Ind. Eng. Chem. Res. 2004, 43, 6955-6963
6955
Reverse Flow Operation with Reactor Side Feeding: Analysis, Modeling, and Simulation Yogi W. Budhi, Jozef H. B. J. Hoebink,* and Jaap C. Schouten Laboratory of Chemical Reactor Engineering, Eindhoven University of Technology, P.O. Box 513, Den Dolech 2, 5600 MB Eindhoven, The Netherlands
The novel concept of reverse flow operation with reactor side feeding is studied for selective oxidation of NH3 to produce either N2, N2O, or NO. During normal reverse flow operation, where the feeds are alternately introduced from either end of the reactor, the conversion is always lower when compared to steady state, once-through operation at the same residence time. The novel concept of reverse flow operation with reactor side feeding is therefore proposed to avoid the decrease of conversion. The production rate of the desired product can be kept at a high level, which may even exceed the production of normal reverse flow operation and steady state, once-through operation. The reactor behavior was simulated for three regimes (quasi steady state, dynamic, and relaxed steady state) with varying feed positions. The development of spatiotemporal patterns inside the reactor with side feeding shows completely different behavior compared to normal reverse flow operation and steady state, once-through operation, leading to the possibility of conversion and selectivity manipulations. In addition, the proposed concept also indicates better conversion for the dynamic and relaxed steady-state regimes. The influence of shifting the feed positions to the reactor center is most pronounced if the switching time is shorter. This concept also provides an opportunity to prevent the dead gas volume in the center of the reactor if suitable feed positions can be applied. The catalyst effectiveness in this novel concept is therefore much better than during normal reverse flow operation and steady state, once-through operation, particularly in the relaxed steady-state regime. Introduction Periodically changing the flow direction through a reactor, known as reverse flow operation (hereinafter referred to as RFO), is a typical example of forced unsteady-state operation. The latter has been widely explored over the past decades to perform heterogeneous catalytic reactions by utilizing the dynamic properties of both the reactor and the catalyst. RFO has not yet been widely applied for improvement of conversion or selectivity. It was shown to be beneficial for exothermic reactions from a viewpoint of energy saving.1,2 Both theoretical and functional aspects have been studied in detail, as well as the industrial applications in various processes. Rather limited information about reverse flow operation for manipulation of selectivity can be found in the literature. RFO with reactor side feeding has already been considered to trap strongly adsorbing NH3 species during the selective catalytic reduction of NO by NH3.3 Simulations showed that side stream NH3 addition enables better mass trapping compared to addition from the reactor ends as well as better control of axial adsorbate profiles.4 A correlation was found between the reversal time to prevent the escape of the adsorbing species and the concentration of the nonadsorbing species. As a result, RFO with central introduction of adsorbing species could be more efficient than a conventional RFO.5 RFO for NO reduction with central NH3 introduction was also used for safety considerations in relation to ammonium salt formation on the catalyst.6,7 * To whom correspondence should be addressed. Tel.: +3140-2472850. Fax: +31-40-2446653. E-mail: J.H.B.J.Hoebink@ tue.nl.
As the dynamics of reverse flow operation for manipulation of conversion and selectivity are much faster than their equivalents in the case of energy saving, it requires considerably more frequent flow reversals with a switching time in the same order of magnitude as the residence time is. Here the switching time is defined as the time period between two flow reversals. This kind of operation induces a pronounced decrease of the conversion, which is mainly caused by an outflow of highly concentrated, unconverted reactant shortly after reversing the flow direction. This amount of gas actually had a shorter residence time and higher reactant concentration compared to the rest of the reactor’s gas holdup.8 On the basis of RFO simulations of ammonia oxidation, Figure 1 depicts as an example the conversion of NH3 and the selectivities to N2, N2O, and NO as a function of the switching time. To overcome the conversion decrease at low switching time, a new concept of reverse flow operation with periodically lower feed concentration as proposed by Budhi et al.8 seems very promising. By temporarily decreasing the feed concentration before reversing the flow direction, a better conversion is obtained compared to the steady-state operation. Furthermore, the selectivity can be manipulated freely in a significant range. However, this technique induces a decrease of the production rate, particularly at longer periods of lowered feed concentration. Another option to improve the conversion during reverse flow operation is considered here by applying the concept of reverse flow operation with reactor side feeding. The feed gas does not enter the reactor at z ) 0 (flow from left to right) or z ) 1 (flow from right to left) as usual, but at some axial coordinates in between.
