Kinetic and Mechanistic Assessment of Alkanol/Alkanal Decarbonylation and Deoxygenation Pathways on Metal Catalysts

Article abstract of DOI:10.1021/jacs.5b05361

This study combines theory and experiment to determine the kinetically relevant steps and site requirements for deoxygenation of alkanols and alkanals. These reactants deoxygenate predominantly via decarbonylation (C-C cleavage) instead of C-O hydrogenolysis on Ir, Pt, and Ru, leading to strong inhibition effects by chemisorbed CO (CO?). C-C cleavage occurs via unsaturated species formed in sequential quasi-equilibrated dehydrogenation steps, which replace C-H with C-metal bonds, resulting in strong inhibition by H₂, also observed in alkane hydrogenolysis. C-C cleavage occurs in oxygenates only at locations vicinal to the C=O group in RCCO? intermediates, because such adjacency weakens C-C bonds, which also leads to much lower activation enthalpies for oxygenates than hydrocarbons. C-O hydrogenolysis rates are independent of H₂ pressure and limited by H?-assisted C-O cleavage in RCHOH? intermediates on surfaces with significant coverages of CO? formed in decarbonylation events. The ratio of C-O hydrogenolysis to decarbonylation rates increased almost 100-fold as the Ir cluster size increased from 0.7 to 7 nm; these trends reflect C-O hydrogenolysis reactions favored on terrace sites, while C-C hydrogenolysis prefers sites with lower coordination, because of the relative size of their transition states and the crowded nature of CO?-covered surfaces. C-O hydrogenolysis becomes the preferred deoxygenation route on Cu-based catalysts, thus avoiding CO inhibition effects. The relative rates of C-O and C-C cleavage on these metals depend on their relative ability to bind C atoms, because C-C cleavage transitions states require an additional M-C attachment.

Full text of DOI:10.1021/jacs.5b05361

Article
pubs.acs.org/JACS
Kinetic and Mechanistic Assessment of Alkanol/Alkanal
Decarbonylation and Deoxygenation Pathways on Metal Catalysts
Elif I. Gurbuz,†,‡ David D. Hibbitts,† and Enrique Iglesia*,†,‡
̈
̈
‡
†Department of Chemical and Biomolecular Engineering and Division of Chemical Sciences, E. O. Lawrence Berkeley National
Laboratory, University of California at Berkeley, Berkeley, California 94720, United States
S
* Supporting Information
ABSTRACT: This study combines theory and experiment to
determine the kinetically relevant steps and site requirements for
deoxygenation of alkanols and alkanals. These reactants
deoxygenate predominantly via decarbonylation (C−C cleavage)
instead of C−O hydrogenolysis on Ir, Pt, and Ru, leading to
strong inhibition effects by chemisorbed CO (CO*). C−C
cleavage occurs via unsaturated species formed in sequential
quasi-equilibrated dehydrogenation steps, which replace C−H
with C−metal bonds, resulting in strong inhibition by H2, also
observed in alkane hydrogenolysis. C−C cleavage occurs in
oxygenates only at locations vicinal to the CO group in
RCCO* intermediates, because such adjacency weakens C−C
bonds, which also leads to much lower activation enthalpies for oxygenates than hydrocarbons. C−O hydrogenolysis rates are
independent of H2 pressure and limited by H*-assisted C−O cleavage in RCHOH* intermediates on surfaces with significant
coverages of CO* formed in decarbonylation events. The ratio of C−O hydrogenolysis to decarbonylation rates increased almost
100-fold as the Ir cluster size increased from 0.7 to 7 nm; these trends reflect C−O hydrogenolysis reactions favored on terrace
sites, while C−C hydrogenolysis prefers sites with lower coordination, because of the relative size of their transition states and the
crowded nature of CO*-covered surfaces. C−O hydrogenolysis becomes the preferred deoxygenation route on Cu-based
catalysts, thus avoiding CO inhibition effects. The relative rates of C−O and C−C cleavage on these metals depend on their
relative ability to bind C atoms, because C−C cleavage transitions states require an additional M−C attachment.
Ir, Ru, Rh) indicate that H-removal steps occur before C−C or
C−O cleavage, similar to the extensive dehydrogenation that
precedes C−C cleavage of alkanes.20,21 High-resolution
electron energy loss spectra,9 the identity and desorption
dynamics of the products formed from preadsorbed etha-
nol,10−12 and DFT calculations14−17 on single crystals, however,
have not led to a consensus about the structure and the degree
of unsaturation of the relevant intermediates and transition
states. Also, while C−O hydrogenolysis of biomass-derived
furanic species was reported on Cu in gaseous22,23 and
aqueous24,25 media, its turnover rate and the C−C and C−O
bond cleavage selectivities have not been systematically
compared among Cu and Group VIII metals. These unresolved
or unaddressed matters have prevented precise assessments of
the origins of reactivity in C−C and C−O cleavage of alkanols/
alkanals and of how C−C cleavage differs in mechanistic detail
among alkanols, alkanals, and alkanes.
1. INTRODUCTION
Oxygen must be removed from biomass-derived molecules to
render them useful as fuels and chemicals, but such processes
remain challenging because of their significant energy and H2
requirements.1,2 Oxygen removal strategies include C−O
hydrogenolysis, using H2 to remove oxygen as H2O (hydro-
deoxygenation), and C−C hydrogenolysis reactions that form a
shorter carbon backbone and CO (decarbonylation) or CO2
(decarboxylation) as coproducts. Alkanols and alkanals are
ubiquitous in biomass-derived streams, and a more detailed
understanding of the relevant mechanisms by which these
molecules react on metal surfaces is essential to choose catalytic
systems for their efficient conversion in reforming,3 direct fuel
cells,4 and oxidation reactions.5
The conversion of many complex alkanol substrates1,6−8 has
been studied on several metal catalysts in the aqueous phase,
but without detailed mechanistic conclusions, at least in part,
because of the mass transfer limitations and nonideal
thermodynamics, ubiquitous in such aqueous media. Rates of
ethanol and 1-propanol decarbonylation reactions were 10−
100 times larger than C−O hydrogenolysis rates on single
crystals of Pt,13 Pd,9−11 and Ni.12 Density functional theory
(DFT) treatments of the conversion pathways for ethanol,14−17
ethylene glycol,18 and glycerol19 on Group VIII metals (Pt, Pd,
Here, we report turnover rates and their dependence on
reactant and product concentrations for reactions of 1-butanol/
H2 mixtures on supported Ir, Pt, Ru, and Cu clusters at
conditions relevant to their practical use. Decarbonylation
Received: May 23, 2015
© XXXX American Chemical Society
A
DOI: 10.1021/jacs.5b05361
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
◆
◆
▲
▲
Figure 1. Effects of residence time on (a) decarbonylation ( ) and C−O hydrogenolysis ( ) turnover rates, (b) propane ( ), CH4 ( , ×4), and
■
butane ( ) carbon (propane and C1 molecules (CO, CH4) formed in equimolar amounts at all conditions), and (c) decarbonylation turnover rates
with no H2O or CO added ( ), 1 kPa H2O added (X) and 1 kPa CO added ( ) (1 wt % Ir/SiO2 (0.7 nm), 523 K, 2 MPa H2, 5 kPa 1-butanol). 1-
Butanol/butanal pool conversions range from 1.2% to 6.4% at these residence times. Dashed lines represent trendlines.
