Can TiO2 promote the reduction of nitrates in water?

Article abstract of DOI:10.1016/j.jcat.2005.06.015

Monometallic palladium catalysts were synthesized using different titanium supports and tested for the reduction of nitrates from aqueous solutions using hydrogen as a reductant. The Pd/TiO₂ catalysts were characterized by electron paramagnetic resonance (EPR), low-temperature Fourier transform infrared (FTIR) spectroscopy of adsorbed CO, and X-ray diffraction (XRD). The catalysts studied exhibited a high activity for nitrate removal with a lower tendency for nitrite formation than the conventional bimetallic Pd catalysts. Although ammonium formation was greater than desired, the use of a monometallic catalyst for this two-step reduction process is significant and suggests that a single site may be responsible for both reduction stages. The titanium support (particularly the Ti³⁺ centers generated during prereduction in the presence of Pd) appear to play an important role in the nitrate degradation process. The potential role of Pd β-hydride in generating these Ti ³⁺ centers is discussed.

Full text of DOI:10.1016/j.jcat.2005.06.015

Journal of Catalysis 234 (2005) 282–291
www.elsevier.com/locate/jcat
Can TiO2 promote the reduction of nitrates in water?
Jacinto Sá a,b, Thomas Berger b, Karin Föttinger b, Alexander Riss b, James A. Anderson a,∗,
Hannelore Vinek b
a
Surface Chemistry and Catalysis Group, Department of Chemistry, University of Aberdeen, Meston Walk, Old Aberdeen AB24 3UE, UK
b
Institute of Materials Chemistry, Vienna University of Technology, Veterinaerplatz 1, A-1210 Vienna, Austria
Received 10 June 2005; accepted 16 June 2005
Available online 9 August 2005
Abstract
Monometallic palladium catalysts were synthesized using different titanium supports and tested for the reduction of nitrates from aque-
ous solutions using hydrogen as a reductant. The Pd/TiO catalysts were characterized by electron paramagnetic resonance (EPR), low-
2
temperature Fourier transform infrared (FTIR) spectroscopy of adsorbed CO, and X-ray diffraction (XRD). The catalysts studied exhibited a
high activity for nitrate removal with a lower tendency for nitrite formation than the conventional bimetallic Pd catalysts. Although ammo-
nium formation was greater than desired, the use of a monometallic catalyst for this two-step reduction process is significant and suggests that
3+
a single site may be responsible for both reduction stages. The titanium support (particularly the Ti centers generated during prereduction
in the presence of Pd) appear to play an important role in the nitrate degradation process. The potential role of Pd β-hydride in generating
3+
these Ti centers is discussed.
2005 Elsevier Inc. All rights reserved.
Keywords: Pd/TiO ; Low-temperature Fourier transform infrared spectroscopy; Nitrate reduction; Aqueous solutions
2
1. Introduction
in the EU have nitrate concentrations in groundwater ex-
ceeding the imposed guideline level (25 ppm).
The demand for good-quality water has increased with
the increased extent of water usage; however, excessive ap-
plication of nitrate fertilizers in agriculture can consequently
lead to the leaching of nitrates into groundwater and hence
into surface water. Nitrate levels above a certain concen-
tration (50 ppm in the European Union [EU]) are deemed
unacceptable. High nitrate concentrations in drinking water
are harmful because of the reduction of nitrates to nitrites,
which combine with hemoglobin in the blood to form methe-
moglobin; furthermore, nitrates can cause cancer and hyper-
tension through nitrosamine formation [1].
From an environmental standpoint, one of the most ac-
ceptable methods of removing nitrates is to convert them
to N2. Liquid-phase nitrate hydrogenation over a catalyst ap-
pears to be one of the most attractive processes to achieve
this [2–20]. Much of the research on catalytic reduction
has focused on the use of bimetallic catalysts, but more re-
cently, studies involving monometallic catalysts, (in particu-
lar, using semiconducting oxides as supports) have appeared
[21,22].
As part of this strategy, the use of TiO2 impregnated
with Pd was investigated. These types of catalysts have been
widely investigated in liquid-phase hydrogenation processes
[23–26], of which nitrate reduction from aqueous solu-
tions bears some relation. The catalysts were then submitted
to pretreatments involving different reduction temperatures.
The influences of these thermal treatments on the sample
characteristics were evaluated by Fourier transform infrared
(FTIR) spectroscopy using CO as a probe molecule and elec-
The problem related to the contamination of groundwater
is becoming more pronounced. The conclusion is based on
the report of the European Environmental Agency released
in 1998, which estimated that 87% of the agricultural lands
*
Corresponding author.
E-mail address: j.anderson@abdn.ac.uk (J.A. Anderson).
0021-9517/$ – see front matter
doi:10.1016/j.jcat.2005.06.015
2005 Elsevier Inc. All rights reserved.

J. Sá et al. / Journal of Catalysis 234 (2005) 282–291
283
tron paramagnetic resonance (EPR) at low temperature, and
kinetic reactions were investigated using a standard batch re-
actor.
X-Ray diffraction (XRD) patterns were obtained using
a Philips X’Pert PRO diffractometer (Bragg–Brentano θ/θ,
geometry type). The Cu-Kα radiation source, a graphite
(002) monochromator, provides continuous scans in the re-
gion 2θ = 5–80◦, with a step size of 0.02◦ and a scan rate of
5 s/step.
2. Experimental
The EPR sample cell, made of Suprasil quartz glass, was
connected to an appropriate high-vacuum pumping system.
It permitted thermal sample activation at <10−5 mbar under
dynamic vacuum conditions and reduction of the catalysts.
