Catalytic urea hydrolysis in the selective catalytic reduction of NO x: Catalyst screening and kinetics on anatase TiO2 and ZrO2

Article abstract of DOI:10.1039/c2cy20668d

The catalytic hydrolysis of urea was investigated under conditions relevant for the selective catalytic reduction of NO_x (urea-SCR). The hydrolysis activities of the tested catalysts coated on cordierite monoliths were in the order ZrO₂ > TiO₂ > Al₂O ₃ > H-ZSM-5 > SiO₂. A comparison with isocyanic acid (HNCO) hydrolysis on the same catalysts showed that urea decomposition was much slower than HNCO hydrolysis; hence, catalytic urea thermolysis into NH ₃ and HNCO is likely to be the rate-determining step in urea decomposition. Interestingly, a different order of catalyst activities was found in water-free experiments on urea thermolysis: TiO₂ > H-ZSM-5 ≈ Al₂O₃ > ZrO₂ > SiO₂. The widely accepted reaction pathway for urea decomposition, namely urea thermolysis followed by HNCO hydrolysis, seems to be valid on all the tested catalysts except ZrO₂: The high urea hydrolysis activity of the ZrO₂ catalyst compared to its low urea thermolysis activity suggested a different reaction pathway, in which water directly attacks adsorbed urea rather than adsorbed HNCO.

Full text of DOI:10.1039/c2cy20668d

Catalysis
Science&Technology
www.rsc.org/catalysis
Volume 3 | Number 4 | April 2013 | Pages 821–1152
ISSN 2044-4753
COVER ARTICLE
Kröcher et al.
Catalytic urea hydrolysis in the selective catalytic reduction of NO : catalyst
x
screening and kinetics on anatase TiO and ZrO2
2044-4753(2013)3:4;1-9
2

Catalysis
Science & Technology
View Article Online
View Journal | View Issue
PAPER
Catalytic urea hydrolysis in the selective catalytic
reduction of NO : catalyst screening and kinetics on
x
Cite this: Catal. Sci. Technol., 2013,
, 942
3
anatase TiO and ZrO
2
2
Andreas M. Bernhard, Daniel Peitz, Martin Elsener, Tilman Schildhauer and
Oliver Kr o¨ cher*
The catalytic hydrolysis of urea was investigated under conditions relevant for the selective catalytic reduction
of NO
order ZrO
catalysts showed that urea decomposition was much slower than HNCO hydrolysis; hence, catalytic urea
thermolysis into NH and HNCO is likely to be the rate-determining step in urea decomposition. Interestingly,
a different order of catalyst activities was found in water-free experiments on urea thermolysis: TiO > H-ZSM-
E Al > ZrO > SiO . The widely accepted reaction pathway for urea decomposition, namely urea
thermolysis followed by HNCO hydrolysis, seems to be valid on all the tested catalysts except ZrO : The high
urea hydrolysis activity of the ZrO catalyst compared to its low urea thermolysis activity suggested a different
reaction pathway, in which water directly attacks adsorbed urea rather than adsorbed HNCO.
x
(urea-SCR). The hydrolysis activities of the tested catalysts coated on cordierite monoliths were in the
2
> TiO > Al > H-ZSM-5 > SiO . A comparison with isocyanic acid (HNCO) hydrolysis on the same
2
2
O
3
2
3
Received 28th September 2012,
Accepted 29th October 2012
2
5
2
O
3
2
2
DOI: 10.1039/c2cy20668d
2
2
www.rsc.org/catalysis
because the compounds involved in the SCR reaction are
strongly diluted. The dilution corresponds to a low gas partial
Introduction
In the selective catalytic reduction of NO
x
(SCR) with urea, the pressure, which shifts the urea thermolysis reaction to the
dosed urea solution decomposes in the hot exhaust gas and on product side. The thermodynamic equilibrium curve for the
the catalyst to yield the actual reducing agent ammonia (NH ). actual concentrations in our experiment is included in Fig. 5a.
According to the established mechanism, urea first thermolyzes It shows that our catalyst screening was not biased by an
into NH and isocyanic acid (HNCO) in a non-catalytic reaction equilibrium limitation of the urea thermolysis reaction.
1) and the intermediately formed HNCO then hydrolyzes
Yet, most studies on urea decomposition in the SCR process
catalytically (2). HNCO is relatively stable in the gas phase, have either focused on non-catalytic thermolysis,
3
3
(
4
–13
or on
Scattered information in the
a large fraction of the dosed urea remains unreacted before it literature indicates that urea thermolysis (1) can be catalyzed as
1
1,14–21
hence non-catalytic HNCO hydrolysis can be neglected. However, catalytic HNCO hydrolysis.
2
22–25
enters the catalyst. Therefore, catalytic urea thermolysis should well,
but dynamic experiments, such as thermogravimetric
also be considered.
analysis (TGA), have been prevailing so far. Dynamic experiments
Thermodynamics show that urea hydrolysis into NH
3
and are difficult to use for kinetic evaluation due to mass and heat
3
CO
2
is a significantly exergonic reaction at any temperature.
transfer phenomena. Only recently, we have reported about
On the other hand, an eventual equilibrium limitation of the catalytic urea thermolysis (1) under steady state conditions over
26
urea thermolysis reaction into NH3 and HNCO needs to be different metal oxides. Also, a theoretical study of Todorova et al.
discussed. At 150 1C, the thermolysis of pure urea gas is (2011) proposes catalytic urea thermolysis. These authors
0
ꢀ1 3
endergonic with DG = 18.5 kJ mol
.
Only above 260 1C, calculated the reaction energy barriers for the hydrolysis of
DG becomes negative. The hypothetical thermolysis of solid guanidine on anatase TiO2 (101) using density functional
urea into gaseous NH and HNCO is even more endergonic with theory (DFT) and found adsorbed urea as an intermediate on
0
3
3
0
ꢀ1
3
0
27
DG = 54.7 kJ mol at 130 1C. However, these DG values do the TiO2 surface. From the adsorbed urea NH3 can be
not represent the situation in the urea-SCR application, eliminated, leaving HNCO on the catalyst surface. HNCO
2
1,27
then hydrolyzes via carbamic acid as another intermediate.
