Active sites and reactive intermediates in the hydrogenolytic cleavage of C-C bonds in cyclohexane over supported iridium

Article abstract of DOI:10.1016/j.jcat.2012.08.005

Hydrogenolysis of cyclohexane has been explored over supported Ir catalysts. The kinetic data are combined with modeling results to assess the structural requirements and the nature of catalytically relevant surface intermediates for endocyclic C-C bond cleavage. The turnover frequency (TOF) for cyclohexane hydrogenolysis showed complex dependence on Ir particle size, while the selectivity to the primary ring opening product, n-hexane, decreased monotonically with decreasing Ir dispersion. The decreasing TOF as the Ir dispersion decreased from 65% to 52% originates principally from the diminishing abundance of low-coordination Ir atoms at particle surfaces. The increase of the TOF with further Ir particle growth is attributed to an increased fraction of terrace planes, or step sites, and a less unsaturated nature of the most abundant reactive intermediate. Selectivities for multiple C-C bond cleavage, yielding C_<6 alkanes, varies with the relative abundance of coordinatively unsaturated Ir atoms and terrace planes. The multiple hydrogenolysis depends additionally upon H₂ pressure, because single and multiple C-C bond scissions are mediated by surface intermediates with different H-deficiencies.

Full text of DOI:10.1016/j.jcat.2012.08.005

Journal of Catalysis 295 (2012) 133–145
Contents lists available at SciVerse ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Active sites and reactive intermediates in the hydrogenolytic cleavage of C–C
bonds in cyclohexane over supported iridium
⇑
Hui Shi, Xuebing Li, Gary L. Haller, Oliver Y. Gutiérrez, Johannes A. Lercher
Department of Chemistry and Catalysis Research Center, Technische Universität München, Lichtenbergstrasse 4, D-85747 Garching, Germany
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 8 June 2012
Revised 14 July 2012
Accepted 6 August 2012
Available online 23 September 2012
Hydrogenolysis of cyclohexane has been explored over supported Ir catalysts. The kinetic data are com-
bined with modeling results to assess the structural requirements and the nature of catalytically relevant
surface intermediates for endocyclic C–C bond cleavage. The turnover frequency (TOF) for cyclohexane
hydrogenolysis showed complex dependence on Ir particle size, while the selectivity to the primary ring
opening product, n-hexane, decreased monotonically with decreasing Ir dispersion. The decreasing TOF
as the Ir dispersion decreased from 65% to 52% originates principally from the diminishing abundance
of low-coordination Ir atoms at particle surfaces. The increase of the TOF with further Ir particle growth
is attributed to an increased fraction of terrace planes, or step sites, and a less unsaturated nature of the
most abundant reactive intermediate. Selectivities for multiple C–C bond cleavage, yielding C<6 alkanes,
varies with the relative abundance of coordinatively unsaturated Ir atoms and terrace planes. The multi-
ple hydrogenolysis depends additionally upon H2 pressure, because single and multiple C–C bond scis-
sions are mediated by surface intermediates with different H-deficiencies.
Keywords:
Iridium
Cyclohexane
Ring opening
Hydrogenolysis
Structure sensitivity
Ó 2012 Published by Elsevier Inc.
1. Introduction
butanes over Ir(111) and Ir(110)-(1 ꢀ 2) surfaces [12–14]. In con-
trast, little change in the TOF was observed for n-butane hydrogen-
The hydrogenolysis of cycloalkanes holds immense practical
and fundamental relevance. Selective ring opening is an important
process following deep hydrogenation of aromatics during the
upgrading of light cycle oil (LCO) for blending in the diesel pool
[1–3]. From a fundamental standpoint, the active site requirements
and the elementary steps involved in the hydrogenolysis of paraf-
finic C–C bonds remain partly unresolved for acyclic alkanes even
after decades of studies [4], not to mention the much less explored
cycloalkanes in this respect.
olysis on Ir/SiO2 [11], as also for cyclopentane ring opening on Ir/
Al2O3 [15].
Compared to the extensively studied ring opening reaction of
methylcyclopentane (MCP) [19–26], far less fundamental knowl-
edge is available for the direct endocyclic C–C bond cleavage of
the more stable C6-naphthenic structures, which are more abun-
dant in LCO feeds. Due to the low reactivity of the cyclohexyl ring,
bifunctional catalysts, which supply an acid-catalyzed ring con-
traction step to more reactive cyclopentyl backbones, are more of-
The structure sensitivity of hydrogenolysis of C–C bonds in al-
kanes was first demonstrated by Sinfelt et al. for ethane hydrogen-
olysis over Ni–Cu alloys [5]. Since then, alkane hydrogenolysis is
usually assumed to require a large ensemble, though the evidence
arising from particle size effects is actually sparse and contradic-
tory [6–18]. Although iridium is known to possess high hydrogen-
ten employed [1,27–32]. Catalyst optimization in such a
‘‘bifunctional’’ scenario requires both a delicate metal–acid balance
[29,32] and tuning of metal particle size. This complexity hampers
attempts to establish an unequivocal structure–performance rela-
tionship with respect to the metal component.
Pt is effective for ring opening in combination with acidic zeo-
lites [27], but, in the absence of appropriate Brønsted acidity, cat-
alyzes predominantly dehydrogenation and isomerization
reactions of six-membered rings [4]. Monometallic Ru- and Pd-
based catalysts exhibit too high and too low ability, respectively,
to cleave C–C bonds. Non-acidic Ir-based catalysts hold promise
among the monofunctional metal catalysts for achieving selective
ring opening of six-membered naphthenes in a single, direct scis-
sion [1], as long as C–C bond ruptures at unwanted positions and
for multiple times are minimized. In a recent study, Resasco and
coworkers found intriguing trends concerning different endocyclic
olysis activity, the structure sensitivity for C–C
(r-) bond
hydrogenolysis over Ir catalysts is less explored and far from estab-
lished. Foger and Anderson reported ranges of less than three times
in specific activities for ethane hydrogenolysis on Al2O3- and SiO2-
supported Ir catalysts with average particle sizes of 1.5–7.0 nm
[11]. Across a much wider variety of reaction conditions, Engstrom
et al. found structure sensitivity for hydrogenolysis of ethane or
⇑
Corresponding author. Fax: +49 89 28913544.
E-mail address: johannes.lercher@ch.tum.de (J.A. Lercher).
0021-9517/$ - see front matter Ó 2012 Published by Elsevier Inc.
http://dx.doi.org/10.1016/j.jcat.2012.08.005

