A molecular mechanism for the chemoselective hydrogenation of substituted nitroaromatics with nanoparticles of gold on TiO2 catalysts: A cooperative effect between gold and the support

Article abstract of DOI:10.1021/ja076721g

Nanoparticles of gold on TiO₂ are highly chemoselective for the reduction of substituted nitroaromatics, such as nitrostyrene. By combining kinetics and in situ IR spectroscopy, it has been found that there is a preferential adsorption of the r

Full text of DOI:10.1021/ja076721g

Published on Web 12/01/2007
A Molecular Mechanism for the Chemoselective
Hydrogenation of Substituted Nitroaromatics with
Nanoparticles of Gold on TiO2 Catalysts: A Cooperative
Effect between Gold and the Support
Merce` Boronat,† Patricia Concepcio´n,† Avelino Corma,*,† Silvia Gonza´lez,‡
Francesc Illas,‡ and Pedro Serna†
Instituto de Tecnolog´ıa Qu´ımica, UniVersidad Polite´cnica de Valencia-CSIC, AV. de los
Naranjos s/n, 46022 Valencia, Spain, and Departament de Qu´ımica F´ısica and Institut de
Qu´ımica Teo`rica i Computacional, (IQTCUB) UniVersitat de Barcelona and Parc Cient´ıfic de
Barcelona, C/Mart´ı i Franque`s 1, E-08b028 Barcelona, Spain
Received September 14, 2007; E-mail: acorma@itq.upv.es
Abstract: Nanoparticles of gold on TiO2 are highly chemoselective for the reduction of substituted
nitroaromatics, such as nitrostyrene. By combining kinetics and in situ IR spectroscopy, it has been found
that there is a preferential adsorption of the reactant on the catalyst through the nitro group. IR studies of
nitrobenzene, styrene, and nitrostyrene adsorption, together with quantum chemical calculations, show
that the nitro and the olefinic groups adsorb weakly on the Au(111) and Au(001) surfaces, and that although
a stronger adsorption occurs on low-coordinated atoms in gold nanoparticles, this adsorption is not selective.
On the other hand, an energetically and geometrically favored adsorption through the nitro group occurs
on the TiO2 support and in the interface between the gold nanoparticle and the TiO2 support. Such
preferential adsorption is not observed with nanoparticles of gold on silica which, contrary to the Au/TiO2
catalyst, is not chemoselective for the reduction of substituted nitroaromatic compounds. Therefore, the
high chemoselectiviy of the Au/TiO2 catalyst can be attributed to a cooperation between the gold nanoparticle
and the support that preferentially activates the nitro group.
Introduction
the selective hydrogenation of the carbonyl group. However,
selectivities are high only at low conversion levels, and gold
Supported gold catalysts have shown interesting possibilities
for selective oxidations,1 carbon-carbon bond formation, and
reactions with alkynes and alkenes among others.2-4 A more
modest role has been played by supported gold catalysts for
selective hydrogenations. Only in the case of R,â-unsaturated
aldehydes such as crotonaldehyde, acroleine, citral, and pent-
3-en-2-one5 is Au supported on ZrO2 and ZnO able to catalyze
still cannot compete with supported Pt doped with Sn, or with
Pt on CeO2, for the selective hydrogenation of carbonyls in the
presence of CdC double bonds, especially at high levels of
conversion.6,7
We have recently shown that Au/TiO2 has unique catalytic
behavior for the chemoselective hydrogenation of the nitro group
in substituted nitroaromatics.8 Interestingly, none of the sup-
ported Pt catalysts reported up to now could achieve the degree
of selectivity of gold, unless other catalytic functions were
present in the homogeneous phase.9 While the unique behavior
of gold was clearly demonstrated,10 why this occurs and which
† Universidad Politecnica de Valencia-CSIC.
‡ Universitat de Barcelona and Parc Cient´ıfic de Barcelona.
(1) (a) Haruta, M.; Yamada, N.; Kobayashi, Τ.; Iijima, S. J. Catal. 1989, 115,
301-309. (b) Landon, P.; Collier, P. J.; Papworth, A. J.; Kiely, C. J.;
Hutchings, G. J. Chem. Commun. 2002, 2058-2059. (c) Abad, A.;
Concepcion, P.; Corma, A.; Garcia, H. Angew. Chem., Int. Ed. 2005, 44,
4066-4069. (d) Αbad, Α.; Αlmeda, C.; Corma, A.; Garcia, H. Chem.
Commun. 2006, 3178-3180. (e) Ηayashi, Τ.; Τanaka, Κ.; Ηaruta, M. J.
Catal. 1998, 178, 566-575. (f) Davis, R. J. Science 2003, 301, 926-927.
(g) Hughes, M. D.; et al. Nature 2005, 437, 1132-1135. (h) Prati, L.; Rossi,
M. Stud. Surf. Sci. Catal. 1997, 110, 509-516. (i) Coluccia, S.; Martra,
G.; Porta, F.; Prati, L.; Rossi, M. Catal. Today 2000, 61, 165-172. (j)
Enache, D. I.; Edwards, J. K.; Landon, P.; Solsona-Espriu, B.; Carley, A.
