Selective photoreduction of nitric oxide to nitrogen by nanostructured TiO2 photocatalysts: Role of oxygen vacancies and iron dopant

Article abstract of DOI:10.1021/ja302246b

Conventional TiO₂-based photocatalysts oxidize NO_x to nitrate species, which do not spontaneously desorb and therefore deactivate the catalyst. We show that the selectivity of this reaction can be changed by creating a large concentration of oxygen vacancies in TiO₂ nanoparticles through thermal reduction in a reducing atmosphere. This results in the photoreduction of nitric oxide (NO) to N₂ and O₂, species which spontaneously desorb at room temperature. The activity of the photoreduction reaction can be greatly enhanced by doping the TiO₂ nanoparticles with Fe³⁺, an acceptor-type dopant that stabilizes the oxygen vacancies. Moreover, the photoinduced reduction of Fe³⁺ to Fe²⁺ provides a recombination pathway that almost completely suppresses the formation of NO₂ and thus enhances the selectivity of the reaction for N₂ formation. Gas chromatography confirms that N₂ and O₂ are formed in a stoichiometric ratio, and the activity for NO decomposition is found to be limited by the concentration of oxygen vacancies. A series of internally consistent reaction equations are proposed that describe all experimentally observed features of the photocatalytic process. The observed influence of oxygen vacancies on the activity and selectivity of photoinduced reactions may lead to new routes toward the design of highly selective photocatalysts.

Full text of DOI:10.1021/ja302246b

Article
pubs.acs.org/JACS
Selective Photoreduction of Nitric Oxide to Nitrogen by
Nanostructured TiO2 Photocatalysts: Role of Oxygen Vacancies and
Iron Dopant
Qingping Wu and Roel van de Krol*
Materials for Energy Conversion and Storage (MECS), Department of Chemical Engineering, Delft University of Technology, P.O.
Box 5045, 2600 GA Delft, The Netherlands
ABSTRACT: Conventional TiO2-based photocatalysts oxidize
NOx to nitrate species, which do not spontaneously desorb and
therefore deactivate the catalyst. We show that the selectivity of
this reaction can be changed by creating a large concentration of
oxygen vacancies in TiO2 nanoparticles through thermal
reduction in a reducing atmosphere. This results in the
photoreduction of nitric oxide (NO) to N2 and O2, species
which spontaneously desorb at room temperature. The activity
of the photoreduction reaction can be greatly enhanced by
doping the TiO2 nanoparticles with Fe3+, an acceptor-type
dopant that stabilizes the oxygen vacancies. Moreover, the photoinduced reduction of Fe3+ to Fe2+ provides a recombination
pathway that almost completely suppresses the formation of NO2 and thus enhances the selectivity of the reaction for N2
formation. Gas chromatography confirms that N2 and O2 are formed in a stoichiometric ratio, and the activity for NO
decomposition is found to be limited by the concentration of oxygen vacancies. A series of internally consistent reaction
equations are proposed that describe all experimentally observed features of the photocatalytic process. The observed influence of
oxygen vacancies on the activity and selectivity of photoinduced reactions may lead to new routes toward the design of highly
selective photocatalysts.
material’s ability to remove NOx from air. To avoid
deactivation, the nitrates need to be washed away by rain.14
The resulting nitric acid is corrosive and could pollute the soil
when the concentration becomes too high. The removal of
NOx from air without deactivation and secondary pollution is
therefore an urgent and demanding challenge.
One of the most promising ways to resolve this problem is to
change the selectivity of the photocatalytic reaction so that NOx
is converted to N2 and O2. No deactivation would occur for this
photoreduction reaction since nitrogen and oxygen readily
desorb from the surface.15 As reported by Anpo et al.,16 the
selectivity toward NO photoreduction can be greatly improved
by reducing the coordination number of Ti4+ from its usual
value of 6 (TiO6 octahedra) to 4 (TiO4 tetrahedra). This has
been successfully achieved by depositing isolated TiO4 clusters
inside the cavities of zeolite-Y with ion beam implantation.16,17
However, the large-scale application of zeolites with ion beam
implantation techniques is economically unattractive.
