Niobium metallocenes deposited onto mesoporous silica via dry impregnation as catalysts for selective epoxidation of alkenes

Article abstract of DOI:10.1016/j.jcat.2012.11.015

A series of niobium catalysts for the selective epoxidation was synthesised by post-synthesis modification of a commercial silica, starting from niobocene dichloride through solventless organometallic precursor dry impregnation (OM-DI) or conventional liquid-phase grafting technique. OM-DI showed to be cheaper, more versatile, less time-consuming and avoided the use of environmentally unfriendly chlorinated solvents. Nb-SiO₂ catalysts displayed an excellent performance in the epoxidation of limonene, using aqueous hydrogen peroxide as oxidant. Niobium-silica catalysts were obtained via OM-DI for the first time in this occasion. They showed conversions up to 78% and chemoselectivity to epoxide of 98%. An unexpected regioselectivity to exocyclic epoxide was also observed.

Full text of DOI:10.1016/j.jcat.2012.11.015

Journal of Catalysis 298 (2013) 77–83
Contents lists available at SciVerse ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Niobium metallocenes deposited onto mesoporous silica via dry impregnation
as catalysts for selective epoxidation of alkenes
Alessandro Gallo a,1, Cristina Tiozzo a, Rinaldo Psaro a, Fabio Carniato b, Matteo Guidotti a,
⇑
a CNR-Istituto di Scienze e Tecnologie Molecolari, via G. Venezian 21, 20133 Milano, Italy
b Dipartimento di Scienze e Innovazione Tecnologica and Nano-SiSTeMI Interdisciplinary Centre, Università del Piemonte Orientale ‘‘A. Avogadro’’,
Viale T. Michel 11, 15121 Alessandria, Italy
a r t i c l e i n f o
a b s t r a c t
Article history:
A series of niobium catalysts for the selective epoxidation was synthesised by post-synthesis modifica-
tion of a commercial silica, starting from niobocene dichloride through solventless organometallic pre-
cursor dry impregnation (OM-DI) or conventional liquid-phase grafting technique. OM-DI showed to
be cheaper, more versatile, less time-consuming and avoided the use of environmentally unfriendly chlo-
rinated solvents. Nb–SiO2 catalysts displayed an excellent performance in the epoxidation of limonene,
using aqueous hydrogen peroxide as oxidant. Niobium–silica catalysts were obtained via OM-DI for
the first time in this occasion. They showed conversions up to 78% and chemoselectivity to epoxide of
98%. An unexpected regioselectivity to exocyclic epoxide was also observed.
Received 23 August 2012
Revised 13 November 2012
Accepted 14 November 2012
Available online 23 December 2012
Keywords:
Niobium–silica
Titanium–silica
Dry impregnation
Epoxidation
Ó 2012 Elsevier Inc. All rights reserved.
Limonene
Hydrogen peroxide
Heterogeneous catalysis
Mesoporous silica
Exocyclic limonene epoxide
1. Introduction
commercial large scale so far [7]. On the contrary, the study of nio-
bium-based silicates for liquid-phase oxidation deserves a more
Titanium-based mesoporous silica catalysts have displayed, in
last decades, potentially interesting results in the selective epoxi-
dation of alkenes with organic hydroperoxides [1–7]. This kind of
oxidants is generally able to selectively epoxidise the alkenes via
radical-free heterolytic oxidation mechanisms. The use of them
over porous supports with large or very large pore openings (from
2 to 50 nm) has opened the way to the potential synthesis of epox-
ide from bulky and highly functionalised alkenes of interest for the
fine and high added-value chemistry. Unfortunately, most of these
catalysts suffer from rapid deactivation in the presence of water
and aqueous media, and the use of hydrogen peroxide therefore
leads to poor results [8–11]. The structural instability of
titanium–silica systems, in fact, grows in parallel with the hydro-
philicity of the surface and the defectiveness of the silica matrix
[12,13]. Since mesoporous silica materials are non-crystalline sol-
ids with a high hydrophilic nature, all these intrinsic drawbacks
have limited any possible exploitation of these systems on a
careful attention, even though a lower number of research teams
have focused their attention on them [14,15]. The stability and
the robustness of Nb–SiO2 catalysts towards metal leaching with
respect to other transition metals, such as vanadium and chro-
mium, is an important advantage for their use with hydrogen per-
oxide and/or in water-containing media [16–19]. In most literature
reports, niobium–silica catalysts are tested in the epoxidation of
simple alkenes as probe substrates [20–26]. Only few groups
focused on the epoxidation of more complex and functionalised
alkenes [27,28]. The epoxidation of cyclic unsaturated terpenes is
an interesting benchmark reaction as it provides a broad series
of data in terms of catalyst activity as well as chemo- and
regioselectivity.
For these reasons, one major target of the present report is to
prepare and study a series of niobium-based catalysts and to com-
pare their catalytic behaviour with reference titanium-based cata-
lysts, in the liquid-phase epoxidation of a doubly unsaturated
cyclic terpene, such as limonene (Scheme 1). Particular attention
has been paid to the synthesis method of the solid catalysts to be
used in the reaction. Single-site heterogeneous catalysts, especially
obtained by post-synthesis grafting of titanocene precursors, have
previously shown promising results in the epoxidation of alkenes
⇑
Corresponding author. Fax: +39 02 5031 4405.
E-mail address: m.guidotti@istm.cnr.it (M. Guidotti).
Present address: Department of Chemical Engineering, University of California,
1
Santa Barbara, CA 93106-5080, USA.
0021-9517/$ - see front matter Ó 2012 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.jcat.2012.11.015

78
A. Gallo et al. / Journal of Catalysis 298 (2013) 77–83
various temperatures (T = X °C in M/SiO2-DI-X). The mixture was
O
calcined under oxygen at 500 °C for 2 h to obtain the final catalyst.
For Nb/SiO2 and Ti/SiO2 catalysts obtained via liquid-phase
grafting (Nb/SiO2-liq and Ti/SiO2-liq), Nb(Cp)2Cl2 or Ti(Cp)2Cl2,
respectively, was grafted onto the surface of the silica adapting
the procedure developed by Maschmeyer et al. for Ti only [29,41]
and extending it to niobium precursors.
oxidant
+
O
endocyclic
exocyclic
2.2. Catalyst characterisation
Scheme 1. Epoxidation of limonene.
