58

61

64

67

70

73

76

0 10 20 30 40 50 60 70 80

t [min]

T [

°C]

Ta(1)

Ta(2)

Ta(3)

Ta(4)

Ta(5)

Ta(6)

Model prediction for Ta(1)


Figure 11: Comparison between experimental result and model prediction for the experiment at 25.1 cm/s.

CHAPTER 5

115

55

57

59

61

63

65

67

0 10 20 30 40 50 60 70 80 90 100 110

t [min]

T [

°C]

Ta(1)

Ta(2)

Ta(3)

Ta(4)

Ta(5)

Ta(6)



Figure 12: Comparison between experimental result and model prediction for the experiment at 42.7 cm/s.

Figure 13: Model prediction of the weight and number averaged molecular weight and polydispersity as a function of time during the experiment at 25.1 cm/s.

0

2

4

6

8

10

12

14

16

18

20

0 10 20 30 40 50 60 70 80

t [min]

Mw a

nd M

n [k

g/m

ol]

2.55

2.56

2.57

2.58

2.59

2.60

2.61

2.62

Mw/M

n [-

]

Mw

Mw/Mn

Mn

CHAPTER 5

116

Injection of hydrogen In figure 14 the normalized temperature profiles obtained from the experiments with injection of hydrogen and the prediction of the model are presented. The model over predicts the temperature increase after hydrogen injection considerably. Moreover, the temperature before hydrogen injection is over predicted. The deactivation observed during the experiments, both before and after the injection of hydrogen, is higher than predicted by the model.

The calculated polydispersity versus time of the formed polymer is shown for each experiment in figure 15. Directly after the injection of hydrogen, the polydispersity starts to increase. The molecular weight distributions of the final polymers are presented in figure 16. It can be seen that the distribution moves from lower average molecular weights (hydrogen injection after 12.5 minutes) to higher average molecular weights (hydrogen injection after 30 minutes). The polymer obtained from the experiments is a mixture of the start-up powder and the low molecular weight polymer produced during the experiment. The molecular weight of the start-up powder is about 300 kg/mol, but with a rather high polydispersity (about 7). The molecular weight distribution of the start-up powder is given in figure 17. The GPC measurements of the mixed powders (start-up material and own produced polymer) were not very reproducible due to the low content of own produced polymer in the samples obtained from the experiments (less than 10%). For one experiment (hydrogen injection

Figure 14: Comparison between experimental results and model predictions for the experiments with injection of hydrogen.

-2

-1

0

1

2

3

4

5

6

7

8

0 10 20 30 40 50 60 70

t [min]

∆∆T [

°C]





CHAPTER 5

117

after 24 minutes), a significant broadening at the low molecular tail of the original curve could be detected, see figure 17. A fit of the molecular weight distribution of the start-up powder has been subtracted from this curve in order to isolate the molecular weight distribution of the own produced powder. Figure 18 presents the normalized subtracted distribution together with the prediction of the model for this experiment. Note that the errors of the subtracted distribution are quite large. Under these circumstances the experimental data are quite well represented by the model.

Figure 15: Model prediction of the polydispersity as function of time for the experiments with injection of hydrogen.

2

2.2

2.4

2.6

2.8

3

3.2

3.4

3.6

0 10 20 30 40 50 60 70 80

t [min]

Pol

ydis

pers

ity

[-]

.





CHAPTER 5

118

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0 1 2 3 4 5 6 7 8

log Mw

Y

Mixed powder

Fit of high molecular part

Start-up powder

Figure 17: MWD of the start-up powder, MWD of polymer obtained from an experiment with injection of hydrogen (start-up powder and produced polymer), MWD of start-up powder used for subtraction.