F.; Herzing, A. A.; Watanabe, M.; Kiely, C. J.; Knight, D. W.; Hutchings,
G. J. Science 2006, 311, 362-365. (k) Corma, A.; Domine, M. Chem.
Commun. 2005, 4042-4044.
(5) (a) Bailie, J. E.; Hutchings, G. J. Chem. Commun. 1999, 2151-2152. (b)
Radnick, J.; Mohr, C.; Claus, P. Phys. Chem. Chem. Phys. 2003, 5, 172-
177. (c) Μilone, G.; Tropeano, M. L.; Gulino, G.; Neri, G.; Ingoglia, Q.;
Galvano, S. Chem. Commun. 2002, 868-869. (d) Fordham, P.; Βesson,
M.; Gallezot, P. Stud. Surf. Sci. Catal. 1997, 108, 429-436.
(6) (a) Concepcio´n, P.; Corma, A.; Silvestre-Albero, J.; Franco, V.; Chane-
Ching, J. Y. J. Am. Chem. Soc. 2004, 126, 5523-5532. (b) Sepu´lveda-
Escribano, A.; Coloma, F.; Rodriguez-Reinoso, F. J. Catal. 1998, 178, 649-
657. (c) Abid, M.; Touroude, R. Catal. Lett. 2000, 69, 139-144. (d)
Silvestre-Albero, J.; Rodriguez-Reinoso, F.; Sepu´lveda-Escribano, A. J.
Catal. 2002, 210, 127-136.
(2) (a) Carrettin, S.; Guzman, J.; Corma, A. Angew. Chem., Int. Ed. 2005, 44,
2242-2245. (b) Gonza´lez-Arellano, C.; Corma, A.; Iglesias, M.; Sa´nchez,
F. J. Catal. 2006, 238, 497-501. (c) Gonza´lez-Arellano, C.; Abad, A.;
Corma, A.; Garc´ıa, H.; Iglesias, M.; Sa´nchez, F. Angew. Chem., Int. Ed.
2007, 46, 1536-1538.
(7) (a) Coloma, F.; Sepu´lveda-Escribano, A.; Fierro, J. L. G.; Rodriguez-
Reinoso, F. Appl. Catal. A 1996, 148, 63-80. (b) Margitfalvi, J. L.; Vanko´,
Gy.; Borba´th, I.; Tompos, A.; Ve´rtes, A. J. Catal. 2000, 190, 474-477.
(8) Corma, A.; Serna, P. Science 2006, 313, 332-333.
(3) Ηashmi, A. S. K.; Hutchings, G. Angew. Chem., Int. Ed. 2006, 45, 7896-
7936.
(4) Ηashmi, A. S. K. Chem. ReV. 2007, 107, 3180-3211.
(9) Siegrist, U.; Baumeister, P.; Blaser, H.-U. Chem. Ind. 1998, 75, 207-219.
9
16230
J. AM. CHEM. SOC. 2007, 129, 16230-16237
10.1021/ja076721g CCC: $37.00 © 2007 American Chemical Society

Chemoselective Hydrogenation of Nitroaromatics on Au/TiO2
A R T I C L E S
active sites are involved in the chemoselective reduction of
substituted nitroaromatics to the corresponding anilines remain
unknown.
In the present work, we have studied, at the molecular level,
the interactions between the reactants and the catalyst surface
by combining in situ IR spectroscopy, realistic quantum
chemical modeling, and kinetic experiments. It will be shown
that, during nitrostyrene hydrogenation with Au/TiO2, there is
a cooperative effect between gold and TiO2 such that H2 is
dissociated on Au and nitrostyrene is adsorbed selectively on
the support through the nitro group only, especially on the gold
atoms located at the boundary between TiO2 and Au, these very
specific adsorption sites being responsible for the high chemose-
lectivity observed.
Figure 1. Evolution of yield versus conversion during the 3-nitrostyrene
hydrogenation using 2 wt % Pt/TiO2 (O), 1.5 wt % Au/TiO2 (9), and 2 wt
% Pd/TiO2 (4).
edges. In the geometry optimizations, the two innermost atomic layers
were kept as in the bulk to provide an adequate metallic environment
to the topmost layers, which are fully relaxed. As representative of
nanoscale Au particles, a Au38 cluster (∼1 nm diameter) was placed in
a 20 Å × 20 Å × 20 Å cubic box, and the position of all atoms was
always fully relaxed. The same cubic box was used to calculate the
energy, geometry, and frequencies of the isolated nitrostyrene molecule.