The Nb and Ti content of the prepared samples was determined
by inductively coupled plasma optical emission spectroscopy ICP-
OES (ICAP 6300 Duo, Thermo Fisher Scientific) after mineralisation
of the samples in a microwave digestion apparatus (Milestone MLS
1200; maximum power 500 W) with a mixture of hydrofluoric (aq.
40 wt.%) and fuming nitric acid.
Diffuse reflectance UV–vis spectra were recorded in the range
from 200 to 800 nm on pure samples using a Praying – Mantis Dif-
fuse Reflectance Accessory (Harrick Sci) mounted in an Evolution
600 spectrophotometer (Thermo). A SpectralonÒ disc was used as
reference material for background measurement. The spectra were
recorded under air at room temperature on the samples after cal-
cination at 500 °C in air and cooled under dynamic vacuum.
Raman spectra were recorded in the range 4000–50 cmÀ1 at
4 cmÀ1 resolution with 2000 scans, using a RFS 100 Bruker FT-
Raman spectrometer with a 1064 nm wavelength excitation laser.
Laser power was set at 100 mW. The Raman spectra were elabo-
rated by using OPUS 5.0 software realised by Bruker Optics
Company.
[4,29–32]. Such catalytic systems are typically obtained by chemi-
sorption of the precursor under liquid-phase conditions in chlori-
nated solvents. So, in order to avoid the use of hazardous and
environmentally unsustainable media (in particular, the use of tri-
chloromethane during the grafting procedure) and to simplify,
whenever possible, the preparation steps, an alternative solvent-
less synthesis method based on the dry impregnation of organo-
metallic precursors (OM-DI) over silica has been here
experimented and evaluated. Thanks to this technique, the finely
ground solid OM precursor is mixed and stirred together with
the oxidic support in the same vessel under reduced pressure
(10À6 bar) at various temperatures. The OM precursor is therefore
locally vaporised and deposited on the surface of the support. A
comparable technique has already been applied to the preparation
of transition and post-transition metal oxidic catalysts [33–35]. In
particular, the absence of the solvent prevents a possible blocking
of the pores, avoids the filtration step and usually allows a low
level of contamination by undesired compounds eventually
present in the solvent [36]. Many authors have successfully
investigated the use of metal chlorides and metal alkoxides as
precursors [37–39]. In this report, a series of catalysts obtained
by OM-DI have been prepared starting from bis(cyclopentadi-
enyl)niobium dichloride, and their features were compared to
those of titanium-systems obtained from bis(cyclopentadi-
enyl)titanium dichloride. In fact, despite the fact that niobocene
and titanocene dichlorides have showed, at theoretical level,
suitable characteristics for the application in solventless vapour-
deposition techniques [40], to the best of our knowledge, their
use in the synthesis of catalysts through this method has not been
reported yet.
X-ray diffractograms (XRD) were collected on unoriented
ground powders with a Thermo ARL ‘XTRA-048 diffractometer
with a Cu Ka (k = 1.54 Å) radiation. Diffractograms were recorded
at room temperature between 2° and 70° 2h degrees with a step
size of 0.02°, and a rate of 1° 2h minÀ1. Pure Nb2O5 (Aldrich,
99.99%) was used for comparison.
2.3. Catalytic tests
All catalysts were pretreated in air at 500 °C and cooled to room
temperature under vacuum prior to use. The epoxidation tests on
(R)-(+)-limonene (97% Aldrich; 98% e.e.) were carried out in a
round-bottom glass batch reactor in an oil bath at 90 °C with mag-
netic stirring (ca. 800 rpm) under inert atmosphere. Acetonitrile
(Aldrich, HPLC grade) was used as solvent. Aqueous hydrogen per-
oxide (H2O2; 50% Aldrich) or tert-butyl hydroperoxide (TBHP;
5.5 M in decane; Aldrich) was used as solvent, with an oxidant to
substrate molar ratio of 2: 1 and 1.1: 1, for H2O2 and TBHP, respec-
tively. The total volume of the mixture was 5 mL. Samples were
taken after reaction times of 1, 2, 3, 6 and 24 h and analysed on
GC chromatograms (Agilent 6890 Series; HP-5 column,
30 m  0.25 mm; FID detector), with mesitylene (Fluka, puriss.
P99%) as internal standard. GC-peaks were identified by compar-
ison with peaks of genuine samples of reference standards and by
means of GC–MS. Peaks of regioisomers of the epoxides (endocy-
clic and exocyclic limonene epoxide; Scheme 1) were confirmed
and quantified by 1H NMR signals in the range 3.5–2.5 ppm. Data
of catalytic performance (in Tables 2 and 3) are an average of at
least three measurements. A standard deviation of 2%, 4% and
2 hÀ1 has to be considered on average for the conversion, selectiv-
ity and specific activity values, respectively. In the case of some
promising catalytic systems (Nb/SiO2-DI-RT), fully comparable re-
sults were obtained over different batches of catalyst, prepared in
multiple times. Neither significant auto oxidation nor support-
catalysed contributions to epoxidation were recorded with nio-
bium- and titanium-free siliceous supports, limonene conversion
being about 10–12% with no remarkable epoxide formation. The
2. Experimental details
2.1. Catalyst preparation
All Nb/SiO2 and Ti/SiO2 catalysts were prepared starting from
one type of commercially available Davisil silica (Grace Davison,
LC60A), with 60–200 l
m particle size, 605 m2 gÀ1 specific surface
area, 1.11 mL gÀ1 total pore volume and a mean mesopore diame-
ter broadly centred around 5.5 nm. The silica was pretreated in air
at 500 °C for 1 h and at 500 °C for 1 h under dynamic vacuum
(10À6 bar) and cooled to room temperature under vacuum.
