Heat and Mass Transfer Effects in Multistage Polymerization

José Bonilla-Cruz , Tania Lara-Ceniceros , Enrique Saldívar-Guerra , Enrique ... Antônio G. Mattos Neto , Márcio Nele , Príamo A. Melo , José Carlos P...
0 downloads 0 Views 2MB Size
Ind. Eng. Chem. Res. 1995,34, 3466-3480

3466

Heat and Mass Transfer Effects in Multistage Polymerization Processes: Impact Polypropylene Jon k Debling and W.Harmon Ray* Department of Chemical Engineering, University of Wisconsin-Madison, Madison, Wisconsin 53706

Polyolefin products are often produced over solid catalysts in multistage reactor processes. As the polymerization proceeds, polymer encapsulates the original catalyst particle and expands with polymer yield. Ultimately a polymer particle 10-50 times as large as the original catalyst particle is produced. I n each reactor a completely different polymer product may be produced adding to the previously formed polymer from prior reactors. During this time the morphology of the polymer particle may change dramatically and have a pronounced effect on the reactionl diffusion process in the catalyst pellet. In this paper we present a model of particle growth for multilayered “heterophasic”polymer products produced in a series of reactors. Simulation studies for impact polypropylene produced in a loop-fluidized bed process are shown. It is shown that reactor residence time distribution, catalyst deactivation, and particle porosity play a significant role in determining the product quality of impact copolymer. At high copolymer contents, severe diffusion diffusion limitations are predicted, producing a sticky copolymer on the surface of the granule and a propylene-rich copolymer in the interior.

I. Introduction Many important polyolefin products (linear low density polyethylene, high density polyethylene, isotactic polypropylene, impact polypropylene, ethylene-propylene copolymer, etc.) are produced by the polymerization of olefins over solid catalysts such as supported ZieglerNatta, chromium oxide, and recently metallocene based systems. The role of the catalyst system is 2-fold. First, the catalyst must be able to produce a polymer of desired molecular properties with sufficient activity as to remain innocuous in the tinal resin. Second, the catalyst particle must act as a template for the growing polymer particle which will encapsulate the catalyst as polymerization proceeds, ultimately producing a granule 1050 times as large as the initial catalyst. With the onset of polymerization, a typical catalyst quickly shatters into small microfragments consisting of active metal and support. These small microfragments are typically between 0.001 and 5.0 pm diameter. Monomer, chain transfer agent, and cocatalyst must diffuse into the porous polymer particle to the catalyst surface where polymerization occurs via a sequential coordinated chain insertion mechanism on active sites (carbon-transition metal bond) present on the surface of the catalyst active metal. With growth of the polymer, the polymer particle is continuously exchanging mass and heat with its surroundings. The physical process of particle growth involving reactioddiffision in a growing particle is complex but can have a dramatic effect on the properties of the polymer product and on process operation. The requirements of today’s modern olefin polymerization processes place high demands on the catalyst system as indicated in Tables 1 and 2. There have been a number of models proposed for the growth of the catalyst/polymer particle; however, two main models have emerged: the continuous flow model and the multigrain model. They differ in the assumptions made regarding the nature of the polymer particle morphology and its influence on the diffusion process. The continuous flow model assumes that the growing particle is nonporous and that the polymerhatalyst

* Author to whom all correspondence should be addressed. 0888-5885/95/2634-3466$09.00l0

Table 1. Desirable Catalyst Properties for Solid Catalyzed Olefin Polymerizationa high surface area a b high porosity with a large number of cracks evenly distributed throughout the granule C high mechanical strength to withstand breakdown during preparation yet low enough to permit shattering during polymerization free access of monomer throughout the d catalyst particle e homogeneous distribution of the active sites f able to produce highly stereoregular polymer (for PP)to avoid atactic polymer removal high activity to avoid catalyst residue removal g h able to produce polymer with desired properties (Le., MWD, composition, etc.) controlled particle morphology large enough catalyst particle size to avoid the production of fines but small enough to prevent overheating in the reactor ~