INTRODUCTION
■
Most of the world’s energy consumption is based on the
oxidation of fossil fuels in air. Examples are internal combustion
engines in cars and turbines for power generation plants. These
processes produce massive amounts of greenhouse gases, such
as CO2 and NOx. NOx (a mixture of NO and NO2)1 is formed
when atmospheric nitrogen and oxygen react as a result of the
high temperatures that are reached during fuel combustion.2,3
Over the past few decades, atmospheric NOx concentrations
have greatly increased because of the growing number of
automobiles and growing industrial activities.4 This is reason
for concern, since the emission of NOx causes damage to
human lung tissue and contributes to the formation of acid
rain.5 TiO2, one of the best-known semiconductor photo-
catalysts, can decompose NOx at room temperature and
ambient pressure and has been widely studied for this
purpose.6−9 When TiO2 is irradiated with photon energies
exceeding its band gap of ∼3.2 eV,10 electrons are excited from
the valence band (VB) to the conduction band (CB), resulting
in the formation of electron−hole (e−−h+) pairs. A certain
fraction of these charge carriers are able to reach the surface of
the TiO2 and are captured by surface-adsorbed species on Ti
sites to form superoxide anions and hydroxyl radicals.11,12 The
free radicals are very active and can react with NO to form
nitrates.13 The main problem of this approach, however, is that
these nitrates cannot spontaneously desorb. These species
therefore deactivate the photocatalyst’s surface, reducing the
In this paper, we propose a new strategy to change the
photocatalytic selectivity of TiO2 based on the creation of a
large and stable concentration of oxygen vacancies. We will
show that this indeed results in the photoreduction of NO to
N2 and O2 and that the photo-oxidation reaction can be largely
Received: March 7, 2012
Published: May 18, 2012
© 2012 American Chemical Society
9369
dx.doi.org/10.1021/ja302246b | J. Am. Chem. Soc. 2012, 134, 9369−9375

Journal of the American Chemical Society
Article
with an online chemiluminescence-based NOx analyzer (Teledyne
Instruments, model 200E) with a measurement range of 0−2 ppm.
The samples were placed in a 10 cm wide reactor trough, 5 mm below
a fused silica optical window, consistent with international standard
NEN-ISO 22197-1:2007. A continuous 1 L/min flow of ∼1000 ppb
NO in air, pure N2, or pure He (Linde Gas Belgium NV) passed over
the sample surface, which was irradiated by a UV light source (75 W
facial tanner, Philips HB172) with an intensity of ∼2 mW/cm2. The
NO and NOx (= NO + NO2) concentrations were continuously
recorded every 10 s, while the NO2 concentration was automatically
calculated by the analyzer from the concentration difference between
NOx and NO.
The N2 and O2 concentrations were measured by online gas
chromatography. The gas chromatograph (Shimadzu, GC-2014) was
equipped with a pulsed discharge detector (PDD) operating at 493 K.
The photocatalytic products were separated by a combination of
Porapak Q and GsBP-PLOT molsieve columns using pure helium
(99.9999%) as the carrier gas. To exceed the detection limit of the gas
chromatograph, a flow of 100 ppm NO in He was used as the target
pollutant for GC measurements.
suppressed. A series of reaction mechanisms that explain these
observations will be given.
EXPERIMENTAL SECTION
■
Synthesis of Fe-doped TiO2 thin films. A simple, template-free
sol−gel method was employed for synthesizing pure TiO2 and Fe-
doped TiO2 (Fe/TiO2) colloidal solutions.18 Briefly, titanium
isopropoxide (TTIP; Acros, 98+%) was dropwise added to ultrapure
water, slightly acidified with nitric acid, under vigorous stirring. For the
preparation of Fe-doped TiO2 samples, a certain amount of
Fe(NO3)3·9H2O was dissolved into the aqueous solution before
addition of the TTIP. After hydrolysis of TTIP in the aqueous
solution, an opaque suspension was obtained, which contains TiO2
and propanol as the main reaction products. A homogeneous colloidal
solution was produced after evaporation of the propanol at 333 K in a
rotary evaporator. Fe-doped TiO2 powders were synthesized with
different Fe concentrations ranging from 0% Fe (undoped) to 1% Fe
(atomic Fe/Ti ratio). Both TiO2 and Fe/TiO2 mesoporous films were
fabricated by tape casting 300 μL sample solutions onto blank glass
substrates (Schott Borofloat 33, 10 × 5 cm2). The solutions were made
by mixing 2 mL of the colloid (density 0.130 g/mL) with 180 μL of
10% Triton X-100 in H2O and 0.02 g of polyethylene glycol. After tape
casting, the films (area 10 × 3.8 cm2) were dried in air to evaporate the
water and fired at 723 K in air to remove any remaining organic
components. The same method, but without the tape casting step, was
used to convert part of the Fe/TiO2 colloidal solutions into powders.