10.1021/ie049702d CCC: $27.50 © 2004 American Chemical Society Published on Web 09/29/2004
6956
Ind. Eng. Chem. Res., Vol. 43, No. 22, 2004
Mathematical Model
Figure 1. Simulated conversion and selectivity for ammonia oxidation in normal RFO as a function of the switching time. Temperature, residence time, and O2/NH3 ratio are 800 K, 10 s, and 1:1, respectively.8
The reactor outlets still flow out from the normal positions at z ) 0 or z ) 1, depending on the flow direction. Such an approach might be feasible to avoid the decrease of conversion, since the gas with high reactant concentration, which entered the reactor shortly before the flow direction is changed, now still may contact the catalyst when flowing out of the reactor. Moreover, the “dead” gas volume in the center of the reactor, which exists if the switching time between flow reversals is smaller than the residence time, could be possibly avoided as well if suitable feed positions are applied. These fundamental aspects are addressed in this paper through the numerical simulation of a reverse flow operation with reactor side feeding for the selective oxidation of ammonia. This reaction was chosen as a model reaction, as it produces either nitrogen (N2), nitrous oxide (N2O), or nitric oxide (NO), which provides an interesting selectivity issue. Computer simulations were carried out with a kinetic scheme, constructed from literature data on the basis of elementary steps.9 Focus was on a comparison of reactor conversion and selectivity, obtained in reverse flow operation with both normal feeding and side feeding, and also on comparison with steady state, once-through operation. Three kinds of regime were considered: the quasi-steady-state, dynamic, and relaxed steady-state regimes. It is shown that reactor side feeding provides a way to overcome most of the conversion decrease, met during normal RFO, while clearly different selectivities can be obtained.
A comparison between a conventional reverse flow reactor and a reverse flow reactor with side feeding is schematically shown in Figure 2. Note that in the reverse flow reactor with side feeding, the feeds are positioned at some axial coordinates between the reactor ends. The white and black valves indicate that the position is open or closed, respectively, depending on the flow direction. During the first half of each cycle, the flow passes through in the order 1 f 2 f R f 3 f 4 and during the second half of each cycle, the flow passes through in the order 1 f 2′ f R f 3′ f 4. The time period of flow in one direction through the fixed bed reactor is typically referred to as the switching time. A symmetric switching time is usually applied, meaning that the periods of forward and backward flows are equal. Flow reversals lead to periodically moving temperature and concentration profiles, e.g., spatiotemporal patterns. Therefore, accumulation terms should be included in the mass and energy balances that describe the behavior of the fixed bed reactor operated in reverse flow mode. The volumetric heat capacity of a solid catalyst is commonly 3 orders of magnitude higher than that of a gas. This means that in the conventional RFO application for energy saving, the concentration profiles have basically reached a steady-state condition and the mass balance can be modeled in pseudo steady state accordingly. In the present work, dealing with more frequent flow reversals, concentration movements along the reactor bed are considered and the temperature response is in a relaxed steady-state regime of operation, leading to real steady state. Taking this into account, heat accumulation in the gas phase can be excluded from the model, as it influences only the reactor behavior shortly after each flow reversal. In this preliminary work, the energy balance was completely discarded. An imposed gas-phase temperature profile along the reactor axis as used by Budhi et al.8 was applied. This profile was calculated for one case, involving heat balances, and it was assured that a permanent cyclic regime had been established. Note that the reactor wall temperature and heat loss through the reactor wall are irrelevant if the temperature profile is known. Figure 3 shows the scheme with reactor side feeding, defining the axial positions of the feeds and the distinct zones in the reactor. Feeds are positioned at axial coordinates z ) La/Lr and z ) 1 - La/Lr in terms of dimensionless reactor length. The regions 0 e z e La/ Lr and 1 - La/Lr e z e 1 are referred to as “interrupted flow regions”, where the flow is alternately stopped or in one direction only. In the “reversed flow region”, La/
Figure 2. Principle diagram of a reverse flow reactor: (a) normal operation; (b) with reactor side feeding.