Figure 16: Model prediction of the MWD of the final polymer for the experiments with injection of hydrogen.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 1 2 3 4 5 6

log Mw

Y





CHAPTER 5

119

Discussion The developed reactor system enables us to control the solids mixing required to study the polymerization of olefins in a small-scale fluidized bed reactor. The temperature gradients caused by catalyst segregation observed in a previous paper2 have not been found anymore. Moreover, elutriation of fines has not been observed despite the high gas velocities applied in the draft tube. The slugs rising through the draft tube (containing the whole particle size distribution) probably take the smaller particles to the annulus, thereby preventing serious elutriation. The 3 zones in the reactor with different heat transport properties may help to systematically broaden the molecular weight distribution. However, the temperature gradients observed during a series of experiments in which the gas velocity was varied, were too small to see a significant broadening. The presented model uses the measured solids circulation characteristics as input data. If the model is to be used for scale-up, a model for the solids circulation rate should be included, which describes the circulation rate as function of the gas velocity, reactor geometry, etc. Furthermore, the model should be improved by implementing the

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 1 2 3 4 5 6

log Mw

Y

Model prediction

Subtracted and normalized GPC curve

Figure 18: Model prediction of the MWD of a polymer from an experiment with injection of hydrogen and the distribution obtained after subtraction of the distribution of the start-up powder.

CHAPTER 5

120

dynamics of absorption and desorption of monomer in the polymer particles with related energy effects, thereby respecting the (changing) particle size distribution. Conclusion A small-scale fluidized bed reactor has been equipped with a draft-tube and conical bottom section to control the solids mixing inside the reactor. The solids circulation rate has been measured at different gas densities and gas velocities. It appeared that the particles in the draft tube and annulus part of the reactor move in plug flow. Slugging takes place in the draft tube and a moving bed flow in the annulus. Segregation and elutriation of the fine particles, which was a dominant effect without the draft tube, could not be detected anymore during operation with the draft tube. Semi-batch polymerizations have been executed at different gas velocities, thereby varying the temperature gradients in the reactor from important to almost absent. Hydrogen was injected at different moments during the experiment in a special series of experiments. Directly after the injection of hydrogen, temperatures are increasing due to reactivation of dormant sites causing a higher polymerization rate. Because the relative temperature increase after hydrogen injection is independent of the injection time, one may conclude that the relative concentration of dormant sites is constant during the experiment. A compartment model has been developed, which is able to describe the general instationary temperature profiles in the reactor and related molecular weight distribution of the polymer. The underprediction of the temperature gradient over the annulus is probably due to the inaccuracies in the input data for the solids circulation rate. The solids circulation rates have been measured under non-polymerization conditions while during polymerization, a sticky polymer is formed which limits the mobility of the powder. The prediction of the model for the molecular weight distribution could be checked for a single experiment. In this case the prediction of the molecular weight distribution is in reasonable agreement with the experimental result. Acknowledgement - This work has been funded by BRITE-EURAM Project CATAPOL (BE 96-3022).

The author wishes to thank DSM Research for the GPC measurements and materials they provided. We

greatly acknowledge the technical team of the High Pressure Laboratories. N.F. Geijsen, M. Poortenga,

J.M. Rutten and M.J.M. Hattink are acknowledged for their contribution in the experimental part.

CHAPTER 5

121

Notation A Cross-sectional area m2 C Amount of non-activated catalyst kg C* Amount of activated catalyst kg

Cm Monomer concentration in polymer kg/m3 Cp Heat capacity J/kg⋅K d Thickness m Eact,d Activation energy for deactivation J/mol Eact,i Activation energy for initiation J/mol Eact,p Activation energy for propagation J/mol ∆Hr Heat of reaction J/kg j Chain length kd Reaction rate constant for deactivation sec-1 ki Reaction rate constant for initiation sec-1 kp Reaction rate constant for propagation m3/kg.s Mn Number averaged molecular weight g/mol Mw Weight averaged molecular weight g/mol

mm Mass fraction of monomer - N Total number of compartments - Nn nth moment of the MWD - p Pressure bar q Chain transfer probability - R Gas constant J/mol.K Rp Reaction rate kg/s t Time s T Temperature K T∞ Temperature in air thermostat K u Velocity m/s Umf Minimum fluidization velocity m/s Ue Inlet gas velocity m/s V Volume m3 X Mass fraction -

djy Density function of the instantaneous MWD -

Greek α Heat transfer coefficient W/m2.K ε Void fraction - ρ Density kg/m3 φm Mass flow kg/s