The EPR spectra were recorded with a Bruker EMX 10/12-
spectrometer system in the X-band mode. Measurements
at 77 K were performed with an ER 4131 VT variable-
temperature accessory. The presented spectra were obtained
by accumulating 10–30 scans to obtain acceptable signal-to-
noise ratios. The g values were determined on the basis of
a DPPH standard. The electron center signal was measured
at 77 K at a fixed-resonance magnetic field value for the re-
spective EPR transitions.
Two sets of experiments were performed on the EPR.
Each set consisted of a consecutive reduction of the calcined
catalysts in 1 mbar of pure H2. The temperatures used to per-
form the reductions were 298, 473, and 623 K. After reduc-
tion, one of the sets was allowed to cool in the presence of
H2 before the call was evacuated, whereas in the other proce-
dure, the cell was evacuated at the temperature at which the
reduction was carried out and cooled under dynamic vac-
uum. Cell evacuation took 10 min. O2 (10 mbar) was dosed
after each reduction step for both procedures to locate the
particular species, because adsorbed oxygen strongly modi-
fies the signals resulting from surface species [30].
TiO2 P25 (P25) with a surface area of 49 m2/g (N2, 77 K,
BET method [27]) support was supplied by Degussa, and the
monocrystalline titania (anatase (Ana) with a surface area of
93 m2/g (N2, 77 K, BET method [27]) and rutile (Rut) with
a surface area of 16 m2/g (N2, 77 K, BET method [27])) sup-
ports were self-prepared through chemical vapor deposition.
Briefly, the TiO2 particles were produced by means of ther-
mal decomposition of titanium isopropoxide vapor in a flow
reactor system by the metal organic chemical vapor deposi-
tion method [28,29]. The product powder was then gradually
annealed (10 K/min) to 870 K under high-vacuum condi-
tions (P < 10−5 mbar) and treated in vacuum for 2 h.
The resultant material was separated into two parts. The
first part was oxidized with 3 Torr O2 at 870 K to remove
organic remnants from the precursor material. The second
was oxidized with 3 Torr O2 at 1070 K for the same reason
and also to allow a crystal-phase transformation.
All supports were pretreated in a flow of 40 ml/min of
pure O2 for 5 h at 870 K before being impregnated with Pd
salt. This was done mainly to remove some retained mois-
ture from the supports, especially from the commercial sup-
port (P25). After this thermal treatment, the supports were
impregnated with Pd(NO3)2 · 2H2O (Merck) using the wet
impregnation method. A ratio of 10 ml of water per gram
of support was used, and the weight of the salt was calcu-
lated to obtain a final weight percentage of noble metal in
the catalysts of 2. The resultant catalysts where then cal-
cined at 770 K to remove nitrate and then gradually reduced
(10 K/min) under a 50-ml/min flow of pure H2 at different
temperatures (298, 473, and 623 K) for 1 h.
Monometallic Pd/γ -Al2O3 and bimetallic Pd–Cu/γ -
Al2O3 were synthesized using incipient wetness method, as
described [18], using γ -Al2O3 (Sigma Aldrich) and the cor-
responding nitrate salts, Pd(NO3)2 · 2H2O and Cu(NO3)2 ·
3H2O (Merck). The 5 wt% Pd catalyst was dried overnight
at 373 K and calcined in air at 723 K. For the bimetallic
catalyst, a subsequent impregnation was carried out with
Cu(NO3)2 · 3H2O using the same procedure; the resulting
bimetallic catalyst was dried overnight at 373 K and cal-
cined in air at 723 K. The catalyst was reduced for 1 h at
623 K in pure H2. The final metal content of the catalysts
was 5 wt% of Pd and 1.25 wt% of Cu. The bimetallic cat-
alyst was synthesized for comparative purposes in kinetic
measurements, because these catalysts are the most widely
studied in the nitrate hydrogenation process. The metal con-
tent of the catalysts was evaluated by atomic absorbance
spectroscopy (AAS) and found to be in agreement with the
nominal values.
Spin concentrations were estimated by numerical dou-
ble integration of the first-derivative spectra and compared
with a frozen Cu2+ aqueous solution, measured in an identi-
cal sample tube, as described elsewhere [31]. Thus a similar
level of accuracy in the results is expected. Cu2+ provides a
suitable comparison because it belongs to the same row as Ti
and has the same number of unpaired electrons as Ti3+ (one
1
2
unpaired electron); that is, both have S = .
FTIR spectra were recorded on a Bruker IFS 28 with a
resolution of 4 cm−1. The spectrometer cell was connected
to a vacuum system working in the 10−6 mbar range and to
a heating system. The cell could be used for in situ pretreat-
ment of samples, adsorption of gases, and measurements
at low temperatures. The samples were pressed into self-
supporting wafers, which were placed inside a ring furnace
in the vacuum cell. CO adsorption measurements were car-
ried out using 50 mbar of pure CO. The presented spectra
were obtained at 100 K by accumulating 100 scans.
The catalysts were investigated after in situ consecutive
reduction with 100 mbar of pure H2 for 1 h, using the same
temperatures of the EPR experiments, and cell evacuation
for 30 min at the reduction temperature. Because the volume
of the FTIR cell was greater than that of the EPR tube, the
reductions were carried out with a higher hydrogen pressure
for the latter, to normalize the number of moles of H2 ex-

284
J. Sá et al. / Journal of Catalysis 234 (2005) 282–291
posed to the catalyst for both experiments. The catalyst in
calcined form was used as the starting material. Changes in
the sample resulting from exposure to O2 were also investi-
gated. Catalysts were exposed to 10 mbar of pure O2 at room
temperature (290 3 K) for 1 h. The cell was then evacuated
for 15 min at same temperature before the samples were ex-
posed to CO.
