Condensation and esterification reactions of alkanals, alkanones, and alkanols on TiO2: Elementary steps, site requirements, and synergistic effects of bifunctional strategies

Article abstract of DOI:10.1016/j.jcat.2016.05.026

Rates and selectivity of TiO₂-catalyzed condensation of C₃ oxygenates (propanal, acetone) are limited by ubiquitous effects of side reactions, deactivation, and thermodynamic bottlenecks. H₂ together with a Cu function, present as physical mixtures with TiO₂, circumvents such hurdles by scavenging unsaturated intermediates. They also render alkanols and alkanals/alkanones equivalent as reactants through rapid interconversion, while allowing esterification turnovers by dehydrogenating unstable hemiacetals. Oxygenates form molecules with new C-C and C-O bonds and fewer O-atoms at nearly complete conversions with stable rates and selectivities. Kinetic, isotopic, and theoretical methods showed that rates are limited by α-C-H cleavage from carbonyl reactants to form enolate intermediates, which undergo C-C coupling with another carbonyl species to form α,β-unsaturated oxygenates or with alkanols to form hemiacetals with new C-O bonds, via an intervening H-shift that forms alkoxide-alkanal pairs. Titrations with 2,6-di-tert-butylpyridine, pyridine, CO₂, and propanoic acid during catalysis showed that Lewis acid-base site pairs of moderate strength mediate enolate formation steps via concerted interactions with the α-H atom and the enolate moiety at transition states. The resulting site-counts allow rigorous comparisons between theory and experiments and among catalysts on the basis of turnover rates and activation free energies. Theoretical treatments give barriers, kinetic isotope effects, and esterification/condensation ratios in excellent agreement with experiments and confirm the strong effects of reactant substituents at the α-C-atom and of surface structure on reactivity. Surfaces with Ti-O-Ti sites exhibiting intermediate acid-base strength and Ti-O distances, prevalent on anatase but not rutile TiO₂, are required for facile α-C-H activation in reactants and reprotonation of the adsorbed intermediates that mediate condensation and esterification turnovers.

Full text of DOI:10.1016/j.jcat.2016.05.026

Journal of Catalysis 340 (2016) 302–320
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Condensation and esterification reactions of alkanals, alkanones, and
alkanols on TiO : Elementary steps, site requirements, and synergistic
2
effects of bifunctional strategies
1
⇑
Shuai Wang, Konstantinos Goulas , Enrique Iglesia
Department of Chemical and Biomolecular Engineering, University of California at Berkeley, Berkeley, CA 94720, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
Rates and selectivity of TiO -catalyzed condensation of C oxygenates (propanal, acetone) are limited by
2
3
Received 14 April 2016
Revised 27 May 2016
Accepted 29 May 2016
ubiquitous effects of side reactions, deactivation, and thermodynamic bottlenecks. H
2
together with a Cu
function, present as physical mixtures with TiO , circumvents such hurdles by scavenging unsaturated
2
intermediates. They also render alkanols and alkanals/alkanones equivalent as reactants through rapid
interconversion, while allowing esterification turnovers by dehydrogenating unstable hemiacetals.
Oxygenates form molecules with new CAC and CAO bonds and fewer O-atoms at nearly complete con-
versions with stable rates and selectivities. Kinetic, isotopic, and theoretical methods showed that rates
Keywords:
TiO
2
are limited by
CAC coupling with another carbonyl species to form
hemiacetals with new CAO bonds, via an intervening H-shift that forms alkoxide–alkanal pairs. Titrations
with 2,6-di-tert-butylpyridine, pyridine, CO , and propanoic acid during catalysis showed that Lewis
acid–base site pairs of moderate strength mediate enolate formation steps via concerted interactions
with the -H atom and the enolate moiety at transition states. The resulting site-counts allow rigorous
a
-CAH cleavage from carbonyl reactants to form enolate intermediates, which undergo
Aldol condensation
Esterification
Oxygenate catalysis
Enolates
Bifunctional stabilization
Density functional theory
Site titrations
a,b-unsaturated oxygenates or with alkanols to form
2
a
comparisons between theory and experiments and among catalysts on the basis of turnover rates and
activation free energies. Theoretical treatments give barriers, kinetic isotope effects, and esterification/-
condensation ratios in excellent agreement with experiments and confirm the strong effects of reactant
substituents at the
ing intermediate acid–base strength and TiAO distances, prevalent on anatase but not rutile TiO
required for facile -CAH activation in reactants and reprotonation of the adsorbed intermediates that
mediate condensation and esterification turnovers.
a
-C-atom and of surface structure on reactivity. Surfaces with TiAOATi sites exhibit-
2
, are
a
Ó 2016 Published by Elsevier Inc.
1
. Introduction
2 3 2
Metal oxides (e.g., MgO [12,13], MgOAAl O [14–17], ZrO
[
6,18], CeO AZrO [19,20], and TiO [21–23]) catalyze aldol con-
2
2
2
Aldol condensations of carbonyl compounds form conjugated
b-unsaturated oxygenates with longer chains and lower O-content
a
,
densation reactions. On these oxide catalysts, reactions are limited
by thermodynamics [15], leading to low equilibrium concentra-
than in the reactant molecules [1–3]. Such products can be used as
tions of a,b-unsaturated carbonyl compounds; these compounds
chemical intermediates or converted via H
2
-mediated deoxygena-
tend to undergo subsequent condensation and cyclization reac-
tion reactions to alkanes and alkenes, and are useful as fuels and
as chemical intermediates within existing market infrastructures
tions that form unreactive residues and block active sites [22,23].
With added H , metal particles supported on the oxide catalysts
2
[
4–11]. Aldol condensations represent a potential enabling strategy
(e.g., Pt [17], Pd [17,24], and Cu [25]) hydrogenate these unsatu-
rated primary products to more stable saturated products, thus
removing thermodynamic bottlenecks and inhibiting deactivation;
these effects of the hydrogenation function, however, have not
been assessed quantitatively or systematically using physical mix-
tures that allow the independent optimization of the condensation
and hydrogenation functions.
for upgrading biomass-derived feedstocks, which contain fewer
C-atoms and more O-atoms than their desired conversion products
[
2,3].
⇑
Corresponding author.
E-mail address: iglesia@berkeley.edu (E. Iglesia).
Current address: Catalysis Center for Energy Innovation, University of Delaware,
1
Aldol condensations on oxides involve enolate formation via
CAH activation at the
a-position to the carbonyl group (C@O) in
Newark, DE 19716, USA.
http://dx.doi.org/10.1016/j.jcat.2016.05.026
0
021-9517/Ó 2016 Published by Elsevier Inc.

S. Wang et al. / Journal of Catalysis 340 (2016) 302–320
303
alkanals and alkanones, subsequent nucleophilic attack by another
carbonyl compound to form a new CAC bond in the aldol formed,
and dehydration of the aldol intermediate and desorption as a gas-
eous conjugated enal or enone [26], in steps resembling those in
their homogeneous reaction analogs [25]. The inhibition of acetone
condensation rates on MgO by added pyridine and acetic acid indi-
cates that both acid and basic sites may be involved [12]; infrared
These mechanistic conclusions and practical bifunctional strategies
appear to be general for C –C alkanals and alkanones and their
respective alkanols on TiO surfaces and to condensation and ester-
ification reactions catalyzed by monoclinic and tetragonal ZrO
[29], thus providing general predictive guidance for the complex
reaction networks involved in the upgrading of their biomass-
derived mixtures.
2
5
2
2
spectra have shown that basic O-atoms abstract the a-H-atoms in
carbonyl compounds, while acid centers can stabilize enolates via
interactions with the O-atoms of the enolates [18,27]. The kinetic
relevance of these steps remains unconfirmed by kinetic, isotopic,
or spectroscopic methods and based solely on theoretical treat-
2
. Methods
2.1. Catalyst synthesis and characterization
2 2
ments on ZrO and CeO surfaces [28].
Cu/SiO
position–precipitation method [30], and used as a co-catalyst
together with H to improve aldol condensation rates, selectivity,
and stability. Colloidal silica (30 wt.%, 21.8 g, LUDOX SM-30), Cu
NO O (99.99%, 6.0 g, Sigma–Aldrich) and urea (99%, 4.7 g,
ꢁ2.5H
Aldrich) were dissolved in deionized water (17.9 M resistivity,
2
(ꢀ20 wt.% Cu) was prepared using a homogeneous de
The ubiquitous thermodynamic and deactivation hurdles have
prevented detailed kinetic and isotopic analyses of the elementary
steps that mediate these reactions. These matters also preclude the
identification and counting of the active sites during catalysis,
essential to compare catalysts based on turnover rates, and an
accurate measurement of the activation free energies, which are
required for theoretical benchmarking of proposed mechanisms
and turnover rates.
2
(
3
)
2
2
X
3
1
HNO
00 cm ) and the suspension pH was adjusted to 2–3 using a
ꢂ3
3
solution (0.5 mmol cm ; Sigma–Aldrich, 99%). This colloidal
ꢂ1
suspension was then heated to 363 K (at 0.167 K s ) and held for
In this study, we use physical mixtures of Cu/SiO
P25) (a mixture of anatase (TiO (a)) and rutile (TiO
phases, anatase/rutile ratio of 3), pure TiO (a), or TiO (r) as bifunc-
tional catalysts for the aldol condensation of C oxygenates (propa-
nal, 1-propanol, acetone, and 2-propanol). The presence of a Cu
function and of H leads to higher turnover rates and much slower
2
with TiO
2
(-
2
0 h while stirring (12 Hz). The powders were recovered by vac-
2
2
(r)) crystal
uum filtration and washed with deionized water until the filtrate
pH was 7, treated in ambient stagnant air at 383 K for 20 h, and
2
2
3
3
ꢂ1 ꢂ1
then heated in flowing dry air (99.999%, 1.67 cm g
s , Praxair)
ꢂ1
to 723 K (at 0.167 K s ) and held for 5 h. These samples were trea-
2
3
ꢂ1 ꢂ1
ted in flowing 10% H
2
/He (99.999%, 5.56 cm g
s , Praxair) by
deactivation by overcoming thermodynamic hurdles and scaveng-
ing unsaturated precursors to unreactive residues. The Cu function
also enables the equilibration of alkanals and alkanones with their
respective alkanols, thus allowing their interchangeable use as
reactants. The unprecedented stability conferred by the Cu func-
tion allows detailed kinetic measurements, rigorous comparisons
ꢂ1
heating to 573 K (at 0.033 K s ) and held for 2 h, and passivated
3
ꢂ1 ꢂ1
in flowing 1% O
at ambient temperature for 1 h before exposure to ambient air.
Metallic Cu particles in the resulting Cu/SiO catalysts were
identified using powder X-ray diffraction (XRD) measurements
Cu radiation, k = 0.15418 nm, 40 kV, 40 mA, Bruker D8
2
/He mixtures (99.999%, 0.83 cm g
s , Praxair)
2
(
Ka
2 2
of areal rates on TiO (a) and TiO (r), site titrations during catalysis
Advance; diffractogram in Supporting information (SI)). The mean
(
with CO , 2,6-di-tert-butylpyridine (DTBP), pyridine, and propa-
2
crystallite size (d) of Cu was estimated using the Scherrer equation
noic acid), and mechanistic studies of the elementary steps and
active site structures that mediate these reactions.
ꢂ1
(
(
d = 0.90ꢁkꢁ(bcosh) ), in which b is the full width at half maximum
2
FWHM) of the diffraction peak at 2h. For the resulting Cu/SiO cat-
Acid–base site pairs of intermediate strength and site distance
alysts, the d value of Cu was determined to be 7.9 nm based on the
strongest Cu diffraction peak from the (111) plane (2h = 43.3°) [8].
prevalent on TiO
for its higher reactivity and stability than TiO
2
(a), probed by titrations during catalysis, account
(r), as evidenced
2
2
ꢂ1
TiO
2
catalysts, including P25 (TiO
2
(P25), 99.8%, 50 m g , ana-
(TiO (a), 99.7%,
(TiO (r), 99.5%,
Aldrich), were treated in flowing air (99.999%,
from the reversible rate inhibition by pyridine and propanoic acid,
but the absence of inhibition by selective titrants of only basic
tase:rutile = 3:1 mass, Degussa), anatase TiO
2
2
2
ꢂ1
ꢂ1
ꢂ1 ꢂ1
2
1
1
40 m g
,
,
Alfa Aesar), and rutile TiO
2
2
(
2
CO ) or only Brønsted acid (DTBP) sites. The number of acid–base
2
60 m g
site pairs was measured by titration with propanoic acid during
condensation reactions, thus allowing the first rigorous measure-
ments of intrinsic turnover rates and activation free energies. Con-
version rates were proportional to reactant pressures for all
carbonyl species and showed normal kinetic isotope effects, con-
3
ꢂ1
.67 cm g
s
, Praxair) by heating to 673 K (at 0.167 K s ) and
and Cu/SiO physical mixtures (TiO + Cu/
= 0.1–2.0 mass) were obtained by crushing
and mixing the two catalysts with a mortar and pestle, and then
holding for 3 h. TiO
SiO , (Cu/SiO )/TiO
2
2
2
2
2
2
pressed, crushed, and sieved to retain 180–250 lm particles.
sistent with the kinetic relevance of a-CAH cleavage to form eno-
late intermediates; these species then react with a carbonyl
compound to form condensation products or (in the case of alka-
nals) with a terminal alkanol to form esters. These conclusions
were confirmed by the observed effects of alkanol/alkanal ratios
on the ratio of esterification/condensation rates.
2.2. Catalytic rate measurements
2 2 2
TiO and TiO + Cu/SiO mixtures (10–200 mg) were loaded into
a quartz tubular reactor with plug-flow hydrodynamics. The
loaded amounts were chosen to maintain differential reactant con-
Density function theory (DFT) treatments of plausible conden-
sation and esterification elementary steps on TiO
2
(a) cluster mod-
versions (<10%). These samples were treated in flowing 10% H
2
/He
3
ꢂ1 ꢂ1
els are consistent with these conclusions and give excellent
agreement with measured turnover rates and isotope effects for
acetone and propanal reactants and with the observed effects of
substituents on enolate formation rates and on condensation/
(99.999%, 5.56 cm g
s , Praxair) by heating to 543 K (at
ꢂ1
0.0833 K s ), holding for 2 h, and then cooling to reaction temper-
atures (453–523 K) before rate measurements. Temperatures were
set with an electronic controller (Watlow, Series 988) and mea-
sured with a K-type thermocouple (0.05 cm diameter, 16 cm
length, Omega) held at the outer reactor wall.
2
esterification rate ratios. Similar treatments on TiO (r) cluster
models show that condensation rates are limited by reprotonation
and desorption of the dimer species formed in carbonyl-enolate
coupling steps; such markedly different reactivities in rutile and
3
C oxygenates, including 1-propanol (>99.9%, Sigma–Aldrich),
propanal (>97%, Sigma–Aldrich), 2-propanol (>99.5%, Sigma–Aldrich),
and acetone (>99.9%, Fisher), were introduced into a He stream
(99.999%, Praxair) using a syringe pump (Cole Parmer, 74900
anatase phases of TiO
acid–base strength of exposed TiAO site pairs on TiO
2
reflect the spatial separation and strong
(r) surfaces.
2

