This article appeared in a journal published by Elsevier.

The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
 Author's personal copy

Applied Catalysis B: Environmental 88 (2009) 420–429

Contents lists available at ScienceDirect

Applied Catalysis B: Environmental

journal homepage: www.elsevier.com/locate/apcatb

A study of copper-exchanged mordenite natural and ZSM-5 zeolites as SCR–NOx

catalysts for diesel road vehicles: Simulation by neural networks approach
Mona Lisa Moura de Oliveira a, Carla Monteiro Silva a,*, Ramon Moreno-Tost b, Tiago Lopes Farias a,
Antonio Jiménez-López b, Enrique Rodrı́guez-Castellón b
a
IDMEC-Institute of Mechanical Engineering, Instituto Superior Técnico, Technical, University of Lisbon, Av. Rovisco Pais, 1049-001, Lisbon, Portugal
b
Departamento de Quı´mica Inorgánica, Cristalografı´a y Mineralogı´a, Unidad Asociada al Instituto de Catálisis, CSIC, Universidad de Malaga, Campus de Teatinos, 29071, Málaga, Spain

A R T I C L E I N F O A B S T R A C T

Article history: Copper-catalysts, based on the ZSM-5 (CuZSM5) and Cuban natural Mordenite (CuMORD) zeolites have
Received 7 August 2008 been prepared by a conventional ion-exchange method and their catalytic activity in the selective
Received in revised form 21 October 2008 catalytic reduction (SCR) of NO was studied using ammonia in presence of H2O and SO2. A commercial
Accepted 25 October 2008
catalyst SCR (CATCO) based on V2O5–WO3–TiO2, was also studied as reference. This paper presents
Available online 5 November 2008
experimental results using catalysts without the toxic vanadium and exploits a neural network based
approach to predict NOx conversion efficiency of three SCR catalysts. The derived mathematical functions
Keywords:
are integrated in a numerical model for diesel road vehicle simulation to simulate diesel vehicles
SCR
equipped with such SCR catalysts. The main results indicate that despite of toxic vanadium and N2O
NOx
Diesel road vehicle formation, CATCO shows the better NOx conversion efficiencies. However, CuMORD does not form N2O
Zeolites and have better performance than the CuZSM5. The simulation results show lower level of NOx for heavy-
Neural-networks duty and light-duty diesel vehicles compared with homologation load cycles.
Road vehicle simulator ß 2008 Elsevier B.V. All rights reserved.

1. Introduction not provoke an increase in fuel consumption or sensitivity to

sulphur as lean NOx trap technology does [3,4]. In particular, SCR of
The main environmental problems for diesel engines are NO using urea as reductant offers significant fuel savings (around
emissions of Nitrogen Oxides (NOx) and particulates (PM). 8% with respect to EGR systems), stable performance over
According to European Union statistics from 2003, transport extended mileage and compatibility with any engine setting
(mainly road transport) contributes to 54% of total NOx emissions, [2,3]. As a matter of fact, availability of low sulphur fuel is gradually
and industry to 23% [1]. NOx emissions from the exhaust gas of a increasing within the EU, with the implementation of Euro IV
diesel engine are usually composed of more than 90% of NO [2]. NO making it possible to fill up with low-sulphur diesel at most filling
is involved in the formation of ground-level ozone, acid rain, and stations in the EU countries (limited to a maximum of 10 ppm in
contributes to the greenhouse effect [3]. Due to its adverse effects 2009). This will be very important to assure the proper working of
in the human health and in the environment, more restricted the SCR system [5,6].
legislations concerning the emissions of NO from stationary and For light-duty diesel vehicles (LDD) a chassis dynamometer test
mobile sources have come out. However, advances in technology is used for emission testing and certification in Europe. It is
related to engines cannot further reduce NOx emissions so it is composed of four ECE (Economic Commission for Europe) driving
necessary to utilize after-treatment systems. cycle, Urban Driving Cycles (UDC), simulating city driving, and one
The SCR of NOx is the chosen technology by the majority of Extra Urban Driving Cycle (EUDC), simulating highway driving
diesel bus and truck manufacturers in Europe to meet the conditions. The cold-start version of the test, introduced in 2000, is
legislation (see Table 1, Figs. 1 and 2). This is due to the high also called New European Driving Cycle (NEDC), as it is shown in
NOx reduction achievable compared to the other NOx reduction Fig. 1 [5].
technology (exhaust gas regulations, EGR). It has been proven by The ESC (European Steady Cycle) test cycle was introduced in
track record that stationary applications and the SCR systems do Directive 1999/96/EC amending Directive 8/77/EEC, together with
the ETC (European Transient Cycle) and the ELR (European Load
Response) tests, for emission certification of heavy-duty diesel
* Corresponding author. engines (HDD). ESC test cycle comprises 13 steady-state loads, and
E-mail address: carla.silva@ist.utl.pt (C.M. Silva). the final result is a weighted average, as shown in Fig. 2a. The ETC

0926-3373/$ – see front matter ß 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.apcatb.2008.10.015
 Author's personal copy

M.L.M. Oliveira et al. / Applied Catalysis B: Environmental 88 (2009) 420–429 421

Table 1
Legislation for diesel vehicles NOx emissions in Europe.

