Simulating the Effects of Poisoning on the Rate of the Oxidation of Ammonia over a V2O5/TiO2 Monolithic Diesel SCR Catalyst Using a Multichannel Model

The background to this study is the need to find out if some reactions of O2 oxidation of ammonia oxidation are important in the Selective Catalytic Reduction (SCR) of NO by NH3. The objective of the study was to shed light on the influence of poisoning on these reactions over a diesel SCR catalyst by compounds in the exhaust gases. The method used was to experimentally determine the amounts of products formed at several temperatures and compared them to simulated values. About 700 ppm NH3 was oxidized by 2% O2 in helium yielding N2, N2O, and NO at increasing temperatures. Comparisons are given for a 4.56% vanadia on titania fresh catalyst and the ones used for 890 and 2299 h. The kinetics was simulated using a multichannel model of the monolithic catalyst. The experimental values of the products were nicely fitted by the kinetic model where all three ammonia oxidation reaction rates were of the first order in the concentration of ammonia. The fit was somewhat better for the non-isothermal case than the isothermal one. The deactivation reduces the activation energies for the formation of all products. Effects of flow and concentration maldistribution are shown to be present but are quite small. The temperature increase is 1.30 K for the most active catalyst at the highest temperature (733 K). The use of the multichannel model shows that quite considerable deviations in inlet ammonia concentrations are obtained over the catalyst cross section. This means that the catalyst is not used to its full potential.