3
04
S. Wang et al. / Journal of Catalysis 340 (2016) 302–320
series); all transfer lines after the liquid introduction point were
held at 433 K to avoid condensation of reactants and products.
2.4. Density functional theory methods
He and H
2
flows were metered using electronic mass flow con-
The most thermodynamically stable and naturally exposed sur-
trollers (Porter, Model 201) to vary reactant pressures and space
velocities. Isotopic experiments were conducted similarly but
2 2
faces for TiO (a) and TiO (r) are the (101) and (110) ones, respec-
tively [31], which provide useful models to study catalytic
processes on the corresponding oxides [32]. In particular, the local
characteristic of catalytically-relevant properties for the Ti and O
sites exposed on these surfaces enables oxide clusters as primary
6 2
using acetone-d (>99.9%, Aldrich) and D (99.8%, Praxair) instead
of the respective protium analog.
Concentrations of reactants and products in the effluent were
measured by on-line gas chromatography (Agilent 6890). Specifi-
cally, oxygenates and hydrocarbons were detected by flame ion-
ization after chromatographic separation in a methyl silicone
models to represent these surfaces [33]. Here Ti
abstracted from the (101) surface of bulk anatase crystal struc-
tures (tetragonal, I4 /amd, a = b = 0.378 nm, c = 0.949 nm) [31]
and the (110) surface of rutile crystals (tetragonal, P4 /amd,
5
O19 clusters were
1
capillary column (Agilent HP-1, 50 m, 0.32 mm ID; 1.05
lm film);
4
other gaseous species (CO, CO , H O, and H ) were separated in a
2
2
2
a = b = 0.459 nm, c = 0.296 nm) [31] and used as model surfaces
in all density functional theory (DFT) calculations (Scheme 1).
Eighteen H-atoms were attached to the dangling O-atoms at the
periphery of each cluster in order to maintain charge neutrality.
The orientation of the OAH bonds was chosen to maintain that
of the TiAO bonds cleaved to form the clusters, so as to avoid extra-
neous H-bonding interactions at reactant and transition states. The
packed column (Agilent Porapak-Q, 4.8 m, 80–100 mesh) and
detected by thermal conductivity. Chemical species were identi-
fied using known standards and speciation by mass spectrometry
after chromatographic separations (HP 5972 GC/MS). Molecular
weight (MW), determined by the molecular ion peak in mass
spectrometry, was used to estimate the molecular formula of a
product. For instance, aldol condensation of alkanals (MW = N)
5 19
five Ti-atoms and the three inner O-atoms in each Ti O H18 clus-
forms
which can be hydrogenated sequentially to alkanals (MW = 2 N-
6) and alkanols (MW = 2 N-14). The Reaction rates are reported
either as rates per TiO surface area (areal rates) or as rates per
a
,b-unsaturated carbonyl compounds (MW = 2 N-18),
ter were allowed to relax during all geometry optimizations, but
the terminal OH groups were frozen to maintain the local coordi-
nation environment of each extended parent surface. These clus-
ters are qualified for DFT treatments of aldol condensation on the
1
2
surface TiAO pair site (turnover rates), for which the exposed
TiAO pair sites were measured by the titration methods described
next in Section 2.3.
2
corresponding TiO surfaces, because of insignificant effects of
neighboring atoms upon catalytic properties of TiAO site pairs
[34] and absence of lateral interaction between adsorbed reactants
at low surface coverages that are prevalent at examined catalytic
conditions as shown in Section 3.4.
2.3. Titration of surface sites by probe molecules during catalysis
DFT simulations of aldol condensation and esterification routes
were carried out at the hybrid B3LYP functional level of theory [35]
using the Gaussian 09 program [36], the standard 6-311G(d,p)
basis set for C, O, and H-atoms [37], and the effective core potential
LANL2DZ basis set for Ti-atoms [38]. An ultrafine grid (99,590) was
set for numerical integration; the Berny geometry algorithm [39]
was selected to optimize structures of reactants, transition states
and products involved in each elementary step with convergence
Pyridine (>99.9%, Sigma–Aldrich), 2,6-di-tertbutylpyridine
(
(
>97%, Aldrich), propanoic acid (>99.5%, Sigma–Aldrich), and CO
2
50% CO
2
/He, 99.999%, Praxair) were used as titrants to assess
the involvement of various types of exposed surface sites in aldol
condensation turnovers. In these experiments, acetone (4 kPa)
was first introduced on a TiO
2 2
+ Cu/SiO physical mixture (1:1
mass) at 523 K and 40 kPa H . After acetone condensation rates
2
ꢂ8
ꢂ5
ꢂ1
reached steady state, the liquid feed was switched to a mixture
of acetone (4 kPa) with each liquid titrant (0.005 titrant/acetone
molar). The reactants were then switched back to pure acetone
criteria of 1.0 ꢃ 10 Ha for energy and 1.5 ꢃ 10 Ha Bohr for
the maximum residual forces on each atom. The optimized struc-
tures were confirmed by frequency calculations at the same com-
putational level in order to ensure that they represent a stable
energy minima (i.e., no imaginary frequency) or a saddle point
for transition states (i.e., one imaginary frequency). The Grimme
D3BJ dispersion correction [40] and the counterpoise correction
for the basis set superposition error (BSSE) [41] were taken into
account in electronic energy calculations. A fitted scaling factor
of 0.9682 was applied to revise DFT-derived vibrational frequen-
cies because of the neglected anharmonicity effects in the theoret-
ical treatment [42]; these revised frequencies were used to derive
vibrational partition functions based on the rigid rotor harmonic
oscillator (RRHO) approximation and to calculate zero-point
energy and thermal corrections [43]. Exceptions were made for
(
4 kPa) to assess the reversibility of titrant binding. In the case of
CO (4 kPa), the titrant was added as a separate gaseous stream
2
after condensation rates reached constant values at 4 kPa acetone.
The concentrations of reactants, products and titrants were mea-
sured using the chromatographic protocols described above. The
number of accessible TiAO site pairs ((TiAO)
TiO catalysts was determined from the amount of propionic acid
required to fully suppress condensation rates, using acid/(TiAO)
stoichiometries of 1 for TiO (a) and TiO (P25) that adsorb car-
boxylic acids molecularly and 0.5 for TiO (r) that adsorbs car-
s
) for the examined
2
s
2
2
2
boxylic acids dissociatively via bidentate carboxylate modes as
revealed from infrared evidence [29]. Specifically, these titrations
were carried out at lower reaction temperatures (453 K) to sup-
press hydrogenation of propionic acid to propanal and 1-
ꢂ1
low-frequency modes (<100 cm ) of loosely bound adsorbates,
for which a free-rotor model was used instead of the RRHO model
in order to avoid significant errors in estimation of vibrational par-
tition functions [44]. Local atomic charges were calculated using
the natural bond analysis as implemented in the Gaussian 09 pro-
gram [45].
2
propanol on Cu/SiO and subsequent condensation and esterifica-
tion of these formed products. This consumption of propionic acid
during the titration was taken into account by measuring these
products formed from propionic acid; this was made possible by
the use of an acid for which hydrogenation forms products distinct
ꢂ
+
The affinities of surfaces for OH and H gaseous ions were used
to probe the acid and basic strengths of exposed acid–base pairs on
from the main reactants. The amounts of such products (nconsume
)
and unreacted acid (noutlet) in the effluent were then subtracted
from the amount of propanoic acid introduced (ninlet) in order to
determine the amount of propanoic acid irreversibly retained by
2
these TiO surfaces. Hydroxide anion affinity (EHA) for an acid (A) is
defined as the energy released upon attachment of a gaseous
ꢂ
ꢂ
hydroxide anion (OH ) onto an isolated A at the surface to form
the TiO
2
surfaces (nuptake) as given by
a bound anion (AOH ):
nuptake ¼ ninlet ꢂ noutlet ꢂ nconsume
ð1Þ
E
HA ¼ EAOH
ꢂ
ꢂ E
A
ꢂ EOH
ꢂ
ð2Þ

S. Wang et al. / Journal of Catalysis 340 (2016) 302–320
305
5 19 2 2
Scheme 1. Structures of Ti O H18 clusters abstracted from anatase TiO (101) and rutile TiO (110) surfaces.
where E
i
represents the electronic energy of species i. Similarly, the
2 2
faster on TiO (a) than on TiO (r), leading to similar areal rates,
proton affinity (EPA) for a base (B) is defined as the energy released
mechanistic conclusions, and active surfaces for TiO (P25) and
2
+
upon attachment of a gaseous proton (H ) onto an isolated B at the
TiO (a) catalysts.
2
+
surface to form a bound cationic moiety (HB ):
Aldol condensation of propanal on TiO (P25) (523 K; 40 kPa H ,
2
2
ꢀ
20% conversion of propanal) formed 2-methyl-pent-2-enal and
E
PA ¼ EHB
þ
ꢂ E
B
ꢂ EH
þ
ð3Þ
C
9
unsaturated alkanals (such as 2,4-dimethyl-hepta-2,4-dienal
and its skeletal isomers) at areal rates of 0.066 and 0.017
l
mol
where E
i
represents the electronic energy of species i. EHA and EPA
ꢂ2 ꢂ1
C-atom m
s , respectively (Table 1). These molecules are the
are used here as descriptors for acid and basic strengths of Lewis
centers, respectively. These affinities of the Ti and O sites exposed
on Ti O H18 clusters were determined using the above calculation
5 19
protocols but with the standard 6-311+G(d,p) basis set for the O and
H-atoms, because these more diffuse functions are required to accu-
rately describe these charged systems [46].
expected products from primary and secondary propanal conden-
sation events (Scheme 2) [50]. Their unsaturated nature and the
absence of 1-propanol, even in the presence of H
TiO (P25) cannot catalyze C@C or C@O hydrogenation via either
activation or H-transfer pathways at these reaction conditions.
2
, indicate that
2
H
2
The approach to equilibrium factor (
of C
H O is given by
g
1
) for aldol condensation
n
carbonyl compounds to form C2n unsaturated carbonyls and
3
. Results and discussion
2
K
aldol
3.1. C
3
oxygenate reactions on TiO
2
and CuATiO
2
catalysts
2C H2nO ¢ C2n
H
O þ H O
ð4Þ
ð5Þ
n
4nꢂ2
2
TiO
2
(P25) is used here to examine rates, selectivities, and deac-
oxygenates on TiO and CuATiO
catalysts, but the trends and conclusions are similar, except for dif-
ferences in reactivity, for TiO (a) and TiO (r) catalysts, as discussed
later. TiO (P25), a TiCl flame hydrolysis product, consists of a
physical mixture of anatase (TiO (a)) and rutile (TiO (r)) crystal-
lites with an anatase/rutile mass ratio of ꢀ3 [47,48]. The mean
diameters of the TiO (a) and TiO (r) crystals in TiO (P25), derived
PC2nH
ꢁ PH2O
1
4nꢂ2
O
g1
¼
ꢁ K
tivation rates in reactions of C
3
2
2
2
P
aldol
C
n
H
2n
O
1
where Kaldol is the equilibrium constant. The g value was deter-
2
2
mined to be 0.3 for propanal (at 20% conversion, 523 K; Table 1)
using available thermodynamic data (Section S2, SI), indicating that
condensation rates are influenced significantly by thermodynamic
2
4
2
2
constraints. Forward condensation rates (r
rates (r ) by
f
) are related to measured
2
2
2
from transmission electron micrographs, are 25 and 85 nm, respec-
n
tively [48], indicating that anatase surfaces are ꢀ10-fold larger
r
n
¼ r ð1 ꢂ g1
f
Þ
ð6Þ
2
than for rutile crystals in TiO (P25). Our thermal treatments in
air before catalysis (673 K, 3 h) retained these properties, consis-
tent with the higher temperatures required for anatase–rutile tran-
sition (ꢀ800 K) [49] and with the infrared evidence for the
Fig. 1 shows that the areal forward rates for propanal reactions on
TiO (P25) decreased with time in the semi-logarithmic manner pre-
2
scribed by first-order deactivation processes with a deactivation
ꢂ1
prevalence of undissociated acetic acid, typical of TiO
2
(a), instead
(P25) surfaces
29]. As shown in Section 3.3, condensation turnovers are much
constant (k
d
) of 0.078 ks . An inverse relation between acetalde-
of acetate species prevalent on TiO
2
(r), on TiO
2
2
hyde aldol condensation rates on TiO (a) and the concentration of
[
irreversibly-adsorbed residues [22] suggests that site blockage by

