Mechanochemical activation of MoS2-Surface properties and catalytic activities in hydrogenation and isomerization of alkenes and in H2/D2 exchange

Article abstract of DOI:10.1016/j.jcat.2008.10.005

High-energy ball milling has been employed to convert a microcrystalline, stoichiometric (S/Mo = 2), catalytically completely inactive MoS₂ (prepared by high-temperature decomposition of ammonium tetrathiomolybdate, ATM) into an active catalyst, the activity profile of which was studied with test reactions (ethylene hydrogenation, H₂/D₂ scrambling, cis-trans isomerization of cis-but-2-ene, double-bond isomerization of 2-methyl-1-butene). Structural and surface properties of the materials were studied by XRD, SEM, TEM, XPS, nitrogen physisorption, oxygen chemisorption, and isotope exchange with D₂ (quantity of exchangeable surface hydrogen). The reaction rates obtained after mechanochemical activation were compared with data from a reference MoS₂ made by low-temperature decomposition of ATM. Ball-milled MoS₂ became active for most of the test reactions (except double-bond isomerization) only after reductive treatments that were effective also with the reference MoS₂. After identical treatments, the ball-milled MoS₂ was much more active in hydrogenation than the reference MoS₂, whereas H₂/D₂ scrambling proceeded more slowly, and cis-trans isomerization not at all. On the basis of earlier conclusions about the site selectivity of these reactions, the activity pattern indicates that mechanochemical activation led to a site structure dominated by sites with multiple vacancies whereas single-vacancy sites were not present at all, which is in agreement with TEM results showing highly defective bent nanoslab structures after ball milling. The results confirm that ethylene hydrogenation, H₂/D₂ scrambling and cis-trans isomerization of cis-but-2-ene may be employed as test reactions for sites on MoS₂ surfaces whereas data for the double-bond shift in 2-methyl-1-butene suggest that the Bronsted sites catalyzing this reaction are related to structural defects rather than to the regular MoS₂ structure.

Full text of DOI:10.1016/j.jcat.2008.10.005

Journal of Catalysis 260 (2008) 236–244
Contents lists available at ScienceDirect
Journal of Catalysis
www.elsevier.com/locate/jcat
Mechanochemical activation of MoS2—Surface properties and catalytic activities in
hydrogenation and isomerization of alkenes and in H2/D2 exchange
Mykola Polyakov a, Sylvio Indris b,1, Stefanie Schwamborn a, Aliaksei Mazheika a,2, Martha Poisot c,3, Lorenz Kienle d,
Wolfgang Bensch c, Martin Muhler a, Wolfgang Grünert a,∗
a
Laboratory of Industrial Chemistry, Ruhr University Bochum, P.O. Box 102148, D-44780 Bochum, Germany
Institute of Physical Chemistry and Electrochemistry, Leibniz University Hannover, Hannover, Germany
Institute of Inorganic Chemistry, Christian Albrechts University Kiel, Kiel, Germany
Institute of Materials Science, Christian Albrechts University Kiel, Kiel, Germany
b
c
d
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 4 June 2008
Revised 9 September 2008
Accepted 7 October 2008
Available online 23 October 2008
High-energy ball milling has been employed to convert a microcrystalline, stoichiometric (S/Mo = 2),
catalytically completely inactive MoS2 (prepared by high-temperature decomposition of ammonium
tetrathiomolybdate, ATM) into an active catalyst, the activity profile of which was studied with test
reactions (ethylene hydrogenation, H2/D2 scrambling, cis–trans isomerization of cis-but-2-ene, double-
bond isomerization of 2-methyl-1-butene). Structural and surface properties of the materials were studied
by XRD, SEM, TEM, XPS, nitrogen physisorption, oxygen chemisorption, and isotope exchange with
D2 (quantity of exchangeable surface hydrogen). The reaction rates obtained after mechanochemical
activation were compared with data from a reference MoS2 made by low-temperature decomposition of
ATM. Ball-milled MoS2 became active for most of the test reactions (except double-bond isomerization)
only after reductive treatments that were effective also with the reference MoS2. After identical
treatments, the ball-milled MoS2 was much more active in hydrogenation than the reference MoS2,
whereas H2/D2 scrambling proceeded more slowly, and cis–trans isomerization not at all. On the basis
of earlier conclusions about the site selectivity of these reactions, the activity pattern indicates that
mechanochemical activation led to a site structure dominated by sites with multiple vacancies whereas
single-vacancy sites were not present at all, which is in agreement with TEM results showing highly
defective bent nanoslab structures after ball milling. The results confirm that ethylene hydrogenation,
H2/D2 scrambling and cis–trans isomerization of cis-but-2-ene may be employed as test reactions for
sites on MoS2 surfaces whereas data for the double-bond shift in 2-methyl-1-butene suggest that the
Brønsted sites catalyzing this reaction are related to structural defects rather than to the regular MoS2
structure.
Keywords:
Molybdenum sulfide
Active sites
Ball milling
TEM
Hydrogenation
H2/D2 scrambling
Double-bond isomerization
© 2008 Elsevier Inc. All rights reserved.
1. Introduction
(hydrogenation, evacuation) or they form under the influence of
the reaction medium, which will produce a defect structure close
The catalytic activity of solid catalysts is supported by sites
which are often conceived to be surface defects. Highly exposed
metal atoms are known to exhibit extraordinary, sometimes un-
desired activities in metal-catalyzed reactions. Vacancies in the
external anion layer are the prerequisite for oxide and sulfide cata-
lysts to show their activities in many reactions. Such coordinatively
unsaturated sites are either produced by thermal pretreatments
to the equilibrium structure under the prevailing conditions.
Mechanochemical activation is an interesting alternative to
thermal routes of active site formation. The defect structure pro-
duced by the treatment (e.g., milling) typically at low temperature
is far from the equilibrium situation. For reactions proceeding un-
der conditions too mild to induce mobility in the surface layer,
catalytic activities and activity patterns in complex reactions may
be expected to deviate strongly from those seen on thermally ac-
tivated catalyst surfaces. Unusual activities may be obtained even
under conditions sufficient to bring surface defect distributions to
equilibrium because these may deviate from distributions on the
surfaces of perfect solids due to bulk structural defects left behind
by the mechanical treatment.
Corresponding author. Fax: +49 234 321 4115.
E-mail address: w.gruenert@techem.rub.de (W. Grünert).