~~

~

Galli et al. (1981, 1984); Galli and Ali (1987); Schill and Buchner (1988). (1

Table 2. Desirable Polymer Particle Properties for Solid Catalyzed Olefin Polymerizationa a particle size large enough (200-5000 pm) for good powder handling, fluidization; optimum goal, particles large enough to be sold directly from reactor without pelletizing b narrow particle size distribution minimal amount of fines ( 7500 s is due to catalyst deactivation and increasing particle size allowing larger surface area for diffusion. In Figure 14, the volume fraction copolymer across the particle radius is shown for all cases A-C. In cases B and C where the catalyst has remained in the loop reactor longer than in case A, the catalyst has deactivated considerably more. The amount of homopolymer is also much greater. As a result, the copolymer content is much lower and the distribution across the macroparticle much more uniform in both cases than in case A. For case B, the pore volume of the particle is completely filled by copolymer; however this requires over 2 h of polymerization in the copolymer reactor. For case C, however, the particle still contains some voids even after 3 h of polymerization in the copolymer unit (cf. Figure 9). Diffusion limitations are not present. 111.2. Effect of Catalyst Particle Size. In reality, a typical olefin catalyst consists of a distribution of catalyst sizes that can often be described by a log normal relationship shown in eqs 47 and 48. To study the effect of catalyst particle size distribution on copolymer content, simulations were performed for catalyst diameters ranging from 10 to 125pm. This size range is more than adequate to describe a typical supported TiClflgC12 catalyst system used industrially with weight mean size of 50 pm and spread parameter u of 0.25. All simulated catalyst particles were assumed t o have the same activity and kinetic parameters.

1 1 P(dcat)= -

)

-(ln dcat- In JcaJ2

(47) (48)

In Figure 15 steady state results are shown for the case where diffusion limitations are expected, i.e., 1h in the loop reactor and 2 h in the copolymer reactor. As evident in Figure 15, the total ethylene content in the particle (expressed as mole fraction in the total resin) is a function of the initial catalyst diameter with higher ethylene contents originating from the smaller catalyst pellets. This can be explained by the following. At equal residence times in the homopolymer reactor, each catalyst particle exits with the same yield (g of polypro-

-1

Weight Mean Diameter

0

Particle radial position

Figure 14. Volume fraction copolymer across the particle radius for cases A-C.

0.15

0

I

0.8

t

E

:

. -0

20

40

60

80

100

120

140

Catalyst Diameter ( microns )

Figure 15. Catalyst size distribution and ethylene content as a function of catalyst size: 1h in loop, 2 h in fluidized bed reactor.

0

0.2

0.4

0.6

0.8

1

Particle radial position

Figure 16. Effect of catalyst deactivation on copolymer content: 2 h in loop, 2 h in fluidized bed reactor. Catalyst diameter = 50 ium.

pylene/g of catalyst), active sites, and homopolymer porosity. Thus,, the homopolymer granules have the same capacity (as given by the remaining porosity) to contain copolymer. When the pores of the homopolymer particle are filled with copolymer, the onset of diffusion limitations begins. However, given the same activity, larger diffusion resistances are expected for larger catalyst sizes. 111.3. Effect of Catalyst Deactivation. The effect of catalyst deactivation on copolymer content across the polymer particle is illustrated in Figure 16 for various spontaneous deactivation rate constants. Simulations are shown for a catalyst of 50 pm polymerizing for 2 h in the loop reactor and 2 h in the fluidized bed. For fixed residence times, increasing gradients in copolymer content across the particle are seen as the catalyst deactivation rate decreases. With the use of the base conditions indicated in Table 5 with deactivation rate constant k d = 8.6 x s-l, the composition of copolymer is uniform across the particle at approximately 20 wt %. However, with no catalyst deactivation the content of copolymer increases from -20 wt % in the center of the granule to over 30 wt % on the surface. At high deactivation rates (kd = 17.2 x s-l), copolymer content is not only uniform across the particle but almost half the level observed in the base simulation. The role of catalyst deactivation can be easily understood by understanding the kinetic and physical limita-