Heavy-duty vehicle Cycles Light-duty vehicle [g/km] Cycles

[g/kWh]

2000 (Euro III) 5.0 ESC & ETC 2000 (Euro III) 0.50 NEDC
2005 (Euro IV) 3.5 ESC & ETC 2005 (Euro IV) 0.25 NEDC
2008 (Euro V) 2.0 ESC & ETC 2009 (Euro V) 0.18 NEDC
2013 (Euro VI)a 0.4 ESC& ETC 2014 (Euro VI)a 0.08 NEDC
a
Under discussion [5].

cycle has been developed by the FIGE Institute, Aachen, Germany,

based on real road cycle measurements of heavy-duty vehicles [6].
Different driving conditions are represented by three parts of the
ETC cycle, including urban, rural and motorway driving (see
Fig. 2b) [5,7].
In Europe, the ETC test has been required, since the date
mentioned above, for all new gas engine types, and for all new
diesel engine types utilizing advanced emission control technol-
ogy, meaning NOx and/or particulate aftertreatment [5,7].
The amount of NOx in the exhaust of the engines before the SCR
catalyst can be measured in a dynamometer/test bench using NOx
gas analysers or can be estimated by simulation codes. The urea
solution is dosed into the exhaust gases of the vehicle suffering
decomposition and hydrolysis to form ammonia. If the ammonia/
urea introduced into the system does not match the NOx to be
converted, then there will be some ammonia slip. In practice
slightly less than 100% NOx conversion is usually targeted [3,4].
Anatase supported vanadia-tungsta (i.e. V2O5 and WO3)
catalysts are the most frequently used for removal of NOx from
the out-gases of power plants, heat plants, nitric acid factories,
stationary diesel engines and now also for lean exhaust gases of
mobile diesel engines [4]. These typical catalysts show high
efficiency conversion for NO at low temperatures. However, at high
temperatures (>400 8C) this catalyst tends to form nitrous oxides
(N2O), a greenhouse gas [1,8]. A reduction in NOx conversion
efficiency can also be observed by a direct oxidation of ammonia
[9]. Fig. 2. (a) ESC (load and engine speed modes of 13-mode) test cycle used in heavy-
duty engine and (b) simulates ETC cycle for engines vehicle speed versus time of
Due to its toxicity, vanadium-based SCR catalysts will be FIGE cycle on chassis dynamometer [5].
removed from the HDD market in few years, due to its toxicity. On
European passenger cars, however, it is expected that stringiest
emission limits will be attained using non-toxic Zeolite-based vanadium is listed in the California Proposition 65 List as
technologies. Any application in USA and Japan (i.e. passenger cars potentially carcinogenic substance [10]. On the so-called ‘‘emer-
and heavy-duty trucks) will use zeolite-based technologies, since ging markets’’ such as China, India, and Brazil, vanadium-based
catalysts perhaps will be still in use for the next years [11,13].
As water vapour is an unavoidable component of diesel exhaust
and its presence could also deactivate Cu-based catalysts [12], it is
understood that the existence of acid sites on the catalyst surface
provides points of attack and subsequent dealumination. It is of
interest, therefore, to evaluate how far the catalyst could tolerate
the presence of water vapour by reducing the availability of acid
sites (i.e. increasing the Si/Al ratio of the parent zeolite) [14,15].
Some studies on the other hand suggested that gold on alumina
catalysts are highly selective toward the formation of N2 and have a
peak of their NOx reduction activities at 200 8C. Such materials,
combined with a suitable high-temperature catalyst (i.e., CuZSM5
and/or Ag/alumina) [15], may broaden the operating window for
an efficient NOx removal under the low temperatures (below
150 8C) recorded during the driving cycle.
The CuMORD catalyst (synthetic) has been studied, though it
was shown to be efficient for NO reduction using hydrocarbons as a
reducing agent [16]. The natural Mordenite-based catalysts were
proven to be an excellent for SCR–NOx with ammonia or methane
as reducing agent, but they may suffer deactivation by water, SO2
Fig. 1. NEDC driving cycle for light-duty diesel vehicles. or HCl [16,17].
 Author's personal copy