Interestingly, an additional reaction pathway was proposed,
where water directly attacks the adsorbed urea. In this more
Paul Scherrer Institut, OVGA/112, 5232 Villigen PSI, Switzerland.
2
7
E-mail: oliver.kroecher@psi.ch; Fax: +41 (0)56 310 21 99; Tel: +41 (0)56 310 20 66
9
42 Catal. Sci. Technol., 2013, 3, 942--951
This journal is c The Royal Society of Chemistry 2013

View Article Online
Paper
Catalysis Science & Technology
Fig. 1 Reaction scheme of catalytic urea decomposition.
direct reaction (3), carbamic acid is formed as an intermediate as
well, but the reaction step of intermediate HNCO formation is
skipped. Fig. 1 illustrates the two reaction pathways. The term
‘‘urea decomposition’’ is used to indicate any combination of
reactions that consume urea.
Urea thermolysis: CO(NH
2
)
2
- NH
3
+ HNCO
+ CO
2
(1)
(2)
(3)
HNCO hydrolysis: HNCO + H
2
O - NH
3
Fig. 2 The reactor used to perform catalytic urea decomposition, with the
thermocouples (T) upstream and downstream of a catalyst-coated monolith
sample.
Direct urea hydrolysis: CO(NH + H O - 2NH + CO
2
)
2
2 3 2
In the present study, we compare water-free (dry) catalytic
urea thermolysis (1) on different catalysts with catalytic urea
hydrolysis (wet), while keeping in mind that there may be two Table 1 Catalyst samples tested
different reaction pathways for urea hydrolysis. Additionally, we
BET surface/ Considered
used a pseudo first-order kinetic model to analyze urea thermolysis
2
ꢀ1
Material
Producer, product name
m g
p/p
0
on all the catalysts tested and to analyze direct urea hydrolysis
on ZrO
Anatase
Crystal Global, DT-51
89
0.05–0.3
2
.
TiO
ZrO
Al
H-ZSM-5
SiO
2
2
MEL Chemicals, XZO881
Condea, Disperal S
67
200
420
380
0.05–0.3
0.05–0.3
0.01–0.1
0.05–0.3
2
O
3
Experimental
Setup
S u¨ d-Chemie, H-MFI 27
Davison Catalysts,
2
s
Davicat SI 1452
Experiments were carried out using the lab-scale setup
described in ref. 28–30 Briefly, the prepared urea solutions
were constantly dosed by a gas-assisted spray nozzle into a
tubular glass reactor with an inner diameter of 20.4 mm as
shown in Fig. 2. Catalyst-coated cordierite monoliths were
directly fitted into the reactor, 9 cm downstream of the spray
nozzle, where the urea was well distributed on the reactor cross
section. We did not observe catalyst deactivation or deposit
formation in most of our experiments. Only dry experiments
with the H-ZSM-5 catalyst induced slight deactivation. The
absence of deposits is attributed to the high spray quality
realized in our setup. The model gas was mixed from pure
gases using electronic mass flow controllers (Brooks 5850S);
water was provided pulsation-free by oxidation of hydrogen on
a Pt-catalyst. Deviating from ref. 30 the reactor exit was heated
to 170 1C, the extraction capillary to the FTIR spectrometer was
heated to 150 1C and the glass wool condenser was removed.
binder (Ludox AS-40, 10% of the catalyst mass). The monoliths
were calcined at 550 1C for 5 h. The specific surface areas of
catalyst powders were measured by nitrogen physisorption on a
Quantachrome Autosorb 1-c instrument. We prepared the
samples for the physisorption measurements by calcining dried
catalyst suspensions under the same conditions as the coated
monoliths.
Measuring conditions
The tests were typically conducted at a space velocity (GHSV) of
ꢀ
1
91 000 h
using the catalyst coated monoliths with a cell
density of 600 cpsi shown in Table 2. The base feed of the
ꢀ1
Table 2 Catalyst coated monoliths used (the GHSV was calculated at 500 L h
at STP)
Catalyst coating and characterization
Monolith
name
Active mass/
mg
Length/
cm
GHSV/
h
Loading/
g L
ꢀ1
ꢀ1
Cordierite monoliths (Corning) with a cell density of 600 cells per
31
TiO2
45
85
540
24
52
64
2
1.06
6
2
2
2
2
2
91 000
170 000
30 000
91 000
91 000
91 000
91 000
91 000
8.3
29
33
4.4
9.4
12
9.1
10
square inch (cpsi), a wall thickness of 0.15 mm, 216 channels,
0 mm long and 18 mm in diameter were coated with the catalyst
powders shown in Table 1.
The TiO powder was suspended in a diluted aqueous NH
solution, the other powders in pure water. We improved the
TiO
TiO
TiO
2
2
2
b
c
d
2
ZrO
Al
H-ZSM-5
2
2
3
2
O
3
48
55
stability of the coating using a commercial colloidal silica SiO
2
This journal is c The Royal Society of Chemistry 2013
Catal. Sci. Technol., 2013, 3, 942--951 943

View Article Online
Catalysis Science & Technology
Paper
concentration was stable within a few percent. From the HNCO
generator output before and after the experiment, the values
at any time during the experiment were calculated by linear
interpolation.
Analytics
The gaseous compounds NH , CO , HNCO, H O, ethanol,
3
2
2
methanol and urea were measured by Fourier transform infra-
red (FTIR) spectroscopy using a multi-component analysis
2
6,30,32
method developed in-house.
The quantification of urea
by FTIR spectroscopy was precise in the water-free experiments,
2
6
but the presence of water induced scattering. The product gas
was also analyzed using a liquid-quench of the product gas,
followed by HPLC analysis as described in ref. 30, 33, and 34.