Ind. Eng. Chem. Res., Vol. 34, No. 10, 1995 3477 tions of the catalyst system. In eq 38 a correlation for microparticle compressibility as a function of homopolymer yield was presented. By use of eqs 38,41, and 42, an expression for homopolymer void fraction (eh) as a function of homopolymer yield (yh)can be found for the simulations where z = 1 - 0.00349(# - 1):

This value can be a considered a physical limit for incorporating copolymer into the homopolymer granule, representing the approximate amount of copolymer that would fill all of the macroparticle pore volume and mark the onset of strong monomer diffusion limitations. Thus, a t this point the weight fraction copolymer in the polymer product is given by

-Porosity limit ..

h

--

.

-

' ~

. kd=0.0000215s-l

o

- .kd=O.OOOO43 S-l

.O

. ' kd = O.oooO86 S-1

0

- .kd = O.OOO172 S-l

10000 20000 3oooO 40000 50000 6oooO Homopolymer yield ( g/g catalyst )

0

Figure 17. Effect of homopolymer yield on maximum copolymer incorporation. 1 ...................

0.8

. . .o . . Deactivation limit kd = O.oooO86 s-1

The yield of homopolymer from the first reactor (yh)can also be related t o the residence time in the loop reactor (zh),deactivation rate constant (kt),and initial activity (R!o)by

Since the active sites are continuously deactivating with time in each reactor,

c" = ci eXp(-kdt)

(52)

the maximum yield of copolymer (YeMax) can be expressed as a function of the residence time in the loop reactor:

(53) = 5.3591 g/(gs) With the use of eqs 51-53 with and RiO= 2.60194 g/(gs)Figure 17 is developed showing maximum copolymer content as a function of homopolymer yield. The maximum content as limited by homopolymer porosity given by eq 50 is plotted as a solid line, and the maximum content as limited by catalyst deactivation by eqs 51-53 is shown as dotted lines. At low yields of homopolymer the catalyst particle is limited in its ability to incorporate copolymer by particle porosity rather than kinetic activity. At high yields, the catalyst is expected to be limited by kinetics. The transition from diffusion limited t o kinetically limited occurs at the intersection of the two curves. Under most conditions catalyst activity is more likely to govern copolymer content than porosity and diffusion. The results of Figure 16 are then easy to explain. For slow catalyst deactivation rates, the particles become diffusion limited causing a gradient in copolymer composition from the surface to the center of the granule. Catalysts that deactivate rapidly do not become diffu-

0

loo00

2oooO

3oooO

40000

Time ( s )

Figure 18. Effect of residence time in the loop reactor on maximum copolymer incorporation.

sion limited, and smooth profiles of copolymer content across the particle are observed. The effects of catalyst deactivation and residence time in the loop reactor are shown in Figure 18. The results suggest that catalyst particles that exit the loop reactor after less than approximately 2 h are limited in their ability to incorporate copolymer by diffusion. Particles with residence times greater than 2 h are limited by catalyst activity. This explains the results found in section 111.1.

N.Discussion It has been shown that the morphological development of impact polypropylene can have a strong effect on the polymerization process. The formation of rubbery copolymer within the homopolymer matrix decreases granule porosity and reduces monomer diffusivity. Normally, in the gas environment, monomer diffusion limitations are not expected. However, once the particle pores are occluded with rubber, significant diffusion resistances can be expected. Ethylene is expected t o be more diffusion limited than propylene because of its higher reactivity, thereby producing a propylene-rich copolymer in the center of the granule. Although not discussed in this paper, the effect of monomer diffusion limitations on polymer molecular weight is not significant.