422 M.L.M. Oliveira et al. / Applied Catalysis B: Environmental 88 (2009) 420–429

Recent studies have reported the use of natural mordenite ion Phase II: Derive mathematical correlation based on experi-
exchanged with Cu(II), Zn(II) and Ag(I). Exchanged Cuban mental data through ANN (details in Section 3).
Mordenite Natural zeolite was successfully utilized in the SCR– Phase III: Integrate the derive correlations in a road vehicle
NOx using ammonia as the reducing agent, demonstrating that simulator [20,21]: Simulate conversion efficiency, NH3 slippage
Cu(II)-exchanged sample may yield an extremely high perfor- and N2O formation throughout several driving cycles (details
mance of NO conversion in the presence of 100 ppm of SO2. It also below). Section 4 shows these results.
displayed good water tolerance and was working without any
diffusion problems, with very low contact times. The positive 2. Experimental
influence on the active chemical state of copper ions and
improvement of microporosity for Cu-Mordenite Cuban could be 2.1. Catalyst preparation
the key factors explaining the better performance of the sample in
the SCR of NO [15,17]. Two types of zeolites were selected, a commercial ZSM-5
Artificial neural networks (ANNs) have been used for several supplied by Zeolist International Co., with an atomic ratio Si/
system modelling. ANN has been mainly used in classification Al = 20, and a natural Mordenite, with an atomic ratio Si/Al = 5,
problems, in character recognition, in negotiation problems in from Palmarito deposit (Santiago de Cuba, Cuba). The catalysts
information processing, in control and automation and in were prepared by ion-exchange with an aqueous solution of
prediction problems [18]. Thus, ANN has been used to model copper(II) acetate. The copper loading was 2.0 and 4.0 wt.% for
the behaviour of one catalyst under different reaction conditions ZSM5 and MORD, respectively. The ion-exchange was carried out
for a specific reaction. Have successfully been applied to at room temperature using the homoionic ammonium forms NH4+-
conventional catalyst modelling and design of solid catalyst ZSM5 and NH4+-MORD in contact with the metal salt solution
[18,19]. Those applications include: design of ammoxidation of during 24 h, and pH 6. Finally, the catalysts were calcined at 550 8C
propylene catalyst, analysis and prediction of results of the for 6 h with a heating rate of 2 8C min1. The detailed preparation
reduction of NO over zeolite, and modelation the kinetics of a method is described elsewhere [12,16,22].
chemical reaction using methods not based in kinetic model based
in catalyst under different reaction conditions (partial pressure, 2.2. Catalyst characterization
contact time and temperature) [19]. Therefore, ANN consists of a
numeral of simple interconnected processing units, called neurons, The copper content was determined by atomic absorption
which are analogous to the biological neurons. Have been used due techniques (A.A.), using a Peking-Elmer 3100 spectrometer.
to its ability of supplying fast answers to a problem and Powder XRD (X-ray diffraction) patterns were obtained with a
generalising their answers, providing acceptable results for Siemens D500 diffractometer, equipped with a graphite mono-
unknown samples [19]. chromator and using Cu Ka radiation. The diffractograms were
Authors’ research was concentrated on the analysis of the recorded in the range 58 < 2u < 508 with a scanning speed of
catalysts based on copper-zeolites proposed for the possible use in 0.018 min1. The BET (Brunauer–Emmett–Teller) surface area of
SCR–NOx systems for the after-treatment of the exhaust gases from the catalysts was determined by nitrogen adsorption at 77 K using
the diesel vehicles. The catalysts were prepared by ion-exchange a Micromeritics ASAP 2020 apparatus. Hydrogen temperature-
over Natural Cuban Mordenite and ZSM-5 zeolites support. Also programmed reduction (H2-TPR) experiments were carried out
the catalyst stability in the presence of H2O and SO2 were between 50 and 800 8C using a flow of 10% H2/Ar (48 cm3. min1)
investigated for the both prepared and commercial catalysts. and a heating rate of 10 8C min1. Water produced in the reduction
Finally, a neural network-based approach to predict NOx conver- reaction was eliminated by passing the gas flow through a cold
sion efficiency of copper-zeolite and commercial catalyst was finger (80 8C). The H2-consumption was controlled by an on-line
implemented to derive and input a mathematical function into a gas chromatograph (Shimadzu GC-14A) provided with a thermal
numerical model for diesel road vehicle simulation [20,21]. conductivity detector (TCD).
The main innovation aspect of this research consists in the use of Temperature-programmed desorption of ammonia (NH3-TPD)
ANN to derive experimental correlations and integrate them in a was carried out to evaluate the total acidity of the catalysts. The
diesel engine vehicle simulator. The ANN was constructed using the catalysts were pretreated in a flow of He at 550 8C before the
experimental dataset derived under the typical exhaust of a diesel adsorption of ammonia at 100 8C, the NH3-TPD was performed
engine. The ANN was successfully trained accordingly to the between 100 and 550 8C with a heating rate of 10 8C min1 by
experimental data in SCR–NOx system to predict transient changes using a helium flow. The evolved ammonia was analysed by online
in NOx emissions level in HDD and LDD vehicle-operating conditions gas chromatography (Shimadzu GC-14A) provided with a TCD.
according to loads/engine speeds observed in driving cycles. X-ray photoelectron spectra were collected using a Physical
This is an important methodology for predicting NOx conver- Electronics PHI 5700 spectrometer with non-monochromatic Mg Ka
sion efficiency (%), the NOx out emissions, the NH3 slip and N2O radiation (300W, 15 kV, 1253.6 eV) and with a multi-channel
emissions (in ppm, g/s, g/km or g/kWh) in a vehicle equipped with detector. Spectra of powder samples were recorded in the constant-
a SCR system under several traffic conditions, including those of pass energy mode at 29.35 eV, using an analysis area of 720 mm in
homologation cycles. diameter. During processing of the XPS spectra, binding energy
The approach adopted used pre-euro III input conditions for values were referenced to the C 1s peak (284.8 eV) from the
diesel vehicles (HDD and LDD), equipped them with three different adventitious contamination layer. Spectra of C 1s and Cu 2p regions
SCR possibilities and predicted SCR system behaviour in several were first recorded for 10 min. This short acquisition time was used
roadway conditions. to avoid, as much as possible, reduction of Cu2+ species by the action
The used methodology was divided in three steps/phases as of the X-ray excitation source as previously reported [22].
follows:
2.3. Catalyst activity
Phase I: Measure experimentally huge amount of data for input
gases, exhaust gases and conversion efficiency of copper based The catalyst activity of the catalysts was tested in a flow micro-
SCR systems (detailed explanation in Section 2). reactor consisting of a Pyrex glass reactor (0.30 in. o.d.) connected
 Author's personal copy

M.L.M. Oliveira et al. / Applied Catalysis B: Environmental 88 (2009) 420–429 423

on-line to a quadrupole mass spectrometer (Balzers GSB 300 02).

