Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Chemical Engineering Science 59 (2004) 5277 – 5282

www.elsevier.com/locate/ces

From a fixed bed Ag–alumina catalyst to a modified reactor design: how to


enhance the crucial heterogeneous–homogeneous reactions in HC-SCR
K. Arvea,∗ , E.A. Popovb, c , M. Rönnholma , F. Klingstedta , J. Elorantab , K. Eränena ,
D. Yu. Murzina
a Laboratory of Industrial Chemistry, Process Chemistry Centre, Åbo Akademi University, Biskopsgatan 8, 20500 Turku/Åbo, Finland
b Department of Chemistry, University of Jyväskylä, Survontie 9, 40500 Jyväskylä, Finland
c On leave from Institute of Energy Problems of Chemical Physics, Chernogolovka, Russia

Received 27 February 2004


Available online 28 October 2004

Abstract
A highly active Ag/alumina catalyst for continuous reduction of NO to nitrogen with n-octane under lean conditions was prepared. It
was observed in the reactor set-up experiments for optimization of the converter, that surface generated gas phase species are crucial for
obtaining high conversion. EPR and matrix isolated FTIR studies at low temperature (10–18 K) were performed for identification of the
radicals. Experimental data, observed at steady state conditions in the temperature range 300–550 ◦ C, was used to produce an artificial
neural network model of the catalytic converter with four catalyst beds.
䉷 2004 Elsevier Ltd. All rights reserved.

Keywords: HC-SCR; Ag/alumina; Catalysis; Design; Reactor

1. Introduction of NOx using hydrocarbons. Earlier it has been shown


(Eränen et al., 2000; Hoost et al., 1997; Jen, 1998; Kung and
The worldwide concern about the increased carbon diox- Kung, 2000; Martínez-Arias et al., 2000; Miyadera, 1993;
ide emissions from traffic has resulted in the development Shimizu et al., 2001) that Ag/alumina exhibits high and
of new, more fuel efficient vehicles. Lean burn engines, run- stable reduction of NOx to N2 with various hydrocarbons.
ning on diesel or gasoline, operate under large excess of The drawback, however, is the simultaneous formation of
oxygen, which result in lower fuel consumption and thus CO as well as the poor low temperature activity of the cat-
the amount of CO2 emitted is also decreased. However, be- alyst. Especially, the insufficient low temperature activity is
cause of the highly oxidising conditions, reduction of nitro- a significant disadvantage, as a large part of the emissions
gen oxides (NOx ) is not possible with ordinary three way will be emitted during cold starts and idling periods. Eränen
catalysts (TWC) designed to work at stoichiometric condi- et al. (2003) showed that placing a commercial oxidation
tions. Therefore, selective catalytic reduction by hydrocar- catalyst behind the Ag/alumina bed could fully treat the
bons (HC-SCR) has attracted much attention as a potential problem with CO production. They also observed a dramatic
method to reduce NOx emissions. change in the NOx to N2 activity over the catalytic system,
In this paper, we have investigated the application of depending on the distance between the Ag/alumina and the
a silver–alumina catalyst for selective catalytic reduction oxidation catalyst. When the two catalysts were mixed or
placed directly after each other, the overall conversion was
much lower than in the case where the distance between
∗ Corresponding author. Tel.: +358 2 215 4431; fax: +358 2 215 4479. the beds was gradually extended up to 33 mm. This proves
E-mail addresses: karve@abo.fi (K. Arve), eloranta@jyu.fi that the reduction of NOx is not purely a heterogeneous cat-
(J. Eloranta). alytic surface reaction, but a combination of heterogeneous
0009-2509/$ - see front matter 䉷 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2004.08.036
5278 K. Arve et al. / Chemical Engineering Science 59 (2004) 5277 – 5282

