Aldol condensation among acetaldehyde and ethanol reactants on TiO2: Experimental evidence for the kinetically relevant nucleophilic attack of enolates

Article abstract of DOI:10.1016/j.jcat.2018.02.030

Combinations of rate measurements as functions of reactant pressures, in situ infrared spectroscopy, comparisons of kinetic isotope effects, and rate inhibition effects provide experimental evidence that aldol condensation of acetaldehyde proceeds by kinetically relevant nucleophilic attack of a reactive enolate upon an acetaldehyde molecule over anatase TiO₂. Steady-state turnover rates of aldol condensation measured as a function of the pressures of C₂H₄O, C₂H₅OH, H₂O, and H₂ between 503 K and 537 K show that rates reflect a second order dependence on C₂H₄O pressure and an inverse second order dependence on the C₂H₅OH pressure at the lowest C₂H₄O-to-C₂H₅OH ratios. Infrared spectra obtained in situ show that the exposed cationic Ti-atoms that facilitate aldol addition on TiO₂ surfaces are saturated with C₂H₅OH? species and C₂H₄O? coverages are much smaller. In addition, aldol rates increase when C₂D₄O replaces C₂H₄O as a reactant, which likely reflects an inverse, secondary isotope effect caused by rehybridization of C-atoms at the transition state that forms a C–C bond between two reactive intermediates derived from acetaldehyde. These results suggest that the kinetically relevant step is a bimolecular surface reaction, specifically the nucleophilic attack of an enolate onto a vicinal C₂H₄O? species. This conclusion is consistent also with aldol condensation rates that decrease with an inverse second order dependence on pyridine (C₅H₅N) pressure, because C₅H₅N displaces C₂H₄O from the two Lewis acid sites involved in the kinetically relevant step (confirmed by in situ FTIR). Comparisons to recent reports on the mechanism of this reaction on anatase TiO₂ indicate that the presence of high coverages of C₂H₅OH? causes nucleophilic attack to become the kinetically relevant step by significantly reducing the number of enolate-acetaldehyde reactant pairs upon the surface.

Full text of DOI:10.1016/j.jcat.2018.02.030

Journal of Catalysis 361 (2018) 290–302
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Aldol condensation among acetaldehyde and ethanol reactants on TiO2:
Experimental evidence for the kinetically relevant nucleophilic attack of
enolates
⇑
Hongbo Zhang, Malek Y.S. Ibrahim, David W. Flaherty
Department of Chemical and Biomolecular Engineering, University of Illinois Urbana Champaign, Urbana, IL 61801, United States
a r t i c l e i n f o
a b s t r a c t
Article history:
Combinations of rate measurements as functions of reactant pressures, in situ infrared spectroscopy,
comparisons of kinetic isotope effects, and rate inhibition effects provide experimental evidence that
aldol condensation of acetaldehyde proceeds by kinetically relevant nucleophilic attack of a reactive eno-
late upon an acetaldehyde molecule over anatase TiO2. Steady-state turnover rates of aldol condensation
measured as a function of the pressures of C2H4O, C2H5OH, H2O, and H2 between 503 K and 537 K show
that rates reflect a second order dependence on C2H4O pressure and an inverse second order dependence
on the C2H5OH pressure at the lowest C2H4O-to-C2H5OH ratios. Infrared spectra obtained in situ show
that the exposed cationic Ti-atoms that facilitate aldol addition on TiO2 surfaces are saturated with
C2H5OH⁄ species and C2H4O⁄ coverages are much smaller. In addition, aldol rates increase when
C2D4O replaces C2H4O as a reactant, which likely reflects an inverse, secondary isotope effect caused
by rehybridization of C-atoms at the transition state that forms a CAC bond between two reactive inter-
mediates derived from acetaldehyde. These results suggest that the kinetically relevant step is a bimolec-
ular surface reaction, specifically the nucleophilic attack of an enolate onto a vicinal C2H4O⁄ species. This
conclusion is consistent also with aldol condensation rates that decrease with an inverse second order
dependence on pyridine (C5H5N) pressure, because C5H5N displaces C2H4O from the two Lewis acid sites
involved in the kinetically relevant step (confirmed by in situ FTIR). Comparisons to recent reports on the
mechanism of this reaction on anatase TiO2 indicate that the presence of high coverages of C2H5OH⁄
causes nucleophilic attack to become the kinetically relevant step by significantly reducing the number
of enolate-acetaldehyde reactant pairs upon the surface.
Received 10 December 2017
Revised 19 February 2018
Accepted 27 February 2018
Keywords:
Aldol addition
CAC bond formation
Nucleophilic attack
Biomass upgrading
In situ spectroscopy
Kinetics
Ó 2018 Elsevier Inc. All rights reserved.
1. Introduction
mass derived products. One such pathway is the Guerbet reaction
of ethanol, which utilizes aldol condensation to form CAC bonds
The fermentation of glucose, fructose or sucrose obtained from
biomass sources can be used to provide ethanol [1–3] or butanol,
albeit with lower efficiencies [4]. The supply of ethanol surpasses
its current demand as a fuel or fuel additive, and therefore, cat-
alytic upgrading of ethanol and other small alcohols via the Guer-
bet reaction [5] or similar networks of dehydrogenation,
condensation, and hydrogenation steps have attracted significant
attention due to their potential to produce larger aldehydes, enals,
and alcohols that are useful chemical intermediates or fuels [6–9].
The mechanisms for CAC bond formation among aldehyde and
alcohol reactants have been studied extensively for metal oxide
catalysts due to the potential of these reactions for upgrading bio-
among aldehyde intermediates [5,10]. A number of catalysts cat-
alyze the Guerbet reaction (or subsets of the intermediate steps)
including MgO [11–13], MgxAlOy (hydrotalcites) [14,15], MgxCeOy
[16], Mg3(PO4)2 [17], ZnxZrOy [18], hydroxyapatites (HAP) [17,19–
24], and TiO2 [19,25–31]. On these surfaces, aldol condensation of
acetaldehyde proceeds by deprotonation of adsorbed acetaldehyde
at the a-carbon to generate an enolate; nucleophilic attack of the
enolate to the carbonyl group of a vicinal acetaldehyde molecule
to form the CAC bond; protonation of the C4-intermediate to gen-
erate the aldol molecule; and subsequent dehydration to form 2-
butenal [25,31,32]. Following aldol condensation, additional reac-
tions among 2-butenal and other co-reactants form 2-buten-1-ol,
n-butanal, butanol and C4+ aldehydes and alcohols as well as aro-
matics (e.g., 2- and 4-methylbenzaldehydes and alcohols)
[5,19,20,24,25,28,33]. The identity of the proposed kinetically rele-
⇑
Corresponding author.
E-mail address: dwflhrty@illinois.edu (D.W. Flaherty).
https://doi.org/10.1016/j.jcat.2018.02.030
0021-9517/Ó 2018 Elsevier Inc. All rights reserved.

H. Zhang et al. / Journal of Catalysis 361 (2018) 290–302
291
vant steps (and the identity of the active sites associated with
these steps) for aldol condensation differs between oxide catalysts
(e.g., TiO2, MgO, HAP, and MgxAlOy) [15,19,20,24,25,28].
from enolate to molecularly adsorbed acetaldehyde to form a
CAC bond in the presence of reactant mixtures containing
acetaldehyde and ethanol. Notably, this bimolecular step occurs
on surfaces that are largely covered by persistent carbonaceous
residues (ꢁ95% of sites), with active sites primarily covered by
ethanol-derived species, and involves acetaldehyde-derived spe-
cies stabilized at proximate Lewis acid sites. Consequently, rates
of nucleophilic attack are more sensitive to the density and prox-
imity of catalytically active Lewis acid sites than monomolecular
pathways such as adsorption or deprotonation. This insight sug-
gests that dramatic decreases in the coverage of deactivating car-
bonaceous will increase the relative coverage of acetaldehyde
and shift the kinetic regime to one limited by monomolecular
deprotonation, whereas, decreases in only the coverage of ethanol
and ethoxide may result in rates that are limited by the desorption
of strongly bound reactants and the subsequent adsorption of
acetaldehyde.