134
H. Shi et al. / Journal of Catalysis 295 (2012) 133–145
C–C bond cleavage modes of di-substituted cyclohexanes over a
series of monofunctional Ir catalysts on non-acidic or weakly acidic
supports [33]. This study pointed to the possibility of maximizing
C–C bond ruptures at hindered positions, thus raising cetane num-
ber, by tailoring Ir dispersion and support characteristics. Never-
theless, the kinetic features of such systems, especially those
associated with changes in H2 pressure, remain unexplored.
A detailed kinetics study on the hydrogenolysis of cyclohexane,
a model six-membered naphthenic molecule, has been undertaken,
therefore, using Ir particles of different sizes supported on non-
acidic carriers and a wide range of reaction conditions. The cata-
lytic consequences of Ir dispersion and H2 pressure have been ad-
dressed in regard to turnover rates and selectivities to ring
opening, and internal and terminal modes of multiple hydrogenol-
ysis. Mechanistic and structural origins for cyclohexane hydrogen-
olysis on Ir particle surfaces are discussed. To augment previous
knowledge accumulated on bifunctional Ir catalysts, where a deli-
cate acid-metal balance is more important than metal characteris-
tics [29,32,34], the present study explicitly addresses the catalytic
role of metallic Ir in the hydrogenolysis of endocyclic C–C bonds.
determined by ICP-OES, the fractional Ir dispersion obtained from
H2 chemisorption measurements, and a designation of the Cl-con-
taining precursor, respectively. In addition, a sample was prepared
on SiO2 support (Aerosil 200, Evonik Degussa) using (NH4)2IrCl6
and incipient wetness impregnation.
2.2. Catalyst characterizations
2.2.1. H2 chemisorption and N2 physisorption
Weakly and strongly adsorbed fractions of H2 uptakes were
determined volumetrically by a Sorptomatic 1990 instrument. Cat-
alyst samples were pretreated in 100 kPa of static H2 at 723 K for
2–3 h and then evacuated at the same temperature for 0.5 h before
measurements. H2 adsorption was conducted at 307 K in a pres-
sure range of 0.5–13.2 kPa with an equilibrating time of 2–5 min
for each pressure increase step. After completing the first isotherm,
the sample was evacuated to 10ꢁ4 kPa for 1 h and a second iso-
therm measured physisorption. The second isotherm was sub-
tracted from the first one. The amount of chemisorbed hydrogen
was determined by extrapolating the linear part of the difference
isotherm (P > 6.5 kPa) to zero pressure. Dispersions (D), defined
as the fraction of Ir atoms exposed at surfaces, were calculated
by assuming Hstrong/Ir = 2 [36–38]. This ratio was rationalized by
multiple hydrogen adsorption on coordinatively unsaturated sur-
2. Experimental section
2.1. Catalyst preparation
face Ir atoms [38]. The strong H2 adsorption on bare c-Al2O3, pre-
treated in the same way as the catalysts, was negligible.
The porosity of the catalyst samples was measured by N2 phys-
isorption at 77 K on the Sorptomatic 1990 instrument. Prior to the
measurements, samples were outgassed at 523 K for 2 h. The spe-
cific surface area was calculated by applying BET theory [39] to the
adsorption isotherms over a relative pressure range (p/p0) between
0.03 and 0.10, assuming a cross-sectional area of 0.162 nm2 for a
N2 molecule.
2.1.1. Ir/Al2O3 catalysts from Cl-free precursor
A Cl-free precursor, iridium(III) acetylacetonate (Ir(acac)3, Al-
drich, 97%), was dissolved in toluene [35]. Fumed c-alumina (Alu
C, Evonik Degussa) was the support. In a typical procedure of incip-
ient wetness impregnation, aliquots of the solution were added
dropwise to c-Al2O3 while stirring with a quartz rod. The as-pre-
pared sample was dried in synthetic air (Westfalen, 20.5vol.% O2
in N2) at 393 K overnight and then treated in H2 (Westfalen,
99.999%) at 723 K for 5 h. The reduced material was divided into
six portions treated at varying temperatures (473–873 K) for 5 h
in dry synthetic air flow and then again in H2 at 723 K for 4 h. A
temperature increment of 1 K minꢁ1 was applied for all thermal
treatments, whereas flow rates of 100cm3 (STP) minꢁ1 (g-sam-
ple)ꢁ1 were applied for all gases. Catalyst samples are denoted as
m%Ir(D)–1/Al2O3, where m%, D and 1 are the actual Ir loading by
wt.% determined by ICP-OES, the fractional Ir dispersion measured
by H2 chemisorption and a designation of the Cl-free precursor,
respectively.
2.2.2. Transmission electron microscopy
Transmission electron microscope (TEM) images of the materi-
als were recorded with a JEM-2010 JEOL transmission electron
microscope operating at 120 kV. Before being transferred to the
vacuum system, the samples were ground, suspended in ethanol
and ultrasonically dispersed for 2 min. Drops of the dispersions
were applied on a copper grid-supported carbon film.
2.2.3. ICP-OES
The Ir contents of the samples were determined by inductively-
coupled plasma optical emission spectroscopy (ICP-OES). Before
measurement, 60–100 mg of the solid sample was mixed with
hydrofluoric acid (2.0 cm3, 10 wt.%), hydrochloric acid (0.3 cm3,
38 wt.%) and nitric acid (0.1 cm3, 69 wt.%), and boiled until the
solution was clear. The Ir concentration of the solution was mea-
sured on a Spectroflame FTMOA81A ICP-OES spectrometer (Spec-
tro Analytic Instruments). Standard solutions containing 0, 50,
100, and 200 ppm Ir in diluted nitric acid (25 cm3, 6.5 wt.%) were
used for calibration.
2.1.2. Ir/Al2O3 catalysts from Cl-containing precursor
Ammonium hexachloroiridate ((NH4)2IrCl6, Aldrich, 99.99%)
was used as another Ir source to examine the potential effect of
precursor on the catalytic performance. The procedure for incipient
wetness impregnation was the same as for Series-1 catalysts ex-
cept that the precursor was dissolved in doubly-distilled deionized
water. After impregnation, the samples were Parafilm covered at
ambient conditions overnight, and then dried in synthetic air at
393 K overnight and reduced in H2 at 723 K for 5 h. This batch
was then divided into four aliquots, which were treated at different
temperatures (673–823 K) for 5 h in dry synthetic air flow and re-
duced at 723 K for 4 h in pure H2 flow. A high-Ir-loading sample
was prepared by wetness impregnation using the same (NH4)2IrCl6
solution at a fivefold excess volume relative to that in incipient
wetness impregnation. After evaporating the solution in a rotavap
and drying at 393 K for 12 h, the high-loading sample was reduced
at 723 K for 4 h in pure H2 flow without air-calcination. A temper-
ature increment of 1 K minꢁ1 was applied for all thermal treat-
ments, whereas flow rates of 100 cm3 (STP) minꢁ1 (g-sample)ꢁ1
were applied for all gases. Samples thus obtained are denoted as
m%Ir(D)–2/Al2O3, m%, D and 2 are the actual Ir loading by wt.%
2.2.4. XRD
The XRD diffractograms of selected catalysts were recorded with
a Philips X’Pert Pro System (Cu Ka1-radiation, 0.154056 nm) oper-
ating at 45 kV and 40 mA. Measurements were carried out using a
step size of 0.017° (2h) and 1400 s as count time per step.
2.3. Steady-state kinetic measurements
The hydrogenolysis of cyclohexane (CH) was studied in a stain-
less steel packed-bed tubular reactor. Compacted and crushed cat-
alyst powders sieved to 180–280
lm
were diluted with
appropriate amounts of acid-washed quartz sand of identical sieve

H. Shi et al. / Journal of Catalysis 295 (2012) 133–145
135
fractions to ensure sufficient bed lengths for plug-flow hydrody-
namics. A K-type thermocouple was attached to the external sur-
face of the reactor within the isothermal region to measure and
control reaction temperatures. Less than 1 K difference was found
between temperatures measured on the surface of the reactor or
within the catalyst bed. Preliminary tests had confirmed the ab-
sence of transport artifacts (Fig. S1). Prior to reactions, catalysts
were pretreated in a pure H2 flow of 20 cm3 (STP) minꢁ1 (from
room temperature to 623 K, 10 K minꢁ1 ramp and holding for
2 h) and then cooled to the reaction temperature (503–563 K).
CH was introduced using a flow of He (Westfalen, 99.996%) passing
through a series of bubblers (in a thermostatted water bath) which
contained liquid CH (Aldrich, P99%). The He flow carrying CH va-
por saturated at the bath temperature mixed with a flow of H2
(Westfalen, 99.999%) in a co-current mode and contacted the cat-
alyst bed to initiate the reaction. Flow rates of gases and total pres-
sures of the reactor system were modulated by electronic mass
flow controllers (Bronkhorst) and back pressure regulators (Bronk-
horst). The kinetic measurements were conducted under total
pressures of 0.1–2.0 MPa with the H2/CH ratio varying from 20 to
350. The products were analyzed online by a Hewlett–Packard
low-loading (0.50–0.54 wt.% Ir) samples. The Ir/Al2O3 samples pre-
pared from Cl-free precursor featured a sharp loss of Ir dispersion
from 52% to 24% between 623 and 673 K, possibly because of en-
hanced mobility of surface iridium oxide close to the Tammann
temperature of IrO2, 686 K. By contrast, the Ir dispersion of those
prepared from Cl-containing precursor showed a relatively gradual
decrease and began to approach that of the Cl-free counterpart at
high calcination temperatures (>800 K). The presence of chlorine
is responsible for re-dispersion of Ir and Pt in O2-containing atmo-
sphere [40,41].
BET surface areas and Ir dispersions for all catalysts used in the
kinetic studies are listed in Table S1 (Supporting information). Up
to an Ir loading of ca. 2 wt.%, the surface areas of the catalysts were
lower by at most 10% relative to bare Al2O3 support (104 m2 gꢁ1).
Neither air-calcination at 393–873 K nor H2-reduction at 723 K al-
tered to a remarkable extent the BET surface areas of the Ir/Al2O3
samples.
Representative TEM micrographs are presented in Fig. S2. Small
(1–2 nm) Ir particles seemed to be prevalent on high-dispersion
catalysts (Fig. S2a and b). For 1.92%Ir(0.11)/Al2O3 (Fig. S2f), a sur-
face-weighted mean particle size was 11.7 6.9 nm by sampling
210 Ir particles counted from 24 different areas. This value corre-
sponds to D = 0.09 (D = 0.99/d for spherical Ir particles [42]). For
low-loading (0.50–0.54 wt.% Ir) samples, however, a sufficient
number of visible particles were not available for analysis of parti-
cle size distribution. The XRD patterns of selected catalysts
(D = 0.235, 0.11, 0.035) are shown in the Supporting information
(Fig. S3). The average crystallite sizes calculated according to the
Debye–Scherrer equation were in good agreement with the aver-
age particle sizes estimated from TEM and H2 chemisorption mea-
surements. A detailed discussion on the consequence of assumed
H/Irs stoichiometry to the results reported in this study has been
placed in the Supporting information.
6890 Plus GC equipped with a DB Petro column (320
lm ꢀ 100 m)
connected to a flame ionization detector. All lines downstream
from the saturator to the GC were heated to 393 K to ensure no
condensation of either the reactant or the products. Initial rates
and selectivities were reported as values derived by extrapolation
of measured rates and selectivities to zero contact time. There was
no detectable reaction on bare supports or in the absence of the
catalyst. During high pressure experiments, deactivation of low-
dispersion catalysts, to the maximum extent of 15%, was observed
after 5 days below 543 K. In the majority of cases, the deactivation
was kept below 10% during the whole run lasting more than 10 h.
The slow deactivation (ca. 10% after >10 h, see Section 2) sug-
gests that the particle growth must have been not remarkable dur-
ing the reaction. TEM photos were taken for catalysts after
reactions; the result of one spent sample (Fig. S2c, Supporting
information) hardly showed any difference in the particle size
compared to that before reaction. Two spent catalysts (D = 0.16
and 0.52) were collected after several catalytic runs in order to
have a sufficient amount for H2 chemisorption. The results re-
vealed <10% decrease in dispersion for both samples. Therefore,
the Ir/Al2O3 catalysts were stable under elevated H2 pressures,
moderate temperatures, and high H2/HC ratios.
3. Results
3.1. Catalyst characterizations
The effect of calcination temperature on Ir dispersion is shown
in Fig. 1. Increasing the calcination temperature from 393 to 873 K
decreased the Ir dispersion from ca. 60% to lower than 20% for the
Hstrong/Ir = 2
3.2. An overview of reaction channels in cyclohexane conversion
60
Methylcyclopentane (MCP), a potential product from ring con-
traction (RC) of cyclohexane (CH), was observed only in traces
(<0.3% within converted fractions) over all Ir catalysts studied
(including Ir black). This is indicative of a negligible contribution
of indirect ring opening (RO) following RC catalyzed by acid func-
tions. Do et al. also concluded that indirect RO makes little contri-
bution to the overall conversion of dimethylcyclohexane on Al2O3-,
SiO2-, and TiO2-supported Ir catalysts [33]. If MCP were formed and
rapidly ring-opened, methylpentanes would have been found in
much larger quantities compared to n-hexane, due to the well-
known preference of Ir to cleave unsubstituted C–C bonds [19].
Here, n-hexane was >95% within ROPs. This was equally true over
catalysts synthesized with or without Cl. Thus, residual chlorine in
the catalyst, associated either with Al2O3 or with surface Ir atoms,
did not introduce an additional acid-catalyzed pathway for CH con-
version. The reactions detected were isomerization, dehydrogena-
tion, and hydrogenolysis (Scheme 1). Their relative significance
depended on the H2 pressure and temperature, as well as the Ir dis-
persion. The basic thermodynamic and kinetic characteristics of
40
20
0
373
573
773
973
Calcination temperature (K)
Fig. 1. Ir dispersion as a function of calcination temperature in synthetic air. Filled
circles represent the dispersions of low-loading Ir/Al2O3 catalysts prepared from Cl-
free precursor and empty squares correspond to those prepared from
containing precursor.
a Cl-