For Nb/SiO2 and Ti/SiO2 catalysts obtained via OM-DI (Nb/SiO2-
DI and Ti/SiO2-DI), the silica support was rehydrated with high
purity deionised (18 MO cm) water (MilliQ Academic, Millipore)
for 2 h and then dried at the rotary evaporator. The sample was cal-
cined at 300 °C for 1 h in air and overnight under dynamic vacuum.
Either niobium or titanium precursor, bis(cyclopentadienyl)nio-
bium(IV) dichloride (Nb(Cp)2Cl2; 95% Aldrich) or bis(cyclopentadi-
enyl)titanium(IV) dichloride (Ti(Cp)2Cl2; 97% Fluka), was finely
milled and added under inert atmosphere in solid phase to the sil-
ica. The solid mixture was stirred overnight under static vacuum at

A. Gallo et al. / Journal of Catalysis 298 (2013) 77–83
79
Table 1
formation of peroxyimidic acid, due to the concomitant presence of
Synthesis details and metal content of the series of Ti/SiO2 and Nb/SiO2 catalysts.
acetonitrile and hydrogen peroxide, can be excluded under these
conditions. After all tests, the presence of residual hydrogen perox-
ide or TBHP was checked and confirmed by iodometric assays and
titrations.
In order to check the leaching of niobium or titanium species,
the solid catalyst was removed from the liquid mixture by centri-
fugation, and the resulting solution was tested for further reaction.
In the tests for the recovery of the catalyst, the solid was separated
by filtration and thoroughly washed with fresh acetonitrile and
then with methanol (Fluka, HPLC grade). The filtered solid was
dried at 100 °C, weighed, activated again at 500 °C in dry air and
then reused in a new test as described above.
Catalyst
Synthesis
method
Synthesis
temperature (°C)
Metal loading
(wt.%)a
Ti/SiO2-DI-120
Ti/SiO2-liq
Dry impregnation
Liquid-phase grafting Room temperature 1.75
120
1.66
Nb/SiO2-DI-170 Dry impregnation
Nb/SiO2-DI-150 Dry impregnation
170
150
80
1.96
1.71
1.49
Nb/SiO2-DI-80
Nb/SiO2-DI-RT
Nb/SiO2-liq
Dry impregnation
Dry impregnation
Liquid-phase grafting Room temperature 1.30
Room temperature 1.75
a
Determined by digestion and ICP-OES analysis.
In all catalysts, the desired expected metal loading was 2 wt.%,
according to the amount of metallocene precursor added in the
dry impregnation mixture. In fact, 2 wt.% is generally, for Ti and
Nb, the optimal loading for the epoxidation of a large variety of al-
kenes [27,31,41,43]. Nevertheless, the actual metal content
showed that only a part of the expected metal was present (Table
1). A fraction of the precursor was lost, likely deposited on the
walls of the glass vessel where the dry impregnation process was
carried out or taken by the pump during the evacuation procedure.
Also, in this case, however, the elemental analysis of different ali-
quots of powder sampled randomly on the same batch of catalyst
showed a uniform dispersion of the metal species on silica support.
In a parallel way, two Nb/SiO2-liq and Ti/SiO2-liq catalysts were
prepared via metallocene grafting in liquid phase and were used as
comparison for OM-DI systems. It is worth noting that Ti/SiO2 pre-
pared by grafting titanocene in liquid phase is a widely studied sys-
tem in literature [4,29–31], whereas the Nb/SiO2 analogue has
been prepared, to our best knowledge, for the first time in this
work. From a merely synthetic point of view, the dry impregnation
approach is simpler and cheaper, as it does not require the use of a
hazardous and environmentally unfriendly solvent (in particular,
chloroform to dissolve properly metallocene dichlorides) or of a
base (triethylamine), to trigger the chemisorption of (cyclopentadi-
enyl)metal moieties onto the surface silanols. In addition, the poor
solubility of Nb(Cp)2Cl2 in chloroform (lower than Ti(Cp)2Cl2)
causes a less efficient use of the organometallic precursor, and an
amount lower than expected of Nb was therefore grafted onto
the final solid (1.30 wt.% vs. expected 2 wt.%).
3. Results and discussion
3.1. Catalyst preparation
The first goal of this work is to prepare niobium-containing sil-
ica catalysts (and some reference titanium catalysts) via a versatile
solventless deposition of metallocene dichloride precursors under
controlled reduced pressure (10À6 bar).
For this aim, a dry impregnation approach was chosen. A cheap
commercially available non-ordered mesoporous silica was chosen
as support. Its specific surface area is high (ca. 600 m2 gÀ1), thus
providing a broad surface for the deposition of the organometallic
precursor. Niobocene dichloride has been preferred as a precursor
to the more common niobium alkoxides, since the former is less
reactive, less moisture-sensitive and more easy-to-handle than
the latter.
The silica support was previously calcined at 500 °C to clean the
surface and remove completely any organic impurity eventually
present as a contaminant. However, such treatment may lead to
a marked condensation of the silanol groups and reduce exces-
sively the number of free –OH moieties available for chemisorption
of the metallocene precursor. So, the support was first rehydrated
with water and then treated at a milder temperature (300 °C) [42].
During the deposition process, the temperature at which the dry
impregnation occurred was varied as the only parameter. In the
case of Nb/SiO2-DI systems, a large set of temperatures was tested,
spanning from 170 °C down to 25 °C (room temperature), and a
broad series of Nb/SiO2-DI catalysts was therefore obtained (Ta-
ble 1). For Ti/SiO2-DI systems, 120 °C was chosen as optimal depo-
sition temperature [40].
3.2. Catalyst characterisation
The pristine silica support (SiO2) and the catalysts prepared by
both dry impregnation (DI) and liquid-phase grafting (liq) (after
calcination) showed type IV isotherms with an H2 hysteresis loop
typical of solids with disordered structural pore size distribution.