The quenching solution was the HPLC eluent itself: 5 mM
phosphate buffer, set to pH 7 by NaOH. Three liquid samples
were collected consecutively at a selected temperature at inter-
vals of 10 min each. HPLC analysis was used to check the
HNCO and urea concentration measured by FTIR spectroscopy
and to measure urea in the presence of water. Furthermore, the
HPLC method is capable of quantifying potential emissions of
byproducts such as biuret, triuret, cyanuric acid and ammelide
Fig. 3 Urea conversion in the empty reactor.
reactor was composed of 10% O and 0% or 5% H O in N , at a
2
2
2
ꢀ
1
total flow rate of 500 L h
pressure (STP: 0 1C, 1013 mbar).
A concentration of 100 ppm urea in the gas phase was
at standard temperature and
ꢀ
1
realized by spraying 4 wt% urea in ethanol (70.4 mL min ),
ꢀ
1
5
wt% urea in methanol (56.3 mL min ) or 15 wt% urea in
ꢀ
1
water solution (14.5 mL min ). Urea solutions were prepared by
dissolving urea (Merck, Z99.5% purity) in ethanol (Merck,
Z99.9% purity), methanol (VWR, o0.05% water) or de-ionized
water. Dosing of the organic solvents resulted in gas phase
concentrations of 0.31% ethanol or 0.36% methanol. When
urea in water solution was dosed, the water concentration in
the base feed was reduced to maintain an effective water
concentration of 5% in the reactor feed.
3
3,34
(Fig. 4),
but none of these byproducts were observed in the
present study, except for the biuret decomposition experi-
ments. The C-balance could usually be closed within ꢁ5%
2
accuracy by summing up the HNCO, urea and CO amounts.
2
The CO yield in the dry experiments due to hydrolysis by water
traces was mostly below 5%.
Non-catalytic urea decomposition could be neglected
because very little urea conversion was observed in the empty
reactor up to 200 1C (Fig. 3) and most of the data points
considered for the Arrhenius analyses were below 200 1C.
In addition to dry and wet urea decomposition, we performed
HNCO hydrolysis experiments (100 ppm HNCO, see the next
section) and biuret hydrolysis experiments with 2% biuret
Calculation of product yields and urea conversions
The yields of the reaction products NH , HNCO and CO were
3
2
calculated based on the N- or C-balance. Using the concentra-
tions was allowed, as all species were diluted to r200 ppm in
the model exhaust gas matrix:
NH
3
yield = c(NH3,out)/(2 ꢂ c(ureain))
ꢀ1
(Fluka, p.a.) in de-ionized water solution (95.3 mL min liquid
dosing rate, resulting in 50 ppm biuret).
HNCO yield = c(HNCOout)/c(ureain)
HNCO generation for HNCO hydrolysis experiments
CO yield = c(CO2,out)/c(ureain)
2
Urea was the reactant in most of the experiments. Additionally,
experiments with HNCO as the reactant were carried out for
comparison. HNCO was generated by thermolysis of cyanuric
acid in a separate reactor as described in ref. 1. Cyanuric acid
was sublimed at about 280 1C and thermolyzed downstream
urea slip = c(ureaout)/c(ureain
)
Urea conversion (X) was always calculated according to
X = 2 ꢂ NH yield ꢀ CO yield
3
2
2
over stainless steel nuts at 380 1C. The carrier gas flow (dry N )
ꢀ
1
through the HNCO generator was set to 20 L h at STP. After
leaving the HNCO generator, its product gas was immediately
Although it would have been easiest to calculate the urea
conversion based on the urea slip, the equation X = 2 ꢂ NH
yield ꢀ CO yield was used because of the following reasons: (a)
3
ꢀ
1
mixed with the main gas stream (total gas flow = 500 L h at
STP). The typical composition of the resulting gas mixture was
2
1
00 ppm HNCO, E15 ppm NH , E20 ppm CO and E3 ppm
3 2
CO. Before starting an experiment, the HNCO generator
was allowed to stabilize for about 1 h, while its output was
monitored by FTIR spectroscopy (bypassing the reactor containing
the catalyst-coated monolith). After the experiment, the HNCO
Fig. 4 Examples for urea decomposition byproducts that can be quantified by
generator output was measured again. Typically, the HNCO the applied HPCL method.
9
44 Catal. Sci. Technol., 2013, 3, 942--951
This journal is c The Royal Society of Chemistry 2013

View Article Online
Paper
Catalysis Science & Technology
ꢀ
1
NH and CO could be measured more precisely than urea; (b) a 9 g L , which is about 20 times lower than in a real application
3
2
moderate relative measuring error of the large urea slip at low in exhaust gas aftertreatment. Also, we chose a rather low urea
conversions would have resulted in a large relative error of the concentration of 100 ppm in the gas phase to allow measure-
conversion rates, required for the Arrhenius analyses; (c) the ments at low temperatures around 150 1C without urea
urea conversion is still indicated irrespective of the presence of condensation.
water (see explanation in the next two paragraphs) and (d) urea
condensation in the reactor would have indicated wrong urea thermolysis (a and b) and urea hydrolysis in the presence of 5%
conversion. water (c and d). Dry conditions, required for studying urea
Fig. 5 shows the results of a catalyst screening for dry urea
For the dry experiments, it would have been acceptable to thermolysis without hydrolysis, were realized by using ethanol
calculate the conversion based on the HNCO yield alone. and methanol solutions of urea. Fig. 5a and b also compare
However, intermediately formed HNCO, which is hydrolyzed urea thermolysis in the presence of ethanol, which was already
2
6
due to water traces, is missing in the HNCO yield. A ‘‘true’’ reported elsewhere, with thermolysis in the presence of
HNCO yield, which is corrected for the HNCO missing due to methanol. Changing the solvent from ethanol to methanol
2
HNCO hydrolysis, can be calculated by adding the CO yield influenced the thermolysis reaction only slightly, indicating
to the HNCO yield (provided that there is no direct urea that these water-free, polar solvents are suitable for studying
hydrolysis). The HNCO quantification by FTIR spectroscopy is the urea thermolysis only. In both cases, no compounds
highly sensitive, but more vulnerable to systematic errors than originating from side-reactions due to the presence of the
the quantification of the other compounds because calibration solvent could be detected by FTIR spectroscopy and the mass
of the reactive HNCO is difficult. The HNCO quantification by balance could often be closed more precisely than in the
2
6
HPLC is robust, but time consuming and less sensitive than the hydrolysis experiments.
quantification by FTIR spectroscopy. Hence, it was a better In the absence of water (Fig. 5a and b), the urea thermolysis
option to use the NH yield to calculate the yield of the NH products NH and HNCO were formed with a high selectivity.
produced by urea thermolysis alone (thermolysis-NH yield), Adding the urea slip (Fig. 5b) to the NH or HNCO yield
which is in principle equivalent to the ‘‘true’’ HNCO yield.