◆
■
All samples were dried in flowing dry air (Praxair, 99.99%, 5.0 cm3
occurs with much higher turnover rates than C−O hydro-
g−1 s−1), then treated under flowing dry air and 50% H2/He (Praxair,
genolysis on small clusters of Ir, Pt, and Ru (∼1 nm mean
99.999%, 1.0 cm3 g−1 s−1) to reduce the metals and obtain various
cluster sizes. The details of the drying, treatment, and reduction
diameter), and no other C−C cleavage reactions occur. DFT-
derived activation barriers indicate that the presence of C−O
conditions for each catalyst (Ir (1.0% wt), Ir (2.0% wt), Ir (3.0% wt),
double bonds (in alkanol/alkanal-derived intermediates)
Ru (0.5% wt), Pt (1.0% wt), and Cu (5.0% wt)) may be found in the
significantly weakens adjacent C−C bonds, resulting in lower
barriers for C−C cleavage vicinal to −CO groups (decarbon-
ylation) than for C−C cleavage at other backbone positions. H2
pressure inhibits decarbonylation and has no effect on C−O
hydrogenolysis, and these H2 effects are consistent with
kinetically relevant transition states of decarbonylation that
are more dehydrogenated than those for C−O hydrogenolysis,
confirmed here using DFT methods. The ratio of C−O
hydrogenolysis to decarbonylation for 1-butanol is much lower
than unity on small (∼1 nm) Ir, Ru, and Pt catalysts and
approaches unity only on larger (7 nm Ir) clusters, indicating
that C−O hydrogenolysis occurs on low-index terrace sites that
become more prevalent on larger clusters, while decarbon-
ylation is much less sensitive to surface coordination. C−O
hydrogenolysis is the predominant route on Cu because CO*
and C* form weaker bonds on Cu than Group VIII metals,
consistent with much lower decarbonylation turnover rates on
Cu (10−100 times smaller) than on Ir, Pt, and Ru catalysts.
Supporting Information (SI). All samples were cooled to ambient
temperature and passivated in flowing 0.5% O2/He (Praxair, 99.99%,
1.0 cm3 g−1 s−1) for 3 h before exposure to ambient air.
The number of exposed Ir, Pt and Ru atoms in each catalyst was
determined from volumetric uptakes of H2 at 298 K using previously
reported procedures21 and assuming adsorption stoichiometries of 1/1
for H/Ms.28 The mean cluster diameter for each sample was estimated
from the measured dispersion by assuming hemispherical crystallites
and the atomic density of the bulk metals.29,30 Cluster sizes for Ir
(1.0% wt, 140% dispersion), Ir (2.0% wt, 26% dispersion), Ir (3.0% wt,
13% dispersion), Ru (0.5% wt, 98% dispersion), and Pt (1.0% wt,
155% dispersion) were determined to be 0.7, 3.6, 7.0, 0.6, and 0.7 nm,
respectively. These cluster sizes were confirmed with transmission
electron microscopy (TEM) in previous work.20,21 For Cu, the cluster
size distribution was determined to be 5.1 nm by TEM imaging
corresponding to a fractional dispersion of 0.20 for the case of
hemispherical clusters with the density of bulk Cu metal.30 The details
of the TEM measurements and dispersion calculations can be found in
the SI.
2.2. Catalytic Reactions of 1-Butanol. 1-Butanol (Sigma-
Aldrich, 98%) conversion rates and selectivities were measured in a
stainless steel tubular flow reactor (3/8″ O.D.) with plug-flow
hydrodynamics at 483−523 K, 1−5 kPa 1-butanol, and 1−3 MPa
H2. The details of the temperature, pressure, gas, and liquid flow
controls can be found in the SI. Catalysts were mixed with SiO2 (Cab-
O-Sil HS-5, washed with deionized water and treated in flowing dry air
at 793 K for 5 h). Decarbonylation and C−O hydrogenolysis turnover
rates (Figure 1) were unaffected by dilution (SiO2/Ir-SiO2 = 0−4),
indicating that neither bed temperature gradients nor contributions
from the support influenced measured turnover rates. All samples were
treated in flowing H2 (Praxair, 99.999%; 50 cm3 g−1 s−1) at ambient
pressure by heating to 673 K at 0.083 K s−1 and holding for 2 h before
rate measurements.
2. EXPERIMENTAL METHODS
2.1. Synthesis and Characterization of SiO2-Supported Ir, Pt,
Ru, and Cu Clusters. SiO2 (Davisil 646, 300 m2 g−1) was treated in
flowing dry air (Praxair, 99.99%, 5.0 cm3 g−1 s−1) by heating to 823 K
(at 0.03 K s−1) and holding for 5 h. Ir (1.0% wt), Ru (0.5% wt) and Cu
(5% wt) precursors were deposited onto the treated silica using
incipient wetness impregnation with aqueous solutions of Ir (H2IrCl6,
Strem Chemicals, 99%), Ru (Ru(NO) (NO3)3, Alfa-Aesar, 32% wt
Ru), and Cu (Cu(NO3)2·2.5H2O, Strem Chemicals, 99.5%), which
were mixed with triethanolamine (TEA, Sigma-Aldrich, 97%) (20:1
mol) solutions in deionized (DI) water (17.9 MΩ resistivity) before
impregnation; these procedures have been reported to lead to small
clusters of uniform size for Ru-based catalysts and were adapted here
for the synthesis of Ir and Cu.21,26 Ir−SiO2 samples with different Ir
contents (2.0% and 3.0% wt) were prepared by varying the
concentration of TEA and H2IrCl6 in the impregnating solution
while maintaining a 20:1 molar ratio. Pt/SiO2 (1.0% wt) was prepared
by strong electrostatic adsorption of Pt(NH3)4(NO3)2 (Sigma-Aldrich,
99.9%) onto SiO2 (Davisil 646) using a NH4OH solution.27
Reactant and product concentrations were measured by gas
chromatography (Agilent GC, 5890) using a methyl silicone capillary
column (HP-1, Agilent, 50 m × 0.32 mm × 1.05 μm) with a flame
ionization detector and a packed column (Porapak Q, Alltech, 80/100,
12′ × 1/8′) with a thermal conductivity detector. All rates are reported
at differential conversions (<5%) in reference to the alkanal/alkanol
reactant pool to avoid axial concentration gradients. Turnover rates are
normalized by the number of surface metal atoms determined by H2
B
DOI: 10.1021/jacs.5b05361
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
a
Scheme 1. Reaction Pathways Involved in 1-Butanol Conversion
a
Stoichiometries are based on the use of 1-butanol (instead of its equilibrated butanal dehydrogenation product) as the reactant.
◆
▲
Figure 2. Decarbonylation ( ) and C−O hydrogenolysis ( ) turnover rates on 1 wt % Ir/SiO2 (0.7 nm) at 523 K. (a) 1-Butanol pressure effects at
2 MPa H2 and 1 kPa CO. (b) H2 pressure effects at 5 kPa 1-butanol and 1 kPa CO. (c) CO pressure effects at 5 kPa 1-butanol and 2 MPa H2.
Dashed lines are representative of turnover rates calculated using the regressed kinetic and thermodynamic parameters and the proposed rate eqs
(eqs 2 and 6 for decarbonylation and C−O hydrogenolysis, respectively).
chemisorption for Ir, Pt, and Ru and by TEM for Cu. Selectivities are
reported as the percentage of the C atoms in the 1-butanol/butanal
equilibrated pool converted that appears within a given product.
lower H2 pressures (<0.1 MPa, 400−523 K). Propanol, ethanol,
and ethane were not detected as products, indicating that only
C−C bonds vicinal to the O atom are cleaved, consistent with
2.3. Computational Methods. Periodic DFT calculations were
1-propanol reactions on Pd11 and Pt.13 CO2 was also not
performed using the Vienna ab initio simulation package (VASP)31−34
detected, indicating CO did not react with H2O formed in situ.
with planewaves constructed using projector augmented-wave (PAW)
potentials.35,36 Exchange and correlation energies were determined
using revised Perdew−Burke−Ernzerhof (RPBE) form of the
generalized gradient approximation (GGA).37−39 Transition-state
structures were obtained for each elementary reaction using the
nudged elastic band (NEB) method40,41 and the dimer method.42 The
details of these electronic structure calculations may be found in the
SI.