surface reactions and homogeneous gas phase reactions. silver content in the precursor solution prior and after the
Cavani and Trifiro (1999) have shown, that heterogeneous impregnation.
surfaces can generate free radicals with further desorption
of them into the gas phase. In addition, Klingstedt et al. 2.2. Activity tests
(2004) have earlier shown that by dividing the catalyst bed
in four layers instead of one single bed a higher overall con- The crushed and sieved Ag/alumina (0.4 g, 250–500 m)
version of NOx was achieved and moreover, the conversion catalyst either as a single bed or divided into four layers,
increased in the low temperature region. was tested in a fixed bed quartz reactor inserted in an oven
For homogeneous reactions, in particular radical reac- equipped with a temperature controller. A temperature range
tions, the size and shape of the reactor are of crucial impor- of 150–600 ◦ C with sampling at steady state conditions with
tance. For example, residence time controls the formation GHSV = 60 000 h−1 and total flow rate of 550 ml/min was
of the critical concentration of radicals for the reaction to used in activity tests. The gas mixture consisted of 500 ppm
occur, or ensures that the concentration of radicals cannot NO, 375 ppm octane, 6 vol% O2 , 10 vol% CO2 , 350 ppm
reach the critical level leading to explosion. Therefore, it is CO, 12 vol% H2 O and He as balance. All the gases were of
necessary to understand how different species on the cata- high purity (AGA) and were introduced into the reactor by
lyst surface and in the gas phase are involved in the reaction means of mass flow controllers (Brooks 5850). The addition
pathway and for this purpose kinetic data is essential. How- of octane took place using a syringe pump (CMA 102/Mi-
ever, before any accurate kinetic modelling can be done, crodialysis). Oxygen was fed separately into the reactor to
identification of the intermediates, by-products and possible avoid oxidation of NO before the catalyst bed. The efflu-
spectators is needed. Experimental detection of free radicals ent gas was analysed by a chemiluminescence NOx -analyser
formed in heterogeneous processes cannot be done only with (API model 200AH) and a GC (HP 6890 series) equipped
surface sensitive methods, but should be applied together with a GS Q column, a GS Molesieve column (J&W Scien-
with some specific procedures, for instance matrix isola- tific) and FI as well as TC detectors. High purity calibration
tion of radicals, combined with IR and EPR spectroscopy, gases (AGA) were used for calibration of the NOx -analyser
photoelectron spectroscopy, multiphoton or resonance ioni- and the gas chromatograph.
sation methods (Andrews and Moskovits, 1989; Eloranta et Special experiments were carried out to ensure that the
al., 1998). After the identification of the radicals it is possi- kinetics is measured in the absence of external and internal
