Numerical simulation of the kinetics of the selective catalytic reduction of NO by NH3

Numerical simulation of the kinetics of the selective catalytic reduction of NO by NH₃

J. P. Bodanese; M. Santos

Centro de Ensino de São José, Universidade do Vale do Itajaí , 88.122-000, São José, SC, Brazil

ABSTRACT

In this work we study a model of a catalytic reaction among three monomers in order to understand the kinetics of the selective catalytic reduction of NO by ammonia (4NO + 4NH₃ + O₂® 4N₂ + 6H₂O). Our model takes into account the formation of the intermediate species in the global scheme of the reaction. Using the Monte Carlo method we simulated the model on a square lattice and determined its phase diagram.

1 Introduction

In the last years, severe regulations have been imposed in order to control the emissions of the nitrogen oxides from mobile and stationary sources, of these types of molecules. The principal argument to decrease the emission of nitrogen oxides is based upon the fact that these molecules have a hazardous effect on the health of the people as well as the environment [1,2]. In general, the studies to lower the emissions of nitrogen oxides only consider reduction of nitrogen monoxide, because it accounts for approximately 95% of all emitted oxides. However, as the decomposition of NO into N₂ and O₂ is thermodynamically unstable [2], the use of a catalyst is necessary to facilitate and accelerate the reaction. In this reaction, a molecule of nitrogen monoxide is adsorbed on a catalytic surface and it dissociates into O and N atoms. Then two N(O) atoms combine to form N₂ (O₂). Different types of catalysts have been used to decompose nitrogen monoxide, but all can be classified in three groups: noble metals, metallic oxides and zeolites [3]. Unfortunately, any of these catalysts is poisoned by oxygen atoms over a wide range of temperatures. An alternative to prevent the emission of nitrogen monoxides is the selective catalytic reduction (SCR) of these molecules by ammonia [2,4,5], notably when the gases are emitted by stationary sources. In this de-pollution process, NH₃ is used as a reducing agent to produce water and gaseous nitrogen. However, if oxygen is also added to the gaseous mixture that is adsorbed in the catalytic surface, conversion to N₂ and H₂O is enhanced [1]. Experimental and theoretical studies have been used in order to determine the kinetics of the reduction of nitrogen monoxide by ammonia on catalytic surfaces [6-8]. In both cases, there are some controversies, relative to the mechanism by which the NO molecules are decomposed. In order to provide information about the mechanisms concerned in the kinetics of the selective catalytic reduction of nitrogen monoxide by NH₃, we have studied, through Monte Carlo simulations [9,10], some possible paths for the overall reaction that occurs on the catalytic surface. Our approach is based on the Ziff, Gulari and Barshad model (ZGB model)[11] for the irreversible oxidation of carbon monoxide on a catalytic surface. Besides, as the gaseous reservoir contain three types of molecules (O₂ , NH₃ and NO), we also used the ideas proposed by Schmidt and Santos [12] to study the poisoning of a catalytic surface when a third nonreactive monomer is present in the gaseous reservoir.

Our paper is organized as follows: in section II we present some possible paths for the reduction of NO by NH₃ in presence of O₂ as well as numerical results obtained in each case. In addition, we describe some details of the simulations. As we will see, a small modification in the reaction mechanism changes completely the dynamical behavior of the model. Finally, in section III, we address our main conclusions.

2 Models and numerical results

The process in which ammonia NH₃ is used to decompose nitrogen monoxide NO, on a catalytic surface to produce N₂ and water is called by selective catalytic reduction (SCR). The overall reaction for the reduction of NO in the presence of oxygen and ammonia is given by [3,9]

The first problem that appears in the study of this reaction is related to the mechanism that control the reaction scheme. In fact, several kinetic studies have been reported where the rate equations for this reaction is based on either the Eley-Rideal (ER)[13,14] or Langmuir-Hinshelwood (LH)[15,16] mechanisms. More recently, in an extensive review about the current state of research on NO_x catalysis, Fritz and Pitchon [3] presented two possible reaction schemes for nitrogen monoxide reduction by NH₃. The ER mechanism involves a NO molecule in the gas phase directly reacting with an ammonia-adsorbed species, that is,

where the index a indicates that the molecules are adsorbed on the surface, g that the molecules are in the gas-phase and E indicates an empty site. On the other hand, for the LH mechanism, the particles of nitrogen monoxide must also be adsorbed on the catalytic surface to react with ammonia. In this case we have