Within the past several years, conflicting reports have proposed
that the adsorption of aldehydes, the desorption of aldol products
[19], and deprotonation of aldehydes/ketones are kinetically rele-
vant (i.e., rate determining) steps upon primarily anatase TiO2 cat-
alysts. Wang et al. [25] demonstrated that the consumption rates
of equilibrated pools of propanal-propanol reactants increase in
proportion to the concentration of propanal and do not depend
on the pressure of propanol over TiO2 (physically mixed with Cu
nanoparticles supported on silica to equilibrate dehydrogenation
reactions), which together with DFT estimates for activation ener-
gies suggests that rates of propanal-propanol consumption are lim-
ited by deprotonation of aldehydes on largely unoccupied active
sites [25]. These conclusions were supported also by the observa-
tion that aldol addition rates among acetone-2-propanol reactants
exhibit a significant, normal kinetic isotope effect (KIE; kH/kD = 2.4)
[25]. More recently, Wang et al. reported similar results for aldol
condensation reactions among mixtures of C2- and C4-aldehydes
and ketones over TiO2 or ZrO2 combined with Cu-SiO2 using the
same combination of methods, and concluded that deprotonation
is kinetically relevant also in these systems [28]. In contrast, Young
et al. measured aldol condensation rates on anatase TiO2 alone (i.e.,
in the absence of Cu co-catalysts) and observed that rates show a
first order dependence on acetaldehyde pressure (in the absence
of ethanol), a negligible KIE when acetaldehyde is replaced with
perdeuterated acetaldehyde (kH/kD = 0.98), and an inverse depen-
dence on the ethanol pressure [19]. These results were interpreted
as evidence that adsorption of acetaldehyde and, perhaps, the des-
orption of products (e.g., 2-butenal) are kinetically relevant for
acetaldehyde conversion on surfaces primarily saturated by higher
carbon number products [19]. These conclusions were supported
also by microcalorimetric measurements that showed heats of
adsorption for both acetaldehyde and ethanol onto TiO2 approach
100 kJ molꢀ1 at low coverages, which suggest that aldehydes, alco-
hols, or more strongly binding species saturate active sites of TiO2
during catalysis [19]. The striking disagreement between the con-
clusions of these studies suggests that seemingly small differences
in the preparation of the catalyst or the conditions for catalysis
(e.g., the presence of a second catalytic function (Cu) or a potential
co-reactant (ethanol)) may yield significant changes in both sur-
face chemistry and the coverage of active species
[19,25,27,28,31,34–36]. Moreover, comparisons of these previous
results indicates that these small changes may in fact influence
the prevailing kinetic regime for aldol condensation on titania cat-
alysts and change the identity of the elementary steps that deter-
mine reactant conversion rates [19,25–28,30,31,34].
2. Experimental methods and catalyst characterization
2.1. Preparation of TiO2 powder catalysts
Titania (TiO2) catalysts were prepared using a variation of a sol-
gel method previously reported [37]. Briefly, 26 cm3 of ethanol
(C2H5OH, Decon Laboratories, 200 proof), 4 cm3 of acetic acid (H3-
CCOOH, Sigma Aldrich, >99.0%), and 20 cm3 of tetrabutyl titanate
(TBT, Sigma Aldrich, >99.0%,) were combined, and the derived solu-
tion was stirred for at least 30 min to achieve homogeneity. A sec-
ond mixture was prepared from mixing 8 cm3 of deionized (DI)
water (Barnstead E-Pure, 17.6 MX
) with 12 cm3 of C2H5OH and
12 cm3 of H3CCOOH and was subsequently added to the initial
solution in a dropwise fashion (over a period of 1 hr) at ambient
temperature under stirring to create an uniform gel. The resultant
gel was aged at ambient conditions for 24 h and was then dried in
an oven at 393 K for at least 24 h. The dried material was heated to
673 K at 10 K minꢀ1 and held for 4 h under flowing air (200 cm3
minꢀ1, Airgas, >99.999%). This solid was then separated into three
batches, which were either stored or further treated to 773 K for 4
h or 873 K for 4 h in flowing air with the intent to change the crys-
tallinity and surface area of the TiO2. The final catalysts are
denoted as TiO2, 673, TiO2, 773 or TiO2, 873, where the subscript indi-
cates the maximum temperature, in Kelvin, used during thermal
treatment.
2.2. Structural characterization of TiO2 catalysts
Powder X-ray diffraction (XRD) patterns were collected on a
Here, we seek to elucidate the mechanism and site require-
ments for aldol addition of acetaldehyde on anatase TiO2 and pro-
vide plausible explanations for the apparent discrepancies
Siemens Bruker D5000 using Cu Ka radiation (k = 0.1542 nm, 40
kV, 30 mA) between 10° and 60° (2h) with 0.05° steps and a scan-
ning speed of 2° minꢀ1. The XRD patterns of all three materials are
shown in the Supporting Information (Fig. S1) and all detected
peaks match anatase phase of TiO2 from the literature [38]. The
mean diameters of the crystalline grains for the three TiO2 samples
were determined following the Scherrer equation applied to the
anatase (1 0 1) feature at 25.4° [39] and are given in Table 1.
Brunauer-Emmett-Teller (BET) surface area measurements
were performed by Micromeritics Corporation using N2 adsorption
and desorption isotherms were measured at 77 K (Micromeritics,
ASAP2020). The samples (>200 mg) were degassed at 573 K under
high vacuum condition for 2 h prior to isotherm measurement, and
the amount of N2 adsorbed was determined at 29.3 kPa and the
specific surface areas were calculated using BET theory. Table 1
shows that the specific surface area of TiO2 catalysts decrease from
106 to 88 m2 gꢀ1 with oxidative treatment temperatures that
increase from 673 to 873 K.
between
the
conclusions
of
previous
investigations
[19,25,27,28,31,34,35]. The conclusions presented here are sup-
ported by a combination of complementary methods that include
measurements of rates as functions of independently controlled
reactant pressures, identification of the prevalent active surface
intermediates by in situ infrared spectroscopy (IR), observation of
inverse kinetic isotope effects upon substituting C2D4O for C2H4O
in reactant streams, and the determination of active site functions
and numbers by in situ adsorption of probe molecules. The kinetic
studies suggest that the kinetically relevant step for aldol addition
involves a bimolecular process between two acetaldehyde derived
intermediates on active sites that are predominantly covered by
ethanol (or ethoxide), as shown by in situ IR. Collectively, these
findings provide compelling evidence that the mechanism for aldol
condensation proceeds by kinetically relevant nucleophilic attack

292
H. Zhang et al. / Journal of Catalysis 361 (2018) 290–302
Table 1
Physical and surface chemical characterization of TiO2 catalysts.
Sample
Oxidative
treatment
temperature [K]
Number of NH3
binding sites
Number of active
Surface area
Crystallite
structured
Crystallite
size [nm]d
b
c
sites [mmol (g TiO2)ꢀ1
]
[m2 gꢀ1
]
a
[mmol (g TiO2)ꢀ1
]
TiO2,
TiO2,
TiO2,
673
773
873
0.42 0.01
0.22 0.02
0.13 0.01
0.14 0.02
0.15 0.03
0.08 0.03
106
95
88
2
2
1
anatase
anatase
anatase
13
15
22
673
773
873
a
b
c
Ammonia binding sites were determined from 15NH3-14NH3 exchange experiments (Section 2.4).
Lewis acid base pairs were determined from in situ butyric acid titrations (Section 2.6).
Surface areas were determined from BET analysis (Section 2.2).
d
The crystallographic phase and crystallite sizes were determined from X-ray diffractograms and analysis of the peak width for the (1 0 1) anatase feature using the
Scherrer equation, respectively (Section 2.2).
2.3. In situ transmission Fourier transform infrared spectroscopy
during aldol condensation
monitor the change in the partial pressures of 14NH3, 15NH3, and
Ar in the reactor effluent as a function of time, which was used
(together with the volumetric flow rate of the effluent stream) to
calculate the number of NH3 adsorption sites on each catalyst
and the rate of NH3 desorption from those sites. Sites that bind
NH3 are acidic and a fraction of these sites participate in the aldol
addition reaction, as discussed later.
TiO2 (>200 mg) was placed into the quartz reactor and heated to
623 K for 3 h under flowing He (50 cm3 minꢀ1) to remove moisture
and volatile species from the surface. The reactor was then cooled
down to 423 K and held at 423 K for 30 min in flowing He, after
which, the catalyst was exposed to dilute 15NH3 (0.12 mol %
15NH3, 55 cm3 minꢀ1) for a period of ꢁ2 h to achieve a constant
coverage of 15NH3 on TiO2, as shown by QMS measurements. Sub-
sequently, the inlet gas stream was switched rapidly (<20 ms)
using a 4-way pneumatic valve to replace the 15NH3-He mixture
with dilute 14NH3 with an Ar tracer (0.12 mol% 14NH3, 0.12 mol%
Ar in He, 55 cm3 minꢀ1). The flow rate and the partial pressures
of NH3 isotopologues in the two gas mixtures were equal to ensure
that the total coverage of all NH3 isotopes remains constant before
and after the isotope switch. Intensities of each molecule from the
effluent, such as NH3 (m/z = 17, 16 and 15), 15NH3 (m/z = 18, 17 and
16), Ar (m/z = 40), He (m/z = 4), N2 (m/z = 17 and 34), O2 (m/z = 16
and 32) and H2O (m/z = 16, 17 and 18), were recorded continuously
with the QMS.
The coverages of surface intermediates derived from gaseous
C2H4O, C2H5OH and pyridine (C5H5N) were probed using in situ
transmission Fourier transform infrared (FTIR) spectroscopy within
a custom-made transmission cell, described previously [20,40].