3
3
3
3
3
usually allowed for closing the mass balance. Only at the
lowest temperatures, the urea slip decreased due to urea
condensation.
‘
‘true’’ HNCO yield: HNCO yield + CO yield
2
Under hydrolysis conditions (Fig. 5c and d), NH and CO2
3
thermolysis-NH
3
yield: 2 ꢂ NH
3
yield ꢀ CO
2
yield
were the final products, however, only ZrO always showed a
2
high selectivity for CO . By contrast, TiO , Al O and H-ZSM-5
2
2
2 3
If only urea thermolysis takes place, the HNCO yield, the
showed significant local maxima in the HNCO yield at 165 1C,
80 1C and 190 1C, respectively. Local maxima of reaction
intermediates are typical for sequent reactions. SiO showed
‘
‘true’’ HNCO yield and the thermolysis-NH yield are identical
3
1
except for measuring errors. In the wet experiments, the chosen
definition of the urea conversion based on the thermolysis-NH
yield is valid as well. In fact, the thermolysis-NH yield indicates
the urea conversion irrespective of the product selectivity. Even
if direct urea hydrolysis (3) takes place, the thermolysis-NH
2
3
poor thermolysis and even lower hydrolysis activity. The hydro-
lysis activity was so low that even in the presence of water the
HNCO yield increased steadily with increasing temperature
3
3
(Fig. 5d).
yield correctly indicates the urea conversion, but, of course,
In light of the very high HNCO hydrolysis rates on anatase
without indicating anymore how much NH was produced by
14
3
2
TiO reported by Hauck et al. (2007), the intermediate HNCO
urea thermolysis. Anyway, if CO is produced with a very high
2
peaks shown in Fig. 5d may be surprising. Fig. 6 shows the
hydrolysis of 100 ppm HNCO on some of the previously tested
catalysts for urea hydrolysis. Indeed, we found HNCO hydro-
lysis to be very fast on TiO and even much faster on ZrO . The
2 2
HNCO slip on TiO2 at 165 1C was only 3%, whereas a local
maximum of 37% HNCO yield was observed at 165 1C during
selectivity, the best option is to simply base the urea conversion
on the CO
hydrolysis on ZrO
based on the thermolysis-NH
2
yield. We tested using the CO
, but the result was very close to the result
yield. Hence, we chose to use
2
yield to analyze urea
2
3
only one definition of the urea conversion. Using a uniform
definition of the urea conversion is simple and provides
excellent comparability between the results obtained under
different conditions (dry or wet and different catalysts).
urea hydrolysis on the same TiO -coated monolith (Fig. 5d).
2
Apparently, HNCO hydrolysis was strongly inhibited by the
presence of urea.
Fig. 7 shows a comparison of the HNCO yields observed in
the presence of water and without water. To indicate the urea
conversion irrespective of the product selectivity, Fig. 7 also
Results and discussion
Catalyst screening
shows the thermolysis-NH yield. If only urea thermolysis takes
3
Catalytic urea decomposition is a fast reaction. To realize place, the HNCO yield and the thermolysis-NH3 yield are
reaction conditions with little diffusion limitation, small identical except for measuring errors. Due to the strong inhibi-
amounts of the different catalyst powders were coated on tion of the HNCO hydrolysis reaction by the presence of urea,
cordierite monoliths with a high cell density of 600 cpsi. The the HNCO yield obtained on TiO
2
under hydrolysis conditions
typical catalyst loading of about 50 mg corresponded to only was only slightly lower than the HNCO yield without water up to
This journal is c The Royal Society of Chemistry 2013
Catal. Sci. Technol., 2013, 3, 942--951 945

View Article Online
Catalysis Science & Technology
Paper
Fig. 6 HNCO hydrolysis, using the same catalyst-coated monoliths as were used
for the experiments shown in Fig. 5.
higher without water at most temperatures, however, at the
lowest temperature investigated (160 1C), the difference was
small (without water: 6%, with water: 5% HNCO yield, see
Fig. 5).
Surprisingly, the HNCO yields obtained on the H-ZSM-5
2 3
(Fig. 7b) and on the Al O (Fig. 7c) catalyst were higher in the
presence of water than without water at some temperatures. On
H-ZSM-5 at 180 1C, the HNCO yield increased dramatically from
3
3% without water to 93% with water. On SiO
was higher with water over the whole temperature range
Fig. 7d).
The simplest explanation for the low HNCO yields in the dry
2
, the HNCO yield
(
experiments would be inhibition of urea thermolysis by the
organic solvent used to dose the urea. However, it is plausible
that the adsorption strength of the solvents on the metal oxide
catalysts decreases with their polarity in the order H
methanol > ethanol. Moreover, the concentration of the organic
solvents was about 15 times lower than the H O concentration
H O: 5%, methanol: 0.36%, ethanol: 0.31%). Hence, the
2
O >
2
(
2
organic solvents are likely to inhibit urea adsorption less
strongly than H O.
2
Another potential effect of the organic solvents that has to
be checked is the presence or absence of side-reactions like
condensation into ethers or substitution of the OH group with
NH . Our results show that the organic solvents did not induce
3
side-reactions because the C-balance of the dry urea thermo-
lysis reaction could be well closed by summing up the HNCO,
urea and CO yields. Also, we did not observe solvent-related
2
2
6
byproducts by FTIR spectroscopy or HPLC analysis.