1-Butanol decarbonylation and C−O hydrogenolysis turnover
rates were unaffected by the presence of additional SiO2 present
as physical mixture (SiO2/Ir-SiO2 = 0−4, Cab-O-Sil HS-5,
washed with deionized water and treated in flowing dry air at
793 K for 5 h), indicative of monofunctional reactions. Propane
and butane selectivities depended weakly on residence time and
reactant pool conversion (Figure 1b), even though they form
directly from equilibrated butanol-butanal mixtures, because
their formation rates are inhibited to different extents by the
CO formed in decarbonylation reactions. CH4 selectivity was
low and increased with residence time (Figure 1b), together
with a decrease in CO selectivity, because of methanation
reactions. Propane and C1 molecules (CO, CH4) formed in
equimolar amounts at all conditions.
Decarbonylation and C−O hydrogenolysis turnover rates
decreased with increasing residence time (Ir/SiO2, Figure 1a),
even at very low reactant pool conversions (2−6%), indicating
that products (CO and/or H2O) inhibit these reactions. Rates
were unaffected by added H2O (Figure 1c) at pressures (1 kPa
H2O) above those prevalent during 1-butanol reactions (<0.003
kPa H2O); therefore, H2O-derived intermediates do not
account for the observed product inhibition. CO added to 1-
butanol/H2 reactants (1 kPa vs <0.07 kPa during reaction;
Figure 1c) led to more than a 10-fold decrease in rates; rates
were no longer influenced by conversion when CO was present.
Turnover rates were stable with time on stream (Figure S1),
indicating that CO* did not irreversibly poison the Ir metal, but
3. RESULTS AND DISCUSSION
3.1. 1-Butanol Turnover Rates and Selectivity on Ir/
SiO2. 1-Butanol can react reversibly on Ir/SiO2 to form butanal
and H2 (dehydrogenation) with an approach to equilibrium
(η):
P Pbutanal
Kdeh
1
H2
C4H9OH ←→ C4H8O + H2, η =
Pbutanol Kdeh
(1)
value of 0.7−1 even at the low 1-butanol conversions in this
study (0.8% to 6.4%), allowing butanol and butanal to be
treated as a lumped reactant pool in all reported turnover rates
and selectivities. 1-Butanol can also form (i) propane and CO
(decarbonylation; C−C hydrogenolysis), (ii) butane and H2O
(C−O hydrogenolysis), and (iii) CH4 and H2O (methanation)
via a secondary reaction of CO and H2 (Scheme 1).
1-Butanol decarbonylation rates were >10 times higher than
C−O hydrogenolysis rates (Ir/SiO2; Figure 1a, 0.7 nm clusters;
473−523 K, 1−3 MPa H2), as also reported for etha-
nol9−14,17,43 and propanol13 on Pt, Pd, Ru, and Ni catalysts at
C
DOI: 10.1021/jacs.5b05361
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
instead inhibited rates through quasi-equilibrated adsorption/
desorption events. These data show that CO strongly inhibits
rates by competitive adsorption, leading to kinetically
detectable CO* coverages, even at very low conversions (2−
6%) and CO pressures (0.04−0.07 kPa). Such CO inhibition
effects are also evident in aqueous and gas-phase reforming of
oxygenates (ethylene glycol, glycerol, sorbitol)18,44 on Pt and
Ni surfaces and in ethanol decarbonylation on Pt surfaces;16
these strong inhibitory effects of CO* remain influential even at
the higher and more relevant H2 pressures (1−3 MPa) and low
conversions of the present study.
3.2. Effects of 1-Butanol, H2, and CO Pressures on
Decarbonylation Turnover Rates. Decarbonylation turn-
over rates were proportional to 1-butanol pressure (Figure 2a)
and inversely dependent on H2 (Figure 2b) and CO (Figure
2c) pressures on Ir/SiO2. The observed decrease in decarbon-
ylation rates with increasing H2 pressure is consistent with the
involvement of unsaturated species, formed via quasi-
equilibrated dehydrogenation steps, as intermediates in the
formation of the transition states that mediate kinetically
relevant steps.
where the denominator terms account for the relative coverages
of vacant sites (*) and (CO*) species and KCO is the
equilibrium constant for CO* adsorption. The α term in eq 2
represents an effective rate constant for decarbonylation:
α = KBuOHKOH,1KCH,1KCH,2K−1.5kCH,3
H2
(3)
with the rate and equilibrium constants defined for the
elementary steps in Scheme 2. The α term reflects the
contributions of C−H activation in C2H5CH2C*O species
(Scheme 2, step 2.6) as the sole kinetically relevant step. The
values of α (2.9 0.13 kPa0.5 s−1) and KCO (6.3 × 10−1 2.4 ×
10−2 kPa−1) were obtained by regressing all rate data (Figure 3,
Scheme 2 depicts a sequence of elementary steps that
accounts for the observed kinetic effects of alkanol, CO, and H2
Scheme 2. Proposed Intermediates and Sequence of
Elementary Steps Involved in Decarbonylation of 1-Butanol
a
on Ir Clusters
◆
Figure 3. Measured decarbonylation ( ) and C−O hydrogenolysis
▲
(
) turnover rates on 1% wt Ir/SiO2 and those predicted from
◆
regression of the rate data to the functional forms of eqs 2 ( ) and 6
▲
(
), respectively, at 523 K.
523 K, 0.7 nm Ir) to the functional form of eq 2. Fractional
CO* coverages derived from this KCO value range from 0.30 to
0.85 as CO pressures increased from 0.8 to 8 kPa at 523 K.
Alternate decarbonylation pathways, such as those involving
kinetically relevant C−C activation in C2H5CH2CO* (instead
of C−H activation in C2H5CH2C*O) would have led to rates
also given by eq 2, but are ruled out here because of their
significantly larger DFT-derived barriers compared with the
steps in Scheme 2 (discussed in detail below). Eq 2 assumes
that H* species are minority surface species, even at 1−3 MPa
H2 pressures (Figure 3). This conclusion is consistent with the
insensitivity of C−O hydrogenolysis rates to H2 pressure and
with the similar effects of H2 pressure on Ir particles with
different mean diameters (0.7−7 nm), which suggest that H*
coverages are not kinetically detectable at these conditions (1−
5 kPa 1-butanol, 1−3 MPa H2, 0.8−8 kPa CO at 523 K), as
further discussed in the SI (Figures S2 and S3).
Previous rate data and theoretical assessments for alkane
hydrogenolysis20,21 have indicated that one or more C atoms
undergo C−H activation to form C−M bonds before C−C
cleavage. The requirement for bound C atoms and the
equilibrated nature of alkanol-alkanal interconversions (Section
3.1) suggest the involvement of RCH*O, RC*O, and RC*OH
species (where R = CH3CH2, CH3CH*, and CH3C*) in
decarbonylation (C−C cleavage). DFT calculations were
carried out for these species instead of 1-butanol/butanal to
decrease conformational diversity and computational intensity
a
The * denotes an unoccupied surface site; kx and Kx are kinetic
constants for forward steps and equilibrium constants, respectively.
Quasi-equilibrated steps are noted by a circle over the double arrows.
pressures on 1-butanol decarbonylation rates (Figure 2) and
which is consistent with DFT calculations described below.
Steps 2.1 and 2.2 in Scheme 2 involve quasi-equilibrated H2
dissociation and molecular adsorption of 1-butanol, respec-
tively. Adsorbed butanol undergoes quasi-equilibrated O−H
and C−H activation events (2.3 to 2.5) to form acyl
intermediates (C2H5CH2C*O, the * appears to the right of
an atom bound to a surface atom) and an irreversible C−H
activation at the α-C atom to form C2H5CH*C*O species
(2.6). The C2H5CH*C*O species undergo another C−H
activation to form C2H5C*C*O (2.7), which cleaves its C−C
bond to form C3H5* and CO* (2.8).
These elementary steps and reversibility assumptions
(Scheme 2) lead to the rate equation:
α(BuOH)
r
[L]
=
(H2)1.5(1 + KCO(CO))2
(2)
D
DOI: 10.1021/jacs.5b05361
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Scheme 3. DFT-Calculated Effective Enthalpy Barriers (ΔH⧧, black, bold font), Intrinsic Activation Enthalpies (ΔHact, blue, left
a
of slash), and Reaction Enthalpies (ΔHrxn, red, right of slash) for Propanol Decarbonylation Pathways
a
Reversible steps are represented by double arrows (↔) and irreversible steps are represented by single arrows (→). Reaction numbers used
throughout the text are given within circles (①, e.g.) on top of each reaction arrow. Structures for reactant, transition, and product states for all
reactions shown in this scheme are available in the SI (Figures S4−S15).