Ind. Eng. Chem. Res., Vol. 43, No. 22, 2004 6957
was set to a low value of 10-7 m2 s-1, and it was verified that this value did not affect the simulation results. Therefore, the flow pattern is essentially plug flow. The pressure drop along the reactor bed was ignored. The steady-state situation for once-through operation was used as initial condition (eqs 8 and 9), while conventional Danckwerts boundary conditions were applied for the gas-phase equations (eq 10 for forward flow, eq 11 for backward flow).
Figure 3. Reverse flow reactor with side feeding, showing the feed positions. The regions 0 e z e La/Lr and 1 - La/Lr e z e 1 refer to the “interrupted flow regions”, where the gas passes through in one direction only. The region La/Lr < z < 1 - La/Lr is called the “reversed flow region”, where the flow direction is alternating. Numbers 2, 2′, 3, and 3′ refer to inlets and outlets, as in Figure 2. Table 1. Global Reactions of Ammonia Oxidationa 4NH3 + 3O2 f 2N2 + 6H2O 4NH3 + 4O2 f 2N2O + 6H2O 4NH3 + 5O2 f 4NO + 6H2O
(1) (2) (3)
a The detailed kinetic scheme based upon elementary steps was taken from Rebrov et al.9
Lr < z < 1 - La/Lr, the flow direction is periodically alternated. Table 1 presents the global reactions considered in this study. The reactor model equations are shown in Table 2. The unsteady continuity equations of the gasphase components i are partial differential equations (PDEs), while for the species j, adsorbed on the catalyst surface, these are ordinary differential equations (ODEs) and algebraic equations. Detailed kinetics for Pt/Al2O3 catalyst was used as presented by Rebrov et al.,9 who proposed a dual-site mechanism with different active centers, which are called hollow and top sites for O- and N-containing species, respectively. The complete model was solved numerically with the software package FlexPDE.10 Table 3 presents the operating conditions that were used. A dynamic, heterogeneous, one-dimensional model was applied (Table 2, eqs 4-7). Any mass transfer resistance was ignored. Axial dispersion was included only for numerical stability. The dispersion coefficient
Residence Time Distribution In the quasi-steady-state and dynamic regimes of normal RFO, the switching time is larger than the gas residence time, which means that between subsequent flow reversals, the gas volume of the fixed bed is flown through completely and a dead gas volume is not present. In the relaxed steady-state regime, however, the switching time is smaller than the gas residence time. As a result, the gas volume in the central part of the reactor is not flown through anymore between flow reversals. Although the conversion in this dead volume is high in itself and may even reach full conversion, normal RFO in this situation causes nevertheless a decrease of the overall reactor conversion. The relaxed steady-state regime is indeed cumbersome for normal RFO. When applied for manipulation of conversion and selectivity, the relaxed steady-state regime is not interesting at all due to the existence of the dead gas volume leading to the ineffective use of catalyst. A flushing method, i.e., alternation between RFO and once-through operation, may improve the reactor conversion, but values as would be obtained in once-through steady-state operation still cannot be reached.8 In RFO with reactor side feeding, it is possible to avoid such dead gas volume if a suitable switching time at fixed feed positions is applied. So, the ineffective use of the catalyst can be avoided accordingly. The dead gas volume in the reactor can be prevented if the region between the two feed points (at La and at Lr - La) with a length
Lmin ) Lr - 2La
(12)
is completely flown through during the switching time. The minimum time required to pass this “reversed flow
Table 2. Model Equations for the Heterogeneous One-Dimensional Model mass balances for gas-phase component i 2 ∂pi 1 ∂pi Def ∂ pi σ )+ 2 2 + FbRT ∂t τ ∂z b L ∂z r
∑r
(4)
k