The pores size distribution in the support and in the derived cata-
lysts covered a wide range of values from 2.0 to 8.0 nm. The depo-
sition of niobium on the SiO2 surface led to a reduction in the
specific surface area, from 605 m2 gÀ1, for the support, to 481
and 447 m2 gÀ1 for Nb/SiO2-liq and Nb/SiO2-DI samples, respec-
During the dry impregnation process, the higher the tempera-
ture, the faster was the colouring of the white support due to the
deposition and (partial) decomposition of the metal precursor onto
the silica.2 At the end, in all cases, the freshly prepared catalysts (be-
fore calcination) showed a uniform colour (brown-greenish for nio-
bium systems or brown-red for titanium ones) in all parts,
suggesting an even deposition of the metal species. The mapping
for Nb element by SEM-EDS analysis on Nb/SiO2-DI samples shows,
at micrometric level, a uniform concentration of metal species on the
support (Supplementary Information; Fig. S1).
By means of a final calcination in oxygen, the organic cyclopen-
tadienyl ligands and the partially decomposed carbonaceous moi-
eties (eventually formed during the deposition step) were burnt
away, and the active Ti(IV) and Nb(V) centres on silica were ob-
tained. Some investigations on the mechanism of the grafting via
dry impregnation technique and on the nature of the intermediate
species are still in progress and will be discussed in a further study.
tively, and a reduction in the pores volume, from 1.11 cm3 gÀ1
,
for the support, to 0.87 and 0.85 cm3 gÀ1 for Nb/SiO2-liq and Nb/
SiO2-DI, respectively. These values indicate that, in both cases,
the deposition of the active species occurred mainly onto the inter-
nal surface of the porous solids, causing a moderate shrinkage of
the non-ordered porosity.
The calcined samples underwent DR UV–vis spectroscopic anal-
ysis in order to have an evaluation on the site isolation and the
coordination geometry of the metal sites, by monitoring the li-
gand-to-metal charge-transfer bands of Nb(V) and Ti(IV) (Fig. 1).
For all Nb/SiO2 and Ti/SiO2 systems, no absorption bands were
detected above 400 nm. No large domains of Nb2O5 or TiO2 were
2
The colouring of the support was extremely rapid for Nb/SiO2-DI-170 and it
turned light brown in few minutes, while the dry impregnation at room temperature
led to a pale green material in few hours.

80
A. Gallo et al. / Journal of Catalysis 298 (2013) 77–83
Fig. 2. X-ray patterns of Nb/SiO2-liq (a), Nb/SiO2-DI-80 (b) and bulk Nb2O5
(c) samples.
Ti/SiO2-liq analogues obtained via liquid-phase grafting tech-
niques, especially when they were prepared at room temperature.
The chemical nature of the aggregates, responsible of the UV
band observed at higher wavelengths (at ca. 350 nm), was con-
firmed by collecting XRD pattern and Raman spectrum of the Nb/
SiO2-DI sample prepared at higher temperatures (80 °C). X-ray dif-
fractogram of Nb/SiO2-DI-80 in comparison with that of Nb/SiO2-
liq solid is reported in Fig. 2.
Nb/SiO2 sample prepared via liquid-phase grafting showed an
X-ray profile characterised by the presence of a wide band centred
at 22° 2h, typical of amorphous silica (Fig. 2). No peaks attributed
to the presence of Nb2O5 crystalline aggregates were clearly
detectable in this sample, suggesting that the Nb centres were
homogeneously distributed on the silica surface, in agreement
with the spectroscopic results. Different features were observed
for the sample prepared by dry impregnation at 80 °C (Fig. 2). In
this specific case, along with the presence of the band typical of
the amorphous silica (band at 22° 2h), two weak X-ray peaks at
22.7° and 28.6° h were clearly visible. By comparing the diffracto-
gram of Nb/SiO2-DI-80 with the one of a commercial Nb2O5 with
orthorhombic structure [50], it was possible to observe that these
two peaks can be ascribed to the presence of Nb2O5 nanometre size
aggregates on the silica surface, in agreement with DR UV–vis data.
Raman spectroscopy is a powerful method to identify the sur-
face metal oxide species on oxide supports, providing information
at molecular level. On the light of these considerations, Nb/SiO2-
DI-80 sample was submitted to Raman spectroscopy characterisa-
tion in order to identify the nature of Nb2O5 species present on the
silica surface and to support the X-ray diffraction results described
above.
Fig. 1. DR UV–vis spectra of dry Ti/SiO2-DI-120 (a), Nb/SiO2-DI-RT (b), Nb/SiO2-DI-
80 (c), Nb/SiO2-DI-150 (d), Nb/SiO2-DI-170 (e) and Nb/SiO2-liq (f) after calcination
(before catalytic test).
therefore present on the catalysts [44,45]. As a general feature, all
samples showed an absorption maximum in the range
210–250 nm that is assigned to the charge-transfer transition be-
tween oxygen atoms and the metal centres (Nb(V) and Ti(IV)) in
tetrahedral coordination (Fig. 1) [23,46,47]. This indicates that
the metal centres are isolated, and both synthetic methods are
suitable to obtain the deposition of evenly dispersed Nb(V) and
Ti(IV) sites.
Nevertheless, Nb/SiO2 samples prepared by dry impregnation
showed a shoulder centred at 350 nm due to the presence of
Nb2O5-like nano-aggregates formed during the dry impregnation
process. It is important to note that this shoulder increased in
intensity passing from Nb/SiO2 sample prepared at room tempera-
ture to samples obtained at higher temperatures (Nb/SiO2-DI-80,
Nb/SiO2-DI-150 and Nb/SiO2-DI-170). This feature suggests that
some Nb2O5-like aggregates with nanometre size were formed
during the dry impregnation procedure carried out at higher tem-
peratures due to a higher extent of decomposition under those
conditions [45,48,49]. So, in these cases, when the deposition of
the metallocenic precursor was extremely rapid, it led to a less
controlled and irregular formation of relatively large oxidic nio-
bium domains.
Fig. 3 shows the Raman spectra of bulk Nb2O5 sample in com-
parison with Nb/SiO2-DI-80 and Nb/SiO2-liq solids.