The activities and selectivities of the catalysts in the
hydrogenation of nitrate ions were tested in a slurry ther-
mostated batch reactor with a capacity of 1 l, equipped with
a mechanical glass stirrer. In a typical run, 450 mg of the cat-
alyst was charged into a reactor containing 0.75 l of distilled
water. The content of the reactor was flushed first for 5 min
with N2 and then for 60 min with CO2 and H2. The reac-
tion was started when a solution of nitrate salt was added to
the vessel to achieve a concentration of 100 ppm of nitrates
(KNO3, Merck). A constant feed of CO2 (100 ml/min) as a
pH buffer and H2 (90 ml/min) as a reducing agent were sup-
plied using mass flow controllers. The reactions were carried
out at atmospheric pressure, at 298 K, and using a stirring
rate of 500 rpm. A typical kinetic run took 4 h.
oxidizing characteristics of the treatment were sufficient to
allow the phase transition of part of the anatase to rutile.
3.2. EPR characterization
EPR was performed to evaluate the formation of Ti3+
centers and their stability in the presence of O2. Two sets of
experiments were followed by EPR. In the first set, the cata-
lysts were allowed to cool under an atmosphere of hydrogen
after the in situ reduction. Fig. 1 shows the EPR spectra ob-
tained for the catalysts Pd_P25. It should be emphasized that
when the experiment was performed using Pd_Ana, similar
results were observed (not shown). In contrast, no signals
were observed with the pure supports in the absence of Pd.
The spectra are very similar in term of both intensity and lo-
cation of the signal for all reduction temperatures. The center
of the main peak is located at g = 1.995 and is attributed
to Ti3+ centers [30,32,33]. The g-tensor value was signifi-
cantly higher than values normally reported for Ti3+ in solid
TiO2 (g ≈ 1.990) [30,31], which would appear to be related
to the presence of Pd. A similar effect on the g-tensor was
observed by Huizinga et al. [34] for platinized TiO2, where
a value of 1.992 was measured and was attributed to Ti3+
formed in the vicinity of platinum clusters.
No perturbation in the intensity of the peaks was ob-
served when 10 mbar of O2 (not shown) was dosed. This
insensitivity to oxygen, the narrow line width, and the high
g-tensor value clearly indicate that the signal derives from
bulk Ti3+ centers, that is, from trapped electrons in the sup-
port lattice. This implies a surface reoxidation as the Ti3+
centers generated are located at the surface. This process in-
volves charge and/or mass transfer from the surface to the
bulk. Because the transport of surface-reduced Ti cations
to the bulk is largely unfavorable under the experimental
conditions used [35], it is assumed that the process occurs
through a charge transfer. Based on the present results, it is
not possible to distinguish between interstitial and substitu-
tional Ti3+. Despite the extensive discussion of this issue,
no agreement has yet been reached. Because this issue is be-
The degradation of nitrates and the formation of nitrites
and ammonia were measured at constant time intervals usin−g
high-performance liquid chromatography (HPLC) for NO3
−
and NO2 and a colorimetric test supplied by Merck for
NH4+. The colorimetric tests were performed in a Shimadzu
UV–vis spectrophotometer operating at λ = 690 nm, and the
HPLC measurements were performed in a Spectra Physics
device with an UV detector operating at λ = 210 nm. The
column used was an IonoSphere 5A, and the mobile phase
was a phosphate buffer with concentration of 0.02 mol and a
pH of 5.3.
3. Results and discussion
3.1. XRD characterization
The percentages of each crystal phases present in each
TiO2 support investigated are summarized in Table 1. It can
be seen that the anatase/rutile ratio for the sample P25 is
almost 1, instead of the expected value of 3. This is be-
cause of the thermal treatment applied to the support before
the impregnation with Pd. The temperature applied and the
Table 1
Percentages of the crystal phases in the TiO
2
Catalyst
Reduction temperature (K)
Anatase (%)
Rutile (%)
Pd_P25 298
Pd_P25 473
Pd_P25 623
Pd_Ana 473
Pd_Ana 623
Pd_Rut 473
Pd_Rut 623
298
473
623
473
623
473
623
55
55
55
95
95
0
45
45
45
5
5
100
100
Fig. 1. EPR of Pd_P25 collected at 77 K, reduced at different temperature
0
and cooled in 1 mbar of H .
2

J. Sá et al. / Journal of Catalysis 234 (2005) 282–291
285
Pd β-hydride is assumed to decompose at 343–363 K [21,
36,37]. This implies that the presence of Pd β-hydride is pos-
sible only when the catalysts are cooled in an atmosphere
of hydrogen (first set of experiments) and/or the catalysts
are reduced and evacuated at 298 K (second set of experi-
ments). These conditions are those under which Ti3+ center
formation was observed. Therefore, it is proposed that Ti3+
center formation occurs mainly through hydride reduction
rather then through hydrogen spillover. This does not invali-
date the hypothesis of Ti4+ reduction by hydrogen spillover;
however, under the conditions described here, its contribu-
tion was the less significant of the two processes.
During catalyst reduction, both adsorbed and absorbed
hydrogen species are formed; both of these processes are
reversible. Throughout the increase in temperature, the hy-
dride (potentially the active species involved in the reduction
of the support) dissolves [38,39], taking with it the electrons
trapped at the Ti3+ centers. In the case that the catalysts are
allowed to cool in hydrogen (first set of experiments), fur-
ther formation of hydride is allowed, and the signal related
to Ti3+ centers is observed. The insignificant increase in the
signal with increased reduction temperature is an indication
that the reduced support due to hydrogen spillover is not very
pronounced. In the second series of experiments, hydride
formation during cooling was prevented, and thus hydride
decomposition occurred, leading to removal of the electrons
trapped at Ti3+ centers, as confirmed by the disappearance
of the signal related to Ti3+ centers when the catalysts were
reduced and evacuated at higher temperatures.
For comparison, the spin concentrations resulting from
the calculations using the frozen Cu2+ aqueous solution
were 2.5 × 1019 spins/g catalyst for Pd_Ana and 3.1 ×
1018 spins/g catalyst for Pd_P25. Using this rough estima-
tion, it is possible to determine the percentage of support
that was partially reduced under the experimental conditions
applied. The values obtained were ∼0.4% for Pd_Ana and
∼0.1% for Pd_P25.