Possibly, the presence of water on the catalyst surface
accelerated the urea thermolysis by facilitating proton transfer
reactions. This assumption of water assisting in the urea
thermolysis is supported by a theoretical study by Alexandrova
and Jorgensen (2007) that suggests water to act as a proton
shuttle for the formation of the zwitterionic intermediate
Fig. 5 Screening of catalysts for urea decomposition. (a, b) Dry experiments;
solid lines: ethanol, dashed lines: methanol. (c, d) Hydrolysis with 5% water.
(
a) includes the NH
3
yield according to the thermodynamic equilibrium of the
+
ꢀ
35
H N C(O)N H in the mechanism for urea thermolysis in aqueous
3
urea thermolysis reaction.
36
solution.
Table 3 summarizes the catalyst activity results from Fig. 5
60 1C (Fig. 7a). Above 165 1C, the HNCO yield decreased again and 6. Interestingly, ZrO , which showed the highest hydrolysis
1
2
due to HNCO hydrolysis. On ZrO , the HNCO yield was much activity, showed only low thermolysis activity (see also Fig. 9b).
2
9
46 Catal. Sci. Technol., 2013, 3, 942--951
This journal is c The Royal Society of Chemistry 2013

View Article Online
Paper
Catalysis Science & Technology
Table 3 Qualitative summary of the catalyst activities shown in Fig. 5 and 6
Catalyst
TiO
Urea thermolysis
Urea hydrolysis
HNCO hydrolysis
2
High
Low
Moderate
Moderate
Very low
High
Very high
Low
Moderate
Very low
High
ZrO
H-ZSM-5
Al
SiO
2
Very high
Low
1
7,37
2
O
3
oTiO
2
17,37
2
Very low
The high hydrolysis activity of the ZrO
low thermolysis activity indicates that urea hydrolyzed directly
3) on the ZrO surface without intermediate HNCO formation.
Alternatively, if HNCO was formed on ZrO during urea hydro-
2
catalyst in spite of its
(
2
2
lysis as a short lived intermediate, the low thermolysis activity
of ZrO might be due to a low proton transfer capability of the
2
dry ZrO2 surface. Anyway, since urea hydrolysis on the ZrO2
catalyst always produced CO
2
with a high selectivity, a kinetic
model with only one rate-determining step should be sufficient
to describe urea hydrolysis on ZrO .
2
Kinetics
For comparison of the urea conversions on the different
catalysts, we calculated pseudo first-order rate constants (k1)
using the equation
ꢃ
V
3
ꢀ1 ꢀ1
s
k1 ¼ ꢀ lnð1 ꢀ XÞ; cm g
W
where V* is the actual gas volume flow rate, W is the catalyst
weight and X is the urea conversion. By using this equation, we
assume a first-order dependency with respect to urea, whereas
all the other reaction orders are approximated by zero. The
1
same approach was used by Kleemann et al. (2000) to calculate
pseudo first-order rate constants of the HNCO hydrolysis (first-
order with respect to HNCO, water present in excess). Fig. 8
shows the resulting Arrhenius plot of the urea thermolysis data
presented in Fig. 5a.
To experimentally confirm that the urea conversion on the
most active catalyst in our screening, TiO , was not limited
2
by the thermodynamic equilibrium of the urea thermolysis
Fig. 7 Dry and wet urea decomposition. The data are from the same experi-
ments as those shown in Fig. 5. Dashed lines and empty symbols represent the
NH
HNCO. (a) TiO
conducted with urea in methanol solution (TiO
solution (Al and SiO ).
3
-yield by thermolysis, solid lines and filled symbols show the FTIR data for
, (b) H-ZSM-5, (c) Al and (d) SiO . Dry experiments were
and H-ZSM-5) or urea in ethanol
2
2
O
3
2
2
Fig. 8 Arrhenius analysis of the urea thermolysis experiments presented in
2
O
3
2
Fig. 5a.
This journal is c The Royal Society of Chemistry 2013
Catal. Sci. Technol., 2013, 3, 942--951 947

View Article Online
Catalysis Science & Technology
Paper
reaction, we performed an experiment with increased active the reactor exit was found to be almost the same with sprayed
mass. As expected, increasing the active mass from 45 mg (curve urea solution as with NH gas over a very broad parameter
‘TiO ’’ in Fig. 8) to 540 mg (curve ‘‘TiO c’’ in Fig. 8) on range, when the distance between the nozzle and catalyst was
3
‘
2
2
3
0
the monolith increased the HNCO yield from 22% to 56%, properly chosen.
respectively. Since the pseudo first-order rate constants shown
ꢅ Fig. 5c shows the CO yield obtained by urea hydrolysis in
in Fig. 8 are normalized to the active mass, the rates for the long (‘‘TiO
‘TiO c’’-monolith were below those of the ‘‘TiO ’’-monolith. shorter monolith with the higher amount of the catalyst
Still, the pseudo first-order rate constants at 150 1C were quite showed a higher CO yield at low temperatures up to 180 1C,
similar for the two TiO -coated monoliths due to the higher urea as expected for gaseous urea but not for urea aerosols. If
conversion on the ‘‘TiO c’’-monolith. Below 150 1C, the pseudo evaporation of the urea aerosols would not have been complete
2
2
2 2
’’) and short (‘‘TiO b’’) TiO -coated monoliths. The
‘
2
2
2
2
2
first-order rate constants from the ‘‘TiO c’’-monolith were prob- upstream of the monolith, aerosols would have preferably
2
ably lowered due to urea condensation, whereas, above 150 1C, slipped through the channels of the short monolith resulting
they were lowered due to mass transport limitation.
in a lower conversion at all temperatures.