Experimental Section
Catalyst and Catalytic Experiments. The Au/TiO2 and the Au/
Fe2O3 catalysts, with 1.5 wt % and 4.5 wt % loadings of gold,
respectively, were supplied by the World Gold Council (reference
catalysts, Type A and D). Other gold catalysts, i.e., 1.5 wt % Au/C
and 1.6 wt % Au/SiO2, were prepared following recently described
procedures,11 the average gold particle size being between 3.5 and 4.0
nm in all cases. Finally, a 2 wt % Pt/TiO2 catalyst was prepared by the
incipient wetness technique using H2PtCl6 as platinum precursor. This
catalyst was reduced under H2 flow during 5 h at 200 °C before its use
in the reaction.
Catalytic experiments were carried out in 2 mL glass reactors,
through the following procedure: First, an appropriate amount of the
catalyst is placed into the reactor, together with the reaction mixture
(1 mL). The system is then purged with H2, before heating, in order to
completely remove oxygen from the reactor. Under H2 atmospheric
pressure, the reaction mixture is heated at 120 °C with magnetic stirring,
and then the pressure is set to 9 bar of H2, this being the starting reaction
time. Aliquots were taken from the reactor at different times and
analyzed by GC and MS-GC measurements.
The TiO2 (anatase) support was modeled by a (2 × 2) supercell slab
model representing the most stable (001) surface. It contains three TiO2
layers (nine atomic layers), with the two uppermost TiO2 layers being
also fully relaxed. Finally, Au particles supported on anatase were
modeled by a supercell approach in which the anatase slab model unit
cell described above was sufficiently enlarged so as to support a Au13
particle in such a way that Au particles are kept ∼8 Å apart from each
other and there is a vacuum region larger than 15 Å between vertical
repeated slabs. In order to stabilize the Au13 particle on top of the TiO2
support, an oxygen vacancy was artificially introduced in the anatase
model just below the metallic particle. The largest supercell contains
172 atoms, and calculations require access to a large parallel super-
computer.
All calculations were carried out using the VASP code,12 with
exchange correlation effects being described by the Perdew-Wang
(PW91)13 version of the generalized gradient approximation (GGA).
The density was expanded in a plane wave basis set with a 415 eV
cutoff for the kinetic energy to represent the valence electron density;
a larger cutoff of 515 eV was used in some cases to test the convergence
of adsorption energies and geometries with respect to this parameter.
The effect of the core electrons in the valence density was taken into
account by means of the projected augmented wave (PAW) method.14
The Brillouin zone of the gold surface unit cells was described with a
3 × 3 × 1 k-points grid within the Monkhorst-Pack scheme, while the
rest of the calculations (Au38 nanoparticle, TiO2, and Au13/TiO2) were
carried out for the Γ k-point. The resulting structures were further
characterized as minima by a pertinent frequency analysis calculation.
These vibrational frequencies were calculated by diagonalizing the block
Hessian matrix corresponding to displacements of the coordinates of
all the atoms of nitrostyrene molecule.
IR Experiments. FTIR spectra were collected with an FTS-40A
spectrometer using a quartz cell connected to a vacuum line. The
samples were evacuated at 10-2 mbar for 20 min prior to the adsorption
experiment. Nitrobenzene, styrene, and nitrostyrene were adsorbed at
increasing pressures from 0.1 to 20 mbar.
Theoretical Calculations. Density functional calculations were
carried out with a series of models of increasing complexity, the final
system being sufficiently close to the experimental situation, although
still modeling ultra-high-vacuum conditions. Hence, adsorption of
nitrostyrene on different Au modelssnamely, single-crystal Au(111)
and Au(001) surfaces, a monatomic row model which contains low-
coordinated gold atoms, an isolated Au38 nanoparticle, and a Au13
nanoparticle supported on TiO2shas been explicitly considered.
Adsorption on the titanium oxide support has also been considered.
The Au(111) and Au(100) were modeled by periodic slabs containing
four atomic layers and a vacuum region larger than 20 Å between
vertical repeated slabs. Large enough supercellss(4 × 4) for Au(111)
and (3 × 3) for Au(001)swere used to avoid interaction between the
adsorbates. The monatomic row model was constructed from a (5 × 5)
supercell slab model for the Au(111) model with five atomic layers
and with two rows and one row missing in the topmost and second
Results and Discussion
Catalytic Experiments. Kinetic results in Figure 1 clearly
show that, with Au/TiO2, it is possible to achieve very high
chemoselectivity toward the hydrogenation of the nitro group
in nitrostyrene (eq 1) when working at levels of conversion close
to 100%.
layers, respectively. Although the Au monatomic row model (AuMAR
)
may appear somewhat artificial, it provides an adequate model for low-
coordinated Au atoms appearing in stepped surfaces and nanoparticle
(10) (a) Corma, A.; Concepcio´n, P.; Serna, P. Angew. Chem., Int. Ed. 2007, 46,
7266-7269. (b) Corma, Α.; Serna, P.; Garc´ıa, H. J. Am. Chem. Soc. 2007,
119, 6358-6359.