i ) NH3, O2, N2, N2O, NO, H2O mass balances for species j, adsorbed on the catalyst surface dθj/dt ) ∑rk θts ) 1 - ∑θjts θhs ) 1 - ∑θjhs j ) {NH3}, {NO}, (O), (OH), (N), with {j} as top site, (j) as hollow site initial conditions, t ) 0, z ∈ [0, 1] pi ) piss θj ) θjss boundary conditions, t > 0
forward flow:
backward flow:
|
∂pi ∂z
) 0,
z)0
|
∂pi ∂z
z)0
) 0,
|
∂pi ∂z
)
z)La/Lr
|
∂pi ∂z
uLr (p ° - pi), Def i )
z)1-La/Lr
and
uLr (p ° - pi), Def i
|
∂pi ∂z and
(5) (6) (7) (8) (9)
(10)
)0 z)1
|
∂pi ∂z
z)1
)0
(11)
6958
Ind. Eng. Chem. Res., Vol. 43, No. 22, 2004
Figure 4. Residence time distribution during reverse flow operation with reactor side feeding. Table 3. Catalyst Properties and Operating Conditions 5.97 × 10-3 0.05 0.45 0.1-0.495 30 × 10-3 2-15 800 1 1.35 × 10-3 1 2:1
catalyst mass (kg) Pt catalyst (%) void fraction (mg3/mr3) feed position La/Lr (-), varied total reactor length (m) switching time ts (s), varied inlet temperature Tin (K) total pressure (atm) superficial velocity u (m/s) NH3 in feed (mol %) O2/NH3 ratio in feed
region” is defined as the minimum switching time (ts,min) and can be written as
ts,min ) b
Lr - 2La u
(13)
with Lr being the total reactor length and u the superficial velocity. With the definition of the residence time for oncethrough operation being τ0 ) bLr/u, eq 13 is rewritten as
(
ts,min ) τ0 1 - 2
)
La Lr
(14)
The residence time distribution of RFO with side feeding is of interest to obtain insight in the achievements of this type of operation. It will be considered here for situations where the switching time is equal to or larger than the minimum switching time. Case La ) Lr/3. The minimum switching time equals τ0/3 according to eq 14. The minimum residence time, τmin, has two contributions: one from flowing through an interrupted flow region and another from standing in this region, while flow is through the other interrupted flow region. So τmin equals 2τ0/3. The maximum residence time, τmax, has additional contributions from flowing through the reversed flow region in forward and backward direction. With each of them being τ0/3, the maximum residence time is 4τ0/3. Between τmin and τmax, the distribution is uniform, as shown in Figure 4. The average residence time equals τ0, as it should. Case 0 < La < Lr/3. In this case, the minimum switching time is always larger than τ0/3. The minimum residence time equals the residence time in the interrupted flow region:
ts,min )
bLa La ) τ0 u Lr
(15)
The reversed flow region adds between 0 and 2ts,min to this minimum residence time, but part of the gas stands an extra time ts,min in the interrupted flow region. This part concerns the fraction La/(Lr - 2La) of the gas in the reversed flow region and has residence times between 3ts,min - τmin and 3ts,min + τmin. The latter is also the maximum value of the residence time, which consists of contributions from flowing up and down the reversed flow region plus flowing and standing in the interrupted flow region. So in this case, the residence time is bimodal. For small values of La/Lr, there exists a small amount of gas with a considerably larger residence time. With increasing La/Lr, the two parts of the distribution function approach each other till they merge at La/Lr ) 1/3 (see Figure 4). Since the average residence time should be equal to τ0, the value of the uniform distribution can be obtained from normalization. Case Lr/3 < La < Lr/2. The minimum switching time ts,min is now smaller than τ0/3, meaning that all gas coming from the reversed flow region will stand in the interrupted flow region for at least one minimum switching time. The value of δ ) int[La/(Lr - 2La)] determines how many times the gas flow will be interrupted for ts,min seconds, when the gas passes the interrupted flow region. The minimum residence time τmin ) δts,min + bLa/u ) δts,min+ τ0La/Lr. The maximum residence time τmax is 2ts,min larger. If La/(Lr - 2La) is an integer, the distribution of residence times is uniform between τmin and τmax; i.e., it has a width of 2ts,min. If La/(Lr - 2La) is not an integer, the residence time distribution becomes