3
06
S. Wang et al. / Journal of Catalysis 340 (2016) 302–320
Table 1
Reaction rates of propanal and 1-propanol on TiO
2
and on TiO
2
+ Cu/SiO
2
.a
TiO
Catalyst
2
TiO
2
+ Cu/SiO
2
2 2
TiO + Cu/SiO
Reactant
Propanal
0.18
19.5
0.18
<0.01
Propanal
0.80
3.6
0.17
0.63
1-Propanol
0.80
3.6
0.16
0.64
Inlet pressureb (kPa)
2
ꢂ1
Residence time (mTiO2 s (mol C-atom)
)
c
Propanal pressure (kPa)
1
-Propanol pressurec (kPa)
d
ꢂ3
ꢂ2 ꢂ1
Net formation rate (10
lmol C-atom m
s
)
e
C
C
6
oligomer
66
17
<0.1
<0.1
<0.1
91
15
8
1.2
0.1
90
13
9
1.5
0.2
oligomere
9
Propyl propionate
Propene
Propane
f
ꢂ3
ꢂ3
ꢂ2 ꢂ1
r
r
n
(10
(10
l
mol C-atom m
s
)
83
118
106
106
103
103
g
ꢂ2 ꢂ1
f
lmol C-atom m
s
)
g
g
g
g
1
2
2
2
(propanal)h
(1-propanol)
(2-methyl-pentanol)
(2-methyl-pentan-3-ol)
0.3
<0.01
1.0
0.8
<0.01
1.0
0.9
i
<0.01
<0.01
<0.01
i
i
0.9
0.9
a
b
c
d
e
f
TiO
2
(P25) or TiO
2
(P25) + 20 wt.% Cu/SiO
2
(1:1 mass), 523 K, 40 kPa H
2
, ꢀ20% conversion.
Inlet reactant pressure.
Average pressure in the reactor.
Extrapolated exponentially to zero time.
Derived from propanal condensation (Scheme 2).
Net propanal condensation rates.
Forward propanal condensation rates (Eq. (6)).
Approach to equilibrium factor for aldol condensation (Eq. (5)).
Approach to equilibrium factor for hydrogenation to alkanol (Eq. (8)).
g
h
i
Scheme 2. Reaction network of aldol condensation and esterification for propanal and 1-propanol reactants on TiO
2
2
and Cu/SiO catalysts. The concentrations of the
molecules shown in brackets were below the experimental detection limit (0.001 kPa).
sequential condensation products accounts for the observed deacti-
vation. The ubiquitous thermodynamic constraints and rapid deac-
1-propanol–propanal mixtures, as shown by the approach to equi-
librium factor (g2; Table 1):
tivation, taken together, render monofunctional TiO
2
surfaces
K
hydro
unsuitable as practical condensation catalysts and also preclude
any rigorous kinetic or mechanistic analysis of the elementary
states and active sites that mediate such reactions on oxide
surfaces.
C
n
H
2nO þ H
2
¢ C
n
H
2nþ1OH
ð7Þ
ð8Þ
P
Cn
H
2nþ1
OH
1
g
¼ P
ꢁ K
2
H2 ꢁ PCnH2n
O
hydro
Propanal reactions were also examined on TiO
physical mixtures (1:1 mass; 523 K, 40 kPa H ). Propanal hydro-
genation on the Cu function (Scheme 2) led to equilibrated
2 2
(P25) + Cu/SiO
2
where Khydro is the equilibrium constant of the hydrogenation reac-
tion, and obtained from available thermodynamic data (Section S2,

S. Wang et al. / Journal of Catalysis 340 (2016) 302–320
307
2
propanal reactants on TiO (P25) did not depend on the amount of
2 2
Cu present in these physical mixtures (0.5–2.0 (Cu/SiO )/TiO mass
ratio; Fig. 2), consistent with the ability of the Cu function to fully
and quickly equilibrate alkanol–alkanal mixtures at the prevalent
H
2
pressures and with the sole kinetic relevance of TiO
condensation and esterification steps.
and C condensation products also formed from 1-propanol–
propanal reactants on TiO (P25) + Cu/SiO , even at low conversions
12–20%). They consist of hydrogenated forms of the condensation
products observed on TiO (P25). For instance, 2-methyl-pent-2-
2
-catalyzed
C
6
9
2
2
(
2
enal was hydrogenated to 2-methyl-pentanal and 2-methyl-
pentanol, and to their respective isomers (2-methyl-pentan-3-one
and 2-methyl-pentan-3-ol) (Scheme 2) to form equilibrated
2 2 2
mixtures at each H pressure on TiO (P25) + Cu/SiO mixtures with
their corresponding approach to equilibrium constants close to
unity (Table 1). The hydrogenation of 2-methyl-pent-2-enal, the
primary propanal condensation product, circumvents the thermo-
dynamic bottlenecks prevalent on TiO
condensation reactions essentially irreversible (
2
catalysts, thus rendering
< 0.01, Table 1)
: from 0.083 to
g
n
1
and leading to higher areal conversion rates (r
ꢂ2 ꢂ1
0
.106
lmol C-atom m
s ; Table 1). These areal rates, when cor-
Fig. 1. Time-on-stream profiles for forward rates of aldol condensation (▲) on
rected by their approach to equilibrium, become the same on the
TiO
2
(P25) and for forward rates of aldol condensation (d) and esterification (r) on
(P25) + Cu/SiO (1:1 mass) (0.18 kPa average propanal pressure in the reactor,
). Dashed lines represent exponential regressed fits of the data with
first-order deactivation constants shown beside the lines.
monofunctional TiO (P25) and on the bifunctional mixture. The
2
TiO
2
2
presence of the Cu function led to a marked decrease in first-
5
23 K, 40 kPa H
2
ꢂ1
d
order deactivation constants (k ) from 0.078 to 0.001 ks
(
Fig. 1), seemingly as the result of the scavenging of the unsatu-
rated products typically implicated in the formation of unreactive
residues that block active sites on TiO surfaces.
Propyl propionate was not detected during propanal reactions
on TiO (P25), but formed from the equilibrated 1-propanol–
propanal mixtures prevalent on TiO (P25) + Cu/SiO . Esterifica-
tion/condensation rate ratios increased from 0.16 to 0.36 with
increasing the (Cu/SiO )/TiO ratio (0.5–1.0) and then reached a
2
2
2
2
2
2
constant value for ratios larger than unity (Fig. 2). Pool conversion
rates (the combined esterification and condensation rates), how-
ever, were unaffected by the amount of Cu/SiO in the physical
2
mixture (Fig. 2). These data suggest that the Cu function is required
also to complete esterification turnovers and that the Cu surface
area required to remove thermodynamic or kinetic bottlenecks
exceeds those required for condensation routes. The constant con-
version rates, however, indicate that condensation and esterifica-
tion turnovers share a common intermediate, formed in the
kinetically-relevant step, which is subsequently involved in con-
densation and esterification events that determine selectivity, but
not alkanol–alkanal pool conversion turnover rates. We surmise
here, and confirm later by kinetic, isotopic, and theoretical meth-
ods, that unstable hemiacetals form via reactions of alkanols with
alkanal-derived enolate species [51], with the latter formed in the
kinetically-relevant step; such reactions are disfavored by thermo-
dynamics, but dehydrogenation to more stable esters on Cu scav-
enges the products to form more stable esters, thus rendering
such reactions first reversible, leading to higher esterification to
condensation ratios, and ultimately irreversible and no longer
consequential for selectivity at higher Cu contents (Fig. 2). The
Fig. 2. Effects of (Cu/SiO
-propanol–propanal reactant pool (j) and on esterification/condensation rate ratios
▲) (TiO (P25) + 20 wt.% Cu/SiO , 523 K, 0.8 kPa 1-propanol, 40 kPa H ). Dashed
curves indicate trends. Areal rates are given on the basis of TiO surface areas.
2 2
)/TiO mass ratios on areal conversion rates of the
1
(
2
2
2
2
SI). These equilibrated mixtures can be rigorously treated as a reac-
tant chemical lump in all kinetic analyses. Their kinetic equivalence
is evident from the identical aldol condensation and esterification
rates measured at a given propanal pressure when using either
constant ratios at high Cu contents ((Cu/SiO
2 2
)/TiO P 1.0, Fig. 2)
rigorously reflect the respective kinetic rates of alkanol and alkanal
1
4
-propanol or propanal as the inlet reactant (0.8 kPa reactant;
0 kPa H ; Table 1). Henceforth, conversions and selectivities are
reactions with enolates and thus the intrinsic reactivity of TiO
surfaces in such reactions, as discussed in detail in Section 3.4.
2
2
based on the formation of products from this equilibrated reactant
lump. Cu surfaces can catalyze aldol condensation and esterification
of 1-propanol–propanal reactants [50], which form products identi-
Acetone condensation reactions were also examined on
TiO (P25) and TiO (P25) + Cu/SiO mixtures (523 K; 40 kPa H ).
Primary condensation events led to equilibrated mixtures of
4-methyl-pent-3-en-2-one (mesityl oxide) and 4-methyl-pent-4-
2
2
2
2
cal to those observed on TiO
These rates on Cu surfaces, however, are >10-fold lower than those
on TiO (P25) catalysts at all conditions used here; such Cu contribu-
tions were subtracted from reported rates on TiO (P25) + Cu/SiO
mixtures when so required. Areal conversion rates of 1-propanol–
2 2
(P25) + Cu/SiO mixtures (Table 1).
en-2-one (isomesityl oxide) and to secondary C
(e.g., 4,6-dimethyl-hepta-3,5-dien-2-one and its skeletal isomers;
Scheme 3); the presence of Cu/SiO and H rendered acetone reac-
tants and the unsaturated condensation products present in
9
alkanone products
2
2
2
2
2

3
08
S. Wang et al. / Journal of Catalysis 340 (2016) 302–320
2 2
Scheme 3. Reaction network of aldol condensation for acetone and 2-propanol reactants on TiO and Cu/SiO catalysts. The concentrations of the molecules shown in
brackets were below the experimental detection limit (0.001 kPa).
Table 2
Reaction rates of acetone and 2-propanol on TiO
2
and on TiO
2
+ Cu/SiO
TiO
2
.a
Catalyst
TiO
2
TiO
2
+ Cu/SiO
2
2
+ Cu/SiO
2
Reactant
Acetone Acetone
2-Propanol
0.80
2.1
Inlet pressureb (kPa)
0.70
1.8
0.80
2.1
Residence time
2
ꢂ1
(
m
TiO2 s (mol C-atom)
)
c
Acetone pressure (kPa)
0.70
<0.01
0.72
0.08
0.71
0.09
2
-Propanol pressurec (kPa)
d
ꢂ3
ꢂ2 ꢂ1
Net formation rate (10
l
mol C-atom m
s )
C
C
6
oligomere
oligomere
261
220
<0.1
<0.1
343
210
80
345
202
93
9
Propene
Propane
4.2
4.8
f
ꢂ3
ꢂ3
ꢂ2 ꢂ1
r
r
n
(10
(10
l
mol C-atom m
s
)
481
534
553
553
547
547
g
ꢂ2 ꢂ1
f
lmol C-atom m
s )
g
g
g
1
2
2
(acetone)h
(2-propanol)
(4-methyl-pentan-2-ol)
0.1
<0.01
<0.01
<0.01
1.0
1.0
<0.01
1.0
0.9
i
i
a
TiO
2
(P25) or TiO
2
(P25) + 20 wt.% Cu/SiO
2
(1:1 mass), 523 K, 40 kPa H
2
, ꢀ20%
conversion.
b
Inlet reactant pressure.
Mean pressure in the reactor.
Extrapolated exponentially to zero time.
Derived from acetone condensation (Scheme 3).
Net acetone condensation rates.
Forward acetone condensation rates (Eq. (6)).
Fig. 3. Forward rates of aldol condensation as a function of time-on-stream on
TiO (P25) (▲) and on TiO (P25) + Cu/SiO (d) (0.70 kPa average acetone pressure in
the reactor, 523 K, 40 kPa H ). Dashed lines represent exponential regressed fits of
c
d
e
f
2
2
2
2
the data with first-order deactivation constants shown next to each data set.
g
h
i
resides at a tertiary C-atom (Scheme 3), thus precluding dehydro-
genation to form an ester.
Approach to equilibrium factor for aldol condensation (Eq. (5)).
Approach to equilibrium factor for hydrogenation to alkanol (Eq. (8)).
In summary, the presence of Cu as a hydrogenation–dehydro
genation function and of H
rates of propanal or acetone on TiO
measured rates by scavenging primary enal or enone products to
more saturated ones (e.g. C alkanals, alkanones, and alkanols);
2
did not influence forward conversion
equilibrium with their saturated analogs, as described below. For
instance, 4-methyl-pentan-2-one and 4-methyl-pentan-2-ol
formed are in equilibrium with each other and with mesityl oxide
2
surfaces, but led to higher
(Scheme 3).
6
these reactions also suppressed the formation of larger unsatu-
rated molecules that form unreactive residues, thus inhibiting
deactivation to nearly undetectable levels. The rapid equilibration
of alkanol–alkanal (alkanone) mixtures renders the two molecules
indistinguishable in condensation reactions.
2
Acetone condensation rates on TiO (P25) were affected by equi-
librium constraints (
g
1
= 0.1, 523 K, 0.8 kPa acetone; Table 2) and
ꢂ1
decreased rapidly with time on stream (k
Equilibrium constraints were negligible on TiO
< 0.01; Table 2), however, and deactivation was essentially sup-
d
= 0.14 ks ; Fig. 3).
(P25) + Cu/SiO
2 2
(g
1
ꢂ1
pressed (k
d
= 0.002 ks ; Fig. 3). The Cu function also enabled the
= 1;
formation of equilibrated 2-propanol–acetone reactants (
g
2
3.2. Equivalence of alkanols and alkanones/alkanals and oxygen
removal and chain growth at practical conversions
Table 2), resulting in identical condensation rates for pure acetone
and 2-propanol reactants at each given acetone pressure, set by
2
H
-propanol–acetone-H
; Table 2). Esters were not detected among the products of 2-
propanol–acetone reactants, because the OH group in hemiacetals
2
equilibration (0.8 kPa reactant; 40 kPa
Here, we show that alkanol streams can be converted to prod-
ucts with longer C-chains and lower oxygen content via condensation
and esterification reactions on bifunctional TiO (P25) + Cu/SiO
2 2
2