To test the presence or absence of mass transfer limitation
in the Arrhenius plot (Fig. 8), we calculated ZextDaII values using catalyst-coated monolith at a larger distance from the spray
ꢅ In a first series of urea hydrolysis experiments, we placed a
the equation
nozzle. Then we placed an inert cordierite foam between the
spray nozzle and the catalyst. Neither of these measures
increased the urea conversion. If the urea aerosols evaporated
slowly, an inert structure upstream of the catalyst should have
improved the urea evaporation, which should also have
ꢀ
ꢁ
3=2
reff
D1;2
T
ZextDaII ¼
with kgðTÞ ꢄ
; cm s 1
ꢀ
kgðTÞac1;g
dg T0
3
8
where reff is the observed reaction rate, k is the gas phase increased the urea conversion on the catalyst.
g
mass transport coefficient, a is the geometric monolith surface
area, T is the actual temperature and T is the temperature for dosed by the spray nozzle evaporated upstream of the catalyst.
the tabulated value of the binary diffusion coefficient (D1,2). If the urea aerosols did not evaporate, at least the aerosol slip
These observations suggest that most of the urea aerosols
0
The ZextDaII value is a measure of the external mass transport through the catalyst-coated monoliths was insignificant.
limitation of a reaction (diffusion of gaseous urea in the
monolith channels). A ZextDaII value below 0.1 means the mass Determination of apparent activation energies
3
8
transport limitation is insignificant. The calculation of the
ZextDaII values was based on the following assumptions:
ꢅ complete evaporation and mixing of the dosed urea with
the model gas
Fig. 9a shows an Arrhenius plot of the urea thermolysis on
anatase TiO and calculated ZextDaII values for this experiment.
2
ꢅ laminar gas flow
1
ꢅ diffusion length (d
g
) = channel width
4
ꢅ The binary diffusion coefficient (D1.2) of urea in the model
gas was approximated by the binary diffusion coefficient of SO
2
2
ꢀ1
38
in air: Durea,modelgas E D
= 0.122 cm s at 298 K. The
SO ,air
2
values for slightly different gas matrices do not differ signifi-
cantly. The binary diffusion coefficient of SO was chosen for
2
the calculations with urea due to the quite similar molecular
mass and due to the not completely different geometry.
ꢅ conversion at the catalyst exit used for calculating the
concentration of gaseous urea (c1,g
)
Please note that using the conversion at the catalyst exit for
calculating c1,g means making a worst case assumption,
because the low c1,g at the catalyst exit leads to the assumption
of a small urea concentration gradient. Another assumption
made has to be discussed: the assumption of complete urea
evaporation. Unfortunately, we could not directly measure if
and where the urea aerosols evaporate in our reactor. However,
the vapor pressure of urea is more than high enough to allow for
3
3,39
complete urea evaporation under the conditions applied.
Indeed, several indications suggest that the high spray quality in
our setup allowed for quantitative urea evaporation upstream of
the catalyst.
ꢅ During method development, the performance of the SCR
Fig. 9 Arrhenius analyses of urea decomposition on anatase (a) TiO
. Urea in methanol solution was used for the dry experiments.
2
and
reaction with respect to NO
x
reduction and NH
3
emissions at (b) ZrO
2
9
48 Catal. Sci. Technol., 2013, 3, 942--951
This journal is c The Royal Society of Chemistry 2013

View Article Online
Paper
Catalysis Science & Technology
An ZextDaII value below 0.1 indicates insignificant mass trans-
3
8
port limitation. The pseudo first-order rate constants of the
urea hydrolysis were calculated likewise and plotted in Fig. 9a.
We are aware of the fact that the simple kinetic model is not
applicable for the two-step urea hydrolysis on TiO , where a
2
significant amount of the intermediate HNCO can desorb into
the gas phase. However, due to the definition of the conversion
as X = 2 ꢂ NH
3
yield ꢀ CO
2
yield, the NH
hydrolysis is subtracted and only the NH
3
produced by HNCO
produced by urea
3
thermolysis is taken into account. Therefore, the pseudo first-
order rate constants obtained for hydrolysis conditions (5%
water) should be suitable for comparison. Comparison of the
pseudo first-order rate constants shows that the hydrolysis was
somewhat slower than thermolysis below 150 1C, whereas the
hydrolysis was faster above 160 1C. Maybe, the hydrolysis was
slower below 150 1C due to competitive adsorption of urea and
2
Fig. 10 Biuret hydrolysis on the TiO d-monolith sample.
In the empty reactor, higher biuret conversions were
water, whereas the hydrolysis was faster above 160 1C because observed (not shown) than urea conversions shown in Fig. 3.
the surface coverage was lowered by HNCO hydrolysis or This result is surprising because biuret is more stable against
8
because adsorbed water accelerated the thermolysis by acting non-catalytic decomposition than urea. The high biuret
as a proton shuttle.
conversion in the empty reactor was probably caused by the
Based on the ZextDaII values and the apparent linearity of the low vapor pressure of biuret. Unlike urea, which seemed to
Arrhenius plots in the temperature range between 150 1C and evaporate quickly and completely in our experiments, part of
170 1C, the urea thermolysis data points in this temperature the biuret may have stuck to the reactor wall until it decom-
range were used for a linear regression (highlighted in Fig. 9a). posed. In other words, the residence time and the heat transfer
From the obtained regression line an apparent activation from the reactor wall to the reactant were increased for biuret
ꢀ
1
energy (E
a
) of 90 kJ mol and a pre-exponential factor (A) of compared to urea. The decomposition products urea, HNCO
1
3
ꢀ1
6ꢂ 10
s
were calculated for urea thermolysis on TiO2 and NH have higher vapor pressures than biuret and could
3
(Fig. 9a).
therefore easily leave the reactor. As mentioned, the biuret
The activity of the ZrO2 catalyst (Fig. 9b) was analyzed conversion in the empty reactor was quite high. Still, the TiO
likewise. If urea hydrolyzes in one step on ZrO according to catalyst largely increased the biuret conversion. At 190 1C,
reaction (3) as proposed in the introduction, the kinetic model where the urea yield with the TiO catalyst peaked at 0.29
used for the urea thermolysis can be applied for urea hydrolysis mol/mol-biuret, the yield in the empty reactor was only 0.08
on ZrO as well. The corresponding values derived from Fig. 9b mol/mol-biuret, which strongly indicates catalytic biuret
are E
2
2
2
2
ꢀ
1
13 ꢀ1
a
= 100 kJ mol , A = 4 ꢂ 10
s
for the thermolysis decomposition.