We observe that, in both reaction schemes, the catalytic surface is poisoned by the intermediate species H. To regenerate the surface, the oxygen gas is added to the gaseous reservoir, and two new reactions can occur on the catalytic surface

This last reaction occurs by two steps

Here, we note the presence of other intermediate species. We assume that the reactions among the adsorbed species are instantaneous, and that the rates of collisions of the gaseous oxygen, nitrogen monoxide and ammonia with the catalytic surface, here denoted by Y_O, Y_NO, and YNH₃, respectively, depend on their mole fractions in the gas phase. They are related by

In order to perform Monte Carlo simulations for the two mechanisms (ER and LH) we use a square lattice of linear size L with periodic boundary conditions. The unit of time in the simulations is the Monte Carlo step (MCs), in which N = L² adsorption steps occur. We start our simulation for the ER mechanism with an empty lattice, for fixed values of YNH₃ and Y_NO. At random, we choose a site on the catalytic surface and a particle (NO, NH₃, or O₂) to react on it. If NO is chosen, and the site is occupied by ammonia, the two molecules react to yield the intermediate species H on that site. After this, we scan its neighborhood in order to verify the possible reactions of H with OH and O, respectively. Otherwise, if the chosen site is empty or occupied by any other species, the trial ends. If the chosen particle is ammonia the simulation proceeds as follows: first we verify if the site is empty or occupied. If it is occupied, the trial ends again, but if it is empty, the NH₃ particle is immediately adsorbed. As the molecule of ammonia does not react, we pass to the next trial of adsorption. Now, if the chosen particle to hit on the catalytic surface is O₂, and if the chosen site is occupied, we stop the adsorption step. On the other hand, if it is empty, the following procedure is adopted: we check the neighborhood of the chosen site. If one of the four first neighbor sites is empty, this site, and the chosen one, are occupied by the two oxygen atoms. If there are two or more first neighbor sites that are empty, the adsorption site is randomly chosen. After the adsorption, we scan the neighborhood of the two oxygen atoms in order to see the possible reactions of O with H to yield OH. After this, we chek the possibility that the last molecule reacts, in its neighborhood, with another H to yield a water molecule.

Following the procedure outlined above, we performed Monte Carlo simulations for L = 256 and for fixed values of the collision rates. We find that the densities of the various types of molecules on the surface reach a stationary state. Moreover, these values depend on the values of the collision rates. For instance, in Fig. 1, we plot the density of empty sites and also the coverages of the other species versus the normalized collision rate of the NO molecules for YNH₃ = 0.3. Y_max = 1.0 – YNH₃ is the maximum value assumed by Y_NO (in this case Y_max= 0.7). As we can see, all the types of molecules can coexist on the surface depending on the value of the collision rates. After a critical value of this parameter, the catalytic surface is poisoned and the only species remaining on the surface is H. In addition, for values of Y_NO/Y_max smaller than the critical one, the density of empty sites is different from zero, characterizing an active phase for this catalytic model. However, this quantity can not be used to characterize the transition between the active and inactive phases. In fact, while the density of ammonia on the catalytic surface is different from zero, it is possible to produce N₂ and H₂O. Therefore, we use the density of ammonia as the relevant parameter to characterize the transition between the active and inactive phases. In Fig. 2 we show how the density of ammonia changes with Y_NO/Y_max for some values of YNH₃. We note that the increase in the collision rate of ammonia corresponds to the decrease of the critical value Y_NO/Y_max at which the transition occurs. The phase transition is continuous for any values of Y_NO and YNH₃. Taking all these results we can construct the phase diagram for the model of the SCR of NO by ammonia when the kinetics of the reaction is governed by the ER mechanism. In Fig. 3, we show the phase diagram for this model. As we can see, it contains a set of points separating an active phase from an inactive one. A final result for this model is related to the production of the N₂ and H₂O gases. As we can see in Fig. 4, the production rates of these gases, in the stationary regime, exhibit a maximum value as a function of Y_NO/Y_max. As we expected from the kinetics of the reaction, the quantity of water produced is larger than the production of N₂. A maximum value of this production rate also depends on the collision rate of ammonia, as shown in Fig. 5, in which we plot the N₂ production rate versus Y_NO/Y_max for some values of YNH₃.