The catalysts were pelletised with a particle size of 0.25–0.40 mm.
About 150 mg of pelletised solids were set into a reactor and
plugged with glass wool. The activities of the copper-catalysts
were compared with a commercial V2O5/WO3/TiO2 catalyst
(CATCO). Before the catalytic runs, the catalysts were pre-treated
at 350 8C in situ for 3 h under a helium flow (30 cm3 min1).
The reaction mixture typically consisted of 1000 ppm NO,
1000 ppm NH3, and 2.5 vol.% O2 (balanced with He). Water was
added to the feed, passing helium through a saturator with
deionized water. Moreover, 50 ppm of the SO2 was added to the
feed to study its effect in the catalytic performance. The flows were
independently controlled by channel mass flow-meters (Brooks).
The system was operated at the atmospheric pressure and with a
total flow of the 150 cm3 min1. The gas hourly space velocity
(GHSV) was about 48,000 h1; under these conditions, both
external and internal diffusion limitations were absent. We Fig. 3. ANN architecture of SCR model.
studied the behaviour of the catalysts for the temperatures ranged
between 100 and 550 8C. The analysis of reactants and products
(NO, NH3, O2, N2, N2O, H2O and SO2) was monitored by mass Fig. 3 shows an overview of the structural organization of the
spectrometer mentioned above. ANN used.
To extrapolate the data measured in the reactor to a real ANN design is constituted by an input layer with seven neurons,
catalyst, the space velocity parameter was used. Space velocity of a a single hidden layer with three neurons, and one output layer with
real catalyst in the diesel vehicle exhaust was referred to the wash one neuron, for each output considered. Stochastic back propaga-
coat volume of the SCR catalyst that is 20% of the catalyst volume tion with the Levenberg–Marquardt algorithm (MatLab1) was
[23]. The typical engine-out data was converted to typical space used to train the network. The activation functions used are
velocity referred to the active part of the catalyst, and these values
of space velocity were the ones used in the experimental part of the 1
f 0 ¼ log sigðnÞ ¼ (2)
research: ð1 þ expðnÞÞ

EMF ðg s1 Þ R ðJ mol

1
K1 Þ T ðKÞ 1 For the hidden layer and for the output layer:
GHSV ¼ 1
 1

M E ðg mol Þ PðJ l Þ V ðlÞ  0:2 2
1
f 1 ¼ tan sigðnÞ ¼ (3)
 3600 ðs h Þ (1) ð1 þ expð2nÞÞ  1

EMF stands for exhaust mass flow, ME stands for molar mass of The derived correlation for the ANN is
exhaust flow, R is the universal constant for gases, T is the
temperature of exhaust gases, and P is the exhaust gas pressure. y0i ¼ log sigðw0n;i  xn þ b0i Þ (4)

3. Simulation y1 ¼ tan sigðw0i;1  y0i þ b11 Þ (5)

3.1. Artificial neural network approach

% NO; % NH3 ; N2 O
X X   
Following a similar approach to engine-out emissions simula- ¼ tan sig w0i;1  log sig w0n;i  xn þ b0i þ b11 (6)
tion in road vehicle simulators (steady-state fuel consumption and
emissions database maps), the large amount of steady-state data Different weights and biases are generated for each different
obtained in the experimental part of the research was used for the output (y1).
neural network approach. This way mathematical correlations
were obtained and directly implemented in the road vehicle 3.2. Validation
simulator. This is a first approach to solve the lack of simulation
models for SCR in such road vehicle simulators. A total of more than 577 sets of measurements were performed,
The measured results obtained from the experimental appa- of which 85% were used to train and validate several networks. The
ratus described above for copper-zeolite catalyst and commercial networks with low sum squared error were used to obtain the
catalyst were used to train an ANN. The main goal is to obtain the correlation function between the inputs and outputs. In face of,
bias and weights for each neuron and the activation functions for stochastic back propagation was used to train the network, the
hidden and output layers with the lowest mean squared error number and size of the hidden layers being determined empirically
(MSE). Fifteen percent of the data was not used to train the ANN, according to error value and stability observed. A single hidden
but instead to test it. The best ANN will be used to construct the layer containing three neurons proved the best solution for the
correlation function between the inputs and outputs, and will be application. The 15% were used to test the SCR ANN model.
integrated in a road vehicle fuel consumption and emission model. The errors when compared to the training data for the network
The relevant inputs considered for the outputs NOx conversion used. The statistical analysis of the results indicates that the R2
efficiency, NH3 conversion efficiency and N2O formation were: value for the test, training and validation data was higher, more
space velocity (SV, h1), exhaust gas temperature (T, 8C), oxygen than 0.67, and this indicates that ANN-based model developed in
concentration (O2%, v/v), water concentration (H2O%, v/v), sulphur this work can predict the (%) NO conversion efficiency, out NH3
dioxide concentration (SO2 ppm), nitrogen oxide concentration (ppm) and N2O (ppm) formation to SCR system of a diesel vehicle.
(NO ppm), ammonia concentration (NH3 ppm). Schematically Table 2 shows the obtained R2 values for each derived correlation.
 Author's personal copy

424 M.L.M. Oliveira et al. / Applied Catalysis B: Environmental 88 (2009) 420–429

Table 2
R2 value for the test, training and validation data.

Catalyst R2test R2training R2validation

CATCO NOx 0.73 0.74 0.80

NH3 0.80 0.82 0.80
N2O 0.94 0.84 0.90

CuMORD NOx 0.85 0.88 0.90

NH3 0.84 0.90 0.85

CuZSM5 NOx 0.85 0.93 0.90

NH3 0.80 0.90 0.80
N2O 0.67 0.70 0.70

Fig. 4 shows an example of simulated and measured data for NO

conversion efficiency.