Introduction
The Selective Catalytic Reduction (SCR) technique is today a required method for removing NO x from mobile applications in many countries. Marine engines run under conditions, which resemble the ones in stationary diesel power plants. In 2017, there were hundreds of SCR units installed on diesel engines on ships [1]. The lubricating oil for the diesel engine is often a source for deactivation components for the catalyst. The Ca, Zn, P, and S will deactivate the SCR catalyst, as shown in previous publications [2,3]. In those studies, it was shown how the activity and selectivity in the SCR reaction system are influenced by poisoning.
In 1993 Ozkan et al. [4] studied the role of ammonia oxidation in the SCR over vanadia catalysts. They found that N 2 , N 2 O, and NO were the products. In their extensive study [5] using nitrogen labelling, they discovered that the ammonia species giving N 2 and N 2 O have relative long residence times on the surface. The NO producing species are short-lived though. Thus, three ammonia oxidation reactions are assumed. Duffy et al. [6], using V 2 O 5 /TiO 2 catalysts with varying vanadia contents, also used isotopic labelling studies to conclude that below 300°C 14 N 15 N is always the first product from 15 NO and 14 NH 3 . At higher temperatures, the product distribution is susceptible to the vanadia content of the catalyst. At 500°C, 70% of the product is 14 NO. Pure V 2 O 5 produces significantly more 14 N 15 NO, and at lower temperatures than a 1.4 wt% V 2 O 5 /TiO 2 catalyst. As part of an SCR study, the kinetics of the partial ammonia oxidation giving N 2 was studied between 250 and 300°C by Efstathiou et al. [7]. They studied the oxidation of 1000 ppm NH 3 by 2% O 2 with helium as background gas over an 8 mol% V 2 O 5 /TiO 2 catalyst. The selectivity to N 2 was from 98.2 to 97.6% decreasing with temperature. The rate of the ammonia oxidation was only from 1.8 to 2.4% of the SCR reaction rate. Djerad et al. [8] included all three NH 3 oxidation reactions in their study on the effect of oxygen on the reaction rates. Using a 3%V, 9% W on titania catalysts, the impact of oxygen was small between 2 and 15% O 2 . A maximum in the formation of N 2 was observed at 425°C, N 2 O was formed above 300°C but NO only above 425°C. They found that the creation of N 2 O from ammonia oxidation is essential in the SCR process.
In the study by Usberti et al. [9], low concentrations, 33, 150, and 333 ppm, of ammonia was oxidized by 3.5% O 2 in the presence of 2% H 2 O. The background gas was nitrogen, so this product could not be detected. They explain their results by oxidation reactions leading to NO, which reacts further with NH 3 by the normal SCR reaction. Their findings contradict the ones in this study which explain the results by only ammonia oxidation reactions.
Chae et al. [10] modelled and simulated the SCR process using experimental data from a catalytic bed. They developed models for extruded as well as wash coated monoliths. They showed that even with a catalyst layer thickness of 54 ffm, the effectiveness factor was only 0.6 for the SCR reaction, indicating diffusion resistance in a V 2 O 5 -WO 3 /TiO 2 catalyst. The ammonia oxidation to NO was used in their study. Nova et al. [11] studied the thermal deactivation of commercial V 2 O 5 -WO 3 /TiO 2 deNO X catalysts. In the direct oxidation of NH 3 using the Temperature Programmed Reaction (TPR) experiments with 840 ppm NH 3 and 2% O 2 in helium, they observed N 2 from 327°C and N 2 O from 427°C. Both compounds increase continuously with temperature up to 500°C. No nitric oxide (NO) was formed on the catalyst calcined at 500°C. When calcined at 800°C, NO was created in addition to N 2 and N 2 O. No kinetic expressions were given though.
Chen et al. [12] performed a simulation study on a catalyst bed using the commercial program COMSOL Multiphysics. They optimized the performance of the reactor using only the SCR and the direct oxidation of NH 3 to N 2 (SCO) reactions. Yun and Kim [13] modelled SCR over vanadia-based catalysts from heavy-duty diesel exhaust gases. The chemical reactions they used were the SCR, the fast SCR where NO plus NO 2 react with NH 3 to N 2 , and the oxidation of NH 3 to NO. They used the assumption that simulation of one channel would represent the performance of the whole monolith. Using the SCR and the SCO, Om et al. [14] modelled a monolith reactor for the SCR process. In 2017 Millo et al. [15] simulated the SCR system using a 1-D fluid-dynamic simulation code with the SCR and the SCO reactions. Åberg et al. [16] performed a detailed study using MATLAB comparing a reference model, including all critical parameters, to simplified models. The reactions were the NH 3 oxidation to NO, the SCR, and the fast SCR. The model was able to predict the catalyst output during the European Transient Cycle (ECT cycle). A recent paper [17] presented a method to simulate the kinetics of the SCR reactions as well as many of the NH 3 oxidation ones to obtain novel and better monolith channel designs. The results are detailed showing, among others, the temperature profile in the catalyst bed when converting 1500 ppm NO with NH 3 in the SCR process. They also consider the shape of the monolith channel showing that the most effective one is a hexagonal one.
In a recent paper [3], results are presented from the oxidation of ammonia over fresh and deactivated vanadia on titania catalysts using simulation with a single-channel model. The results were presented mostly as isothermal cases. It is therefore of interest to study a multichannel model under non-isothermal conditions. The multichannel model presented in this paper shows the details of how the flow distribution affects the activity and selectivity in the ammonia oxidation. Effects of maldistribution, on the reactor performance, of temperature, and component concentrations will also be shown.
The deactivation procedure used decreased the surface area sharply after 890 h. Only a small further reduction in area is obtained at 2299 h of deactivation. The catalyst used is a thin layer vanadia supported on cordierite where the effects of deactivation will be visible because of the small amount of catalyst present. At the same time, the 20 ffm thick catalyst layer is thin enough for internal diffusion effects to be minimal.
The data used in the present study were measured at two increasing degrees of poisoning by the compounds most likely to be present on catalyst after commercial use. Therefore, the result could be used in the design of new catalysts for diesel engines which are deactivated by compounds in the lubrication oil.

Preparation and Deactivation of the Catalyst Samples
The catalysts used were made in house by a Swedish catalyst manufacturer and consisted of 5.64% V 2 O 5 on TiO 2 (Rhone Poulenc DT (8)) on the support of cordierite from Corning with 400 Cells Per Square Inch (CPSI). Details on the preparation, deactivation and characterization can be found in Odenbrand [3]. The accelerated test was supposed to simulate the regular running of a truck for up to 500 000 km (given by the catalyst manufacturer).