Raman spectrum of bulk diniobium pentoxide was character-
ised by a strong band centred at 700 cmÀ1, associated with the
stretching modes of different niobia polyhedra in orthorhombic
phase, in agreement with X-ray results and as reported in the liter-
ature [50]. This peak was completely absent in the Raman spec-
trum of Nb/SiO2-liq sample, but it is clearly visible (yet with a
lower intensity with respect to pure Nb2O5 solid) in the Raman
profile of niobium-containing silica prepared by dry impregnation
at 80 °C. Such result confirms that the dry impregnation methodol-
ogy, when it was carried out at higher temperatures, led to the for-
mation of a small amount of nanometric Nb2O5 aggregates with
orthorhombic structure on the surface of the siliceous support.
Such relatively larger aggregates were due to the fact that, at
Finally, Nb/SiO2-DI and Ti/SiO2-DI catalysts displayed rather
comparable spectroscopic patterns to their Nb/SiO2-liq and

A. Gallo et al. / Journal of Catalysis 298 (2013) 77–83
81
ness of Ti–O–Si bonds in mesoporous titanium–silica systems
against hydrolysis and to their aggregation in the presence of water
[47]. Indeed, some low initial formation of epoxide (ca. 7% yield)
was detected in the first hour of reaction, but then, it rapidly disap-
peared, as it transformed into a wide range of secondary products.
Conversely, niobium–silica materials did withstand the presence of
aqueous H2O2 and showed a remarkable epoxidation activity even
after 1 h of reaction. In particular, in all the tests, aqueous H2O2
was added in one aliquot at the beginning of the reaction, and no
slow dropwise addition of the oxidant was adopted. In fact, such
peculiar protocol of oxidant addition has been successful in the
epoxidation of various alkenes over Ti–silica catalysts, as it mini-
mises the unproductive decomposition of hydrogen peroxide and
keeps the local water concentration as low as possible [53–55].
This was not the case when Nb/SiO2 solids were used as catalysts,
and no dropwise oxidant addition protocol was thus necessary.
In terms of specific activity of the catalysts (moles of limonene
converted per mole of metal present in the catalyst in the unit of
time), Nb/SiO2 catalysts showed better performances than Ti/SiO2
ones. Actually, the limonene conversion on Nb/SiO2 reached high
values during the first hour of reaction, and then, it attained its
maximum within the fourth hour, without any noteworthy
improvement in the following 24 h (Supplementary Information;
Fig. S2). The loss in activity after the forth hour of reaction can
be ascribed to the gradual formation of oxidised side products on
the surface of the catalyst, leading to a gradual deactivation of nio-
bium sites [23]. The amount of organic by-products on the surface
of the spent catalyst was relatively moderate (C, H, N elemental
analysis showed some 3–4% of carbon on the washed, used cata-
lyst), but such deposits could be removed completely only by cal-
cination at high temperature (above 400 °C) in air.
Fig. 3. Raman spectra of Nb/SiO2-liq (a), Nb/SiO2-DI-80 (b) and bulk Nb2O5
(c) samples.
higher temperatures, the particles of the organometallic precursor
may, at some extent, undergo decomposition before they can sub-
limate and deposit onto the activated silica support surface. Never-
theless, in order to obtain catalysts where the metal centres are as
disperse as possible, milder dry impregnation temperatures, thus
lower deposition rates, appeared to be the ideal ones.
3.3. Catalytic performance
The catalysts were tested and compared in the liquid-phase
epoxidation of limonene. By studying this reaction, it is possible
to evaluate not only the activity and chemoselectivity of the cata-
lysts, but also their regioselectivity (with respect to the formation
of the endocyclic or exocyclic epoxide; Scheme 1). In addition, two
oxidants were used and compared: hydrogen peroxide, one of the
most promising reagents for environmentally sustainable oxida-
tion processes and tert-butyl hydroperoxide (TBHP), an oxidant
that is particularly suitable for use under anhydrous conditions
[51,52].
Iodometric assays on the final mixtures revealed that in none of
the tests, hydrogen peroxide had been the limiting agent of the
reaction, and the oxidant efficiency (the amount of oxidised prod-
ucts obtained per amount of consumed oxidant) was, on average,
rather good with hydrogen peroxide (above 60%). For this reason,
the loss in catalyst activity described above cannot be ascribed to
the complete consumption (via unproductive disproportionation)
of hydrogen peroxide, since, at the end of all tests, a non-negligible
amount of oxidant was always detected.
On titanium, on the contrary, the initial reaction rate was lower,
and longer reaction times (as long as 24 h) were necessary to attain
with Ti/SiO2 + TBHP systems the same results as those obtained
with Nb/SiO2 + H2O2 in only 3 h. For this reason, reaction times of
1 and 3 h were chosen as the most significant one for a direct com-
parison of the catalytic properties. The oxidant efficiency was al-
ways very high (higher than 90%) with TBHP. Thanks to this, it
was possible to use a lower excess of oxidant with TBHP (1.1:1
From the direct comparison of niobium- and titanium-based sil-
ica catalysts (Table 2), Nb/SiO2 systems displayed the best perfor-
mance with hydrogen peroxide as oxidant, whereas Ti/SiO2
systems showed good results in the presence of TBHP only. The
better results in epoxidation achieved over mesoporous
titanium–silica catalysts in the presence of TBHP rather than with
H2O2 are widely described in previous literature [7,10] and
references therein and are generally ascribed to the poor robust-
Table 2
Catalytic performance of Ti/SiO2 and Nb/SiO2 catalysts in the liquid-phase epoxidation of limonene.
Catalyst
Oxidant
Ca 1 h (%)
Sb 1 h (%)
Endo/exoc 1 h
Ca 3 h (%)
Sb 3 h (%)
Actd 1 h (hÀ1
)
Nb/SiO2-DI-RT
Nb/SiO2-DI-RT
Nb/SiO2-liq
H2O2
TBHP
H2O2
TBHP
H2O2
TBHP
H2O2
TBHP
58
7
98
27:73
n.d.
24:76
76:24
n.d.
98:02
91:09
92:08
62
14
74
30
13
55
18
77
97
71
>98
80
10
>98
20
29
4
46
20
3
13
3
14
n.d.
