Hydrocarbon reactions on MoS2 revisited, II: Catalytic properties in alkene hydrogenation, cis-trans isomerization, and H2/D2 exchange

Article abstract of DOI:10.1016/j.jcat.2008.03.006

MoS₂ prepared by thermal decomposition of ammonium tetrathiomolybdate in inert gas at temperatures up to 773 K was activated by treatments involving thermoevacuation at 723 K and/or reduction at 573 K. The effect of different activations on the activity in ethylene hydrogenation, cis-trans isomerization of 2-butene, and H₂/D₂ isotope exchange (all measured at temperatures around 473 K) was compared, taking into account the extent of Mo exposition as measured by oxygen chemisorption. It was found that activation had a widely varying impact on the test reactions, indicating that these proceed on sites with different numbers of vacancies. Hydrogenation activity was boosted by two orders of magnitude when reduction of the catalyst at 573 K was followed by thermoevacuation at 723 K, whereas this change was moderate with H₂/D₂ exchange and negative with cis-trans isomerization. For the latter reaction, mere thermoevacuation was a suitable activation, and the impact of subsequent catalyst reduction was negative, whereas appreciable activity in the former reactions was obtained only when thermoevacuation was combined with a subsequent reduction. The data suggest that all three reactions proceeded on different sites, probably 3 vacancies per Mo for olefin hydrogenation, 2 vacancies per Mo for H₂/D₂ exchange, and 1 vacancy per Mo for cis-trans isomerization, with the latter two reactions involving adjacent {single bond}SH groups. Sites with greater Mo exposure than stated may be suitable for H₂/D₂ exchange but not for cis-trans isomerization. With increasing severity of the activation treatments, sites with a small number of vacancies appeared to combine and migrate toward the rims or to escape into the bulk. Therefore, very high hydrogenation activity may be seen on surfaces with an oxygen chemisorption capacity far from the maximum value occurring after less drastic treatments. Our findings imply that a combination of chemisorption studies with test reactions may be a valuable tool for surface characterization of sulfide catalysts.

Full text of DOI:10.1016/j.jcat.2008.03.006

Journal of Catalysis 256 (2008) 137–144
www.elsevier.com/locate/jcat
Hydrocarbon reactions on MoS2 revisited, II: Catalytic properties in alkene
hydrogenation, cis–trans isomerization, and H2/D2 exchange
Mykola Polyakov a, Martha Poisot b, Wolfgang Bensch b, Martin Muhler a, Wolfgang Grünert a,∗
a
Laboratory of Industrial Chemistry, Ruhr University Bochum, D-44780 Bochum, Germany
b
Institute of Inorganic Chemistry, Christian Albrechts University Kiel, Kiel, Germany
Received 3 February 2008; revised 10 March 2008; accepted 10 March 2008
Available online 14 April 2008
Abstract
MoS prepared by thermal decomposition of ammonium tetrathiomolybdate in inert gas at temperatures up to 773 K was activated by treatments
2
involving thermoevacuation at 723 K and/or reduction at 573 K. The effect of different activations on the activity in ethylene hydrogenation, cis–
trans isomerization of 2-butene, and H /D isotope exchange (all measured at temperatures around 473 K) was compared, taking into account
2
2
the extent of Mo exposition as measured by oxygen chemisorption. It was found that activation had a widely varying impact on the test reactions,
indicating that these proceed on sites with different numbers of vacancies. Hydrogenation activity was boosted by two orders of magnitude when
reduction of the catalyst at 573 K was followed by thermoevacuation at 723 K, whereas this change was moderate with H /D exchange and
2
2
negative with cis–trans isomerization. For the latter reaction, mere thermoevacuation was a suitable activation, and the impact of subsequent
catalyst reduction was negative, whereas appreciable activity in the former reactions was obtained only when thermoevacuation was combined
with a subsequent reduction. The data suggest that all three reactions proceeded on different sites, probably 3 vacancies per Mo for olefin
hydrogenation, 2 vacancies per Mo for H /D exchange, and 1 vacancy per Mo for cis–trans isomerization, with the latter two reactions involving
2
2
adjacent –SH groups. Sites with greater Mo exposure than stated may be suitable for H /D exchange but not for cis–trans isomerization. With
2
2
increasing severity of the activation treatments, sites with a small number of vacancies appeared to combine and migrate toward the rims or to
escape into the bulk. Therefore, very high hydrogenation activity may be seen on surfaces with an oxygen chemisorption capacity far from the
maximum value occurring after less drastic treatments. Our findings imply that a combination of chemisorption studies with test reactions may be
a valuable tool for surface characterization of sulfide catalysts.
© 2008 Elsevier Inc. All rights reserved.
Keywords: Molybdenum sulfide; Active sites; Hydrogenation of ethene; Cis–trans isomerization; H /D isotope exchange
2
2
1. Introduction
was studied extensively by Tanaka et al. [5]; some of their ex-
periments are textbook examples today, although some of their
conclusions are coming under increasing question. Although it
is obvious from Tanaka’s work that hydrogenation requires a
higher degree of coordinative unsaturation than cis–trans iso-
merization, Jalowiecki et al. reported that the latter proceeds
on sites both of lower and higher unsaturation than required for
the hydrogenation reaction, but most effectively on the most ex-
posed Mo sites [6]. Another serious issue is the doubt about the
existence of hydrogen directly bound to Mo sites (i.e., Mo–H)
recently raised in theoretical papers [7–12].