S. Wang et al. / Journal of Catalysis 340 (2016) 302–320
309
Fig. 4. Average number of C-atoms (a) and O/C ratio (b) of liquid effluents as a function of alkanol–alkanal (alkanone) pool conversions for pure 1-propanol (s) and 2-
propanol ( ) reactant feeds (TiO (P25) + 20 wt.% Cu/SiO (1:1 mass), 523 K, 0.8 kPa 1-propanol or 2-propanol, 40 kPa H ). Dashed curves indicate trends.
D
2
2
2
catalysts at practical conversions. The average number of
C-atoms of the effluent stream (including unreacted lumped
reactants) increased from 3.0 (1-propanol) to 7.4 as the conversion
of the lumped (equilibrated) 1-propanol–propanal reactants
increased to 84% (523 K; 40 kPa H
molecules present in the effluent stream decreased from 0.33 (in
-propanol) to 0.15 (Fig. 4b). The products present at all conver-
2
; Fig. 4a); the O/C ratio in the
1
sions and the trends in Fig. 4 are consistent with the CAC and
CAO bond formation and O-removal steps characteristic of con-
densation and esterification reactions (Scheme 2). Similar results
for 2-propanol reactants are shown in Fig. 4 and elsewhere for
2 5
C –C alkanols [29]; the conclusions, practical consequences, and
catalyst requirements apply broadly to conversion strategies for
alkanols and carbonyl compounds and their mixtures, which are
ubiquitous in biomass-derived streams.
These bifunctional strategies, however, are much less effective
when noble metals (e.g. Pt and Pd) are used instead of Cu as the
metal function, because their higher decarbonylation activity leads
to the concurrent formation of CO, which adsorbs strongly and
inhibits the metal function [52]. Cu surfaces, in contrast, do not
catalyze decarbonylation to detectable extents and would, in any
case, adsorb CO much more weakly, making it the most effective
and least costly choice as the metal function in these cascade reac-
tion strategies.
Fig. 5. Areal forward acetone condensation rate as a function of time-on-stream on
TiO (P25), TiO (a) and TiO (r) at 503 K (TiO + 20 wt.% Cu/SiO (1:1 mass), 0.6 kPa
acetone, 40 kPa H ). Dashed lines represent exponential regressed fits of the data
with first-order deactivation constants shown next to each data set.
2
2
2
2
2
2
ꢂ2
ꢂ1
3.3. Site requirements of aldol condensation on TiO
2
surfaces
tively) and areal rates (0.146 and 0.161 lmol C-atom m ks ,
respectively); TiO
2
(r), in contrast, exhibited much faster deactiva-
ꢂ1
Rigorous mechanistic assessments of elementary steps and site
requirements require a reliable measure of the number and the
chemical properties of plausible active sites on TiO surfaces. These
tion (k
d
0.059 ks
)
and lower initial rates (0.025
(P25) and TiO
(r) show that anatase sur-
faces predominantly account for condensation turnovers. It seems
plausible that measured rates on TiO (r) may reflect, in fact,
anatase-like surfaces on some rutile crystallites or the presence
of small crystals of TiO (a) [29,49], undetectable by X-ray
diffraction, but contributing disproportionately to surface areas
in TiO (P25). DFT treatments confirmed that activation barriers
lmol C-
ꢂ2
ꢂ1
atom m ks ). The similar behavior of TiO
and the much lower reactivity of TiO
2
2
(a)
2
2
site counts can then be used to express reactivity in terms of turn-
over rates, thus allowing comparisons among surfaces differing in
chemical composition or crystallographic exposure; they also
make it possible to compare free energies derived from theory
and experiment in attempts to choose among alternate mechanis-
tic hypotheses. Such site counts require, in turn, that we titrate
surface sites with molecules showing different affinities for the
various types of sites prevalent on oxides (Lewis and Brønsted
acids, bases) in order to assess their abundance and their involve-
ment in catalytic turnovers.
2
2
2
for aldol condensation are much larger on TiO
surfaces (Section S8, SI).
Site titrations using probe molecules that bind selectively to
acid or basic surface sites were carried out to determine their
respective surface densities and the extent to rates which are influ-
enced by their binding on TiO (P25). CO [13], 2,6-di-tert-
2 2
butylpyridine (DTBP) [53], pyridine [12], and propanoic acid [12]
were chosen as titrants for basic sites, Brønsted acid sites, Lewis
2 2
(r) than on TiO (a)
Fig. 5 compares areal acetone condensation rates on TiO
pure TiO (a) and pure TiO (r) (503 K; 0.6 kPa acetone; 40 kPa H
all present as physical mixtures with Cu/SiO . TiO (P25) and
TiO (a) showed similar stability (k 0.004 and 0.006 ks , respec-
2
(P25),
2
2
2
),
2
2
ꢂ1
2
d

3
10
S. Wang et al. / Journal of Catalysis 340 (2016) 302–320
acid sites, and acid–base site pairs, respectively. Each titrant was
introduced together with acetone (4 kPa) on TiO (P25) + Cu/SiO
1:1 mass) at 523 K and 40 kPa H ; Cu did not influence titrant
carbonyl O-atom and a Ti center and between their acidic H-atom
and a vicinal O-atom on TiO surfaces [29]. These propanoic acid
inhibition effects are much stronger (and much more slowly
2
2
2
(
2
uptakes or cause significant chemical conversions of these titrants
to other species.
CO , even at high concentrations (4 kPa; CO /acetone = 1.0), did
2 2
not detectably influence reaction rates (Fig. 6), indicating that
strong basic sites are not involved in condensation turnovers. DTBP
reversed) than for pyridine titrants, indicating that acid–base TiAO
site pairs are likely to account for the reactivity of TiO
In this context, basic O-atoms act as H-abstractors, required to acti-
vate -CAH bonds in the alkanal (alkanone) reactants [25], with Ti-
2
(a) surfaces.
a
atoms acting to stabilize the transition state that mediate enolate
formation in the kinetically-relevant elementary step via interac-
tions with the carbonyl group in the alkanal (alkanone) reactants.
The strong interactions of propanoic acid with acid–base site
(
DTBP/acetone = 0.005) also did not affect rates (Fig. 6), showing
that Brønsted acid sites are not involved. In contrast, pyridine
(
1
pyridine/acetone = 0.005) decreased rates slightly from 2.2 to
ꢂ2 ꢂ1
.8
lmol C-atom m
s
(Fig. 6); rates gradually recovered, how-
2
pairs on TiO allow measurements of the number of such sites at
ever, upon pyridine removal (over ꢀ3 ks; Section S3, SI). Thus,
weak Lewis acid sites that reversibly bind pyridine appear to be
required for condensation turnovers.
TiO surfaces during acetone condensation catalysis. In doing so,
2
we use lower reaction temperatures (453 K) to ensure irreversible
binding and to suppress any hydrogenation of propionic acid to
propanal or 1-propanol and subsequent condensation and esterifi-
cation of these formed products. The number of bound acid mole-
cules was determined by accounting for trace amounts of such
products formed from propionic acid that are chemically distinct
from those formed from acetone reactants, and of unreacted acid
in the effluent (Section 2.3).
Propanoic acid strongly decreased condensation rates (2.2–
ꢂ2 ꢂ1
0
.6
lmol C-atom m
s
(Fig. 6), even at concentrations much
lower than for acetone reactants (propanoic acid/acetone = 0.005);
such inhibition was gradually and fully reversed, but over periods
(
ꢀ8 ks; Section S3, SI) much longer than typical turnover times.
Infrared spectra show that carboxylic acids bind on TiO (a) sur-
faces without dissociation via concerted interactions between their
2
Areal acetone condensation rates on TiO
almost linearly with the amount of adsorbed propanoic acid
Fig. 7a). Slight deviations from these linear trends reflect the
2
(P25) decreased
(
growing scarcity of vicinal Ti centers, required for CAC coupling
steps (Section 3.4), as TiAO site pairs are titrated by propanoic acid.
These data show that TiAO site pairs are titrated stoichiometrically
and irreversibly by propanoic acid and that such sites are involved
in the kinetically-relevant elementary steps that mediate aldol
condensation reactions.
The number of titrants required to fully suppress condensation
rates (by linear extrapolation of the data in Fig. 7) gives the number
of TiAO pairs on each surface. Their number on TiO
2
(P25)
ꢂ2
(
3.7 titrants nm , Fig. 7a) is slightly smaller than the number of
TiAO site pairs determined from the anatase crystal structure for
ꢂ2
ꢂ2
its low-index planes (5.2 nm
7
,
(101); 5.6 nm
,
(100);
ꢂ2
.0 nm , (001)), plausibly because of the presence of some per-
manent OH groups, required to preserve favorable O and Ti coordi-
nation on surfaces of small crystallites [29].
Fig. 6. Effects of CO
introductions on areal forward acetone condensation rates on TiO
Cu/SiO (1:1 mass) at 523 K (4 kPa acetone, 4 kPa CO 20 Pa 2,6-di-tert-
butylpyridine, 20 Pa pyridine, 20 Pa propanoic acid, 40 kPa H
2
,
2,6-di-tert-butylpyridine, pyridine, and propanoic acid
Propanoic acid titration data were also measured on TiO
TiO (r). The number of site pairs (per surface area) was nearly
2
identical in TiO (a) (4.0 titrants nm , Fig. 7b) and TiO (P25)
2
(a) and
2
(P25) + 20 wt.%
2
2
2
,
ꢂ2
2
).
2
Fig. 7. Forward acetone condensation rates on (a) TiO
uptake) (TiO + 20 wt.% Cu/SiO (1:1 mass), 0.8 kPa acetone, 20 Pa propanoic acid, 20 kPa H
with propanoic acid. Dashed lines are linear regression fits.
2
(P25), (b) TiO
2
(a), and (c) TiO
2
(r) at 453 K before propanoic acid introduction (vs. time) and after (vs. cumulative titrant
2
2
2
). The rates are normalized by surface TiAO site pairs measured from titration