(11) (a) Comotti, M.; Della Pina, C.; Matarrese, R.; Rossi, M. Angew. Chem.,
Int. Ed. 2004, 43, 5812-5815. (b) Budroni, G.; Corma, A. Angew. Chem.,
Int. Ed. 2007, 45, 3328-3331.
(12) (a) Kresse, G.; Furthmu¨ller, J. Phys. ReV. B 1996, 54, 11169-11186. (b)
Kresse, G.; Hafner, J. Phys. ReV. B 1993, 47, 558-561.
(13) (a) Perdew, J. P.; Chevary, J. A.; Vosko, S. H.; Jackson, K. A.; Pederson,
M. R.; Singh, D. J.; Fiolhais, C. Phys. ReV. B 1992, 46, 6671-6687. (b)
Perdew, J. P.; Wang, Y. Phys. ReV. B 1992, 45, 13244-13249.
(14) Blo¨chl, P. E. Phys. ReV. B 1994, 50, 17953-17979.
9
J. AM. CHEM. SOC. VOL. 129, NO. 51, 2007 16231

A R T I C L E S
Boronat et al.
Table 1. Turnover Frequency of Au/TiO2, Au/SiO2, and Pt/TiO2
Catalysts for the Hydrogenation of Styrene and Nitrobenzenea
TOF (mol converted h-1 mol-1 metal)
substrate
styrene
nitrobenzene
1.5% Au/TiO2
1.6% Au/SiO2
2% Pt/TiO2
160
110
660
365
20
200
To explain this result, one can assume, to a first approximation,
that the intrinsic rate for the reduction of the nitro group on the
gold catalyst is much larger than that of the olefinic group. To
check this hypothesis, hydrogenations of nitrobenzene and
styrene were carried out separately. Turnover frequency (TOF)
values were calculated from initial reaction rates as the number
of molecules transformed per hour and per gold atom. The
results presented in Table 1 indicate that the TOF for hydro-
genating nitrobenzene is 2.2 times larger than that for styrene
with the Au/TiO2 catalyst. In contrast, on both Au/SiO2 and
Pt/TiO2 catalysts, styrene is hydrogenated more rapidly than
nitrobenzene. While these results are consistent with a higher
chemoselectivity of the Au/TiO2 system to hydrogenate a
substrate such as 3-nitrostyrene, they cannot explain the 98%
chemoselectivity for the hydrogenation of the nitro group on
Au/TiO2. To explain the very high selectivity of gold, a second
kinetic factor that may influence selectivity was also consid-
ered: that is, the preferential adsorption of the nitro versus the
olefinic group when there is competitive adsorption. This
hypothesis was tested by studying the reactivity of styrene in
the presence of nitrobenzene.
a TiO2 is shown to play an important role for activating the nitro group
in the gold catalysts. In contrast, the platinum sample is shown to be
especially active for catalyzing the CdC hydrogenation. Reaction condi-
tions: 120 °C, 9 bar (gold catalysts); 40 °C, 3 bar (platinum catalyst).
Feeding (1 mL): 8.5 mol % of substrate, 90.5 mol % of toluene (solvent),
and 1 mol % of o-xylene (internal standard).
Table 2. Turnover Frequency of the Au/TiO2 Catalyst for the
Hydrogenation of Styrene and Nitrobenzene at Different Feeding
Compositionsa
feeding (mol %)
nitrobenzene
TOF (mol converted h-1 mol-1 Au)
styrene
styrene
nitrobenzene
0
4.25
8.5
8.5
4.25
0
0
16
160
365
405
0
a While the double-bond reduction rate notably decreases in the presence
of a nitro group, this one keeps a very similar reaction rate, independent of
the styrene concentration. Reaction conditions: 120 °C, 9 bar. Feeding (1
mL): 8.5 mol % of substrate (styrene + nitrobenzene), 90.5 mol % of
toluene (solvent), and 1 mol % of o-xylene (internal standard).
Results in Table 2 indicate that, although styrene hydrogena-
tion can occur on Au/TiO2, its reaction rate is strongly inhibited
by the presence of nitrobenzene, while the hydrogenation of
nitrobenzene is practically not influenced by the presence of
styrene. Therefore, from the kinetic experiments, one can
conclude that the high chemoselectivity observed with Au/TiO2
not only is due to a higher intrinsic activity for hydrogenating
the nitro with respect to the olefinic group but also, especially,
is a consequence of the preferential adsorption of nitrostyrene
through the nitro group on the Au/TiO2 catalyst. This conclusion
is further supported by in situ IR competitive adsorption results.
Thus, when nitrobenzene is adsorbed on the TiO2 support or
on the Au/TiO2 catalyst (Figure 2a), the asymmetric νas(NO2)
IR vibration frequency shifts from 1552 cm-1 in the gas phase
to 1526 cm-1, while no shift for the aromatic ring vibration
frequencies (1620, 1606, 1585 cm-1) is observed. This suggests
that nitrobenzene interacts with the catalyst through the nitro
group. In this case, the support plays an important role, since
FTIR measurements of nitrobenzene adsorption on Au/SiO2
(Figure 2b), in which only the interaction between nitrobenzene
and the gold nanoparticles can be observed, indicate that the
nitro group adsorbs weakly on metallic gold.