The catalytic properties of MoS2 have been studied in great
detail because of MoS2’s role in hydrorefining catalysis. Owing
to this application, most attention has been paid to reactions of
molecules containing S and N in aromatic systems [1–4]. Be-
yond the cleavage of C–S and C–N bonds, MoS2 catalyzes a
number of hydrocarbon reactions as a hydrogenation or cis–
trans rearrangement of double bonds, which are part of the
reaction networks in hydrorefining, and H/D exchange with hy-
drocarbons or between H2 and D2. The latter group of reactions
The present work was guided by the concept of using the
well-studied hydrocarbon reactions as chemical probes for the
surface properties of MoS2 and other chalcogenide phases.
*
Corresponding author. Fax: +49 234 321 4115.
E-mail address: w.gruenert@techem.rub.de (W. Grünert).
0021-9517/$ – see front matter © 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2008.03.006

138
M. Polyakov et al. / Journal of Catalysis 256 (2008) 137–144
We report some relevant results obtained with MoS2 prepared
by low-temperature (ꢀ773 K) decomposition of ammonium
tetrathiomolybdate (ATTM). In Part I of this work [13], we
reported that reductive activation of such materials in hydro-
gen at 573 K (with or without previous thermoevacuation at
773 K) produced appreciable amounts of coordinatively unsat-
urated Mo sites even though the S/Mo ratio did not fall below 2.
With increasing reduction temperature, the extent of coordi-
native unsaturation (characterized by oxygen chemisorption)
passed a maximum, which casts some doubt on attempts to
relate the S/Mo stoichiometry to the MoS2 vacancy structure.
Because the properties of MoS2 batches prepared through the
same synthesis route are not well reproducible, catalytic inves-
tigations should be performed only on batches that have been
well characterized.
Here, in Part II of this work, we report the catalytic prop-
erties of a MoS2 batch that was characterized after different
activations in Part I [13]. We found that olefin hydrogenation,
cis–trans isomerization, and H2/D2 exchange were affected in
different ways by variations in the severity of the activation
treatment and hence the site structure on the surface, suggest-
ing that these reactions require different degrees of Mo expo-
sure.
Fig. 1. Scheme of the catalytic setup allowing rapid switch between pulse and
flow operation, retrieval of sensitive catalyst states into a glove box, and con-
venient switch between activation and reaction modes. Cycle pump and dosing
valve for pulse chemisorption studies omitted.
included in a reaction cycle (gray lines), which can be opened
by valve A to allow operation in flow mode. Valve C allows
separation of the reactor from the cycle. The corresponding part
can be detached at the joints to transfer a sample to a glove box
(or vice versa, to present an air-sensitive catalyst for catalytic
investigation). The six-way valve B serves to switch the reac-
tor between the reaction cycle and the vacuum pump, to ensure
minimum contamination of the sample surface after thermo-
evacuation treatments. This setup also was used to measure
oxygen chemisorption (OCS) in the pulse regime (flow mode;
dosing valve omitted in Fig. 1) or to measure the exchange-
able hydrogen on the catalyst (see Part I [13]). After measuring
OCS, the catalyst was reactivated for subsequent catalytic stud-
ies by reduction (10% H2/He) at 473 K for 1 h. Exploratory
studies [13] have found that this treatment restores the OCS ca-
pacity without creating new adsorption sites.
All reactions were studied in the recycle mode to cover a
wide range of reaction rates, including very low ones. Only
some deactivation tests were performed in the flow mode. The
reaction cycle was first evacuated (with the transfer loop under
helium) and then filled with the required gases, which were cy-
cled while the reactor was heated to the desired temperature.
After switching the transfer loop containing 0.1 g of MoS2 into
the cycle, the kinetic run was started. Mixing effects were seen
in the concentration measurements only within the first minute
because of the good pump performance.
2. Experimental
2.1. Materials
All of the data presented here, with a few exceptions that will
be mentioned in the text, were obtained with one MoS2 batch.
The preparation procedure for this batch, labeled MoS2(A1),
was described in detail in Part I [13]. It is based on the thermal
decomposition of ATTM in flowing nitrogen at temperatures up
to 773 K. This catalyst was activated by different treatments de-
scribed in Part I [13] as well: V723, evacuation at 723 K for 4 h;
V
723/R573, reduction in H2 (10% in He) for 3 h at 573 K af-
ter previous V723; R573, reduction in H2 (10% in He) for 3 h
at 573 K without previous V723. In some cases, thermoevacua-
tion was performed after the reduction step (i.e., R573/V723). In
cases where the activation temperature differed from the stan-
dard value, this temperature was included in the designation
(e.g., V723/R473, indicating reduction only at 473 K).
In Part I [13], we reported that the surface of MoS2(A1)
was initially contaminated by ammonium sulfate species, which
could be completely removed by V723/R573 or R573, whereas
after V723, residual surface Mo(VI) oxide species were ob-
served. The catalytic data reported herein were mostly obtained
on surfaces cleaned by V723/R573 or R573. For data obtained
after V723 or V723/R473, possible effects of surface impurities
also were considered.
For the hydrogenation of ethylene, the cycle was filled
with 100 mbar of an ethylene/He mixture (30% ethylene) and
100 mbar H2 (pure or mixed with D2, 99.7%), subsequently
with He, up to a total pressure of 1 bar. After the transfer loop
was included, initial concentrations of 2.4 vol% ethylene and
8 vol% H2 (H2/D2) were obtained. The reaction products in
the kinetic runs were analyzed continuously by calibrated mass
spectrometry. Some earlier runs designed to study the influ-
ence of D2 on the reaction rate involved gas chromatographic
All gases and gas mixtures used were provided by Messer
GmbH, except for Ar, which was provided by Air Liquide.