ble to apply rigorous kinetic modelling to optimise the re- diffusion limitations. The impact of internal mass-transfer
actor design. Alternatively, neural network models, which resistance over the Ag/alumina was tested using particle
are flexible mathematical tools helping in predicting dif- sizes of 1.0–1.41 mm, 0.250–0.500 mm and 0.09–0.180 mm
ferent behaviours in processes and the interaction between with equal mass of catalyst. The external mass transfer was
multiple variables can be applied. Rönnholm et al. (2003) verified using three different lengths of the Ag/alumina bed:
have reported a trained neural network model over a single 0.35, 0.7 and 1.05 cm and keeping the gas flow/catalyst
bed Ag/alumina catalyst at different temperatures. The pre- volume ratio constant. The results confirmed that there are
sented model predicted the reactor concept when a simple no internal and external mass transfer limitations over the
model was applied, but a complex model resulted in over- Ag/alumina catalyst.
fitting. This study presents a neural network model trained
for the optimal four bed reactor concept. Improved conver-
sion values due to more complete gas phase reactions are 2.3. EPR and FTIR experiments
implemented as output data for the neural network model.
For EPR measurements two bulbs containing: (a)
1000 ppm NO, 750 ppm n-octane in Ar and (b) 60 000 ppm
2. Experimental O2 in Ar were prepared. The size of the bulbs was 1 dm3
and the total pressure was 1000 mbar. The mixtures were led
2.1. Catalyst preparation in separate lines to the reactor inlet, where they were mixed
and thus the concentrations were 500 ppm NO, 375 ppm
An alumina supported silver catalyst was prepared by im- C8 H18 and 30 000 O2 . The temperature of the catalyst was
pregnation (24 h) of a commercial (LaRoche) alumina sup- 350 ◦ C and flow rate of 2 Torr/min was fixed by pressure
port (< 250 m), which is a mixture of , ,  and pseudo- controllers (MKS Baratron Capacitance Manometers, type
 phases (289 m2 /g), with aqueous silver nitrate solution 622). The deposition time was 15 min resulting in a total
(0.022 M, pH ≈ 5) of high purity (J.T. Baker). After impreg- deposition of 30 Torr. Behind the catalyst bed the effluent
nation the catalyst was filtered and dried for 24 h at room gas was trapped immediately in the growing argon matrix at
temperature. Thereafter, the catalyst was dried 3 h at 100 ◦ C 18 K and the spectrum was scanned using EPR spectrome-
and calcination was carried out for 3 h at 550 ◦ C in air. Sil- ter (Bruker ESP-380). After this the matrix was annealed to
ver content, 1.91 wt%, was analysed with XRF technique 34 K and re-scanned. A reference run without the catalyst
(XRF: X-MET 880, Outokumpu electronics) by measuring was measured using the same temperature and gas mixtures.
K. Arve et al. / Chemical Engineering Science 59 (2004) 5277 – 5282 5279