4. Results and discussion

4.1. Characteristics of the prepared catalyst Fig. 4. Simulated and measured data for the test values of NO conversion efficiency
for CuMORD.
Table 3 shows the BET surface areas and Cu contents (wt.%). The
results show, as expected, a slight decrease in the BET surface areas
as in the pore volume regarding the support, after the incorpora-
tion of Cu into the framework. In contrast, when we considered the
surface area and pore volume for the CuMORD sample, they hardly
change regarding the natural Mordenite zeolite, due to its
unidimensional structure [16,17,22].
The X-ray diffraction patterns of the prepared catalysts and
supports based ZSM-5 and natural mordenite zeolites showed any
loss of crystallinity of the natural mordenite and MFI phase and the
absence of the characteristic reflection lines of copper(II) oxide
species [12,14,17]. The absence of metallic oxides points out the
successful incorporation of metallic cations by ion-exchange on
the exchange sites.
The reduction of copper species was studied by H2-TPR. In
general, the reducibility of copper(II) is function of both the nature
of the support and the dispersion degree of the active phase. Fig. 5
shows the normalized H2 consumption curves, obtained by
dividing the H2 consumption signal by the amount of copper
present in the catalyst. These catalysts have different H2
consumption peaks at different temperatures. The CuZSM5
catalyst presents a H2 complex consumption band extending from
150 to 250 8C, and the CuMORD catalyst presents a H2 consump- Fig. 5. TPR curves of copper-zeolite catalysts.
tion band from 120 to 250 8C. For this catalyst, different copper
ions seem to be present in different environments, with broad
bands of H2 consumption appearing with various shoulders. The
first species to be reduced are copper oxides in the lowest
temperature peak.
As determined by NH3-TPD, the acidity of these catalysts is
depicted in Fig. 6. From this plot, the addition of Cu produces a
sharp decrease in the acidity in relation to the support. The greatest
desorption of ammonia occurs in the temperature range 200–
300 8C for the CuMORD catalyst, while it is similar to temperature
range 100–200 8C for CuZSM5 catalyst. From the results presented
in Fig. 7 it can be observed that for typical space velocity values,

Table 3
Copper content and textural parameters for the supports and the catalysts.

Sample [Cu] (wt.%) SBET (m2 g1) Vmicroporea (cm3 g1)

ZSM5 – 286 0.159

MORD – 167 0.085
CuZSM5 1.98 240 0.124
CuMORD 4.01 178 0.089
a
Determined by the Dubinin–Astakhov method, applied to microporous solids
(0.01  P/P0  0.05). Fig. 6. NH3 desorbed as a function of the temperature for copper–zeolite catalysts.
 Author's personal copy

M.L.M. Oliveira et al. / Applied Catalysis B: Environmental 88 (2009) 420–429 425

Fig. 9. NO conversion over Cu-exchanged zeolites and commercial catalyst (CATCO).

Fig. 7. NO conversion evolution with space velocity. Experimental conditions: 1000 ppm NO, 1000 ppm NH3, and 2.5 vol.% O2 (balanced
with He), flow rate 150 cm3 min1 and GHSV = 48,000 h1.

94%, but the catalyst CuMORD is more active at lower temperature

than the catalysts CuZSM5.
The real diesel exhaust gas contains water vapour and SO2, so
that their effects on catalyst activity must be taken into account.
We performed the corresponding studies at 350 8C by adding in the
feed 10% H2O (v/v) and 50 ppm of SO2 as shown in Figs. 10 and 11,
respectively.
Figs. 10 and 11 show very high NO conversion degree at 350 8C
for CuMORD (99.6% and 100%) and CATCO (91.8% and 99.8%) in the
presence of H2O vapours and SO2, respectively. Furthermore, no
appreciable deactivation after 10 h of reaction was observed. The
CuZSM5 catalyst demonstrates a slight decrease (around 20%) in its
catalyst activity under the presence of SO2. An introduction of 10%
water vapour into reaction also caused a slight deactivation of the
CuZSM5 catalyst and a slight decrease in the selectivity towards
nitrogen.
Table 4 summarizes the experimental results for the catalysts
tested. CATCO catalyst shows a trend to form N2O during SCR
Fig. 8. Cu 2p core level spectra for CuMORD and CuZSM5 catalyst. reaction. The CuZSM5 catalyst displayed high catalytic activity,
with an appreciable formation of N2O and with a significant
deactivation in the presence of the water and SO2. Formation of
there was always some amount of NO conversion indicating that N2O may be explained by oxidation of NH3, as described else where
diffusion of NH3 and exhaust gases occurred. The effect of space
velocity was one of the main inputs of the neural network
approach model.
In order to know the copper species on the surface of copper(II)-
exchanged zeolites, the catalysts have been studied by XPS. The Cu
2p3/2 signal is composed of two peaks at 933.4 and 935.6 eV, where
the former is assigned to the presence of CuO species, and the latter
to Cu(II) ions coordinated to superficial oxygen atoms of the
zeolites [15,16,22], as it is shown in Fig. 8.

4.2. Catalyst activity measurements

Fig. 9 displays the NO conversion degree as a function of

reaction temperature for the catalysts tested. The high SCR activity
of CATCO (commercial catalytic) is clearly discernible, especially at
low temperature, up to 250 8C with 100% of NO conversion [12,22].
The results show that the CuMORD is an active catalyst at
250 8C, but at temperatures above 400 8C the catalytic oxidation of
NH3 is observed [15,22]. This reflects the non-selective reduction Fig. 10. Water tolerance of the Cu-exchanged zeolites and a commercial catalyst.
of NO. Similar results of NO conversion efficiency were obtained in Experimental conditions: 1000 ppm NO, 1000 ppm NH3, and 2.5 vol.% O2, 10 vol.%
the standard test for CuZSM5. At 350 8C, the NO conversion is about H2O (balanced with He), flow rate 150 cm3 min1 and GHSV = 48,000 h1.
 Author's personal copy

426 M.L.M. Oliveira et al. / Applied Catalysis B: Environmental 88 (2009) 420–429

standard classification of the vehicles after SCR fitting. Other cycles

were used such as: UDDS (Urban Dynamometer Driving Schedule),
HWFET (HighWay Fuel Economy Test), a constant speed of 120 km/
h HDC (Highway Driving Cycle) [5,7] and areal measured Porto city
(Portugal) bus line driving cycle, were simulated and NOx emission
values again compared with Euro emission limits.
Table 5 shows some characteristics of each driving cycles, such
as total duration, average speed, distance, and so forth.
Fig. 12 shows non-homologation driving cycles. As example,
Fig. 12a shows the heavy-duty driving cycle of Porto city in
Portugal, with simulated maximum exhaust temperatures reach-
ing 550 8C, with an average of 200 8C.
Emissions simulated for light-duty vehicles are expressed in
grams of pollutant per unit of travelled distance, g/km, and
emissions for heavy-duty vehicle are expressed in grams per
mechanical energy delivered by the engine, g/kWh.