2.2. Catalytic measurements
The catalytic data were measured in the setup depicted (in
simplified form) in Fig. 1. The catalytic microflow reactor was

M. Polyakov et al. / Journal of Catalysis 256 (2008) 137–144
139
product analyses. The isomerization of cis-but-2-ene was per-
formed in the presence of H2, because it is known that this
isomerization is extremely slow without hydrogen [5,14]. First,
100 mbar of a mixture containing 500 ppm cis-but-2-ene in He
and 100 mbar of H2 were dosed, resulting in 0.004 vol% butene
and 8 vol% H2 after the addition of He and inclusion of the
transfer loop. The products were analyzed by gas chromatogra-
phy. Analogously, for the study of the H2/D2 exchange, 50 mbar
of H2 and of D2 were dosed, resulting in a concentration of
4 vol% for each. The development of concentrations in the cy-
cle was analyzed by calibrated mass spectrometry. A first-order
reaction rate law, as reported by Wilson et al. [15], was suitable
for representing the data at conversions up to 20%. A series of
experiments also was conducted with a lower initial concentra-
tion [0.8 vol% H2 (D2)].
3. Results
3.1. Hydrogenation of ethylene
(a)
The hydrogenation of ethylene was studied at temperatures
of 440–480 K. For samples activated by V723, a measurable re-
action rate was found only at 450 K. With the other activations,
hydrogenation occurred already at room temperature, but strong
deactivation was noted. This deactivation most likely was re-
lated to the deposition of carbonaceous material, because when
the catalyst was treated with 10% H2/He at 573 K, hydrocar-
bons of different carbon number (including heavier molecules
than ethane) were identified in the desorbing gas by GC. The
kinetic data for conversions <20% could be fitted by a rate law
(pseudo-) first-order in ethylene.
Fig. 2 summarizes the data obtained in an Arrhenius plot,
and Table 1 gives rate constants for T = 473 K, initial reac-
tion rates related to the OCS capacity, and activation energies.
Measurements after V723 were complicated by a parallel acti-
vation of the catalyst due to the presence of hydrogen. The final
state of this activation (i.e., the activity achieved after reduc-
tion of the thermoevacuated surface in 10% H2/He at 473 K for
3 h) are shown in Fig. 2a (dashed line). Because this activation
process competes strongly with the rate measurement, data for
ethylene hydrogenation after V723 are reported only for a single
low temperature, 450 K.
Activation energies were derived for several activation treat-
ments. There appears to be no relationship between pretreat-
ment and activation energy, EA, although the scatter in the data
measured after the R573 treatment casts some doubt on the cor-
responding EA value. But for V723/R473, a similar activation
energy of 41 kJ/mol was obtained. Therefore, it appears safe
to conclude that the activation energy was 35–40 kJ/mol irre-
spective of the pretreatment. This suggests that the same type
of site was operative after the activations, with only the number
of sites changing.
(b)
Fig. 2. Ethylene hydrogenation over MoS . (a) (Pseudo-) first-order rate con-
2
stants measured after different activation treatments. V , evacuation (723 K,
723
4 h); V /R , reduction in H (10% in He, 3 h, 573 K) after previous V ;
723 573
2
723
V
/R , 3 h reduction H (10% in He) at 473 K after V
(after this the
723
723 473
2
surface may contain residual impurities [13]); R , reduction in H (10%
573
2
in He, 3 h, 573 K) without previous V ; R /V , evacuation (723 K,
723
573 723
4 h) after previous R . MoS batch used—MoS (A ) from Part I [13].
573
2
2
1
(b) Batch hydrogenation of ethylene with different isotopic composition of hy-
drogen. T = 463 K, activation by V /R . MoS batch used—MoS (B)
R
723 573
2
2
from Part I [13].
degree of Mo exposure (i.e., OCS capacity) in the latter case.
However, the most striking result is the effect of thermoevacu-
ation subsequent to the reduction treatment (R573/V723). Such
thermoevacuation decreased the degree of coordinative unsat-
uration while simultaneously the activity was boosted by two
Remarkably, the rate constants at 473 K varied by almost
three orders of magnitude depending on the treatment of the
catalyst (Table 1). Fig. 2a shows that reduction after previ-
ous thermoevacuation (V723/R573) led to a more active catalyst
state compared with mere reduction (R573) despite the greater

140
M. Polyakov et al. / Journal of Catalysis 256 (2008) 137–144
Table 1
Table 2
a
a
Ethylene hydrogenation: summary of kinetic data (Fig. 2)
Isomerization of cis-but-2-ene: summary of kinetic data (Fig. 3)
4
a
2
3
a
4
Activation
OCS,
µmol g
10 k at 473 K ,
−1 −1
10 r /OCS at
E
,
Activation
OCS,
µmol g
10 k at 473 K ,
−1 −1
10 r /OCS at
E ,
A
0
A
0
−1
b
−1
−1
−1
b
−1
−1
kJ mol
l g
s
473 K , s
kJ mol
l g
s
473 K , s
V
V
V
R
R
R
6.5
11
28
60
29
42
1.1
5.3
34
9.4
770
91
1.7
4.8
12
1.5
254
21
n.d.
41
34
36
n.d.
n.d.