Figure 4. (A) ΔHact (open symbols, difference between transition states and relevant precursors); (B) ΔH⧧ (closed symbols, difference between
⧧
■
▲
●
transition states and gas-phase propanol); and (C) ΔG (at 483 K, 1 bar) and for C−C activations of RCH*O ( ), RC*OH ( ), and RC*O ( )
intermediates where R = CH3CH2, CH3CH*, and CH3C*.
and because C−C hydrogenolysis occurs exclusively at the C−
C bond vicinal to the −OH group (Section 3.1).
occurs with smaller ΔHact and ΔH⧧ than for C−C cleavage in
RCH*O (Figures 4 and 5). ΔHact and ΔH⧧ values decreased as
H atoms are removed from the α-C atom in both RCH*O and
RC*O species, because H-removal occurs with the concomitant
formation of C−M bonds (Figure 5), which weakens C−C
bonds, as also shown for C−C cleavage steps alkane
hydrogenolysis.20,21,45,47−49
Adsorbed propanal can react with a vacant site and undergo
C−C cleavage to form vicinal CH3CH2* and CH*O* species
(Scheme 3, Step 2); ΔHact values (151 kJ mol−1) for this step
are much larger than C−H bond activation at the α-C position
in adsorbed propanal to form an CH3CH*CH*O and H*
(Step 3, ΔHact,3 = 88 kJ mol−1). Subsequent C−H activation at
the α-C position would then form CH3C*CH*O (Step 7,
ΔHact,7 = 48 kJ mol−1), thus C−H activations at the α-C are
more facile than direct C−C activation of propanol. Intrinsic
(ΔHact, enthalpy differences between transition states and the
corresponding precursors) and effective (ΔH⧧, difference
between the transition state and propanol in the gas-phase)
enthalpy barriers for C−C cleavage reactions both decrease as
more H atoms are removed from the α-C atom (Figures 4 and
5) for all RCH*O species (R = CH3CH2, CH3CH*, CH3C*).
Propanal can also undergo C−H activation at the carbonyl C
atom in RCH*O to form RC*O species (CH3CH2C*O,
CH3CH*C*O, and CH3C*C*O). C−C cleavage in RC*O
Successive C−H activation events in propanol-derived
species can form RC*OH species (R = CH3CH2, CH3CH*,
CH3C*), which can undergo C−C cleavage (Scheme S1, SI).
C−C cleavage in RC*OH (R = CH3CH2, CH3CH*, CH3C*)
occurs with higher ΔHact and ΔH⧧ than in RC*O (Figures 4
and 5). ΔHact and ΔH⧧ values for C−C cleavage in RC*OH
both decreased as α-C atoms lose H atoms, as also found for
RC*O and RCH*O species (Figure 4). Transition states for
C−C cleavage in RC*OH exhibit much longer C−O bonds
than in RC*O (Figure 5), and the bonding geometry at the C
atom for RC*OH and RC*O species indicates a shift from sp3
to sp2 hybridization (from RC*OH to RC*O). These changes
in local geometry indicate that CO bonds are present in
E
DOI: 10.1021/jacs.5b05361
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
⧧
rmax
[L]
kBT
h
(−ΔG /RT) (alkanol)
1
=
e
[1 + KCO(CO)]2
(H2)y/2
(4)
where kB and h are the Boltzmann and Planck constants,
respectively, and y represents the number of H atoms removed
before C−C cleavage. Since rates depend on Gibbs free
energies (ΔG⧧), entropy (ΔS⧧) was accounted for by
calculating the relative magnitudes of the C−C cleavage in
the nine intermediates in Figures 4 and 5. ΔS⧧ values increase
with increasing extent of dehydrogenation because H2 (g) is
evolved upon forming each transition state. C−C cleavage in
CH3C*C*O (with the lowest ΔH⧧, Figure 4B) shows the
largest ΔS⧧ values and the smallest ΔG⧧ (Figure 4C),
indicating that C−C cleavage occurs almost exclusively via
CH3C*C*O intermediates.
In alkane hydrogenolysis on metals, C−C cleavage steps are
irreversible and kinetically relevant, while C−H activation steps
are quasi-equilibrated.20,21 If this were the case for alkanol/
alkanal decarbonylation reactions, C−C cleavage via
CH3C*C*O species would be given by
α(BuOH)
r
[L]
=
(H2)2.5(1 + KCO(CO))2
(5)
inconsistent with measured rates (r ∼ (H2)−1.5; eq 2). These
DFT treatments, taken together with measured rates, indicate
that while C−C cleavage may indeed occur preferentially in
CH3C*C*O species during decarbonylation, some preceding
C−H activation steps may not achieve equilibrium. The
measured H2 effects on decarbonylation rates implicate the
activation of a C−H bond in CH3CH2C*O* species (Step 9)
as the kinetically relevant decarbonylation step. These
inferences are made rigorous by the seamless coupling of
theory and experiment in this study; mechanistic conclusions
would have been inaccurate if based on kinetic analysis or
theoretical treatments alone. Next, we assess the differences in
C−C activation events among alkanols/alkanals and alkanes.
3.3. Effects of Proximity to Carbon−Oxygen Bond on
the Cleavage Rates of C−C Bonds. C−C bonds in alkanols
cleave after four H atoms are removed from the C atoms
involved in the cleaved C−C bond based on experimental and
DFT results shown in Section 3.2, as for C−C hydrogenolysis
Figure 5. Transition-state structures, ΔH⧧ (black, boldface), and
ΔHact (blue) for C−C activations or RCH*O*, RCOH* and RCO*
intermediates. Bond distances (in pm) are shown in yellow; “x2”
indicates a C−M bond with a similar ( 1 pm) bond length which
cannot be seen at the viewing angle used.
RC*O but not in RC*OH; these CO bonds weaken adjacent
C−C bonds, thus favoring C−C bond cleavage in RC*O over
similar reactions in RC*OH.
C−C cleavage in CH3C*C*O (Step 14) shows the lowest
ΔHact (48 kJ mol−1) and ΔH⧧ (123 kJ mol−1) values among all
nine C−C cleavage steps considered (Steps 2, 5, 6, 12, and 14
in Scheme 3, and Steps 27, 30, and 33 in Scheme S1). The
relative rates of C−C cleavage in these nine intermediates
(RCH*O, RCOH*, and RCO* with R = CH3CH2, CH3CH*,
and CH3C*) were evaluated by determining the maximum
possible rate of each C−C cleavage step. This maximum rate
requires that all previous steps be quasi-equilibrated, thus
resulting in a thermodynamically limited pool of reactant
precursors. Transition-state theory formalisms then give these
maximum rates for each C−C cleavage steps as
1
1
2
2
of C−1C, C−2C, and C−2C bonds (1C and C represent
primary and secondary carbons, respectively, with C−C
activation as the kinetically relevant step) in acyclic and cyclic
alkanes.20,21,45,50 In contrast with alkanes, C−C hydrogenolysis
in alkanols involves: (i) O−H activation to form CO bonds
that weaken vicinal C−C bonds; (ii) C−H activation, instead of
C−C activation, as kinetically relevant step; and (iii) the
Figure 6. Transition-state structures, ΔH⧧, and ΔHact for C−C cleavage via decarbonylation (CH3C*−CO*⧧) and via RC*−CH*⧧ transition states
(R = −CH3, −CH2OH, −CHO). Important bond distances (in pm) are shown in yellow; “x2” indicates that a bond which cannot be seen at the
viewing angle used has a similar bond length.
F
DOI: 10.1021/jacs.5b05361
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
selective activation of C−C bonds adjacent to CO groups.