Structural Characterization. The crystal structures of Fe-doped
TiO2 and pure TiO2 nanoparticles were analyzed by X-ray diffraction
(XRD; Bruker D8 Advance) using Co Kα radiation (λ = 0.178897
nm). The specific surface areas of the powder samples were
determined by Brunauer−Emmett−Teller (BET) adsorption measure-
ments on a Quantachrome Autosorb-6B instrument at 77 K in liquid
nitrogen. Prior to these measurements, the samples were pretreated in
vacuum at 623 K for 16 h. The Raman spectra were recorded by a
Renishaw Raman imaging microscope (system 2000). A 514.5 nm Ar+
laser line with a power output of 20 mW was used for excitation. A
Leica DMLM optical microscope with a Leica PL floutar L500/5
objective lens was used to focus the beam on the sample. The 520
cm−1 peak of a Si wafer served as a wavelength reference for the
Raman spectra.
RESULTS AND DISCUSSION
■
To investigate the influence of oxygen vacancies on the
decomposition of NO over undoped TiO2, the photocatalytic
activity for stoichiometric TiO2 (as-prepared TiO2 that is fully
oxidized by a heat treatment at 723 K in air) and reduced TiO2
(heat treatment at 723 K in a 2% H2/Ar atmosphere) are
compared. The results are shown in Figure 2, and we briefly
Activity Measurements. The photocatalytic activity of the
samples was evaluated in a continuous-flow setup (Figure 1) equipped
Figure 2. Photocatalytic degradation of NO in air for (a) oxidized and
(b) reduced TiO2 under UV light irradiation.
describe the overall features here before presenting more
detailed reaction schemes later in this paper. The photocatalytic
reactions are initiated by illuminating the sample with UV light
after 1 h of equilibration in a ∼1 ppm NO atmosphere in the
dark. For oxidized TiO2, the NO concentration immediately
decreases by ∼800 ppb upon illumination (Figure 2a). This is
due to the reaction of gas-phase NO species with adsorbed
−
superoxide anions (O2 ), resulting in the formation of nitrate
groups. The superoxide anions are formed by the reduction of
adsorbed O2 by photoexcited electrons. Since the nitrate groups
do not spontaneously desorb, the surface slowly saturates and
the NO concentration at the reactor outlet increases again.14,19
Figure 1. Schematic diagram of the continuous-flow NOx photo-
catalytic setup. Air, pure nitrogen, or helium was used as the carrier
gas.
9370
dx.doi.org/10.1021/ja302246b | J. Am. Chem. Soc. 2012, 134, 9369−9375

Journal of the American Chemical Society
Article
Integration of the NOx peak area below the 1040 ppb baseline
in Figure 2a yields a total number of 0.69 × 1018 nitrate
adsorbates, which corresponds to ∼5% of the total number of
surface sites (∼1.3 × 1019 cm−2, based on 0.019 g of TiO2, a
BET surface area of 71 m2/g, and a surface site density of 1 ×
1015 cm−2 for 5-fold-coordinated Ti4+ at the lowest energy
{101} surface20). At the same time, 50% of the NO is photo-
oxidized to NO2 at the illuminated TiO2 surface via reaction of
NO with photogenerated hydroxyl radicals. This undesired
reaction (the toxicity of NO2 exceeds that of NO) also occurs
in air, but under ambient conditions the process is very slow.
Clearly, illuminated TiO2 strongly catalyzes this reaction. After
∼200 min, the surface is fully saturated with nitrate groups and
no net change in the NOx concentration occurs anymore.
Reduced TiO2 shows a slightly higher activity for the photo-
oxidation of NO to NO2 than oxidized TiO2 under otherwise
identical conditions (Figure 2b). We attribute this to the
presence of oxygen vacancies that are formed during the
reduction treatment.21 Using standard Kroger−Vink notation,22
̈
the corresponding reaction can be written as
Figure 3. (a) Shift of the anatase (101) peak as a function of the Fe
concentration in Fe/TiO2 nanoparticles. The XRD pattern in the inset
shows that anatase (main peak) is the dominant phase, with only a
small amount of rutile (peak at 32°). (b) Position of the anatase Eg
Raman peak (inset) as a function of the Fe concentration for Fe/TiO2
(this study) and as a function of the oxygen stoichiometry for undoped
TiO2 (Parker et al.29).
TiO2
OO× XooooY V•O• + 2e +
O2(g)
1
′
(1)
2
In-plane vacancies such as these, or step or kink sites, are well-
known to be able to enhance the catalytic activity for certain
reactions by providing energetically favorable ad- or desorption
sites.23−26
A less pronounced but arguably more important observation
from the data in Figure 2b is that, after the surface is saturated
with NO3− groups, the sum of the NO and NO2 concentrations
is no longer equal to the initial NO concentration of 1040 ppb.