136
H. Shi et al. / Journal of Catalysis 295 (2012) 133–145
Iso
Iso
Sec-HyL
DeH
RO
+
HyD-iso
+
Multi-HyL
+
CH4
Scheme 1. Reaction network of CH conversion over Ir/Al2O3 catalysts in presence of high H2 pressure. Minor modes of multiple-scissions in primary and secondary
hydrogenolysis are not shown for brevity. DeH = dehydrogenation, RO = ring opening, Iso = isomerization (metal-catalyzed), HyD-iso: Hydroisomerization (acid-catalyzed),
Multi-HyL: multiple hydrogenolysis, Sec-HyL: secondary hydrogenolysis.
the first two minor reaction channels are described in the Support-
ing information, while the hydrogenolytic scission of C–C bonds is
the focus in the sections to follow.
(a)10
8
3.3. Structure sensitivity of cyclohexane hydrogenolysis: rates and
selectivities
6
Cyclohexane turnovers showed >95% selectivity to RO and frag-
mentation at H2 pressures sufficiently high to suppress aromatiza-
4
2
0
tion and isomerization (Section S2, Supporting information). The
hydrogenolysis reactions possess high equilibrium constants (on
the order of 102 for RO, and of 106–108 for fragmentation in the
studied temperature range) and are, therefore, irreversible in light
of their overall remoteness from equilibrium (approach-to-equilib-
rium index [43]:
g g
RO < 10ꢁ3 and frag < 10ꢁ9) at all conditions.
n-Hexane was practically the exclusive ROP (>98% within
C6-isomers) from CH, whereas lighter n-alkanes, arising from mul-
tiple C–C bond cleavages of CH and secondary hydrogenolysis of
re-adsorbed n-hexane, appeared in small quantities depending on
the temperature, H2 pressure, Ir dispersion, and contact time.
0
20
40
60
Ir dispersion (%)
(b) 12
3.3.1. General trend of the activity-dispersion relation
10
8
Fig. 2 displays the mass-specific rates and turnover frequencies
(TOFs) for hydrogenolysis of CH over Ir/Al2O3 catalysts as a func-
tion of Ir dispersion. Fig. S9 shows the CH turnover rates at 543 K
over all studied Ir catalysts, including a purchased Ir black
(D = 0.01) and 1.0%Ir/SiO2 (D = 0.20). At both temperatures, the
turnover rate first declined with decreasing Ir dispersion in a range
corresponding to average particle sizes between 1 and 2 nm, then
increased again when further decreasing Ir dispersion to lower
than 30% (Fig. 2b and S9). Noting the higher mass-specific rate
on Ir(0.16)/Al2O3 than on Ir(0.24)/Al2O3 (Fig. 2a), we exclude the
reason for the variation of TOFs in the low-dispersion range to be
related to an inaccurately determined number of surface Ir atoms.
In marked contrast with the current monofunctional catalytic sys-
tem, the TOFs for tetralin hydroconversion (RO plus RC) over
bifunctional Ir/amorphous silica–alumina catalysts remained
essentially constant in an Ir size range of 1.5–7.8 nm, suggesting
that metal component in these catalysts was not involved in the
rate-controlling step [32].
6
4
2
0
0
20
40
60
Ir dispersion (%)
Fig. 2. Apparent dispersion sensitivities of (a) total-Ir-mass-based rates and (b)
surface-exposed-Ir-based rates for cyclohexane (CH) hydrogenolysis at 523 K,
3.1 kPa CH and 0.56 MPa H2 on 0.50%Ir(D)–1/Al2O3 catalysts.

H. Shi et al. / Journal of Catalysis 295 (2012) 133–145
137
3.3.2. Product selectivities: primary and secondary pathways
100
90
80
70
60
50
(a)
Space velocities were varied to probe the kinetic primary and sec-
ondary nature of different C–C bond cleavage products. The selectiv-
ities to n-hexane and n-pentane/methane extrapolated to zero
conversion (Figs. 3 and S10) indicate that these are primary prod-
ucts, irrespective of the catalyst. In contrast, the selectivity trends
for n-butane, propane, and ethane indicate that they are produced
mainly via secondary reactions of re-adsorbed n-hexane. However,
their formation via multiple hydrogenolysis of CH before desorbing
the primary product n-hexane seems also plausible, in view of their
non-zero selectivities at zero contact time. The selectivities to n-
pentane and methane were insensitive to conversions over catalysts
of low Ir dispersions. This suggests that secondary reactions of re-ad-
sorbed n-hexane proceed mainly via internal C–C bond cleavages
while largely bypassing terminal cleavage pathway.
Clearly, the catalysts that contain primarily small and large Ir
particles substantially differ between each other in terms of the
initial selectivities to n-pentane and methane, pointing to en-
hanced probability of terminal hydrogenolysis (of intermediately
formed n-hexane before desorption) on larger Ir particles. The in-
creased preference for successive terminal scission over lower-dis-
persion catalysts was manifested in the larger disparity of the
initial pentane to methane ratio, for example, 0.6–0.8 (methane
excess).
Ir-1/Al2O3
Ir-2/Al2O3
Ir black
0
20
40
60
Ir dispersion (%)
25
20
15
10
5
(b)
The initial selectivity to n-hexane increased with increasing Ir
dispersion, as shown in Fig. 4a. The selectivity shown on an iridium
(a)
11
C6
80
Ir-1
Ir-2
Ir black
/Al2O3
/Al2O3
9
C5
C3
60
40
20
0
7
5
3
1
0
0
20
40
60
C4
C2
Ir dispersion (%)
Fig. 4. Initial selectivities to (a) n-hexane and (b) n-pentane in hydrogenolysis of
cyclohexane (CH) as a function of Ir dispersion at 523 K, 3.1 kPa CH and 0.56 MPa H2
over Ir/Al2O3 catalysts prepared from Cl-free (h) and Cl-containing (e) precursors
and over commercial Ir black (s).
C1
30
10
14
C6
18
22
26
black (D = 0.01) was the same as that on supported Ir particles with
the largest average size (D = 0.04). By contrast, the initial selectivity
to n-pentane, which is formed via terminal hydrogenolysis of the
ring-opened intermediates before desorption, decreased with
increasing dispersion (Fig. 4b). For both n-hexane and n-pentane,
plateaus were observed toward the low and high ends of the stud-
ied dispersion range. These plateaus suggest the selectivity limits at
given reaction conditions over terrace planes and low-coordination
atomic arrangement, which are the prevalent surface structures on
large and small particles, respectively. Initial selectivities to ethane,
propane, and n-butane that are formed via internal cleavages re-
mained low (ca. 4–6%), and also decreased slightly with increasing
dispersion.
Conversion (%)
(b)
6
4
2
0
60
40
20
0
C1
C3
C4
C5
C2
3.4. H2-pressure dependences of rates, reaction orders and selectivities
over Ir particles of different sizes
2
4
6
8
10
12
Conversion (%)
Fig. 5 depicts the impact of Ir dispersion on the conversion rate
as a function of H2 pressure. The H2 pressure, at which the maximal
rates were observed, as well as the steepness of rate decline after
the optimum pressure, differs significantly between smaller
(D > 0.50) and larger particles (D < 0.25). The activity ratios among
Fig. 3. Product selectivities as a function of conversion for the hydrogenolysis of
cyclohexane (CH) over (a) 0.50%Ir(0.65)–1/Al2O3 and (b) 0.50%Ir(0.16)–1/Al2O3 at
523 K, 3.1 kPa CH and 0.56 MPa H2. The contact time varied from 3.2 to 7.5 s
ðmol-Irs molꢁ1Þ in (a) and from 0.9 to 1.9 s (mol-Irs molCꢁH1) in (b).
CH

