(432e) Kinetic Modeling of Industrial-Scale Ldpe Reactors

Various reactor types are operated for the production of low-density polyethylene (LDPE) to meet customer demands for different application properties. The most widely used reactor types for the high-pressure polymerization of ethylene on industrial scale are tubular reactors and continuous stirred autoclaves. Within these general reactor constructions many different modifications were developed giving a higher flexibility to adjust process conditions, conversion and polymer properties. Multi-chamber autoclaves and multi-injection tubular reactors are operated preferentially to realize various process conditions inside one reactor to generate tailor-made LDPE. Due to their commercial relevance, it is of economic and scientific interest to understand, describe and predict the polymerization behavior of ethylene in dependence of reactor types and operation parameters. In this context mathematic modeling is a powerful tool to discover influence factors on the polymeric microstructure determining the application properties of LDPE. Besides heat balance and mixing simulations the kinetic representation of the full reaction network plays the decisive role in the description of the free-radical polymerization of ethylene. As many different reaction steps with varying rate coefficients are published for this reaction, there is no generally accepted approach postulated. Especially, the conversion of autoclave to tubular models often necessitates an adaption of the kinetic model approaches, because autoclave LDPE reactors exhibit totally different polymeric microstructures than their tubular analogues. Therefore, the selection of the most representative set of kinetic rate coefficients for one process of interest is quite challenging.  

The work to be presented illustrates the development of a kinetic model using the example of the peroxide-initiated polymerization of ethylene in an industrial-scale multi-zone autoclave. Within the software package PREDICI the reaction network as well as reactor geometry and heat balance can be implemented in one model simulating microstructural polymer properties like the molecular weight distributions (MWD), molar mass averages and branching densities [1].

To test the impact of rate coefficients and reaction steps on the modeled microstructure of LDPE and to find the most suitable combination, commonly used kinetic data sets are applied to the considered autoclave model [2]-[4]. The main criterion for the assessment of a certain reaction network is the right simulation of the autoclave-characteristic high-molecular weight shoulder in the MWD. As shown by a comparison of rate sets and their effect on polymer properties, the formulation and quantification of the following reactions can be determined as crucial for the prediction of autoclave-specific MWD:

• intermolecular transfer reactions (low- and high-molecular species)
• β-scission
• terminal double bond propagation

The influence of the first aspect is significant for the correct estimation of the maximum and the dispersity of the MWD, whereas the second point is simply responsible for the degradation of high-molecular weight chains. Mostly, the rate coefficients for these transfer and β-scission reactions are fitted to achieve the best agreement with experimental data (MWD and branching densities) neglecting terminal double bond propagation. But reactions on polymeric double bonds resulting from transfer to α-olefines, β-scission and disproportionation could play a decisive role, especially for the development of a distinctive high-molecular weight shoulder in the MWD. Sensitivity studies on terminal double bond propagation kinetics demonstrate how this additional mechanism affects the polymeric microstructure and how it could be implemented to match experimentally found LDPE properties.

The extended kinetic approach is tested and validated on a number of processes operated in the chosen multi-zone autoclave. Finally, the kinetics are transferred to an industrial tubular reactor model to check the universal applicability of this approach.

[1]   M. Wulkow, Macromol. React. Eng. 2008, 2, 461-494.

[2]   M. Busch, Macromol. Theo. Simul. 2001, 10, 408.

[3]   P. Iedema, M. Wulkow, H. C. J. Hoefsloot, Macromolecules 2000, 33, 7179.

[4]   P. Pladis, C. Kiparissides, Chem. Eng. Sci. 1998, 53, 3315.