V
V
V
R
R
6.5
11
28
60
29
2.8
1.0
1.3
4.1
1.3
7
n.d.
48
54
51
n.d.
723
723
/R
/R
/R
/R
1.5
0.75
1.1
0.75
723 473
723 473
723 573
723 573
573
573
/V
/V
573 723
573 723
673
a
MoS batch used—MoS (A ) from Part I [13].
−1
Extrapolated with an activation energy of 52 kJ mol (V , R /V ),
723 573 723
2
2
1
a
b
MoS batch used—MoS (A ) from Part I [13].
Extrapolated with an activation energy of 35 kJ mol
/V ), or interpolated from Arrhenius curves.
2
2
1
b
−1
(V , R
,
673
or interpolated from Arrhenius curve.
723
R
573 723
activity decreased when thermoevacuation was applied after
the reduction treatment (R573/V723), which boosted the hydro-
genation activity dramatically. In cis–trans isomerization, the
activity levels after V723/R573 and R573/V723 treatments were
comparable (Fig. 3).
It is remarkable that the rate constant at 473 K obtained
after different activation treatments differed only slightly (Ta-
ble 2). Indeed, mere thermoevacuation V723 led to an even
higher activity compared with the same V723 treatment fol-
lowed by reduction. Correspondingly, the activity related to the
OCS capacity (i.e., the rater per vacancy) was highest after ther-
moevacuation. Because the initial contaminations on the MoS2
batch used were not completely removed by the V723 treatment,
with Mo(VI) oxide species retained on the surface [13], this ex-
periment was repeated with a different MoS2 batch for which
a clean surface had been demonstrated by XPS (MoS2(A2)
in [13]). The result of the catalytic test is shown in the inset
of Fig. 3. Again, the isomerization activity was higher after
mere thermoevacuation and decreased on subsequent reduction.
The dashed line in the inset is the regression line for the activi-
ties measured after V723/R573 on our standard MoS2 (obtained
from the main panel). Obviously, this was somewhat more ac-
tive than MoS2(A2) (by a factor of ≈2), but the response to the
activation condition was the same and thus was not determined
Fig. 3. Cis–trans isomerization of 2-butene over MoS . First-order rate con-
2
stants measured after different activation treatments. MoS batch used—
2
MoS (A ) from Part I [13]. For explanation of labels see Fig. 2a. Inset: Ex-
2
1
*
periments with a different MoS batch (MoS (A ) from [13]) with initially
2
2
clean surface; dashed line, V /R
regression curve from main panel.
723 573
orders of magnitude. For the same OCS capacity, the rate con-
stant after R573/V723 was as much as 30-fold that obtained
with the reverse order of treatment (V723/R573). The effect
was markedly larger than the change resulting from simply an
increase in reduction temperature (R673), which also led to en-
hanced hydrogenation activity with a diminished OCS capacity.
The ethylene hydrogenation was checked for a kinetic iso-
tope effect by performing the reaction in H2/D2 mixtures of
different H2/D2 ratios (using a MoS2 batch labeled “MoS2(B)”
in Part I [13]). The results of the experiments performed at
463 K are displayed in Fig. 2b. It is obvious that the presence
of D2 did not change the reaction rate, because the curves for
100% D2 and 10% D2 are identical; the slight deviation with
the 1:1 mixture is due to experimental reasons rather than to a
kinetic isotope effect.
by oxide impurities remaining on the standard MoS2 after V723
.
As shown in Fig. 3, the activation energy of cis–trans iso-
merization did not depend on the activation procedure and
amounted to 51–54 kJ mol−1. As for ethylene hydrogenation,
this finding suggests that the isomerization site was the same
on all surfaces.
3.3. H2/D2 isotope exchange
The kinetic data of the H2/D2 isotope exchange are sum-
marized in Fig. 4 and Table 3. During the experiments, which
were conducted mainly at 423–473 K, no deactivation was
noted. The comparison between different activation procedures
is shown in Fig. 4a. The ranking of rates achieved after the ac-
tivations differed from that of the other two reactions. Whereas
mere thermoevacuation rendered the least active surface as in
the case of ethylene hydrogenation, the activities achieved af-
ter reduction at 573 K were virtually identical without or with
previous thermoevacuation (Fig. 4a). Thermoevacuation after
3.2. Isomerization of cis-but-2-ene
Cis–trans isomerization of cis-but-2-ene was studied at 423–
473 K (Fig. 3, Table 2). Deactivation phenomena were weak
under the conditions used here, and a 1-h treatment in 10%
H2/He at 473 K was sufficient to restore standard activity af-
ter each run. For this reaction, the reduction activation (R573
)
yielded the highest catalytic activity, as opposed to ethylene
hydrogenation. Again in strong contrast to hydrogenation, the
R
573, yielding a dramatic activity gain in hydrogenation, re-

M. Polyakov et al. / Journal of Catalysis 256 (2008) 137–144
141
Table 3
a
H /D scrambling: summary of kinetic data (Fig. 4)
2
2
3
a
Activation
OCS,
µmol g
10 k at 473 K ,
−1 −1
r /OCS at
473 K , s
E ,
A
kJ mol
0
−1
b
−1
−1
l g
s
V
V
V
R
R
6.5
11
28
60
29
1.5
16.8
18.8 [15.8]
18.3 [≈16]
34.5 [12]
0.4
2.8
n.d.
31
723
/R
/R
723 473
1.2
0.5
2.1
15
c
723 573
5
5
17.5 [28]
n.d.