Fig. 11. The influence of SO2 on NO conversion over Cu-exchanged zeolites and 4.4. Heavy-duty vehicle
commercial catalyst. Experimental conditions: 1000 ppm NO, 1000 ppm NH3,
50 ppm SO2, and 2.5 vol.% O2 (balanced with He), flow rate 150 cm3 min1 and Table 6 shows the performance of the SCR system for ESC and
GHSV = 48,000 h1.
ETC European homologation cycles (see Fig. 2) [5]. For ESC test, an
average NOx reduction of 71, 75 and 67% were achieved for CATCO,
[22]. CuMORD catalyst does not form N2O in the conditions CuMORD and CuZSM5, respectively.
analysed, and shows the highest N2 selectivity (100%). Some These results indicate that it is possible to convert more than
authors reported that in a copper(II)-exchanged natural Morde- 60% of the engine-out emissions of NOx and meet Euro IV emission
nite, copper(II) ions seem to be present in a structural configura- standards in ESC cycle for a HDD equipped with SCR system using a
tion where the formation of a Cu–H2O complex on the catalyst copper–zeolite or commercial catalyst. The CATCO showed the
surface is complicated, allowing to avoid catalyst deactivation by higher conversion efficiency. The drop in average conversion
water [16,22]. efficiency under ETC transient cycle conditions for Cu-containing
zeolite catalysts may be related with copper-zeolite catalyst
4.3. Simulation results deactivation occurring due to a change of active component, i.e.,
the copper ions [13,16]. Table 7 summarizes the results of NOx
Both heavy (18,000 kg, 10 l, 173 kW) and light-duty (1050 kg, emission, N2O formation and NH3 slippage for non-homologation
1.7 l, 60 kW) vehicles with and without a SCR–NOx system were driving cycles.
simulated in current European homologation cycles (for the engine Some studies have shown ammonia slip together with NOx after
and vehicle, respectively). NOx emissions values were compared the catalyst measured on the FIGE cycle [6,24]. In this study, at first
with emissions limits to analyse the corresponding emission the ammonia slip is well below the 10 ppm goal but then it rises to

Table 4
Catalytic data at 350 8Ca.

Catalyst NO conversion (%) NH3 conversion (%) N2O formation (ppm) N2 selectivity (%)

CATCO Standard 100 100 32.0 97.0

H2O 91.8 92.4 65.0 93.0
SO2 99.2 97.6 43.0 91.0

CuMORD Standard 100 94.3 0.00 100

H2O 99.6 100 0.00 100
SO2 100 100 0.00 100

CuZSM5 Standard 98.1 94.0 47.0 95.1

H2O 73.8 72.6 41.0 94.6
SO2 78.9 77.9 0.00 100
a
Feed conditions general (standard test and stability): 1000 ppm NO, 1000 ppm NH3, 2.5% (v/v) O2, balance with helium; flow rate = 150 mL min1, 0.150 g catalyst and
GHSV = 48,000 h1.

Table 5
Driving cycles used for vehicle (heavy and light) type approval in this work.

Drive cycle UDDS (urban) HWEFT (road) NEDC Porto bus line HDC ETC

Maximum speed (km/h) 91 96 120 38 120 91

Average speed (km/h) 32 78 34 11 120 60
Distance (km) 12 17 11 8 10 29
Time (s/min) 1369/23 765/13 1180/20 2466/41 300/5 1799/30
Maximum acceleration (m/s2) 1.48 1.43 1.06 2.23 0 3.83
Max downgrade (%) 0 0 0 15 0 0
Max upgrade (%) 0 0 0 10 0 0
Average grade (%) 0 0 0 3 0 0
Idle (s) 259 6 298 696 0 75
 Author's personal copy

M.L.M. Oliveira et al. / Applied Catalysis B: Environmental 88 (2009) 420–429 427

Fig. 13. Instantaneous emissions of Porto City driving cycle for CuMORD catalyst:
(a) emission NOx in g/s, (b) ammonia slip (ppm) and temperature.

a maximum of 30 ppm and on average NOx is reduced to 85% [24].

Also in this work showed that catalyst temperature is one of the
parameters affecting the conversion efficiency and the results
suggest that at elevated temperatures the urea injection strategy is
too aggressive in relation to the catalyst size [13,20,25].
Fig. 13 shows an example of the instantaneous emissions of NOx
before and after the SCR system for HDD vehicles equipped with
CuMORD catalyst. Significant NOx reductions were attained by
using the SCR system in the driving cycle real, but as also results
suggest that in one Real cycle (Porto City) at elevated temperatures
the NH3 slip can be around 600 ppm (see c). This would indicted
that the ammonia introduced into the system does not match the
NOx to be converted, then there has be some ammonia slip, as
Fig. 12. Others driving cycles and temperature used in this study: (a) Porto city, (b) desecrated in the literature [3,4,24].
UDDS and (c) HWEFT.