Present address: Forschungszentrum Karlsruhe, Institut für Nanotechnologie,
*
1
P.O. Box 3640, 76021 Karlsruhe, Germany.
2
Present address: Research Institute for Physical Chemical Problems of the Be-
larusian State University, 14 Leningradskaya Street, 220050 Minsk, Belarus.
3
The use of high-energy milling in catalyst preparation and
activation has a long tradition [1], e.g. for synthesis of mixed
Present address: Universidad del Papaloapan, Circuito Central 200, Parque In-
dustrial Tuxtepec 68301, Oaxaca, Mexico.
0021-9517/$ – see front matter © 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2008.10.005

M. Polyakov et al. / Journal of Catalysis 260 (2008) 236–244
237
phases (perovskites [2], Ce–Zr mixed oxides [3], hydrogen stor-
age materials [4]), for the preparation of Ziegler–Natta polymer-
ization catalysts [5], and for the activation of known catalytic
phases, e.g. vanadium phosphates for selective oxidation [6], per-
ovskites for catalytic combustion [7], SiO2-supported MoOx for
selective methane oxidation [8], or TiO2 for photocatalytic hydro-
carbon combustion [9]. Mechanochemical stages have been con-
sidered even for the preparation of catalysts for ammonia [10] and
methanol synthesis [11].
up to 1273 K under oxygen atmosphere. The gases emitted (SO2,
CO/CO2, H2O, NO/NO2) were separated gas-chromatographically
and detected with a thermo-conductivity cell. X-ray powder pat-
terns were recorded with a Phillips X’Pert diffractometer (MoS2,
m-MoS2) or with a STOE STADI-P instrument (LT-MoS2) using
monochromatized CuK
tron microscopy images were obtained with a Phillips ESEM XL30
Scanning electron microscope.
radiation (λ = 1.54056 Å). Scanning elec-
α1
Transmission electron microscopy studies were performed with
Less work has been done with sulfide materials. Kuriki et al.
activated crystalline MoS2 in different mills and characterized the
products by XRD, TEM and a sedimentation technique (for par-
ticle size distribution) [12]. They found the activity in 1-meth-
ylnaphthalene hydrogenation drastically increased. According to
Kouzo et al., the activity (on an equivalent weight of Mo basis) of
mechanochemically activated MoS2 in dibenzothiophene dehydro-
desulfurization exceeded that of Al2O3-supported MoS2 and com-
peted even with an industrial Co–Mo/Al2O3 catalyst [13]. A strong
preference of the hydrogenative pathway over the direct desulfu-
rization route in the milled MoS2 indicated differences in the site
distribution between supported and mechanically activated MoS2.
In the present paper, a completely inactive microcrystalline MoS2
was activated by high-energy ball milling and characterized by
different methods, including TEM. Its catalytic activities in test
reactions known to proceed on different sites (ethylene hydro-
genation, cis–trans isomerization of cis-but-2-ene, H2/D2 isotope
exchange [14–16], and double-bond isomerization of 2-methyl-1-
butene) were compared with those found over a nano-crystalline
MoS2 from a low-temperature preparation. It will be shown that
the mechanochemical treatment creates a site structure favoring
highly exposed Mo sites at the expense of sites with a low num-
ber of vacancies.
a
Philips CM 30ST microscope (LaB6 cathode, 300 kV, CS
=
1.15 mm), to which the samples were inserted on copper/holy
carbon grids. SAED (selected area electron diffraction) and PED
(precession electron diffraction [19–21]) were carried out using
a diaphragm which limited the diffraction to a circular area of
2500 Å in diameter. HRTEM micrographs were evaluated (includ-
ing Fourier filtering) with the programs Digital Micrograph 3.6.1
(Gatan) or Crisp (Calidris).
N2 physisorption isotherms were measured with a Quantachrome
Autosorb-1 MB instrument. Prior to the measurements, the sam-
ples were evacuated at 523 K if not stated otherwise. Pore size
distributions were derived from the desorption branches using the
BJH formalism [22]. For the initial MoS2, the BET surface area was
determined by Kr physisorption. X-ray photoelectron spectra (Mo
3d, S 2p, O 1s, N 1s, and C 1s lines) were recorded with a Sci-
enta/Specs/Prevac Surface analysis system using monochromatized
Al radiation. Samples were transferred from the glove box to the
load lock of the instrument by means of a sample transfer shuttle
(Prevac, with in-house modifications), cf. [18].
Catalytic and related data were measured in a setup character-
ized by facile switching options between batch (cycle) and flow
regimes and between thermoevacuation treatment and reaction
(for more details see [14]). The setup allowed to perform pulse
oxygen chemisorption (OCS) prior to catalysis (reactivation after
OCS by reduction in 10% H2/He at 473 K for 1 h) and to load
or withdraw air-sensitive samples from (or to) a glove box. MoS2
(typically 0.1 g) was investigated after three different activation
treatments: (a) evacuation (rotary pump) at 723 K for 4 h (“V723”),
(b) thermoevacuation with subsequent reduction in H2 (10% in He)
at 573 K for 3 h (“V723/R573”), and (c) reduction in H2 (10% in He)
for 3 h at 573 K without previous thermoevacuation (“R573”). Af-
ter the activation, one or more of the following procedures were
performed:
2. Experimental
2.1. Materials
The initial MoS2 was prepared by a high-temperature decompo-
sition of ammonium tetrathiomolybdate ((NH4)2MoS4, ATM). The
precursor was heated in flowing Ar to 723 K in 15 h, after 24 h
at this temperature further to 1273 K in 22 h, and slowly cooled
down after 60 h at 1273 K (total duration of synthesis—7 days).
1.5 g of this MoS2 were milled in a high-energy ball mill with a
corundum vial and ball (Spex 8000, ball-to-powder weight ratio—
3:1). The milling was performed for 8 h in Ar atmosphere without
additives. The mechanically activated MoS2 (“m-MoS2”) was fur-
ther handled in Ar and stored in a glove box to avoid oxidation
of its surface. A sample of m-MoS2 was suspended in ethanol and
treated in an ultrasonic bath (Sonorex RK 103H of Bandelin Co.,
Germany) at 35 kHz and 100 W for 1 h to break supposed particle
aggregations (“m-MoS2(us)”).
– Oxygen chemisorption (OCS) in pulse mode at 273 K, with mass-
spectrometric detection (Pfeiffer QME 125).