Figure 2. FTIR spectra of nitrobenzene adsorption at 25 °C on (a) Au/
TiO2 and (b) Au/SiO2.
This is further confirmed by studying the competitive adsorption
of nitrobenzene and styrene with a 1:1 molar ratio (Figure 3c).
In this case, bands associated with nitrobenzene are exclusively
observed, indicating a preferential adsorption of the nitro group
from nitrobenzene versus the double bond in the styrene. Finally,
FTIR nitrostyrene adsorption experiments on TiO2 and Au/TiO2
(Figure 4) show that nitrostyrene adsorbs selectively through
the nitro group.
From the FTIR adsorption studies, it can be concluded that
(i) both the nitro and the olefinic groups can adsorb on Au/
TiO2, though the interaction through the nitro group is stronger
and competes very favorably with the adsorption of the olefinic
group, and (ii) the nitro group adsorbs weakly on metallic gold
and more strongly on the TiO2 support.
Styrene adsorption on both TiO2 and Au/TiO2 samples leads
to similar FTIR spectra (Figure 3b). The IR bands at 1601, 1494,
and 1450 cm-1 are related to the stretching vibrations of the
phenyl ring, and the bands at 1630 and 1416 cm-1 are attributed
to the ν(CdC) and δ(dCsH) vibrations of the vinyl group,
respectively. Interestingly, only a very small shift (<5 cm-1
)
of the IR bands associated with the olefinic groups occurs with
respect to the molecule in the gas phase, indicating an even
weaker interaction of the double bond with the surface. In other
words, when comparing the IR spectroscopy results for the
adsorption of nitrobenzene and styrene, it appears quite clear
that a stronger adsorption/interaction of the nitro group with
respect to the olefinic group on either TiO2 or Au/TiO2 occurs.
9
16232 J. AM. CHEM. SOC. VOL. 129, NO. 51, 2007

Chemoselective Hydrogenation of Nitroaromatics on Au/TiO2
A R T I C L E S
model or with the Au38 nanoparticle. For the AuMAR, it is found
that both the nitro group and the CdC double bond interact
moderately with low-coordinated Au atoms situated in the edge
of the monatomic row, the calculated adsorption energies being
8.7 and 12.2 kcal mol-1, respectively. The adsorption through
the CdC double bond is of π nature, with a concomitant
increase of 0.05 Å in the CC bond length. Analysis of the final
geometry shows that the main interaction occurs through the
CdC double bond, but it also reveals that one of the C atoms
of the benzene ring is quite strongly interacting with a Au atom
in the monatomic row edge (Figure 6). This coordination
provokes a coupling of the different vibrational modes in the
1650-1500 cm-1 region, all of them associated with the nitro
group and the carbon skeleton of the molecule. As a conse-
quence, the characteristic spectroscopic band at 1638 cm-1
,
corresponding to the pure CdC stretching mode, shifts to
∼1510 cm-1 (Table 4) and includes important contributions
from other modes, while the vibrational modes including
large contributions from the nitro group appear at 1574 and
1545 cm-1. Adsorption through the nitro group in the same
AuMAR model is only 3.5 kcal mol-1 less favorable but
significantly activates the NO2 group, since the optimized NO
distances are now 0.04 Å larger than the corresponding values
for the gas-phase molecule (Table 3). This geometry distortion
is large enough to provoke a change in the asymmetric νas(NO2)
vibration frequency from 1537 cm-1 in the gas phase to
∼1470 cm-1 (Table 4) and with significant mixing with the
vibrational modes of the aromatic ring, while other modes,
Figure 3. FTIR spectra of adsorption of nitrobenzene (a), nitrostyrene (b),
and a mixture of nitrobenzene and styrene with a 1:1 molar ratio (c) at
25 °C on TiO2 or Au/TiO2.
clearly dominated by NO contributions, appear at ∼1300 cm-1
.
Nitrostyrene’s interaction with the Au38 nanoparticle model is
stronger, as reflected by the higher adsorption energies obtained
and the larger increase in the N-O and C-C bond lengths
(Table 3). Again, the interaction through the CdC double bond
is more favorable than the interaction through the nitro group,
but both groups appear to be highly activated. In the case of
the CdC interaction, the adsorption is of di-σ nature, each of
the carbon atoms of the double bond interacts with one Au atom,
and the C-C distance increases to 1.47 Å, which is close to
the standard value for a C-C single bond. It is important to
remark that, as shown in Figure 7, there is also a noticeable
distortion of the Au38 nanoparticle due to the interaction with
some of the C atoms of the benzene ring. As could be expected,
the characteristic vibration frequency associated with the CdC
stretching mode does not show up simply because, as mentioned
above, the carbon-carbon bond changes its character from
double- to single-bond. Moreover, the NO2 group’s asymmetric
vibration couples with some of the modes associated with the
aromatic ring, as in the case of interaction with the AuMAR
model, but here two modes dominated by the nitro group’s
Figure 4. FTIR spectra of nitrostyrene adsorption at 25 °C on (a) Au/
TiO2 and (b) TiO2.