CHAPTER 5

122

Sub- or superscripts a Annulus av Average value c Cone comp Compartment d Draft tube e Conditions at reactor entry g Gas i Compartment number m Monomer p Polymer r Reactor s Solids w Wall 0 Initial value Abbreviations MAO Methylaluminoxane MWD Molecular weight distribution TIBA Tri-isobuthylaluminium Literature 1. Matsen, J.M., Powder Techn., 88, 237-244, (1996) 2. Meier, G.B., Roos, P., Weickert, G., van Swaaij, W.P.M., submitted to AIChE J.,

(2000) 3. US 5,698,642: Montell technology, (1997) 4. Meier, G.B., Weickert, G., van Swaaij, W.P.M., submitted to J. Pol. Sci., (2000) 5. Ji, H., Tsutsumi, A., Yoshida, K., J. Chem. Eng. Japan, 31(5), 842-845, (1998) 6. Alappat, B., Rane, V.C., Ind. J. Eng. Mat. Sci., 2, 113-117, (1995) 7. Song, B.H., Kim, Y.T., Kim, S.D., Chem. Eng. J., 68, 115-122, (1997) 8. Berruti, F., Muir, J.R., Behie, L.A., Can. J. Chem. Eng., 66, 919-923, (1988) 9. Yang, W.C., Keairns, D., Can. J. Chem. Eng., 61, 349-355, (1983) 10. Floyd, S., Choi, K.Y., Taylor, T.W., Ray, W.H., J. Appl. Polym. Sci., 32, 2935-2960,

(1986) 11. Meier, G.B., Weickert, G., van Swaaij, W.P.M., submitted to J. Appl. Pol. Sci.,

(2000) 12. Daubert, T.E., Danner, R.P., “Data compilation of pure compounds”, DIPPR Project,

AIChE, New York, (1985)

SAMENVATTING

123

Samenvatting Dankzij de voortdurende ontwikkeling van katalysatoren en processen zijn polyolefinen belangrijke plastics geworden. Polyolefinen kunnen tegenwoordig tegen lage productiekosten en met zeer variërende producteigenschappen geproduceerd worden. Polypropyleen wordt nu voor het grootse deel in gasvormig en/of vloeibaar propyleen gemaakt. Het merendeel van het kinetiekonderzoek wordt echter in de slurry-fase uitgevoerd. Bijzonder schaars is het onderzoek dat bij industrieel relevante procescondities wordt uitgevoerd, d.w.z. bij hoge druk en relatief hoge temperatuur. De bekendste industriële technologie voor de gasfase polymerisatie van olefinen is de gefluïdiseerde bed reactor. Dit proces wordt bedreven bij een dru k van 10 – 30 bar. De conversies per gang door de reactor worden laag gehouden, 1- 3%, om concentratie en temperatuur gradiënten, die beide een directe invloed op de polymeereigenschappen hebben, te voorkomen. De deeltjes die in 1 tot 3 uur verblijftijd worden verkregen hebben een brede deeltjesgrootteverdeling. Er zijn geen experimentele gegevens beschikbaar in de open literatuur over dit type reactor m.b.t. olefine polymerisatie. In dit proefschrift worden de resultaten gepresenteerd die onder andere verkregen zijn met een gemodificeerde fluïd bed reactor. Hiermee worden verschillende aspecten bestudeerd van de gasfase polymerisatie van propyleen met een heterogene metalloceen katalysator bij condities die de industriële condities benaderen of daar sterk vanaf wijken. Om dit te kunnen bestuderen is een experimenteel gevalideerd kinetisch model een vereiste. Dit kinetische model, afgeleid van experimenten bij isotherme en isobare condities, kan dan gebruikt worden om de niet isotherme experimenten in het fluïd bed te beschrijven met de daarbij behorende beschrijving van de molecuulgewichtsverdeling van het polymeer. Andere aspecten die in dit proefschrift aan de orde komen zijn de elektrostatische oplading van deeltjes, de deeltjes menging en de elutriatie van de kleine deeltjes. Polymerisatie kinetiek In zowel gasvormig als vloeibaar propyleen is propyleen gepolymeriseerd in geroerde semi-batch reactoren met rac-Me2Si[Ind]2ZrCl2 / MAO / SiO2(Grace) als metalloceen katalysator. Gasfase polymerisaties zijn uitgevoerd bij temperaturen tussen 40 en 80°C en drukken tussen 5 en 25 bar. Polymerisaties in vloeibaar propyleen zijn uitgevoerd bij temperaturen tussen 40 en 80°C met waterstofconcentraties tussen 0 en 2.2% in de gasfase boven de vloeistof. De kinetiek is beschreven met een vereenvoudigd model dat een eerste orde reactie aanneemt m.b.t. het aantal actieve plaatsen en de concentratie van het monomeer in het polymeer. De propyleenconcentratie in het amorfe gedeelte van het