The weakening of C−C bonds by the vicinal CO moieties is
evident from the DFT-calculated ΔHact values for C−C
cleavage in CH3C*C*O (48 kJ mol−1) and CH3C*C*H (65
kJ mol−1). These ΔHact values lead to much smaller ΔH⧧ values
for decarbonylation (123 kJ mol−1) than for alkane hydro-
genolysis (170 kJ mol−1), consistent with 1-butanol/butanal
decarbonylation turnover rates that are >104 times larger than
for C−C hydrogenolysis in n-butane on Ir (0.7 nm) at similar
conditions (523 K, 2 MPa H2).21
Here, we also examine the C−C cleavage of alkanols at
bonds without neighboring O atoms. Intrinsic activation
enthalpies (ΔHact) are much higher for C−C cleavage in
CH*C*CH2OH (90 kJ mol−1) and in CH*C*CHO (97 kJ
mol−1) than for the respective propane-derived intermediates
(65 kJ mol−1 for CH*C*CH3) (Figure 6). Decarbonylation has
the lowest ΔH‡ among these C−C bond activations (Figure 6),
consistent with the exclusive formation of C3 and C1 molecules
from 1-butanol reactants (Section 3.1).
with parameters defined in Scheme 4. Regressing all rate data to
the functional form of eq 6 shows that the elementary steps in
Scheme 4 accurately describes the measured C−O cleavage
rates at all 1-butanol, H2 and CO pressures (Figure 3). The
parameter β in eq 7 is 6.0 0.59 ×10−7 kPa−1 s−1 at 523 K; the
K
CO value was slightly smaller for C−O hydrogenolysis (0.47
0.06 kPa−1) than for decarbonylation (0.63 0.02 kPa−1). The
variance in KCO values and the relationship between the types
of sites involved in each reaction are discussed in Section 3.5.
Scheme 5 shows possible pathways and the corresponding
enthalpic barriers for C−O hydrogenolysis of 1-propanol.
Scheme 5. DFT-Derived Effective Enthalpy Barriers (ΔH⧧,
black, bold font), Intrinsic Activation Enthalpies (ΔHact,
blue, italic font), and Reaction Enthalpies (ΔHrxn, red) for
a
Propanol C−O Hydrogenolysis Pathways
3.4. Effects of 1-Butanol, H2, and CO Pressures on C−
O Hydrogenolysis Turnover Rates. Measured C−O
hydrogenolysis rates are proportional to 1-butanol pressure
(Figure 2a) and inversely dependent on CO pressure (Figure
2c) on Ir/SiO2, as also observed for decarbonylation rates
(Figure 2a,c). C−O hydrogenolysis rates do not depend on H2
pressure, while decarbonylation reactions are inhibited by H2
(cf. Figure 2b). These data indicate that kinetically relevant C−
O hydrogenolysis transition states contain the same number of
H atoms as 1-butanol, making them much less dehydrogenated
than the transition states for decarbonylation, as also proposed
based on theoretical treatments of ethanol reactions on
Pt(111).17
Scheme 4 depicts a plausible sequence of elementary steps
for C−O hydrogenolysis of 1-butanol; these steps are
a
Rate dependences on H2 are based on the alkanol reactant. Reversible
steps are represented by double arrows (↔), and irreversible steps are
represented by single arrows (→). Reaction numbers used throughout
the text are encircled (①, e.g.) above each reaction arrow. Structures
for reactant, transition, and product states for all reactions shown in
this scheme are available in the SI (Figures S4−S15). Color version
available online.
Scheme 4. Proposed Sequence of Steps and Intermediates in
C−O Hydrogenolysis of 1-Butanol on Ir Clusters
a
Molecularly adsorbed propanol can react directly with a vicinal
vacant site by cleaving its C−O bond and forming
CH3CH2CH2* and OH* (Step 16). This direct activation
would lead to rates with the functional form of eq 6, but with a
different chemical meaning of β. The intrinsic activation barrier
for this step (ΔHact,16 = 171 kJ mol−1) is much larger than for
cleaving the O−H bond in 1-butanol to form propoxyl
(CH3CH2CH2O*) and H* (Step 17, ΔHact,17 = 73 kJ mol−1)
or for cleaving the C−H bond in 1-butanol at the O-bound C
atom to give hydroxypropyl species (CH3CH2CH*OH) and
H* (Step 18, ΔHact,18 = 83 kJ mol−1). Thus, C−H and O−H
activation steps occur before the C−O bond cleaves on Ir
surfaces.
The reaction network in Scheme 5 includes six different C−
O activation steps (Steps 16, 19, 21, 22, 25, and 26). Out of the
six C−O activation routes, there are two pathways (Steps 16
and 21) involving transition states that contain eight H atoms
and would lead to rates that are unaffected by H2 pressure, as
observed experimentally (Figure 2). C−O cleavage is mediated
in molecularly adsorbed propanol (via reaction with a vicinal *)
in Step 16 in the first route and in CH3CH2CH*OH species
a
The * denotes an unoccupied surface site; kx and Kx are kinetic
constants for forward steps and equilibrium constants, respectively.
consistent with the observed effects of alkanol and H2 pressures
on rates (Figure 2) and with theoretical estimates described
below. Molecularly adsorbed butanol undergoes quasi-equili-
brated C−H activation (4.3) to form CH3CH2CH*OH, which
then reacts in an irreversible step with H* with concerted O−H
formation and C−O cleavage to form vicinal H2O* and
CH3CH2CH* (4.4). The corresponding rate equation is
β(BuOH)
r
[L]
=
(1 + KCO(CO))2
(6)
(7)
with β given by
β = KBuOHKCHkC−O
G
DOI: 10.1021/jacs.5b05361
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
◆
■
Figure 7. (a) Effect of Ir cluster size on decarbonylation ( ) and C−O hydrogenolysis ( ) turnover rates (523 K, 5 kPa 1-butanol, 2 MPa H2, 1 kPa
CO). (b) Effect of CO pressure on the ratio of C−O to decarbonylation (523 K, 5 kPa 1-butanol, 2 MPa H2). Dashed lines represent trends.
(via reaction with an additional H*) in Step 21 in the second
route. The ΔH⧧ value for H*-assisted CH3CH2CH*OH
ysis turnover rates increased by a factor of 30 with increasing
cluster size (0.7−7 nm), while C−C hydrogenolysis turnover
rates decreased about 5-fold (Figure 7a) at conditions (1 kPa
CO) that lead to significant CO* coverages (based on
regressed KCO values in Sections 3.2 and 3.4), which implicates
the preferential involvement of exposed atoms of different
coordination in these two reactions. C−O hydrogenolysis
occurs preferentially on exposed atoms with high coordination
that prevail on low-index planes and larger clusters,52 while
lower-coordination atoms appear to favor C−C hydrogenolysis,
despite C−C hydrogenolysis rates of alkanes occurring
preferentially on higher-coordinated atoms.50 The ratio of C−
O to C−C hydrogenolysis rates increases with increasing CO
pressure (Figure 7b; 0.7 nm Ir clusters), suggesting that C−O
hydrogenolysis occurs on atoms that bind CO* more weakly
than atoms involved in C−C hydrogenolysis, consistent with
CO* binding energies that are weaker on low-index metal
planes than on corner and edge atoms.53
activation (ΔH⧧ = 74 kJ mol−1) is much smaller, however,
21
than for direct C−O activation in molecularly adsorbed
propanol (Step 16, ΔH⧧ = 171 kJ mol−1), leading to a very
small contribution from Step 16, as can be determined from the
rate ratio of these two steps:
r
r21
16
(ΔH⧧21−ΔH⧧16)/RT
A
16 e
A21
=
(8)
with an exponential term much smaller than unity (2.1 × 10−10
at 523 K). These results indicate that C−O hydrogenolysis of
alkanol-alkanal reactants proceeds predominantly via H*-
assisted C−O activation in CH3CH2CH*OH species. The
C−O bond in propanol weakens upon formation of a C−M
bond during C−H activation to form CH3CH2CH*OH; a
vicinal H* then interacts with the lone pair in the O atom to
assist the cleavage of the C−O bond. This mechanism is
consistent with the independence of C−O hydrogenolysis rates
on H2 pressure (Figure 2).