138
H. Shi et al. / Journal of Catalysis 295 (2012) 133–145
5
0.8
0.6
0.4
0.2
0
(a)
(a)
6
4
2
0
C6
C5
C4
C3
D = 0.65
D = 0.61
D = 0.58
D = 0.52
4
3
2
1
0
0
3
6
9
12
0
3
6
9
12
H2 partial pressure (102 kPa)
H2 partial pressure (102 kPa)
(b)
8
6
4
2
0
4.0
3.0
2.0
1.0
0.0
(b)
D = 0.235
D = 0.16
D = 0.04
C6
12
8
C5
C4
C3
4
0
0
3
6
9
12
H2 partial pressure (102 kPa)
0
3
6
9
12
H2 partial pressure (102 kPa)
Fig. 5. Turnover rate dependences on H2 pressure over 0.50%Ir(D)–1/Al2O3 catalysts
with fractional dispersions of (a) 0.52–0.65 and (b) 0.035–0.235 at 523 K, 3.1 kPa
cyclohexane (CH).
Fig. 6. Rate dependences on H2 pressure of the ROPs, n-hexane (ꢀ) and fragments,
n-pentane (h), n-butane (4), and n-propane (s) over Ir-1/Al2O3 catalysts with Ir
dispersion of (a) 0.65 and (b) 0.035, respectively. Reaction conditions were 523 K,
3.1 kPa CH. Rates were reported as values obtained after extrapolation to zero
contact time. Formation rates of CH4 and C2H6, revealing similar trends as their
pairing products (C5H12 and C4H10, respectively) with changes in H2 pressure, are
omitted for the sake of clarity.
different catalysts that contain small Ir particles were constant at
any given H2 pressure (Table S2, Supporting information), indicat-
ing the same H2 pressure dependences of TOFs for all of these small
particles. For large Ir particles (Fig. 5b), the H2 pressure for maxi-
mum TOFs shifted higher (0.3–0.6 MPa). Moreover, the H2 pressure
dependence of TOFs became less pronounced compared to those
exhibited by small particles. In addition, the increase in H2 pres-
sure is accompanied by an increase in the reaction order with re-
spect to CH partial pressure (Fig. S11), suggesting a decrease in
the coverage of the CH-derived reactive species by increasing the
chemical potential of H2. It appears that small particles sense this
H2-pressure effect on coverage more strongly than the large
particles.
terminal hydrogenolysis on larger particles and the slightly diver-
gent trends over small and large particles in response to the H2
pressure increase. However, the selectivities to internal multiple
cleavages consistently decrease with increasing H2 pressure,
hardly dependent on the particle size (Fig. 7c).
3.5. Temperature dependences of rates and selectivities over Ir
particles of different sizes
Analyzing formation rates of different products separately in
relation to the particle size and the H2 pressure reveals marked
differences (Fig. 6). Over small Ir particles, the formation rate of
n-hexane exhibits a higher optimum H2 pressure than the maxima
of all other products (Fig. 6a). In contrast, the H2 pressure depen-
dence of n-pentane (and also methane) formation parallels that
of n-hexane over large particles, both showing much higher opti-
mal H2 pressures than the other fragments (C2–C4) formed via
internal multiple-scissions (Fig. 6b). The less pronounced depen-
dence of the rate on H2 pressure for n-hexane than for smaller al-
kanes leads in consequence to its enhanced selectivity at higher H2
pressure on both small and large particles (Fig. 7a). The selectivity
gap greater than 20% between small and large particles persists at
all H2 pressures studied (Fig. 7b), mainly due to the enhanced
Table 1 compiles the kinetic parameters for product formation
routes (kinetic primary) at varying H2 pressures over small and
large Ir particles. The Arrhenius plots of TOFs for CH hydrogenoly-
sis over typical small (D = 0.61) and large (D = 0.16) Ir particles are
presented in Fig. S12. The Ea,app decreased with particle size for
each cleavage mode at the same conditions, but increased on the
same particle size with increasing H2 pressure. The Ea,app for ring
opening to n-hexane were 20–30 kJ molꢁ1 lower than for C5 forma-
tion. Hydrogenolysis of internal C–C bonds had even higher activa-
tion energies than n-hexane and n-pentane formation. As a result,
n-hexane selectivity decreased with increasing reaction tempera-
ture (Fig. S13, Supporting information). The slope of the n-hexane
selectivity with the temperature showed that the selectivity shifts

H. Shi et al. / Journal of Catalysis 295 (2012) 133–145
139
by the greater pre-exponential factors in the respective case. Com-
pensation phenomena are common in the hydrogenolysis of al-
kanes [44]. Pre-exponential factors were similar on sufficiently
large Ir particles (e.g., 107ꢁ108 sꢁ1 for ring opening at 0.56 MPa
H2 on catalysts of D = 0.035–0.30).
100
90
80
70
60
(a)
D > 0.50
4. Discussion
Hydrogenolysis of cyclohexane over these Ir catalysts yields n-
hexane as the sole mechanistically primary product. Some n-hex-
ane formed in this process undergoes multiple hydrogenolysis be-
fore vacating the surface. Consequently, a fraction of the smaller
alkanes formed from these non-desorbed n-hexane molecules are
discerned as apparent primary products due to their non-zero
selectivities at zero contact time (Figs. 3 and S10).
D < 0.20
0
0
0
5
10
15
15
15
4.1. Cleavage patterns of cyclohexane in primary and secondary
pathways
H2 partial pressure (102 kPa)
30
25
20
15
10
5
(b)
Conceptually, as the concentration of available surface hydro-
gen (H⁄) decreases, the ring-opened intermediate state has a higher
chance of experiencing further C–C bond scissions before H⁄-facil-
itated desorption. Thus, reaction variables that deplete the ad-
sorbed hydrogen increase the selectivities to multiple C–C bond
cleavage. The binding energy of H⁄ is higher on step edges than
on terraces [45,46]. Consequently, the steady-state H⁄ coverage
should be higher on small Ir particles with a relatively larger frac-
tion of these defect sites compared to large ones [11,38]. This, in
turn, suggests that the observed lower selectivity to hydrogenoly-
sis of the intermediately formed n-hexane on small Ir particles
(Fig. 4b) is related to higher H⁄ concentrations, in keeping with
the proposal by Paál and coworkers [47]. Alternatively, the proba-
bility of finding neighboring sites to accommodate reactive states
for multiple C–C bond scissions is expected to increase on terrace
planes over large particles, resulting in lower RO selectivities.
Primary and secondary pathways showed different preferences
to terminal and internal cleavage modes for multiple hydrogenol-
ysis (Section 3.3.2). Based on the observed changes of product
selectivities with contact time (Fig. 3), we infer that the active spe-
cies derived from CH first undergoes a single C–C bond scission
leading to adsorbed n-hexane. This intermediate has a certain
probability of undergoing further cleavage of a terminal C–C bond,
as well as a lower probability of hydrogenolysis at internal CII–CII
(bi-secondary) bonds. This low selectivity to hydrogenolysis of
D < 0.20
D > 0.50
0
5
10
H2 partial pressure (102 kPa)
16
12
8
(c)
internal C–C bonds is likely related to the interaction of the C1,6
-
atoms in the proposed structure of the intermediate with the Ir
surface (Scheme 2). Both interacting C-atoms might assume a sor-
bate configuration (probably through bonding at a threefold hol-
low site in a quite rigid sp3 configuration) that renders access of
inner carbon atoms (C2,3,4,5) to neighboring metal atoms highly
improbable. For achieving cleavage of the terminal C–C bond, only
one of the C1,6-atoms needs to be mobile via partial detachment
from the surface (Scheme 2), while both Ir–C bonds need to be-
come flexible for cleavage of an inner C–C bond. It was also ob-
served that the apparent activation energies were significantly
larger for products from internal C–C bond scissions, increasing
progressively with increasing H2 pressure (Table 1). This reflects
the lower coverages of reactive species for multiple C–C bond scis-
sions at internal positions and the greater suppression effects of H2
pressure on their coverage.
D < 0.20
4
D > 0.50
0
5
10
H2 partial pressure (102 kPa)
Fig. 7. Initial selectivities to (a) n-hexane, (b) n-pentane, and (c) lumped C2–C4
products as a function of H2 pressure over small (D = 0.52–0.65) and large Ir
particles (D = 0.014–0.16) at 523 K, 3.1 kPa CH.
On large Ir particles, the probability of terminal C–C bond cleav-
age in kinetic primary pathways is higher than that for hydrogen-
olysis of internal C–C bonds (Fig. 3b), whereas the difference
between two modes is much smaller on small particles (Figs. 3a
and S10a). This might point to the fact that on small particles the
were independent of the particle size. Also shown in Table 1, the
typically larger activation energies for multiple hydrogenolysis
on all particles and for all rates on small particles are compensated