SAMENVATTING

124

semi-kristallijne polymeer, d.w.z. bij het actieve centrum van de katalysator, is experimenteel bepaald. Bij lage drukken kan de wet van Henry worden gebruikt om de concentratie te beschrijven, bij hogere drukken wordt de Flory-Huggins vergelijking gebruikt. Dalende waarden voor de Flory-Huggins interactie parameter zijn gevonden met stijgende temperaturen. De sorptie gegevens zijn gebruikt om de relatieve reactiesnelheden in gasvormig en vloeibaar propyleen te vergelijken. Lagere reactiesnelheden zijn gevonden voor de gasfase polymerisaties. De gevonden activeringsenergieën uit experimenten in beide fasen zijn ongeveer gelijk. In een andere serie van experimenten is propyleen in de gasfase gepolymeriseerd bij verschillende temperaturen, drukken en waterstofconcentraties met dezelfde katalysator, echter op een andere silica drager. De reactiesnelheden zijn beschreven met een kinetisch model dat zowel de activering als de deactivering van de katalysator beschrijft. Bij hogere temperaturen, drukken en waterstofconcentraties zijn relatief lagere opbrengsten gevonden die geïnterpreteerd zouden kunnen worden als het gevolg van een thermische runaway op deeltjes schaal. De polymeer monsters zijn met behulp van een GPC analyse op de molecuulgewichtsverdeling geanalyseerd. Deze verdeling is beschreven met een model gebaseerd op twee actieve centra. Bij constante temperatuur blijken de ketenoverdrachtswaarschijnlijkheden van het eerste en tweede centrum alleen afhankelijk van de waterstofconcentratie gedeeld door de monomeerconcentratie. Polymerisatie in de FBR Op het hogedruk laboratorium van de Universiteit Twente is een fluïd bed reactor op kleine schaal gebouwd voor de katalytische polymerisatie van olefinen onder druk. De menging en segregatie van deeltjes zijn bestudeerd. Het bleek dat de menging relatief snel is vergeleken met de verblijftijd van de katalysator in een polymeriserend systeem. De kleine deeltjes in de reactor accumuleren echter in het bovenste gedeelte van het bed. Verder werd een laag van fijne deeltjes met een in de tijd groeiende dikte op de wand aangetroffen, veroorzaakt door electrostatische oplading. Deze deeltjes werden weer gemengd na een injectie van een anti-statisch poeder. Semi-batch polymerisaties zijn uitgevoerd bij verschillende fluïdisatiesnelheden. De gevonden vertic ale temperatuurgradiënten blijken hoofdzakelijk door de segregatie van de katalysator veroorzaakt te worden. Vooral bij lage fluïdisatiesnelheden bleken onder reactieve omstandigheden de menging en segregatie van deeltjes anders te zijn vergeleken met niet reactieve omstandigheden door het verschil in de interactie tussen de deeltjes. Ondanks het feit dat gefluïdiseerde bedden nauwelijks opgeschaald kunnen worden met behulp van experimenten die uitgevoerd zijn in reactoren met een diameter kleiner dan 30 cm, kunnen kleine fluïd bed reactoren toch nuttig zijn om electrostatische oplading en de gevolgen van onvolledige menging te bestuderen. De gevonden balans tussen segregatie en menging mag echter niet vergeleken worden met de situatie op industriële schaal. Het