– Determination of exchangeable surface hydrogen (H-Exc); the cat-
alysts were cooled to 473 K in 10% H2/Ar where the carrier
gas was replaced by pure Ar keeping the temperature until
the m/e = 2 QMS signal was constant. After cooling to room
temperature, the reactor was reheated up to 573 K in a slow
flow of D2 (10% in Ar), and the quantity of HD evolved was
determined by calibrated mass spectrometry.
– Catalysis. All reactions were studied in recycle mode. The pro-
cedure has been described in detail in [14]. Briefly, the reaction
cycle was filled with the required reaction mixture while the
catalyst was heated to the desired temperature in a separate
reactor loop under helium. The kinetic run was started by
switching the reactor loop into the cycle, which caused mixing
perturbations only in the first minute. In Table 1, the reactions
studied are listed together with the initial reactant concentra-
tions resulting after mixing of reaction cycle and catalyst loop
contents. It should be noted that all test reactions were per-
formed after activations at temperatures above the reaction
temperature, therefore, changes of the site structure during
reaction were not expected. Moreover, except for H2/D2 scram-
bling the test reactions suffered deactivation by hydrocarbon
The data will be compared with results obtained with
a
MoS2 material prepared via low-temperature decomposition of
(NH4)2MoS4 in inert gas as described earlier in the literature [17].
Detailed information on the preparation of this low-temperature
(LT-)MoS2 (maximum temperature involved—773 K), its surface
properties and catalytic behavior have been published earlier
[14,18]. The data cited in the present paper were almost exclu-
sively measured with a batch labeled MoS2(A1) in [14,18], only in
one case (Fig. 8a), additional results from another batch prepared
under a slightly different temperature regime (MoS2(B) in [14,18],
labeled LT-MoS2* here) will be given.
2.2. Methods
Chemical elemental analyses were made with a EURO Vector EA
Combustion analyzer using 2–3 mg of sample, which were heated

238
M. Polyakov et al. / Journal of Catalysis 260 (2008) 236–244
Table 1
vated MoS2 was extremely rapid, with equilibrium conversions
of nearly 90% achieved already after 4 min in some cases, and
exhibited strong deactivation which prevented measuring pre-
cise reaction rates even in flow mode. The data given below
for this reaction have therefore to be classified as estimates
rather than as accurate data. Nevertheless, due to the large
differences between the materials studied, some relevant con-
clusions could be drawn.
Test reactions and parameters of their experimental realization.
Reaction
Initial concentrations
(vol% in He)
Product
analysis
QMSa
QMSa
GCb
Ethylene hydrogenation
H2/D2 scrambling
Ethylene—2.4, H2—8
H2—4, D2—4
Cis–trans isomerization of
cis-but-2-ene
Cis-butene—0.004, H2—8
GCb
3. Results and discussion
Double-bond isomerization of
2-methyl-1-butenec
2-methyl-1-butene—≈0.1
a
3.1. Physicochemical characterization
Calibrated mass spectrometry.
30 m 0.5 mm GSQ capillary column, 400–410 K.
To 2-methyl-2-butene.
b
c
Elemental analysis showed that mechanical milling caused a
considerable sulfur loss in the MoS2 material. While the initial
MoS2 had a S/Mo ratio of 2.01, this quantity fell to 1.8 after the
mechanochemical activation. For as-synthesized LT-MoS2, a Mo/S
ratio of 2.40 was determined, which fell to 2.04 ... 2.11 after the
thermal activations employed here [18].
The X-ray diffractograms of the different MoS2 samples are com-
pared in Fig. 1. The effect of milling on the crystalline MoS2
is quite obvious: all initially narrow reflections are lowered and
broadened. The effect of the mechanical treatments on the coher-
ence lengths is, however, moderate and LT-MoS2 gives signals with
much larger peak widths. The coherence lengths estimated for the
(002) reflection (stacking direction) via the Scherrer equation are
≈28 nm before and ≈17 nm after milling, whereas the data of
LT-MoS2 result in a primary particle size in the order of 3 nm.
The moderate effects of the mechanochemical treatment may be
explained by the observation that the coherence length was far
below the crystallite size already in the initial MoS2 (30 nm vs.
30 μm), i.e. stacking faults were abundant already in the seem-
ingly perfect crystals (vide infra).
In Fig. 2, scanning electron micrographs of the crystalline MoS2 in
the initial state (a) and after milling (b) are compared with images
of LT-MoS2. The initial MoS2 consists of nice hexagonal platelets,
the diameter of which varies in a broad range from 10 to 50 μm.
After milling, no more regular shapes are left, and irregular chunks
of matter are mixed with chips of μm size. The LT-MoS2 consists of
flat bars evocating of wooden boards (c), with a rough and porous
surface.
Fig. 1. X-ray diffractograms of MoS2 prepared via different routes.
deposition to a certain extent (cf. also [14], and below), hence,
only initial reaction rates could be used to characterize the
surface.
Except for the 2-methyl-1-butene isomerization, kinetic data
for conversions up to 20% were used to evaluate first-order
rate constants [14], including H2/D2 scrambling where a first-
order kinetics has been reported earlier in the literature [23].
2-Methyl-1-butene isomerization on mechanochemically acti-
Fig. 2. Scanning electron micrographs of MoS2 materials employed: (a) (microcrystalline) MoS2, (b) m-MoS2 made by high-energy ball milling of MoS2 (cf. a), (c) and (d)
LT-MoS2, made by low-temperature decomposition of ATM, in different magnifications.

M. Polyakov et al. / Journal of Catalysis 260 (2008) 236–244
239
Fig. 3. TEM micrographs of microcrystalline MoS2: (a) bright-field survey image of a typical microcrystal in [001] zone axis orientation, insert—PED pattern, (b) high-resolution
micrographs and Fourier transforms recorded on different areas of the same microcrystal, left—weak misfit, right—strong misfit.
Fig. 4. TEM micrographs of m-MoS2: (a) high resolution micrograph (with enlarged
cutout) and SAED pattern recorded on nanoslabs, (b) characteristic crystal defects of
nanoslabs.
Fig. 5. TEM micrographs of LT-MoS2: (a) SAED pattern and high resolution micro-
graph recorded on an aggregation of nanoslabs, (b) split nanoslab projecting from
an aggregation of bent nanoslabs.