At this point, and to go one step further toward understanding
the molecular interactions at the surface, quantum chemical
calculations were performed to model the adsorption of the
reactant on different types of gold surfaces, on isolated gold
nanoparticles, on the TiO2 support, and on Ti-O-Au interface
sites as described in the previous section. Analysis of the
interaction energies, geometry deformation, and vibrational
frequencies should allow us to reach definitive conclusions about
the nature of the catalyst active sites and the modes of reactant
adsorption.
atomic displacements, which appear at 1562 and 1509 cm-1
,
can be clearly identified. In contrast, interaction of nitrostyrene
with the Au38 nanoparticle through the NO2 group results in a
significant increase of one of the NO distances to 1.30 Å. Here,
the NO2-related modes, appearing at 1494 and 1439 cm-1, are
even more strongly coupled to those of the aromatic ring, the
ones clearly dominated by NO motions appearing now at 1395
DFT Calculations. 1. Nitrostyrene Adsorption on Gold.
The present density functional calculations predict that nitrosty-
rene interacts weakly with Au(111) and Au(001) surfaces, in
agreement with experiment (Au/SiO2). The calculated adsorption
energies are small (Table 3), the adsorbed molecule lies parallel
to the Au surface at a distance of ∼3-4 Å (Figure 5), and the
optimized NsO and CdC distances are almost identical to those
obtained for the isolated molecule in the gas phase. A quite
different situation is obtained for the interaction of the nitrosty-
rene molecule with the low-coordinated Au atoms of the AuMAR
and 1371 cm-1
.
From the results corresponding to the interaction of nitrosty-
rene with low-coordinated Au atoms, either in the extended
AuMAR or in the Au38 particle, it can be concluded that both the
9
J. AM. CHEM. SOC. VOL. 129, NO. 51, 2007 16233

A R T I C L E S
Boronat et al.
Figure 5. Nitrostyrene adsorption on perfect Au(111) (left) and Au(001) (right) surfaces.
Table 3. Adsorption Energies (kcal/mol) and Optimized Values of the Most Important Distances (Å) of Nitrostyrene Adsorbed on the
Different Systems Considered
Eads
rNO
rCC
rOAu
rCAu
rOTi
isolated
Au(111)
Au(001)
AuMAR (NO2)
AuMAR (CdC)
Au38 (NO2)
Au38 (CdC)
TiO2 (strongly)
TiO2 (weakly)
Au13/TiO2
1.24, 1.24
1.24, 1.24
1.24, 1.24
1.28, 1.28
1.24, 1.24
1.30, 1.26
1.25, 1.24
1.27, 1.24
1.24, 1.24
1.39, 1.39
1.34
1.34
1.35
1.34
1.39
1.34
1.47
1.34
1.34
1.34
-2.5
-2.6
4.09, 4.14
3.36, 4.04
2.32, 2.32
3.61, 3.75
3.07, 3.52
-8.7
-12.2
-20.0
-28.9
-42.4
-24.0
-15.4
2.25, 2.41
2.15, 2.29
2.31, 2.53
2.14, 2.39
3.71, 4.01
2.06, 2.14
2.32, 2.33
nitro group and the CdC double bond are activated, but without
any preferential activation toward one or the other. Conse-
quently, the preferential adsorption of the nitro group on the
gold particle surface cannot explain the chemoselective hydro-
genation and the kinetic results presented above. Indeed, when
nitrostyrene hydrogenation is carried out on gold supported on
an “inert” carrier such as SiO2, the results presented in Table 5
clearly show that, while active, this catalyst is not selective for
the hydrogenation of the nitro group, in agreement with the
prediction from the present theoretical study. Taking into
account all results reported up to now, we have to consider the
possibility that the support, either by itself or through its
interaction with the gold nanoparticles, could be responsible
for the selective adsorption and reactivity of the nitro groups.
Therefore, we have extended the theoretical adsorption studies
to supported gold nanoparticles.
2. Nitrostyrene Adsorption on TiO2 and Au/TiO2. When
nitrostyrene approaches the TiO2 surface, the CdC interaction
with the Ti atoms of the oxide support is possible only if the
molecule is parallel to the TiO2 surface. However, the repulsion
between the aromatic ring and the O atoms of the oxide surface
is stronger, and all attempts to locate a minimum corresponding
to the adsorption through the CdC on TiO2 evolved to a
structure in which the interaction occurs via the nitro group and,
more precisely, between the oxygen atoms of the nitro group
and the Ti atoms of the surface (Figure 8). In this structure, the
molecule lies perpendicular to the surface, with the two O atoms
of the nitro group 2.14 and 2.39 Å from the nearest surface Ti
atoms, and with a lengthening of one of the N-O bonds from
1.24 to 1.27 Å. The characteristic spectroscopic feature associ-
ated with the CdC stretching mode appears at 1641 cm-1 (Table
4), and the vibrational modes associated with the nitro group
are strongly coupled to those of the aromatic ring and appear
at 1484 and 1459 cm-1, far away from the characteristic
frequency expected for the nitro group of the isolated molecule.