Measurements of Catalyst Activity and Selectivity
Activities and selectivities were measured on 1 cm mini monoliths as described before [17] by mass spectrometry using helium as a background gas. The piece five from the inlet was used in the ammonia oxidation experiments. This piece was chosen because of its low but still noticeable degree of poisoning. The activity and selectivity in the oxidation of ammonia with O 2 were measured at about 700 ppm NH 3 and 2% O 2 . The pressure was 1.24 bar (Space velocity 45 000 h -1 ). Temperatures were from 380 to 460°C (653 to 733 K) at 20°C intervals. Helium was used as a background gas containing about 3000 ppm of Ar for internal calibration. The background concentrations of N 2 and H 2 O were 37.7 ± 0.2 and 31.4 ± 2.5 ppm respectively showing the accuracy of the mass spectrometric data. The inlet concentrations of NH 3 were 729, 746, and 702 ppm for the fresh catalyst, the ones used for 890 h, and 2299 h. For N 2 O, the following inlet concentrations were measured 1.2, 1.0, and 2.0 ppm, respectively. Also, for NO, they were 4.6, 3.8, and 4.0 ppm. The actual levels were used in the simulation of each series of experiments.

The Multichannel Model of the Catalyst Monolith
The full monolith contains 81 channels. The multichannel model uses the geometry shown in Figures 1 and 2. In the simulations, one-eighth of the entire reactor is used. The experimental setup includes an inlet tube with 2 mm inside diameter which feeds the gas to the reactor and is included in the multichannel model. High inlet velocities in the center of the reactor could cause problems with maldistribution of reactants at the inlet of the monolith. This phenomenon can quickly be addressed using the simulation program.

Stationary Modelling of the Catalytic Reactor
COMSOL Multiphysics ver. 5.5 was used in the simulations in a similar way as described before [18,19] but with a different monolithic model with physical data given in Table 1. The processor was an Intel i7 9700K octa-core operating at 3.6 GHz using a maximum of 64 GB of a 2666 MHz memory. The physic-controlled mesh was used with the "coarser" element size (The total number of elements is 241,914) for preliminary studies. When a physics-controlled mesh is used, three boundary layers close to the walls are introduced by the program to get a fine resolution of the velocity profile in those positions. The simulations were performed for non-isothermal cases. The heat was conducted to the reactors outside, according to Q 0 =h*(Tin-T). The heat transfer constant (h) was taken as 100 W/m 2 /K [18]. The solution for all cases was obtained by performing two separate stationary studies. The process was simulated in a stationary mode in 2 studies. In the stationary study, one, used to get starting values for study two, the calculations were performed in 3 steps. The Free and Porous Media Flow node was calculated in step 1 at 733 K. In step 2 (733 K), the following nodes were used, Free and Porous Media Flow, Chemistry, and Transport of Diluted Species. Finally, in step 3 (733 K), all above nodes were used in addition to Heat Transfer in Porous Media. In the stationary study two, all nodes were used, and initial values of variables were taken from study one step three. The temperature was decreased from 733 downwards to 613 or 653 K in stages of 20 K using an auxiliary step procedure. The stationary solver used was PARDISO with a relative tolerance of 0.001. The variables u and p were solved in segregated step one, all concentrations in step two, and the temperature in step three.
The fitting procedure is such that first, a mesh size "coarse" (The total number of elements is 425,512) is used. Nitrogen is first fitted. Then simultaneously N 2 O and NO. The mesh size was refined until the constant values of the outlet concentrations were obtained. Finally, the mesh size is changed to "normal" (The total number of elements is 772,241), and the improvement in results is studied. The process is guided by the relative total sum of squares of errors (see below) which is minimized. Manual changes in kinetic parameters are performed to fit the simulated values to the experimental ones according to the method described before [18,19].

Chemistry
This part of the program is used to define and calculate the chemical rates. The R ij are given as Arrhenius expressions in Table 2. In all cases, the e ect of oxygen, present in excess ff (2%), is included in the rate constant. In this part, data are supplied to calculate transport and thermodynamic data as a function of the temperature. Helium is used as a solvent.

Reaction number Global reaction Rate expression
The heat of reaction (kJ/mol) The reactions are demanding more and more oxygen and are being more important at increased temperatures; the higher the number of the reaction is (Table 2). It is only reactions 2, 4 and 5 that have a significant influence on the kinetics of ammonia oxidation because of deficient concentrations of NO in the system. Thus, reactions 1 and 3 are omitted in this study.