TEM micrographs of the MoS2 materials studied are presented in
Figs. 3–5. Thin platelets with square dimensions of several microns
and (001) top and bottom surfaces were the predominant mor-
phology observed in the microcrystalline MoS2 (Fig. 3a), nanoslabs
in their preferred [uv0] zone axis orientation were never seen.
Diffraction patterns, when taken on selected areas with uniform
bright field contrast, correlated well with MoS2–2H (cf. insert).
Variations of the bright-field contrast based on the misfit of ro-
tated (001) layers and other planar defects occurred frequently
(stripes in Fig. 3a). In such cases, Moiré contrast was clearly seen
in the high resolution mode, and diffraction patterns and Fourier
transforms exhibited satellites next to the main structure reflec-
tions. These phenomena varied locally inside the same crystallite
as depicted in Fig. 3b for areas with weak (left) and strong (right)
misfit. This explains the above-mentioned strong divergence be-
tween crystallite size and coherence lengths observed by XRD.
In m-MoS2 a majority of aggregated MoS2 nanoslabs was found
together with extended platelets apparently left over from the ini-
tial material. The rotational disorder of the nanoslabs produced

240
M. Polyakov et al. / Journal of Catalysis 260 (2008) 236–244
Fig. 6. Porosity analysis of MoS2 activated by high-energy ball milling, compared with MoS2 prepared by low-temperature decomposition of ATM (LT-MoS2): (a) development
of BET surface area with temperature of thermoevacuation for two batches; (b, c, d) pore-size distributions (BJH, desorption branch) of: LT-MoS2 (b), m-MoS2 (c), and m-MoS2
after ultrasonic treatment (c) after evacuation at different temperatures.
diffuse intensity on concentric circles in the diffraction patterns (cf.
insert to Fig. 4a), which was superimposed by the Bragg reflections
of the platelets. Note that due to the orientation of the nanoslabs
along zone axes [uv0] intensity for reflections 002 (d(002)exp. =
6.17 Å, calculated [24]—6.15 Å) was significantly excited, see arrow
in Fig. 4a. HRTEM micrographs (see enlarged cutout in Fig. 4a, and
Fig. 4b) imaged the consecutive (001) layers as parallel lines with
a distance of 6.15 Å. As a rule, the nanoslabs were strongly bent
and contained dislocations. Additionally, splitting of the nanoslabs
and a stepwise truncations of the (001) layers occurred at the sur-
face, see arrows in Fig. 4b. Such characteristic surface structure
produced sharp edges as also seen at the tips of the nanoslabs,
cf. Fig. 4a.
completely stable under vacuum at 723 K. After ultrasonic treat-
ment in ethanol, the resulting m-MoS2(us) exhibited completely
different properties (Fig. 6d). The enhancement of the BET sur-
face area by the ultrasonic treatment was due to contributions
from a wide range of pore sizes including apparently micropores
(where the BJH formalism gives erroneous results). This unspecific
effect may at least partly have been caused also by variations in
the intra-particle voids, which can contribute to the capillary phe-
nomena. No preferred pore width was generated by the ultrasonic
treatment, and further thermoevacuation created such pore system
only to a minor extent. This suggests that the reactivity of the ma-
terial was significantly changed by the treatment.
XP spectra of MoS2 have been recorded in the initial state and
after milling to check for possible contaminations by the milling
process. The spectrum of the initial MoS2 was shown in [18]
(cf. Fig. 3 there, label “MoS2(cryst)”). It exhibited only signals at-
tributable to the sulfide. Small satellites, which were observed at
the low binding-energy side of all lines, may arise from a differ-
ential charging effect, but do not indicate contaminations. After
milling, no impurity elements were found, and the satellites men-
tioned had disappeared (spectra not shown). In the S 2p region, the
sulfide signal was accompanied by a weak signal at a binding en-
ergy typical of sulfate. This shows that surface oxidation could not
be completely avoided, which is not unexpected for a highly defec-
tive surface. It has been shown in [18] that reduction in hydrogen
at 573 K effectively removes surface sulfate species, therefore the
surface was clean in all experiments that involved such activation.
Oxygen chemisorption capacities and concentrations of exchange-
able hydrogen after the activation treatments are presented in Ta-
ble 2. The initial MoS2 did not adsorb oxygen after any of the
thermal activations applied, and exchangeable hydrogen was not
observed. After the mechanochemical treatment, sites for oxygen
chemisorption were created only by the standard activations. OCS
capacities and amounts of exchangeable hydrogen were lower than
on LT-MoS2. Whereas the data for exchangeable hydrogen seem to
be related between m-MoS2 and LT-MoS2, the former rendering
35–45% of the latter, there was no such relation in their OCS ca-
pacities (65% after V723/R573, 30% after R573). Taking into account
Morphology and real structure of the MoS2 species in sample
LT-MoS2 were strongly related to the nanoslabs observed in m-
MoS2, however, extended crystals were rarely observed. Therefore,
the SAED patterns recorded on aggregates exclusively showed dif-
fuse rings (Fig. 5a) with diameters correlating well with MoS2.
The high resolution micrographs of Fig. 5 depict aggregates of
nanoslabs built from 3–7 (001) layers and with average aspect ra-
tios up to 10. Again, bending and splitting of the nanoslabs was
frequently observed, cf. arrow in enlarged cutout of Fig. 5b. These
structural defects well account for the small coherence lengths
found for LT-MoS2 by XRD whereas the XRD line shapes of m-
MoS2 may be understood as a superposition of narrow and broad
contributions (residual platelets and nanoslabs, respectively).
−1
The BET surface area of MoS2 was as low as 0.3 m2 g
and did
not change significantly upon thermoevacuation at 723 K. After
milling, the BET surface area was considerably larger (7 m2 g−1),
but this value is far below data reported by previous authors after
−1
−1
wet milling (98 m2 g [13], 120 m2 g [12]). Ultrasonic treatment
−1
of m-MoS2 increased its BET surface area to 24 m2 g
(Fig. 6a).
It has been reported in [18] that outgassing of LT-MoS2 at
higher temperatures causes a pronounced enhancement of the
BET surface area by formation of new mesopores, in particular of
sizes around 4 nm. The corresponding data have been included in
Figs. 6a and 6b. The milled m-MoS2 showed a similar behavior
(Fig. 6c) although this sample originated from a material that was

M. Polyakov et al. / Journal of Catalysis 260 (2008) 236–244
241
Table 2
The initial microcrystalline MoS2 was inactive in all reactions
including the acid-catalyzed 2-methyl-1-butene isomerization. The
same is true for the ball-milled MoS2 after ultrasonic treatment as
mentioned above.