140
H. Shi et al. / Journal of Catalysis 295 (2012) 133–145
Table 1
Kinetic parameters over Ir/Al2O3 catalysts with varying dispersions.
a
b
H2 pressure (102 kPa)
Ea,app (kJ molꢁ1
)
Apparent pre-exponential factor, Aapp (sꢁ1
)
C6
C5
C4
C3
C6
C5
C4
C3
0.50%Ir(0.65)–1/Al2O3
3.3
5.6
11.1
5.6
3.3
5.6
11.1
5.6
11.1
5.6
8.3
5.6
121
154
179
146
116
138
158
95
102
87
95
85
94
147
174
199
170
145
168
180
98
118
105
117
106
116
175
216
–
187
218
–
8.2 ꢀ 1010
1.1 ꢀ 1014
1.1 ꢀ 1016
4.6 ꢀ 1012
2.9 ꢀ 109
2.8 ꢀ 1011
8.3 ꢀ 1012
4.3 ꢀ 107
1.8 ꢀ 108
1.9 ꢀ 107
1.1 ꢀ 108
3.1 ꢀ 107
2.2 ꢀ 108
2.6 ꢀ 1012
5.8 ꢀ 1014
4.3 ꢀ 1016
9.3 ꢀ 1013
1.6 ꢀ 1011
1.7 ꢀ 1013
7.0 ꢀ 1013
3.0 ꢀ 107
2.5 ꢀ 109
5.0 ꢀ 108
7.1 ꢀ 109
1.6 ꢀ 109
1.4 ꢀ 1010
8 ꢀ 1014
4 ꢀ 1018
–
3 ꢀ 1016
1 ꢀ 1019
–
c
c
0.50%Ir(0.52)–1/Al2O3
0.54%Ir(0.46)–2/Al2O3
208
155
197
210
158
208
2 ꢀ 1017
2 ꢀ 1012
1 ꢀ 1016
–
5 ꢀ 1017
5 ꢀ 1012
1 ꢀ 1017
–
c
c
–
–
0.54%Ir(0.30)–2/Al2O3
0.50%Ir(0.16)–1/Al2O3
0.50%Ir(0.035)–1/Al2O3
160
176
157
167
152
169
170
210
163
200
154
185
8 ꢀ 1012
1 ꢀ 1014
3 ꢀ 1012
5 ꢀ 1013
8 ꢀ 1012
2 ꢀ 1014
1 ꢀ 1014
2 ꢀ 1017
6 ꢀ 1012
1 ꢀ 1017
2 ꢀ 1013
1 ꢀ 1016
8.3
a
Reaction conditions were 3.1 kPa CH and 513–553 K; these reported values, with maximum error of 5 kJ molꢁ1 for butanes and propane and 3 kJ molꢁ1 for hexanes and
pentanes, were derived by taking the slopes from semilogarithmic Arrhenius plots of formation rates of different products as a function of reciprocal temperature.
b
The pre-exponential factors were derived from the r (TOF) = Aapp ꢀ exp(ꢁEa,app/RT). The errors depend on the accuracy of values for activation energies. Therefore, the
relative standard deviation is 30–60% for n-hexane and n-pentane and 300–500% for n-butane and propane.
c
These values are >200 kJ molꢁ1 but are likely overestimated due to the inaccuracy of the low extrapolated initial rates at lower temperatures.
Scheme 2. Terminal C–C bond cleavage prevails in chemisorbed n-hexane intermediate formed from cyclohexane ring opening on flat surfaces (C1–C2 bond cleavage as an
example). Carbon atoms at 2–5 positions in the starting molecule cannot access the surface effectively.
sp3 configuration of the ring-opened intermediate is less favored
(lower concentration of threefold hollow sites) than the more flex-
ible bonding in twofold and in atop configurations [48]. As dis-
cussed above, the favorable adsorbed state after the first C–C
bond scission enabling further hydrogenolysis is one in which a
higher flexibility of the metal–carbon bond is assumed.
In stark contrast to kinetic primary pathways, less than 15 mol%
of kinetic secondary C–C bond scission reactions yielded n-pen-
tane/methane on small Ir particles, and only ca. 1–2 mol% on large
particles (Figs. 3 and S10). As the secondary reactions largely by-
passed the cleavage of a terminal C–C bond, the associated ad-
sorbed states of n-hexane (primary products) directly reacting
further and those formed upon its re-adsorption must be different
from each other. This suggests that upon re-adsorption on the me-
tal, the activation of the C–H bond in the n-hexane molecule on the
surface becomes more important in affecting the C–C bond cleav-
age reactivities through the coverages of relevant intermediates.
values (D < 0.20), shifts the optimal H2 pressure to higher values
and leads to less pronounced variations for primary ring opening
(n-hexane) and for the subsequent terminal hydrogenolysis (n-
pentane) on large (average size >5 nm) Ir particles. This is radically
different from the steep decline after the rate maxima on small (1–
2 nm) particles. It appears that small Ir particles respond to the H⁄
coverage variations more than large particles (Fig. S11), possibly
reflecting that both CH-derived reactive species and H⁄ strongly
compete for the low-coordination sites.
Alternatively, the concentrations of the H-deficient reactive
intermediates and the concentration of vicinal H⁄ may dictate the
H2 pressure at which the maximum rate emerges. Conceptually,
an increase in H⁄ coverage will decrease the population of such
deeply dehydrogenated species. Therefore, we reason that the low-
er pressure dependence and the higher optimum H2 pressures on
large particles (Figs. 5b and 6b) point to a less H-deficient nature
of the most abundant reactive intermediates (MARIs) than on small
particles (for a more rigorous discussion, refer to Sections 4.4 and
4.5).
4.2. Implications of H2 pressure effects on product formation rates and
apparent activation energies in cyclohexane hydrogenolysis
The functional relation between Ea,app and reactant pressures
was first derived by Temkin [50] and expressed as follows:
Although the surface concentration of H-adatoms (H⁄) plays a
crucial role in hydrogenolysis [49], a detailed mechanistic analysis
under high H2 pressures has been lacking. In the present work, the
decrease in Ir dispersion, especially from high (D > 0.50) to low
Ea;app ¼ Ea;int
þ
R
n
iDHi
ð1Þ
where ni is the order in the reactant i. The enthalpy change for H2
chemisorption is negative, and the reaction order in H2 is also