Catalyst powders (50–100 mg) were pressed into self-supporting
wafers and placed within a temperature-controlled stainless-
steel cell equipped with CaF2 windows and connected to a gas
manifold, which enabled in situ use of FTIR during thermal treat-
ments and during catalysis. The temperature of the cell was con-
trolled using resistive heating cartridges connected to an
electronic PID controller (Watlow, EZ-Zone), and the temperature
of the wafer was measured with a K-type thermocouple (Omega)
located within the cell and in contact with the wafer. Catalyst
wafers were pretreated to at least 473 K for 1 h under flowing He
(100 cm3 minꢀ1), and then cooled prior to acquiring in situ FTIR
spectra. In experiments that involved adsorption of C5H5N (Fisher
Scientific, >99%), the system was first stabilized at the reaction
conditions (0.3–0.9 kPa C2H4O, 2.0–2.7 kPa C2H5OH, 15 kPa H2,
443–503 K) until all peak intensities became constant (usually
about 2 h), and then C5H5N (0.1–15 kPa) was introduced while
maintaining the partial pressures of C2H4O, C2H5OH and H2. Spec-
tra were recorded continuously by the spectrometer (Bruker, Ten-
sor 37).
The number of accessible Lewis acid sites per gram of TiO2
(mmol (g TiO2)ꢀ1) is determined from the number of moles of
14NH3 (or 15NH3) that adsorb (or desorb) from a known mass of
TiO2 (MTiO2) under steady-state flow of NH3 ([14NH3*]ss), which
are calculated by using the relationship [41]:
2.4. Steady-state isotopic transient kinetic analysis of 15NH3-14NH3
exchange
A custom-made steady-state isotopic transient kinetic analysis
(SSITKA) system (inspired by the design of Goodwin and coworkers
[41]) was used to determine the ammonia adsorption capacity and
rates of ammonia desorption from TiO2 surfaces at constant cover-
age by measuring exchange capacities and rates between 15NH3
and 14NH3. These experiments provide estimates for the total num-
ber of exposed Ti cations, which are weakly acidic. The quartz reac-
tor (12 mm outer diameter) was placed in a vertically-aligned
temperature-controlled tubular furnace (Lindberg Hevi-Duty,
55035-A). The temperature of the catalyst bed was measured with
a K-type thermocouple placed in contact with the outer wall of the
reactor adjacent to the catalyst bed and was controlled by a PID
controller (Watlow, EZ-Zone). Two separate gas manifolds were
used to control individually the flow rates of the labeled and unla-
beled gas mixtures. The flow rates of 0.12 mol% 15NH3 (>98%
labeled 15N, Cambridge Isotope Laboratories Inc.) in He (SJ Smith,
99.999%) and 0.12 mol% 14NH3 (99.0%, from Airgas) mixed with
0.12 mol% Ar (Airgas, 99.999%) in He (SJ Smith, 99.999%) were set
by mass flow controllers (MFC, 601 series, Porter), which were
designed for use with corrosive gases. An online quadrupole mass
spectrometer (QMS; Pfeiffer Vacuum, Thermostar) was used to
!
R
ꢀ
ꢁ
tt
t0
I
ðtÞ ꢀ IArðtÞ ꢄ dt
14NH3
1
½
14NH3ꢂꢃss
¼
tNH3
ꢂ
ð1Þ
I
ðt0Þ
MTiO2
14NH3
where [14NH3*]ss is the total number of 14NH3 binding sites (Lewis
acid sites), t0 and tt represent the starting and ending time of the
ammonia exchange (e.g., from 15NH3 to 14NH3), IxðtÞ represents
the intensity of the QMS signal of species x as a function of time,
and tNH3 represents the molar flow rate of NH3. The IAr signal is
monitored and used to account for the effects of diffusional broad-
ening of the gaseous species during flow through the reactor [41].
The NH3 exchange was determined to be complete when the
14NH3 partial pressure changed by less than 0.01% of the original
d
ð
½
14NH3ꢃ=½14NH3
ꢃ
0Þ
< 0:0001ksꢀ1). Example data
value within one ks (
dt
for TiO2, 673 are included in the Supporting Information (Section S1,
Fig. S2). The measured numbers of ammonia adsorption sites on
each TiO2 catalyst are summarized in Fig. 1 and Table 1.

H. Zhang et al. / Journal of Catalysis 361 (2018) 290–302
293
(GC, Agilent, 6850) equipped with a capillary column (DB-WAX, 30
m) and a flame ionization detector (FID).
The retention time for each component was determined using
m ꢅ 0.25 mm ꢅ 0.25
l
1.0
0.8
0.6
0.4
0.2
0.0
1.0
0.8
0.6
0.4
0.2
0.0
known chemical standards (Sigma-Aldrich, C2ꢀC8 n-alcohols and
SSITKA
C2ꢀC4 n-aldehydes) and was confirmed with
a
gas-
chromatograph mass-spectrometer (GC–MS, QP2010 Ultra, Shi-
madzu). Rates were measured at differential conversion (<10%) of
the limiting reactant to identify primary reactions and to ensure
that depletion of reactants across the bed minimally influenced
measured rates. The number of the CAC bond formation events
is calculated by multiplying the moles of formed C2i species (i ꢆ
2) by (i-1). For example, C6, C8, and larger C2i products are formed
by two, three, and (i-1) CAC bond formation events, respectively.
Aldol condensation rates are reported as the number of CAC bond
formation events per second per mole of exposed Ti sites ((mol
Titration
C
ꢆ4)(mol Tis s)ꢀ1), where exposed Ti atoms are determined by
butyric acid titrations, described below. Selectivities are reported
based on moles of carbon contained within each category of prod-
ucts or product species (C-%). Carbon balances closed typically to
90% or better and reflected the condensation or oligomerization
of higher boiling point products (e.g., aromatics and Cꢆ8 alcohols)
within the gas transfer lines.
The effects of isotope substitution on CAC bond formation rates
were measured on TiO2 catalysts which had reached steady-state
rates at given set of conditions (0.3 kPa C2H4O, 1.15 kPa C2H5OH,
15 kPa H2 at 503 K). Then C2H4O (0.3 kPa) was replaced with
C2D4O (Sigma Aldrich, ꢆ99 atom% D, ꢆ98% chemical purity) while
maintaining all other concentrations (0.3 kPa C2D4O, 1.15 kPa
C2H5OH, 15 kPa H2). Subsequently, C2H5OH (1.15 kPa) was further
substituted with C2D5OD (0.3 kPa C2D4O, 1.15 kPa C2D5OD, 15 kPa
H2). Finally, the reaction was returned to the initial conditions (0.3
kPa C2H4O, 1.15 kPa C2H5OH, 15 kPa H2) to correct for deactivation
that occurred during the course of the experiment (<10%).
Reported rates (e.g., Fig. 4) have been corrected for deactivation
using the method described.
BET
500 600 700 800 900 1000 1100
Oxidative Treatment Temperature (K)
Fig. 1. The change in the number of Lewis acid sites (j, determined by SSITKA of
14NH3-15NH3 mixtures), the number of active sites ( , determined from in situ
butyric acid titrations), and BET surface areas
( ) as a function of oxidative
treatment temperature. The dashed lines were included to guide the eye.
2.5. Measurements of catalytic rates and selectivities for acetaldehyde
and ethanol conversion
The rates of aldol condensation catalyzed by TiO2 surfaces were
measured within a tubular glass reactor (12 mm outer diameter)
with a packed-bed configuration. The reactor was placed in a
vertically-aligned temperature-controlled tubular furnace (Applied
Test Systems, 3210). The temperature of the catalyst bed was mea-
sured with a K-type thermocouple (Omega) that was placed in
direct contact with the wall of the reactor immediately adjacent
to the catalyst bed and was controlled by a PID controller (Watlow,
EZ-Zone). The TiO2 catalysts (0.05–0.1 g, mechanically mixed with
silica to achieve a catalyst bed of at least 0.3 g) were treated in situ
by heating at 10 K minꢀ1 to 573 K and holding at 573 K under He
flow (ꢁ100 cm3 minꢀ1) for 2 h to remove adsorbed species before
measuring catalytic rates. Flow rates of He (Airgas, >99.999%)
and H2 (Airgas, >99.999%) were set using mass flow controllers
(Parker Porter, Model 601). Liquid phase reactants (C2H5OH, Decon
Laboratories, 200 proof; acetaldehyde, C2H4O, Fluka, 99.5%; and 2-
butenal, C4H6O, Sigma Aldrich, 99%, predominantly trans) were fed
using two syringe pumps (Legato 110, KD Scientific) and were
vaporized inside the 1/400 stainless steel transfer lines heated to
>373 K using electrical heating tape. Volatile liquid reactants
(e.g., C2H4O) were fed as a 10–20 vol% mixture in C2H5OH. The flow
rates of reactant gas and liquid mixtures were varied to control the
partial pressure and surface residence time of the reactants inde-
pendently during experiments. The total pressure of the system
was maintained at 101 kPa by co-feeding He with the other reac-
tants. Before measuring the change in CAC bond formation rates
as a function of reactant pressures or temperature, the system
was allowed to reach steady-state at a set of common conditions
(0.3 kPa C2H4O, 2.7 kPa C2H5OH and 15 kPa H2 at 503 K), which
took approximately 6 h (Fig. S4). Measured rates were corrected
for deactivation by measuring rates at a given set of conditions
multiple times throughout the course of an experiment and by
using linear interpolation to estimate the number of active sites
present at any given time during the experiment (a detailed expla-
nation is included in the Supporting Information, Fig. S5).