About 1% of the NOx (10 ppb) has disappeared and must have
shift was first observed for reduced (undoped) anatase TiO2 by
Parker et al.,29,30 who were able to relate the size of the shift to
the oxygen stoichiometry. Parker’s results are plotted in Figure
3b (dashed line) on the same x-axis scale, using the assumption
that one oxygen vacancy is formed for every two Fe ions (see
eq 2). The slopes of the data points and the dashed line are
identical. Together with the XRD data, this confirms that the
Fe dopants are indeed fully charge-compensated by oxygen
vacancies.27
−
been converted to another species. Since NO3 and NO2 are
the only stable oxidation products after NO conversion, we
attribute the 1% “lost” NOx to the formation of a reduction
product, such as N2O or N2.
Although the 1% lost NOx can be accurately and very
reproducibly measured, the amount is rather small. We
attribute this to the fact that reaction 1 is reversible, which
causes part of the reduced TiO2 to be reoxidized while exposed
to air. Before exploring further evidence for the photoreduction
reaction and possible mechanisms that cause it, we first need to
stabilize the oxygen vacancies in reduced TiO2. This can be
achieved by doping the TiO2 nanoparticles with Fe, a process
that we recently studied in detail.27 The Fe3+ substitutes for
Ti4+ ions in the lattice, and the effective negative charge of this
acceptor-type dopant is compensated by the positively charged
oxygen vacancies. The reaction for the dissolution of iron in
TiO2 can be written as follows:
The degradation of NO over Fe-doped TiO2 is shown in
Figure 4a. The initial stage of the reaction after the light is
turned on is similar to that observed for undoped TiO2: a fast
TiO2,773K
Fe2O3 ⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯→ 2FeTi + VO•• + 3OO×
′
(2)
Evidence for the formation of oxygen vacancies upon Fe
doping is provided by a combination of X-ray diffraction
(XRD) and Raman measurements (Figure 3). The XRD
patterns show that the d101 lattice spacing decreases linearly
with increasing Fe concentration (Figure 3a), which proves that
Fe ions are indeed incorporated as dopants in the TiO2 lattice.
The fact that the lattice spacing decreases, even though Fe3+ is
slightly larger than Ti4+,28 is due to the formation of oxygen
vacancies.27 Further evidence for the presence of oxygen
vacancies is given by Raman spectroscopy. As indicated in
Figure 3b, the Raman peak position of the anatase Eg mode
shows a linear shift with increasing Fe concentration. Such a
Figure 4. Photocatalytic degradation of NO over 1% Fe/TiO2 (a) in
air and (b) in pure N2.
9371
dx.doi.org/10.1021/ja302246b | J. Am. Chem. Soc. 2012, 134, 9369−9375

Journal of the American Chemical Society
Article
decrease of the NO concentration (adsorption at photo-
generated superoxide sites) followed by a slower increase of the
−
NO signal as the surface becomes saturated with NO3 groups
(deactivation). However, the amount of NOx that is
presumably “lost” to photoreduction has now increased to
3%. This is an increase of a factor of 3 compared to that of
undoped reduced TiO2. We attribute this large increase to the
higher concentration of oxygen vacancies, which are now
stabilized by the presence of charge-compensating Fe acceptors.
A second important observation in Figure 4a is that the
formation of NO2 is almost completely suppressed after 300
min. After steady-state conditions are reached, only 3% of the
NO is continuously converted to NO2. This is much less than
the 50−60% observed for undoped TiO2 (Figure 2).
Apparently, Fe doping strongly suppresses the activity for
NO2 formation, but also increases the selectivity toward
photoreduction from 0% to 50% (3% reduction + 3%
oxidation).
To further prove the ability of 1% Fe-doped TiO2 to
photoreduce NO, pure N2 (99.999%) was used instead of air as
the carrier gas for the NO pollutant. Figure 4b shows that, even
in the absence of oxygen, 4.5% of the NOx continuously
disappears after reaching steady-state conditions. This is again
consistent with what would be expected for an overall
photoreduction reaction.
Figure 5. Photocatalytic conversion of NO to N2 and O2 over 1% Fe-
doped TiO2. The sample was irradiated with UV light, and the target
pollutant was 100 ppm NO in He.
lower fractional conversion. A control experiment with
undoped TiO2, also shown in Figure 5, did not show any N2
or O2 evolution. This clearly shows that Fe doping of TiO2
changes its photocatalytic selectivity for NO degradation from
oxidation to reduction.
On the basis of all these observations, we propose the
following series of reactions to describe the various processes
that occur.