The calculated adsorption energy is quite large, 42 kcal/mol,
suggesting that this strongly chemisorbed nitrostyrene molecule
might be just a spectator and not the reactive species. Here, it
is worth pointing out that the experimental IR measurements,
previously described for the interaction of nitrostyrene with the
TiO2 support and the Au/TiO2 catalyst, show a clear band at
1532 cm-1 which, in the light of the present theoretical results,
cannot be assigned to nitrostyrene strongly chemisorbed through
the nitro group. In order to assign this experimentally observed
band, several geometry optimization calculations were carried
out, starting from a variety of conformations. In this way, a
9
16234 J. AM. CHEM. SOC. VOL. 129, NO. 51, 2007

Chemoselective Hydrogenation of Nitroaromatics on Au/TiO2
A R T I C L E S
measured ∆νas(NO2) frequency shifts (-13 and -19 cm-1
,
respectively) suggests that the nitrostyrene molecules observed
by IR spectroscopy are precisely those weakly adsorbed on the
TiO2 support. Moreover, it further validates the present com-
putational approach and reinforces the conclusions described
above for the interaction of nitrostyrene with low-coordinated
Au atoms.
The model calculations described so far indicate that, on one
hand, interaction of nitrostyrene with pure Au cannot explain
the preference for the nitro group activation observed in the
catalytic experiments and, on the other hand, the support itself
plays only a marginal role in sticking the molecule to the surface.
We are left with the possibility that the adsorption can occur
on the gold atoms interacting with the support. To analyze this
possibility, the adsorption of nitrostyrene on a more realistic
model of the catalyst, involving a gold nanoparticle supported
on TiO2, has been studied (Figure 9). Nitrostyrene was situated
perpendicular to the TiO2 surface, at the interface between the
gold nanoparticle and the oxide support. In this case, and as
described above, a strong interaction between the O atoms of
the nitro group and the Ti atoms of the support is observed,
with calculated Ti-O distances of 2.06 and 2.14 Å. However,
for the Au/TiO2 system, there is another noticeable interaction
between the O atoms of the nitro group and two Au atoms at
the nanoparticle edge. Interestingly enough, these are low-
coordinated Au atoms, directly bonded to four or five other Au
atoms, and at 2.6-2.7 Å from an O atom of the support. The
calculated adsorption energy at this site is 15.4 kcal/mol,
considerably smaller than that on the extended TiO2 support,
and almost as large as that obtained for the interaction of
nitrostyrene with the isolated nanoparticles. However, there are
two important differences between these two cases. First, the
interaction of nitrostyrene with the isolated Au38 nanoparticle
induces a large distortion of the metallic particle, but it does
not modify the geometry of the Au13 supported cluster. Second,
and most importantly, nitrostyrene adsorption on the Au13-on-
TiO2 supported model strongly activates the nitro group while
leaving the CdC almost unaffected. Both NO bond lengths
increase to 1.39 Å, and the conformation of the N atom displays
a noticeable distortion from planarity, evidenced by the C-N-
O-O dihedral angle, which switches from 180° to 125°.
Unsurprisingly, the calculated frequency associated with the pure
CdC stretching vibrational mode appears at 1640 cm-1 (Table
4), while the frequencies dominated by the nitro group atomic
displacements are shifted to 1227 and 1138 cm-1 and include
large contributions from modes related to the aromatic ring
vibrations. In other words, the model calculations strongly
suggest that, while nitrostyrene can be adsorbed and activated
through either the NO2 or the CdC groups on isolated gold
nanoparticles, a different situation appears when this molecule
interacts with Au nanoparticles supported on TiO2. In fact, in
this latter situation, selective adsorption and proper activation
of the nitro group appear to occur naturally, with the nitro group
oxygen atoms interacting with two low-coordinated Au atoms
located at the nanoparticle edge in contact with the support.
This very selective adsorption and activation on the Au/TiO2
with respect to isolated Au nanoparticles or with respect to Au/
SiO2 catalysts can explain the high chemoselectivity observed
for that catalyst. This is further confirmed by additional catalytic
experiments carried out for 3-nitrostyrene hydrogenation using
Figure 6. Nitrostyrene adsorption on the edge of a monatomic row on a
Au(111) surface through the nitro group (left) and the CdC double bond
(right).
Figure 7. Nitrostyrene adsorption on a Au38 nanoparticle through the nitro
group (left) and the CdC double bond (right).
