Reports: ND952789-ND9: Realizing the Selectivity of Exothermic Partial Oxidation Reactions

Anthony G. Dixon, Worcester Polytechnic Institute

1-Introduction Multitubular fixed bed reactors are widely used in the chemical industry, especially for partial oxidation reactions. Due to the fast and highly exothermic nature of these reactions strong species concentration and temperature gradients exist in the reactor, leading to loss of selectivity. A greenhouse gas, carbon dioxide, is produced as a by-product of the non-selective side reactions. A CFD integrated multiscale method is developed for transport phenomena in both the solid and fluid phases, for ethylene and n-butane oxidation reactions. CFD is coupled with microkinetic models and aniotropic pellet diffusivity to study selectivity and catalytic activity loss inside the reactor, and their connections to elementary reaction steps, temperature, and catalyst structure. To date, no CFD study has coupled detailed reaction kinetics inside the catalyst particles with transport phenomena in the reactor. 2-Coupling microkinetics with CFD Two different methods were developed to couple microkinetic models with CFD. Initially, the proposed microkinetic models were reduced to a single general reaction rate expression model. Later, a method was developed and implemented to couple the full microkinetic model with CFD. The procedure of developing a general rate expression for a microkinetic model is  straightforward. The approach includes several kinetic and mathematical assumptions to overcome the complexity of the existing models. A reduced model for ethylene oxidation (EO) based on a literature microkinetic model was obtained. Furthermore, literature reduced models developed for n-butane oxidation were used for the maleic anhydride reactor. To our knowledge these are the most comprehensive kinetic models existing in the literature for the reactions of interest. For the full microkinetic model of EO, initially a sensitivity analysis was performed to evaluate the crucial parameters to the model. Next, the model was solved over a wide range of temperatures (T) and species partial pressures (Pi). Reaction rates were then mapped into quadratic splines, and spline coefficients were stored in a UDF as 1-D arrays. Finally, as CFD simulations were carried out, for each computational cell, splines were constructed based on the corresponding T and Pi  to evaluate reaction rates. 3-CFD simulations Two models are used for our studies as shown in Figure-1.

Computational parameters for different cases are summarized in Tables 1-3.

  3-1-Ethylene oxidation Based on a reduced microkinetic model, simulations were carried out for a randomly packed bed of 120 spheres (Figure 2).

 



It is shown that coverages were strongly dependent on species mass fraction and temperature gradient inside the particles. Furthermore, due to high reaction rate on the outer surface of the particles and diffusion limitations, oxygen mass fraction was highly reduced inside the particles. High temperature and low oxygen content caused further combustion of EO (product). This happened when the elementary step that was responsible for EO production went in the reverse direction (Figure 2b).  Reduced microkinetic models (often used in reaction engineering) are usually developed based on assumptions such as quasi-equilibrium (QE) and rate determining step. As mentioned, oxygen partial pressure was reduced significantly inside some of the particles. This raises an important question: are all of the elementary steps that were assumed to be at QE, still equilibrated? To answer this question we developed the second method that integrated CFD and microkinetics without any assumption about elementary steps. As shown in Figure 3, there were significant differences (temperature and concentration profiles inside the particles) between two models. This highlighted the importance of the second approach. 3-2-Butane oxidation Simulations were carried out on a 9-sphere model for n-butane oxidation to maleic anhydride. Significant temperature increase was observed between different rows of particles. Oxygen surface coverages were evaluated inside the particles. An important connection between surface site distribution of oxygen and selectivity loss (maleic combustion rate) was observed. As can be seen in Figure 4-(f) particle shapes and reactor configuration can contribute to selectivity loss. Simulations on a randomly packed bed of 120 spheres are being carried out, to investigate the further impact of the above parameters on selectivity loss in maleic production.


4-Improving the Effective Pellet Diffusivity In previous studies, it was assumed that the effective diffusivity of species inside the catalyst particles is constant. A simplification of the Dusty Gas Model derived by Hite and Jackson (1977) was used to obtain a Fickian form to model diffusion inside the solid particles.
 

(1)

 

(2)

  It is assumed here that the ratio of fluxes is given by the ratio of stoichiometric coefficients. However, this is a questionable assumption in 3D simulations. To examine the validity of that assumption, an anisotropic diffusivity tensor user-defined function is made and implemented into the CFD method. Hite and Jackson’s model is used without the need for any further assumptions.
 

(3)

  where α = x, y, z and i = species index. Equations (1), (2) and (3) yield the diffusivity tensor. The only assumption here is that the pore directions of the catalyst are aligned with principal coordinates of the system. This results in a diagonal matrix with xx, yy, zz components. 


The developed method is implemented in a 9-spherical particle model for ethylene oxidation, and results are compared with the isotropic constant diffusivity model. It is observed that the reaction rate and species concentration profiles inside the particles deviate from those calculated by the previous method. Therefore, effects of species fluxes and pellet structure cannot be neglected in modeling the reaction inside the solid particle.   5- Impact of Research A CFD integrated multiscale method is developed which has improved the previous method based on global kinetics by adding detailed surface microkinetics and anisotropic effective diffusivity. By investigating the effects of elementary reaction steps and transport phenomena on each other can address the selectivity loss, and go one step closer to using CFD as an ultimate tool for optimized design of catalytic reactors. The research has allowed the PI to broaden his CFD research into microkinetics, and enabling international collaborations with other research groups. The PhD student Behnam Partopour was able to focus on research for the first two years of his graduate study.