S. Wang et al. / Journal of Catalysis 340 (2016) 302–320
311
ꢂ2
(
3.7 titrants nm , Fig. 7a), consistent with the prevalence of ana-
tase surfaces in TiO (P25) [47,48]. Site densities on TiO (r)
0.7 titrants nm , Fig. 7c), however, were much smaller than on
anatase and also than those expected from the rutile crystal struc-
attack on another propanal, a characteristic of aldol condensations
catalyzed by strongly basic oxides or liquid bases [1,12].
2
2
ꢂ2
(
Fig. 8a shows that combined turnover rates (per TiAO pair) for
the conversion of 1-propanol–propanal reactants to condensation
and esterification products are proportional to propanal pressure,
but insensitive to 1-propanol pressure. The ratio of esterification
to condensation rates, however, depends linearly on the 1-
propanol/propanal ratio present in the equilibrated reactants
(Fig. 8b). These trends indicate that the formation of common
enolate-like species from propanal limits the combined formation
rates of condensation and esterification products as described in
Section 3.1. Moreover, the insensitivity of conversion rates to 1-
propanol concentrations indicates that reactions of enolates to
form new CAC or CAO bonds cannot be kinetically-relevant,
because their rates would differ for the reactions of these enolates
with alkanols (to form hemiacetals/esters) and with alkanals (to
form aldols), thus rendering conversion rates dependent on both
alkanol and alkanal pressures. We conclude, therefore, that enolate
formation must be the sole kinetically-relevant step, with product
selectivities determined by the relative reactivity of enolates with
alkanols and alkanals and by their respective pressures, consistent
with DFT-derived barriers reported below (Section 3.5).
ꢂ2
ꢂ2
ꢂ2
ture (5.2 nm , (110); 7.4 nm , (100); 8.0 nm , (011)). We sur-
mise that such low densities of accessible TiAO site pairs reflect
the placement of unreactive acetone-derived species during ace-
tone reactions before titrant introduction; such species prevent
access to sites for dissociating OH bonds in propanoic acid or
CAH bonds in acetone, a conclusion consistent with the faster
deactivation of TiO (r) surfaces and with DFT treatments that
implicate strong acid–base site pairs on TiO (r) in the difficult
a-
2
2
reprotonation and desorption of the CAC coupling products (Sec-
tion S8, SI). Initial acetone turnover rates (normalized by the num-
ber of catalytically-relevant acid–base site pairs) are similar on
ꢂ1
ꢂ1
TiO
2
(r) (1.9 (TiAO ks)
,
Fig. 7c), TiO
2
(P25) (2.2 (TiAO ks)
,
ꢂ1
Fig. 7a) and TiO
2
(a) (2.3 (TiAO ks) , Fig. 7b). These data are consis-
tent with condensation turnovers that occur exclusively on anatase
surfaces, making turnover rates and site densities on TiO (r)
merely a consequence of anatase-like surfaces present as minority
species.
Taken together, these assessments of the number and type of
active sites during catalysis indicate that TiAO acid–base site pairs
of intermediate acid and basic strengths on anatase surfaces are
required for efficient condensation turnovers. This evidence and
the ability to determine intrinsic site reactivities allow theoretical
treatments based on the posited site structures, as well as the
accurate benchmarking of DFT-derived activation free energies
against measured values, while also providing confirmatory evi-
dence for the kinetic relevance of the proposed elementary steps
2
Scheme 4 depicts condensation and esterification elementary
steps consistent with these kinetic effects on rates and selectivities.
The kinetically-relevant formation of an enolate from propanal via
a-CAH bond cleavage occurs irreversibly on TiAO site pairs that
remain sparsely covered during catalysis; at such sites, the lattice
O-atoms act as the base and the Ti centers act as the Lewis acid
centers that bind the O-atom in the enolate formed (Step 2;
Scheme 4). These enolates then couple with a propanal bound at
a vicinal Ti center to form a CAC bond (Step 4; Scheme 4) and
(
Section 3.5).
the aldol species dehydrate to form
a,b-unsaturated carbonyl
3
.4. Kinetic dependence of condensation and esterification reactions
products (Steps 6 and 7; Scheme 4). These enolates can also react
with a 1-propanol bound at the vicinal Ti center to form a CAO
bond via an equilibrated H-shift between the enolate and 1-
propanol (Step 9; Scheme 4) and a subsequent CAO coupling reac-
tion between the resulting propanal and alkoxide species to form a
hemiacetal (Step 10; Scheme 4), as suggested by DFT treatments
(Section 3.5). Propyl propionate is finally formed via dehydrogena-
tion of the OH group in the hemiacetals on the Cu function (Step
13; Scheme 4).
and consequences for the kinetic relevance of elementary steps
The effects of propanal and 1-propanol pressures on aldol con-
densation and esterification rates were examined on TiO
Cu/SiO physical mixtures (1:1 mass; 523 K). 1-Propanol–
propanal-H equilibration was achieved at all reaction conditions
Fig. 8a). The condensation products (Scheme 2) contain new
CAC bonds formed exclusively at the -C position in one of the
propanal reactants, as expected from enolate-like intermediates
formed via -H abstraction and their subsequent nucleophilic
2
(P25)
+
2
2
(
a
The elementary steps in Scheme 4 lead to an equation for the
enolate formation turnover rate:
a
Fig. 8. Effects of propanal pressure on turnover rates for the conversion of 1-propanol–propanal equilibrated reactants (a) and of 1-propanol/propanal ratio on esterification
to condensation rate ratios (b) (TiO (P25) + 20 wt.% Cu/SiO (1:1 mass), 523 K, 0.8 kPa 1-propanol, 10–80 kPa H ). Dashed lines in (a) and (b) represent the regressed fits to the
functional forms of Eqs. (10) and (16), respectively. The solid line in (a) represents a linear regression fit.
2
2
2

3
12
S. Wang et al. / Journal of Catalysis 340 (2016) 302–320
A. Enolate formation from propanal
B. Aldol condensation reaction between enolate and propanal
C. Esterification reaction between enolate and 1-propanol
Scheme 4. Aldol condensation and esterification reaction pathways, illustrated here for 1-propanol–propanal reactants, on bifunctional TiO
2 2
+ Cu/SiO catalysts. Steps
occurring on TiAO site pairs and on Cu surfaces are represented by solid and dashed arrows, respectively. Reaction numbers used throughout the text are in parentheses (e.g.,
(
1)) below each reaction arrow; k
j
and K
j
are kinetic constant of the forward step and equilibrium constant for Step j, respectively.
0
r
enolate
r
aldol þ rester
k
enolate
K
1 þ
al
P
K
propanal
which is related, in turn, to k
through
¼
¼
P
ð9Þ
enolate
½
LꢄTiO2
½Lꢄ
i
P
i
ꢀ
ꢁ
TiO2
h
k T
B
0
0
z
D
G
¼ ꢂRT ln
k
ð12Þ
enolate
enolate
where [L]TiO2 is the number of active TiAO site pairs, renolate, raldol
and rester are the respective enolate formation, aldol condensation,
and esterification rates, kenolate is the rate constant for enolate for-
,
0z
D
G
represents the difference between the free energy of the
enolate
z
mation from propanal, and K
i
is adsorption constant for species i
transition state (TS) for enolate formation (G
) and of a gaseous
enolate
on TiAO site pairs (e.g., propanal (Kal), 1-propanol (Kol) and prod-
carbonyl reactant (Gcarbonyl; propanal in this case) and a bare TiAO
site (GTiAO) as shown in Scheme 5:
ucts). The first-order dependence of renolate on propanal pressure
P
indicates that (K
i
P
i
) is much smaller than unity and that TiO
2
sur-
0
z
¼ Gz
D
G
ꢂ Gcarbonyl ꢂ GTiAO
ð13Þ
enolate
enolate
faces, lacking strong basic or acid sites, are essentially bare during
reactions of equilibrated alkanal–alkanol mixtures. As a result, mea-
sured rates are accurately described by
0
0z
The measured values of k
and
D
G
for propanal reactants
enolate
enolate
(
from the data in Fig. 8a) are shown in Table 3.
r
enolate
propanal ¼ k0enolatePpropanal
The formation of condensation and esterification products from
a common enolate species (Scheme 4) leads to the rate equations:
¼
k
enolate
K
al
P
ð10Þ
½
LꢄTiO2
0
r
aldol
LꢄTiO2
where k
enolate formation.
represents the apparent first-order rate constant for
¼ kCCK Ppropanalhenolate
ð14Þ
enolate
al
½
The formalism of transition state theory allows these rates to be
r
ester
0z
¼
ol propanol enolate
CO H
k K K P h
ð15Þ
related to an activation free energy (
Section S4, SI):
DG
; derivation details in
enolate
½
LꢄTiO2
Here, kCC and kCO are kinetic constants for the reaction of enolates
present at a fractional coverage henolate on the active TiAO site pairs
with propanal or 1-propanol to form CAC or CAO bonds, respec-
D
G0
z
r
enolate
k
B
T
enolate
ꢂ
¼
h
e
RT
P
propanal
ð11Þ
½
LꢄTiO2

S. Wang et al. / Journal of Catalysis 340 (2016) 302–320
313
0
z
Scheme 5. Schematic reaction coordinate diagram for enolate formation from gaseous propanal on a TiAO site pair of TiO
2
.
D
G
represents the experimentally accessible
enolate
apparent free energy barrier of the enolate formation (Eq. (12)).
Table 3
Values of k
0
ol H
and (kCOK K )/(kCCKal) from rate data regressed to the respective functional forms of Eqs. (10) and (16) and the corresponding DG values for propanal and acetone
enolate
reactants.
0
ꢂ1
0z
c
ꢂ1
z
d
Reactant
k
(TiAO ks)
D
G
(kJ mol
)
(kCOK
ol
K
H
)/(kCC
K
al
)
DD
G
enolate
enolate
ester—aldol
ꢂ1
(
kJ mol
)
Propanala
Acetoneb
Acetone-d
188 ± 5
86.4 ± 0.8
35.3 ± 0.2
124 ± 1
134 ± 1
138 ± 1
0.101 ± 0.03
10 ± 1
b
6
a
b
c
From data in Fig. 8.
From data in Fig. 9.
From Eq. (12).
d
From Eq. (21).
tively; K
H
is equilibrium constant for the H-shift from 1-propanol to
Eqs. (17) and (18), taken together, show that the (kCO
K
ol
K
H
)/(kCC
K
al
)
)
z
enolates on TiAO site pairs (Step 9, Scheme 4). The esterification/
condensation ratios derived from Eqs. (14) and (15) agree with their
measured dependence on 1-propanol/propanal ratios (Fig. 8b) and
are given by
coefficient reflects the free energy barrier difference (DD
G
ester—aldol
z
z
between
D
G
and
D
G
as given by
ester
aldol
!
z
kCOK KH
ol
DD
G
ester—aldol
¼
exp
ꢂ
ð21Þ
k
CC
K
al
RT
r
r
ester
aldol
k
CO
K
ol
K
H
P
propanol
¼
ð16Þ
k
CC
K
P
al propanal
in which
The measured slope in Fig. 8b thus reflects the value of the
DD
G
zester—aldol ¼ Gzester ꢂ Gazldol þ Gal ꢂ Gol
ð22Þ
0
(k
CO
ol H
K K )/(kCCKal) coefficient in Eq. (16). As in the case of kenolate,
k
CO
K
ol
K
H
and kCC
K
al in Eqs. (14) and (15) depend on free energy
z
The data in Fig. 8b, used in Eq. (21), give a DD
1
G
value of
ester—aldol
barriers:
ꢂ1
0 ± 1 kJ mol (at 523 K; Table 3) for 1-propanol–propanal reac-
!
z
tants, reflecting a modest preference for enolate reactions with pro-
panal over those with 1-propanol on TiAO site pairs at anatase
surfaces.
The mechanistic interpretation of acetone condensation rates is
similar to that shown above for propanal reactions, albeit without
parallel esterification reaction routes from the enolates formed.
The approach to equilibrium between 2-propanol and acetone
k
h
B
T
D
G
ester
k
k
CO
K
ol
al
K
H
¼
exp
ꢂ
ð17Þ
ð18Þ
RT
!
z
k
h
B
T
D
G
aldol
CC
K
¼
exp
ꢂ
RT
z
where
D
G
represents the free energy of the CAO coupling tran-
was also near unity at all conditions (TiO (P25) + Cu/SiO2 (1:1
mass); Fig. 9). Acetone condensation rates were proportional to
ester
2
z
sition state (G
) relative to a gaseous 1-propanol (Gol) and an
ester
adsorbed enolate on a TAO site pair (Genolate) (Scheme 6):
acetone pressures, insensitive to 2-propanol pressures, and inde-
pendent of whether acetone pressures were achieved from acetone
or 2-propanal as reactants, because of their fast interconversions
on the Cu function (Fig. 9).
z
¼ Gz ꢂ Gol ꢂ Genolate
D
G
ð19Þ
ester
ester
z
and
D
G
represents the free energy of the CAC coupling transition
) relative to a gaseous propanal (Gal) and the bound eno-
aldol
z
Scheme 7 depicts acetone condensation pathways similar to
state (G
aldol
those shown above for propanal reactions on TiO
2
surfaces
late (Genolate) (Scheme 6):
(
Scheme 4); the elementary steps include -H abstraction from
a
z
¼ Gz
acetone to form enolates (Steps 1–2), CAC coupling between eno-
late and a vicinal bound acetone (Steps 3–4), and dehydration to
D
G
ꢂ Gal ꢂ Genolate
ð20Þ
aldol
aldol