Fig. 2. EPR of Pd_Ana, collected at 77 K, reduced at different temperatures
with 1 mbar of H and evacuated before cooling.
2
yond the scope of the present work, we can conclude only
that the Ti3+ centers have bulk characteristics and that they
appear to be located in the vicinity of Pd.
In the second set of experiments, the cell was evacuated
at the reduction temperature and the catalysts were cooled
under dynamic vacuum. Fig. 2 shows the EPR spectra ob-
tained for the catalysts Pd_Ana; the measurements made
using Pd_P25 (not shown) gave similar results. The signal
related to Ti3+ (g = 1.991) was measurable only when the
catalysts had been reduced with 1 mbar of H2 at 298 K and
only when Pd was present. Another important observation
was that the shape of the profiles remained the same after
the catalysts were exposed to O2 (not shown). The very nar-
row line width indicates that Ti3+ occupies a single type
of site within the anatase. Following the foregoing reason-
ing, the Ti3+ centers are assumed to be present within the
support lattice. In this set of experiments an additional sig-
nal was detected that was diminished in intensity with an
increase in reduction temperature such that it remained un-
detected when the catalysts were reduced at 623 K. Because
the g factor of the signal is significantly greater than 2, the
signal must derive from another type of species, which in
this case is assumed to be Pd, because no other chemical
species are present in the catalysts. The signal is assigned to
residual, incompletely reduced paramagnetic Pd+ species.
Furthermore, the catalyst color was visibly darker with an
increase in the reduction temperature, indicating a reduction
of Pd to its metallic state.
3.3. Adsorption of CO on reduced Pd/TiO2 catalysts
FTIR measurements using CO as a probe molecule were
performed to identify the species present on the surface of
the catalysts, their oxidation states, and their susceptibility
to the presence of oxygen in situ after reduction. Because
of the weak nature of CO adsorption on exposed sites of
the support, such as CO bonded to Ti3+ [40], experiments
were conducted with the sample at 100 K. The results of
the interactions of CO with the surface species of the cat-
alysts supported on P25 and anatase after in situ reduction
and evacuation are shown in Figs. 3a and b. The spectra in
each figure were obtained for the same sample submitted
consecutively to different reduction temperatures and CO
adsorption. CO was previously evacuated at room temper-
ature before subsequent reduction; thus it is assumed that
the reduction procedures were identical for the FTIR and
EPR experiments. Similar studies were also performed for
Any discussion of these results must begin by explain-
ing why the support can be reduced at such low temperature.
The explanation for this must be related to the presence of
Pd in the catalysts. It is known that Pd promotes reduction
of the support, allowing reduction to occur at a lower tem-
perature [23]. This effect is due to the spillover of hydrogen
and/or the presence of Pd β-hydride (absorbed hydrogen).
Because these two mechanisms are possible, the question is
which one is responsible for the formation of the Ti3+ cen-
ters measured?

286
J. Sá et al. / Journal of Catalysis 234 (2005) 282–291
tion on exposed Ti species (band centered at 2180 cm−1
)
and the decreased intensity of bands related to adsorption
of CO on exposed Pd sites (bands centered at 2100, 1995,
and 1940 cm−1). The band centered between 2176 and
2180 cm−1 is assigned to Ti4+–CO β-acid sites [40–42],
that is, five coordinated Ti ions. The shift of the band to
high wave numbers with an increase in reduction tempera-
ture is related to dehydroxylation of the surface occurring
during sample evacuation. It should be remembered that the
catalysts were evacuated at the reduction temperature for
30 min to remove any water formed during the reduction
process. The band was more pronounced in the presence
of Pd. A slight decrease in band intensity was observed for
samples containing Pd when the reduction was carried out at
higher temperatures and exposed to oxygen. But in the case
of the support alone, the intensity decreased strongly after
the addition of oxygen. In the case of catalysts reduced at
lower temperature (298 K), the band decreased by 30%. It is
possible to conclude that the noble metal has two functions:
promoting the formation of coordination vacancies and pre-
venting their recombination with molecular oxygen.
(a)
A small band located at 2210 cm−1 was observed more
clearly in spectra of the catalysts containing anatase as a sup-
port. This is attributed to another acidic cation, the α-Ti4+
acid sites [40,41], that is, four coordinated Ti ions or Ti with
two coordination vacancies. Oxygen dosing had no effect on
band intensity. Furthermore, this band was not observed for
the support alone in the absence of the metal, which sup-
ports the argument that Pd promotes the development and
stabilization of coordination vacancies. A consequence of
the decreased coordination number of terminal Ti4+ is an
increased surface positive charge of the catalysts. It should
be mentioned that the interaction of CO with the α-Ti4+ is
known to be much stronger and that its corresponding band
intensity does not depend on the temperature at which the
measurements were performed, whereas the intensity of the
band representing the β sites has a temperature dependence
(i.e., more intense when the measurements are performed at
lower temperatures) [40].
(b)
Fig. 3. FTIR spectra following exposure of samples (a) Pd_P25; and (b)
Pd_Ana to 50 mbar of CO at 100 K, for samples reduced at different tem-
peratures.
The band between 2150 and 2158 cm−1 is assigned to CO
adsorbed on Ti–OH [40]. The weak bands, which are visible
in the spectra of the Pd/TiO2 anatase catalyst at 2144 and
2138 cm−1, can be related to the same sites but with slightly
different environments. They are apparent with greater inten-
sity then normally expected, because the measurements were
performed at low temperature. These features decreased with
increasing reduction temperature due to the dehydroxylation
process. The final band, which can be attributed to adsorp-
tion at exposed Ti cations, is located at 2166 cm−1 and is
assigned to Ti4–(CO)2 species [40]. This band was observed
only in the sample containing anatase as a support.