The other four C−O activation routes in Scheme 5 (via
Steps 19, 22, 24, or 25) involve kinetically relevant transition
states with fewer H atoms than alkanol reactants and would
thus lead to C−O hydrogenolysis rates that would decrease as
H2 pressure increases, in contrast with experiments (Figure 2b).
Nevertheless, we consider such steps in detail in the SI (Section
S7) for completeness and to confirm that these theoretical
treatments can indeed rule out such sequences of elementary
steps based on their activation free energies, in addition to
comparisons with experiments. DFT-predicted effective free
energy barriers confirm that H*-assisted activation of
CH3CH2CH*OH is preferred over all other C−O activation
routes. H*-assisted C−O activation proceeds with lower
intrinsic enthalpy barriers than unassisted C−O activation in
ring opening hydrogenolysis of heterocyclic oxygenates,51 but
was not considered for C−O activation in alkanol reactants in
previous work, which only considered “unassisted” C−O
cleavage reactions.14,17
These effects of Ir cluster size on decarbonylation and C−O
hydrogenolysis turnover rates reflect the effects of surface
coordination on the lumped rate constants α (eq 2) and β (eq
6) for decarbonylation and C−O hydrogenolysis, respectively,
and the CO adsorption constants (KCO). The lumped rate
constants depend upon the free energies of surface-bound
transition states (G⧧
for decarbonylation or G⧧
for C−
CH,3
C−O
O hydrogenolysis) and gas-phase species (see Section S8 (SI)
for more details). Bond-order conservation considerations
would suggest that low-coordination atoms bind all adsorbed
species (including transition states) more strongly;54 thus, C−
O and C−C hydrogenolysis rates would be higher on smaller
clusters, but only at conditions of low CO* coverages, for
which α and β solely determine turnover rates, but not for the
turnover rates reported in Figure 7a measured at conditions (1
kPa CO) that lead to significant CO* coverages.
At high CO* coverages, it is the relative effects of
coordination on α or β, and on KCO (and thus on G⧧
or
CH,3
G⧧C−O, and GCO*) that determine their preference for atoms
with a given coordination (see Section S8 (SI) for more
details). Higher decarbonylation turnover rates on smaller
particles indicate that low coordination atoms favor the binding
3.5. Effects of Ir Cluster Size on Decarbonylation and
C−O Hydrogenolysis Turnover Rates. C−O hydrogenol-
H
DOI: 10.1021/jacs.5b05361
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
of its kinetically relevant transition state over the two CO*
species that must be removed to bind it, while the opposite is
the case for C−O hydrogenolysis reactions on Ir clusters. These
considerations are illustrated by a reaction coordinate in Figure
S16 and can be related to the differences in the binding
strength of the two transition states. The kinetically relevant
transition state for decarbonylation has two C−M bonds (of
204 and 232 pm) and one H−M bond (163 pm) (Figure 8a),
even higher because the Langmuir−Hinshelwood formalism
used to derive eqs 2 and 6 assume that rate and equilibrium
constants (α and KCO in eq 2) remain unchanged as CO*
coverages increase. Yet, CO* species impose repulsive forces
on vicinal CO* or transition states, rendering such species less
stable, as confirmed by theoretical treatments.53,55 Such CO*
coverage effects would cause the α and KCO parameters in the
rate equation (eq 2) to become smaller as CO* coverages
increase with increasing CO pressure and thus invalidate the
assumptions underlying the derivation of these Langmuir−
Hinshelwood rate equations. If increasing CO* coverages
decrease ΔGads,CO values (and thus KCO) more than they
decrease ΔGeff,C−C (and thus α, eq 3), KCO would be
underestimated by eqs 2 and 6, indicating that actual CO*
coverages are closer to 1 ML than those reported above (0.4
ML). Thus, the expectation for high CO* coverages suggests
that the kinetically relevant transition state for decarbonylation
has a smaller “footprint” on the catalyst surface than a pair of
coadsorbed CO* species, partially due to the intact C−C bond
of the transition state (whereas a pair of coadsorbed CO*
species are not covalently bound to one another). These
speculations also reveal the difficulties of applying Langmuir−
Hinshelwood based rate equations to systems with strong co-
adsorbate interactions that may confound kinetic analyses and
lead to inconsistencies between coverages determined by
kinetics, spectroscopy data, and theory.
3.6. Comparisons of 1-Butanol Turnover Rates and
Selectivities on Pt, Ru, Ir, and Cu. Decarbonylation
consumes less H2 stoichiometrically than C−O hydrogenolysis,
making the former a seemingly attractive deoxygenation route.
Decarbonylation, however, forms CO molecules that strongly
inhibit both deoxygenation routes. We consider in this section
C−O hydrogenolysis selectivities on Ru, Pt, and Cu metal
surfaces. Decarbonylation rates were 10−100 times higher than
C−O hydrogenolysis rates on Ru/SiO2 (0.6 nm), Pt/SiO2 (0.7
nm), and Ir/SiO2 (0.7 nm) (Figure 9b), indicating that
decarbonylation is the main decomposition pathway on Group
VIII metals.15,17 CO strongly inhibited decarbonylation and C−
O hydrogenolysis rates on Pt and Ru (Figure 9a) as on Ir,
Figure 8. Structures for the kinetically relevant transition states for (A)
decarbonylation (Step 2.6 of Scheme 2) and (B) C−O hydrogenolysis
(Step 4.4 of Scheme 4). Notable bond distances shown in pm.
whereas the transition state for C−O hydrogenolysis contains
just one C−M bond (213 pm) and one H−M bond (183 pm)
(Figure 8b). The number and lengths of the bonds each
kinetically relevant transition state forms to the surface indicate
that the decarbonylation transition state is more strongly bound
to the surface than the transition state for C−O hydrogenolysis,
which proceeds via activation of the C−O bond by vicinal H*
instead of metal-atom insertion into a C−H bond, as in
decarbonylation.
The KCO values regressed from eqs 2 and 6 for decarbon-
ylation (0.63 0.02 kPa−1) and C−O hydrogenolysis (0.47
0.06 kPa−1) suggest that coverages are 0.3−0.4 ML at 1 kPa CO
and 523 K on Ir clusters (0.7 nm), indicating that the observed
size effects reflect the stabilization of transition states and CO*
on crowded surfaces. In fact, CO* coverages are likely to be
Figure 9. (a) Effect of CO (generated from 1-butanol/butanal conversion) pressure on decarbonylation turnover rates. (b) Effect of 1-butanol/
◆
▲
butanal pool conversion on decarbonylation/C−O hydrogenolysis rate ratio on 1 wt % Ir/SiO2 (0.7 nm, ), 1 wt % Pt/SiO2 (0.7 nm, ), and 0.5
■
wt % Ru/SiO2 (0.6 nm, ) (483 K, 2 MPa H2, 5 kPa 1-butanol). Dashed lines represent trends.
I
DOI: 10.1021/jacs.5b05361
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Figure 10. Decarbonylation turnover rate (a) and methane/propane ratio (b) with changing 1-butanol/butanal pool conversion on 1 wt % Ir/SiO2
◆
■
▲
(0.7 nm, ), 1 wt % Pt/SiO2 (0.7 nm, ), and 0.5 wt % Ru/SiO2 (0.6 nm, ) (483 K, 2 MPa H2, 5 kPa 1-butanol). Dashed lines represent trends.