Transport of Diluted Species
This part of the program is used to calculate the concentrations of all components (NH 3

Free and Porous Media Flow
This part is used to calculate the velocity and the pressure from the inlet and the outlet conditions. The inlet condition is u=vin perpendicular to the inlet surface. The outlet condition is p=p tot -p ref . The reference pressure is the atmospheric pressure. The Chemistry node calculates density and dynamic viscosity.
ρis the density (kg/m 3 ) p is the pressure (Pa) I is the moment of inertia (kgm 2 )

Heat Transfer in Porous Media
This part is used in the non-isothermal cases and calculates the temperature in the system. Heat, evolved by the chemical reactions in the catalyst, and transported through the system causes the temperature to increase. Q (q in the equation above) is the heat that flows from the reactor to its outside. ρC p u*∇T+∇*q=0 q=-k eff ∇T C p is the specific heat capacity at constant pressure (J/kgK) T is the absolute temperature (K) k eff is the effective thermal conductivity (W/mK) q is the heat transferred (J/m 2 s)

General Simulation Conditions
The physical properties used in the simulations are given in Table 1. Incompressible flow with a normal inlet velocity of vin in the direction of the reactor axis was used for the Free and Porous Media Flow node. The total flow (Q in ) at NTP was 0.9 l/min was introduced into the large tube containing the catalyst block trough a stainless-steel tube with an inner diameter of 4 mm in the multichannel model. The circular inlet area of the tube (A in ) was 1.265*10 -5 m 2 . The reference pressure (p ref ) was 1 bar, and the total one (p tot ) at the reactor exit was 1.236 bar. Thus, v in =Q in /A in *Tin/273.15/p tot *p ref (2.593 m/s at 733 K (460°C)). The program calculates the varying inlet velocities at different inlet temperatures (T in ).
The thickness of the catalyst layer was to 21 µm as described before [3]. The catalyst layer was supposed to cover all inside walls of the monolith. The reactor, made of quartz glass, is 100 mm long with the bottom of the catalyst monolith (1 cm long) positioned 60 mm from the inlet on a plug of glass wool. The monolith was surrounded by quartz wool to stop gas bypassing the catalyst. Only a small part of the simulated system is shown to detail the size of the mesh used in the calculations.

Criteria for Evaluation of the Best Fit
When comparing simulations, it is necessary to use the relative sum of squares of errors (SSE reltot ) of the whole data set. Each set of experimental concentration data is first divided by its maximal value (at T=733 K) yielding N 2 rel=N 2 (Tin)/N 2 exp (733

Catalyst Characteristics
The deactivated catalysts have been analyzed for their surface compositions of Ca, P, Zn, and S [2] usually compounds present in the fuel and especially lubrication oil. From the results shown in Table 3 one can assume that Ca is present at the very uttermost layers of the catalyst surface since its content increases with time of use. Part of the S also stays at the surface probably as CaSO 4 . CaSO 4 has been shown before to deactivate similar catalysts [19]. P, Zn, and S do also penetrate the catalyst layer and therefore their contents on the surface are lower at the longest times. The contents of P and Zn on the surface after 2299 h are about 25% of that after 890 h. Even so, the experimental rates for most of the reactions are lowest on the catalyst used for 2299 h. Thus, one must assume that these compounds penetrate deeper into the catalyst layer causing the lower activity.
There is a strong influence of the deactivation procedure on the surface area and the pore volume ( Table 3). The catalyst used for 890 h has only 61% of its original surface area left.
After 2299 h the surface area decreases to 55% of its original value. The pore volume is not so drastically reduced, but by13% after 890 h and by 27% after 2299 h. Thus, the surface area seems to stabilize after some time while the pore volume continues to decrease on deactivation almost proportionally to the time of use. Earlier studies show that the smallest pores are eliminated by sintering or pore filling with poisons [2].