Oxygen chemisorption capacity (OCS) and concentration of exchangeable hydrogen
(H-exc) after different activation treatments.
Activation,
sample
Surface area
OCS
(μmol g
H-exc
−1
−1
−1
(μmol g )
(m2
g
)
)
V
723/R573
MoS2
m-MoS2
LT-MoS2
3.2.1. Ethylene hydrogenation
≈0.3a
30a
0
18
28
0
40
115
The mechanochemical treatment resulted in a catalyst that out-
performed LT-MoS2 in ethylene hydrogenation after identical ther-
mal activations. The activation energies found for both catalysts,
however, were identical in the limits of experimental error (Fig. 7a,
Table 3), which suggests that only one type of active site is re-
sponsible for the hydrogenation on all surfaces, and the different
activities might be discussed in terms of site abundance. m-MoS2
activated by V723/R573 could well compete even with the most ac-
tive states found with the low-temperature preparation: As shown
in [14], reversal of thermoevacuation and reduction of LT-MoS2
(i.e., R573/V723) led to a catalyst surface with an almost 25 fold
higher hydrogenation activity, but low activity in cis–trans isomer-
70a,b
R573
MoS2
n.d.
n.d.
21b
0
17
60
0
48
109
m-MoS2
LT-MoS2
a
After V723
From [18].
.
b
that after V723/R573 the BET surface area of m-MoS2 was ca. 40%
that of LT-MoS2 (Fig. 6, data after evacuation at 723 K, cf. Table 2),
one may conclude that after this activation the density of vacan-
cies was larger on m-MoS2 while that of exchangeable hydrogen
was similar on both samples. It should be kept in mind, how-
ever, that opposed to the OCS capacity the amount of exchangeable
hydrogen exhibited extreme variations between different LT-MoS2
batches [18], hence, any conclusion related to this quantity should
be drawn with caution.
After the ultrasonic treatment of m-MoS2, oxygen chemisorp-
tion was not observed either after V723/R573, or after R573. Prob-
ably, the ethanol in which the sample was suspended underwent
strong interactions with the surface which prevented the exposi-
tion of Mo ions by the treatments employed. As a consequence,
the ultrasonic treatment suppressed any catalytic activity as well.
ization of cis-but-2-ene (see also footnote b in Table 3). The rate
−2
constant of m-MoS2 after V723/R573 (3 × 10
l g−1 s−1, at 473 K)
was of the same order of magnitude as the highest seen with LT-
−2
MoS2 (after R573/V723, 7.7×10
l g−1 s−1), and due to the smaller
OCS capacity of the former (cf. Table 1), the “rate per vacancy”
r0/OCS obtained with m-MoS2 after V723/R573 was about 2/3 of
the highest ever seen with LT-MoS2 (see footnote b in Table 3).
Extensive studies on adsorption and reactivity of LT-MoS2 in the
test reactions used here have led us to propose that ethylene hy-
drogenation can be used as a test reaction for Mocus with three
vacancies (3M sites). Analogously, cis–trans isomerization activity
is indicative of 1M sites while H2/D2 scrambling may proceed both
on 2M and 3M sites [14,18]. Given the invariability of the activation
energies in ethylene hydrogenation (Fig. 7a, Table 3), the increased
hydrogenation activity after identical activations suggests a shift
towards higher exposed sites in the Mocus distribution of m-MoS2
as compared with LT-MoS2. This is what should be expected from
the S/Mo ratios, which remained ꢀ2.0 after all activations per-
formed with LT-MoS2 while that of m-MoS2 was 1.8 already before
any thermal treatment. It should be noted that m-MoS2 achieved
reaction rates per vacancy that are comparable but still smaller
than the highest seen with LT-MoS2. This gives rise to the spec-
ulation that the surface of m-MoS2 after V723/R573 might contain
higher exposed Mo sites (4M), which might be inactive for hydro-
3.2. Catalysis
In the following, the catalytic behavior of mechanochemically
activated microcrystalline MoS2 is compared with that of LT-MoS2.
For reactions catalyzed by coordinatively unsaturated ions (Mocus),
which are poisoned by oxygen, more information about the cat-
alytic properties of LT-MoS2 is given in [14]. Our data on the
isomerization of 2-methyl-1-butene, which is an acid-catalyzed re-
action and not poisoned by oxygen [15,16] are published here first
for both catalysts.
Fig. 7. Relation between activation procedure and catalytic activity for ball-milled MoS2(m-MoS2) and LT-MoS2, made by low-temperature decomposition of ATM: (a) ethylene
hydrogenation; (b) H2/D2 scrambling.

242
M. Polyakov et al. / Journal of Catalysis 260 (2008) 236–244
Table 3
Catalytic activity of m-MoS2 and LT-MoS2: comparison of kinetic data.
Reaction
Quantity
LT-MoS2
m-MoS2
R573
V723/R573
R573
V723/R573
Ethylene hydrogenation
103k (l g−1 s−1) at 473 Ka
0.94
1.5
36
3.4b
12b
34
6.7
38
n.d.
30
160
36
102r0/OCS (s−1) at 473 Ka
−1
EA (kJ mol
)
H2/D2 scrambling
103k (l g−1 s−1) at 473 Ka
r0/OCS (s−1) at 473 Ka
18c
19c
1.25
15c
8.4c
1.8c
75c
0.64
0.13
(130)
c
c
0.55
−1
EA (kJ mol
)
17.5c
2-Butene isomerization
2-Methyl-1-butene isomerization
103k (l g−1 s−1) at 473 Ka
4.1
1.1
51
1.3
0.75
54
0
0
–
0
0
–
104r0/OCS (s−1) at 473 Ka
−1
EA (kJ mol
)
103k (l g−1 s−1) at 473 Ka
−1
0.52
49
1.75
57
11
48
EA (kJ mol
)
a
Interpolated from Arrhenius curve.
b
−1
.
Highest ethylene hydrogenation activity achieved by reversal of treatments (R573/V723)—103k (473 K) = 77 l g−1 s−1, 102r0/OCS = 254 s
c
Influenced by mass transport limitations, see text.
genation. The inactivity of 4M sites has earlier been proposed by
Jalowiecki et al. [25].