For FTIR experiments two gas mixtures: (a) 1500 ppm Table 1
C8 H18 , 12 vol% O2 and Ar as balance and (b) 2000 ppm The NOx to N2 activity over a single bed Ag/alumina catalyst vs. 4-layer
bed
NO and Ar as balance were prepared. The total pressure
in the glass bulbs was 1000 mbar. The gas mixtures were Temp. NOx to N2 conversion (%)
led in separate lines to the reactor inlet, where they were (◦ C)
Single bed 4-layer
mixed. Flow rates were fixed to 1 Torr/min according to the
method described for EPR. The total amount of deposition 150 0 4.1
200 0 3.4
was 50 Torr and the effluent was trapped immediately be- 250 0 4.9
hind the catalyst in the growing argon matrix at 10 K in 300 6.5 36.4
the cryostat (Displex DE-202S (10 K)). After the deposition 350 33.3 72.3
the spectra were recorded using FTIR spectrometer (Nicolet 400 62.3 92.0
Magna-IR 760 ESP). The spectra were scanned using cata- 450 72.2 91.7
lyst temperatures 250 and 400 ◦ C during the deposition and
500 74.5 78.1
550 66.0 60.7
following gas mixtures: octane + O2 , NO + O2 and octane + 600 46.2 40.3
O2 + NO. A background spectrum was scanned before each
experiment.
vs. catalyst divided in four beds (1.5 cm between the beds),
2.4. Activity tests for neural network modelling keeping the total amount of the catalyst constant. With the
single bed arrangement an overall conversion of 60% was
To apply a neural network model on the system, sepa- reached in the temperature range 150–600 ◦ C. However, the
rate activity tests were performed applying the four beds visible NOx to N2 reduction begins first at 350 ◦ C, where the
reactor concept in the temperature range 300–550 ◦ C with conversion is approximately 35%. When the activity pattern
50 ◦ C steps. In these experiments data points were obtained was recorded over the same catalyst, divided in four layers,
in (1) series including variations of either nitrogen monox- thus enabling the gas phase reactions, a clear difference in
ide or octane concentration and keeping the concentration activity patterns was observed (Table 1). With this modifica-
of O2 constant and (2) by changing the C8 H18 /NO concen- tion of the reactor system the activity was undisputable im-
trations pair-wise and keeping the O2 concentration con- proved. The conversion started at lower temperature and in
stant and (3) keeping both NO and C8 H18 at a constant addition the conversion increased in the temperature range
level and varying the incoming O2 concentration. Following 300–550 ◦ C.
gas concentrations were applied: 375–500–1000 ppm NO, From the results it can be concluded that the overall con-
134–187.5–375–500–750 ppm C8 H18 and 1–12 vol% O2 . version rate over the catalyst is strongly dependent on the
residence time behind the catalyst bed. Moreover, the results
2.5. Neural network modelling from a reference run done with an empty reactor showing
zero activity in terms of NOx to N2 conversion prove, that
The results obtained from the concentration variation ex- the HC-SCR over a Ag/alumina catalyst using n-octane as a
periments were investigated using a neural network model. reducing agent is initiated on the catalyst surface and con-
Training and testing of the networks were performed by tinues in the gas phase forming molecular nitrogen.
means of the NNDT software (Saxén and Saxén, 1995), us-
ing the Marquardt algorithm and cross validation. The ex- 3.2. EPR and FTIR
periments were divided into one group with 83 experiments
and 4 outputs for the training and another group with for the The first intensity spectrum of EPR was obtained from
testing with 8 experiments and 4 outputs. the matrix formed when NO, octane and oxygen where
The input and output nodes had linear activation func- fed through the Ag/alumina catalyst at 350 ◦ C and trapped
tions, whereas sigmoids from 0 to 1 were used for the hidden in the argon matrix at 18 K. Characteristic for the spectra
nodes. All input and output data were scaled to the interval were triplets, which most probably mean that they belong to
[0,1] and all the weights were initialized by a random gen- species, containing N nucleus. The main component trapped
erator before starting the training of the network. The inputs in the matrix at 18 K was identified as NO2 . The identifica-
were temperature, NOin , octanein and O2 and the outputs tion of NO2 is based on a reference run without the catalyst.
were N2 out , COout , O2 out and CO2 out . NO2 trapped in argon matrix at 18 K has pronounced broad-
ening and anisotropy. Line-shape simulation of the experi-
mental spectrum has confirmed its assignment as randomly
3. Results and discussion
oriented NO2 . The best fit was obtained by choosing the
3.1. Activity tests principal values of g- and A-tensors (1.991, 2.0025, 2.0052)
and (45.5, 63.0, 50.7 G), respectively.
Table 1 depicts the activity pattern recorded over the When the matrix was annealed to 34 K and scanned again,
Ag/alumina catalyst placed in the reactor tube as a single bed a new spectrum was obtained. It comprised a triplet, centered
5280 K. Arve et al. / Chemical Engineering Science 59 (2004) 5277 – 5282