−
Photo-Oxidation of NO to NO2 and NO3 . The reaction
of photoinduced electrons and holes with surface-adsorbed
oxygen and hydroxyl groups results in the formation of
superoxide anions and hydroxyl radicals:11,12
It should be noted that, even in a N2 atmosphere, the NO
concentration briefly decreases during the first 50 s after the
light is turned on (Figure 4b). As outlined above, this is due to
the reaction of NO with adsorbed O2− species, resulting in the
formation of NO3 . The O2 species are formed while the
sample is exposed to air and ambient light prior to the
experiment and do not spontaneously desorb after replacement
of the air in the reactor chamber with N2. Further inspection of
Figure 4b shows that no NO2 is formed during the initial and
steady-state phases of the photoreaction. This indicates that the
selectivity for photoreduction is 100% for 1% Fe-doped TiO2 in
the absence of oxygen.
TiO2
hν ⎯⎯⎯→⎯ e− + h+
−
−
(3)
k1
−
Ti(O2)ads + e− → Ti(O2 )ads
(4)
Ti−OH− + h+ → Ti−OH•
(5)
The superoxide anion can directly oxidize nitric oxide to a
nitrate adsorbate:
−
Ti(O2 )ads + NO(g) → Ti(NO3−)ads
(6)
A final observation from Figure 4b is the immediate
desorption of a small amount of NO after the UV light is
turned off. This indicates the presence of a small amount of
unreacted, adsorbed NO at the TiO2 surface. From integration
of the desorption peak, the amount of desorbed NO
corresponds to ∼0.04% of the total number of TiO2 surface
sites.31 This is a much smaller fraction than the number of
adsorbed nitrate species mentioned before, which suggests that
At the same time, photo-oxidation of NO by hydroxyl radicals
leads to the formation of NO2:
Ti−OH• + NO → Ti−H + NO2(g)
(7)
While the Ti−H bond seems unusual, density functional theory
(DFT) calculations suggest that it can indeed exist at the (001),
(100), and (101) surfaces of anatase (the hydrogen binds to
undercoordinated Ti atoms, resulting in a Ti−H bond length of
∼1.75 Å and a slightly outward relaxation of the Ti atoms).33 At
the initial stage of the reaction (before the steady state is
reached), the net overall reaction can thus be summarized as
−
NO quickly reacts to NO3 after being adsorbed.
To support our assumption that the lost NO in Figure 4 is
photocatalytically reduced, gas chromatography measurements
were carried out to identify the chemical nature of the reaction
product(s). This indeed revealed the presence of N2 and O2,
while no other species (such as N2O) were detected. The
absence of N2O can be explained by the absence of lateral
interactions between adsorbed NO species at these low NO
concentrations.32 As shown in Figure 5, the concentrations of
N2 and O2 gradually increase until steady-state conditions are
reached after ∼80 min. Both gases evolve in a stoichiometric
ratio, which is consistent with the absence of N2O formation. It
should be noted that the sum of the N2 and O2 concentrations
is ∼1% of the initial NO concentration, i.e., 4.5 times less than
the fraction of NO converted in Figure 4b. We attribute this to
the 100-fold larger total concentration of NO used for the GC
experiment. Such a large concentration may saturate the total
number of available reaction sites at the surface, explaining the
Ti(O2)ads + Ti−OH− + hν + 2NO(g)
→ Ti(NO3−)ads + Ti−H + NO2(g)
(8)
−
The sum reaction 8 indicates that NO2 and NO3 are
simultaneously produced, consistent with the data shown in
Figure 2a.
Since the adsorbed nitrate groups do not spontaneously
desorb, reactions 4 and 6 will stop after a while and the catalyst
slowly deactivates. Since NO2 formation via photogenerated
holesreactions 5 and 7persists after deactivation, there
must be an alternative reaction path for the photogenerated
electrons under steady-state conditions. A likely pathway is the
following:12,34
9372
dx.doi.org/10.1021/ja302246b | J. Am. Chem. Soc. 2012, 134, 9369−9375

Journal of the American Chemical Society
Article
formation of N2 and O2 is due to tetrahedrally coordinated Ti4+
seems unlikely.