For spectra of catalysts containing Pd, three different
types of adsorbed species due to the interaction of CO with
different Pd sites can be identified. The highest-frequency
band at 2100 cm−1 is attributed to CO linearly bonded to Pd
[24,40,43]. Bands at 1995 and 1940 cm−1 are assigned to
Fig. 4. Modifications to CO adsorption following exposure of the catalyst
to 10 mbar of O . The spectra presented are the difference between the
2
spectrum of CO adsorption before and after dosing of O . FTIR collected
2
at 100 K and adsorption carried out with 50 mbar of CO on Pd_P25.
the supports in the absence of Pd (results not shown). Fig. 4
shows the perturbation in the intensity of bands due to ad-
sorbed CO resulting from exposure to O2.
Two features are evident in the spectra of the catalysts
studied as a function of the increased reduction temperature:
the increased intensity of the bands corresponding to adsorp-

J. Sá et al. / Journal of Catalysis 234 (2005) 282–291
287
compressed and isolated bridge carbonyls, respectively [24,
40,43,44]. All bands decreased in intensity with increasing
reduction temperature. The decreased number of exposed Pd
sites available for CO adsorption could be a consequence of
the onset of strong metal support interaction (SMSI) and/or
a sintering process.
The term SMSI was first proposed by Tauster et al.
[45,46] to indicate the modification of the adsorption abil-
ity of metals from groups 8, 9, and 10 when supported on
partially reducible oxides such as TiO2 and reduced at ele-
vated temperatures (generally 773 K). The standard criterion
for SMSI is a significant reduction in H2 or CO adsorption
capacity with elevated reduction temperature without an ac-
companying increase in crystallite size [23,45,46].
with exposed Ti species when the samples are considered
in the SMSI state (473 and 623 K), whereas for the support
in the absence of metal, band intensity decreased by around
30%.
Before concluding the characterization section, we note
that FTIR measurements were not possible for the rutile sup-
ported catalysts using the methodology applied for the other
two types of catalyst. This failure was related to the difficulty
in obtaining adequate IR transmission through the sample
after the reduction pretreatments. Therefore, we present no
characterization for this type of catalyst. However, for the
sake of providing a useful comparison in terms of the role
of the support, we conducted activity measurements using
these samples.
The SMSI phenomenon can be explained either by the
decoration of the metal surface with partially reduced ox-
ides [47–53] or by an electron transfer process between the
support and the dispersed metal [47,54,55]. Both of these
situations lead to decreased chemisorbtion of CO on the
metal. In the first situation this is due to the decrease in the
surface metal atoms, whereas in the second it is due to elec-
tronic modification, which leads to a weakened adsorbate–
adsorbent bond.
In the present study, FTIR measurements of chemisorbed
CO were carried out at 100 K, which facilitates CO adsorp-
tion and maximizes coverage, as confirmed by the appear-
ance of bands due to CO bonded to some Ti species, which
are otherwise highly unstable at room temperature [40]. Un-
der these experimental conditions, diminished intensity of
bands due to adsorbed CO is thus not expected should SMSI
manifest itself in terms of an electron transfer process lead-
ing to weakened adsorption. It is more likely that the dimin-
ished CO uptake, as indicated by the reduced intensity of CO
absorption bands with increasing reduction temperatures, is
a consequence of the SMSI state, which in this case results
from decoration of Pd particles by partially reduced TiO2.
Sintering is expected to play at most a minor role in re-
duced CO uptake; this is based on the assumption that the
catalysts were previously calcined at a temperature above the
3.4. Catalytic performance
The results of activity measurements, maximum nitrite
concentration, and selectivity of the catalysts to nitrogen are
plotted in Figs. 5a–c, with values summarized in Table 2.
The overall degradation profiles of nitrate via nitrite to nitro-
gen and ammonium over all catalysts are compared in Fig. 6.
The activities were calculated from the slope of the initial
linear section of the conversion curve (up to 20 min reaction
time); the selectivities were calculated using the ammonium
concentration remaining at the end of the catalytic test (4 h).
The TiO2-supported catalysts displayed a relatively high
activity for the removal of nitrates. In some cases, higher val-
ues than those of the reference PdCu bimetallic catalyst were
obtained. These results are surprising given literature reports
that TiO2 might be less promising as a support [6]. Another
important observation is that the nitrite concentration was
very low or reached undetectable levels for the monometal-
−
lic catalysts. The final concentration of NO2 was 0 for all
catalysts except Pd_Ana623 (Table 2), for which a level of
0.3 ppm was recorded. According to the reaction mechanism
proposed by Wärna˘ et al. [13], nitrates are reduced in a step-
wise fashion, initially to nitrites over bimetallic sites, before
migrating to monometallic sites where subsequent nitrite re-
duction takes place. Generally, this two-step process leads to
a release of NO2− into the solution, which may be further re-
duced at a later stage. In this way a potentially relatively high
concentration of nitrites in solution during reaction might
exist for the bimetallic catalysts, consistent with data given
here for the PdCu sample (Fig. 5b). Because the final nitrite
concentration was either very low or mostly undetectable for
the monometallic catalysts (Table 2), and the maximum ob-
tained (Fig. 5b) were also very low for these samples, we
assume that the active sites for the reduction of nitrates to
nitrites are the same as for subsequent steps (nitrites toward
nitrogen or ammonium) and that in general, the release of ni-
trites into solution followed by subsequent reduction is not
the predominant pathway.
palladium Tammann temperature (T = 593 K, T ∼ 0.5Tmpt
)
[56,57], which would be expected to lead to a greater degree
of sintering than occurs during the reduction procedure. An
acceptable criterion for the occurrence of sintering is that the
treatment temperature is above the Tammann temperature.