2.6. Estimates for active site counts by butyric acid titration during
aldol condensation
The number of active sites (e.g., Lewis acid base pairs) were
measured on each TiO2 catalyst by using butyric acid (BA, Sigma
Aldrich, >99.0%) as a selective inhibitor during catalysis. All TiO2
catalysts were loaded into the reactor, treated in flowing He
in situ, and allowed to reach steady-state at a set of common con-
ditions (0.3 kPa C2H4O, 2.7 kPa C2H5OH, 15 kPa H2 at 453 K), which
took approximately 6 h (Fig. S4). The in situ site titration was initi-
ated by introducing 60 Pa BA while maintaining the concentration
of all other species (0.3 kPa C2H4O, 2.7 kPa C2H5OH, 15 kPa H2, 453
K) and by simultaneously measuring the cumulative BA uptake and
the change in the reactant conversions as a function of time. The
number of active sites was determined by assuming that the each
adsorbed BA titrated a single Lewis acid base site pair (e.g., in the
form of bidentate butyrate and a proton) and by calculating the
cumulative amount of BA bound to the surface of TiO2 using:
Z
tt
ꢀ
ꢁ
½Lꢃ ¼
rBA ꢀ rBA ꢀ rBA
ꢄ dt
ð2Þ
out
cons
in
t0
where [L] represents the total number of active sites on the sur-
face, t0 and tt represents the times at which the titration began and
finished (which was determined when BA partial pressure from the
effluent changed by less than 0.01% of the original value within one
d
½BAꢃ=½BAꢃ
ð
0Þ
< 0:0001ksꢀ1; [BA] is determined from the FID detec-
ks (
dt
The identity and concentrations of reactants and products in the
reactor effluent stream were measured using a gas-chromatograph
tor by integrating the peak area of BA and multiplied by a predeter-
mined sensitivity factor), and rBA , rBA , and rBA represent the
cons
out
in

294
H. Zhang et al. / Journal of Catalysis 361 (2018) 290–302
molar rate at which BA enters the reactor, reacts to form gaseous
products (e.g., ethyl butyrate, as determined by GC), and leaves
(i.e., 2-butenal, 2-buten-1-ol, butanol, C6-C8 alcohols and aromat-
ics) is greater. The increased selectivity towards aldol addition
pathways likely reflects changes in barriers for the elementary
steps involved (compared to those for esterification) or changes
in the relative coverages of the intermediates involved in the two
pathways as a result of increasing the oxidative heat treatment
the reactor outlet. An example of BA titration over TiO2,
is
873
included in Fig. S3. These titrations are used to report aldol con-
densation rates as the number of C ꢀ C bond formation events
per second per mole of Ti sites exposed ((mol Cꢆ4)(mol Tis s)ꢀ1).
temperature. Second, the C products are dominated by C6-C8
ꢆ4
alcohols and aldehydes and C8 aromatics, all of which must form
by sequential aldol addition reactions [20,42]. The prevalence of
these species, along with only trace quantities of 2-butenal, show
that aldehydes and enals formed in situ undergo additional aldol
addition steps with high rates. Third, the turnover rates for C2 con-
version determined over TiO2, 673 (6.0 (mmol C2 consumed) (mole
Tis s)ꢀ1) is ꢁ15 times higher than that on TiO2, 873 (0.4 (mmol C2
consumed) (mole Tis s)ꢀ1) at these conditions. The cumulative dif-
ferences shown in Table 2 strongly suggest that the surface chem-
istry of TiO2 changes considerably as a result of differences in
oxidative treatment temperatures (673–873 K) that do not change
the crystallographic phase (Table 1) and that decrease the specific
surface and surface density of active sites only moderately (Fig. 1),
as discussed below.
3. Results and discussions
3.1. Effects of thermal treatment on product Selectivities
Table 2 shows that reactions within mixtures of C2H4O and
C2H5OH over TiO2, 673 generate a broad distribution of species that
includes mostly C4 alcohols and aldehydes (ꢁ25C-%; e.g., 2-
butenal, 2-buten-1-ol, and butanol), C6ꢀC8 alcohols and aldehydes
(ꢁ11C-%),
C8
aromatics
(ꢁ25C-%;
e.g.,
2-
and
4-
methylbenzaldehydes and alcohols), and a small percentage of
unknown products (< 5 C-%, such as C10+ products, unknown enals,
branched products and some alkenes) at ꢁ10% conversion of C2H4O
(0.3 kPa C2H4O, 2.7 kPa C2H5OH, 15 kPa H2, 503 K). The amount of
butanal formed was immeasurable, which suggests that this spe-
cies is rapidly consumed to form C products by additional aldol
ꢆ6
3.2. Effect of thermal treatments on catalyst structure and site counts
condensation or to form 1-butanol by the Meerwein-Ponndorf-
Verley (MPV) reaction with ethanol [20]. Esterification produces
not only ethyl acetate (ꢁ37% C-%) by a primary reaction pathway
between the alcohol and the aldehyde [25] but also d-
hexalactone (ꢁ21C-%), which likely forms by intramolecular ester-
ification of a C6 aldol addition product. Secondary reactions
become more significant as acetaldehyde conversions increase
from 10 to 75% (Fig. S6), and ethyl acetate and C4 alcohols and alde-
hydes undergo cross-coupling reactions to form increasing
amounts of C6-C8 alcohols, aldehydes, aromatics and lactones.
Table 2 shows also that the distribution of products formed dif-
fers significantly between TiO2, 873 and TiO2, 673. First, the percent-
age of products formed by esterification pathways involving
nucleophilic attack from alkoxides to aldehydes [25] (e.g., ethyl
acetate and d-hexalactone) on TiO2, 873 (17C-%) are lower than on
The synthesized TiO2 were treated to progressively higher tem-
peratures (523–1073 K) in air to determine how subtle differences
in the thermal treatments may influence bulk and surface proper-
ties. The XRD patterns for the TiO2 catalysts (Fig. S1) treated at 673
K, 773 K, and 873 K show reflections only consistent with the ana-
tase phase, which is consistent with previously determined struc-
tures of unsupported TiO2 or dispersed TiO2 nanoclusters on
supports thermally treated to temperatures equal to or less than
873 K [38,43]. Although the crystallographic phase does not
change with increasing oxidative treatment temperatures from
673 K to 873 K, Table 1 shows that the mean diameters of the ana-
tase crystallites increase and the surface areas (determined by N2
physisorption) decrease. These structural properties may influence
catalysis, but more direct methods are needed to probe the number
and structure of the active sites for the aldol condensation reaction.
The catalytic behavior of TiO2, 673 differs significantly from that
of TiO2, 873 (Table 2), and the reasons for these differences may be
related to changes in the surface density of Lewis acid sites (which
are thought to participate in the reaction) or active sites (which we
propose also include a Lewis acid site; Section 3.5). Fig. 1 shows
that the number of accessible Lewis acid sites (Ti4+, estimated by
SSITKA with 14NH3-15NH3 mixtures) decreases by nearly an order
of magnitude (i.e., from 0.9 mmol (g of TiO2)ꢀ1 to ꢁ0.1 mmol (g
of TiO2)ꢀ1) as the temperature of the oxidative treatment increases
from 523 K to 873 K. The number of catalytically active sites on the
surface, however, decreases only by a factor of two (i.e., from ꢁ0.2
mmol (g of TiO2)ꢀ1 over TiO2, 523 to ꢁ0.08 mmol (g of TiO2)ꢀ1 over
TiO2, 873), as determined by in situ titrations with butyric acid. The
BET surface areas for TiO2 samples treated over the range 673–873
K change in ways more similar to the number of active sites (Fig. 1)
than the estimates for exposed Ti4+ species from SSITKA. Compar-
isons among these methods (Fig. 1) clearly shows that only a frac-
tion of the exposed Ti cations are catalytically active and that
within the range of temperatures often used for oxidative treat-
ments (673–873 K), the number of active sites changes by a rela-
tively small amount (ꢁtwo-fold) [19,25,28,31].
TiO2,
(ꢁ58C-%) and the fraction of aldol addition products
673
Table 2
Products formed with a mixture of C2H4O and C2H5OH at 503 K (0.3 kPa C2H4O, 2.7
kPa C2H5OH, 15 kPa H2).