A more plausible explanation is that oxygen vacancies act as
catalytic centers that capture the oxygen end of the NO
k2
Ti−H + O2(g) + e− → Ti−OOH−
(9)
(10)
(11)
Ti−OOH− + Ti−H → Ti−OH− + Ti−OH•
Ti−OH• + e− → Ti−OH−
Under steady-state conditions, reactions 3, 5, 7, and 9−11 all
occur simultaneously and can be summed up to give
molecules:
••
V
+ 2e− + NO(g) → Osurf −N
O(surf)
(17)
Since the captured NO molecules have no net charge, mobile
oxygen vacancies39 are able to diffuse close to the captured
molecule and capture another NO molecule on a neighboring
site. Alternatively, the Osurf−N species themselves may be
sufficiently mobilefor example, via a surface vacancy diffusion
mechanismto meet each other. They can then react to form
N2:
TiO2
2NO(g) + O2(g) + 2hν ⎯⎯⎯→⎯ 2NO2(g)
(12)
Suppression of NO2 Formation. The presence of Fe3+
markedly changes the reaction mechanism for NO2 formation,
as illustrated by Figure 4a. During the initial stage of
illumination, NO2 is again formed via the mechanism described
by reaction 8. However, instead of regenerating Ti−H sites via
reactions 9−11 after deactivation, the photogenerated electrons
can now reduce Fe3+:
2Osurf −N → 2O + N2(g)
(18)
surf
Since N2 has a higher thermodynamic stability than NO
(ΔG0f (NO) = +87.6 kJ/mol), reaction 18 will be exothermic.
The released energy helps to release the trapped oxygen atoms
from the lattice sites, resulting in the formation of O2:
k3
Fe3+ + e− → Fe2+
(13)
Fe2+ is a well-known adsorption site for nitric oxide, leading to
the formation of mono- or dinitrosyl species:35,36
2O → 2VO× + O2(g)
(19)
surf
Fe2+ + nNO(g) → Fe2+(NO)n,ads
(14)
The photogenerated holes (h+) can then reoxidize the neutral
oxygen vacancies to their normal 2+ state:
where n equals 1 or 2, respectively. This species can be oxidized
again via adjacent hydroxyl radicals or by direct capture of
photogenerated holes:
VO× + 2h+ → V•O•
(20)
Summing up reactions 3 and 17−20 gives the overall
Fe2+(NO)n,ads + Ti−OH• → Fe3+(NO)n,ads + Ti−OH−
photoreduction reaction
(15a)
2V•O•
2NO(g) + 4hν ⎯⎯⎯→ N2(g) + O2(g)
Fe2+(NO)n,ads + h+ → Fe3+(NO)n,ads
(21)
(15b)
Repeating the experiment of Figure 4b for a longer time
showed no change in the fraction of NO that is lost to
photoreduction; a value of 4.5% was found even after 1050 min.
This corresponds to a turnover number of ∼2 nitric oxide
molecules per oxygen vacancy site, which confirms that this
defect acts as a catalytic center.
Before any further attempts can be made to improve the
photocatalytic activity of the material, it is important to identify
the rate-limiting factor of the overall photoreduction process.
Possible external factors include the illumination intensity and
the concentration of oxygen vacancies. The influence of the UV
light intensity on NO decomposition for 1% Fe-doped TiO2 is
shown in Figure 6. Clearly, the photon flux is not a limiting
factor in this range of light intensities.
The oxidation of Fe2+ to Fe3+ reduces the degree of π back-
bonding, which weakens the Fe3+−NO bond and results in
desorption:18
Fe3+(NO)n,ads → Fe3+ + nNO(g)
(16)
The sum of reactions 3, 5, 13, 14, 15a, and 16 represents a NO-
mediated recombination mechanism that explains why so little
NO2 is formed over Fe-doped TiO2 after steady-state
conditions are reached.
The formation of Fe2+(NO)n species via reaction 14 is
supported by the immediate release of NO to the gas phase
(reaction 16) when the UV light is turned off, as shown in
Figure 4b. Note that this observation also rules out reaction
15b, since no holes are available in the dark. Reaction 15a
therefore seems a more likely pathway for reoxidation of Fe2+.
Photoreduction. The formation of N2 and O2 (Figure 5)
can occur via at least two different routes. One possibility is that
a small amount of tetrahedrally coordinated Ti is formed in the
Fe-doped TiO2 samples. This species has been reported as the
active site for catalytic decomposition of NO into N2 and O2 at
Ti-modified zeolites by Anpo and co-workers.17,37 Since most
Ti ions at the TiO2 surface are 5-fold-coordinated, a single
oxygen vacancy created at or near the surface could indeed lead
to a 4-fold-coordinated Ti4+ center. However, the Ti−O bond
length would still be similar to the 1.93 Å found in bulk TiO2,
whereas the tetrahedrally coordinated TiO4 centers that are
reported to photoreduce NO to N2 have a significantly smaller
bond length (∼1.78 Å).38 Such a strong reduction in bond
length would require very high oxygen vacancy concentrations,
much higher than those present in our 1% Fe-doped TiO2.27
On the basis of these arguments, the possibility that the
The influence of oxygen vacancies on the photocatalytic
activity is investigated by repeating the experiment of Figure 4b
with half the concentration of oxygen vacancies (0.5% instead
of 1% Fe; all other conditions are the same). As shown in
Figure 6. Influence of the UV light intensity on the photocatalytic
activity for NO degradation of 1% Fe-doped TiO2 in pure N2.