Fit of the Simulated and Experimental Concentrations
As Figure 3 shows is it possible to nicely fit the experimental concentrations of N 2 , N 2 O and NO, as direct reaction products of the oxidation of ammonia, to the ones simulated by the model. Figure 3 top left shows the concentration of N 2 in the reactor exit and its variation with the temperature for the fresh catalyst. It varies from 85 to 156 ppm for the fresh catalyst, from 89 to 155 ppm for the catalyst used for 890 h, and from 45 to 80 ppm for the catalyst used for 2299 h. Considering that two ammonia are needed to produce one N 2 the conversions are 13.2 to 32.8%, 13.9 to 31.6%, and 12.8 to 33.3% for the three catalysts. Thus, there is not a large difference in the conversions of NH 3 to N 2 among the catalysts.
The products N2O and NO are present as 32 and 13 ppm, 54 and 24 ppm, and 44 and 38 ppm for the three catalysts at the highest temperature. Corresponding conversions of NH 3 Table 4 shows that the activation energy decreases when the catalyst gets more deactivated, as explained before [20,21]. All three reactions behave similarly in this respect. The poisoning has the most substantial effect on the formation of NO with an increased amount of NO on the poisoned catalyst.

Kinetic Parameters
Kinetic parameters do not change very much when non-isothermal conditions are introduced using the "normal" mesh size. When the "fine" mesh size is used, exceptionally long simulation times are needed and large amounts of memory, slowing down the process too much for only a small change in parameters obtained (not shown here).
We compare our values to Chen and Tan [12] who also used COMSOL Multiphysics in the simulation of, in their case, a V 2 O 5 on TiO 2 catalyst bed. Their value of E 2 is 84.4 kJ/mol for the formation of N 2 (R 2 ) when corrected for diffusion limitations while our apparent value is 64 kJ/mol for the fresh catalyst. The values of k 2 (A 2 *exp(-E 2 /(R*T)) were 6.73*10 7 and 3.10*10 6 s -1 , respectively. The rate constant at 653 K is thus 11.9 for Chen and 23.5 s -1 in this study. Remarkably similar values. Are the rates of formation of N 2 large enough to be influenced by washcoat diffusion? The thickness of the catalytic layer is 21 µm, and the Thiele modulus is 0.0293 leading to an effectiveness factor of 0.99975 at 460°C. Thus, all lower temperatures will be diffusion free too. The high value of the effectiveness factor in our study can be compared to the data of Chae et al. [10] where they state that for a wash coated 2 wt% V and 6 wt% W, V 2 O 5 -WO 3 /TiO 2 catalyst with a 54 µm catalyst layer the effectiveness factor in the SCR reaction was 0.6 at 300°C and 0.2 at 460°C. The rates of the oxidation of ammonia are lower than the rates of the SCR reaction and can explain the difference.
Efstathiou and Fliatoura [7] determined the apparent activation energy of 61.4 kJ/mol for the oxidation of 1000 ppm NH 3 with 2% O 2 in helium for an 8 mol% (16.5 wt%) V 2 O 5 /TiO 2 catalyst from experimental rates. The rate of N 2 formation at 300°C was 17.2*10 -6 mol/g/min compared to ours calculated one of 7.62*10 -2 mol/(m 3 s) (1.99*10 -6 mol/g/min) for the fresh catalyst. Thus, rates are in the same range. The difference can be explained by that their catalyst contained more V 2 O 5 than ours.
The formation rates of N 2 (R 2 ), as shown in Table 4 [22] to ours, since they also used first-order dependence on NH 3 concentration. They also simulated a monolithic structure with a catalyst layer. Their values of k 2 and E 2 were 6.8*10 7 and 85 compared to ours of 3.1*10 6 s -1 and 64 kJ/mol, respectively. We calculate the rate R 2 at 653 K from Salehi's [22] data to 0.159 compared to ours of 0.347 mol/m 3 /s. They are in the same range.
Om et al. [14] studied, using Fluent, the SCR in a monolith isothermally. The concentrations of NO and NH 3 were in the range of 1000 ppm, so we believe that the simulation should have been non-isothermal for more precise results. Their data were based on an experimental study on a catalytic filter by Schaub et al. [23], so the values of the rate parameters should be close to intrinsic values, and k 2 and E 2 were 6.73*10 7 and 85.4 kJ/mol. At the same time, ours were 3.1*10 6 and 64 kJ/mol, respectively ( Table 4). Calculation of nitrogen formation rates at 653 K from their data gave a value of 0.183 while our value was 0.19 mol/m 3 /s. The rates are remarkably close. At 460°C their rate was 0.91 and ours 1.24 mol/m 3 /s. Millo et al. [15] simulated the SCR on a filter catalyst for automotive applications. Besides the standard SCR, the fast and slow SCRs, the oxidation of NH 3 to N 2 was used as model reactions. The activation energy (E 2 ) was 144.6 kJ/mol, much higher than our value of 64 kJ/mol. Their rate expression was r 2 =k 2 '*c NH3 *c O2 includes the dependence on the oxygen concentration making direct comparisons hard. Their experiments also included water which would decrease the rate of all reactions.
There is a maximal rate for all oxidation rates but the oxidation to NO for the catalyst used for 890 h. Formation of new acidic sites, from sulphur species, is a probable cause for this maximum. The uncorrected, non-isothermal, rate of formation of N 2 (R 2 ) increases from 1.228 via 1.297 and decreases to 0.868 mol/m 3 /s from 0 via 890 to 2299 h of use. The first increase is 5.6%. The decrease at 2299 h is 29.3% from the fresh one.