SAMENVATTING

125

is daarom noodzakelijk om de verticale menging in deze kleine reactor te verbeteren om representatiever te kunnen zijn. De laboratoriumopstelling is daartoe uitgerust met een “draft tube” en een conus om de verticale menging te controleren. In deze aangepaste opstelling worden de deeltjes in de draft tube met behulp van een relatief hoge gassnelheid omhoog getransporteerd. Vervolgens bewegen de deeltjes als een “moving bed” in de annulus omlaag, om daarna weer de conus binnen te treden. De circulatiesnelheid van het poeder neemt versneld toe met toenemende gassnelheid. Vergeleken met experimenten zonder draft tube, werd een sterke afname van elutriatie waargenomen. Semi-batch propyleen polymerisaties zijn uitgevoerd bij verhoogde druk. Bij deze experimenten lag de nadruk op het creëren van verschillende zones in de reactor met verschillende temperaturen. De temperatuurgradiënten in de reactor kunnen gevarieerd worden met behulp van de circulatiesnelheid van het poeder. Een injectie van waterstof tijdens een experiment veroorzaakte een instantane stijging van de polymerisatiesnelheid, waarschijnlijk door het activeren van “dormant sites”. De irreversibele deactivatiesnelheid van dormant en actieve centra lijken identiek te zijn. Verder blijkt waterstof zeer effectief te zijn voor het verbreden van de molecuulgewichtsverdeling. Een reactormodel is ontwikkeld om de temperatuur- en concentratieprofielen in de reactor te kunnen beschrijven met de daarbij behorende molecuulgewichtsverdeling van het ontstane polymeer.

SAMENVATTING

126

DANKWOORD

127

Dankwoord Met het afronden van dit proefschrift kan ik met veel genoegen terugkijken op een zeer interessante en leerzame periode. Een ieder die tot de totstandkoming van dit proefschrift heeft bijgedragen wil ik hartelijk bedanken voor de vaak belangeloze inzet. Een aantal van hen wil ik echter niet ongenoemd laten. Tijdens mijn afstudeerproject was mij de enthousiaste werkwijze van de polymerisatie club van IPP niet ontgaan. Ik ben dan ook professor Westerterp zeer erkentelijk voor de geboden mogelijkheid een promotieonderzoek uit te voeren in zijn vakgroep. De dagelijkse begeleiding lag in handen van professor Guenter Weickert. Grote waardering heb ik voor zijn enthousiasme en passie voor het vak. De talrijke discussies, kritische blik gecombineerd met zijn vele experimentele en modelmatige ideeën zijn van onschatbare waarde geweest. Professor van Swaaij ben ik zeer dankbaar voor zijn ondersteuning in de eindfase van het project. Veel dank ben ik hem dan ook verschuldigd voor de bijdrage aan mijn proefschrift via discussies en zeer snelle correctierondes. De twee belangrijkste opstellingen waarvan ik op het Hogedruk Laboratorium gebruik heb gemaakt waren deels al gedurende de promotieprojecten van Job Jan Samson en Peter Roos opgebouwd. Karst van Bree ben ik zeer dankbaar voor de vakkundige wijze waarop de talloze wijzigingen zijn doorgevoerd. Ook de andere leden van het ondersteunende team op het lab ben ik zeer erkentelijk: Fred, Geert, Gert en Arie bedankt! Een groot gedeelte van het werk dat in dit proefschrift beschreven staat is in het kader van doctoraal opdrachten uitgevoerd door studenten. Bart Wijers, Nathan Kuper, Rik Uiterwijk, Robert Smit, Jeroen Dunnewijk, Japke ten Have, Oskar Slotboom, Michiel Bergstra en Wilma Aanstoot ben ik zeer dankbaar voor hun bijdrage aan de kinetiek gerelateerde onderwerpen. Arjan van Klaveren, Marc Poortenga, Niels Geijsen, Marjorie Hattink en Jasper Rutten wil ik danken voor hun bijdrage aan de onderwerpen aangaande de fluid bed reactor en Sander Breeveld voor het modelleren van een enkel groeiend polymeer deeltje. Alle partners binnen het Brite Euram project Catapol wil ik bedanken voor de nuttige bijeenkomsten waar de voortgang van het project werd besproken. Deze bijeenkomsten, meestal gehouden op zeer aangename locaties, hebben ook zeker mijn blikveld verbreed. DSM ben ik dank verschuldigd omdat zij dit project mede financieel mogelijk hebben gemaakt. Ik heb de discussies met Stan Mutsers, Tjaart Molenkamp, Gerwin Wijsman,