3478 Ind. Eng. Chem. Res., Vol. 34, No. 10, 1995

The onset of diffusion limitations caused by pore occlusion causes rubbery copolymer to be produced primarily on the surface of the granule. Prior t o this point, the copolymer is expected to be evenly distributed throughout the polymer particle. This has the potential for causing particle sticking and agglomeration. The presence of a rubber segment lying on the surface of the polymer particle has been claimed t o be the source of sticking in gas phase reactors according to Kakugo et al. (1987). The stickiness of the particles is also said to become greater as the rubber content increases but is also catalyst dependent. Galli and Haylock (1991) have suggested that precise control of particle morphology via regulation of the catalyst morphology is essential in controlling particle sticking and proper fluidization. To avoid sticking problems as the level of copolymer increases, Galli and Haylock suggest the Himont catalyst is capable of forming a thick crust of homopolymer on the surface of the catalyst that acts as a protective shield containing the rubbery copolymer. The rubbery phase must then be homogeneously dispersed within the homopolymer matrix of elevated isotacticity and the rubber domain size controlled (-1 pm diameter perfectly amorphous) in order to realize the optimum stiffness to impact resistance balance (Galli and Ali, 1987). Although the total copolymer capacity would still be limited by homopolymer porosity, preventing sticking may allow the reactor unit to operate at higher copolymer contents before sticking problems become too severe. Poisoning the surface of the copolymer granule may also reduce sticking in the gas phase reactor. An important concern in the production of impact polypropylene is the prediction of homopolymer porosity which appears t o determine the capacity of the particle to contain copolymer in the downstream reactor. Often the homopolymer powder bulk density is taken to be an indication of the homopolymer porosity and hence copolymer capacity. In a continuous process, a balance between polymer yield and rubber content must be made. Increasing homopolymer yield reduces the pore volume available to contain the rubbery phase and the number of active sites available for copolymerization. This limits the incorporation of rubber into the particles. Decreasing homopolymer yield allows the production of higher rubber content particles a t the expense of total polymer yield. Reactor residence time distribution also has a significant impact on product uniformity and process operation. Processes that use multiple reactors in series to narrow the residence time distribution of the catalyst particles will produce a product with a more uniform distribution of properties such as copolymer composition distribution and content. Using multiple reactors in series for the homopolymer stage has the particular advantage of reducing the number of catalyst particles that leave the homopolymer stage with short residence times. Particle sticking may be related to particles that have had short residence times in the homopolymer stage and are still very active. Thus the copolymer content of this fraction of particles could become quite extreme. Furthermore, it is probable that these catalyst particles have increased chances for overheating and melting in the gas phase reactor. It is well-known that prepolymerizing (polymerization under mild conditions to increase the size and surface area of the particle) very active or large catalyst particles is used to reduce fracturing and melting of catalyst particles in industrial processes. We suspect that, in general, processes with

narrow reactor residence time distributions may be able to operate a t higher copolymer contents for a given catalyst system, because of the tendency to reduce the very high copolymer content granules that may lead to operational problems in the reactor. The effect of reactor residence time distribution on product composition distribution of impact polypropylene is studied in detail (Zacca et al., 1995a,b; Debling et al., 1995). It is interesting to comment on the role of particle morphology on the polymerization process. One can envision a scenario where a dense nonporous layer of polymer forms on the surface of the polymer particle blocking the diffusion of monomer and reducing the reaction rate in the interior of the particle. Such a condition may occur if the polymerization conditions cause the outer layers of polymer to fuse together. One possible consequence of this would be the formation of hollow particles (which are found industrially). For impact polypropylene, the presence of a rubbery phase on the surface of the particle could also cause a similar effect, blocking monomer access to the interior parts of the granule and limiting rubber incorporation. These effects can be simulated with the current version of the multigrain model and are discussed elsewhere (Debling, 1996).

Nomenclature Indices i = index over diffusing species used in the simulation; i = l-Nspecies (number of monomers = Nmon) k = index over different catalytic site types; k = 1-4 n = index over discretized radial shells along the macroparticle radius; n = l-N, j = index over polymer phases present in the partic1e;j = l-Nphases System or Constant Parameters t = time, s dCat= diameter of original catalyst particle, cm rcat= radius of original catalyst particle, cm V,,, = volume of solid catalyst cm3 of cat r,, = radius of catalyst crystallite, cm [Me] = concentration of active metal in the catalyst, mol of metal/cm3 of catalyst EO = initial void fraction of catalyst particle, cm3 of voids/ cm3 of catalyst