In only one case was the reduction temperature sufficiently
high for sintering to occur on the basis of this criterion. Fur-
thermore, there was no increase in the ratio between bridge
CO and linear CO carbonyls on Pd, which could also be in-
terpreted in terms of morphological changes that generally
accompany any particle growth process [24].
Establishment of the catalyst system in an SMSI state al-
lows stabilization of the generated coordination vacancies or
acidic sites of Ti, which is not expected in the absence of
strong interactions between metal and support. This is illus-
trated in Fig. 5, where the dosing of oxygen leads to only
minor decreases in the bands due to the interaction of CO
As mentioned earlier, the accepted mechanism for the re-
duction of nitrates to nitrites requires promotion of the noble
metal by addition of a second metal, that is, a bimetallic site,

288
J. Sá et al. / Journal of Catalysis 234 (2005) 282–291
(a)
(b)
(c)
−
Fig. 5. Catalytic performances of the catalysts: (a) activity (after 20 min reaction); (b) maximum concentration of NO ; (c) selectivity to N .
2
2
Table 2
Activity and selectivity of the different catalysts in the hydrogenation of 100 ppm nitrates
−
(ppm)
−
f
(ppm)
Catalyst
Activity
(mol/(min g )) × 10
[NO
]
[NO
]
Selectivity towards N
(mol%)
max
2
2
2
−5
cat
Pd_P25 298
Pd_P25 473
Pd_P25 623
Pd_Ana 473
Pd_Ana 623
Pd_Rut 473
Pd_Rut 623
1.9
5.1
8.0
5.2
6.1
6.2
3.4
0.1
5.2
0.0
0.0
1.76
0.0
1.1
0.0
0.0
0.0
7.8
0.0
0.0
0.0
0.0
0.3
0.0
0.0
0.0
0.0
31
56
0
49
7
53
11
78
77
Pd/γ -Al O
Pd-Cu/γ -Al O
2
3
2
3
−
−
f
[NO
]
, maximum nitrite concentration observed during the run; [NO ] , final nitrite concentration.
max
2
2

J. Sá et al. / Journal of Catalysis 234 (2005) 282–291
289
ered. The activity is increased with increasing temperature
at which the catalysts were reduced (except where rutile was
used as the support); however, the EPR results show only a
minor increase in the Ti3+ signal when the reduction tem-
perature was increased. Therefore, it is assumed that another
parallel process is occurring.
FTIR experiments confirm the presence of the SMSI con-
dition and also suggest that it is related to a decoration of
Pd particles by partially reduced TiO2 suboxide species. The
chemical structure of the partially reduced TiO2 is believed
to be Ti4O7 [39] from studies using electron diffraction and
lattice spacing measurements. This Ti4O7 phase was de-
tected during reduction of the TiO2 support, which was not
observed in the absence of Pt. This electronically enriched
local structure contains potential active sites for nitrate re-
duction. In fact, the development of an SMSI state, apart
from the enhanced support Lewis acid site density, was the
only significant change observed with an increase in reduc-
tion temperature. It can be assumed that the development of
the SMSI state was directly related to the increased activ-
ity, despite a reduction in the total number of exposed metal
surface sites. This apparent paradox regarding activity en-
hancement due to formation of an SMSI state but a reduction
in the total number of sites is not without precedent [26,60].
Tacke and Vorlop [7] found the reaction order to be 0.7
with respect to nitrate and independent of hydrogen partial
pressure, provided that the latter pressure was >1.0 bar. Pin-
tar et al. [9] suggested that the reaction was first order at
lower nitrate concentrations and zero order at higher con-
centrations, which is an indication that nitrate adsorption fol-
lows a Langmuir-type process at the respective active centers
on the catalyst surface. Therefore, a great excess of metal
sites would be expected for the reduction of nitrates. SMSI
is expected to lead to only a decreasing number of potential
active sites, not to poisoning of the catalyst. The possibil-
ity that the remaining sites were much more active and/or
that new, active sites (e.g., at the Pd–support interface) were
created should not be discounted [25].
The rate-determining step in the nitrate conversion process
is the reduction of nitrates to nitrite. In this case this step is
controlled by generation of the Ti3+ centers, because in the
nitrate reduction profiles for these catalysts, a steep decrease
in concentration is observed during the initial stages of the
reaction. This decrease appears to be related to the rapid
consumption of previously stabilized electrons, followed by
hydroxylation of the acid sites, and a significant decrease in
the activity beyond that expected for a first-order consump-
tion of nitrate from the reaction media.
In terms of catalyst selectivity, the catalysts supported on
TiO2 exhibit relatively poor selectivity, which can be jus-
tified by the known strong hydrogenation character of this
type of catalyst as well as the fact that high activity is of-
ten compensated by poorer catalyst selectivity, due in this
case to overreduction. Furthermore, the results indicate that
the overall reaction is not strongly dependent on the amount
of nitrite accumulated in the liquid phase, as has been sug-
Fig. 6. Nitrates conversion profiles as a function of time for all catalysts.
because monometallic catalysts supported on Al2O3 show
poor activity [18] (Table 2). Previous studies [18,58] have
demonstrated that nitrates might be reduced to nitrites on
metallic copper according to a redox reaction, leading to the
oxidation of copper species. Subsequently, the role of the
noble metal is to activate hydrogen and enable copper reduc-
tion. A similar type of mechanism could well be envisaged
here; however, in this case the promoting role of the noble
metal involves generation/stabilization of reduced species of
the titanium support.
After nitrates are introduced to a solution, they are ad-
sorbed at exposed acid sites (oxygen vacancies) of the sup-
port through electrostatic interactions. These sites are ex-
pected to be located in the vicinity of the Pd particles, be-
cause otherwise it is difficult to envisage a scenario in which
they would have the required stability. A similar type of ad-
sorption site has been postulated by Epron et al. [22].