Catalyst
TiO2,
6.0
TiO2,
0.4
673
873
Turnover rates (mmol C2 consumed) (mol Tis s)ꢀ1a
C2 conversion (%)b
Acetaldehyde conversion (%)
Carbon balance (%)c
Selectivity among Cꢆ4 products (%)
Ethyl acetate
2-Butenal
2-Buten-1-ol
Butanol
d-Hexalactone
Hexanal
4.5ꢇ0.2
10.2ꢇ0.4
90
4.0ꢇ 0.1
9.8 ꢇ0.2
89
30ꢇ1
16 ꢇ1
trd
17ꢇ 1
1.0ꢇ 0.1
2:4ꢇ0.1
17ꢇ0.5
trd
7.9ꢇ 0.3
trd
1.6 ꢇ0.1
10 ꢇ0.1
5.2ꢇ0.3
40ꢇ2
16 ꢇ1
<4.5
Hexanol
3.3ꢇ0.1
20ꢇ 1
5.3ꢇ0.2
<4.0
C8 aromaticse
Octanol
Othersf
a
Determined using summed molar flow rates of ethanol and acetaldehyde.
Measured by considering the net decrease in the partial pressure of ethanol and
b
acetaldehyde.
c
Systematically low values for the carbon balance reflect condensation of C8 and
3.3. Effects of reactant pressures on initial CAC bond formation rates
larger products in gas transfer lines and uncertainty in GC sensitivity factors for
these species.
Fig. 2 shows the changes in summed rates of formation of Cꢆ4
d
Trace amounts detected.
C8 aromatics include 2- and 4-methylbenzaldehydes and alcohols.
Likely include C6+ branched alcohols and aldehydes.
e
molecules by aldol addition on TiO2,
and TiO2,
that result
673
873
f
from independent changes in the pressures of C2H4O, C2H5OH,

H. Zhang et al. / Journal of Catalysis 361 (2018) 290–302
295
Fig. 2. Aldol addition rates over (a) TiO2, 673 as a function of C2H4O partial pressure at 503 K (j) and 523 K (d) (2.7 kPa C2H5OH, 15 kPa H2) and as a function of C2H5OH partial
pressure at 503 K ( ), 523 K ( ) and 537 K ( ) (0.3 kPa C2H4O, 15 kPa H2); (b) TiO2, 673 as a function of H2O partial pressure at 503 K (j) and 523 K (d) (0.3 kPa C2H4O, 2.7 kPa
C2H5OH, 15 kPa H2) and as a function of H2 at 503 K ( ) (0.3 kPa C2H4O, 2.7 kPa C2H5OH); (c) TiO2, 873 as a function of C2H4O partial pressure at 503 K (j), 523 K (d) and 537 K
(▲) (2.7 kPa C2H5OH, 15 kPa H2) and as a function of C2H5OH partial pressure at 503 K ( ), 523 K ( ) and 537 K ( ) (0.3 kPa C2H4O, 15 kPa H2). The dashed lines are included
to guide the eye.
H2, and H2O. On surfaces of TiO2, 673 at 523 K, aldol addition rates
increase with a nearly second order dependence on C2H4O pressure
(r ꢁ [C2H4O]1.9) over the full range of C2H4O pressures tested (0.1–
0.8 kPa) at 523 K (Fig. 2a), while at a lower temperature (503 K),
rates exhibit a second order dependence on C2H4O pressure at rel-
atively low pressures (<0.55 kPa C2H4O) and become constant
when C2H4O pressures are increased further (2.7 kPa C2H5OH, 15
kPa H2). Additionally, rates decrease with an inverse second order
dependence on C2H5OH at 523 K and 537 K (0.3 kPa C2H4O, 15 kPa
H2), which decreases to a negative first order dependence on
C2H5OH pressure at 503 K. Fig. 2b shows that aldol addition rates
The influence of reactant pressures on aldol addition rates on
TiO2, 873 (Fig. 2c and Fig. S7) qualitatively resemble the functional
dependencies observed on TiO2, 673, which implies that the differ-
ence in the oxidative treatment temperature does not change the
mechanism of the reaction. Fig. 2c shows that aldol addition rates
on TiO2, 873 increase as the square of the C2H4O pressure at the low-
est pressures (0.05–0.13 kPa C2H4O), transition into a region that
resembles a first order dependence (0.13–0.3 kPa C2H4O), and
gradually approach a sub-linear dependence on C2H4O at higher
pressures (0.3–0.7 kPa C2H4O) at otherwise constant conditions
(2.7 kPa C2H5OH, 15 kPa H2, 503 K). Aldol addition rates at 523 K
and 537 K show a similar dependence on C2H4O pressure over
the range of 0.2 kPa to 0.7 kPa C2H4O. In contrast, rates depend
inversely on C2H5OH pressure and exhibit a nearly negative first
order dependence at 537 K and a weaker dependence (r ꢁ [C2H5-
on TiO2,
are inversely proportional to the partial pressure of
673
H2O (i.e., r ꢁ [H2O]ꢀ1) and depend only weakly on the partial pres-
sure of H2 (i.e., r ꢁ [H2]ꢀ0.2) at similar temperatures (0.3 kPa C2H4O,
2.7 kPa C2H5OH).

296
H. Zhang et al. / Journal of Catalysis 361 (2018) 290–302
OH]ꢀ0.3 to ꢀ0.7) as temperatures decrease to 523 K and 503 K (0.3
kPa C2H4O, 15 kPa H2). Notably, the oxidative treatment tempera-
ture of TiO2, 873 matches that of TiO2 samples used by Young et al.
[19] and is more similar to those used by Wang et al. (Degussa, P25,
face species present during steady-state aldol addition over TiO2,
established before and after switching to C2D4O and C2D5OD
673
mixtures from the reactant feed as a function of time at 503 K.
The rapid appearance of the
m(OAD) feature corresponds with a
25% rutile-75% anatase) [25,28] than the preparation of TiO2,
.
commensurate decrease in (OAH) (not shown) that reflects rup-
m
673
Finally, CAC bond formation rates on TiO2,
decrease with H2
ture and reformation bonds between lattice O-atoms and H/D-
atoms abstracted from the reactants prior to the kinetically rele-
vant step for aldol condensation. The inset of Fig. 3a shows the
873
to
(r ꢁ [H2]ꢀ0.3) and H2O (r ꢁ [H2O]ꢀ1/2
ꢀ1) pressures (Fig. S7) in
ways that are analogous to the trends observed on TiO2, 673
.
Overall, the manner in which the aldol addition rates on these
TiO2 samples depend on the pressures of C2H4O, C2H5OH, H2O,
and H2 (Fig. 2 and Fig. S7) can be summarized as:
ratio of the intensities for the
m
as(CD3) peak (2231 cmꢀ1) [49] to
that for the
m
as(CH3) feature (2981 cmꢀ1) as a function of time,
and this comparison suggests that ꢁ4% of the ethanol-derived sur-
face intermediates are reactive. In other words, more than 95% of
the surface of TiO2, 673 is covered by unreactive aliphatic hydrocar-
bons, alcohols, or alkoxides that do not turnover or desorb on the
timescale of this experiment (0.6 h) [19,20,44,45].
0ꢀ2
½C2H4Oꢃ
rCꢀC ꢁ kapp
ð3Þ
1ꢀ2
0ꢀ1
0:2ꢀ0:3
½C2H5OHꢃ ꢂ½H2Oꢃ ꢂ½H2ꢃ
where kapprepresents the apparent reaction constant for CAC bond
formation and [X] represents the pressure of species X in the reac-
tor. Notably in the limit of the smallest ratios of C2H4O to C2H5OH
pressures, the form of Eq. (3) suggests that two C2H4O-derived
intermediates participate in the kinetically relevant step on active
sites that are mostly covered by C2H5OH-derived species. With
increasing ratios of C2H4O to C2H5OH pressures, the sensitivity of
rates on the pressures of these reactants decreases in a manner con-
sistent with competitive adsorption processes and an increasing
coverage of C2H4O-derived surface intermediates. In the following
section, comparisons of FTIR spectra taken in situ provide direct evi-
dence that support this interpretation of the rate dependencies and
that disprove other mechanistic interpretations.