bimodal, because part of the gas stands an extra time ts,min in the interrupted flow region. Whether or not La/(Lr - 2La) is an integer becomes irrelevant, if La approaches Lr/2, meaning that the minimum switching time decreases to zero. For sufficiently large values of δ, the residence time distribution is uniform between τ0 - ts,min and τ0 + ts,min with an average value of τ0. The limit situation of La approaching Lr/2 corresponds with half of the flow rate moving from the middle of the reactor to one side, while the other half moves to the other side. This situation shows a residence time distribution, which is exactly the same as met with once-through operation. For first-order reactions the conversion is fully determined by the residence time distribution. This means that side feeding around La ) Lr/3 provides opportunities to influence the conversion in a positive way, because of the bimodal character of the distribution function. Part of the gas resides longer in the reactor, while another part has a shorter residence time. For reaction orders which are different from 1, the reaction kinetics has an influence as well. It is obvious that the reactor selectivity will be affected as well. Results and Discussion As already mentioned in the Introduction, a major drawback of normal RFO is a decrease of the conversion in comparison to steady state, once-through operation and a minor influence on the product distribution, which are the major reasons to apply reactor side feeding. In this section, simulation results of a reverse flow operation with reactor side feeding are presented. First, its behavior at different regimes is discussed and compared to the behavior of normal RFO. Second, the effect of the feed position on conversion and selectivity is presented. Third, the conversion and selectivity under conditions of minimum switching time is demonstrated.
Ind. Eng. Chem. Res., Vol. 43, No. 22, 2004 6959
Figure 5. Gas-phase NH3 profiles as a function of time at discrete positions in the reactor during one cycle. The flow direction (left and right) is shown by arrows. Left figures relate to normal RFO, right figures to RFO with reactor side feeding at positions z ) 0.1 and 0.9. Operating conditions as mentioned in Table 3: (a1, a2) quasi-steady-state (ts ) 15 s); (b1, b2) dynamic (ts ) 10 s); (c1, c2) relaxed steady state (ts ) 2 s).
Behavior of RFO with Reactor Side Feeding. The simulation of normal RFO is used as a reference for the comparison of RFO with reactor side feeding. Figure 5 shows typical results of the NH3 partial pressure as a function of time for three different regimes during one complete cycle. Each regime is represented by a characteristic switching time. Normal RFO with the feed position at the outer ends of the reactor (z ) 0 or z ) 1, depending on the flow direction) is shown in parts a1, b1, and c1 of Figure 5. For reactor side feeding with feeds at z ) 0.1 and 0.9, parts a2, b2, and c2 of Figure 5 provide results at similar conditions.
In the quasi-steady-state regime, the outlet concentration during normal RFO (Figure 5a1) is able to reach the steady-state value as shown by the horizontal part of the profiles. A short while after reversing the flow direction, the outlet concentration of NH3 is high because the high concentration of NH3 that just entered into the reactor is flown out again. This typical behavior causes the decrease of conversion. In general, the timeaverage behavior is not so different from the steadystate performance. When reactor side feeding is applied (see Figure 5a2), the concentration profiles between the inactive inlet point and outlet point are also able to
6960
Ind. Eng. Chem. Res., Vol. 43, No. 22, 2004
Figure 6. Gas-phase NH3 profiles as a function of time at discrete positions in the reactor during one cycle. The feed positions are at z ) 0.4 and 0.6. The flow direction (left and right) is shown by arrows. Operating conditions as mentioned in Table 3: (a) quasi steady state (ts ) 15 s) and (b) relaxed steady state (ts ) 2 s).