6000 Am=1 G
Pmw=0.2 mW
4000 Tanneal=34 K
Trecord=22 K
2000
Intensity (Arb. units)

-2000

-4000

-6000

-8000

-10000 ge

3300 3350 3400 3450 3500


Magnetic field (Gauss)

Fig. 1. EPR spectrum upon warming the deposit to 34 K. 1000 ppm NO,
750 ppm octane and 6 vol% O2 .

Fig. 2. FTIR spectra upon the switching from reactant mixture. (a) n-octane
and O2 , 250 ◦ C, (b) NO and O2 , 250 ◦ C and (c) n-octane, NO and O2 ,
at g = 2.0079 with hyperfine splitting constant Aiso = 8.3 G 250 ◦ C.
(Fig. 1). Its appearance clearly manifests the presence of
active intermediates, which are first generated on the cata-
lyst surface, subsequently desorbed into the gas phase and
finally trapped in Ar matrix. In argon matrix, the thermal peak was formed in the root of the NO peak at 1871 cm−1 .
activated diffusion starts above 30 K promoting low temper- This peak can be attributed to (NO)2 complex (Colthup
ature chemical reactions and leading to formation of new et al., 1990) due to dimerisation of NO upon annealing.
radical species. An important observation was made when the matrix was
It should be stressed here that at such low temperatures annealed further to 30 K. In this spectrum the (NO)2 peak
any chemical reactions between valence saturated species vanished totally and a new clear peak at 1832 cm−1 was
are strongly supressed. Additionally, translational diffusion formed. This peak is attributed to asymmetric N2 O3 (Wang
of rather heavy molecules such as NO2 is extremely slow. and Quin, 1999).It is worth mentioning that this peak was
Therefore, the triplet in Fig. 1 could be tentatively ascribed not visible at 10 K. N2 O3 is in equilibrium with NO and
to formation of NO3 radical, via reaction of NO2 + O. Such NO2 and the equilibrium favours N2 O3 formation at low
assignment would indicate that there is O atom isolated in temperatures (Hägg, 1963). On the other hand, reaction
the matrix. Another possible explanation would be a reac- between NO and NO2 to form N2 O3 should be very slow
tion between NO and some small hydrocarbon clusters with at the temperature of the matrix. Anyway the detection of
formation of a nitroxyl-like radical O–N–(R, R ) with un- N2 O3 indicates that both NO and NO2 are trapped in the
paired electron density centered on oxygen atom. matrix. In addition, as no NO2 was detected in the matrix
The FTIR spectra recorded from the argon matrix are at 10 K and N2 O3 appears when annealed it is reasonable
presented in Fig. 2. When octane and oxygen were fed to suggest, that there are also O atoms trapped in the
at 250 ◦ C through the catalyst (Fig. 2a), CO2 bands were matrix, which react with NO to form NO2 . Formation of
recognised from the characteristic very strong peaks at isolated O atoms was also proposed based on our EPR
2349 cm−1 and at 667 cm−1 in the gas phase (Shimanouchi, studies.
1972a). Deposition of nitrogen oxide and oxygen (Fig. 2b) At 400 ◦ C with the complete gas mixture passing through
led to a very strong peak at 1610 cm−1 and at 750 cm−1 , the catalyst following new peaks were observed compared to
which correspond almost perfectly to characteristic peaks the spectrum scanned at catalyst temperature 250 ◦ C: 2282,
for NO2 at 1618 and 750 cm−1 in the gas phase reported 2278, 1100, 833, 727 and 611 cm−1 . Two of the peaks,
by Shimanouchi (1972b). When the complete (C8 H18 , O2 2278 and 727 cm−1 , belong to cyanogen isocyanate, NC-
and NO) gas mixture was passed through the catalyst (Fig. NCO, (NIST Chemistry Webbook,). The rest of the peaks
2c), a new spectrum could be observed. The following new (2282, 1100, 833, 611 cm−1 ) are present also in the back-
peaks were detected: CN stretch at 2218 cm−1 and NO ground spectrum as rather sharp features and therefore can-
(N2 O2 ) stretch at 1871 cm−1 . The appearance of a CN group not be used for reliable identification. In (Eränen et al., 2004)
is in good agreement with observation made by Eränen it was shown that amine and ammonia are easily formed
et al. (2003), where they isolated nitrile species evolving from R-NCO over Ag/alumina in excess of oxygen. Amine
from the Ag/alumina catalyst surface during HC-SCR at and ammonia then react in the gas phase with activated forms
400 ◦ C. Upon annealing of the matrix to 20 K a satellite of NOx after the catalyst bed.
K. Arve et al. / Chemical Engineering Science 59 (2004) 5277 – 5282 5281

N2 Average model, temperature effects,


octane 750/500/200 ppm and
NO 1000/500 ppm at 60 000 ppm oxygen
1.0
0.9
0.8

NO to N2 conversion
octane 750 ppm, NO
NH3 and + Activated 0.7 1000 ppm
R-NH2 NOx 0.6 octane 500 ppm, NO
1000 ppm
0.5
octane 500 ppm, NO 500
0.4 ppm
0.3 octane 200 ppm, NO 500
0.2 ppm
0.1
0.0
250 350 450 550 650
o
Temperature/ C
R-CN R-NCO R-NO2 ad-NOx
Fig. 4. The modelled conversion of nitrogen oxide at different temperatures
Ag/alumina 250 - 400 °C giving the average value model for four beds. Based on experiments at
temperatures from 300 ◦ C to 550 ◦ C.