573
/V
573 723
Measurements with c (H ) = c (D ) = 4 vol%, in brackets data for c (H ) =
0
2
0
2
0
2
c (D ) = 0.8 vol%. For mass transfer influence on data—see text.
0
a
2
MoS batch used—MoS (A ) from Part I [13].
Extrapolated with an activation energy of 15 kJ mol or interpolated from
2
2
1
b
−1
Arrhenius curves, in brackets data from Fig. 4b, extrapolated with an activation
−1
energy of 28 kJ mol
.
c
−1
63 kJ mol for 303 K < T < 373 K (Fig. 4c).
hibited a clearly greater increase of the rate constant with
temperature and an activation energy of ≈30 kJ mol−1. This
may suggest the presence of mass transport limitations. Be-
cause the standard procedures used to identify such effects (i.e.,
decreased catalyst particle size for pore diffusion limitations,
increased linear gas velocity for external mass transport limi-
tations) could not be readily performed under the present ex-
perimental conditions, the problem was tackled by decreasing
the intrinsic reaction rates; experiments were performed with
lower H2 (D2) pressure (Fig. 4b) or over a very wide temper-
ature range (Fig. 4c). With reduced H2 (D2) pressure, a higher
activation energy of 30 kJ mol−1 indeed was observed (Fig. 4b,
Table 3) in a series performed after R573 activation. A compar-
ison between activation procedures also was made under these
conditions. Whereas the identity of activities achieved by R573
and V723/R573 was fully reproduced, an unexpected behavior
was observed with the R573/V723 sequence; the resulting re-
action rates were once larger, once smaller than those found
after R573 and V723/R573. These differences were moderate but
clearly significant (Figs. 4a, 4b).
(a)
(b)
In the Arrhenius plot measured after R573 with 0.8 vol% H2
(D2) (Fig. 4b), a curvature can be seen; however, this is not sig-
nificant at the given level of experimental scatter. Rate constants
covering a wide temperature range of 303–473 K [activation
R
573, 4 vol% H2 (D2)] are presented in Fig. 4c. The increased
activation energy at low temperatures, and consequently the ef-
fect of mass transport at high temperatures, are quite obvious;
the activation energy in the low-temperature range was on the
(c)
order of 60 kJ mol−1
.
Fig. 4. H /D isotope exchange over MoS . First-order rate constants mea-
sured after different activation treatments. MoS batch used—MoS (A ) from
2
2
2
The uncertainty about the actual activation energy may have
caused some error in the extrapolation of the rate constants to
473 K for the compilation in Table 3. But this error is not large,
because the temperature differences encountered are small.
2
2
1
Part I [13]. (a) c (H ) = c (D ) = 4 vol%; (b) c (H ) = c (D ) = 0.8 vol%;
0
2
0
2
0
2
0
2
(c) study of wide temperature range, c (H ) = c (D ) = 4 vol%. For explana-
0
2
0
2
tion of labels see Fig. 2a.
sulted in only a moderate increase in H2/D2 scrambling activity
compared with V723/R573 and R573
4. Discussion
.
The results for H2/D2 exchange differed from those of the
other reactions also in terms of divergence of activation en-
ergies after different treatments. Whereas rather low activa-
tion energies (15–18 kJ mol−1) were found after the R573 and
The discussion in this section includes the characterization
results reported in Part I [13]. As mentioned in the Introduc-
tion, XPS demonstrated that the surface of the MoS2 batch used
here (MoS2(A1) from Part I [13]) was clean only after acti-
vations including reduction at 573 K (e.g., V723/R573, R573),
V
723/R573 treatments, the data measured after V723/R473 ex-

142
M. Polyakov et al. / Journal of Catalysis 256 (2008) 137–144
whereas after V723, some Mo(VI) oxide remained in the near-
surface region, with sulfur present exclusively as sulfide. After
treatments (R673, R573/V723) despite decreasing OCS capacity
indicates a role of defect diffusion as well. Likely, smaller de-
fects combine to form larger ones and migrate to the rim; other
defects may escape into the bulk crystallite.
Cis–trans isomerization exhibited the least effect of activa-
tion treatment on the rate constant; it varied by a factor of ca. 3
(Table 2), compared with a factor of 700 in ethylene hydrogena-
tion and a factor of 12 in H2/D2 scrambling (Tables 1 and 3).
The response to the reductive activations differed from that with
ethylene hydrogenation; whereas thermoevacuation before re-
duction (V723/R573) favored high hydrogenation activity, this
treatment led to a less active state in cis–trans isomerization.
V
723/R473, the purity of the surface was not checked; how-
ever, the rather small OCS capacity may suggest that residual
oxide species might have interfered with vacancy formation.
However, because all reactions discussed here require coordina-
tively unsaturated Mo sites and do not proceed on nonactivated
MoS2 [14], we include data from the V723 and V723/R473 acti-
vations in the discussion.
The edge planes of MoS2 are capable of exposing Mo ions in
various site structures, which are differentiated by, for example,
the degree of coordinative unsaturation (1, 2, or 3 vacancies per
Mo—1M, 2M, or 3M sites). It has been proposed that the reac-
tions studied here require sites of different Mo exposure, based
on product distributions in tracer experiments combined with
plausibility arguments [5]. According to these studies, olefin
hydrogenation proceeds on 3M sites, H2/D2 exchange requires
3MH sites (i.e., 3M sites with one vacancy already occupied by
adsorbed H), and cis–trans isomerization requires 2MH sites.