4.5. Light-duty vehicle

Table 6
NOx performance of the SCR systems in ESC and ETC cycle.
SCR technology could be used on LDD vehicles and passenger
Catalyst Engine-out After SCR NOx conversion Euro cars without major investments in the urea distribution system.
(g/kWh) (g/kWh) (%) standard The application to LDD vehicles and passenger cars presents
ESC additional challenges. Their operational characteristics are quite
CATCO 11 3.26 71 IV different from HDD vehicles as they typically operate at high speed
CuMORD 11 2.73 75 IV and low load with lower exhaust temperatures in and this can be
CuZSM5 11 3.61 67 IV
accomplished by co-feeding NO2, obtained from partial conversion
ETC of NO on a pre-oxidizing catalyst [4,25]. However, in this work it
CATCO 5.43 1.36 80 V has been possible a preliminary study.
CuMORD 5.43 3.23 40 IV
Table 8 shows the performance of SCR system for NEDC driving
CuZSM5 5.43 3.49 31 IV
cycle is used in this study (see Fig. 1).
 Author's personal copy

428 M.L.M. Oliveira et al. / Applied Catalysis B: Environmental 88 (2009) 420–429

Table 7
Performance of the SCR system for heavy-duty vehicle.

Catalyst Cycles driving (see Fig. 14) NOx in (g/kWh) NOx out (g/kWh) % NO conversion Average N2O formation (104 g/km) NH3 slip (g/km)

CATCO UDDS 6.18 1.50 85.6 0.20 2.29

HDC 0.14 0.06 54.0 0.02 0.07
HWEFT 5.60 1.60 74.6 1.03 1.90
Real Urban Circuit of Porto 6.25 1.20 90.4 0.40 3.02

CuMORD UDDS 6.18 2.40 40.0 0.00 3.00

HDC 0.14 0.06 53.9 0.00 0.12
HWEFT 5.60 2.51 48.0 0.00 1.90
Real Urban Circuit of Porto 6.25 2.15 35.0 0.00 3.51

CuZSM5 UDDS 6.18 3.80 27.0 0.13 4.10

HDC 0.14 0.05 59.5 0.02 0.08
HWEFT 5.60 3.32 36.2 7.04 2.84
Real Urban Circuit of Porto 6.25 3.93 23.0 0.24 6.40

Table 8
Performance of the SCR system in NEDC.

Catalyst Engine-out Aftertreatment NOx conversion Euro

(g/km) SCR-NH3 (g/km) (%)

CATCO 0.59 0.088 90.7 V

CuMORD 0.59 0.346 27.0 III
CuZSM5 0.59 0.413 21.0 III

Highest average conversion efficiency of NOx emissions was

90% for CATCO catalyst. Table 9 shows exhaust NOx emissions for
all catalysts studied and simulated driving cycles. A pre-euro III
LDD can comply with EuroV after fitting CATCO and to Euro III after
fitting Cu-zeolite catalyst.
Fig. 14 exemplify the instantaneous NOx emissions before
and after the SCR system and ammonia slip, for LDD vehicle
equipped with SCR using CuMORD. In face of Urea decomposi-
tion does not reach completion in the gas phase at temperatures
that are typical for light-duty diesel exhaust (below 300 8C). In
fact, only about 20% or less of the urea decomposes to HNCO and
NH3 in the gas phase at 330 8C, and only about 50% decomposes
at 400 8C. HNCO has been shown to be very stable in the gas
phase, requiring an oxide surface to catalyze its decomposition
to NH3 [4,24–26]. Ammonia control can be assured by an
appropriated control strategy and accurate control of the
dosing system or by using an oxidation catalyst downstream
of the SCR catalyst to oxidise the ammonia [1]. In this case, the
ammonia slip (see Fig. 14b) is less than 300 ppm which is in
accordance with data found in the literature for commercial
catalyst [25,26].
For LDD vehicles no experimental data in the literature was
found for N2O formation after SCR systems. Generally more N2O is
formed for over-oxidized cases when urea is injected compared
with NH3 injection [24,27]. In this work has been show N2O Fig. 14. Instantaneous emissions of road driving cycle for CuMORD catalyst: (a) NOx
formation from CATCO and CuZSM5 catalysts. emissions in ppm and (b) ammonia slip (ppm).

Table 9
Performance of SCR system for light-duty vehicle.

Catalyst Cycles driving NOx in (g/km) NOx out (g/km) % NO conversion Average N2O formation (g/km)  104 NH3 slip (g/km)

CATCO UDDS 0.59 0.07 90.7 6.32 0.20

HDC 0.74 0.27 65.0 0.41 0.29
HWEFT 0.39 0.09 78.6 2.60 0.19
UDDS 0.59 0.32 29.0 0.00 0.14

CuMORD HDC 0.74 0.50 55.0 0.00 0.25

HWEFT 0.39 0.24 32.0 0.00 0.17
UDDS 0.59 0.34 26.0 0.21 0.36

CuZSM5 HDC 0.74 0.30 60.0 0.32 0.29

HWEFT 0.39 0.26 28.2 2.13 0.23
 Author's personal copy

M.L.M. Oliveira et al. / Applied Catalysis B: Environmental 88 (2009) 420–429 429