HC + O 2 + NO
1.0

Fig. 3. A simplified reaction scheme for C8 H18 -SCR over Ag/alumina


0.9
catalyst.

0.8

0.7
Based on the results obtained in this study as well as in
Estimated N2 out scaled

(Eränen et al., 2004) following simplified heterogeneous– 0.6


homogeneous reaction mechanism is proposed (Fig. 3).
In this mechanism different surface species such as nitro- 0.5
compounds, R-NCO and nitriles are formed through re-
0.4
action of hydrocarbons + NO + O2 over the Ag/alumina
and are further transformed into amine and/or ammonia, 0.3
which desorb into the gas phase. The ammonia and amine
species react in the gas phase together with activated NOx 0.2
species to form N2 . The activation of NO takes place over
0.1
the catalyst and is accelerated by suitable reducing agents.
Thus the gas phase reaction is a crucial part of the complete 0.0
HC-SCR mechanism. 0 0.2 0.4 0.6 0.8 1
Experimental N2 out scaled

3.3. Artificial neural network modelling


Fig. 5. N2 out from optimal network versus N2 out from experiment.
97.3% fit.
The training was performed at temperatures of 300, 400,
450, 500 and 550 ◦ C, while the test series were carried out
at 350 ◦ C. The previous model presented for a single bed
concept (Rönnholm et al., 2003) has been modified and ap- 4. Conclusions
plied for the four bed concept (Fig. 4). The rearrangement
of the catalysts beds in the reactor resulted in a neural net- Combined heterogeneous–homogeneous catalytic reac-
work model, which explains the process more accurate. The tions were significantly enhanced at temperatures below
overfitting appearing in the model applied for the single bed 350 ◦ C over a highly active 2 wt% Ag/alumina catalyst by
concept has vanished and in the new model the low and high changing the catalytic reactor construction from a single
temperature regions were fitted much better together (Fig. bed to a four bed concept. Matrix isolation technique was
5). The final artificial neural network consists of two paral- applied together with EPR and FTIR to study the impact of
lel networks, one for simple nitrogen conversion estimation the gas phase reactions on the NOx to N2 conversion. The
and another more complex for all the final products. How- matrix isolated deposition was found to contain low molec-
ever, more experiments with varying distance between the ular weight radicals as well as iso-cyanate species, which
catalyst beds are needed before the model can be used for seem to be crucial for the heterogeneous–homogeneous re-
the final optimization of the reactor construction. This is the action mechanism proposed. An improved neural network
scope of another ongoing study. model, showing the importance of low temperature region
5282 K. Arve et al. / Chemical Engineering Science 59 (2004) 5277 – 5282