3
14
S. Wang et al. / Journal of Catalysis 340 (2016) 302–320
Scheme 6. Schematic reaction coordinate diagram for the formation of CAC and CAO bonds via reactions of propanal-derived enolates with propanal and 1-propanol at TiAO
à
à
site pairs on TiO
Scheme 4), respectively; DD
2
surfaces. [CAO coupling] and [CAC coupling] represent respective transition states for the CAO coupling (Step 10; Scheme 4) and CAC coupling (Step 4;
0
z
G
values are measurable activation free energy barrier differences between esterification and condensation reactions (Eq. (21)).
ester—aldol
0
z
molecule and an unoccupied TiAO site pair (
D
G
, defined by
DG value of
enolate
enolate
0z
Eq. (13)). Measured rate constants (Fig. 9) give a
ꢂ1
1
34 ± 1 kJ mol (523 K, Table 3), which is 10 ± 2 kJ/mol higher than
ꢂ1
for enolate formation from propanal (124 ± 1 kJ mol ; 523 K,
Table 3). The relative reactivities implied by these different activa-
tion free energies are consistent with the higher DFT-derived depro-
tonation energy (DPE) at the
a
-C-atom in gaseous acetone
ꢂ1
ꢂ1
(
1648 kJ mol
)
than in gaseous propanal (1622 kJ mol ), as
described in Section 3.5.
Condensation rates were also proportional to the alkanone
6 2
pressure for perdeuterated acetone (acetone-d ; D used instead
of H to avoid isotopic dilution of reactants) (Fig. 9). Acetone-d
2
6
condensation rates were lower than for undeuterated acetone,
ꢂꢂ
ꢃ ꢄꢂ
ꢃ ꢃ
0
0
with a rate constant ratio
k
k
of 2.4 (523 K,
enolate
H
enolate
D
Table 3); such primary kinetic isotope effects are consistent with
the kinetic relevance of
a
-CAH bond cleavage steps required for
ꢂꢂ
ꢃ ꢄꢂ
ꢃ ꢃ
0
0
enolate formation. This
k
k
value is nearly iden-
enolate
H
enolate
D
tical to that derived from DFT methods (2.5; Section 3.5).
These kinetic effects of alkanal and alkanone pressures show
that condensation and esterification rates are limited by the forma-
tion of enolates from carbonyl reactants. The subsequent enolate
reaction steps, to form CAC or CAO bonds in molecules with the
O-contents and backbones expected from nucleophilic attacks,
determine the products formed but not their combined formation
rates. Next, we assess these mechanistic hypotheses using theoret-
ical treatments, while also examining the reactivity descriptors
that determine the effectiveness of TiAO site pairs for enolate for-
mation and for the selectivity to condensation and esterification
products (Section 3.4) and the very different reactivities of anatase
Fig. 9. Acetone pressure effects on aldol condensation turnover rates with 2-
propanol (h), acetone (j) and acetone-d (▲) as reactants (TiO (P25) + 20 wt.% Cu/
SiO (1:1 mass), 523 K, 0.8–2.0 kPa oxygenates, 40 kPa H for 2-propanol and
acetone, 40 kPa D for acetone-d ). Dashed lines represent the regressed fits to the
functional form of Eq. (23).
6
2
2
2
2
6
form mesityl oxide (Steps 5–7). As in the case of propanal reac-
tants, the linear relation between rates and acetone pressures indi-
cates that
a-H abstraction is the sole kinetically-relevant step; it
2
and rutile TiO surfaces (Section 3.3).
occurs on TiAO site pairs essentially uncovered during catalysis,
thus leading to the rate equation:
3
.5. Density functional theory treatments of aldol condensation and
r
enolate
r
aldol
acetone ¼ k0enolatePacetone
esterification on Ti 18 cluster models
5 19
O H
¼
¼ kenolate
K
one
P
ð23Þ
½
LꢄTiO2
½LꢄTiO2
The condensation pathways in Schemes 4 and 7 involve two Ti
Lewis acid centers and one basic O-atom vicinal to these two Ti-
atoms. These TiAOATi structures were examined using clusters
where [L]TiO2 is the number of TiAO pairs, renolate and raldol are the
respective enolate formation and aldol condensation rates, kenolate
is rate constant for enolate formation from acetone, Kone is acetone
adsorption constant on exposed anatase TiAO site pairs, and k
represents the apparent rate constant for condensation (kenolate
2
constructed from the (101) plane of TiO (a), the most stable ana-
0
enolate
tase surface [31]; this surface consists of sawtooth-like features
with two Ti-atoms and two O-atoms that are distinct in coordi-
nation number (CN) as shown in Scheme 1a. One Ti center is
under-coordinated (Ti5c, CN 5) and the other exhibits octahedral
0
Kone). As in Eq. (12) (for propanal reactants), kenolate reflects the free
energy of the enolate formation TS relative to a gaseous acetone

S. Wang et al. / Journal of Catalysis 340 (2016) 302–320
315
(
(
A) Enolate formation form acetone
B) Aldol condensation reaction between enolate and acetone
Scheme 7. Aldol condensation pathway of 2-propanol–acetone reactants on bifunctional TiO
2
+ Cu/SiO
2
catalysts. Reaction numbers used throughout the text are in
parentheses (e.g., (1)) below each reaction arrow; k and K are kinetic constant of the forward step and equilibrium constant for Step j, respectively.
j
j
coordination, making it coordinatively saturated (Ti6c, CN 6).
Bridging (O2c, CN 2) and in-plane (O3c, CN 3) O-anions link these
or alkanol reactant to complete CAC or CAO formation turnovers;
in contrast, each O3c site binds to two acidic Ti5c sites with identical
bond lengths (0.203 nm). The involvement of these Ti5cAO2c/3cATi5c
sites in the elementary steps of Scheme 4 is assessed next using DFT
methods by calculating enthalpies, entropies, and free energies for
these steps.
Ti centers. Here, we use
5 19
a symmetrical Ti O H18 cluster,
extracted from anatase (101) surfaces, to describe these struc-
tures and their binding and catalytic properties (Scheme 1a),
because these properties on oxides, in contrast to metals, pre-
dominantly reflect the local properties near the binding sites
The proposed propanal condensation and esterification elemen-
tary steps (Scheme 4) were examined at Ti5cAO3cATi5c sites on
[
34]; the low surface coverages prevalent at the examined cat-
alytic conditions (Section 3.4) exclude significate lateral interac-
tions among adsorbed species, which would require multiple
site pairs for accurate treatments. These clusters consist of five
Ti-atoms (2 Ti5c, 3 Ti6c) and three O-atoms (1 O3c, 2 O2c) with
H-atoms placed at the O-atoms in the cluster edge, in order to
terminate the cluster while maintaining charge neutrality
anatase Ti
mation transition states (from propanal; Step 2, Scheme 4) involve
-CAH cleavage (Scheme 8) and concerted interactions with the
basic O3c site, which abstracts the -H atom from propanal ( -H-
3c 0.127 nm), and the acidic Ti5c center, which concurrently inter-
acts with the carbonyl O-atom (Ti5cAO 0.199 nm). The -CAH
5 19
O H18 clusters (Scheme 1a). DFT-derived enolate for-
a
a
a
O
a
(
Section 2.4).
bond at the TS is longer than in propanal reactants (0.140 vs.
0.109 nm, Scheme 8), but much shorter than the combined van
der Waals radii of C and H-atoms (0.29 nm), indicating that the
The strength of the Lewis acid centers in these clusters was
probed using the affinity of each distinct Ti center for gaseous
OH anions (EHA, Eq. (2)), while basic strength for O-atoms was
ꢂ
a-CAH bond is not fully cleaved at the enolate formation TS. As
assessed using their affinity for gaseous protons (EPA, Eq. (3))
the -CAH bond cleaves along the reaction coordinate, the C@O
a
ꢂ1
(
Table 4). The hydroxide anion affinity is ꢂ389 kJ mol for Ti5c
bond in propanal weakens and lengthens, as its O-atom interacts
with the Ti5c site, leading to C@O bonds that are longer at the TS
(0.128 nm) than in propanal reactants (0.122 nm); these bonds
ultimately increase to 0.134 nm in the fully-formed enolate pro-
duct state (Scheme 8).
ꢂ1
centers and less negative than ꢂ1 kJ mol for the coordinatively-
saturated Ti6c centers (Table 4), indicating that only Ti5c centers
could stabilize enolate formation transition states. The proton
affinities for
O
2c and
O
3c sites are similar (ꢂ1009 and
ꢂ1
ꢂ1008 kJ mol , respectively; Table 4); therefore, their involve-
ment and reactivity depend predominantly on their distance from
the active Ti5c centers, because of the concerted interactions
required to stabilize transition states along the reaction coordinate
for enolate formation and for subsequent reactions of such enolate
species (Scheme 4). Each O2c site is linked to one Ti5c site (0.183 nm)
and one Ti6c site (0.184 nm), but resides far from a second Ti5c site
The evolution of these structures on Ti5cAO3c site pairs indi-
cates that the TS lies at an intermediate point along the
a-CAH
bond activation path and that it is neither very late nor very early
along the reaction coordinate. In addition, the local atomic charges
of the structures involved, determined from the natural bond orbi-
tal (NBO) theory using bonding orbitals with maximum electron
density to determine the most possible Lewis electronic structure
[45], indicate that the electron transfer between propanal and
(
0.433 nm), which would be required to bind the second carbonyl
5 19
the Ti O H18 cluster is negligible throughout the enolate forma-
tion reaction coordinate (<0.1 e, Table 5); this confirms that
Ti5cAO3c site pairs act as Lewis acid–base site pairs without signif-
icant redox character.
Table 4
Coordination numbers (CN), hydroxide anion affinities (EHA), and proton affinities
(EPA) for exposed atoms on TiO (a) (101) surface (Scheme 1a).
2
The condensation and esterification elementary steps after the
enolate formation (Scheme 4) were also examined at the same
level of theory; DFT-derived structures of the relevant reactants,
transition states, and products are shown in the SI (Section S5).
These theoretical treatments notably showed that CAO coupling
between enolates and alkanols cannot occur in one elementary
step, but requires instead sequential steps involving hydrogen shift
and CAO bond formation (Steps 9 and 10; Scheme 4).
a
ꢂ1
a
ꢂ1
Site
CN
E
HA (kJ mol
)
E
PA (kJ mol
)
Ti5c
Ti6c
5
6
2
3
ꢂ389
<ꢂ1
O
O
2c
ꢂ1008
ꢂ1009
3c
a
Ti
5 19
O H18 cluster model, B3LYP, 6-311+G(d,p) for C, H, and O-atoms, LANL2DZ
for Ti-atoms.

3
16
S. Wang et al. / Journal of Catalysis 340 (2016) 302–320
5 19
Scheme 8. DFT-derived structures involved in enolate formation reactions from propanal on Ti5cAO3cATi5c sites of the anatase Ti O H18 cluster (B3LYP, 6-311G(d,p) for C, H,
and O-atoms, LANL2DZ for Ti-atoms; only Ti5cAO3cATi5c sites and two O-atoms directly bound with the Ti5c sites are shown for clarity).
Table 5
study and even at much higher propanal pressures (e.g. 0.07 at
Local charges of optimized structures involved in enolate formation of propanal on
Ti5cAO3cATi5c sites of the anatase Ti
5
O
19
H18 cluster obtained from nature bond
10 kPa) and with the observed rate equation and its mechanistic
interpretation (Eq. (10) and Scheme 4).
orbital analysis.a
The TS for the enolate–propanal CAC coupling step (Step 4,
Scheme 4) showed the highest free energy along the enolate reac-
ꢂ1
tion branch leading to condensation products (76 kJ mol
,
z
Fig. 10); therefore, this step accounts for the value of
DG
in Eq.
CAC
(
20). The TS for alkoxide–alkanal CAO coupling, which occurs after
the H-transfer from 1-propanol to the enolate (Step 10; Scheme 4),
showed the highest free energy along the enolate reaction branch
ꢂ1
that leads to esterification products (89 kJ mol , Fig. 10), thus
z
accounting for the value of
DG
in Eq. (19). The value of
CAO
z
DDG
(Eq. (22)) calculated from the free energy difference
⁄
à
⁄
ester—aldol
Atom
Propanal
Transition state
Enolate
z
z
ꢂ1
between
DG
and
DG
is 13 kJ mol (Fig. 10), also agreed well
with the experimentally-measured value (10 ± 1 kJ mol , Table 3).
These propanal condensation pathways (Scheme 4) were also
CAC
CAO
C
C
C
1
2
3
ꢂ0.50
0.59
ꢂ0.57
ꢂ0.51
0.26
0.24
0.13
0.21
0.20
ꢂ0.53
0.47
ꢂ0.58
ꢂ0.60
0.42
0.26
0.16
0.21
0.20
ꢂ0.27
0.17
ꢂ0.59
ꢂ0.64
0.55
0.22
0.15
0.20
0.19
ꢂ1
O
1
5 19
examined at Ti5cAO2cATi5c sites on anatase Ti O H18 clusters
H
H
H
H
H
H
1
2
3
4
5
6
(Scheme 1a); the DFT-derived structures of the reactants, transi-
tion states, and products involved (shown in Section S6, SI) are
similar to those at Ti5cAO3cATi5c sites, but with O2c instead of
O3c sites as the a-H-atom abstractor and with the proton bound
0.19
0.20
0.19
at the O2c site before reprotonation of the bound product of the
enolate–propanal CAC coupling step (Step 5; Scheme 4). The free
energies of transition states involved in enolate formation and its
Sum
0.24
0.21
0.17
a
B3LYP, 6-311G(d,p) for C, H, and O-atoms, LANL2DZ for Ti-atoms.
CAC coupling reaction (Steps 2 and 4; Scheme 4) were both
ꢂ1
ꢀ
15 kJ mol smaller than at Ti5cAO3cATi5c sites (referenced to a
The free energy reaction coordinate diagrams are depicted in
Fig. 10 for aldol condensation and esterification reactions of
-propanol–propanal reactants on Ti5cAO3cATi5c sites of the anatase
5 19 18
Ti 18 cluster. All energies are referenced to a bare Ti O H
5 19
bare Ti O H18 cluster and two gaseous propanal reactants, 523 K
and 1 bar; Section S6, SI).
1
The greater stability of these enolate formation and CAC coupling
transition states on the Ti5cAO2cATi5c sites reflects shorter Ti5cAO2c
bonds (0.183 nm) than in Ti5cAO3cATi5c sites (0.203 nm), even
though Ti5cAO3cATi5c and Ti5cAO2cATi5c sites have identical acid
and base properties (Table 4). The Ti5cAO2c bonds allow more effec-
tive concerted coordination at transition states, as evident from the
5
O
19
H
cluster and the respective gaseous 1-propanol or propanal reac-
tants, and free energies are given at 523 K and 1 bar. The corre-
sponding enthalpy and entropy components are reported in the
SI (Section S5). The free energy of the enolate TS from propanal
ꢂ1
0z
reactants (122 kJ mol ), corresponding to
higher than those for the subsequent CAC and CAO formation
steps, consistent with the kinetic relevance of a common -CAH
bond cleavage in the formation of esterification and condensation
products and in agreement with the experimental measurements
124 ± 1 kJ mol , Table 3). The DFT-derived Gibbs free energy for
adsorbed propanal at the Ti5c site on the Ti
is 1 kJ mol (Fig. 10), which gives, in term, an equilibrium adsorp-
tion constant (Kal):
D
G
in Eq. (13), is
enolate
shorter
a-CAH bond at the enolate formation TS for Ti5cAO2c
(
0.136 nm, Section S6, SI) than for Ti5cAO3c site pairs (0.140 nm,
a
Scheme 8). The reprotonation TS of the aldol precursor (Step 5;
ꢂ1
Scheme 4), however, was 79 kJ mol higher on Ti5cAO2cATi5c than
z
ꢂ1
on Ti5cAO3cATi5c sites (
DG
: 129 vs. 50 kJ mol ; Section S6, SI),
ꢂ1
reprot
(
as a consequence of the much larger distance between the proton
and the aldol precursor on Ti5cAO2cATi5c sites (0.396 nm, Section S6,
SI), compared to the corresponding distance on Ti5cAO3cATi5c sites
5
O
19
H
18 cluster (
D
G
ads
)
ꢂ1
(
0.188 nm, Section S5, SI). On Ti5cAO2cATi5c sites, the reprotonation
ꢀꢂ
D
G
RT
ꢁ
TS had a higher free energy than those involved in enolate formation
ads
K
al ¼ exp
ð24Þ
ꢂ1
ꢂ1
(
(
108 kJ mol ) and CAC coupling (58 kJ mol ) transition states
Section S6, SI), thus making reprotonation the kinetically-relevant
ꢂ1
of 0.008 kPa , consistent with surfaces that are essentially uncov-
ered by intermediates during catalysis at the conditions of this
step and its overall activation free energy higher than on
ꢂ1
Ti5cAO3cATi5c sites (129 vs. 122 kJ mol ). The kinetic relevance of