reach the steady state, while the concentration profiles in regions where flow is interrupted stay in the unsteady state due to continuing reaction. These patterns are clearly shown in Figure 5a2. During the first half-cycle (130 s e t e 145 s), the gas enters the reactor at z ) 0.9, while the products leave the reactor at z ) 0. The NH3 partial pressures at z ) 0.1 and 0 reach the steadystate value, while at the closed outlet, z ) 1, the partial pressure continues to decrease. Such typical profiles are repeated during the next half-cycle, which means that the periodic flow reversal has already attained stable oscillations. The behavior in the dynamic regime seems similar to the quasi-steady-state behavior, apart for the constant steady-state values, which do not exist. Before the profile takes steady-state values, the reactor is already perturbed by reversing the flow direction. In this way, the reactor is always maintained under dynamic conditions, as shown by Figure 5b1. Time-average behavior is therefore expected to be different from the steadystate performance. When reactor side feeding is applied (see Figure 5b2), similar behavior is observed. Just after reversing the flow direction, the outlet concentration profile changes nonmonotonically as a function of time without reaching the steady state. In RFO with reactor side feeding, the outlet concentrations are slightly higher for the dynamic regime (see Figure 5b2) compared to the quasi-steady-state regime (see Figure 5a2). It is caused by the faster switching time in the dynamic regime. It also means that the interrupted flow time in the dynamic regime is shorter than in the quasi-steadystate regime, as reflected by less decay of reactant at positions of interrupted flow region. In the relaxed steady-state regime for normal RFO, the concentration profiles look completely different (see Figure 5c1). All profiles remain in the unsteady state and depend strongly on the ratio of switching time over residence time. Compared to the two former regimes, the outlet concentration in the relaxed steady-state regime is much higher. When reactor side feeding is applied (see Figure 5c2), the concentration levels in the relaxed steady-state regime are also higher than in the other regimes. It should be emphasized, however, that Figure 5 refers to feed positions at z ) 0.1 and 0.9 and that the opposite holds for feeds in the region 0.3 < z < 0.7. Such conditions lead to a lower outlet concentration of NH3 and hence higher conversion of NH3.
Figure 7. Gas-phase NH3 profile along the reactor length at discrete times during one half-cycle. ts ) 10 s and feed positions are at z ) 0.1 and 0.9. Other conditions as mentioned in Table 3.
Figure 6 shows profiles for the quasi-steady-state and relaxed steady-state regimes with feeds more to the center, e.g., z ) 0.4 and 0.6. In the quasi steady state, shifting the feed position toward the reactor’s center causes a higher concentration of NH3 in the interrupted flow region if flown through, while the concentration decay is similar if not flown through (compare Figures 5a2 and 6a). Figure 6b shows for the relaxed steadystate regime definitely lower outlet concentrations when compared to feeds at z ) 0.1 and 0.9 (see Figure 5c2). This behavior is very interesting for manipulation of the catalyst surface coverages, because it is never reached during normal RFO. Figure 6b also shows that during the whole cycle the outlet concentration of NH3 in the relaxed steady-state regime is lower compared to the outlet concentration of NH3 in the quasi-steady-state regime (see Figure 6a). Figure 7 provides a typical example of the NH3 partial pressure as a function of reactor length at discrete times in the dynamic regime. The profile at t ) 130 s is the final profile of the previous half-cycle just before reversing the flow direction. The peak value of this profile expresses the feed gas concentration at z ) 0.9, while the reactor outlet is at z ) 0. When the flow direction is now reversed, the inlet gas is fed at z ) 0.1 and the
Ind. Eng. Chem. Res., Vol. 43, No. 22, 2004 6961
Figure 8. Conversion of NH3 and selectivities to N2, N2O, and NO as a function of the feed position in terms of dimensionless reactor length at different switching times. Operating conditions as mentioned in Table 3.
reactor outlet at z ) 1. The steep gradient at z ) 0.10.2 will become less steep over time (131, 133, and 135 s) until at 140 s the profile’s shape is a mirror image of the profile at 130 s. A lower minimum at t ) 135 s compared to t ) 133 s is caused by the on-going reaction. The profile at 130 s in the region 1-0.9 indicates that the outlet concentration at z ) 1.0 first will rise and then decline when the region from 0.9 to 0.5 reaches the outlet. Finally, the concentration at z ) 1 rises again if the minimum value (see profiles at 133 and 135 s) has passed the outlet. It produces the concentration versus time behavior at the reactor outlet (z ) 1) as shown in Figure 5b2. Effect of the Feed Position on Conversion and Selectivity. Figure 8 shows time-average values of the conversion of NH3 and the selectivities toward N2, N2O, and NO as a function of the feed position at different switching times, which reflect the various operation regimes. In general, there are two effects that have an influence on the reactor conversion when shifting the feed positions to the reactor center. (i) The volume of the interrupted flow region increases, which causes a higher conversion due to longer residence time and less outflow of unreacted gas shortly after a flow reversal. (ii) The volume of the rest of the reactor decreases, which lowers the conversion obtained in this part.
The overall behavior is rather complicated because of the interaction of these effects. In the quasi steady state (see Figure 8a), the conversion of NH3 slightly decreases nonmonotonically when the feed positions are shifted to the reactor center. The behavior in the dynamic regime looks similar as in the quasi steady state, but the general trend is now increasing. The influence of the feed position in the quasi-steady-state and dynamic regimes is not impressing, contrary to the effects in the relaxed steady-state regime. Here, two different switching times of 5 and 2 s were applied. In both cases, the conversion of NH3 shows a large increase when the feed positions are shifted to the reactor center. The effect is most pronounced if the switching time becomes shorter, probably because of the increasing amount of flow interruptions in the interrupted flow region. RFO with reactor side feeding also affects the selectivity of the products. Although the conversion level itself has an influence on the selectivity, the differences in product distributions result from RFO, as shown by a comparison to the steady state at constant conversion.8 Figure 8b-d shows the effect of reactor side feeding on the selectivity. The selectivity to N2 increases slightly in the quasi-steady-state and dynamic regimes if the feed positions shift to the reactor center. It decreases, however, significantly in the relaxed steady-state re-
6962
Ind. Eng. Chem. Res., Vol. 43, No. 22, 2004
Figure 9. Conversion and selectivity at minimum switching time for RFO with reactor side feeding. Conversion and selectivities at steady state are indicated by marks on the axes: (b) NH3 conversion; (0) N2 selectivity; (9) N2O selectivity; (O) NO selectivity. For conditions, see Table 3.
gime, at high flow reversal frequency. The selectivity to N2O also decreases considerably in the relaxed steady-state regime, while the feed position hardly has an influence on the other regimes. Similarly, the selectivity to NO shows minor changes in the quasisteady-state and dynamic regimes if the feed positions shift to the reactor center, but the NO selectivity increases considerably in the relaxed steady-state regime. Therefore, the application of RFO with reactor side feeding may have an important positive effect on the conversion and the selectivity, if NO is the desired product, and production of N2O and N2 should be avoided. Conversion and Selectivity at Minimum Switching Time. For conditions of minimum switching time for each chosen feed position, Figure 9 shows the conversion of NH3 and the selectivities toward N2, N2O, and NO as a function of the dimensionless feed position for half of the reactor. Feed positions and the minimum switching time in Figure 9 relate to each other as shown via the bottom and top axis. This figure shows the locus
of conversion and selectivity at the minimum switching time. All minimum switching times applied in Figure 9 relate to the relaxed steady-state regime, which seems most interesting (see Figure 8). The reactor performance is even better than during normal reverse flow operation or steady-state, once-through operation, particularly when the feed positions are near to the reactor center. The conversion and selectivities during normal RFO under similar conditions are actually the values at the feed position z ) 0 in Figure 9, while steady state is indicated by marks on both axes. The conversion of NH3 increases if the feed positions are shifted to the reactor center. The increase of the NH3 conversion boosts up if the feed positions exceed La/Lr ) 0.3. Simultaneously, the selectivities for N2 and N2O decrease significantly, but the selectivity for NO increases. Apparently, the distribution of residence times for La/Lr > 0.3 is beneficial for ammonia conversion and NO production. This is due to the abundance of adsorbed O on the catalyst surface, as shown in Figure 10, which favors both conversion and NO formation. The left-hand side of Figure 10 shows how the O surface coverage changes during normal RFO over one cycle at different axial positions. The right-hand side shows similar data, but now with reactor side feeding at z ) 0.03 and 0.97. It is clear that side feeding induces coverages of 0.7 and more, while they are 0.6 or periodically much less during normal RFO. Conclusions A novel concept of a reverse flow operation with reactor side feeding was modeled and simulated for quasi-steady-state, dynamic, and relaxed steady-state regimes in the particular case of NH3 oxidation in order to study the effect of flow reversal on conversion and selectivity. The novel concept shows a promising way to affect the distribution of products as well as the conversion in the case of operation in the relaxed steadystate regime. Pronounced changes of conversion and selectivity are observed that cannot be reached by normal RFO or by steady state, once-through operation. Reactor side feeding allows adapting the residence time distribution, leading to higher oxygen surface coverage.
Figure 10. O surface coverage at normal RFO (a) and RFO with reactor side feeding (b) at discrete dimensionless axial positions as a function of time during one cycle. The switching time for normal RFO is 10 s, and it is 9.3 s for RFO with reactor side feeding. Other conditions as mentioned in Table 3.
Ind. Eng. Chem. Res., Vol. 43, No. 22, 2004 6963
The financial support, provided by Novem (The Netherlands Agency for Energy and Environment), STW (Dutch Technology Foundation), and Quality for Undergraduate Education Project, Chemical Engineering, Institute of Technology Bandung, Indonesia, is gratefully acknowledged.
i ) component in the gas phase j ) component in the solid phase k ) elementary step, k ) 1, 2, 3, ..., 13 max ) maximum min ) minimum 0 ) steady state, once-through operation r ) reactor s ) switching
Nomenclature
Literature Cited
Symbols
(1) Matros, Yu. Sh.; Bunimovich, G. A. Reverse-Flow Operation in Fixed Bed Catalytic Reactors. Catal. Rev. Sci. Eng. 1996, 38, 1. (2) Matros, Yu. Sh. Performance of Catalytic Processes under Unsteady Conditions. Chem. Eng. Sci. 1990, 45, 2097. (3) Agar, D. W.; Ruppel, W. Extended Reactor Concept for Dynamic deNOx Design. Chem. Eng. Sci. 1988, 43, 2073. (4) Snyder, J. D.; Subramaniam, B. Numerical Simulation of a Reverse-Flow NOx-SCR Reactor with Side-Stream Ammonia Addition. Chem. Eng. Sci. 1998, 53, 727. (5) Falle, S. A. E. G.; Kallrath, J.; Brockmuller, A.; Schreieck, A.; Giddings, J. R.; Agar, D. W.; Watzenberger, O. The Dynamics of Reverse Flow Chromagraphic Reactors with Side Stream Feed. Chem. Eng. Commun. 1995, 135, 185. (6) Noskov, A. S.; Bobrova, L. N.; Matros, Yu. Sh. ReverseProcess for NOx-off Gases Decontamination. Catal. Today 1993, 17, 293. (7) Mearns, A. M.; Ofosu-Asiedu, J. K. Ammonium Nitrate Formation in Low Concentration Mixtures of Oxides of Nitrogen and Ammonia. Chem. Technol. Biotechnol. 1984, 34a, 350. (8) Budhi, Y. W.; Jaree, A.; Hoebink, J. H. B. J.; Schouten, J. C. Simulation of reverse flow operation for manipulation of catalyst surface coverage in the selective oxidation of ammonia. Chem. Eng. Sci. 2004, 59, 4125. (9) Rebrov, E. V.; de Croon, M. H. J. M.; Schouten, J. C. Development of the Kinetic Model of Platinum Catalyzed Ammonia Oxidation in a Microreactor. Chem. Eng. J. 2003, 90, 61. (10) PDE Solutions Inc. FlexPDE 4. User Guide and Reference Manual; PDE Solutions Inc.: Antioch, CA, 2001.
Acknowledgment
Def ) axial dispersion coefficient (m2 s-1) E(t) ) residence time distribution function L ) reactor length (m) La ) feed position (m) Lr ) total reactor length (m) pi ) partial pressure of gas component i (atm) R ) ideal gas constant (m3 atm mol-1 K-1) rk ) reaction rate of step k (s-1) t ) time (s) ts ) switching time (s) T ) temperature (K) u ) superficial velocity (m s-1) z ) dimensionless axial position Greek Letters b ) bed porosity (mgas3 mreactor-3) Fb ) bed density (kgcat mreactor-3) δ ) number of flow interruptions θj ) surface coverage of component j σ ) catalyst mol site (mol kgcat-1) τ ) residence time (s) Superscripts hs ) hollow site o ) feed ss ) steady state ts ) top site Subscripts b ) bed
Received for review April 13, 2004 Revised manuscript received July 9, 2004 Accepted July 23, 2004 IE049702D