Figure 11. Effect of 1-butanol/butanal pool conversion on (a) decarbonylation turnover rates, and (b) C−O hydrogenolysis/decarbonylation rate
◆
■
●
ratio on 5 wt % Cu/SiO2 (5.0 nm, ), 1 wt % Ir/SiO2 (0.7 nm, ), 3 wt % Ir/SiO2 (7.1 nm, x), 0.5 wt % Ru/SiO2 (0.6 nm, ), (523 K, 2 MPa H2, 5
▲
kPa 1-butanol), and 1 wt % Pt/SiO2 (0.7 nm, ), (498 K, 2 MPa H2, 5 kPa 1-butanol). Dashed lines represent trends.
which is reflected by decreasing turnover rates with increasing
CO pressure (with increasing 1-butanol/butanal pool con-
version). Such inhibition is expected from large CO* binding
energies on these metals (−166 kJ mol−1 on Ir(111), −129 kJ
mol−1 on Pt(111), −162 kJ mol−1 on Ru(0001) calculated in
this work for isolated CO*). Similar considerations lead us to
conclude that strong CO inhibition effects will also prevail on
Rh, Ni, and Pd catalysts (CO binding energies: −153 kJ mol−1
on Rh(111), −119 kJ mol−1 on Ni(111), and −104 kJ mol−1 on
Pd(111), as calculated by DFT).
CO* species formed from decarbonylation can undergo
secondary reactions with H2 to form CH4. The CH4/C3H8 ratio
in the products represents the fraction of CO* that undergoes
secondary methanation reactions, which increases with
increasing conversion (Figure 10b). CH4/C3H8 ratios were
much larger on Ru than on Ir or Pt (483 K, 5 kPa 1-butanol, 2
MPa H2, Figure 10b); such higher CH4 selectivities are
consistent with methanation turnover rates that are ∼10-fold
higher on Ru than on Pt and Ir clusters.56 Surfaces that
scavenge CO* via reactions with hydrogen to form CH4 will
decrease CO pressures and CO* coverages, thus decreasing
CO-inhibition effects (Figure 10a). Methanation, as the CO
scavenging route, however, gives rise to H2 consumption
stoichiometries
C3H7CH2OH + 2H2 → CH4 + C3H8 + H2O
(9)
that are even higher than for C−O hydrogenolysis
C3H7CH2OH + H2 → C4H10 + H2O
(10)
Thus, strategies that combine decarbonylation with methana-
tion are more H2-intensive than C−O hydrogenolysis, a
reaction that altogether avoids CO formation and thus its
strong inhibition effects.
J
DOI: 10.1021/jacs.5b05361
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Measured decarbonylation turnover rates on Cu/SiO2 (5.0
nm) were >100 times lower than on Ir (0.7 or 7.1 nm) and Ru
(0.6 nm) clusters at 523 K and Pt clusters (0.7 nm) at 483 K
(Figure 11a). The lower decarbonylation turnover rates on Cu
may reflect the required formation of two M−C bonds at the
transition state (on Ir, Figure 8A) that mediate this reaction
and the weak nature of such bonds at Cu surfaces; C* binding
energies (BEC*) are significantly lower on Cu(111) (−438 kJ
mol−1) than on Ru(0001) (−698 kJ mol−1), Ir(111) (−688 kJ
mol−1) or Pt(111) (−685 kJ mol−1). These low C−C
hydrogenolysis rates also appear to reflect the much weaker
binding of CO* on Cu (111) (−46 kJ mol−1) than on Ru
(0001), Ir (111), or Pt (111) (−162 kJ mol−1, −166 kJ mol−1,
−129 kJ mol−1, respectively). C−O hydrogenolysis rates, on
the other hand, vary much less with changing metal
composition than those of decarbonylation. The very low
decarbonylation rates on Cu in turn lead to high C−O
hydrogenolysis selectivities (ratios of C−O to C−C hydro-
genolysis >15, Figure 11b). This correlation is plausible given
that the C−C cleavage transition state (on Ir, Figure 8A)
involves two C−M bonds, while the C−O cleavage transition
state includes only one (on Ir, Figure 8B). These C−C cleavage
transition states thus become much less stable than C−O
cleavage transition states (which may have an additional M−O
bond depending on the mechanism) on surfaces that bind C*
weakly, as on Cu surfaces, in comparison to Ir, Pt, or Ru
surfaces. This leads to lower concentrations of the C−C
cleavage transition states compared to that of C−O cleavage
and consequently to higher C−O hydrogenolysis selectivities
on Cu.
Among the metals examined here, only Cu selectively
catalyzes C−O hydrogenolysis. C−O hydrogenolysis preserves
valuable C atoms in the products and avoids strong inhibition
by CO*, making it the preferred route for deoxygenation of
alkanol/alkanal mixtures. Other Group IX metals (Ag and Au)
also bind C* weakly (BEC*: −282 kJ mol−1 and −382 kJ mol−1
on Ag (111) and Au (111)) compared to Cu (111) (BEC*:
−438 kJ mol−1) or Group VIII metals (Ru, Rh, Pd, Os, Ir, and
Pt) (BEC*: − 616 to −702 kJ mol−1) and thus are expected to
exhibit low C−C cleavage rates. Their significantly weaker M−
C bonds on terrace sites are also expected to lead to low C−O
hydrogenolysis turnover rates, however, requiring the presence
of low-coordination surface atoms that prevail on small Ag and
Au clusters to achieve high C−O hydrogenolysis selectivities.
significantly weakens the adjacent C−C bond. DFT calculations
demonstrate that C−O hydrogenolysis proceeds through an
H*-assisted C−O activation of the CH3CH2CHOH* inter-
mediate, leading to rates proportional to alkanol pressure and
independent of H2 pressure, consistent with experimental
measurements. C−O hydrogenolysis turnover rates become
comparable to decarbonylation rates as the cluster size of Ir is
increased from 0.7 to 7 nm because C−O hydrogenolysis
preferentially takes place on high-coordination Ir sites due to
strongly adsorbed CO* blocking corner/edge sites. The use of
large metal clusters, however, is inefficient as the majority of the
precious metal is inaccessible. CO inhibition can instead be
alleviated by coupling decarbonylation with methanation on
metals such as Ru. Methanation, however, has a higher H2
requirement than C−O hydrogenolysis. Group VIII metals are
not ideal deoxygenation catalysts if C−C activations and H2
consumption need to be minimized. Cu, on the other hand, is
unique in its ability to catalyze C−O hydrogenolysis selectively
over C−C hydrogenolysis because of weak binding of C* and
CO*. This is crucial for the conversion of many biomass-
derived intermediates with alkanol/alkanal functionalities to
chemicals and fuels.
ASSOCIATED CONTENT
* Supporting Information
■
S
The Supporting Information is available free of charge on the
ACS Publications website at DOI: 10.1021/jacs.5b05361.
Methodological details, figures containing DFT-derived
structures for all calculations herein, and further
discussions of ruled-out C−O hydrogenolysis mecha-
nisms, particle size effects, and the low coverages of H*
on Ir surfaces (PDF)
AUTHOR INFORMATION
Corresponding Author
*iglesia@berkeley.edu
■
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
This material is based upon work supported by the U.S.
Department of Energy, Office of Science, Office of Basic Energy
Sciences, Chemical Sciences Division of the US Department of
Energy as well as by XSEDE (CHE130022) and EMSL
(Proposal 47800) for computational resources. The authors
acknowledge Mr. Will Knaeble and Dr. Shuai Wang for helpful
discussions.
4. CONCLUSIONS
Decarbonylation is the main decomposition pathway of 1-
butanol on Group VIII metals, such as Ir, Pt, and Ru. The main
product of decarbonylation, CO, is a strong inhibitor, resulting
in decreasing turnover rates with alkanol/alkanal pool
conversion. Analyses of decarbonylation turnover rates indicate
that at the reaction conditions used in this study (1−5 kPa 1-
butanol, 1−3 MPa H2, and 1−10 kPa CO at 523 K), Ir clusters
are covered with chemisorbed CO* species, and the kinetically
relevant transition state has three fewer H atoms than 1-
butanol. These kinetic data, combined with DFT calculations of
reaction and activation enthalpies across a series of RCHO*,
RCOH*, and RCO* intermediates (R = CH3CH2, CH3CH,
CH3C), indicate that C−C cleavage proceeds in CH3C*C*O
intermediates following the kinetically relevant C−H activation
of CH3CH2CO* to form CH3CHCO*. C−C hydrogenolysis in
1-butanol/butanal occurs only at the C−C bond vicinal to the
O atom because the CO double bond in RCCO*
REFERENCES
■
(1) Roman-Leshkov, Y.; Barrett, C. J.; Liu, Z. Y.; Dumesic, J. A.