Reactions that use a different number of vacancies per Mo
also should be distinguishable by their response to the variation
of the degree of Mo exposure. This approach was used in a pre-
vious study [6], where the degree of Mo exposure was assessed
by model calculations on the basis of the experimental S/Mo
stoichiometry, and it was concluded that olefin hydrogenation
requires triply unsaturated Mo ions (3M), whereas (cis–trans)
isomerization proceeds on twofold or fourfold unsaturated Mo.
Our studies followed the same approach. Because of signif-
icant oxygen chemisorption and catalytic activity with MoS2
that is not sulfur-deficient (Part I [13]), we believe that oxy-
gen chemisorption cannot be disregarded as a measure of co-
ordinative unsaturation. The problems caused by this are illus-
trated by comparing hydrogenation activities after V723/R573
and R573/V723 (Fig. 2, Table 1). It is obvious that both acti-
vations led to completely different site structures, even though
the overall degree of Mo exposure (as measured by OCS) was
identical. A detailed explanation for the drastic difference in
the site structure after the treatments requires additional inves-
tigation; however, the pronounced differences in the catalytic
activity patterns after the activations suggest some conclusions
about the site structures and the active-site requirements of the
test reactions.
The most remarkable feature was the high activity after V723
,
which produced inferior activities in the other reactions. For the
cis–trans isomerization, the rate constant after V723 was well
within the range observed for the other activation treatments.
The rate decreased with subsequent catalyst reduction; that is,
the new vacancies formed hindered the isomerization reaction
(Table 2).
The kinetic data for H2/D2 scrambling should be considered
carefully, because most of the reaction rates reported in Fig. 4
and Table 3 were affected by mass transport limitations. Un-
fortunately, the well-characterized MoS2(A1) batch was com-
pleted after the mass transport influence was clearly identified
(Fig. 4c), thus precluding a comparison of activations on low-
temperature kinetic measurements. Nevertheless, we can draw
some important qualitative conclusions based on the existing
data.
Based on the data in Fig. 4 and Table 3, we may infer that
the measurements performed around 470 K at the higher H2/D2
concentration (Fig. 4a) were in the transition region between the
kinetic and the fully mass-transfer controlled regimes, where
the activation energy drops to 4–6 kJ mol−1 [17]. This indicates
that the surface reaction contributed significantly to the exper-
imental rates even under these conditions, and much more so
at the lower initial H2/D2 concentration (Fig. 4b). As long as
a quantitative comparison is avoided, these data can be used to
rank the effects of activation procedures, because experimen-
tal conditions affecting mass-transfer rates (e.g., gas velocities,
particle sizes, bed dimensions) were kept constant during the
measurements.
Therefore, we can safely conclude that the intrinsic H2/D2
scrambling activities achieved by the V723/R573 and R573 acti-
vations were very close; indeed, their similarity was confirmed
under conditions of different mass-transfer impact (Figs. 4a,
4b). This marks a clear difference from hydrogenation, in which
Obviously, olefin hydrogenation was the reaction with the
largest space requirement at the Mo site, with a rate constant ex-
hibiting the most dramatic variation among the reactions stud-
ied, even without considering the R573/V723 activation. There-
3
fore, it is plausible to identify the hydrogenation site as a M
activation by V723/R573 was much more efficient than by R573,
center, as has been proposed earlier [5,6] and as also suggested
by the rim site concept [16]. According to this concept, Mo
ions capable of acquiring the required coordinative unsatura-
tion are located at the rims of MoS2 stacks. When the reaction
rates measured in this study are related to the OCS capacity,
this “turnover frequency” varies by two orders of magnitude.
Because the invariability of the activation energy suggests the
presence of only one type of active site, this indicates that other
sites counted by OCS also may be converted into hydrogenation
sites. The increased hydrogenation activity with the severity of
as well as from cis–trans isomerization, where the reverse was
found. The influence of a reduction subsequent to V723 fur-
ther differentiated H2/D2 scrambling from cis–trans isomeriza-
tion; the former was accelerated by R573 subsequent to V723
,
whereas the latter was attenuated. The data related to the com-
parison between the V723/R573 and R573/V723 sequences are
less clear-cut, with a slight preference for the latter at 4 vol%
H2 (D2) but for the former at 0.8 vol% H2 (D2) (Figs. 4a, 4b).
Although we propose a possible reason for this behavior later,
these findings clearly demonstrate that these procedures had a

M. Polyakov et al. / Journal of Catalysis 256 (2008) 137–144
143
Scheme 1. Cis–trans isomerization of 2-butene.
Scheme 2. H /D exchange, half-cycle. (The analogous reaction with H restores the initial state.)
2
2
2
similar, not different effect on H2/D2 scrambling activity. In
summary, it can be stated that the response of H2/D2 scram-
bling to the surface modification introduced by the different
activation treatments was between that seen for ethylene hy-
drogenation and cis–trans isomerization.