Catalytic biuret decomposition activity is a highly welcome
for the hydrolysis feature of a urea hydrolysis catalyst in an SCR system, because
(
2
(
considering the data points at 180, 190, 195, 200, 205, 210 and
ꢀ1
11 ꢀ1
15 1C) and E = 70 kJ mol , A = 8 ꢂ 10
s
a
considering the data points at 125, 130, 135 and 140 1C).
biuret that may form during low temperature operation will
quickly hydrolyze at elevated temperatures. We have observed
catalytic biuret decomposition before in a recent study on the
Biuret hydrolysis
40
formation and decomposition of byproducts on TiO
2
.
In that
4
0
Byproduct formation was not observed in the experiments study, slow partial catalytic biuret hydrolysis yielding urea
reported above, however, byproducts may have accumulated was observed at only 100 1C. At 160 1C, melamine was cataly-
slowly, which could cause catalyst deactivation after long-term tically hydrolyzed into ammeline, ammelide and cyanuric acid.
operation. The most probable byproducts are biuret, cyanuric Cyanuric acid hydrolysis started at about 200 1C. Hence,
4
0
acid, triuret and ammelide. Since only biuret has a significant temperatures around 200 1C should be sufficient to regenerate
solubility in water, 2% aqueous biuret solution was used in this a TiO2 catalyst after poisoning with biuret, cyanuric acid,
study to investigate byproduct decomposition.
Fig. 10 shows the product distribution of biuret hydrolysis easily reached in urea-SCR applications. The decomposition of
on TiO . Biuret hydrolyzed fairly well in a three-step reaction, more stable byproducts like melem was not investigated in
ammelide, ammeline and melamine. Such temperatures are
2
showing local maxima of the formed intermediates urea and ref. 40 Melem is very unlikely to ever form on TiO , because
2
HNCO. The urea yield peaked at 190 1C and 0.29 mol/ melem can only be formed from melamine at very high
2
2
mol-biuret, whereas HNCO peaked at 200 1C and 0.23 mol/ temperatures around 500 1C, when melamine is already
mol-biuret. When the water concentration in the gas phase was hydrolyzed in the presence of water. Notably, Al O was
2
3
4
1
decreased from 5% to 1.4%, the urea peak increased to reported to catalytically hydrolyze even melem. TiO , which
2
0
2
.36 mol/mol-biuret, whereas the HNCO peak shifted to showed a higher hydrolysis activity than Al O in our study,
2 3
10 1C and increased to 0.52 mol/mol-biuret (not shown).
may thus catalyze melem hydrolysis, too.
This journal is c The Royal Society of Chemistry 2013
Catal. Sci. Technol., 2013, 3, 942--951 949

View Article Online
Catalysis Science & Technology
Paper
Conclusions
Acknowledgements
Experiments on the catalytic hydrolysis of 100 ppm urea to NH
3
Funding by TOTAL (France) is gratefully acknowledged. TOTAL
and CO
hydrolysis activities: ZrO
Dry experiments on the thermolysis of urea to HNCO and NH
revealed a different activity order: TiO > H-ZSM-5 E Al
ZrO > SiO . For the most active catalysts TiO and ZrO the
activation energies for hydrolysis and thermolysis were deter-
mined by means of an Arrhenius analysis of the apparent
pseudo first-order rate constants:
2
in the gas phase showed the following order of was not involved in the study design, collection, analysis or
2
> TiO
2
> Al
2
O
3
> H-ZSM-5 > SiO
2
.
interpretation of data, nor was it involved in the preparation of
the article.
3
2
2 3
O >
2
2
2
2
Notes and references
1
M. Kleemann, M. Elsener, M. Koebel and A. Wokaun, Ind.
Eng. Chem. Res., 2000, 39, 4120–4126.
ꢀ
1
2 M. Koebel, M. Elsener and M. Kleemann, Catal. Today, 2000,
9, 335–345.
Urea thermolysis on TiO
2
:
a
E = 90 kJ mol
5
ꢀ1
3 HSC Chemistry 5.11. Computer program by Outokupu
Research.
Urea thermolysis on ZrO2: Ea = 100 kJ mol
4
E. Abu-Ramadan, K. Saha and X. Li, AlChE J., 2011, 57,
210–3225.
ꢀ
1
Urea hydrolysis on ZrO2: Ea = 70 kJ mol
3
5
H. Aoki, T. Fujiwara, Y. E. Morozumi and T. Miura, Measure-
ments of urea thermal decomposition reaction for NO selective
non-catalytic reduction, Lisbon, Portugal, 1999.
O. Carp, Rev. Roum. Chim., 2001, 46, 735–740.
A. Lundstr o¨ m, B. Andersson and L. Olsson, Chem. Eng. J.,
The found activation energies for the thermolysis of urea are
ꢀ
1
larger than the activation energy of 73 kJ mol reported for the
1
4
2
hydrolysis of HNCO on anatase TiO by Hauck et al. (2007).
6
7
Consequently, our results clearly show that urea thermolysis is
slower than HNCO hydrolysis (Fig. 5 vs. Fig. 6) and urea
thermolysis must be the rate-determining step in catalytic urea
decomposition in the SCR process.
2009, 150, 544–550.
8
P. A. Schaber, J. Colson, S. Higgins, D. Thielen, B. Anspach
and J. Brauer, Thermochim. Acta, 2004, 424, 131–142.
A. Schmidt, Oesterr. Chem. Ztg, 1967, 6, 175–179.
2
On TiO , the urea conversion in the absence of water was
9
quite similar to the urea conversion with water (Fig. 9a). The
main effect of the presence of water was a changed selectivity of
the overall reaction due to HNCO hydrolysis. By contrast,
10 T. J. Wang, S. W. Baek, S. Y. Lee, D. H. Kang and G. K. Yeo,
AIChE J., 2009, 55, 3267–3276.