We now turn our attention to study the model when the LH mechanism defines the kinetics of the overall reaction of the SCR of nitrogen monoxide by ammonia. According to the kinetics above mentioned, we have simulated the model as follows. First we choose, at random, the particle to be adsorbed as well as the site where the adsorption step occurs. If the chosen site is occupied, the trial ends. Otherwise, the chosen particle is adsorbed. If the chosen particle to be adsorbed is the oxygen gas, an additional first neighbor site, also empty, is needed. In this case, the O₂ molecule dissociates on these two sites of the catalytic surface. After the adsorption, we scan the neighborhood of the adsorbed species in order to see the possible reactions. Then, according to this algorithm, we perform Monte Carlo simulations on the square lattice of linear size LxL with L = 256. Surprisingly, we observed that the stationary states of this model are always poisoned for any values of the collision rates of the molecules NO, NH₃ and O₂. Besides, all the species that can be on the catalytic surface can coexist on the surface in the poisoned regime. Thus, despite the great similarity between the two reaction schemes, we find a very different dynamical behavior for the densities on the catalytic surface. To avoid the poisoning of the catalytic surface we changed the reaction scheme of the LH mechanism to include dissociation of both ammonia and nitrogen monoxide but, unfortunately, the inactive stationary state is always reached.

3 Conclusions

We have studied two heterogenous catalytic reaction models among three monomers on a square lattice in order to understand the kinetics of the selective catalytic reduction of NO by ammonia. Using Monte Carlo simulations we have shown that, while the Eley-Riedel mechanism provides an active regime for the system, the Langmuir-Hinshelwood mechanism always leaves the surface completely poisoned for any values of the collision rates.

Acknowledgements

The authors acknowledge the Brazilian agencies FUNCITEC and CNPq by the financial support for this work. W. Figueiredo is thanked for critical reading of the manuscript.

Received on 5 September, 2003

• [1] M. Chiron, Stud. Surf. Sci. Catal. 30, 1 (1987).
• [2] H. Bosch and F. Janssen, Catal. Today 2, 369 (1988).
• [3] A. Fritz and V. Pitchon, Appl. Catal. B 13, 1 (1997).
• [4] F. Nakashjima and I. Hamada, Catal. Today 29, 109 (1996).
• [5] P. Forzatti and L. Lietti, Heter. Chem. Rev. 3, 33 (1996).
• [6] F. Janssen, F. van den Kerkhof, H. Bosch, and T. J. R. Ross, J. Phys. Chem. 91, 5921 (1987).
• [7] V. Tufano and M. Turco, Appl. Catal. B 2, 9 (1993).
• [8] W. S. Kijlstra, D. S. Brands, E. K. Poels, and A. Bliek, Catal. Today 50, 133 (1999).
• [9] P. Forzatti, Catal. Today, 62, 51 (2000).
• [10] D. P. Landau and K. Binder, A guide to Monte Carlo Simulations in Statistical Physics. Cambridge, Cambridge University Press (2000).
• [11] R. M. Ziff, E. Gulari, and Y. Barshad, Phys. Rev. Lett. 56, 2553 (1986).
• [12] D. H. Schmidt and M. Santos, Phys. Stat. Sol. (a) 187, 305 (2001).
• [13] C. U. I. Odenbrand, A. Bahamonde, P. Avilla, and J. Blanco, Appl. Catal. B 5, 117 (1994).
• [14] J. A. Dumesic, N. -Y. Topsře, H. Topsře, Y. Chen, and T. Slabiak, J. Catal. 163, 409, (1996).
• [15] H. -G. Lintz and T. Turek, Appl. Catal. A 85, 13 (1992).
• [16] M. Turco, L. Lisi, R. Pirone, and P. Ciambelli, Appl. Catal. B 3, 133 (1994).

Publication Dates

• Publication in this collection
  01 Sept 2004
• Date of issue
  June 2004

History

• Accepted
  05 Sept 2003
• Received
  05 Sept 2003