Nature 2007, 447, 982.
(2) Chheda, J. N.; Huber, G. W.; Dumesic, J. A. Angew. Chem., Int.
Ed. 2007, 46, 7164.
(3) Davda, R. R.; Shabaker, J. W.; Huber, G. W.; Cortright, R. D.;
Dumesic, J. A. Appl. Catal., B 2005, 56, 171.
(4) Braunchweig, B.; Hibbitts, D.; Neurock, M.; Wieckowski, A.
Catal. Today 2013, 202, 197.
(5) Hibbitts, D. D.; Neurock, M. J. Catal. 2013, 299, 261.
(6) Davda, R. R.; Shabaker, J. W.; Huber, G. W.; Cortright, R. D.;
Dumesic, J. A. Appl. Catal., B 2003, 43, 13.
(7) Boga, D. A.; Oord, R.; Beale, A. M.; Chung, Y.-M.; Bruijnincx, P.
C. A.; Weckhuysen, B. M. ChemCatChem 2013, 5, 529.
K
DOI: 10.1021/jacs.5b05361
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
(8) Kunkes, E. L.; Simonetti, D. A.; West, R. M.; Serrano-Ruiz, J. C.;
(48) Cimino, A.; Boudart, M.; Taylor, H. J. Phys. Chem. 1954, 58,
796.
Gartner, C. A.; Dumesic, J. A. Science 2008, 322, 417.
̈
(49) Sinfelt, J. H.; Taylor, W. F.; Yates, D. J. C. J. Phys. Chem. 1965,
69, 95.
(50) Flaherty, D.; Uzun, A.; Iglesia, E. J. Phys. Chem. C 2015, 119,
2597.
(51) Tan, Q.; Hibbitts, D.; Neurock, M.Unpublished results, 2014
(52) Van Hardeveld, R.; Hartog, F. Surf. Sci. 1969, 15, 189.
(53) Allian, A.; Takanabe, K.; Fujdala, K.; Hao, X.; Truex, T.; Cai, J.;
Buda, C.; Neurock, M.; Iglesia, E. J. Am. Chem. Soc. 2011, 133, 4498.
(54) Shustorovich, E. Adv. Catal. 1990, 37, 101.
(55) Loveless, B. T.; Buda, C.; Neurock, M.; Iglesia, E. J. Am. Chem.
Soc. 2013, 135, 6107.
(9) Davis, J. L.; Barteau, M. A. J. Am. Chem. Soc. 1989, 111, 1782.
(10) Davis, J. L.; Barteau, M. A. Surf. Sci. 1990, 235, 235.
(11) Davis, J. L.; Barteau, M. A. Surf. Sci. 1987, 187, 387.
(12) Gates, S. M.; Russell, J. N.; Yates, J. T. Surf. Sci. 1986, 171, 111.
(13) Sexton, B. A.; Rendulic, K. D.; Hughes, A. E. Surf. Sci. 1982, 121,
181.
(14) Ferrin, P.; Simonetti, D.; Kandoi, S.; Kunkes, E.; Dumesic, J. A.;
Nørskov, J. K.; Mavrikakis, M. J. Am. Chem. Soc. 2009, 131, 5809.
(15) Mavrikakis, M.; Barteau, M. A. J. Mol. Catal. A: Chem. 1998,
131, 135.
(16) Gursahani, K. I.; Alcala,
Catal., A 2001, 222, 369.
(17) Alcala, R.; Mavrikakis, M.; Dumesic, J. A. J. Catal. 2003, 218,
178.
́
R.; Cortright, R. D.; Dumesic, J. A. Appl.
(56) Vannice, M. A. J. Catal. 1975, 37, 449.
́
(18) Kandoi, S.; Greeley, J.; Simonetti, D.; Shabaker, J.; Dumesic, J.
A.; Mavrikakis, M. J. Phys. Chem. C 2011, 115, 961.
(19) Liu, B.; Greeley, J. J. Phys. Chem. C 2011, 115, 19702.
(20) Flaherty, D. W.; Hibbitts, D. D.; Gurbuz, E. I.; Iglesia, E. J. Catal.
̈
̈
2014, 311, 350.
(21) Flaherty, D. W.; Iglesia, E. J. Am. Chem. Soc. 2013, 135, 18586.
(22) Sitthisa, S.; Resasco, D. Catal. Lett. 2011, 141, 784.
(23) Sitthisa, S.; Sooknoi, T.; Ma, Y.; Balbuena, P. B.; Resasco, D. E.
J. Catal. 2011, 277, 1.
(24) Deutsch, K. L.; Shanks, B. H. Appl. Catal., A 2012, 447−448,
144.
(25) Deutsch, K. L.; Shanks, B. H. Appl. Catal., A 2014, 470, 390.
(26) Soled, S. L.; Malek, A.; Miseo, S.; Baumgartner, J.; Kliewer, C.;
Afeworki, M.; Stevens, P. A. Stud. Surf. Sci. Catal. 2006, 162, 103.
(27) Miller, J. T.; Schreier, M.; Kropf, A. J.; Regalbuto, J. R. J. Catal.
2004, 225, 203.
(28) McVicker, G. B.; Baker, R. T. K.; Garten, R. L.; Kugler, E. L. J.
Catal. 1980, 65, 207.
(29) Choi, M.; Wu, Z.; Iglesia, E. J. Am. Chem. Soc. 2010, 132, 9129.
(30) Handbook of Heterogenous Catalysis; Ertl, G., Knozinger, H.,
Schuth, F., Weitkamp, J., Eds.; John Wiley and Sons: New York, 2008.
(31) Kresse, G.; Furthmuller, J. Comput. Mater. Sci. 1996, 6, 15.
(32) Kresse, G.; Furthmuller, J. Phys. Rev. B: Condens. Matter Mater.
Phys. 1996, 54, 11169.
(33) Kresse, G.; Hafner, J. Phys. Rev. B: Condens. Matter Mater. Phys.
1993, 47, 558.
(34) Kresse, G.; Hafner, J. Phys. Rev. B: Condens. Matter Mater. Phys.
1994, 49, 14251.
(35) Blochl, P. E. Phys. Rev. B: Condens. Matter Mater. Phys. 1994, 50,
17953.
(36) Kresse, G.; Joubert, D. Phys. Rev. B: Condens. Matter Mater. Phys.
1999, 59, 1758.
(37) Hammer, B.; Hansen, L.; Norskov, J. K. Phys. Rev. B: Condens.
Matter Mater. Phys. 1999, 59, 7413.
(38) Perdew, J.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77,
3865.
(39) Zhang, Y.; Yang, W. Phys. Rev. Lett. 1998, 80, 890.
(40) Henkelman, G.; Jonsson, H. A. J. Chem. Phys. 2000, 113, 9978.
(41) Jonsson, H. A.; Mills, G.; Jacobsen, K. W. Nudged Elastic Band
Method for Finding Minimum Energy Paths of Transitions; World
Scientific: Hackensack, NJ, 1998.
(42) Henkelman, G.; Jonsson, H. A. J. Chem. Phys. 1999, 111, 7010.
(43) Mavrikakis, M.; Barteau, M. A. J. Mol. Catal. A: Chem. 1998,
131, 135.
(44) Huber, G. W.; Shabaker, J. W.; Dumesic, J. A. Science 2003, 300,
2075.
(45) Goel, S.; Zones, S. I.; Iglesia, E. J. Am. Chem. Soc. 2014, 136,
15280.
(46) Lee, K.; Morikawa, Y.; Langreth, D. C. Phys. Rev. B: Condens.
Matter Mater. Phys. 2010, 82, 155461.
(47) Maire, G.; Plouidy, G.; Prudhomme, J. C.; Gault, F. G. J. Catal.
1965, 4, 556.
L
DOI: 10.1021/jacs.5b05361
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX