CATALYSIS AND AUTOMOTIVEPOLLUTIONCONTROLIV Studiesin Surface Science and Catalysis, Vol. 116 N. Kruse, A. FrennetandJ.-M Bastin (Eds.) 91998 Elsevier Science B.V. All rights reserved.
Kinetic study of the selective catalytic reduction of nitric oxides with hydrocarbon in diesel exhausts Bj6rn Westerberg 1'3, Bengt Andersson 1, Christian Kiinkel2 and Ingemar Odenbrand 2 IDepartment of Chemical Reaction Engineering, Chalmers University of Technology, S-412 96, G6teborg, Sweden 2Department of Chemical Engineering II, Lund University, Institute of Technology, P.O.Box 124, S-221 00, Lund, Sweden 3Competence Center for Catalysis, Chalmers University of Technology, S-412 96, G6teborg, Sweden
ABSTRACT The kinetics of the selective catalytic reduction of nitric oxides (NOx) on a proprietary high temperature catalyst with diesel as the reductant have been studied. The objective was to derive a kinetic model that can be used for real time simulation of the catalyst. In the extension, the real time simulation will be used when controlling the injection of reductant. This is a requirement for achieving a high efficiency and a low fuel penalty. The response time and the NOx conversion level upon transient diesel injection was found to be dependent on the temperature. At temperatures below 570 K very low or no NOx conversion was observed. Above 570 K a small conversion was observed. No direct response upon diesel injection could be distinguished and the NOx conversion was independent on the hydrocarbon concentration. As the temperature was increased the response became apparent and then faster and the conversion level gradually became more dependent on the hydrocarbon concentration. Above 700 K the response was immediate (response time less than 15 s) and the conversion level was directly dependent on the hydrocarbon concentration. It was concluded that the NOx reduction proceeds via the formation of a hydrocarbon intermediate and the successive reaction between the hydrocarbon intermediate and NOx. When this reaction mechanism was modeled many features of the catalyst behaviour were reproduced.
1. INTRODUCTION The diesel engine has many advantages when used as the power source in heavy transport vehicles, but a disadvantage is the emission of pollutants. As the exhaust emission limits become stricter, the need for more effective emission control systems becomes urgent.
318 Particulates, carbon monoxide and hydrocarbons can be removed with particulate filters and oxidation catalysts, but none of these systems can effectively reduce the NOx emissions. High NOx conversion levels (60-70%) have been achieved with systems that use ammonia [ 1] or urea as a reductant, but this technology has some disadvantages. A distribution chain to supply the reductant, and a reservoir to keep it onboard the vehicle will be required. Another necessity is a reliable control system  or an oxidation catalyst  to avoid ammonia slip. A more attractive alternative may be a system that uses the fuel, already available, to reduce NOx emissions. In this study the kinetics of the NOx reduction on a proprietary high temperature catalyst with diesel as the reductant was examined. The objective was to derive a kinetic model that can be used for real time simulation of the catalyst. In the extension, the real time simulation will be used when controlling the injection of reductant. This is a requirement for achieving a high efficiency and a low fuel penalty.
2. EXPERIMENTAL Transient experiments were performed on a 12 1 heavy duty diesel engine, with a 24 1 monolithic catalytic converter connected to the exhaust pipe. The catalytic converter contained two different catalysts, both supplied by Johnson Matthey. These were an 18 I high temperature active (HT) catalyst placed upstream and a 6 1 low temperature active (LT) catalyst placed downstream. The HT catalyst provides the main capacity for NOx reduction. The LT catalyst combusts unreacted hydrocarbon from the HT catalyst and contributes with some NOx reduction at lower temperatures. As this study only concerns the performance of the HT catalyst, the LT catalyst will not be discussed further, and the HT catalyst will be referred to simply as the catalyst. Diesel was injected with an air assisted spray nozzle placed 2 m upstream of the catalyst. A 1.5 mm K-type thermocouple provided the exhaust temperature before the catalyst and a 2 mm K-type thermocouple provided the temperature after the catalyst. Sampling of the exhaust were done before and after the catalyst. The sampled gas was led through heated pipes and passed through a J.U.M. Engineering model 222 heated gas pre filter before passed to the analyzing equipment. With a switching valve before the gas filter, sampling before or after the catalyst was selected. The NOx content was determined with a TECAN CLD 700 EL ht chemiluminescence detector and the hydrocarbon content was determined with a J.U.M. Engineering model VE5 FID detector. To evaluate the catalyst and to provide data for a kinetic model a specially designed test cycle was used. The engine was run at different speeds and loads as specified in table 1. The space velocities ranged from 35 000 to 150 000 h -l. The first step in the cycle was selected to provide equal starting conditions between different runs. Desorbing accumulated hydrocarbon, burning off carbonaceous deposits was done by heating up the catalyst to a temperature equal between different runs. Step 2-6 in the test cycle were selected to provide different NOx concentrations and mass flows and a temperature that varied during the steps. During each step hydrocarbon transients, with a duration of one or two minutes, were introduced by injecting diesel before the catalyst. Two runs of the test cycle were performed. In the first run the flow of injected hydrocarbon (as Cl) was twice the NOx flow in mole/s, and in the second run this ratio was four.
319 Table 1. LSpecificmion of the test cycle. Step Time Speed (min) (rpm) 1 0-10 1920 2 10-20 1920 3 20-30 1920 4 30-40 1260 5 40-50 1260 6 50-60 1260
Load (Nm) 900 100 500 100 500 900
Exhaust Flow (mole/s) 11.2 7.5 9.0 4.5 5.0 6.0
Temperature (K) 450-810 810-530 530-680 680-470 470-590 590-730
NOx (ppm) 450 140 340 170 600 880
3. RESULTS AND DISCUSSION Figure 1 shows the hydrocarbon concentration at the catalyst inlet and outlet during the first and the second run of the test cycle. The hydrocarbon conversion during diesel injection varied from 40% at low temperatures, to 85% at high temperatures. The response in hydrocarbon concentration at the outlet when diesel was injected was slow at low temperatures and fast at high temperatures. When the injection was interrupted a tailing was observed that was more pronounced at lower temperatures. This indicates that the hydrocarbon both adsorb and desorb from the catalyst. 12000
E O. v
~: __ Ou__ tlet2nd
I LJL_J .
, - . 10
Figure 1. Hydrocarbon concentration at catalyst inlet and outlet during the first and the second run of the test cycle. The outlet 2nd, the inlet 1st and the inlet 2nd curves are offset by 2000, 4000 and 6000 ppm respectively.
Figure 2 shows the temperature at catalyst inlet and outlet during the first and the second run of the test cycle. The outlet temperature followed the inlet temperature with a lag of about one minute. Due to external heat losses, the outlet temperature never reached the inlet
320 temperature, except during diesel injection. At high temperatures an outlet temperature peak could be seen during diesel injection. This was an effect of the evolved heat from the hydrocarbon conversion. When injection was done in the 40-50 minutes' interval, this effect could barely be seen. The temperature here was between 560 and 590 K. These temperatures are just below the reported value of the ignition temperature for a fresh Cu zeolite catalyst . Below this temperature no increase in the outlet temperature could be seen during diesel injection. 1100 1000
~-~ 700 600 500 400 0
Figure 2. Temperature at catalyst inlet and outlet during the first and the second run of the test cycle. The outlet 2 nd and the inlet curves are offset by 100 and 200 K respectively.
Figure 3 shows the NOx concentration at the catalyst inlet and outlet during the first and the second run of the test cycle. At the end of the 0-10 minutes' interval the temperature was just above 800 K and the response time when diesel was injected was less than 15 s. The NOx conversion was 8% in the first run and 18% in the second run. In the beginning of the 10-20 minutes' interval the temperature fell rapidly towards 550 K and only during the first three minutes a small NOx conversion (9% in both runs) was observed. In the 20-30 minutes' interval the temperature rise to 680 K and a NOx conversion of 15%, in the first run, and 20%, in the second run, was observed before any diesel had been injected. When diesel was injected the NOx conversion remained at 15% in the first run, and increased to 24%, in the second run. The response time was 45 s. The 30-40 minutes' interval had the lowest temperatures in the test cycle (below 500 K at the end of the interval) and only during the first injection a small NOx conversion were observed (6 and 9% in respective run). In the 40-50 minutes' interval the temperature was increased to 590 K. No direct response upon diesel injection could be distinguished, but a continuos and increasing NOx conversion (from 5 to 10% in both runs) could be observed during the interval. In the last 10 minutes' interval the temperature was increased to 720 K. The first minutes, before any diesel injection was done, a relatively high NOx conversion could be observed. The conversion peaked at 20%, in the first run, and at 26%, in the second run. When diesel was injected the conversion was 20 and 35% in
321 respective run. The response time was 40 s for the first injection and 25 s for the second injection. 2000
E 1600 ck t.--~ 1200
tJ-- Outlet 2 n d ~
Outlet, 1st ~
Time (min) Figure 3. NOx concentration at catalyst inlet and outlet during the first and the second run of the test cycle. The outlet 2 nd and the inlet curves are offset by 400 and 800 ppm respectively.
An interesting trend can be observed when examining the catalyst behaviour at different temperatures. At temperatures below 570 K very low or no NOx conversion was observed. Above 570 K a small NOx conversion was observed. No direct response upon diesel injection could be distinguished and the conversion was independent on the amount of hydrocarbon injected. As the temperature was increased the response became apparent and then faster and the conversion level gradually became dependent on the amount of injected diesel. Above 700 K the response was immediate and the conversion level was directly dependent on the amount of injected diesel. From these observations it can be concluded that the NOx reduction does not proceed via the direct reaction between NOx and hydrocarbon. Instead they suggest that the hydrocarbon first form an intermediate, and that NOx is reduced when it reacts with this intermediate. At lower temperatures the formation of the intermediate is slow, but the consumption is even slower, so the intermediate will accumulate on the surface. When the temperature is increased the accumulated intermediate is consumed accompanied by a simultaneous NOx reduction. If the coverage of the intermediate is high, as it will be after extended times at low temperatures, the NOx reduction can be quite significant. At elevated temperatures the formation and consumption of the intermediate balances. A prolonged time at the same conditions will yield a steady state coverage. As a consequence the response to a change in the hydrocarbon concentration will be slow. At high temperatures the consumption of the intermediate is faster than the formation and th e coverage will be small. The rate limiting step in the NOx reduction is the formation of the intermediate. These observations agrees with the findings of Ansell et. al.  in their study of the classical Cu/ZSM-5 catalyst. They observed that carbonaceous deposits (coke) is deposited on
322 the catalyst when exposed to a lean propene/oxygen feed and that the coke is active in the lean NOx reaction. They also showed that the deposited coke is burnt off in oxygen. They assumed that coke is formed on the acidic sites of the zeolite and that the NOx reduction takes place when NO2 reacts with the coke. They also assumed that NO is converted into NO2 on the Cu sites, and that oxygen is essential in this process. Bennet et. al.  found earlier that the NOx conversion on a Cu/ZSM-5 catalyst is first order dependent on the propene pressure and zero order dependent on the NO pressure. They suggested a mechanism in which the hydrocarbon generates a reactive intermediate capable of reducing NO. An attempt was made to model the studied catalyst. It was assumed that hydrocarbon (HC) adsorbs and forms an intermediate (HC*) which either reacts with NOx or oxygen. It was also assumed that NO and NO2 can be treated as the same species, i.e. NOx. This can be justified if the NOx reduction proceeds via the reaction between the hydrocarbon intermediate and NOz, and if the conversion of NO into NO2 is not a rate limiting step. The oxidation products were assumed to be CO2 and H20. Formation of CO and the successive oxidation to CO2 probably occur as well, but has been omitted in the model. The model also assumes that the reaction rates are independent on the oxygen concentration. S 1 + H C ---> S~ - H C
r~ = k~c~cO,,~
S l - H C --> S~ + H C
r2 = kEOl,nc
S~ - H C --> S~ - H C *
r 3 = k3OI,Hc
r4 = k4Ol,ttc,
r5 = k s c Nox O...c.