Surface Specific Rates
The rates given in Table 4 are recalculated as surface specific rates since the BET surface area shrinks considerably during the deactivation process as described above. When the non-isothermal surface-specific rates are compared at 733 K (Table 5), they are 96.4, 167, and 122.9 *10 -3 mol/m 3 /s/m 2 /g, for the three catalysts. Thus, a 73% increase is observed at 890 h from the fresh catalyst. At 2299 h of use the increase is 27% compared to the fresh one. Thus, both deactivated catalysts are more active than the fresh one per surface area unit. At 653 K the non-isothermal surface-specific rates are 30.8, 56.0, and 51.8 *10 -3 mol/m 3 /s/m 2 /g, respectively.
The non-isothermal surface-specific rates for the formation of N 2 O at 773 K are 23.8, 75.2, and 61.5*10 -3 mol/m 3 /s/m 2 /g, respectively for the three catalysts. Both these reactions have maximum activities at 890 h of use.
The rate of formation of NO behaves differently being 36.4, 144, and 245*10 -4 mol/m 3 /s/m 2 /g at 733 K. Thus, the rate of formation of NO increases continuously with the degree of poisoning. Such a result is in line with them by Due-Hansen et al. [24] who studied the K poisoning of V 2 O 5 -WO 3 /ZrO 2 catalyst in ammonia oxidation. An increased poisoning caused an increased formation of NO.  Table 4 shows that the maximal rate is obtained for the catalyst used for 890 h. This results in the highest temperature increase for this case (Figure 4, red curve, 1.3 K, inlet temperature 733 K). The catalyst inlet is situated 11 cm from V 2 O 5 /TiO 2 Monolithic Diesel SCR Catalyst Using a Multichannel Model the reactor´s inlet. The temperature starts to increase before it reaches the monolith and is 734.1 K at its inlet for the catalyst used for 890 h. It increases to a maximum of 734.3 K inside the monolith. The maximal temperature for the fresh catalyst and for the one used for 2299 h are734.1, and 734.0, respectively. Thus, the temperature increases correlated fine with the rates in Table 4.  Figure 5 shows that at 733 K inlet temperature, the maximal temperature increase (1.30 K) is in the center of the catalyst block and is located at around 4 mm from the exit of the monolith. At 653 K inlet temperature, the increase (0.43 K) is at 2 mm from the outlet of the monolith. The higher the temperature, the closer to the catalyst inlet is the maximal temperature which is expected. A maximal temperature in the center of the catalyst block is expected since the heat of reaction is transported towards the outer side of the block. The temperature outside of the reactor is equal to T in and is cooling the catalyst.    before the monolith. The question arises whether this causes a maldistribution in temperature and inlet NH 3 content. The relative long distance from the inlet tube to the monolith works to make the velocity profile more even. After the catalyst, the flow is linear as it was in the inlet tube. Figure 8 shows how ammonia is distributed over the system for the catalyst used for 890 h. The simulation gives 745.6 ppm at the reactor inlet. The correct inlet value is 745.6 ppm since this is the boundary condition. Because of heat effects the content of NH 3 at the catalyst inlet surface (Figure 8 top right) has decreased to between 728 and 697 ppm. At the outlet of the catalyst (Figure 8 bottom left) the content varies from 393 to 384 ppm with the lowest conversion in the center of the monolith where there is a higher velocity than other places.  The concentration of ammonia at the reactor inlet is, by simulation, 745.6 pm over the whole cross section. 5 mm from the catalyst inlet the concentration varies from 492 to 514 ppm with conversions of 34.0 to 31.1%, respectively (not shown here). At 1 mm from the monolith exit, (Figure 8 top right) the value in the center is 408 ppm and the lowest 389 ppm somewhere inside the monolith. At the catalyst outlet the conversion varies from 48.5 to 47.3% (Figure 8 bottom left). Thus, the variation is not considerable. In the last part of the Figure 8 (bottom right), the reactor exit concentration is shown and it is 386 ppm, equal over the reactor cross section. Thus, here the conversion is 48.2% and it is even over the whole surface. Figure 9 clearly shows the variation of the NH 3 content from 415 ppm in the three channels shown in blue to the left in the monolith to 432 ppm in the center channel (red). The higher content in the center is caused by the shorter residence time there (higher linear flow) and thus a lower conversion of NH 3 . The difference between the lowest and the highest concentration is but 4%. Figure 10 shows the simulated concentration of NO at the monolith exit for the catalyst used for 890 h and at 733 K. The distribution is uneven with the highest temperature in the center part of the catalyst monolith. The highest NO contents are observed in the pink areas in the Figure and are 13.7 ppm for the fresh catalyst, 24.1 ppm for the catalyst used for 890 h, and 38.2 ppm for the catalyst used for 2299 h. The lowest contents span from 13.4 via 23.6 to 37.1 ppm and are always seen in the center of the catalyst where the flow rate is largest. The total temperature increase is only 0.95 K at 733 K for the catalyst used for 2299 h. The low temperature maldistribution results in just a little smaller amount of the product NO in the center of the catalyst monolith as shown in Figure 10.