Whereas the actual reaction kinetics of H2/D2 scrambling over
both types of MoS2 will require additional study, the data given
in Fig. 7b and Table 3 are sufficient in the present context where
only qualitative rankings of activation effects will be discussed.
This ranking is clearly different between LT-MoS2 and m-MoS2:
both activations had approximately the same effect on the for-
mer (which was confirmed in experiments with different degree of
3.2.2. cis–trans isomerization of cis-but-2-ene
The conclusion that the activations used produce predominantly
highly exposed Mocus on the surface of m-MoS2 is further sub-
stantiated by the observations made with cis–trans isomerization.
LT-MoS2 was best activated for this reaction by mild reduction, e.g.,
mere thermoevacuation (V723), whereas an additional subsequent
reduction in hydrogen at 573 K caused the isomerization activity
to decrease [14]. m-MoS2 could not be activated at all for this re-
action neither by R573 nor by V723/R573, no conversion was seen at
473 K in runs of more than 90 min duration. Apparently, no 1M
sites were formed on its surface. This might have been expected
given the low S/Mo ratio of m-MoS2 already before thermal ac-
tivation (1.8), but it has yet to be elucidated if it is due to the
non-equilibrium character of the site structure or would occur also
on a surface where the same stoichiometry would have been cre-
ated thermally.
mass-transport influence [14]) whereas the milder treatment (R573
)
was much more effective with the latter. According to [14], H2/D2
scrambling may occur both on 2M and 3M sites; on this basis, the
fact that the more severe activation (V723/R573) resulted in lower
activity may be explained by an assumption mentioned above: the
formation of inactive (or less active) sites of excessive Mo exposure
(4M). On LT-MoS2, the majority of Mocus appears to be 2M or 3M
irrespective of the activation procedure used, and the different ra-
tio between both sites after the activations explains the differences
in hydrogenation activity at comparable H2/D2 scrambling rates.
On the sulfur-deficient surface of m-MoS2, a loss of active sites ap-
pears to come into play. Apparently, V723/R573 converts 2M into
3M, and 3M into 4M sites, which would explain a loss of scram-
bling activity and a smaller hydrogenation activity than achievable
with LT-MoS2 at optimum activation (vide supra).
3.2.3. H2/D2 scrambling
The case of H2/D2 scrambling (Fig. 7b, Table 3) is somewhat
more complicated because with LT-MoS2 the reaction data ob-
tained under the conditions of the present study were affected by
mass-transport limitations [14]. In a series extending over a wider
3.2.4. Isomerization of 2-methyl-1-butene
The isomerization of 2-methyl-1-butene has been suggested by
Tanaka et al. [16] to be catalyzed by Brønsted sites of MoS2. This
assignment was supported by the observations that this reaction
is neither accelerated by gas-phase hydrogen nor poisoned by oxy-
gen and proceeds on the as-prepared catalyst without any thermal
activation unlike the reactions discussed so far. We have con-
firmed these observations in our own work with different batches
of LT-MoS2 [15].
Beyond these qualitative experiments, a reliable characteriza-
tion of MoS2 with respect to its activity in 2-methyl-1-butene
isomerization appears to be difficult. There was significant de-
activation which clearly affects batch experiments at 373 K (see
runs with LT-MoS2* in Fig. 8a). At lower temperatures, the gas-
chromatographic analysis showed that up to 40% of the hydrocar-
bon dosed into the cycle had been adsorbed on the catalyst surface
at the time of the first analysis (2 min) [15]. As this adsorption was
stronger when the catalysts had undergone thermal activations,
one may conclude that the adsorption occurred mainly on Mocus
sites. However, 2-methyl-1-butene isomerization which requires
acidic sites was also deactivated. With the standard LT-MoS2, we
attempted to stabilize the activity by exposing the catalyst to the
reaction mixture overnight. After this treatment, the catalyst ex-
temperature range (300–480 K) the Arrhenius plot was found to
−1
be curved, with an activation energy of 63 kJ mol
low 373 K instead of 17 kJ mol
found be-
−1
around 470 K [14]. Over m-
MoS2, H2/D2 scrambling was much slower than over LT-MoS2, and
the temperature dependence of the rate constant was larger. The
−1
75 kJ mol
found after the R573 activation are quite near to the
value obtained with LT-MoS2 at lower temperatures. However, op-
posed to the observations with LT-MoS2, the effect of V723/R573
on the scrambling activity was different from that of R573, the re-
sulting activity was smaller and the activation energy even larger,
which can be safely stated despite the low number of data points
−1
available (the activation energy would exceed 75 kJ mol
even
if any one of the three points was omitted). This suggests that
the data measured with m-MoS2 after R573 are still subject to
mass-transfer influences; indeed, the corresponding Arrhenius plot
shows a slight curvature, which is, however, hardly significant at
the present level of experimental accuracy. Thus, despite the large
uncertainties in the activation energy estimated for the V723/R573
activation (130 kJ mol−1), one may conclude that the intrinsic ac-
tivation energy of H2/D2 scrambling over MoS2 is even larger than
−1
75 kJ mol
.

M. Polyakov et al. / Journal of Catalysis 260 (2008) 236–244
243
Fig. 8. Relation between activation procedure and catalytic activity for ball-milled MoS2 (m-MoS2) and LT-MoS2, made by low-temperature decomposition of ATM: (a) iso-
thermal runs (batch recycle reactor) MoS2 of different activation and preparation, (b) temperature dependence of reaction rate measured over m-MoS2 and LT-MoS2 after
different activations. For details about experimental problems and accuracy of data cf. text.
hibited appreciable residual activity, but deactivation could still be
observed. The data presented below were therefore measured with
non-stabilized catalysts (1st or 2nd run) except if specified other-
wise (LT-MoS2, after R573, cf. Fig. 8b).
existence of Mo–H, which would not be a Brønsted site anyway
is questionable in a hydrogen-free atmosphere and at low degrees
of coordinative unsaturation. Therefore, if acidic –SH groups are to
be identified as active sites, some relation between the amount
of exchangeable hydrogen on the MoS2 surface and the isomeriza-
tion activity should exist, but it could not be observed. It has been
shown in [18] that V723 resulted in very low amounts of exchange-
able hydrogen on LT-MoS2, which is plausible for a material that
had not been in contact with hydrogen before and should there-
fore hold also for m-MoS2. Here, however, the highest reaction
rates were observed just after V723. Moreover, after reductive ac-
tivation, the concentration of exchangeable hydrogen was smaller
on m-MoS2 than on LT-MoS2 (Table 2), but the former was more
active.