The explanation for this effect is based on the observation that
two-thirds of adsorbed exchangeable hydrogen was desorbed
while a sample activated in H2 at 573 K was cooled to room
temperature in inert gas (Part I [13], Fig. 6). We may assume
that SH groups adjacent to vacancies were more reactive and
thus belonged to those that were removed during this desorp-
tion process, so that isomerization required the presence of hy-
drogen to restore them. Another important observation of the
Tanaka group is that (in extrapolation to low conversions) D
atoms were not introduced into the butene molecule when cis–
trans isomerization was performed in D2. On a catalyst treated
in hydrogen before the reaction (as necessary to create the Mo–
H groups required in the earlier mechanism), SH groups also
The sometimes drastic differences in the response of the test
reactions to the activation treatments suggest that the differ-
ences between the active sites proposed in the earlier litera-
3
3
ture (e.g., M for hydrogenation, MH for H2/D2 scrambling,
2MH for cis–trans isomerization [15]) may be too subtle to
explain the experimental data obtained here. The formation of
sites with greater sulfur deficiencies requires more drastic con-
ditions. Thus, the tendencies for the formation of 3MH sites and
3M sites should be parallel at least to some extent; however, the
activity patterns in hydrogenation and H2/D2 scrambling were
1
would be available around M sites at the start of the reaction
as long as they were not equilibrated with the D2-containing
gas phase through exchange on adjacent larger Mocus sites and
subsequent spillover.
2
clearly divergent. And although the formation of M as well
2
as of MH sites should be favored by a reduction subsequent
to the mildest activation (V723), in fact the isomerization activ-
ity was suppressed by such a reduction treatment. Indeed, the
elimination of active sites for cis–trans isomerization by a re-
ductive treatment (V723/R573 vs V723; Fig. 3, Table 2) strongly
suggests that this reaction proceeds on sites with less Mo ex-
posure than needed for H2/D2 scrambling, which was boosted
by the same reduction (V723/R573 vs V723; Fig. 4a, Table 3).
Analogously, a thermoevacuation after reduction at 573 K (i.e.,
Whether cis–trans isomerization and H2/D2 exchange also
can proceed on sites with greater degrees of coordinative un-
saturation than required according to Schemes 1 and 2 is rel-
evant aspect of the assignment. The loss of cis–trans isomer-
ization activity on further reduction of the V723-activated sur-
face strongly suggests that the higher exposed Mocus formed
contributed little or no activity. A reactivity pattern with very
high hydrogenation activity and low activity for H2/D2 scram-
bling but also complete absence of cis–trans isomerization was
found in a study with mechanochemically activated microcrys-
talline MoS2 [14]. This observation can only be explained by
a complete inactivity of the highly exposed Mo sites for the
latter reaction. Therefore, we can conclude that cis–trans iso-
merization is specific for the 1M sites. On the other hand, H2/D2
scrambling was little affected by the R573/V723 treatment [en-
hanced or delayed, depending on the H2 (D2) partial pressure],
which in turn produced a drastic increase in hydrogenation ac-
tivity. This suggests that the reaction also may have proceeded
R
573/V723) created new hydrogenation sites (Fig. 2, Table 1)
but destroyed sites for cis–trans isomerization (Fig. 3, Table 2),
whereas it had a moderate and somewhat indistinct effect on
H2/D2 exchange (Figs. 4a and 4b, Table 3). This indicates
that hydrogenation requires Mo sites with higher exposure than
H2/D2 exchange, which is apparently intermediate in its re-
quirement to the degree of unsaturation in the Mo coordination
sphere. Therefore, in agreement with the view that hydrogena-
3
2
tion proceeds on M sites [5,6], we believe that M sites are
1
required for H2/D2 scrambling and that M sites are required
3
for cis–trans isomerization. The hydrogen involved in the lat-
ter reactions might be contributed by neighboring SH (or SD)
groups instead of Mo–H species, which are viewed skeptically
by theoretical chemists [7–12]. Schemes 1 and 2 detail these
reactions.
Of course, these models should explain all of our experi-
mental observations as well as the original models proposed
by Tanaka did. A relevant fact is that cis–trans isomerization
is dramatically suppressed in absence of gas-phase hydrogen
even when the sample was activated in hydrogen previously.
on M sites. Indeed, it would be difficult to understand why a
reaction involving only hydrogen could not proceed in the pres-
ence of an additional vacancy in the Mo coordination sphere,
whereas the more specific adsorption of a hydrocarbon moi-
ety may be perturbed to a structure unfavorable for catalysis
by the excessive space. The conflicting evidence regarding the
effect of R573/V723 on the H2/D2 exchange activities at differ-
ent partial pressures (Fig. 4, Table 3) may indeed arise from
the presence of two types of sites: predominately 3M sites after
R
573/V723 and predominately 2M sites after R573 or V723/R573

144
M. Polyakov et al. / Journal of Catalysis 256 (2008) 137–144
(vide supra), which may respond differently to changes in H2
(D2) partial pressure.
peted well with the other treatments in cis–trans isomerization,
and subsequent reduction deteriorated the catalyst performance
in this reaction.
Another relevant question concerns the origin of the hydro-
gen added to the olefin during ethylene hydrogenation. Given
the picture developed herein, it is not easy to understand why
–SH groups adjacent to Mocus should not participate in eth-
ylene hydrogenation, whereas they are thought to form the
half-hydrogenated state in cis–trans isomerization. On the other
hand, it is known that no d1-ethane is formed on reaction
in H2/D2 mixtures at low conversions [5]; that is, the H/D
atoms added originate from the same H2 (D2) molecule. In
Part I [13], we reported the presence of hydrogenation-active
hydrogen on hydrogen-treated surfaces even after cooling in
inert gas from 573 K. But its quantities determined after dif-
ferent activation procedures showed no correlation with the
hydrogenation activities reported in herein (Table 1). More-
over, comparing the rates of ethane formation from surface
hydrogen (data given in Part I [13]) with the initial rates mea-
sured in the kinetic runs at the same ethane concentrations
demonstrated that the former were much lower than the lat-
ter; with strongly bound surface hydrogen (minimum adsorp-
tion temperature 473–573 K; cf. [13]), reaction rates between
0.01 (V723/R573) and 0.06 µmol g−1 s−1 (R573/V723) were ob-
Our findings imply that all three reactions proceeded on dif-
ferent sites, with the olefin hydrogenation requiring the highest
degree of coordinative unsaturation (i.e., 3 vacancies per Mo).