Table 4. Calculated Vibrational Frequencies (cm-1) of
Nitrostyrene Adsorbed on the Different Systems Considered
νCC
νNO2+"
νNO2
isolated
1638
1636
1516, 1509
1627
1537
1470
1317
1308, 1299
AuMAR (NO2)
AuMAR (CdC)
Au38 (NO2)
Au38 (CdC)
TiO2 (strongly)
TiO2 (weakly)
Au13/TiO2
1574, 1545
1494, 1439
1562, 1509
1484, 1459
1524
1395, 1371
1641
1639
1640
1281, 1268
1319
1227, 1138
new adsorption complex was obtained in which nitrostyrene is
weakly bonded to the TiO2 surface. The calculated adsorption
energy is 24 kcal/mol (Table 3), the distances from the O atoms
of the nitro group to the nearest surface Ti atoms are larger
than 3.7 Å, and the NO distances are only slightly elongated
with respect to those obtained for the isolated molecule. As a
consequence of these small changes in the molecular geometry,
the asymmetric νas(NO2) vibration frequency shifts from
1537 cm-1 in the gas phase to 1524 cm-1 (Table 4). The good
agreement between the calculated and the experimentally
9
J. AM. CHEM. SOC. VOL. 129, NO. 51, 2007 16235

A R T I C L E S
Boronat et al.
Table 5. Catalytic Behavior of Different Supported Gold Materials in the Hydrogenation of 3-Nitrostyrenea
selectivity(%)
ABE
particle
size (nm)b
amount of
conversion (%),
time (h)
catalysts
catalyst (mg)
TOFc
NBE
ABS
AZO
AZOXY
1.5% Au/TiO2
4.5% Au/Fe2O3
1.6% Au/SiO2
1.5% Au/C
3.6
3.5
4.0
4.1
25
20
25
25
173
23
10
6
98.5, 6
25, 2
25, 5
5, 6
0.5
96
95
30
41
3
2
0.5
3
10
17
50
42
a Metal oxides such as TiO2 and Fe2O3 play an important role for preferentially reducing the nitro group, which cannot be achieved by using “inert”
supports such as SiO2 or C. Feeding (1 mL): 8.5 mol % of nitrostyrene, 90.5 mol % of toluene (solvent), and 1 mol % of o-xylene (internal standard). NBE,
3-nitroethylbenzene; ABS, 3-aminostyrene; ABE, 3-aminoethylbenzene; AZO, azostyrene; AZOXY, azoxystyrene. b Averaged values. c Calculated at initial
reaction rate as moles of substrate converted per hour and per mole of Au.
Figure 8. Nitrostyrene adsorption on the TiO2 support.
Figure 9. Nitrostyrene adsorption on the Au/TiO2 catalyst model.
different supports; a summary of the results is shown in Table
5. In agreement with the theoretical study, active supports such
as TiO2 or Fe2O3 lead to selective processes, whereas “inert”
ones such as SiO2 or C produce both the reduction of the double
bond and the nitro functions.
the olefinic group on the gold catalyst, but also to the very
favorable adsorption through the nitro when both groups are
competing.
In the most selective Au/TiO2 catalyst, H2 is dissociated on
gold while nitrostyrene is weakly adsorbed on metallic gold
and more strongly on highly uncoordinated gold atoms.
Although nitrostyrene adsorbs very strongly and selectively
through the nitro group on Ti, the high heat of adsorption (42
kcal‚mol-1
Conclusions
The high chemoselectivity of Au/TiO2 for the hydrogenation
of substituted nitroaromatics to the corresponding anilines, and
more specifically that of nitrostyrene, is due not only to an
intrinsic higher rate for the hydrogenation of the nitro versus
) indicates that the molecules strongly adsorbed on
TiO2 may be spectators. On the other hand, nitrostyrene adsorbs
very selectively through the nitro group on two low-coordinated
Au atoms at the nanoparticle edge bonded to four or five other
9
16236 J. AM. CHEM. SOC. VOL. 129, NO. 51, 2007

Chemoselective Hydrogenation of Nitroaromatics on Au/TiO2
A R T I C L E S
Au atoms and at 2.6-2.7 Å away from an oxygen atom of the
support. In this case, the calculated heat of adsorption (15.4
kcal‚mol-1) and the activation of the NO2 group strongly favor
the reactivity of the adsorbed molecule.
Catalunya (2005SGR00697, 2005 PEIR 0051/69, and Distincio´
per a la Promocio´ de la Recerca Universitaria of F.I.) for
financial support. The authors thankfully acknowledge the
computer resources, technical expertise, and assistance provided
by the Barcelona Supercomputing Center-Centro Nacional de
Supercomputacio´n.
Acknowledgment. We thank the Spanish government (projects
MAT2006-14274-C02-01, CTQ2005-08459-CO2-01, UNBA05-
33-001, and grant FPU AP2003-4635) and the Generalitat de
JA076721G
9
J. AM. CHEM. SOC. VOL. 129, NO. 51, 2007 16237