>98
80
40
96
64
29
10
45
12
52
Nb/SiO2-liq
Ti/SiO2-DI-120
Ti/SiO2-DI-120
Ti/SiO2-liq
62
93
Ti/SiO2-liq
81
Conditions: dry CH3CN solvent; 100 mg cat; 1.0 mmol limonene; 2.0 mmol aq. H2O2 or 1.1 mmol dry TBHP; reflux temperature; batch reactor.
a
Limonene conversion.
Selectivity to limonene epoxide.
b
c
Endocyclic/exocyclic epoxide ratio.
d
Specific activity ([molconverted limonene] [molNb h]À1) after 1 h.

82
A. Gallo et al. / Journal of Catalysis 298 (2013) 77–83
Table 3
Catalytic performance of the series of Nb/SiO2 catalysts (obtained under different synthesis conditions) in the liquid-phase epoxidation of limonene with hydrogen peroxide.
Catalyst
Ca 1 h (%)
Sb 1 h (%)
Ca 24 h (%)
Sb 24 h (%)
Actc 1 h (hÀ1
)
Nb/SiO2-DI-170
Nb/SiO2-DI-150
Nb/SiO2-DI-80
Nb/SiO2-DI-RT
Nb/SiO2-liq
43
47
57
58
64
96
>98
>98
98
61
63
73
66
80
95
92
91
90
77
20
25
35
29
46
>98
Conditions: dry CH3CN solvent; 100 mg cat; 1.0 mmol limonene; 2.0 mmol aq. H2O2; reflux temperature; batch reactor.
a
Limonene conversion.
b
Selectivity to limonene epoxide.
c
Specific activity ([molconverted limonene] [molNb h]À1) after 1 h.
oxidant to limonene molar ratio) than with hydrogen peroxide (2:1
molar ratio). Over Ti/SiO2 systems too, TBHP was not the limiting
reagent, under these conditions.
displayed an inverse trend of the activity as a function of the dry
impregnation temperature. Roughly, the milder the temperature,
the higher was the specific activity of the catalyst and, hence, the
higher the limonene conversion value. In fact, the most active nio-
bium catalysts resulted to be the Nb/SiO2-DI-80 and the Nb/SiO2-
DI-RT systems. Then, a moderate (but non-negligible) diminution
of the specific activity was recorded, and on Nb/SiO2-DI-170, it
was 40% or 30% less than the one obtained on Nb/SiO2-DI-80 or
Nb/SiO2-DI-RT, respectively. Such observation is in good agreement
with DR UV–vis and Raman spectroscopic data, according to which
the milder the dry impregnation temperature, the better was the
dispersion of the niobium sites and the lower the occurrence of
Nb2O5-like aggregates. These relatively larger Nb2O5-like aggre-
gates, indeed, even though they had not a detrimental effect per
se on the catalytic performance, likely led to a lower catalyst activ-
ity, as they caused the diminution of the amount of available active
tetrahedral Nb(V) on the catalyst surface [26,45]. Nevertheless, tak-
ing into account the values of yield to the desired epoxide (in terms
of conversion  selectivity), then, DI-derived catalysts showed
interesting performances as well.
The four Nb/SiO2-DI catalysts showed, on the contrary, fully
comparable results in terms of chemoselectivity and regioselectiv-
ity (ca. 75:25 endocyclic/exocyclic epoxide ratio, over all the cata-
lysts in Table 3). Such behaviour is a clue that the niobium active
sites were obtained with a different dispersion and accessibility,
by varying the temperature of the dry impregnation process, but
they did not change in chemical nature.
All the catalysts (both niobium- and titanium-based ones) were
easily recovered by filtration, washed, calcined and reused in a sec-
ond catalytic run. They kept their epoxidation activity and very
high selectivity to epoxide. The conventional tests to prove the het-
erogeneous nature of the catalytic reaction confirmed no homoge-
neous and catalytically active niobium or titanium species were
formed in the presence of aqueous H2O2 or anhydrous TBHP. The
good stability not only of Ti–SiO2, but also of Nb–SiO2 catalysts
has been already reported in previous works [27,29,48]. It is worth
highlighting, however, that Nb/SiO2-DI-RT and Ti/SiO2-DI-RT
showed fully comparable performance with the results recorded
over fresh catalysts (see also DRS UV–vis spectra after one catalytic
test; Supplementary Information, Fig. S3).
The comparison of the two series of niobium catalysts, namely
prepared via dry impregnation or via liquid-phase grafting, shows
that the dispersion of the metal sites is an important factor in
terms of catalytic-specific activity. Nb/SiO2-liq, in fact, proved to
be the most active catalyst, and this is consistent with the good site
isolation of Nb centres observed by DR UV–vis spectroscopy
(Fig. 1). However, Nb/SiO2-DI catalysts obtained at lower tempera-
tures (RT or 80 °C) showed interesting performances too, in terms
of alkene conversion after 1 h (57% and 58%, respectively), thanks
to their good level of Nb site dispersion, notwithstanding the pres-
ence of a small fraction of nanosized Nb2O5-like aggregates formed
during the dry impregnation process (see Section 3.2). In this case
and under these conditions, the presence of a limited fraction of
niobia nanodomains is not fully detrimental, as it happens when
TiO2 aggregates are present on titanium–silica catalysts, leading
to the undesired decomposition of the oxidant [7,56]. Nb2O5-like
aggregates can, on the contrary, show some epoxidation activity
themselves [57] and do not lead to complete disproportionation
of hydrogen peroxide (H2O2 was not the limiting agent under these
conditions; see above).
With regard to the selectivity to epoxides, the reaction prefer-
entially led to the chemoselective formation of limonene epoxides
(both endo and exocyclic ones), except in the case of Ti/SiO2 + H2O2
systems, in which the poor selectivity values were due to the ex-
tended formation of undesired side products via free-radical
homolytic oxidation processes [58]. In these cases, important
amounts of allylic oxidation products (i.e. carveol and isomeric
compounds) were detected. Conversely, Nb/SiO2-liq, the most ac-
tive system, led to the formation of acid-catalysed secondary prod-
ucts over long reaction times (24 h) and, hence, to lower epoxide
selectivity values (77%; Table 3).