HC* + 0 2
--> S 1 dl- C O 2 -~- HEO
S~ - H C * + N O x --~ S~ + C O z + 1 1 2 0 + N 2
When fitting this mechanism only, large residuals were attained for the hydrocarbon concentration. In order to obtain a better fit a second site with hydrocarbon adsorption and oxidation was introduced. It consisted of the following steps: S 2 + H C ---) S 2 - H C
r 6 ---- k6Cl_1CO2,v
S 2 - H C --) S 2 + H C
r7 = k702.,c
S 2 - HC + 02 --, S 2 + CO z + H 20
r8 = k802,,c
The monolith was modeled with a one dimensional model. The following simplifications have been made in the model: a) b) c) d) e) f)
uniform radial flow distribution negligible radial temperature and concentration profiles no axial diffusion or heat conduction no gas phase accumulation no diffusion resistance in the washcoat transfer of mass and energy between gas and solid is accounted for by coefficients derived from the correlation obtained by Tronconi and Forzatti  g) the monolith is treated as a series of continuously stirred tank reactors
323 The following equations were used to model a differential axial monolith segment: Gas mass balance: (9)
Fi,~_ l - Fi, k - k ~ A k (Cg,i,k -- Cs,i, k ) -- 0
Surface mass balance:
kcAk (cg.~.k --C.,..,,k) = ~_~ vi,,,r, mwc,k
Gas energy balance: F~c p,i ( T~ ,k-, - T~ ,~, ) - hA~, ( T~ ,k - ~,k ) =0
Solid energy balance:
" Ot - hAk (T~'k - ~"k ) + ~-'r"mc'k ( - A H " ) - k f A[ (T"'k - T")
The preexponential factors and the activation energies of the reactions were fitted to the experimental data of the second run of the test cycle. The values of these parameters can be found in table 2.
Table 2. Parameters obtained from fitting the model to experimental data. Reaction Preexponential factor Activation energy number (k J/mole) 1 7.8 x 10 ~ m 3 kg cat. l s l 14 2 9.4 x 101 mole kg cat. -l s -l 51 3 2.4 x 103 mole kg cat. -l s l 69 4 4.0 x 104 mole kg cat. ~ s -I 71 5 2.7 x 108 m 3 kg cat. -l s ~ 97 6 8.4 x 10 3 m 3 kg cat. l s -l 32 7 1.8 x 101 mole kg cat. l s l 26 8 3.9 x 102 mole kg cat. l s l 60
Figure 4 shows the observed and the simulated hydrocarbon concentration at the catalyst outlet during the second run of the test cycle. The standard deviation for the residual is 118 ppm or 18% of the mean HC concentration. The modeled concentration follows the observed
324 with some exceptions. During injection in the 0-10 minutes' interval and during the second injection in the 20-30 minutes' interval the model predicts too low conversion. In the 10-20 minutes' interval the model predicts too low conversion between the injections. There is also a slighter deviation in the conversion during the first and second injection in the 10-20 minutes' and in the 30-40 minutes' interval. There are also deviations at the flanks of the hydrocarbon transients in the 40-50 and the 50-60 minutes' interval. One explanation to the deviations could be that the model treats all hydrocarbon as a single compound. This is a coarse simplification. The diesel fuel itself consists of a variety of larger hydrocarbons which are cracked into shorter ones in the catalyst. All these different hydrocarbons have different adsorption properties and reactivities. An improved model would need to distinguish between different hydrocarbons or at least groups of them. Another improvement would be to account for variations in the oxygen concentration or even include oxygen adsorption in the model.