H. Shi et al. / Journal of Catalysis 295 (2012) 133–145
141
negative after the optimum H2 pressure. The activation (adsorption
plus dehydrogenation) of cyclic hydrocarbons prior to C–C bond
scission is endothermic (weaker C–M bonds than C–H bonds) when
dissociated hydrogen atoms recombine and desorb as gas phase H2
(i.e., no net effect from M–H bond strength), and the reaction order
in CH is fractional and tends to unity with increasing H2 pressures.
It follows from Eq. (1) that over much of the H2 pressure range,
Ea,app would be greater than Ea,int. For catalysts of higher disper-
sions (D P 0.46), the reaction order in H2 goes more negative at
higher H2 pressures, while the reaction order in CH stays always
positive and increases with H2 pressures, both factors weighing
on the increase in apparent activation energies (Table 1). The
slightly changed Ea,app with increasing H2 pressure for hydrogenol-
ysis of CH over large Ir particles (D < 0.20) is qualitatively attrib-
uted to the lower sensitivity of the H2 reaction order in the
studied pressure range (Figs. 5 and 6 and S14). The markedly lower
apparent activation barriers for primary C–C bond cleavage over
the large Ir particles are attributed to the much smaller reaction or-
ders in H2 (ꢁ0.8 to 0.3) and in CH (0.2–0.5) compared to those
(ꢁ2.5 to ꢁ0.8 and 0.4–0.7, respectively) exhibited by high-disper-
sion catalysts. These variations in reaction orders reflect a lower
coverage of H-adatoms and a higher coverage of CH-derived inter-
mediates on the surface of large particles than on small particles.
The intrinsic barrier for CH hydrogenolysis on large particles was
estimated to be 85 10 kJ molꢁ1, by applying Eq. (1) at conditions
where reaction orders in H2 and CH are both close to zero (e.g., at
523 K, 3.1 kPa CH, 0.5–0.6 MPa H2). Despite that the precision of
the Temkin equation may be questioned in fractional-reaction-or-
der cases, the estimation was attempted also for results on small
particles, which led to Ea,int values typically larger than
100 kJ molꢁ1. In comparison, kinetic modeling of rates at different
temperatures (not shown) suggested that the intrinsic activation
energy is about 100 10 kJ molꢁ1 on small clusters, that is,
10–20% larger than the value for the large particles.
by Bond [4] and considered by some others [21,51,53], with a
few modifications of the original site balance equation (Supporting
information). This model assumes a surface H⁄ acting as the split-
ting agent in the rate-controlling C–C bond cleavage. Here, we con-
sider that the H⁄-assisted RDS may be favored over the non-
assisted C–C bond cleavage, which can be understood by kinetic
coupling [59,60] of C–H bond formation in the transition state with
the C–C bond cleavage that brings down the activation barrier.
The parity plots are shown in Fig. S15 for three mechanisms dif-
fering only in the RDS. Tables S2 and S3 present the four fitted
parameters derived from non-linear regression analysis of mea-
sured CH hydrogenolysis rates at 523 K on Ir/Al2O3 catalysts con-
taining small and large particles, respectively. We are aware that
the mechanistic soundness of all assumptions is to be subjected
to the scrutiny of further experimental studies as well as theoret-
ical assessments. Note, however, that most of the other competing
proposals lack precision in predicting the reaction order in CH
compared to the ‘‘H⁄-assisted C–C bond cleavage’’ model
(Fig. S16), and that the ‘‘vacancy-involved’’ RDS assumption results
in much lower values of the H2 adsorption constant (1.6 ꢀ 10ꢁ5
-
kPaꢁ1) and the estimated coverage of H⁄ (0.04–0.11) than the
‘‘H⁄-assisted’’ RDS assumption.
4.4. Structure sensitivity of cyclohexane hydrogenolysis: the
sympathetic branch
The TOF of structure-sensitive reactions may decrease when the
dispersion increases, which is defined as antipathetic structure
sensitivity; if the TOF increases with increasing dispersion, the
behavior is termed as sympathetic structure sensitivity [61]. There
exist cases where the TOF passes through a maximum [61],
whereas the contribution of Del Angel et al. showing that the
TOF for MCP hydrogenolysis had a minimum with varying Rh dis-
persion [62] is the only case we have found in the literature where
the TOF of a reaction passed through a minimum as a function of
dispersion. In this and the next sections, we separately address
the possible mechanistic and structural origins for the sympathetic
and antipathetic branches of TOF variation with Ir dispersion in CH
hydrogenolysis (Figs. 2b and S9) [63].
As expected from the closely resembled H2 pressure depen-
dences (Table S2, Supporting information) and the nearly equiva-
lent H2 pressure values at rate maxima on small particles
(Fig. 5a), the differing TOFs should primarily stem from the scaling
properties, that is, intrinsic rate constants of the RDS. Indeed, mod-
eling results show that the thermodynamic properties (Kcyc, KH,
and x in Scheme 3) governing the equilibrium concentration and
distribution of surface species do not vary to a substantial degree
on high-dispersion catalysts (D = 0.5–0.7), as is also seen from
the similar estimated coverages (Table S3). More remarkably, the
number of C–H bonds cleaved before C–C bond scission is pre-
dicted to be nearly four (x ꢂ 2 (H2 molecules)) on small particles
of D > 0.50 (Table S3), which corresponds exactly to the widely
proposed C2-unit mode, or the so-called dicarbene structure over
Ir [1,11,33,64].
Now that the observed structure sensitivity has been deprived
of coverage effects, it becomes valid to base the interpretation of
the sympathetic branch simply on the decrease in the RDS rate
constant with increasing particle size. At the present stage, we con-
sider that the population of low-coordination atoms is most conse-
quential to catalysis, owing to their greater coordinative
unsaturation. As first quantified by van Hardeveld and Hartog
[65], the percentage of low-coordination atoms, among all types
of surface atoms, decreases sharply in this narrow range (1–
2 nm). Table 2 shows a quantitative estimation of the population
of different types of surface atoms in closed-shell cuboctahedral
Ir particles. For example, with a slight increase in particle size from
4.3. Mechanistic considerations based on a classical microscopic model
for alkane hydrogenolysis
Before discussing the identity of active sites and catalytically
relevant species that give rise to the observed structure sensitivi-
ties and H2-pressure dependent kinetics for endocyclic C–C bond
cleavage, we would like to first outline some mechanistic consider-
ations. It is well accepted that the non-activated dissociative
chemisorption of H2 and the chemisorption of alkanes both remain
quasi-equilibrated on the time scale of hydrogenolysis turnovers
[4,51–57]. We propose that H⁄ and reactive intermediates compete
for the hydrogenolysis sites, a hypothesis seen valid in MCP and
ethane hydrogenolysis over Pt and Ru catalysts, respectively
[21,51].
A generalized catalytic sequence, involving quasi-equilibrated
steps of H2 and hydrocarbon activation, an irreversible C–C bond
rupture rate-determining step (RDS), and kinetically insignificant
re-hydrogenation and desorption steps, is shown in Scheme 3,
and the derivation of the rate equation for this mechanistic se-
quence is shown in the Supporting information [58]. In a rigorous
sense, the molecularity of the RDS remains elusive in hydrogenol-
ysis reactions. Engstrom et al. suggested that the activated C–C
bond breaks without involving other species or empty sites [12].
Zhuang and Frennet found that better fits in kinetic modeling of
a rate expression were obtained when assuming a gas phase or
weakly adsorbed H2 molecule to be required in the RDS [21]. Shang
and Kenney considered no less than ten rate equations and con-
cluded that the involvement of an empty site in the RDS offered
the best option for modeling their results on ethane hydrogenoly-
sis over Ru/SiO2 [51]. For the moment, let us be concerned with a
more reasonable model for alkane hydrogenolysis as suggested

142
H. Shi et al. / Journal of Catalysis 295 (2012) 133–145
Scheme 3. Sequence of steps (not elementary for quasi-equilibrated steps) proposed for cyclohexane hydrogenolysis on Ir/Al2O3 catalysts.
1.6 to 2.2 nm, the percentage of corner atoms could reduce almost
by a factor of 3. This magnitude is similar to the difference in rate
constants, that is, a factor of 3–4, in this dispersion range
(D = 0.52–0.65).
(Fig. 5). Second, the reaction order in CH is lower than that mea-
sured over small particles (Fig. S11). Third, the temperature depen-
dence of rates (i.e., Ea,app) is lower than in the case of high-
dispersion catalysts (Fig. S12). As argued above, the first feature
suggests a less-dehydrogenated nature of the catalytically relevant
species for C–C bond cleavage on large particles. According to a
Langmuir–Hinshelwood formalism, the second feature points to a
higher equilibrium coverage of CH-derived reactive intermediates
and/or a less severe competition with H⁄ adsorption sites. And
the last feature can be understood from the Temkin relation as a
result of smaller contributions from the exothermicity term for
We anticipate that the activity would stop increasing and start
to drop, with further increasing dispersion (D > 0.65), somewhere
before D reaches unity, as a single surface-bound Ir atom is unli-
kely to accommodate the C–C bond and H-adatoms at the same
time. To clarify this point, it would be highly relevant to study
the catalytic performance of site-isolated mononuclear Ir com-
plexes [66,67] and sub-nanoclusters [68,69] in the hydrogenolysis
of cyclohexane. These well-defined structures have been investi-
gated in alkene hydrogenation [66,67], but have not yet been ex-
plored in hydrogenolysis reactions which are more sensitive to
surface structures and more demanding in reaction conditions.
H2 chemisorption and the endothermicity term for CH chemisorp-
tion. Consequently, the apparent energies of activation on large
particles are close to the intrinsic activation barrier for C–C bond
cleavage. Although it is tempting to ascribe the higher intrinsic rate
constants on very large Ir particles to greater intrinsic pre-expo-
nential factors (transition-state entropy changes), rather than to
smaller activation barriers, the accuracy ( 10 kJ molꢁ1) of deter-
mining true activation barriers on small and large particles renders
a conclusive statement difficult.
4.5. Structure sensitivity of cyclohexane hydrogenolysis: the
antipathetic branch
Large Ir particles employed in this work exhibit the following
kinetic features. First, compared to the high-dispersion catalysts,
the H2 pressures for maximum rates are higher, and the suppres-
sion of activity by H2 pressure after the optimum is less distinct
Compatible with the above qualitative analysis, dehydrogena-
tion depths lower (x = 1.2–1.6) than those on small particles were
predicted by kinetic modeling of TOFs measured on low-dispersion