The similarities between the spectrum of the species at the
steady-state and the spectrum of the pure ethanol included in
Figs. S11 and S15 as well as the presence of oxygen within all reac-
tion products (Table 2) suggests that the majority of the exposed Ti
cations are covered by strongly bound, unreactive ethanol or
ethoxide during catalysis. The peaks of the spectra of the species
at the steady-state (Fig. 3a) agree with those for adsorbed ethanol
(or ethoxide) and include features at 2981, 2936, 2905, and 2877
cmꢀ1 corresponding to
modes [19,20,44,45] and at 1486, 1449, 1396 and 1373 cmꢀ1
attributed to (CH2), das(CH3), ds(CH3), and ds(CH2) modes
mas(CH3), mas(CH2), ms(CH3), and ms(CH2)
c
[20,27,48]. Together these observations (Fig. 3a) suggest that etha-
nol (or ethoxide) exists as the MARI on exposed cationic Ti sites
and that the coverage of adsorbed aldehydes (e.g., C2H4O which
3.4. Comparisons of the relative coverages of reactive species by in situ
FTIR
has an intense
Fig. 3b and the accompanying inset shows that the area of the
vibrational feature corresponding to the carbonyl of C2H4O (i.e.,
(C@O), 1720 cmꢀ1) [46,47] increases with respect to those features
for methyl groups (i.e., (CH3)) as the ratio of C2H4O to C2H5OH in
m
(C@O) feature at 1720 cmꢀ1 Fig. S11) is negligible.
m
The identity and the relative coverages of surface intermediates
were probed by in situ FTIR in order to test the proposal that etha-
nol (or ethanol-derived intermediates, such as ethoxide) exist at
greater coverages than other active surface species within the
range of conditions similar to those used during catalysis. These
experiments were used also to compare the prevalent surface
intermediates between TiO2, 673 and TiO2, 873. Fig. 3a shows vibra-
tional spectra obtained in situ during aldol addition over TiO2, 673 at
conditions that result in a clear second order dependence on the
C2H4O pressure and a negative second order dependence on etha-
nol pressure (Fig. 2a, 0.09 kPa C2H4O, 2.7 kPa C2H5OH, 15 kPa H2,
503 K). Spectra obtained at steady-state contain vibrational fea-
tures between 2800 and 3000 cmꢀ1 and 1300–1500 cmꢀ1 that cor-
m
the reactant feed increases during aldol addition reactions on both
TiO2, 673 and TiO2, 873 (0.1–0.9 kPa C2H4O, 2.0 kPa C2H5OH, 443 K;
complete spectra from 1660 to 3100 cmꢀ1 are shown in Fig. S9).
Notably, the intensity of the
m
(C@O) feature is significantly smaller
(C@O) fea-
than that for the as(CH3) peak from C2H5OH, and the
m
m
ture at ꢁ1720 cmꢀ1 is nearly 10-fold more intense than any other
vibrational mode for adsorbed C2H4O (Fig. S11). Therefore, the ratio
of the integrals for the peak at 1720 cmꢀ1 to that at 2980 cmꢀ1
(I1720/I2980) for a given spectrum provides is proportional to the
ratio of the coverage of C2H4O to C2H5OH-derived surface species.
Fig. 3b shows that the ratio I1720/I2980 is less than 0.5 at all condi-
tions (and often much lower on TiO2, 673. We estimate that the
molar extinction coefficient for C2H4O is nearly 2.5 times greater
for C2H4O than for C2H5OH in this comparison (see Fig. S11 and dis-
cussion in SI), and thus, the total number of C2H5OH-derived sur-
face intermediates appear to exist at coverages greater than that
of C2H4O on TiO2, 673 and TiO2, 873 catalysts (and often more than
10-fold greater) over this range of conditions. Detailed information
on deconvolution within the regions between 2800 cmꢀ1 and
3200 cmꢀ1 were included in Fig. S10 and Table S1.
respond to
m(CAH) and d(CAH) modes, respectively, of
hydrocarbons and oxygenates [19,20,44,45]. The lack of easily
observed features between 1600 and 1800 cmꢀ1 (i.e., frequencies
consistent with
m(C@O) modes) indicates that C2H4O and other
aldehydes exist at negligible coverages at these conditions.
[19,45–47]. The spectra of the reactive surface intermediates and
those of the unreactive, persistent surface species were differenti-
ated by replacing the mixture of C2H4O and C2H5OH (0.09 kPa
C2H4O, 2.7 kPa C2H5OH, 15 kPa H2) with perdeuterated reactants
(0.09 kPa C2D4O, 2.7 kPa C2D5OD, 15 kPa H2) while continuously
acquiring FTIR spectra (Fig. 3a, red spectra and inset). The intensi-
ties of absorption features present in the steady-state spectra (i.e.,
peaks at between CAH stretching modes from 2800 to 3000 cmꢀ1
and bending modes 1300–1500 cmꢀ1) decrease by less than 5%
over a period of 0.6 h, which suggests that most of the surface
intermediates present on TiO2 are unreactive. Shortly after the
switch to perdeuterated reactants, several new vibrational features
emerge between 2500 and 2800 cmꢀ1 and 1800–2300 cmꢀ1, which
Fig. 3b shows that values of I1720/I2980 for TiO2, 673 increase with
the molar ratio of C2H4O to C2H5OH in the gas-phase, whereas,
I1720/I2980 for TiO2, 873 increases only slightly over the same range.
Moreover, the values of I1720/I2980 on TiO2, are much smaller
673
than on TiO2, 873 at all conditions. The results in Fig. 3a and trends
in Fig. 3b suggest that the MARI on TiO2, 673 is primarily C2H5OH,
but that C2H4O and C2H5OH exist at more comparable coverages
on TiO2, 873 at equivalent conditions. These results are consistent
also with the weaker dependence of aldol addition rates on
C2H4O and C2H5OH pressures over TiO2, 873 (i.e., r ꢁ [C2H4O]1ꢄ [C2H5-
are attributed to
m(OAD) of surface hydroxyls and stretching
modes of ACD3 and ACD2 groups, respectively [49]. Fig. S8 shows
OH]ꢀ1) in comparison TiO2, 673 (r ꢁ [C2H4O]2[C2H5OH]ꢀ2) at compa-
ꢄ
the complete spectra over the range 1330–3100 cmꢀ1 on the sur-
rable reactant pressures, and the comparisons of rate measurements

H. Zhang et al. / Journal of Catalysis 361 (2018) 290–302
297
(a)
0.038
0.04
0.02
0.00
I2231 cm-1
I2981 cm-1
0
20
40
60
80
Time (min)
species at steady-state
(C-H)
labeled reactive species
(C-D)
(C-H)
(O-D)
2600
3000
2800
2400
2200
2000
1800
1600
1400
Wavenumber (cm-1)
(b)
0.1
TiO2, 873
.
I
-1
A1720: A2980
1720 cm
0.01
I2980 cm-1
TiO2, 673
.
0.001
0.1
1
(C=O)
~1720 cm-1
[C2H4O]: [C2H5OH]
TiO2, 873
TiO2, 873
TiO2, 873
TiO2, 673
TiO2, 673
TiO2, 673
x 10
x 10
3000
2900
2800
2700
2000
1900
1800
1700
Wavenumber (cm-1)
Fig. 3. In situ FTIR measurements that include (a) the spectrum of all surface species present during steady-state aldol addition over TiO2, 673
(
) established after 4 h (0.09
kPa C2H4O, 2.7 kPa C2H5OH, 15 kPa H2, 503 K) and the spectrum of surface species present after switching to a mixture of perdeuterated reactants for 0.6 h (
C2D4O, 2.7 kPa C2D5OD, 15 kPa H2, 503 K). The inset shows the ratio of the intensities of the features at 2231 cmꢀ1 as(CD3)) to those at 2981 cmꢀ1
as(CH3)) . (b) FTIR spectra
of species that form at ratios of the partial pressures of C2H4O to C2H5OH of 0.06 ( ), 0.12 ( ), and 0.45 ( ), on TiO2, 673 as well as 0.06 ( ), 0.12 ( ), and 0.45 (
on TiO2, 873 at 443 K. The inset displays magnified features of 0.06 ( ) and 0.12 (
) multiplied by a factor of 10 within the region between 1780 cmꢀ1 and 1668 cmꢀ1 on
TiO2, 673. Ratios of the areas of the peak at 1720 cmꢀ1 (corresponding to (C@O) of C2H4O) to that at 2980 cmꢀ1 (corresponding to
as(CH3) from ethanol, acetaldehyde and
) (0.09 kPa
(
m
(m
)
v
v
polymerized organics on the surface) for TiO2, 673 (s) and TiO2, 873 (j) as a function of the ratio of the partial pressures of C2H4O to C2H5OH at 443 K was inserted (0.1–0.9 kPa
C2H4O, 2.0 kPa C2H5OH, 15 kPa H2, 443 K.). Dashed lines are included to guide the eye.
and FTIR spectra suggest that greater coverages of C2H4O on TiO2, 873
contribute to the reduced apparent dependence on the pressure of
the reactants (Fig. 2a and 2c). Quantitative arguments that support
these claims can be provided from a rate expression derived from
a system of elementary steps for aldol addition as shown next.
those discussed for TiO2 [19] and proposed previously for physical
mixtures of Cu/SiO2 and TiO2 catalysts [25].
Scheme 1 includes steps for molecular adsorption of C2H4O at
Lewis acid sites (step 1.1), and deprotonation of C2H4O by a neigh-
boring base site to form the enolate (step 1.2). A CAC bond then
forms when the enolate ((CH2dꢀ)CHO*A) nucleophilically attacks
C2H4O adsorbed at an adjacent acid site (CH3CHO*A) (step 1.3).