Because nitrate degradation is a reduction process, the
transfer of electrons is an integral part of the reaction
scheme. These electrons are envisaged as being located at
the Ti3+ centers, generated by the catalyst from the Pd
β-hydride during the flushing of the reaction mixture with
hydrogen before the introduction of nitrates. These centers
are extremely reactive; however, EPR evidence suggests that
they are located within the lattice of the support, which
would make them inaccessible for the reaction unless they
were conducted to the surface, where reaction could take
place. Whether ele−ctrons are transferred from the support
directly to the NO3 or primarily to the metal and from this
to the anion is an open question. The charge transfer from
the support to the metal has been the subject of discussion,
with the suggestion that the electronic interactions present in
a SMSI state could result in a charge transfer from the sup-
port to the metallic particles [47]. Kim et al. [59] observed
a shift in the Pd XRD peak for Pd/TiO2 reduced at 773 K,
which they attributed to negative charge transfer from Ti3+
to the Pd particles.
It seems plausible that the charge transfer is occurring
during the reaction; however, the catalytic results cannot
be fully accounted for if this is the only pathway consid-

290
J. Sá et al. / Journal of Catalysis 234 (2005) 282–291
gested for the bimetallic catalysts [61]. The poisoning re-
sulting from using a solution buffer (CO2) is not as rele-
vant as in the case reported for Pd/CeO2 [22]. This finding
is significant, because it offers opportunities for the design
of further catalyst development based on titanium with the
objective of improving the nitrogen selectivity of these cata-
lysts.
Acknowledgments
The work was partially supported by the Fonds zur
Förderung der Wissenschaftlichen Forschung under Project
FWF P13521. The authors thank Dr. Erich Halwax of the In-
stitut für Chemische Technologien und Analytik of Vienna
University of Technology for performing the XRD measure-
ments and the College of Physical Sciences, University of
Aberdeen for providing a postgraduate research studentship
(to J.S.).
The other important topic relevant to the present dis-
cussion is the structure dependence of the activity. We ob-
serve that all catalysts are active, although not to the same
extent. In the case of the anatase and P25 (mixed-phase)
structures, increased activity with increasing reduction tem-
perature was observed. This is assigned to increased Lewis
acidity of the support, which enhances the initial adsorp-
tion of the nitrate ions. For rutile, different behavior was
observed that may be related to the reduction temperature,
although further speculation is not desirable at this stage,
because performing the IR measurements for these samples
was not possible. Somewhat surprising was the finding that
the highest values for activity were obtained for the cat-
alysts supported on P25, not those supported on the pure
anatase, which would suggest further degrees of complexity
in terms of the optimum support structure/exposed crystal
planes. This is an extremely complex topic of discussion
and beyond the scope of the present work; however, we
believe that the boundaries between the crystal structures
may act as electron-trapping sites involved in the reduction
process.
References
[1] L.W. Canter, Nitrate in Groundwater, CRC Press, Boca Raton, FL,
1997.
[2] K.-D. Vorlop, T. Tacke, M. Sell, G. Strauss, European Patent DE
3830850 A1 (10 September, 1988).
[3] T. Tacke, K.-D. Vorlop, Dechema Biotechnology Conferences, vol. 3,
VCH, Weinheim, 200X.
[4] K.-D. Vorlop, T. Tacke, Chem. Ing. Technol. 61 (1989) 836, (1989)
1007.
[5] K.-D. Vorlop, S. Hörold, K. Pohlandt, Chem. Ing. Technol. 64 (1992)
82.
[6] S. Hörold, K.-D. Vorlop, T. Tacke, M. Sell, Catal. Today 17 (1993) 21.
[7] T. Tacke, K.-D. Vorlop, Chem. Ing. Technol. 65 (1993) 1500.
[8] G. Strukul, R. Gavagnin, F. Pinna, E. Modaferri, S. Perathoner, G.
Centi, M. Marella, M. Tomaseli, Catal. Today 55 (2000) 139.
[9] A. Pintar, J. Batista, J. Levec, T. Kajiuchi, Appl. Catal. B 11 (1996)
81.
[10] A.J. Lecloux, Catal. Today 53 (199) 23.
[11] U. Prüsse, K.-D. Vorlop, Catal. Today 55 (2000) 79.
[12] F. Deganello, L.F. Liotta, A. Macaluso, A.M. Venezia, G. Deganello,
Appl. Catal. B 24 (2000) 265.
[13] J. Wärna˘, I. Turunen, T. Salmi, T. Maunula, Chem. Eng. Sci. 49 (1994)
5763.
4. Conclusion
In answer to the question raised in this paper’s title,
TiO2 would appear to promote the reaction leading to re-
moval of nitrates from aqueous sources. The reaction mech-
anism consists of two parallel processes catalyzed by an
excess of electrons trapped by the support, in the form of
either Ti3+ centers or partially reduced support in intimate
contact with Pd crystallites (SMSI). Decreased amounts of
adsorbed CO were measured when the reduction temper-
ature was increased as a consequence of the SMSI state,
whereas EPR measurements confirmed the presence of Ti3+
centers at 298 K (reaction temperature), possibly resulting
from the generation of Pd β-hydride species. The SMSI
state allowed a stabilization of the Lewis acidic cus species.
Monometallic catalysts showed a low selectivity to unde-
sired nitrites throughout the reaction, indicating that these
intermediates are rapidly reduced, possibly in a single site
mechanism, unlike the bimetallic PdCu samples, in which
nitrite ions are initially released into solution before being
reduced in a second step. Overreduction to ammonia resulted
in a poorer selectivity to nitrogen for the titanium-based
Pd monometallic samples than for the bimetallic analogues.
However, the high activity exhibited by Pd/TiO2 opens up
new opportunities in the design of catalysts for the denitra-
tion process.
[14] S. Naïto, M. Tanimoto, J. Catal. 119 (1989) 224.