H. Shi et al. / Journal of Catalysis 295 (2012) 133–145
143
Table 2
Estimated population of four types of surface atoms in a cuboctahedral model for four particle sizes (i.e., shell number: 3, 4, 5 and 6, respectively)a.
Type
CN
Fraction of surface atoms
Ir55 (1.3 nm, D = 0.76)b
Ir147 (1.6 nm, D = 0.63)b
Ir309 (1.9 nm, D = 0.52)b
Ir561 (2.2 nm, D = 0.45)b
Corner
Edge
(100)
(111)
5
7
8
9
0.286
0.571
0.143
0
0.130
0.521
0.261
0.087
0.074
0.444
0.333
0.148
0.048
0.381
0.381
0.190
a
Numerical formulae: Ns (number of surface atoms) = 10 m2 ꢁ 20 m + 12; Ncorner (number of corner atoms) = 12; Nedge (number of edge atoms) = 24(m ꢁ 2); N(100) (number
of (100) terrace atoms) = 6(m ꢁ 2)2; N(111) (number of (111) terrace atoms) = 4(m ꢁ 3)(m ꢁ 2).
b
D (dispersion) = Ns/Nt, d (Ir particle size) ꢂ 0.99/DIr [42].
catalysts when the same mechanistic sequence was applied
(Table S4, Supporting information). The dehydrogenation depth
could be even less pronounced (smaller x), if the adsorption
strength of H⁄ on these large particles had been underestimated
by kinetic modeling. Any attempts to constrain the dehydrogena-
tion depth at x = 2 and model the other three parameters only lead
to unreasonable estimates of H⁄ adsorption equilibrium constant
and H⁄ coverage (Table S5, Supporting information). It is interest-
ing to note that Engstrom et al. [12] explained the differences in
the apparent reaction kinetics and pre-exponential factors of eth-
ane hydrogenolysis on Ir(111) and Ir(110)-(1 ꢀ 2) surfaces also
by different reaction intermediates, namely, C2H4(ads) on the for-
mer and C2H2(ads) on the latter. That is, an extensively dehydroge-
nated species was the MARI on the Ir(110)-(1 ꢀ 2) surface
containing a large faction (25%) of low-coordination atoms, while
a less unsaturated species was the MARI on terrace sites. The
underlying driving forces for various surface structures to effect
different H-deficiencies of MARIs may root in the different degrees
of coordinative unsaturation of surface atoms.
The antipathetic structure sensitivity is ascribed to a less unsat-
urated and more reactive nature of the carbonaceous MARI and
possibly also an adsorption mode that requires a larger ensemble
size than the dicarbene-like adsorption complex prevalent on
small particles. It is possible that an adsorption state parallel to
the surface exists on large Ir particles predominantly terminated
with terrace planes. Although the (111) and (100) planes also ex-
ist on small particles, the large ensemble size requirement for the
adsorption parallel to the surface would make this mode nearly
inoperative on high-dispersion catalysts.
To the best of our knowledge, there has been no literature on
the structure sensitivity of CH hydrogenolysis over supported Ir
catalysts or model Ir surfaces. Surface science studies by Somorjai
and coworkers were one of the few investigations into structure
sensitivity of CH hydrogenolysis on metal surfaces [16–18]. Over
different Pt model surfaces, these authors showed that Pt(111) ter-
races without steps and kinks are the most active for C–C bond
breaking of CH in 15 Torr CH and 100 Torr H2 [16], opposite to
the trend observed by the same group at very low CH
(5 ꢀ 10ꢁ6 Pa) and H2 (10ꢁ4 Pa) pressures where kinks in the steps
appeared the most active [17,18]. These surface science studies
seem to suggest both surface structures to be active for the C–C
bond hydrogenolysis, and that this apparent contradiction can be
resolved by the different reaction conditions applied. While being
intrinsically active for C–C bond cleavage, low-coordination atoms
might have been irreversibly titrated by carbon deposits when CH
coverage was high, as in the former case [16], rendering them inac-
tive in catalytic turnovers. In the present work, these coordinative-
ly unsaturated atoms are still active on small Ir particles, because Ir
possesses much higher hydrogenolysis activity than Pt, and so, is
more resistant to carbon deposition.
4.6. H-deficiency of reactive intermediates as the selectivity mediator
for single-scission and terminal/internal multiple-scissions
We interpret here the effects of H2 pressure on the ring opening
selectivity (Fig. 7) on the basis of the unsaturation degree of sur-
face intermediates. Within the framework of the kinetic model
(Scheme 3), we introduce different H-deficiencies (the value of x)
of MARIs for single and multiple C–C bond scissions (denoted as
SRO and Smulti, respectively) and also for internal and terminal
modes of multiple-scissions (denoted as Stm and Sim, respectively).
According to step (iii) in Scheme 3, the formation rates of single-
scission (RO) product and multiple-scission products along kinetic
primary pathways are given by the rates of the C–C bond cleavage:
rRO ¼ kROhS hH
ð2Þ
RO
rmulti ¼ kmultihS hH
ð3Þ
multi
The quasi-equilibrium assumption for cyclohexane activation
(step ii, Scheme 3) implies that the MARIs for single and multiple
C–C bond cleavage in kinetic primary pathways also exist in
equilibrium:
ð4Þ
where xRO and xmulti refer to the dehydrogenation depths of the
MARIs for single and multiple C–C bond cleavage, respectively,
and, thus, gives:
multiꢁxRO
Þ
hS ¼ Kꢁ1hS PðHx
ð5Þ
int
RO
multi
where Kint is defined as the equilibrium constant for Eq. (4). Thus, a
functional relation (Eq. (6)) can be established between H2 pressure

144
H. Shi et al. / Journal of Catalysis 295 (2012) 133–145
(a) 3.2
(b) 1
2.8
2.4
2
0.8
0.6
0.4
1.6
1.2
0
1
2
3
0
1
2
3
ln (H2 partial pressure,102 kPa)
ln (H2 partial pressure,102 kPa)
Fig. 8. Different H-deficiencies of MARIs for single-scission (RO) and multiple-scissions indicated by the dependences of rate ratios on H2 pressure over (a) small (D > 0.50,
e4) and (b) large (D < 0.20, jN) Ir particles. Reaction conditions: 523 K, 3.1 kPa CH.
and the rate ratio of single and multiple hydrogenolysis (detailed
derivation in Supporting information).
relation (Eq. (1)), leading to higher apparent activation energies
for internal hydrogenolysis (Table 1).
ꢀ
ꢁ
ꢀ
ꢁ
rRO
rmulti
kRO
kmultiKint
5. Conclusions
ln
¼ ln
þ ðxmulti ꢁ xROÞ ln PH
ð6Þ
The present study investigates Ir catalysts of varying disper-
sions, without the presence of a solid acid, that catalyze exclusively
the direct C–C bond cleavages of cyclohexane.
Therefore, the slope of the ln(rRO/rmulti) ꢁ ln(PH) plot points to
the difference in the H-deficiency between surface intermediates
involved in single and multiple C–C bond scissions, respectively.
The double-logarithmic function is plotted in Fig. 8 for both small
and large Ir particles. A comparison of the slopes in Fig. 8a and b
indicates that H-deficiency (x) is less different between SRO and
Smulti on large than on small particles, thus bringing the reactivities
of single and multiple C–C bond scissions closer together (Fig. 6b).
As is shown in a quantitative way (Table 3), the active intermediate
for ring opening possibly has to further lose more than one hydro-
gen atom, on average, to allow multiple hydrogenolysis on small Ir
particles, in accordance with a previous report on the hydrogenol-
ysis of lower alkanes over Ru/Al2O3 catalysts [70].
A dual-branch sensitivity of the TOFs to the Ir dispersion
(D = 0.01–0.65) was observed for the C–C bond cleavage. The sym-
pathetic structure sensitivity originates from the declining popula-
tion of low-coordination atoms leading to lower rates. On the basis
of the measured reaction barriers, the Temkin relation and model-
ing results, we conclude that reactive intermediates of smaller H-
deficiency are involved in the rate-limiting C–C bond cleavage on
low Ir dispersion catalysts. It is also speculated that a new adsorp-
tion mode of cyclohexane that requires a sizable ensemble is more
favorably formed on terrace planes.
The pathways for multiple hydrogenolysis in kinetically pri-
mary and secondary reactions differ greatly. The relative impor-
tance of the pathways depends mainly on the Ir particle size and
H2 pressure and to a lesser degree on the reaction temperature
in the studied range. The detrimental effects of lower H2 pressures,
larger particle size, and higher temperatures on the selectivity to n-
hexane are mainly attributed to the nature of reactive intermediate
for single and multiple C–C bond cleavage, as well as the availabil-
ity of adsorbed vicinal hydrogen species. The generally high values
of apparent energies of activation for hydrogenolysis (that further
increase with increasing H2 pressure) are a consequence of the
endothermic activation of cyclic reactants. Moreover, the even
higher activation energies for internal C–C bond cleavages are
attributable to an energetically more demanding nature as well
as a low coverage of the reactive intermediates involved in the
rate-determining step.
In contrast, less than one hydrogen needs to be further ab-
stracted for multiple-scission to occur on large particles before
desorption of the ring-opened intermediate (Table 3). Moreover,
the average H-deficiency of active intermediates for internal multi-
ple-scission is greater than single-scission (RO) and terminal mul-
tiple-scission on both small and large particles, accounting for their
lower values of H2 pressure at maximum rates in primary reactions
(propane and butane in Fig. 6) and the constantly decreasing selec-
tivity to internal multiple-scission with increasing H2 pressure
(Fig. 7c). The rates of those pathways that have greater dehydroge-
nation depths before C–C bond rupture exhibit more marked sup-
pression by increasing H2 pressure (Fig. 6). The greater
dehydrogenation depths at internal C–C bonds can be attributed
to the slightly weaker secondary C–H bonds [71]. Finally, a deeper
unsaturation degree of Sim than SRO and Stm would necessitate a
higher contribution of the endothermicity term in the Temkin
Acknowledgments
Table 3
Different extents of H-deficiency of reactive intermediates for single-scission,
terminal and internal multiple-scissions at 523 K on typical small and large Ir
particles.
Hui Shi thanks the Elitenetzwerk Bayern NanoCat for a Ph.D.
grant and financial support. The authors are indebted to Dipl.-
Ing. Xaver Hecht for the help with construction of the reactor setup
and for conducting N2 physisorption and H2 chemisorption
measurements. Skillful assistances from Dr. Marianne Hanzlik
(Technical university of Munich, Germany) and Dr. Andrew
DelaRiva (University of New Mexico, USA) are gratefully
acknowledged during the TEM measurements. We also thank
xm–xRO (H-deficiency difference)a
D = 0.65
D = 0.16
Terminal multiple-scission
Internal multiple-scission
0.5–0.6
0.9–1.0
0.1–0.2
0.4–0.6
a
The difference in dehydrogenation depth (x) of reactive intermediates between
single-scission and multiple-scission (including terminal and internal).