Subsequent protonation of the C4 complex creates the aldol pro-
duct (step 1.4). The aldol product undergoes dehydration to create
2-butenal (step 1.5), then the C@O group of 2-butenal is hydro-
genated by hydrogen transfer from ethanol (e.g., via the
Meerwein-Ponndorf-Verley) mechanism (step 1.6) to create 2-
buten-1-ol, and C@C bonds are hydrogenated by reaction with
H*B (step 1.7) to form n-butanol. Desorption of 2-butenal (step
1.8), 2-buten-1-ol (step 1.9), n-butanol (step 1.10), and water (step
3.5. Mechanism for CAC bond formation by aldol addition
3.5.1. Elementary steps and predominant surface intermediates
Scheme 1 shows a sequence of elementary steps that describe
the kinetic effects for C2H4O, C2H5OH, H2O and H2. This sequence
and the resulting rate equation are consistent with the measured
dependence of aldol rates on reactant concentrations reported
here. The set of elementary steps proposed here are similar to

298
H. Zhang et al. / Journal of Catalysis 361 (2018) 290–302
Scheme 1. Proposed sequence of steps for aldol condensation among acetaldehyde and acetaldehyde-derived species in the presence of ethanol on TiO2 catalysts. Note: *A
and *B denote unoccupied Lewis acid and base sites, respectively; X and X* represents a gas-phase species and the corresponding bound surface intermediate, respectively;
indicates that the elementary step is quasi-equilibrated;
indicates that the elementary step is kinetically relevant and irreversible.
1.11) form the gaseous products and regenerate empty Lewis acid
sites. Simultaneously, C2H5OH adsorbs molecularly on Lewis acid
sites (*A) (step 1.12) and reacts with base sites (*B) to form ethoxide
and hydrogen atoms (H*B) (step 1.13). Gaseous H2 forms by recom-
binative desorption of H*B from adjacent sites (step 1.14). In the
following sections (and the corresponding SI), we compare rate
expressions and the associated assumptions to the measured
kinetic and spectroscopic observations.
observed dependence of aldol addition rates on reactant pressures
(Fig. 2), the prevalent reactive surface intermediates (Fig. 3, i.e.,
primarily C2H5OH with significant coverages of C2H4O only at
the largest C2H4O-C2H5OH ratios on TiO2, 873), and the effects of
isotope substitution (Fig. 4, vide infra) on both TiO2,
and
673
TiO2,
are consistent only with the kinetically relevant nucle-
873
ophilic attack of the enolate to acetaldehyde, despite the fact that
the product distributions and aldol addition rates differ signifi-
cantly between these two materials (Table 2).
When nucleophilic attack from the enolate to the molecularly
adsorbed acetaldehyde (step 1.3) is assumed to be the sole kineti-
cally relevant step and adsorption and deprotonation of acetalde-
hyde and ethanol are quasi-equilibrated (i.e., steps 1.1–1.2 and
steps 1.12–1.13), Scheme 1 shows that the rate of nucleophilic
attack (i.e., r3) depends on the number of molecularly adsorbed
C2H4O ([CH3CHO*A]) and the coverage of the enolate ([(CHd2ꢀ)CHO
*A]) as follows:
3.5.2. Derived rate expressions for aldol condensation
The set of elementary steps proposed in Scheme 1 and previ-
ously published results and proposals regarding aldol condensa-
tion reactions on TiO2 [5,19,25,28] were considered to develop
(and test) six potential rate expressions for aldol condensation
on these TiO2 catalysts (details of the derivations are shown in
section S5 of the SI). In most of the cases, the molecular adsorp-
tion and desorption of reactant and product molecules (i.e.,
acetaldehyde, ethanol, 2-butenal, 2-buten-1-ol, water) are
assumed to be quasi-equilibrated except for the situation that
the adsorption of acetaldehyde (section S5.2) or the desorption
of 2-butenal (section S5.5) was assumed to be kinetically rele-
vant. The six mechanisms vary in the identity of the kinetically
relevant step(s), which are either the nucleophilic attack to form
the CAC bond (step 1.3), the adsorption of C2H4O (step 1.1), the
deprotonation of C2H4O to form the enolate (step 1.2) [25,28],
the protonation of the C4 intermediate to create the aldol product
(step 1.4) [25,28], the dehydration of the aldol molecule to create
2-butenal (step 1.5), or the desorption of 2-butenal (step 1.8)
[19]. Comparisons of the resulting rate expressions to the
k3½CH3CHOꢂAꢃ½ðCHdꢀÞCHOꢂAꢃ
2
r3
¼
ð4Þ
½LAꢃ
where k3 is the rate constant for step 1.3, and [LA] is the
total number of Lewis acid sites on the surface of the catalyst.
The terms for [CH3CHO*A] and [(CHd2-)CHO*A] can be restated as
functions of the gas-phase pressures of the reactants due to
their quasi-equilibrated formation, which leads to the rate
equation:
2
2
k3K2K2½CH3CHOꢃ ½ꢂAꢃ ½ꢂBꢃ
1
r3
¼
ð5Þ
½LAꢃ½HꢂBꢃ

H. Zhang et al. / Journal of Catalysis 361 (2018) 290–302
299
where Ky is the equilibrium constant for step y in Scheme 1. The
site balance for the Lewis acid sites contains the following species:
reasons described due to the assumption of quasi-equilibrium for
H2 desorption. More importantly, restating Eq. (8) in simplified
forms that reflect the different combinations of surface intermedi-
ates observed by in situ FTIR (Fig. 3) reproduces the manner in
which aldol condensation rates depend strongly on both C2H4O
and C2H5OH pressures at the smallest C2H4O to C2H5OH reactant
ratios and become less sensitive to these pressures as the ratio of
the species moves towards unity. For example, rates are propor-
tional to [C2H4O]2 and [C2H5OH]ꢀ2 at the smallest C2H4O to
C2H5OH ratios (Fig. 2a) on TiO2, 673, which lead to predominantly
C2H5OH*-covered (i.e., C2H5OH*A as MARI) reactive Lewis acid sites
(Fig. 3a). In comparison, rates on TiO2, 873 depend on [C2H4O]1 and
[C2H5OH]ꢀ1 when C2H5OH pressures are an order of magnitude
greater than the C2H4O pressure (Fig. 2c), which also results in
more comparable coverages of C2H4O*A and C2H5OH*A reactive
intermediates (Fig. 3b). Both H2O and H2 inhibit rates, because
water competes for Lewis acid adsorption sites (as shown for
HAP catalysts)20 and the dissociative adsorption of co-fed H2
increases the coverage of H*-atoms at the expense of unoccupied
base sites, and subsequently decreases the coverage of the enolate
(via reverse of step 1.2). Overall, the assumption that the nucle-
ophilic attack of surface enolates to acetaldehyde are kinetically
relevant for aldol condensation on TiO2 is consistent with both
the experimentally observed dependencies of rates on reactant
pressures (Figs. 2 and S7) and the coverages of surface intermedi-
ates observed by in situ FTIR (Fig. 3).
½LAꢃ ¼ ½ꢂAꢃ þ ½CH3CH2OHꢂAꢃ þ ½CH3CH2OꢂAꢃ þ ½CH3CHOꢂAꢃ
h
i
þ ðCHd2ꢀÞCHOꢂA þ ½H2OꢂAꢃ
ð6Þ
where the coverage of the aldol product ([C4H8O2*]) and other
4+ intermediates are assumed to be insignificant at the differential
C
conversions used here. Combining the site balance (Eq. (6) and the
rate expression (Eq. (5) and restating surface concentrations, with
the exception of [H*B], in terms of the corresponding gas-phase
species and equilibrium constants (see section S5 in the SI for
the complete derivation) leads to an expression for the rate of aldol
condensation (rC-C):
2
2
rCꢀC
½LAꢃ
k3K1K2½CH3CHOꢃ ½ꢂBꢃ
¼
ꢃ þ K ½CH3CHOꢃ þ OꢃÞ2 ꢄ ½HꢂBꢃ
K12 K13 ½CH3 CH2OHꢃ½ꢂB
K1 K2 ½CH3 CHOꢃ½ꢂB
½H2
ð1 þ K12½CH3CH2OHꢃ þ
ꢃ þ
½HꢂB
ꢃ
½HꢂB
ꢃ
K11
1
ð7Þ
Eq. (7) can be compared to the apparent dependence of aldol
condensation rates on the pressures of reactants (i.e., C2H5OH
and C2H4O) and co-products (e.g., H2O), however, this expression
does not explicitly capture the effect of H2 pressure on rates. Aldol
addition rates that decrease slightly as H2 pressures increase
(Fig. 2b and S7, r ꢁ [H2]ꢀ0.2 to ꢀ0.3) suggest that the recombinative
desorption of H2 from the surface of TiO2 (step 1.14) is reversible
but is not quasi-equilibrated. This behavior can be described by
considering the approach to equilibrium (g) for this process, which
3.5.3. Effects of H/D-Atom substitution on rates of aldol condensation
Aldol condensation rates were measured using a series of three
distinct combinations of H- and D-labeled acetaldehyde (i.e.,
C2H4O, C2D4O) and ethanol (i.e., C2H5OH, C2D5OD) to determine
how isotope substitution changes reaction rates and to test further
the kinetic relevance of elementary steps that involve acetalde-
hyde deprotonation, reactant adsorption or product desorption,
dehydration of the aldol molecule to generate 2-butenal, and
nucleophilic attack of the enolate. Fig. 4 shows that steady-state
aldol condensation rates on TiO2, 673 and TiO2, 873 are 3.6 and 0.2
(mmol Cꢆ4) (mol Tis s)ꢀ1, respectively, when perhydrogenated
combinations of C2H4O and C2H5OH are used as reactants (0.3
kPa C2H4O, 1.15 kPa C2H5OH, 15 kPa H2, 503 K). Aldol condensation
rates increase by a factor of 1.6–1.8 when C2H4O is replaced by
perdeuterated acetaldehyde (C2D4O) in the reactant stream (i.e.,
kH/kD ꢁ 0.6) and subsequently show a negligible change when
C2H5OH is replaced by perdeuterated ethanol (C2D5OD) (i.e., kH/
kD ꢁ 1) while holding all other parameters constant. These results
differ from the kinetic isotope effects reported previously for reac-
tion mixtures containing mixtures of aldehyde and alcohol reac-
tants (e.g., kH/kD = 2.4) [25] and for reactions of pure
acetaldehyde (i.e., kH/kD = 0.98) [19], albeit under somewhat differ-
ent reaction conditions.