On the other hand, there was no correlation between 2-methyl-
1-butene isomerization activity and the concentration of Mocus
either. The reaction proceeded on the initial surfaces which have
no oxygen chemisorption capacity at all, and it was not poisoned
by oxygen. Therefore, catalysis by Mo-organic intermediates or by
Lewis sites can be ruled out as well. The acidity reflected in the
test reaction was apparently a Brønsted acidity, and we believe
that it is related to a type of site not present on regular surfaces,
i.e., it arises from structural defects. The poor reproducibility of
the isomerization rate and the particular activity obtained by the
mechanochemical treatment support this view.
Any more detailed description of the site will, however, re-
main speculative on the basis of the present data. Bending of
nanoslabs, which may affect acidity as well, was present both in
m-MoS2 and in LT-MoS2 and can, therefore, not explain the dif-
ferences observed. Impact-related defect structures in MoS2 which
create more rim Mo sites (Fig. 4b) seem to offer a better approach
to explain differing results. A candidate for the active site might
be Mo-OH defect groups because on all surfaces studied by XPS,
O 1s signals with very low intensity could be detected, which
were too weak and unspecific for quantitative studies [15]. The
higher activity of the initial m-MoS2 as compared to the reduced
state (Fig. 8b) may have originated from the presence of sulfate
species (vide supra). The particular success of V723 in the activa-
tion of m-MoS2 for 2-methyl-1-butene isomerization is not easily
explained. A possible hint may come from former studies where
Mo in higher oxidation states (Mo(VI)) was found after thermo-
Fig. 8a shows the development of the 2-methyl-2-butene yield
with reaction time measured over different MoS2 batches. The
two LT-MoS2 batches employed were very different, the reaction
temperatures required to achieve comparable initial reaction rates
differed by 40 K. For LT-MoS2*, runs with different activation treat-
ments are included, and it can be seen that the treatment had no
significant influence on the activity although there may be a differ-
ence in the deactivation rate (see points at 40 min). A third batch
prepared by the same route (labeled MoS2(A2) in [14], data not
included in Fig. 8a) yielded a third activity level between those
shown in Fig. 6a [15]. Hence, 2-methyl-1-butene isomerization pro-
ceeded over LT-MoS2 at rates varying strongly among different
batches. Over m-MoS2, the reaction was more rapid than over any
of the LT-MoS2 batches studied.
In Fig. 8b, the temperature dependence of 2-methyl-1-butene
isomerization is presented for m-MoS2 and LT-MoS2. Due to the
above mentioned experimental problems, the rate constants plot-
ted should be considered as estimates rather than as accurate
kinetic data. With m-MoS2, reductive activation caused a slight
loss of activity compared with the initial state. Surprisingly, mere
thermoevacuation (V723) led to a real boost in activity, which was
quenched when the surface was subsequently reduced (V723/R573).
The difference between the different activations of LT-MoS2 may
arise from the overnight stabilization applied in the series with
R573 activation (vide supra). The data have been included only to
demonstrate the order of magnitude of the activation energy. All
Arrhenius plots except that of m-MoS2 after V723 have approxi-
mately the same gradient. Therefore, given the large experimental
uncertainties in the kinetic experiments with 2-methyl-1-butene
isomerization (vide supra), it can be concluded that the activation
energy of this reaction does not depend on the pretreatment and
is in the order of 50–60 kJ mol−1. The measurements with m-MoS2
after V723 were apparently affected by mass-transfer limitations.
These results cannot be easily reconciled with the notion of
structural Brønsted sites being active centers for the 2-methyl-1-
butene isomerization. According to theoretical studies [26,27], the
only type of hydrogen-containing species on the MoS2 surface are
edge-plane SH groups, basal-plane SH groups being unstable. The

244
M. Polyakov et al. / Journal of Catalysis 260 (2008) 236–244
evacuation of sulfate-containing MoS2 surfaces [18,28]. The higher
oxidation state might increase the strength of the acid defect sites.
used, and it deviated strongly from that known from a reference
MoS2 made by low-temperature decomposition of ATM. After iden-
tical reductive activations, the mechanochemically treated MoS2
was more active than the latter in ethylene hydrogenation and
in 2-methyl-1-butene isomerization. It was less active in H2/D2
scrambling and had no activity at all for the cis–trans isomeriza-
tion of 2-butene, which was clearly observed with the reference
catalyst. This activity pattern implies that the mechanochemical
treatment of microcrystalline MoS2 left the Mo sites exposed with
a larger degree of coordinative unsaturation than on the surface of
the reference material; in particular, sites with only one sulfur va-
cancy were absent. The plausibility of this conclusion supports the
eligibility of the use of some of these reactions (ethylene hydro-
genation, H2D2 scrambling, cis–trans isomerization of 2-butene) to
detect sites with different degree of coordinative unsaturation. Op-
posed to this, the isomerization of 2-methyl-1-butene seems to be
inappropriate for the detection of acidic sites inherent to the MoS2
surface. Instead, our results suggest that it is catalyzed by acidic
sites related to structural defects.
3.3. Outlook
The data presented above demonstrate that the impact of
mechanical energy transformed
a catalytically completely ac-
tive microcrystalline MoS2 into an active hydrogenation catalyst,
which was to be expected on the background of the literature
[12,13]. More remarkable are the strongly different effects of the
mechanochemical treatment on other test reactions: H2/D2 scram-
bling was activated to a much smaller extent, and while the
hydrogenation activity competed with the best performance ob-
tained with a reference catalyst (LT-MoS2), the scrambling activity
remained far below the standard. cis–trans isomerization of cis-
but-2-ene, which was definitely catalyzed by LT-MoS2, did not
occur at all over m-MoS2. This confirms our earlier conclusion
that these reactions use different sites (hydrogenation—3M sites,
H2/D2 scrambling—2M and 3M sites, cis–trans isomerization—1M
sites [14]) and may be employed to trace these sites on unknown
MoS2 surfaces. On this basis, the site structure of m-MoS2 can be
characterized more in detail: It was dominated by 3M sites, with
2M sites (and, possibly, also 4M sites, see above) present as well,
but without 1M sites in significant quantities.