Finally, in terms of regioselectivity, a surprising preferential
regioselective epoxidation of the exocyclic C@C bond was observed
with Nb/SiO2-DI and Nb/SiO2-liq catalysts in the presence of H2O2
only. In the other cases, with Nb/SiO2 + TBHP and with all Ti–silica
systems, a fully expected preference towards endocyclic C@C bond
was obtained. Actually, the electrophilic oxygen from the activated
oxidant species (H2O2 or TBHP) should always attack preferentially
the trisubstituted endocyclic unsaturation [27], but over Nb/
SiO2 + H2O2, the opposite is true. The reasons causing such inver-
sion in regioselectivity are not fully elucidated. Preliminary inves-
tigations, however, seem to suggest that sterical factors have a
stronger influence on the regioselectivity than electronic ones.
Experimental work is in progress to explicate this unusual result.
As a last point, the influence of the temperature at which dry
impregnation is performed was studied with respect to the catalytic
behaviour (Table 3). Particular attention was paid to niobium-con-
taining silica catalysts, as they showed the most promising features
in epoxidation. The four catalysts, Nb/SiO2-DI-170, 150, 80 and RT,
4. Conclusions
An alternative solventless post-synthesis protocol has been suc-
cessfully proposed and optimised for the preparation of mesopor-
ous silica epoxidation catalysts containing titanium and niobium
sites. The synthesis method, based on an organometallic dry
impregnation (OM-DI) approach of metallocenic precursors, was
particularly convenient with respect to previously described (and
widely studied) liquid-phase grafting techniques, as it is cheaper,
more versatile, less time-consuming and avoids the use of environ-
mentally unfriendly chlorinated solvents and of amine additives. It

A. Gallo et al. / Journal of Catalysis 298 (2013) 77–83
83
[13] L.J. Davies, P. McMorn, D. Bethell, P.C. Bulman Page, F. King, F.E. Hancock, G.J.
Hutchings, J. Catal. 198 (2001) 319.
[14] I. Nowak, M. Ziolek, Chem. Rev. 99 (1999) 3603.
[15] V. Lacerda Jr., D. Araujo dos Santos, L.C. da Silva-Filho, S.J. Greco, R. Bezerra dos
Santos, Adrichim. Acta 45 (2012) 19.
[16] M. Ziolek, I. Sobczak, I. Nowak, P. Decyk, A. Lewandowska, J. Kujawa, Micropor.
Mesopor. Mater. 35–36 (2000) 195.
led to reasonably even dispersions of Nb(V) or Ti(IV) sites on the
mesoporous silica surface. OM-DI-derived catalysts were active
in the selective epoxidation of limonene and were nearly compara-
ble to the catalysts obtained via liquid-phase grafting
methodology.
Ti/SiO2-DI catalysts showed good results in the epoxidation
with anhydrous TBHP. Nb/SiO2-DI catalysts, on the contrary, dis-
played an excellent performance in the presence of aqueous hydro-
gen peroxide. High conversions (up to 78%) and a complete
chemoselectivity to the desired epoxide (>98%) were obtained over
Nb/SiO2-DI-RT catalyst in short reaction times (1 h). Additionally,
Nb/SiO2 systems showed an unexpected peculiar regioselectivity
towards the epoxidation of the less electron-rich exocyclic C@C
double bond of limonene.
For these reasons, niobium-containing mesoporous silica cata-
lysts obtained by post-synthesis methods deserve a deeper atten-
tion. They can indeed find several potential applications in the
selective oxidation of various richly functionalised substrates, for
which their particular selectivity properties can be exploited at
best.
[17] M. Ziolek, A. Lewandowska, M. Renn, I. Nowak, P. Decyk, J. Kujawa, Stud. Surf.
Sci. Catal. 135 (2001) 366.
[18] M. Ziolek, Catal. Today 78 (2003) 47.
[19] M. Selvaraj, S. Kawi, D.W. Park, C.S. Ha, J. Phys. Chem. C 113 (2009) 7743.
[20] Y. Liu, K. Murata, M. Inaba, Chem. Lett. 32 (11) (2003) 992.
[21] I. Nowak, M. Ziolek, Micropor. Mesopor. Mater. 78 (2005) 281.
[22] F. Somma, P. Canton, G. Strukul, J. Catal. 229 (2005) 490.
[23] J.M.R. Gallo, H.O. Pastore, U. Schuchardt, J. NonCryst. Solids 354 (2008) 1648.
[24] A. Held, P. Florczak, Catal. Today 142 (2009) 329.
[25] C.A. Deshmane, J.B. Jasinski, P. Ratnasamy, M.A. Carreon, Catal. Commun. 15
(2011) 46.
[26] M. Ziolek, P. Decyk, I. Sobczak, M. Trejda, J. Florek, H. Golinska, W. Klimas, A.
Wojtaszek, Appl. Catal. A: Gen. 391 (2011) 194.
[27] F. Somma, A. Puppinato, G. Strukul, Appl. Catal. A: Gen. 309 (2006) 115.
[28] M.A. Feliczak-Guzik, I. Nowak, Catal. Today 142 (2009) 288.
[29] T. Maschmeyer, F. Rey, G. Sankar, J.M. Thomas, Nature 378 (1995) 159.
[30] K.K. Kang, W.S. Ahn, J. Mol. Catal. A: Chem. 159 (2000) 403.
[31] M. Guidotti, N. Ravasio, R. Psaro, G. Ferraris, G. Moretti, J. Catal. 214 (2) (2003)
247.
[32] V. Dal Santo, F. Liguori, C. Pirovano, M. Guidotti, Molecules 15 (6) (2010) 3829.
[33] S. Jaenicke, W.L. Loh, Catal. Today 49 (1999) 123.
[34] A. Hagemeyer, Z. Hogan, M. Schlichter, B. Smaka, G. Streukens, H. Turner, A.
Volpe Jr., H. Weinberg, K. Yaccato, Appl. Catal. A: Gen. 317 (2007) 139.
[35] D.P. Debecker, M. Stoyanova, U. Rodemerck, P. Eloy, A. Leonard, B.L. Su, E.M.
Gaigneaux, J. Phys. Chem. C 114 (2010) 18664.