The Advantages of Using a Multichannel Model
The use of the multichannel model has several advantages, some of them have been shown in this example. It is of special importance to use a multichannel model if one has a special arrangement of the inflow as in the example presented here. The total temperature increase is meagre in the present example, but still effects are seen in variation of the conversion over the catalysts surface. It is the high flow in the inlet tube that causes a distinct higher flow also in the center part of the monolith. For more exothermic reactions and larger inlet flows the advantage of using the multichannel model are clear. The only disadvantage is that in this model it was V 2 O 5 /TiO 2 Monolithic Diesel SCR Catalyst Using a Multichannel Model necessary to divide the monolith and reactor into 8 parts and only make simulation on one of them. This was necessary because of large memory demand and large computation time especially with an exceptionally fine mesh.

Conclusions
In the simulation of the of oxidation of ammonia with oxygen using a commercial Finite Element package it was shown that the three products nitrogen, nitrous oxide, and nitric oxide are formed at increasing temperatures. In the simulation, all three rates of oxidation of ammonia were supposed to be of first order in its concentration.
A multichannel model, including all the 9*9 channels, was used to study the effects of flow, concentration, and temperature maldistributions on the results. The effects were clear, albeit at low values. The maximal temperature increase was around 1.3 K at 733 K.
The kinetic parameters were shown to vary with the degree of deactivation. By simulation, the effect of deactivation on the kinetic parameters of the oxidation of ammonia was determined. A decrease of the activation energies with the degree of deactivation was observed for all reactions. Experimental rates were at maximum for the catalyst used for 890 h for the formation of N 2 and N 2 O. The formation of NO increased continuously with the degree of poisoning.
There is quite a substantial reduction in BET surface area during the deactivation process. When the surface-specific rates were compared, the order of rates for the three catalysts remained the same even if absolute values decreased.