They further suggest that H2/D2 exchange could proceed on
sites with 2 vacancies per Mo, with adjacent –SH (SD) groups
being involved in the mechanism, and that cis–trans isomer-
ization requires a single vacancy with an adjacent –SH site.
Sites with larger Mo exposure than stated herein may be suit-
able for H2/D2 exchange, but not for cis–trans isomerization.
With increasing severity of the activation treatments, sites with
small numbers of vacancies appear to combine and migrate to-
ward the rims or escape into the bulk. Therefore, very high
hydrogenation activities may be observed on surfaces where the
oxygen chemisorption capacity is far from the maximum value
observed after less vigorous treatments. These conclusions sug-
gest that the combination of chemisorption techniques and test
reactions may be a valuable tool for surface characterization for
the type of material used in our study.
Acknowledgments
tained, compared with rates between 0.9 µmol g−1 s−1 (R573
)
and 75 µmol g−1 s−1 (R573/V723) in the presence of gas-phase
hydrogen. Likely, the reaction with this surface hydrogen is a
side reaction that is not relevant in the presence of gas-phase
hydrogen. The absence of a kinetic isotope effect in olefin hy-
drogenation (Fig. 2b) implies that all reactions involving H (in-
cluding adsorption) are rapid. On the other hand, no hydrogen
chemisorption could be traced in the pulse mode at temper-
atures up to 473 K in Part I [13], demonstrating that strong
adsorbates were not formed at a sufficient rate under these con-
ditions. But this does not mean that there was no interaction
between gas-phase hydrogen and the Mocus sites; on the con-
trary, we believe that the whole of our data related to olefin
hydrogenation and hydrogen adsorption suggest that the olefin
Financial support was provided by the German Science
Foundation (Grants Gr 1447-15 and Be-1653/11). The authors
thank Thomas Hanft for his help with the experiments.
References
[1] R. Prins, A. Egorova, A. Rothlisberger, Y. Zhao, N. Sivasankar, P. Kukula,
Catal. Today 111 (2006) 84.
[2] E. Furimsky, F.E. Massoth, Catal. Rev.-Sci. Eng. 47 (2005) 297.
[3] P. Grange, X. Vanhaeren, Catal. Today 36 (1997) 375.
[4] H. Topsoe, B. Hinnemann, J.K. Norskov, J.V. Lauritsen, F. Besenbacher,
P.L. Hansen, G. Hytoft, R.G. Eheberg, K.G. Knudsen, Catal. Today 107–
108 (2005) 12.
[5] K.-I. Tanaka, T. Okuhara, Catal. Rev.-Sci. Eng. 15 (1977) 249.
[6] L. Jalowiecki, A. Aboulaz, S. Kasztelan, J. Grimblot, J.P. Bonelle,
J. Catal. 120 (1989) 108.
3
was hydrogenated by a short-lived hydrogen adsorbate on M
sites.
[7] L.S. Byskov, M. Bollinger, J.K. Norskov, B.S. Clausen, H. Topsoe, J. Mol.
Catal. A 163 (2000) 117.
[8] L.S. Byskov, J.K. Norskov, B.S. Clausen, H. Topsoe, J. Catal. 187 (1999)
109.
5. Conclusion
[9] S. Cristol, J.F. Paul, E. Payen, D. Bougeard, S. Clemendot, F. Hutschka,
J. Phys. Chem. B 106 (2002) 5659.
[10] S. Cristol, J.F. Paul, E. Payen, D. Bougeard, S. Clemendot, F. Hutschka,
J. Phys. Chem. B 104 (2002) 11220.
[11] V. Alexiev, R. Prins, T. Weber, Phys. Chem. Chem. Phys. 3 (2001) 5326.
[12] M. Sun, A.E. Nelson, J. Adjaye, Catal. Today 105 (2005) 36.
[13] M. Polyakov, M.W.E. van den Berg, T. Hanft, M. Poisot, W. Bensch,
M. Muhler, W. Grünert, J. Catal. 256 (2008) 126 (this issue).
[14] M. Polyakov, Ph.D. thesis, Bochum, 2007.
[15] R.L. Wilson, C. Kemball, A.K. Galwey, Trans. Faraday Soc. 58 (1962)
583.
[16] M. Daage, R.R. Chianelli, J. Catal. 149 (1994) 414.
[17] M. Baerns, H. Hofmann, A. Renken, Chemische Reaktionstechnik, second
ed., Georg Thieme, Stuttgart, 1992, p. 135.
Activation treatments for unsupported MoS2 consisting of
thermoevacuation and hydrogenation steps create widely diver-
gent Mo exposure (as detected by oxygen chemisorption) and
have a varied impact on the catalytic activity in ethylene hy-
drogenation, cis–trans isomerization of 2-butene, and isotope
exchange between H2 and D2. Reduction at 573 K followed by
thermoevacuation at 723 K resulted in the highest hydrogena-
tion activity by far, but demonstrated no particular advantage
over other treatments for H2/D2 exchange and for cis–trans
isomerization. Mere thermoevacuation at 723 K, resulting in
inferior activities for hydrogenation and H2/D2 exchange, com-