S. Wang et al. / Journal of Catalysis 340 (2016) 302–320
317
Fig. 10. DFT-derived reaction free energy diagram of aldol condensation (solid lines) and esterification (dashed lines) pathways for 1-propanol–propanal reactants on
Ti5cAO3cATi5c sites of the anatase Ti 18 cluster (B3LYP, 6-311G(d,p) for C, H, and O-atoms, LANL2DZ for Ti-atoms; 523 K, 1 bar). TSj and Pj represent the respective
transition state and product of Step j in Scheme 4. Energies are relative to a bare cluster and two gaseous propanal (AL) and one gaseous 1-propanol (OL).
5 19
O H
this reprotonation step, taken together with DFT-estimated negligi-
derived TS structures exhibited a slightly longer
for propanal (0.143 vs. 0.140 nm, Schemes 8 and 9) and a slightly
shorter -H-O3c bond (0.123 vs. 0.127 nm, Schemes 8 and 9), con-
sistent with a TS that occurs somewhat later along the reaction
a-CAH bond than
ble propanal coverages on Ti5c sites as described above (
D
G
ads
ꢂ1
1
kJ mol , Fig. 10), leads to a rate equation:
a
z
DG
r
aldol
k
B
T
reprot
coordinate and with the different deprotonation energies (EDPE
)
ꢂ
2
¼
h
e
RT ðPpropanal
Þ
ð25Þ
of their respective -CAH bonds. These energies are given by
a
½
LꢄTiO2
E
DPE ¼ E
Z
ꢂ
þ EH
þ
ꢂ EHZ
ð26Þ
for aldol condensation rates (raldol) on Ti5cAO2cATi5c sites (deriva-
tion details in Section S6, SI), which is inconsistent with the mea-
sured first-order rate dependence on propanal pressure
ꢂ
, and E+
where EHZ, E
Z
H
are respective electronic energies of a neutral
ꢂ
molecule (HZ), the anion deprotonated from HZ (Z ), and a bare
(
Section 3.4). Moreover, the overall activation free energy difference
gaseous proton. DFT-derived DPE values for the
a
-H-atoms in pro-
ꢂ1
ꢂ1
for propanal condensation on Ti5cAO2cATi5c (129 kJ mol ) and
Ti5cAO3cATi5c (122 kJ mol ) sites leads to a ratio of 0.0004 for con-
panal and acetone were 1622 and 1648 kJ mol , respectively
(Table 6), consistent with the latter TS for the bond with the larger
heterolytic dissociation energy. DFT-derived structures of the reac-
tants, transition states, and products are similar for enolate forma-
tion and CAC coupling steps in acetone and propanal reactions
(Section S7, SI), indicative of the general nature of these condensa-
tion pathways on acid–base site pairs.
Fig. 11 depicts the free energy reaction coordinate diagram for
acetone condensation (Scheme 7; the respective enthalpies and
entropies are shown in Section S7, SI), referenced to a bare
ꢂ1
densation rates on Ti5cAO2cATi5c sites to those on Ti5cAO3cATi5c
sites at 523 K and 0.18 kPa propanal (a typical condition examined
in this study), according to Eqs. (11) and (25); the much lower reac-
tivity of Ti5cAO2cATi5c sites compared to Ti5cAO3cATi5c sites reflects
that modest distances of the O site with the two Ti sites involved in
the TiAOATi sites are required for efficient aldol condensation turn-
overs. This spatial consequence of TiAOATi sites was also found on
2 2
TiO (r) (110) surfaces, which renders TiO (r) (110) surfaces inac-
tive for aldol condensation turnovers (Section S8, SI).
5 19
Ti O H18 cluster and two gaseous acetone molecules. The enolate
Ti5cAO3cATi5c sites were chosen to examine acetone condensa-
tion pathways (Scheme 7) using DFT methods. DFT-derived TS
structures for enolate formation are similar for acetone (Scheme 9)
and propanal (Scheme 8); in both cases, TS structures involve con-
certed interactions of Ti5c and O3c sites with carbonyl O-atoms and
formation TS gives the highest free energy along the reaction coor-
dinate (Fig. 11), indicating that this is the sole kinetically-relevant
step, as found for propanal, consistent with the measured first-
order condensation rate dependence on acetone pressure (Fig. 9)
and the normal kinetic H/D isotope effect (2.4, 523 K, Table 3).
The free energy of the enolate formation TS thus accounts for the
a-H-atoms in the enolate precursors, respectively. Acetone-
5 19
Scheme 9. DFT-derived structures involved in enolate formation reactions from acetone on Ti5cAO3cATi5c sites of the anatase Ti O H18 cluster (B3LYP, 6-311G(d,p) for C, H,
and O-atoms, LANL2DZ for Ti-atoms; only Ti5cAO3cATi5c sites and two O-atoms directly bound with the Ti5c sites are shown for clarity).

3
18
S. Wang et al. / Journal of Catalysis 340 (2016) 302–320
Table 6
0
z
0z
0z
reactants.a
Deprotonation energy (EDPE) of
a
-H-atom and DFT-derived DH
, DS
and DG
for propanal, acetone, and acetone-d
6
enolate
enolate
enolate
ꢂ1
z
ꢂ1
z
ꢂ1 ꢂ1
z
ꢂ1
Reactant
EDPE (kJ mol
)
DH
(kJ mol
)
D
S
(J mol
K
)
DG
(kJ mol
)
enolate
enolate
enolate
Propanal
Acetone
Acetone-d
1622
1648
1648
9
17
21
ꢂ215
ꢂ217
ꢂ217
122
131
135
6
a
Using Ti5cAO3cATi5c sites of the anatase Ti
5
O
19
H18 cluster; 523 K, 1 bar, B3LYP, 6-311G(d,p) for C, H, and O-atoms, and LANL2DZ for Ti-atoms.
5 19
Fig. 11. DFT-derived reaction free energy diagram of acetone condensation on Ti5cAO3cATi5c sites of the anatase Ti O H18 cluster (B3LYP, 6-311G(d,p) for C, H, and O-atoms,
LANL2DZ for Ti-atoms; 523 K, 1 bar). TSj and Pj represent the respective transition state and product of Step j in Scheme 7. Energies are relative to a bare cluster and two
gaseous acetone reactants (ONE).
0
z
ꢂ1
results are included in SI (Section S8). These DFT treatments show
that relevant TiAOATi sites on rutile (110) surface have stronger
acid and basic characters than the active Ti5cAO3cATi5c species
on anatase (101) surface and a longer TiAO distance on rutile than
anatase (0.323 vs. 0.203 nm); these properties lead to the interme-
diates and transition states involved in acetone condensation on
measured
D
G
value (Eq. (12); 134 ± 1 kJ mol , Table 3), which
enolate
ꢂ1
agrees well with the theoretical estimate (131 kJ mol , Table 6).
0z
The DFT-derived
kJ mol larger than for undeuterated acetone (Table 6), as a
DG
enolate
6
value for acetone-d reactants was
ꢂ1
4
result of the larger free energy difference between the enolate
formation TS and the gaseous reactant for deuterated species;
this free energy difference gives a kinetic isotope effect of 2.5
ꢂ1
rutile were more stable than on anatase by >50 kJ mol . In partic-
ꢂꢂ
ꢃ ꢄꢂ
ꢃ ꢃ
ular, the dimer species formed in CAC coupling steps (Step 4;
0
0
k
k
at 523 K, consistent with the measured value
enolate
H
enolate
D
ꢂ1
Scheme 7) are very strongly-bound (ꢂ266 kJ mol , referenced to
(
2.4, Table 3).
ꢅ
ꢆ
5 19
a bare Ti O H18 cluster and two gaseous acetone molecules; Sec-
ꢂ
ꢃ
0
z
0z
DFT-derived enthalpy
DH
and entropy
D
S
contri-
tion S8, SI), and the difficult reprotonation of these stable dimer
enolate
enolate
0
z
species limits the CAC coupling turnovers, which would lead to
butions to activation free energies (
D
G
) are shown in Table 6
enolate
13
0
z
ꢂ1
undetectable rates on rutile (e.g. ꢀ10 -fold lower than on anatase;
for propanal and acetone reactants.
smaller for propanal than acetone, thus accounting for the reactiv-
D
H
values were 8 kJ mol
enolate
5
23 K, 0.8 kPa acetone). Such difficult reprotonation of the aldol
precursors on the relevant TiAOATi sites of rutile reflects the
strong acid and basic characters of the Ti and the O sites, respec-
tively, which stabilizes both reactants involved (the precursor at
Ti centers and the proton at O sites), combined with their large
separation (0.403 nm; Section S8, SI) that is reminiscent of the spa-
tial effects at the less active Ti5cAO2cATi5c sites on anatase. The
free energy barrier from the bound dimer species to the reprotona-
0
z
ity differences between these two molecules (
kJ mol ). The relative reactivity of these C
formation is thus enthalpy-determined, reflecting their similar
entropy changes from a gaseous reactant to the enolate formation
DG
difference
enolate
ꢂ1
9
3
molecules in enolate
0
z
TS.
-CAH bonds are partially cleaved at the enolate formation TS
Schemes 8 and 9). DPE predominantly depends, in turn, on the
effects of substituents on the stability of the anion formed. The
methyl group bound at the -C-atom in propanal leads to more
stable enolate species (via p– conjugation with the C@C bond)
DH
values depend on DPE (a-H) in reactants because their
enolate
a
(
ꢂ1
tion TS was much larger on rutile (289 kJ mol , 523 K; Section S8,
ꢂ1
SI) than at Ti5cAO2cATi5c sites on anatase (116 kJ mol , 523 K;
a
Section S7, SI), in spite of their similar distances between the aldol
precursor and the proton (0.403 vs. 0.396 nm), indicating that
acid–base strength, instead of the distance between the aldol pre-
cursor and the proton, accounts for the difficult reprotonation, and
p
than the methyl group at the carbonyl C-atom in acetone, which
destabilizes the enolate (via repulsion between the electron-
donating methyl group and the O anion in the enolate formed)
thus the unreactive nature of TiO (r) surfaces.
2
[
54].
Aldol condensation pathways were also examined on TiO
using the above DFT methods and a Ti 18 cluster extracted
from the most stable exposed rutile (1 1 0) plane [31]; detailed
In summary, we conclude that aldol condensation turnovers
2
(r)
require acid–base site pairs of moderate acid–base strength,
located within moderate distances to allow the concerted stabiliza-
tion of enolate formation transition states while preventing the
5 19
O H