Although it is not possible so far to determine absolute
amounts of these sites due to missing calibration of the reaction
rates with surfaces of known site structure, the comparison of re-
action rates (related to OCS capacities) is a useful new tool for
the characterization of site structures on the surfaces of MoS2 and,
possibly, on other sulfide catalysts as well. The present study has
left several relevant questions to be answered. First, due to the
blockage of coordinatively unsaturated sites, probably by oxygen,
we had to use thermal activations to clean the surface and find
catalytic activity. These procedures might already have changed
the initial site structure. While we employed standard procedures
in this first attempt, further work would have to deal with pos-
sible impacts of this step and to device more conservative ac-
tivations. Further, a comparison between an equilibrium and a
non-equilibrium site structure would require the S/Mo ratios of
both surfaces to be comparable. For our LT-MoS2, however, data
were available only for S/Mo ≈ 2. From the TEM results presented
above, a direct comparison of m-MoS2 and LT-MoS2 would be ap-
pealing: while the size and curvature of the nanoslabs appears
to be similar, the abundance of defects apparently originating from
mechanical impact (cf. Fig. 4b) may differ and cause different activ-
ities just for reactions that require large sites (e.g., hydrogenation).
Earlier studies with LT-MoS2 gave evidence about coalescence and
disappearance of surface defects on LT-MoS2 [14,18]. Future stud-
ies, which are now under way, may show if and under which
conditions such defect migration will lead to an equalization of
differences in equilibrium and non-equilibrium site structures.
Acknowledgments
Financial support by the German Science foundation is grate-
fully acknowledged (grants No. Gr 1447/15 and Be-1653/11). We
would like to thank Ms. Susanne Buse for the physisorption mea-
surements, Mrs. V. Duppel for practical TEM work and Prof. Dr. Dr.
h.c. mult. A. Simon for enabling the TEM examinations.
References
[1] Y.G. Shirokov, Russ. J. Appl. Chem. 70 (1997) 919.
[2] Q.W. Zhang, F. Saito, J. Alloys Compd. 297 (2000) 99.
[3] A. Trovarelli, F. Zamar, J. Llorca, C. de Leitenburg, G. Dolcetti, J.T. Kiss,
J. Catal. 169 (1997) 490.
[4] T. Ichikawa, N. Hanada, S. Isobe, H.Y. Leng, T. Ichikawa, Mater. Trans. 46 (2005)
1.
[5] J.J.A. Dusseault, C.C. Hsu, J. Macromol. Sci.-Rev. Macromol. Chem. Phys. C 33
(1993) 103.
[6] J. Haber, V.A. Zazhigailov, J. Stoch, L.V. Bogutskaya, I.V. Batcherikova, Catal. To-
day 33 (1997) 39.
[7] S. Soiron, L. Aymard, A. Rougier, J.M. Tarascon, Top. Catal. 16 (2001) 391.
[8] O.V. Krylov, A.A. Firsova, A.A. Bobyshev, V.A. Radtsig, D.P. Shashkin, L.Y. Margo-
lis, Catal. Today 13 (1992) 381.
[9] S. Indris, R. Amade, P. Heitjans, M. Finger, A. Haeger, D. Hesse, W. Grünert,
A. Börger, K.D. Becker, J. Phys. Chem. B 109 (2005) 23274.
[10] C.J.H. Jacobsen, J.Z. Jiang, S. Morup, B.S. Clausen, H. Topsoe, Catal. Lett. 61
(1999) 115.
[11] H.L. Castricum, H. Bakker, B. van der Linden, E.K. Poels, J. Phys. Chem. B 105
(2001) 7928.
[12] Y. Kuriki, K. Uchida, E. Sekreta, S. Ohshima, M. Yumura, F. Ikazaki, Fuel Process.
Technol. 59 (1999) 189.
[13] M. Kouzu, K. Uchida, Y. Kuriki, F. Ikazaki, Appl. Catal. A 276 (2004) 241.
[14] M. Polyakov, M. Poisot, W. Bensch, M. Muhler, W. Grünert, J. Catal. 256 (2008)
137.
[15] M. Polyakov, Ph.D. thesis, Ruhr University, Bochum (2007).
[16] K.-I. Tanaka, T. Okuhara, Catal. Rev.-Sci. Eng. 15 (1977) 249.
[17] D.G. Kalthod, S.W. Weller, J. Catal. 95 (1985) 455.
[18] M. Polyakov, M.W.E. van den Berg, T. Hanft, M. Poisot, W. Bensch, M. Muhler,
W. Grünert, J. Catal. 256 (2008) 126.
4. Conclusions
Microcrystalline MoS2 obtained by high-temperature decompo-
sition of (NH4)2MoS4 (ATM), which was completely inactive in
hydrocarbon transformations and H2/D2 scrambling after standard
reductive activation treatments, has been transformed into an ac-
tive catalyst by high-energy ball milling. The treatment, which
caused the creation of bent MoS2 nanoslabs with a multitude of
defects, resulted in a considerable sulfur deficit (S/Mo = 1.8), an in-
crease of the external surface area and the development of oxygen
chemisorption capacity and hydrogen uptake after standard reduc-
tive activations. The activity pattern observed showed a very differ-
ent effect of the mechanochemical treatment on the test reactions
[19] J. Gjonnes, V. Hansen, A. Krerneland, Microsc. Microanal. 10 (2004) 16.
[20] R. Vincent, P.A. Midgley, Ultramicroscopy 53 (1994) 271.
[21] C. Own, Ph.D. thesis, Northwestern University, Evanston, IL, 2005.
[22] E.P. Barrett, L.G. Joyner, P.P. Halenda, J. Am. Chem. Soc. 73 (1951) 373.
[23] R.L. Wilson, C. Kemball, A.K. Galwey, Trans. Faraday Soc. 58 (1962) 583.
[24] B. Schönfeld, J.J. Huang, S.C. Moss, Acta Crystallogr. B 39 (1983) 404.
[25] L. Jalowiecki, A. Aboulaz, S. Kasztelan, J. Grimblot, J.P. Bonelle, J. Catal. 120
(1989) 108.
[26] L.S. Byskov, J.K. Norskov, B.S. Clausen, H. Topsoe, J. Catal. 187 (1999) 109.
[27] L.S. Byskov, M. Bollinger, J.K. Norskov, B.S. Clausen, H. Topsoe, J. Mol. Catal.
A 163 (2000) 117.
[28] F. Maugé, J. Lamotte, N.S. Nestorenko, O. Manoilova, A.A. Tsyganenko, Catal.
Today 70 (2001) 271.