11 S. D. Yim, S. J. Kim, J. H. Baik, I. S. Nam, Y. S. Mok, J. H. Lee,
B. K. Cho and S. H. Oh, Ind. Eng. Chem. Res., 2004, 43,
the ZrO
2
catalyst showed a much lower urea conversion
in the absence of water than with water, which indicates that
4856–4863.
urea hydrolyzes on ZrO according to a different mechanism,
2
1
1
2 E. F. Zanoelo, Chem. Eng. Sci., 2009, 64, 1075–1084.
3 A. Lundstr o¨ m, B. Waldheim, H. Str o¨ m and B. Westerberg,
presented in part at the Proceedings of the Institution
of Mechanical Engineers, Part D: Journal of Automobile
Engineering, October 1, 2011.
in which water directly attacks adsorbed urea rather than
adsorbed HNCO (Fig. 9b).
Appendix
14 P. Hauck, A. Jentys and J. A. Lercher, Appl. Catal., B, 2007,
a
A
c
diffusion surface area
70, 91–99.
pre-exponential factor in the Arrhenius equation
volume fraction for gas phase concentrations; mass
fraction for liquid phase concentrations
reactant concentration in the gas phase
diffusion length
binary diffusion coefficient
second Damk o¨ hler-number
apparent activation energy
effective reaction rate
apparent pseudo first-order reaction rate constant
gas phase mass transport coefficient
efficiency factor, Zext = reff/r
gas constant
actual temperature
15 P. Hauck, A. Jentys and J. A. Lercher, Catal. Today, 2007,
127, 165–175.
16 O. Kr o¨ cher and M. Elsener, Chem.–Eng. J., 2009, 152,
167–176.
17 G. Piazzesi, Dissertation, ETH Zurich, Switzerland, 2006,
thesis no. 16693.
18 G. Piazzesi, M. Devadas, O. Kr o¨ cher, M. Elsener and
A. Wokaun, Catal. Commun., 2006, 7, 600–603.
19 G. Piazzesi, O. Kr o¨ cher, M. Elsener and A. Wokaun, Appl.
Catal., B, 2006, 65, 55–61.
c
1,g
dg
D1.2
DaII
E
a
r
eff
k
1
k
g
20 G. Piazzesi, D. Nicosia, M. Devadas, O. Kr o¨ cher, M. Elsener
and A. Wokaun, Catal. Lett., 2007, 115, 33–39.
21 I. Czekaj, O. Kr o¨ cher and G. Piazzesi, J. Mol. Catal. A: Chem.,
2008, 280, 68–80.
Z
ext
R
T
T0
V*
W
X
temperature for the tabulated value of D1,2
total flow rate at the actual temperature and pressure
catalyst weight
22 M. Eichelbaum, R. J. Farrauto and M. J. Castaldi, Appl.
Catal., B, 2010, 97, 90–97.
23 H. L. Fang and H. F. M. DaCosta, Appl. Catal., B, 2003, 46,
17–34.
relative conversion
9
50 Catal. Sci. Technol., 2013, 3, 942--951
This journal is c The Royal Society of Chemistry 2013

View Article Online
Paper
Catalysis Science & Technology
2
2
2
2
2
2
3
3
3
4 M. A. Larrubia, G. Ramis and G. Busca, Appl. Catal., B, 2000, 33 A. M. Bernhard, I. Czekaj, M. Elsener, A. Wokaun and
7, L145–L151.
O. Kr o¨ cher, J. Phys. Chem. A, 2011, 115, 2581–2589.
5 A. Lundstr o¨ m, T. Snelling, P. Morsing, P. Gabrielsson, 34 M. Koebel and M. Elsener, J. Chromatogr., A, 1995, 689,
E. Senar and L. Olsson, Appl. Catal., B, 2011, 106, 273–279. 164–169.
6 A. M. Bernhard, D. Peitz, M. Elsener and O. Kr o¨ cher, Top. 35 HSC Chemistry 7. Computer program by Outokupu
Catal., submitted. Research.
7 T. Todorova, D. Peitz, O. Kr o¨ cher, A. Wokaun and B. Delley, 36 A. N. Alexandrova and W. L. Jorgensen, J. Phys. Chem. B,
J. Phys. Chem. C, 2011, 115, 1195–1203. 2007, 111, 720–730.
8 A. M. Bernhard, Dissertation, ETH Zurich, Switzerland, 37 I. Czekaj and O. Kr o¨ cher, Top. Catal., 2009, 52, 1740–1745.
2
submitted, thesis no. 20813.
9 D. Peitz, Dissertation, ETH Zurich, Switzerland, 2012, thesis
no. 20568.
38 M. Baerns, H. Hofmann and A. Renken, Chemische
Reaktionstechnik, Springer, Georg Thieme Verlag, Stuttgart,
1992.
0 D. Peitz, A. Bernhard, M. Elsener and O. Kr o¨ cher, Rev. Sci. 39 A. P. Krasulin, A. A. Kozyro and G. Y. Kabo, Russ. J. Appl.
Instrum., 2011, 82, 084101. Chem., 1987, 60, 96–99.
1 Corning Celcor standard-wall substrates, http://www.corn 40 A. M. Bernhard, D. Peitz, M. Elsener, A. Wokaun and
ing.com/WorkArea/showcontent.aspx?id=6275, 2012.
O. Kr o¨ cher, Appl. Catal., B, 2012, 115–116, 129–137.
2 O. Kr o¨ cher, M. Elsener and E. Jacob, Appl. Catal., B, 2009, 41 Z. Zhan, M. M u¨ llner and J. A. Lercher, Catal. Today, 1996,
8, 66–82. 27, 167–173.
s
8
This journal is c The Royal Society of Chemistry 2013
Catal. Sci. Technol., 2013, 3, 942--951 951