[36] R.N. d’Alnoncourt, M. Becker, J. Sekulic, R.A. Fischer, M. Muhler, Surf. Coat.
Technol. 201 (2007) 9035.
[37] J. Klaas, G. Schulz-Ekloff, N.I. Jaeger, J. Phys. Chem. B 101 (1997) 1305.
[38] C.L. Tavares da Silva, V.L. Loyola Camorim, J.L. Zotin, M.L. Rocco Duarte Pereira,
A. da Costa Faro Jr., Catal. Today 57 (2000) 209.
[39] J. Lu, K.M. Kosuda, R.P. Van Duyne, P.C. Stair, J. Phys. Chem. C 113 (2009)
12412.
Acknowledgments
The authors thank the Italian Ministry of Education, University
and Research through the Project ‘‘ItalNanoNet’’ (Rete Nazionale di
Ricerca sulle Nanoscienze; Prot. No. RBPR05JH2P) and Regione
Lombardia (Project ‘‘Mind in Italy’’) for economical support. They
also thank Prof. Sandro Recchia for SEM-EDS analysis.
[40] H.P. Diogo, M.E. Minas da Piedade, J.M. Gonçalves, M.J.S. Monte, M.A.V. Ribeiro
da Silva, Eur. J. Inorg. Chem. 1 (2001) 257.
Appendix A. Supplementary material
[41] M. Guidotti, L. Conti, A. Fusi, N. Ravasio, R. Psaro, J. Mol. Catal. A 182–183
(2002) 149.
[42] L.T. Zhuravlev, Langmuir 3 (1987) 316.
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.jcat.2012.11.015.
[43] L.J. Burcham, J. Datka, I.E. Wachs, J. Phys. Chem. B 103 (1999) 6015.
[44] B.J. Aronson, C.F. Blanford, A. Stein, Chem. Mater. 9 (1997) 2842.
[45] P. Carniti, A. Gervasini, M. Marzo, J. Phys. Chem. C 112 (2008) 14064.
[46] X. Gao, I.E. Wachs, M.S. Wong, J.Y. Yingy, J. Catal. 203 (2001) 18.
[47] E. Gianotti, C. Bisio, L. Marchese, M. Guidotti, N. Ravasio, R. Psaro, S. Coluccia, J.
Phys. Chem. C 111 (2007) 5083.
[48] M. Trejda, A. Tuel, J. Kujawa, B. Kilos, M. Ziolek, Micropor. Mesopor. Mater. 110
(2008) 271.
[49] J.M.R. Gallo, H.O. Pastore, U. Schuchardt, J. Catal. 243 (2006) 57.
[50] V.S. Braga, J.A. Dias, S.C.L. Dias, J.L. de Macedo, Chem. Mater. 17 (2005) 690.
[51] R.A. Sheldon, J. Mol. Catal. 7 (1980) 197.
[52] M. Guisnet, M. Guidotti, in: A.W. Chester, E.G. Derouane (Eds.), Zeolite
Chemistry and Catalysis. An Integrated Approach and Tutorial, Springer, 2009,
p. 275.
[53] J.M. Fraile, J.I. Garcia, J.A. Mayoral, E. Vispe, Appl. Catal. A: Gen. 245 (2003) 363.
[54] M. Guidotti, C. Pirovano, N. Ravasio, B. Lázaro, J.M. Fraile, J.A. Mayoral, B. Coq,
A. Galarneau, Green Chem. 11 (2009) 1421.
[55] M. Guidotti, E. Gavrilova, A. Galarneau, B. Coq, R. Psaro, N. Ravasio, Green
Chem. 13 (2011) 1806.
[56] R.A. Sheldon, J.K. Kochi, Metal-Catalyzed Oxidations of Organic Compounds,
Academic Press, New York, 1981.
References
[1] A. Corma, M.A. Camblor, P. Esteve, A. Martinez, J.P. Pérez Pariente, J. Catal. 145
(1994) 151.
[2] G. Sankar, F. Rey, J.M. Thomas, G.N. Greaves, A. Corma, B.R. Dobson, A.J. Dent, J.
Chem. Soc. Chem. Commun. (1994) 2279.
[3] P.T. Tanev, M. Chibwe, T.J. Pinnavaia, Nature 368 (1994) 321.
[4] R.D. Oldroyd, J.M. Thomas, T. Maschmeyer, P.A. MacFaul, D.W. Snelgrove, K.U.
Ingold, D.D.M. Wayner, Angew. Chem. Int. Ed. Engl. 35 (1996) 2787.
[5] J.M. Fraile, J.I. Garcìa, J.A. Mayoral, E. Vispe, D.R. Brown, M. Naderi, J. Chem. Soc.
Chem. Commun. (2001) 1510.
[6] S.C. Laha, R. Kumar, J. Catal. 208 (2002) 339.
[7] M.G. Clerici, in: S.D. Jackson, J.S.J. Hargreaves (Eds.), Metal Oxide Catalysis,
Wiley-VCH, Weinheim, Germany, 2009, pp. 705–754.
[8] A. Hagen, K. Schueler, F. Roessner, Micropor. Mesopor. Mater. 51 (2002) 23.
[9] F. Chiker, F. Launay, J.P. Nogier, J.L. Bonardet, Green Chem. 5 (2003) 318.
[10] O.A. Kholdeeva, N.N. Trukhan, Russ. Chem. Rev. 75 (5) (2006) 411.
[11] M. Guidotti, B. Lazaro, in: M. Guisnet, F. Ramoa Ribeiro (Eds.), Deactivation
and Regeneration of Zeolite Catalysts, Imperial College Press, London, UK,
2011, p. 303.
[57] H. Shima, M. Tanaka, H. Imai, T. Yokoi, T. Tatsumi, J.N. Kondo, J. Phys. Chem. C
113 (2009) 21693.
[58] J.M. Fraile, J.I. Garcia, J.A. Mayoral, E. Vispe, J. Catal. 204 (2001) 146.
[12] A. Carati, C. Flego, E. Previde Massara, R. Millini, L. Carluccio, W.O. Parker Jr., G.
Bellussi, Micropor. Mesopor. Mater. 30 (1999) 137.