5. Conclusions [4] Par L.T. Gabrielsson, Topics in Catalysis 28 (2004) 1.

[5] Website dieselnet.com (last accessed on April 2008).
[6] ETC test cycle, developed by the FIGE Institute, Aachen, Germany, FIGE Report 104
Simulation results of homologation cycles show that a pre-euro 05 316, January 1994, 060403 http://www.dieselnet.com/standards/eu/hd.php.
HDD can comply with Euro IV after fitting the SCR system. [7] E. Zervas, G. Bikas, Energy & Fuels (2008) 1.
[8] H. Sjovall, E. Fridell, R.J. Blint, L. Olsson, Applied Catalysis B: Environmental 64
Simulation results in real driving cycles show lower level of NOx (2006) 180.
(up to Euro V) for HDD. Despite having toxic compounds [9] F. Birkhold, U. Meingast, P. Wassermann, O. Deutschmann, Applied Catalysis B:
(vanadium) and N2O formation CATCO shows the better NOx Environmental 70 (2007) 119.
[10] R. Aneja, K. Flathmann, C. Savonen, T. Tindall, SCR Potential and Issues for Heavy
conversion efficiencies. CuMORD do not form N2O and have better
Duty Applications in the USA, in: Proceedings of the Detroit Diesel Corporation
performance than the CuZSM5, so being a possible substitute of 10th Annual DEER Conference, 2 September, 2004.
CATCO. LDD kept under the same regulation limits accordingly [11] J.H. Goo, M.F. Irfan, S.D. Kim, S.C. Hong, Chemosphere 67 (2007) 718.
[12] J.H. Park, H.J. Park, J.H. Baik, I.-S. Nam, C.-H. Shin, J.-H. Lee, B.K. Cho, S.H. Oh,
with the different catalysts (Euro III to Euro V). A pre-euro III LDD
Journal of Catalysis 240 (2006) 47.
can comply with Euro V after fitting CATCO and to Euro III after [13] H. Yahiro, M. Iwamoto, Applied Catalysis A: General 22 (2001) 163.
fitting Cu-zeolite catalyst. [14] K. Eränen, F. Klingstedt, K. Arve, L.-E. Lindfors, D.Yu. Murzin, Journal of Catalysis
The results of the research show that the ANN integrated with a 227 (2004) 328.
[15] R. Moreno-Tost, M.L.M. Oliveira, D. Eliche-Quesada, J. Jimenéz-Jimenéz, A. Jime-
road vehicle model produce average results consistent with nez-Lopes, E. Rodriguez-Castellón, Chemospere 72 (2008) 608–615.
experimental data and can be a powerful tool for catalyst [16] C. Torres-Abreu, M.F. Ribeiro, C. Heriques, F.R. Ribeiro, Catalysis Letters 43 (1997)
simulation of vehicle emission control technology. 25.
[17] G.A. Castro, M.G. Echegarrua, M.A. Perez, R. Moreno-Tost, E. Rodrı́guez Castellon,
A. Jimenez-Lopez, Microporous and Mesoporous Materials 108 (2008) 325–332.
Acknowledgments [18] (a) J. Clark, K. Warwick, Artficial Intelligence Review 12 (1998) 95–115;
(b) K. Huang, F. Chen, D. Lu, Applied Catalysis A 219 (2001) 61–68.
[19] C. Brace, Prediction of Diesel Engine Exhaust Emissions using Artificial Neural
The authors would like to acknowledge the National Council for Networks, in: Proceedings of the Neural Networks in systems Design IMechE
Scientific and Technological Development (CNPq) – Brazil for Seminar S591, 10 June, (1998), pp. 1–11.
financial support for scholarship (GDE/CNPq) of Mona Lisa Oliveira. [20] A. Brooker, K Haraldsson, T. Hendricks, V. Johnson, K. Kelly, B. Kramer, T. Markel,
M. O’Keefe, S. Sprik, K. Wipke, M. Zolot, ADVISOR Documentation, NREL, 2002.
Ministerio de Educación y Ciencia of Spain (project NAN2004-
[21] C.M. Silva, G.A. Gonçalo, T.L. Farias, M. Lopes, Journal Science for the Total
09267-C03-01). The authors would like to acknowledge FCT Environment 367 (2006) 441.
(National Science and Technology Foundation) through the [22] R. Moreno-Tost, J. Santamaria-González, E. Rodriguez-Castellón, A. Jiménez-
López, M. Autié, E. González, M. De las Pozas, Applied Catalysis B: Environmental
national project MMSAFU-Microsimulation Model to Simulate
50 (2004) 279.
Alternative Fuel Usage POCI/ENR/57450/2004. [23] G.C. Koltsakis, A.M. Stamatelos, Chemical Engineering Science 54 (1999) 4567.
[24] G. Lenaers, M. Van Poppel, On-Board Ammonia Measurements for Vehicles with
References Recent Emission Control Technology Exemplified on a SCR Equipped Bus, in:
Proceedings of the 8th International Conference on Energy for a Clean Environ-
[1] F. Klingstedt, K. Arve, K. Eraenen, D.Yu. Murzin, Accounts of Chemical Research 39 ment–Clean Air, 2005.
(2006) 273. [25] E. Tronconi, I. Nova, C. Ciardelli, D. Chatterjee, B.B. Konrad, T. Burkhardt, Catalysis
[2] J.B. Heywood, Internal Combustion Engine Fundamentals, McGraw-Hill Interna- Today 105 (2005) 529.
tional (Editor), Singapore, 1988. [26] I.M. Saaid, A.R. Mohamed, S. Bhatia, React. Kinet. Catal. Lett. 75 (2) (2002) 359.
[3] ACEA Final report on Selective Catalytic Reduction, http://europa.eu.int/comm/ [27] N.J.L. Gense, R.J. Vermeulen, N2O Formation in Vehicles Catalysts: Nederlandse
enterprise/automotive/mveg_meetings/meeting94/scr_paper_final.pdf, June 2003. Organisatie report, 2002.