9373
dx.doi.org/10.1021/ja302246b | J. Am. Chem. Soc. 2012, 134, 9369−9375

Journal of the American Chemical Society
Article
Figure 7, the NO conversion efficiency is 3%, significantly less
than the 4.5% in Figure 4b. This shows that the concentration
thank B. Boshuizen for designing a Labview program to read
the NOx analyzer. Financial support for this work is provided
by the Shell-TU Delft “Sustainable Mobility” program.
REFERENCES
■
(1) Devahasdin, S.; Fan, C.; Li, K.; Chen, D. H. J. Photochem.
Photobiol., A 2003, 156, 161−170.
(2) Roy, S.; Baiker, A. Chem. Rev. 2009, 109, 4054−4091.
(3) Wu, J. C. S.; Cheng, Y. J. Catal. 2006, 237, 393−404.
(4) Taylor, K. C. Catal. Rev.Sci. Eng. 1993, 35, 457−481.
(5) Rodriguez, J. A.; Jirsak, T.; Liu, G.; Hrbek, J.; Dvorak, J.; Maiti, A.
J. Am. Chem. Soc. 2001, 123, 9597−9605.
(6) Maggos, T.; Bartzis, J. G.; Liakou, M.; Gobin, C. J. Hazard. Mater.
2007, 146, 668−673.
(7) Negishi, N.; Takeuchi, K.; Ibusuki, T. J. Mater. Sci. 1998, 33,
5789−5794.
Figure 7. Photocatalytic decomposition of NO for 0.5% Fe/TiO2 in
pure N2.
(8) Lin, Y. M.; Tseng, Y. H.; Huang, J. H.; Chao, C. C.; Chen., C. C.;
Wang, I. Environ. Sci. Technol. 2006, 40, 1616−1621.
(9) Anpo, M. Pure Appl. Chem. 2000, 72, 1265−1270.
(10) Wang, J.; Tafen, D. N.; Lewis, J. P.; Hong, Z.; Manivannan, A.;
Zhi, M.; Li, M.; Wu, N. J. Am. Chem. Soc. 2009, 131, 12290−12297.
(11) Du, Y.; Rabani, J. J. Phys. Chem. B 2003, 107, 11970−11978.
(12) Hoffmann, M. R.; Martin, S. T.; Choi, W.; Bahnemann, D. W.
Chem. Rev. 1995, 95, 69−96.
of oxygen vacancies indeed limits the photocatalytic activity for
NO reduction. Further improvements of this system may
therefore be possible by increasing the concentration of Fe. We
recently showed that Fe dopant concentrations of up to 10% in
TiO2 nanoparticles are possible before segregation of iron oxide
occurs.27
(13) Dalton, J. S.; Janes, P. A.; Jones, N. G.; Nicholson, J. A.; Hallam,
K. R.; Allen, G. C. Environ. Pollut. 2002, 120, 415−422.
(14) Wang, H.; Wu, Z.; Zhao, W.; Guan, B. Chemosphere 2007, 66,
185−190.
CONCLUSIONS
■
We have found that oxygen vacancies in nanosized TiO2 serve
as active centers for the photocatalytic reduction of nitric oxide
into N2 and O2. By doping the material with Fe, the mechanism
of the reaction can be influenced in two distinct ways: (i) Fe3+
is an acceptor-type dopant that stabilizes the oxygen vacancies
through charge compensation, thereby increasing the activity of
the photoreduction reaction, and (ii) Fe3+ can be photoreduced
to Fe2+, providing a recombination pathway that suppresses the
formation of NO2 and thus enhances the selectivity of the
reaction for N2 formation. While the conversion efficiency is
still modest, the Fe/TiO2 photocatalyst does not show any
signs of deactivation. In contrast to the standard DeNOx
catalysts based on TiO2, the conversion is not blocked by
nitrate species that have to be washed away periodically. The
material is also easier and cheaper to synthesize than NO
photoreduction catalysts based on modified zeolites.
(15) Anpo, M.; Zhang, S. G.; Mishima, H.; Matsuoka, M.; Yamashita,
H. Catal. Today 1997, 39, 159−168.
(16) Anpo, M.; Takeuchi, M.; Ikeue, K.; Dohshi, S. Curr. Opin. Solid
State Mater. Sci. 2002, 6, 381−388.
(17) Yamashita, H.; Ichihashi, Y.; Zhang, S. G.; Matsumura, Y.;
Souma, Y.; Tatsumi, T.; Anpo, M. Appl. Surf. Sci. 1997, 121, 305−309.
(18) Wu, Q.; Mul, G.; Van de Krol, R. Energy Environ. Sci. 2011, 4,
2140−2144.
(19) Komazaki, Y.; Shimizu, H.; Tananka, S. Atmos. Environ. 1999,
33, 4363−4371.
(20) Henningsson, A.; Andersson, M. P.; Uvdal, P.; Siegbahn, H.;
Sandell, A. Chem. Phys. Lett. 2002, 360, 85−90.
(21) Liu, H.; Ma, H. T.; Li, X. Z.; Li, W. Z.; Wu, M.; Bao, X. H.
Chemosphere 2003, 50, 39−46.
(22) Chiang, Y. M.; Birnie, D. P.; Kingery, W. D. Physical Ceramics;
Wiley: New York, 1997.
Further improvement of the photocatalytic activity seems
simply a matter of increasing the dopant concentration. For Fe,
dopant concentrations of up to 10% have been reported, which
leaves ample room for further optimization. Alternatively, other
acceptor-type dopants can be explored. On the basis of the
proposed reaction mechanisms, it is important to choose
dopants that can be reduced to the 2+ oxidation state to avoid
the oxidation of NO to NO2. Cr, Co, and Ni are therefore more
suitable choices than, e.g., Al or Ga. Further explorations along
these lines may lead to a new generation of highly selective
photocatalysts.
(23) Gong, X. Q.; Selloni, A.; Batzill, M.; Diebold, U. Nat. Mater.
2006, 5, 665−670.
(24) Beck, T. J.; Klust, A.; Batzill, M.; Diebold, U.; Valentin, C. D.;
Selloni, A. Phys. Rev. Lett. 2004, 93, 0361041−0361044.
(25) Thompson, T. L.; Yates, J. T. Chem. Rev. 2006, 106, 4428−
4453.
(26) Henderson, M. A.; Epling, W. S.; Perkins, C. L.; Peden, C. H. F.
J. Phys. Chem. B 1999, 103, 5328−5337.
(27) Wu, Q.; Zheng, Q.; Van de Krol, R. J. Phys. Chem. C 2012, 116,
7219−7226.
(28) Shannon, R. D. Acta Crystallogr., A 1976, 32, 751−767.
(29) Parker, J. C.; Siegel, R. W. Appl. Phys. Lett. 1990, 57, 943−945.
(30) Parker, J. C.; Siegel, R. W. J. Mater. Res. 1990, 5, 1246−1252.
(31) The number of desorbed NO molecules was determined by
integrating the NOx curve from t = 469 min to t = 471.5 min in Figure
4b. The final concentration of 1040 ppb was used as a baseline. This
corresponds to 5.2 × 1015 NO molecules. On the basis of the mass of
the sample (0.019 g) and a BET surface area of 71 m2/g, this
corresponds to ∼0.04% of the total amount of the sample’s surface
sites.
(32) Diebold, U. Surf. Sci. Rep. 2003, 48, 53−229.
(33) Barnard, A. S.; Zapol, P. Phys. Rev. B 2004, 70, 235403.
(34) Zhang, Z.; Wang, C.; Zakaria, R.; Ying, J. Y. J. Phys. Chem. B
1998, 102, 10871−10878.
AUTHOR INFORMATION
Corresponding Author
r.vandekrol@tudelft.nl
■
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
We gratefully acknowledge Prof. D. Bahnemann (Leibniz
■
Universitat Hannover, Germany) for advice and J. Middelkoop
̈
for practical help with setting up the NOx analysis system. We
9374
dx.doi.org/10.1021/ja302246b | J. Am. Chem. Soc. 2012, 134, 9369−9375

Journal of the American Chemical Society
Article
́
(35) Mul, G.; Perez-Ramírez, J.; Kapteijn, F.; Moulijn, J. A. Catal.
Lett. 2002, 3, 129−138.
(36) King, D. L.; Peri, J. B. J. Catal. 1983, 79, 164−175.
(37) Hu, Y.; Martra, G.; Zhang, J.; Higashimoto, S.; Coluccia, S.;
Anpo, M. J. Phys. Chem. B 2006, 110, 1680−1685.
(38) Yamashita, H.; Ichihashi, Y.; Anpo, M.; Hashimoto, M.; Louis,
C.; Che, M. J. Phys. Chem. 1996, 100, 16041−16044.
(39) Schaub, R.; Wahlstron, E.; Rønnau, A.; Lægsgaard, E.;
̈
Stensgaard, I.; Besenbacher, F. Science 2003, 299, 377−379.
9375
dx.doi.org/10.1021/ja302246b | J. Am. Chem. Soc. 2012, 134, 9369−9375