S. Wang et al. / Journal of Catalysis 340 (2016) 302–320
319
stranding of condensation products by the strong stabilization of
Acknowledgments
aldol precursors and protons by pairs with strong acid and basic
sites, respectively. These conclusions apply also to reactions of
The authors acknowledge with thanks thoughtful technical dis-
cussions with Drs. Eric Doskocil, John Shabaker, Glenn Sunley, and
the rest of the XC program members at BP p.l.c., and with Drs.
2
C –C
5
alkanals and alkanones and for acid–base site pairs on other
) [29]. Notably, the site
2
oxides (monoclinic and tetragonal ZrO
2
pairs of moderate acid–base strength required for condensation
and esterification turnovers also prevent the strong inhibition by
Stanley Herrmann, Prashant Deshlahra, and Elif Gurbuz at UC-
Berkeley throughout the course of this study. The authors are also
grateful for the access to computational facilities provided by the
Extreme Science and Engineering Discovery Environment (XSEDE)
(NSF, grant number ACI-1053575) and the Molecular Graphics and
the Computation Facility at the University of California, Berkeley
(NSF, grant number CHE-0840505). S.W. kindly acknowledges Dr.
Stephanie Kwon (UC-Berkeley) for a careful review of the support-
ing information. This research project was supported as part of the
H
2
O (or CO
high turnover rates even at high conversions. When protected
through these bifunctional strategies conferred by H , TiO (a)
and ZrO allow high and stable condensation and esterification
rates and selectivities at practical conditions.
2
) ubiquitous on strongly basic oxides [55], leading to
2
2
2
2
4
. Conclusions
BP Conversion Consortium (BP-XC ) at the University of California,
Berkeley.
Addition of a Cu cocatalyst and H
constraints and avoids deactivation for TiO
and acetone condensation reactions by scavenging equilibrium-
limited unsaturated condensation products to more stable alka-
nals, alkanones, and alkanols; the hydrogenation–dehydrogenation
2
removes thermodynamic
-catalyzed propanal
2
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.2016.05.026.
3
function conferred by Cu also equilibrates C alkanols and their
carbonyl analogs, providing lumped reactant pools, and catalyzes
esterification for 1-propanol–propanal reactants. This bifunctional
References
2
strategy increases net rates, selectivity and stability of TiO -
[
[
1] G.W. Huber, S. Iborra, A. Corma, Chem. Rev. 106 (2006) 4044–4098.
2] D.A. Simonetti, J.A. Dumesic, Catal. Rev. 51 (2009) 441–484.
catalyzed aldol condensation reactions over a large range of con-
versions, rendering these reactions practical for chain growth and
O-removal of oxygenates.
TiAO site pairs of intermediate acid–base strength and TiAO
distances, prevalent on anatase surfaces (but not rutile surfaces),
principally account for measured condensation and esterification
[3] J.C. Serrano-Ruiz, R.M. West, J.A. Dumesic, Annu. Rev. Chem. Biomol. Eng. 1
(2010) 79–100.
[
4] T.V. Choudhary, C.B. Phillips, Appl. Catal. A 397 (2011) 1–12.
5] M.J.A. Tijmensen, A.P.C. Faaij, C.N. Hamelinck, M.R.M. van Hardeveld, Biomass
Bioenergy 23 (2002) 129–152.
[
[6] J. Sun, R.A.L. Baylon, C. Liu, D. Mei, K.J. Martin, P. Venkitasubramanian, Y. Wang,
J. Am. Chem. Soc. 138 (2016) 507–517.
[
2
rates on TiO ; these active TiAO site pairs were counted accurately
7] N. Li, G.W. Huber, J. Catal. 270 (2010) 48–59.
8] S. Wang, Y. Zhang, H. Liu, Chem. Asian J. 5 (2010) 1100–1111.
via site titration with propanoic acid during catalysis, allowing for
rigorous calculations of turnover rates and activation free energy
barriers. Enolate formation from alkanal (alkanone) reactants on
sparsely covered TiAO site pairs, using basic O sites to abstract
[
[9] Y. Lin, S. Tanaka, Appl. Microbiol. Biotechnol. 69 (2006) 627–642.
10] D.A. Simonetti, J.A. Dumesic, ChemSusChem 1 (2008) 725–733.
11] E.L. Kunkes, D.A. Simonetti, R.M. West, J.C. Serrano-Ruiz, C.A. Gartner, J.A.
Dumesic, Science 322 (2008) 417–421.
[
[
a
-H-atoms and acidic Ti sites to stabilize enolate moieties concert-
[12] J.I. Di Cosimo, V.K. Díez, C.R. Apesteguía, Appl. Catal. A 137 (1996) 149–166.
[
13] W. Shen, G.A. Tompsett, R. Xing, W.C. Conner Jr., G.W. Huber, J. Catal. 286
2012) 248–259.
edly, is kinetically-relevant to both aldol condensation and esteri-
fication turnover rates; the dependence of enolate formation
turnover rates on alkanal (alkanone) reactant pressures reflects
the free energy of the enolate formation transition state with
respect to a gaseous alkanal (alkanone) reactant and a bare TiAO
site pair. Formed enolate species react via nucleophilic attack
either with an alkanal (alkanone) bound on a vicinal Ti site to form
(
[
14] J.I. Di Cosimo, C.R. Apesteguía, M.J.L. Ginés, E. Iglesia, J. Catal. 190 (2000) 261–
275.
[15] J.I. Di Cosimo, G. Torres, C.R. Apesteguía, J. Catal. 208 (2002) 114–123.
[16] G. Torres, C.R. Apesteguía, J.I. Di Cosimo, Appl. Catal. A 317 (2007) 161–170.
[17] A.A. Nikolopoulos, B.W.-L. Jang, J.J. Spivey, Appl. Catal. A 296 (2005) 128–136.
[18] M.I. Zaki, M.A. Hasan, L. Pasupulety, Langmuir 17 (2001) 768–774.
[19] E.L. Kunkes, E.I. Gürbüz, J.A. Dumesic, J. Catal. 266 (2009) 236–249.
[
[
20] A. Gangadharan, M. Shen, T. Sooknoi, D.E. Resasco, R.G. Mallinson, Appl. Catal.
A 385 (2010) 80–91.
21] D.G. Hanna, S. Shylesh, Y. Li, S. Krishna, M. Head-Gordon, A.T. Bell, ACS Catal. 4
(2014) 2908–2916.
an a,b-unsaturated carbonyl compound or with an alkanol bound
on the vicinal Ti site to form an hemiacetal intermediate (via
sequential H-shift and CAO coupling reactions) followed by dehy-
drogenation to esters on Cu; esterification/condensation selectivi-
ties on TiAO site pairs are determined by the differences in free
energies between enolate-alkanal (CAC) and enolate-alkanol
[22] J.E. Rekoske, M.A. Barteau, Ind. Eng. Chem. Res. 50 (2011) 41–51.
[23] S. Luo, J.L. Falconer, J. Catal. 185 (1999) 393–407.
[24] A.C.C. Rodrigues, J.L.F. Monteiro, Appl. Catal. A 362 (2009) 185–192.
[25] M.J.L. Ginés, E. Iglesia, J. Catal. 176 (1998) 155–172.
[
26] E. Iglesia, D.G. Barton, J.A. Biscardi, M.J.L. Gines, S.L. Soled, Catal. Today 38
1997) 339–360.
27] M. Singh, N. Zhou, D.K. Paul, K.J. Klabunde, J. Catal. 260 (2008) 371–379.
(
CAO) coupling transition states and between gaseous alkanal
(
and alkanol reactants.
[
These mechanistic conclusions are consistent with DFT simula-
tions using anatase cluster models on kinetic consequences of the
proposed condensation and esterification elementary steps, and on
both isotope and substituent effects on enolate formation rates.
Similar simulations on rutile cluster models show that available
TiAO site pairs on rutile surfaces are spatially-separated and have
stronger acid–base strength than those on anatase surfaces; aldol
condensation reactions on these site pairs are unfeasible because
of high free energy barriers required for reprotonation of the
CAC coupling products to form aldols. This marked reactivity dif-
ference between anatase and rutile surfaces evident from both
experiments and theory reveals that Lewis acid–base site pairs of
modest acid–base strength and acid–base site distances are pre-
ferred for aldol condensation turnovers.
[28] W. An, Phys. Chem. Chem. Phys. 17 (2015) 22529–22532.
[
[
29] S. Wang, E. Iglesia, Unpublished Work.
30] C.J.G. van der Grift, P.A. Elberse, A. Mulder, J.W. Geus, Appl. Catal. A 59 (1990)
275–289.
[31] U. Diebold, Surf. Sci. Rep. 48 (2003) 53–229.
[
[
32] W-F. Huang, Hs-T. Chen, M.C. Lin, J. Phys. Chem. C 113 (2009) 20411–20420.
33] A.V. Bandura, D.G. Sykes, V. Shapovalov, T.N. Troung, J.D. Kubicki, R.A.
Evarestov, J. Phys. Chem. B 108 (2004) 7844–7853.
[34] R. Erdogan, M.F. Fellah, I. Onal, Int. J. Quant. Chem. 111 (2011) 174–181.
[35] (a) A.D. Becke, J. Chem. Phys. 98 (1993) 5648–5652;
(
b) C. Lee, W. Yang, R.G. Parr, Phys. Rev. B 37 (1988) 785–789.
[
36] M.J. Frisch, G.W. Trucks, H.B. Schlegel, G.E. Scuseria, M.A. Robb, J.R. Cheeseman,
G. Scalmani, V. Barone, B. Mennucci, G.A. Petersson, H. Nakatsuji, M. Caricato,
X. Li, H. P. Hratchian, A.F. Izmaylov, J. Bloino, G. Zheng, J.L. Sonnenberg, M.
Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y.
Honda, O. Kitao, H. Nakai, T. Vreven, J. Montgomery, Jr., J.E. Peralta, F. Ogliaro,
M. Bearpark, J.J. Heyd, E. Brothers, K.N. Kudin, V.N. Staroverov, R. Kobayashi, J.
Normand, K. Raghavachari, A. Rendell, J.C. Burant, S.S. Iyengar, J. Tomasi, M.

3
20
S. Wang et al. / Journal of Catalysis 340 (2016) 302–320
Cossi, N. Rega, N.J. Millam, M. Klene, J.E. Knox, J.B. Cross, V. Bakken, C. Adamo, J.
[44] S. Grimme, Chem. Eur. J. 18 (2012) 9955–9964.
Jaramillo, R. Gomperts, R.E. Stratmann, O. Yazyev, A.J. Austin, R. Cammi, C.
Pomelli, J.W. Ochterski, R.L. Martin, K. Morokuma, V.G. Zakrzewski, G.A. Voth,
P. Salvador, J.J. Dannenberg, S. Dapprich, A.D. Daniels, Ö. Farkas, J.B. Foresman,
J. V. Ortiz, J. Cioslowski, D.J. Fox, Gaussian Inc, Wallingford CT, 2009.
[45] R.L. Martin, J. Chem. Phys. 118 (2003) 4775–4777.
[46] T. Clark, J. Chandrasekhar, G.W. Spitznagel, P.v.R. Schleyer, J. Comput. Chem. 4
(1983) 294–301.
[47] L. Mino, G. Spoto, S. Bordiga, A. Zecchina, J. Phys. Chem. C 116 (2012) 17008–
17018.
[48] T. Ohno, K. Sarukawa, K. Tokieda, M. Matsumura, J. Catal. 203 (2001) 82–86.
[49] J. Zhang, M. Li, Z. Feng, J. Chen, C. Li, J. Phys. Chem. B 110 (2006) 927–935.
[50] M.E. Sad, M. Neurock, E. Iglesia, J. Am. Chem. Soc. 133 (2011) 20384–20398.
[51] K. Inui, T. Kurabayashi, S. Sato, J. Catal. 212 (2002) 207–217.
[52] E.I. Gürbüz, D.D. Hibbitts, E. Iglesia, J. Am. Chem. Soc. 137 (2015) 11984–
11995.
[
37] (a) A.D. McLean, G.S. Chandler, J. Chem. Phys. 72 (1980) 5639–5648;
b) R. Krishnan, J.S. Binkley, R. Seeger, J.A. Pople, J. Chem. Phys. 72 (1980) 650–
54.
(
6
[
[
[
[
38] P.J. Hay, W.R. Wadt, J. Chem. Phys. 82 (1985) 270–283.
39] X. Li, M.J. Frisch, J. Chem. Theor. Comput. 2 (2006) 835–839.
40] S. Grimme, S. Ehrlich, L. Goerigk, J. Comput. Chem. 32 (2011) 1456–1465.
41] (a) S.F. Boys, F. Bernardi, Mol. Phys. 19 (1970) 553–566;
(
b) S. Simon, M. Duran, J.J. Dannenberg, J. Chem. Phys. 105 (1996) 11024–11031.
[53] H. Liu, N. Bayat, E. Iglesia, Angew. Chem. Int. Ed. 42 (2003) 5072–5075.
[54] Y. Chiang, M. Hojatti, J.R. Keeffe, A.J. Kresge, N.P. Schepp, J. Wirzt, J. Am. Chem.
Soc. 109 (1987) 4000–4009.
[
[
42] J.P. Merrick, D. Moran, L. Radom, J. Phys. Chem. A 111 (2007) 11683–11700.
43] D.A. McQuarrie, Statistical Mechanics, University Science Books, Sausalito, CA,
2
000.
[55] H. Hattori, Appl. Catal. A 504 (2015) 103–109.