[15] F. Epron, F. Gauthard, C. Pinéda, J. Barbier, J. Catal. 198 (2001) 309.
[16] S. Kerkeni, E. Lamy-Pitara, J. Barbier, Catal. Today 75 (2002) 35.
[17] A. Pintar, Catal. Today 77 (2003) 451.
[18] J. Sá, H. Vinek, Appl. Catal. B 57 (2004) 247.
[19] J. Sá, S. Gross, H. Vinek, Appl. Catal. A, submitted for publication.
[20] J. Sá, D. Gasparovicova, K. Hayek, E. Halwax, J.A. Anderson, H.
Vinek, Catal. Lett., submitted for publication.
[21] R. Gavagnin, L. Biasetto, F. Pinna, G. Strukul, Appl. Catal. B 38
(2002) 91.
[22] F. Epron, F. Gauthard, J. Barbier, J. Catal. 206 (2002) 363.
[23] Y. Li, B. Xu, Y. Fan, N. Feng, A. Qui, J.M.J. He, H. Yang, Y. Chen,
J. Mol. Catal. A 216 (2004) 107.
[24] J.H. Kang, E.W. Shin, W.J. Kim, J.D. Park, S.H. Moon, J. Catal. 208
(2002) 31.
[25] J. Silvestre-Albero, A. Sepúlveda-Escribano, F. Rodrígues-Reinoso,
J.A. Anderson, J. Catal. 223 (2004) 179.
[26] J.D. Bracey, R. Burch, J. Catal. 86 (1984) 384.
[27] S. Brunauer, P.H. Emmet, E. Teller, J. Am. Chem. Soc. 60 (1938) 309.
[28] S. Benfer, E. Knozinger, J. Mater. Chem. 9 (1999) 1203.
[29] D.M. Schleich, B. Walter, Nanostruct. Mater. 8 (1997) 579.
[30] R.F. Howe, M. Grätzel, J. Phys. Chem. 91 (1987) 3906.
[31] R.F. Howe, M. Grätzel, J. Phys. Chem. 89 (1985) 4495.
[32] D.C. Hurum, A.G. Agrios, K.A. Gray, T. Rajh, M.C. Thurnauer,
J. Phys. Chem. B. 107 (2003) 4545.
[33] P. Meriaudeau, M. Che, P.C. Gravelle, S. Teichner, Bull. Soc. Chim.
France 1 (1971) 13.
[34] T. Huizinga, R. Prins, J. Phys. Chem. 85 (1981) 2156.

J. Sá et al. / Journal of Catalysis 234 (2005) 282–291
291
[35] M.A. Henderson, Surf. Sci. 343 (1995) L1156.
[36] J. Batista, A. Pintar, D. Mandrino, M. Jenko, V. Martin, Appl. Catal.
A 206 (2001) 113.
[49] H.R. Sadeghi, V.E. Henrich, J. Catal. 87 (1984) 279.
[50] A.J. Simoens, R.T.K. Baker, D.J. Dwyer, C.R.F. Lund, R.J. Madon,
J. Catal. 86 (1984) 359.
[37] T.-C. Chang, J.-J. Chen, C.-T. Yeh, J. Catal. 96 (1985) 51.
[38] M. Boudart, H.S. Hwang, J. Catal. 39 (1975) 44.
[39] R.T.K. Baker, E.B. Prestridge, G.B. McVicker, J. Catal. 89 (1984)
422.
[40] K.I. Hadjivanov, G.N. Vayssilov, Adv. Catal. 47 (2002) 307.
[41] C. Morterra, J. Chem. Soc., Faraday Trans. 1 84 (1988) 1617.
[42] V. Bolis, B. Fubini, E. Garrone, C. Morterra, J. Chem. Soc., Faraday
Trans. 1 85 (1989) 1383.
[51] Y.-W. Chung, G. Xiong, C.C. Kao, J. Catal. 85 (1984) 237.
[52] C.S. Ko, R.J. Gorte, J. Catal. 90 (1984) 59.
[53] G.B. Raupp, J.A. Dumesic, J. Catal. 95 (1987) 587.
[54] J.M. Herrmann, M. Gravelle-Rumeau-Maillot, P.C. Gravelle, J. Ca-
tal. 104 (1987) 136.
[55] P. Chou, M.A. Vannice, J. Catal. 104 (1987) 1.
[56] N.M. Rodriguez, S.G. Oh, R.A. Dalla-Betta, R.T.K. Baker, J. Ca-
tal. 157 (1995) 676.
[57] P.O. Thevenin, E. Pocoroba, L.J. Petterson, H. Karhu, I.J. Väyrynen,
S.G. Järas, J. Catal. 207 (2002) 139.
[43] D. Tessier, A. Rakai, F. Bozon-Verduraz, J. Chem. Soc., Faraday
Trans. 88 (1992) 741.
[58] A. Edelmann, W. Schießer, H. Vinek, A. Jentys, Catal. Lett. 69 (2000)
11.
[59] W.J. Kim, J.H. Kang, I.Y. Ahn, S.H. Moon, Appl. Catal. A 268 (2004)
77.
[60] S.Y. Wang, S.H. Moon, M.A. Vannice, J. Catal. 71 (1981) 167.
[61] A. Pintar, T. Kajiuchi, Acta Chim. Slovenica 42 (1995) 431.
[44] K. Choi, M.A. Vanice, J. Catal. 127 (1991) 465.
[45] S.J. Tauster, S.C. Fung, R.L. Garten, J. Am. Chem. Soc. 100 (1978)
170.
[46] S.J. Tauster, S.C. Fung, J. Catal. 53 (1978) 29.
[47] R. Green, P. Morrall, M. Bowker, Catal. Lett. 98 (2004) 129.
[48] J. Santos, J. Phillips, J.A. Dumesic, J. Catal. 81 (1983) 147.