and the gas phase reactions, was developed. The new model Hoost, T.E., Kudla, R.J., Collins, K.M., Chattha, M.S., 1997.
describing the four beds concept was superior compared to Characterization of Ag/-Al2 O3 catalyst and their lean-NOx properties.
the corresponding single bed model. Applied Catalysis B: Environmental 13, 59–67.
Jen, H.-W., 1998. Study of nitric oxide reduction over silver/alumina
catalysts under lean conditions: effects of reaction conditions and
support. Catalysis Today 42, 37–44.
Acknowledgements Klingstedt, F., Eränen, K., Lindfors, L.-E., Andersson, S., Cider, L.,
Jobson, E., Landberg, C., Eriksson, L., Ilkenhans, T., Webster, D.,
The financial support from the European Union through 2004. A highly active Ag/alumina catalytic converter for continuous
the Ammonore project (GRD-CT 2001-00595) is gratefully HC-SCR during lean-burn conditions: from laboratory to full scale
vehicle tests. Topics in Catalysis 30–31, 27–30.
acknowledged. This work is part of the activities at the Åbo Kung, M., Kung, H., 2000. Lean NOx catalysis over alumina-supported
Akademi Process Chemistry Centre within the Finnish Cen- catalysts. Topics in Catalysis 10, 21–26.
tre of Excellence Programme (2000–2005) by the Academy Martínez-Arias, A., Fernández-García, M., Iglesias-Juez, A., Anderson,
of Finland. J.A., Conesca, J.C., Soria, J., 2000. Study of the lean NOx
reduction with C3 H6 in the presence of water over silver/alumina
catalysts prepared from inverse microemulsions. Applied Catalysis B:
References Environmental 28, 29–41.
Miyadera, K., 1993. Alumina-supported silver catalysts for the selective
reduction of nitric oxide with propene and oxygen containing organic
Andrews, L., Moskovits, M., 1989. Chemistry and Physics of Matrix
compounds. Applied Catalysis B: Environmental 2, 199–205.
Isolated species, Elsevier Science B.V., Amsterdam.
NIST Chemistry Webbook. http://webbook.nist.gov/chemistry.
Colthup, N.B., Daly, L.H., Wiberley, S.E., 1990. Introduction to Infrared
Rönnholm, M., Klingstedt, F., Eränen, K., Lindfors, L-E., 2003. Artificial
and Raman Spectroscopy, Academic Press, New York.
neural network modelling applied on NOx reduction with octane in
Eloranta, J., Vaskonen, K., Hakkinen, H., Kiljunen, T., Kunttu, H., 1998.
excess oxygen over Ag/Al2 O3 . Reaction Kinetics and Catalysis Letters
193 nm photodynamics of NO in rare gas matrices. Fluorescence,
78 (2), 331–340.
thermoluminescence and photodissociation. Journal of Chemical
Saxén, B., Saxén, H., 1995. NNDT—A Neural Network Development
Physics 109, 7784–7792.
Tool, Proceedings of the International Conference on Neural Networks
Eränen, K., Lindfors, L.-E., Niemi, A., Elfving, P., Cider, L., 2000.
and Genetic Algorithms (ICANNGA’95), Alés, France, pp. 325.
Influence of hydrocarbons on the selective catalytic reduction of NOx
Shimanouchi, T., 1972a. Tables of Molecular Vibrational Frequencies
over Ag/Al2 O3 : laboratory and engine tests. SAE paper 2000-01-2813.
Consolidated Volume I. National Bureau of Standards, pp. 1–160.
Eränen, K., Lindfors, L.-E., Klingstedt, K., Murzin, D.Yu., 2003.
Shimanouchi, T., 1972b. Tables of Molecular Vibrational Frequencies
Continuous reduction of NO with octane over a silver/alumina catalyst
Consolidated Volume II. Journal of Physical and Chemical Reference
in oxygen-rich exhaust gases: combined heterogeneous and surface-
Data 6 (3), 993–1102.
mediated homogeneous reactions. Journal of Catalysis 219, 25–40.
Shimizu, K-i., Shibata, J., Yoshida, H., Satsuma, A., Hattori, T., 2001.
Eränen, K., Kligstedt, F., Arve, K., Murzin, D.Yu., 2004. On the
Silver–alumina catalysts for selective reduction of NO by higher
mechanism of the selective catalytic reduction of NO with higher
hydrocarbons: structure of active sites and reaction mechanism. Applied
hydrocarbons over a silver/alumina catalyst. Journal of Catalysis 227,
Catalysis B: Environmental 30, 151–162.
328–343.
Wang, W., Quin, Q-Z., 1999. Journal of Photochemistry and Photobiology
Hägg, G., 1963. Allmän och oorganisk kemi, Almqvist & Wiksell Förlag
A 122, 1.
AB, Stockholm.

You might also like