DANKWOORD

128

Gerard Krooshof en Denise Bakker zeer gewaardeerd. Ilco de Vries wil ik graag bedanken voor de GPC analyses en de hulp bij de interpretatie van de resultaten. Bestellingen en reparaties werden uitgevoerd door het Techno Center CT. Hierbij hebben met name Wim Platvoet, Jan Jagt, Bert Kamp en Jan Heezen bijgedragen. Alle leden van de Doctoraal commissies van mijn afstudeerders ben ik erkentelijk voor de tijd die zij hierin hebben gestoken. In het bijzonder wil ik Jochem bedanken voor de gezellige tijd in ons koffiehok met de bijbehorende discussies, die ons enthousiasme voor PP tot grote hoogte heeft doen stijgen. Mijn (oud) collega’s wil ik graag bedanken voor de gezellige sfeer tijdens het werk en onze reizen. Yvonne Bruggert en Gery Stratingh zorgden voor een prima afwikkeling van de administratieve zaken. De paranimfen Jochem “dikke” Pater en Michiel “Waterreus” Bergsma wil ik hierbij bedanken voor hun inzet en talloze voetbal discussies. Een dag die ik niet snel vergeten zal is 2 december 1998. De explosie die op die dag in bunker 7 van het HDL plaatsvond was een, voor sommigen letterlijke, klap in het gezicht. Gelukkig hebben zich geen ernstige persoonlijke ongelukken voorgedaan, maar de impact op allerlei andere vlakken was enorm. Ik heb hier, en met mij waarschijnlijk vele anderen, enorm veel van geleerd. Ik wil graag iedereen die bij de nasleep van dit ongeval betrokken was enorm bedanken voor de goede en zorgvuldige afhandeling. Ik wil bij deze graag mijn ouders, familieleden en vrienden bedanken voor de getoonde interesse en steun. Tot slot wil ik Dynah bedanken voor haar steun, begrip en vertrouwen gedurende de afgelopen jaren.

LEVENSLOOP

129

Levensloop Gerben Meier werd op 22 april 1972 geboren in Stad Delden. Na de lagere school bezocht hij het Twickel College te Hengelo waar hij in juni 1990 het VWO diploma behaalde. In augustus van datzelfde jaar begon hij met de studie Chemische Technologie aan de Universiteit Twente. Tijdens zijn studie liep hij stage bij Akzo Nobel Engineering Obernburg in Duitsland. In juni 1996 sloot hij de opleiding af binnen de vakgroep Industriële Processen en Producten met een onderzoek naar de gasfase polymerisatie van ethyleen met behulp van heterogene metalloceen katalysatoren. In juli 1996 trad hij vervolgens in dienst als assistent in opleiding en verrichtte onder leiding van prof. dr. ir. K.R. Westerterp en prof. dr. G. Weickert het onderzoek dat in dit proefschrift beschreven staat. Na het emeritaat van prof. Westerterp in 1998 werd de supervisie in 1999 overgenomen door prof. dr. ir. W.P.M. van Swaaij. Sinds augustus 2000 is Gerben Meier werkzaam bij Montell te Ferrara in Italië.

Date post:	30-Apr-2020
Category:	Documents
Upload:	others
View:	43 times
Download:	3 times

Fluidized Bed Reactor for Catalytic Olefin Polymerization · FLUIDIZED BED REACTOR FOR CATALYTIC...

Documents