200 E 4000 El. Ex co
32 tO o n,
Figure 4. Observed and simulated hydrocarbon concentration at catalyst outlet during the second run of the test cycle. The simulated curve is offset by 2000 ppm.
Figure 5 shows the observed and the simulated NOx concentration at the catalyst outlet during the second run of the test cycle. The standard deviation for the residual is 33 ppm or 8% of the mean NOx concentration. The model manages to predict the NOx conversion that onsets before diesel injection in the beginning of the 20-30 minutes' and the 50-60 minutes' interval. During injection in the 0-10 minutes' interval and during the second injection in the 20-30 minutes' interval the model predicts too low conversion. This coincides with a predicted too low hydrocarbon conversion. The model also predicts a slightly too low conversion at the beginning and a slightly too high conversion at the end of the 40-50 minutes' interval. In the 50-60 minutes' interval the model deviates at the flanks of the hydrocarbon transients. There are also deviations at the end of each 10 minutes' interval, when the NOx inlet concentration is changed. These deviations could indicate that NOx adsorption and desorption occurs. The agreement between the modeled and the observed NOx concentration is to a large extent influenced by the deviations between the observed and the
325 modeled hydrocarbon concentration. An improvement of the model's ability to predict the hydrocarbon concentration would probably result in better predictions of the NOx concentration. Another improvement would be to distinguish between NO and NO2.
lOO AE 9
G) 0 tO
Figure 5. Observed and simulated NOx concentration at catalyst outlet during the second run of the test cycle. The simulated curve is offset by 400 ppm.
4. CONCLUSIONS It has been concluded that the reduction of NOx on a high temperature catalyst proceeds via the formation of a hydrocarbon intermediate and the successive reaction between the hydrocarbon intermediate and NOx. When this reaction mechanism was modeled many features of the catalyst behaviour were reproduced.
5. N O M E N C L A T U R E A AP
c Cp Cp F h kc kf
mwc ms N
Channel wall area in monolith Peripheral area of monolith Gas concentration Gas heat capacity Solid heat capacity Molar flow Heat transfer coefficient Mass transfer coefficient Heat loss coefficient Mass of washcoat Mass of solid Number of sites
mole/s J/mole K J/kg K mole/s W/m E K m/s W/m 2 K kg kg mole/kg
326 r t Ta Tg Ts -All 0 v
Reaction rate Time Ambient temperature Gas temperature Solid temperature Heat of reaction Coverage Stoichiometric coefficient
mole/s kg washcoat S
K K K J/mole
Index: i j k n v
Specie number Site number Section of monolith Reaction number Vacant site
ACKNOWLEGDEMENTS The authors would like to thank: Johnson Matthey for supplying the catalysts for this study. AB Volvo for providing admittance to their engine laboratory. Bengt Cyr6n and Martin Bruszt for technical assistance. NUTEK for financial support.
1. S.L. Andersson, P.L.T. Gabrielsson and C.U.I Odenbrand, AIChE J., 40(11) (1994) 1911. 2. L. Andersson., "Mathematical Modeling in Catalytic Automotive Pollution Control", Ph.D. thesis, Department of Chemical Reaction Engineering, Chalmers University of Technology, Sweden, 1995. 3. C. Havenith, R.P. Verbeek, D.M. Heaton and P. van Sloten, SAE Technical Paper Series 952652 (1995). 4. K.M. Adams, J.V Cavataio and R.H. Hammerle, Appl. Catal. B, 10 (1996) 157. 5. G.P. Ansell, A.F. Diwell, S.E. Golunski, J.W. Hayes, R.R. Rajaram, T.J. Truex and A.P. Walker, Appl. Catal. N, 2 (1993) 81. 6. C.J. Bennet, P.S. Bennet, S.E. Golunski; J.W. Hayes and A.P. Walker, Appl. Catal. A, 86 (1992) L1. 7. E. Tronconi and P. Forzatti, AIChE J., 38(2) (1992) 201.