H. Shi et al. / Journal of Catalysis 295 (2012) 133–145
145
[33] P.T. Do, W.E. Alvarez, D.E. Resasco, J. Catal. 238 (2006) 477.
[34] J.-W. Park, K. Thomas, J. van Gestel, J.-P. Gilson, C. Collet, J.-P. Dath, M. Houalla,
Appl. Catal. A. 388 (2010) 37.
Dr. George D. Meitzner (Edge Analytical, Inc.) for his critical read-
ing of the manuscript.
[35] S. Nassreddine, G. Bergeret, B. Jouguet, C. Geantet, L. Piccolo, Phys. Chem.
Chem. Phys. 12 (2010) 7812.
[36] G.B. McVicker, R.T.K. Baker, R.L. Garten, E.L. Kugler, J. Catal. 65 (1980) 207.
[37] S. Krishnamurthy, G.R. Landolt, H.J. Schoennagel, J. Catal. 78 (1982) 319.
[38] B.J. Kip, F.B.M. Duivenvoorden, D.C. Koningsberger, R. Prins, J. Catal. 105 (1987)
26.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.jcat.2012.08.005.
[39] S. Brunauer, P.H. Emmett, E. Teller, J. Am. Chem. Soc. 60 (1938) 309.
[40] S. Balcon, S. Mary, C. Kappenstein, E. Gengembre, Appl. Catal. A 196 (2000) 179.
[41] H. Lieske, G. Lietz, H. Spindler, J. Volter, J. Catal. 81 (1983) 8.
[42] M. Choi, Z. Wu, E. Iglesia, J. Am. Chem. Soc. 132 (2010) 9129.
[43] M. Boudart, G. Djega-Mariadassou, The Kinetics of Heterogeneous Catalytic
Reactions, Princeton University Press, Princeton, NJ, 1984.
[44] G.C. Bond, M.A. Keane, H. Kral, J.A. Lercher, Chem. Rev. Sci. Eng. 42 (2000) 323.
[45] B.E. Nieuwenhuys, D.I. Hagen, G.A. Somorjai, Surf. Sci. 59 (1976) 155.
[46] M.K. Oudenhuijzen, J.A. van Bokhoven, J.T. Miller, D.E. Ramaker, D.C.
Koningsberger, J. Am. Chem. Soc. 127 (2005) 1530.
References
[1] G.B. McVicker, M. Daage, M.S. Touvelle, C.W. Hudson, D.P. Klein, W.C. Baird Jr.,
B.R. Cook, J.G. Chen, S. Hantzer, D.E.W. Vaughan, E.S. Ellis, O.C. Feeley, J. Catal.
210 (2002) 137.
[2] A. Corma, V. González-Alfaro, A.V. Orchillés, J. Catal. 200 (2001) 34.
[3] H. Du, C. Fairbridge, H. Yang, Z. Ring, Appl. Catal. A 294 (2005) 1.
[4] G.C. Bond, Metal-Catalysed Reactions of Hydrocarbons, Springer, Berlin, 2005.
[5] J.H. Sinfelt, J.L. Carter, D.J.C. Yates, J. Catal. 24 (1972) 283.
[6] H. Song, R.M. Rioux, J.D. Hoefelmeyer, R. Komor, K. Niesz, M. Grass, P. Yang,
G.A. Somorjai, J. Am. Chem. Soc. 128 (2006) 3027.
[7] R.M. Rioux, H. Song, J.D. Hoefelmeyer, P. Yang, G.A. Somorjai, J. Phys. Chem. B
109 (2005) 2192.
[8] J.R. Anderson, B.G. Baker, Proc. Roy. Soc. A 271 (1963) 402.
[9] D.J.C. Yates, J.H. Sinfelt, J. Catal. 8 (1967) 348.
[10] D.E. Resasco, G.L. Haller, J. Catal. 82 (1983) 279.
[11] K. Foger, J.R. Anderson, J. Catal. 59 (1979) 325.
[12] J.R. Engstrom, D.W. Goodman, W.H. Weinberg, J. Am. Chem. Soc. 110 (1988)
8305.
[13] J.R. Engstrom, D.W. Goodman, W.H. Weinberg, J. Am. Chem. Soc. 108 (1986)
4653.
[14] D.W. Goodman, Catal. Today 12 (1992) 189.
[15] J. Barbier, P. Marecot, Nouv. J. Chim. 5 (1981) 393.
[16] R.K. Herz, W.D. Gillespie, E.E. Peterson, G.A. Somorjai, J. Catal. 67 (1981) 386.
[17] G.A. Somorjai, D.W. Blakely, Nature 258 (1975) 580.
[18] D.W. Blakely, G.A. Somorjai, J. Catal. 42 (1976) 181.
[19] F.G. Gault, Adv. Catal. 30 (1981) 1.
[20] M. Chow, G.B. McVicker, J. Catal. 112 (1988) 290.
[21] Y. Zhuang, A. Frennet, Appl. Catal. A 134 (1996) 37.
[22] D. Teschner, K. Matusek, Z. Paál, J. Catal. 192 (2000) 335.
[23] M. Vaarkamp, P. Dijkstra, J. van Grondelle, J.T. Miller, F.S. Modica, D.C.
Koningsberger, R.A. van Santen, J. Catal. 151 (1995) 330.
[24] W.E. Alvarez, D.E. Resasco, J. Catal. 164 (1996) 467.
[25] M. Chow, S.H. Park, W.M.H. Sachtler, Appl. Catal. 19 (1985) 349.
[26] P. Samoila, M. Boutzeloit, C. Especel, F. Epron, P. Marécot, J. Catal. 276 (2010)
237.
[47] A. Majesté, S. Balcon, M. Guérin, C. Kappenstein, Z. Paál, J. Catal. 187 (1999)
486.
[48] Alternatively, this difference can be attributed to the higher concentration of
more reactive under-coordinated surface Ir atoms (corners, edges) on smaller
Ir particles which discriminate less between the difficulties to cleave the first
C–H bond than the sites with higher coordination on the surface of larger
particles.
[49] V. Ponec, P.G. Menon, Catal. Rev. Sci. Eng. 25 (1983) 229.
[50] M. Temkin, Acta Phsiocochim. URSS 3 (1935) 312.
[51] S.B. Shang, C.N. Kenney, J. Catal. 134 (1992) 134.
[52] J.H. Sinfelt, D.J.C. Yates, J. Catal. 8 (1967) 82.
[53] J.H. Sinfelt, Catal. Rev. Sci. Eng. 9 (1974) 147.
[54] G.C. Bond, R.H. Cunningham, J. Catal. 166 (1997) 172.
[55] S.A. Goddard, M.D. Amiridis, J.E. Rekoske, N. Cardona-Martinez, J.A. Dumesic, J.
Catal. 117 (1989) 155.
[56] A. Cimino, M. Boudart, H. Taylor, J. Phys. Chem. 58 (1954) 796.
[57] D.J.C. Yates, J.H. Sinfelt, W.F. Taylor, J. Am. Chem. Soc. 86 (1964) 2996.
[58] It is implied in the kinetic treatment that the most abundant surface
intermediates are directly involved, not as spectators, in the primary
reaction pathways. Competitive adsorption of two reactants is reflected in
the site balance.
[59] M. Boudart, J. Phys. Chem. 87 (1983) 2786.
[60] M. Boudart, G. Djega-Mariadassou, Catal. Lett. 29 (1994) 7.
[61] M. Che, C.O. Bennett, Adv. Catal. 36 (1989) 55.
[62] G.A. Del Angel, B. Coq, G. Ferrat, F. Figueras, S. Fuentes, Surf. Sci. 156 (1985)
943.
[63] The ensemble size requirement for the C2-unit mode must have been met on
clusters of D = 0.5–0.7. Otherwise, it would have been rendered impossible to
have the TOF increase as dispersion increases. It is also implied that there
would be a maximum to be observed as the Ir dispersion tends further
towards unity.
ˇ
[27] D. Kubicka, N. Kumar, P. Mäki-Arvela, M. Tiitta, V. Niemi, H. Karhu, T. Salmi,
D.Y. Murzin, J. Catal. 227 (2004) 313.
ˇ
[28] D. Kubicka, N. Kumar, P. Mäki-Arvela, M. Tiitta, V. Niemi, T. Salmi, D.Y. Murzin,
[64] F. Garin, G. Maire, Acc. Chem. Res. 22 (1989) 100.
J. Catal. 222 (2004) 65.
[65] R. van Hardeveld, F. Hartog, Surf. Sci. 15 (1969) 189.
[66] J. Lu, P. Serna, C. Aydin, N.D. Browning, B.C. Gates, J. Am. Chem. Soc. 133 (2011)
16186.
[67] A. Uzun, V. Ortalan, N.D. Browning, B.C. Gates, J. Catal. 269 (2010) 318.
[68] A. Uzun, D.A. Dixon, B.C. Gates, ChemCatChem 3 (2011) 95.
[69] A.M. Argo, J.F. Odzak, B.C. Gates, J. Am. Chem. Soc. 125 (2003) 7107.
[70] G.C. Bond, J.C. Slaa, J. Mol. Catal. A 98 (1995) 81.
[29] S. Lecarpentier, J. van Gestel, K. Thomas, J.-P. Gilson, M. Houalla, J. Catal. 254
(2008) 49.
[30] M. Santikunaporn, J.E. Herrera, S. Jongpatiwut, D.E. Resasco, W.E. Alvarez, E.L.
Sughrue, J. Catal. 228 (2004) 100.
[31] S. Rabl, A. Haas, D. Santi, C. Flego, M. Ferrari, V. Calemma, J. Weitkamp, Appl.
Catal. A 400 (2011) 131.
[32] S. Nassreddine, L. Massin, M. Aouine, C. Geantet, L. Piccolo, J. Catal. 278 (2011)
253.
[71] D.F. McMillen, D.M. Golden, Ann. Rev. Phys. Chem. 33 (1982) 493.