leads to an expression for [H*B] as a function of H2 pressure (i.e.,
ꢂ
ꢃ
k
ꢀ14ꢄ½H2ꢃꢄ½ꢂBꢃ1=2ꢀð1ꢀ
gÞ
½HꢂBꢃ ¼
1=2), that may be substituted into Eq.
k14
(7). Rates of aldol condensation will become most sensitive to
the pressure of H2 in the limit that the recombinative desorption
of H*-atoms approaches quasi-equilibrium (i.e.,
g = 1, a condition
that overestimates the rate of dissociative H2 adsorption), which
will lead to the following rate expression for aldol addition:
2
rCꢀC
½LAꢃ
k3K2K2K1=2½CH3CHOꢃ
1
14
¼
K12 K13 ½CH3 CH2 OHꢃK1=2
K1 K2 ½CH3 CHOꢃK1=2
2
1=2
½H2
ð1þK12½CH3CH2OHꢃþ
þ
þK ½CH3CHOꢃþ OꢃÞ ꢄ½H2ꢃ
14
14
½H2 ꢃ1=2
½H2 ꢃ1=2
K11
1
ð8Þ
The form of Eq. (8) accounts for the apparent dependencies on
the pressures of C2H5OH, C2H4O, and H2O (Fig. 2 and S7, summa-
rized in the power law expression shown in Eq. (3), while some-
what overestimating the influence of the H2 pressure for the
6
The values of kH/kD calculated from the data in Fig. 4 are incon-
sistent with the kinetic relevance of most processes within the
aldol condensation reaction network (Scheme 1) except for the
nucleophilic attack of the enolate to acetaldehyde [51–54]. First,
these results strongly suggest that neither deprotonation of
acetaldehyde nor aldol dehydration (by kinetically relevant b C-H
bond rupture as demonstrated by comparisons of dehydration
rates of C2H5OD and C2D5OD, Fig. S17, and prior reports) [50] are
kinetically relevant, because substitution of C2H4O for C2D4O
would likely decrease aldol condensation rates (i.e., kH/kD > 1)
due to primary kinetic isotope effects caused by differences
between the zero point energies of the involved CAD and CAH
bonds. Second, rates of aldol condensation that differ between
reactant mixtures with different H/D-atom labels shows that
molecular adsorption steps or desorption steps are not kinetically
relevant when reactant streams contain both aldehydes and
TiO2, 673
4
2
0
X
14
TiO2, 873
C2H4O + C2H5OH
C2D4O + C2D4O +
C2H5OH C2D5OD
C2H4O + C2H5OH
100
200
300
400
500
600
700
Time (min)
Fig. 4. Aldol condensation rates as a function of time on stream over TiO2,
(h)
673
and TiO2, 873
( , multiplied by a factor of 14 for clarity) with changes in the isotope
labelling of the reactant mixtures of acetaldehyde and ethanol (0.3 kPa C2H4O-
C2D4O, 1.15 kPa C2H5OH-C2D5OD, 15 kPa H2, 503 K). The horizontal dashed lines are
included to guide the eye.

300
H. Zhang et al. / Journal of Catalysis 361 (2018) 290–302
alcohols, because molecular adsorption and desorption would not
involve the activation of CAH/D or OAH/D bonds and are unlikely
to exhibit secondary kinetic isotope effects. Notably, rates of aldol
condensation are indistinguishable for pure reactant streams of
C2H4O or C2D4O at 553 K, as reported previously [19] and as repro-
duced here (Fig. S18), which may indicate that the kinetically rel-
evant step differs in the absence of ethanol. Third, aldol
condensation rates that do not depend on the isotopic label of
ethanol (i.e., C2H5OH or C2D5OD) indicate that protonation of the
C4-intermediate to form the aldol product (step 1.4, Scheme 1) is
not kinetically relevant, because the dissociation of ethanol to form
ethoxide provides a fraction of the H/D-atoms on TiO2 (via step
1.13, Scheme 1) and thus protonation rates would be expected to
show a normal, primary kinetic isotope effect.
The combination of kinetic isotope effects shown in Fig. 4 are
consistent only with a mechanism in which nucleophilic attack
of the enolate to acetaldehyde limits rates of conversion. These rate
comparisons suggest that aldol condensation rates are greater dur-
ing reactions of C2D4O due to either: an inverse, secondary kinetic
isotope effect caused by changes from sp2 to sp3 hybridization of C-
atoms upon nucleophilic attack, or an equilibrium, primary isotope
effect caused by a greater equilibrium constant for acetaldehyde
deprotonation to the enolate for C2D4O compared to C2H4O [54].
Within aldol addition, the inverse, secondary kinetic isotope effect
acetaldehyde or inverse second order dependence on ethanol
(i.e., r ꢁ [C2H4O]2 [C2H5OH]ꢀ2, Fig. 2) on TiO2 surfaces that were
observed to have active sites covered by ethanol (Fig. 3b) in this
investigation. In addition, kinetically relevant deprotonation of
acetaldehyde to form the enolate would result in a normal kinetic
isotope effect (i.e., kH/kD > 1, as observed in the prior studies
[25,28]), whereas, an inverse kinetic isotope effect arises when
C2D4O is substituted for C2H4O here. Consequently, enolate forma-
tion does not appear to be kinetically relevant under the conditions
used here for aldol condensation within acetaldehyde – ethanol
reactant mixtures.
Young et al. concluded that adsorption or desorption processes
limited aldol condensation rates on 2-butenal (MASI) covered sur-
face of TiO2 when the reactant stream contained only acetaldehyde
after measuring rates that initially increased in proportion to the
acetaldehyde pressure before becoming constant (0.05–35 kPa
C2H4O, 533–553 K) and observing a negligible kinetic isotope effect
upon isotopically substituting acetaldehyde (kH/kD = 0.98, 0.26 kPa
C2H4O, 553 K) [19]. In this investigation, we also observed that
aldol condensation rates did not change upon substituting C2D4O
for C2H4O in the absence of ethanol (see supporting information,
Fig. S18). Comparisons of those results to the experimental rate
measurements (Fig. 2) and kinetic isotope effects (Fig. 4) obtained
within mixtures of ethanol and acetaldehyde strongly suggest that
the presence of high coverages of ethanol on active sites may be
responsible for changing the kinetic regime for aldol condensation
perhaps by significantly decreasing the proximity and the abun-
dance of enolate-aldehyde reactant pairs. Therefore, nucleophilic
attack of the enolate to form the CAC bond appears to largely
determine the rates of aldol condensation when reactant streams
contain both aldehydes and alcohols, yet other mechanisms may
dominate when the composition of the gas phase changes the dis-
tribution of surface intermediates significantly.
would reflect rehybridization of the C-atom
a to the carbonyl from
sp2 within the enolate towards sp3 at the transition state for nucle-
ophilic attack to acetaldehyde. This change in C-atom hybridiza-
tion leads to narrower potentials for molecular bending modes
and lower zero point energies at the transition state, which cause
reactions of deuterated reactants to proceed at greater rates than
the fully hydrogenated analogues. Yet, a kH/kD value of 0.6 reflects
an unusually large change secondary isotope effect for which val-
ues of 0.8–0.9 are more common [54]. Alternatively, an equilibrium
isotope effect with use of C2D4O results could lead to larger surface
coverages of reactive enolates because of greater equilibrium con-
stants for acetaldehyde deprotonation. Deprotonation involves
3.6. Identity and number of sites that facilitate aldol condensation
cleavage of the
a
-CAH bond in acetaldehyde and formation of an
Aldol condensation rates that are set by the rate at which
C2H4O*A is attacked by enolates (i.e., C2H3O*A) depend on the cov-
erage of these reactive species (Eq. (5), and consequently, the num-
ber of Lewis acid sites that participate in the catalytic cycle. To test
further the proposal that two Lewis acid sites participate in the
kinetically relevant step, steady-state aldol condensation rates
and the coverage of a pyridine (C5H5N, a titrant for Lewis acid sites)
were measured as a function of the pressure of C5H5N.
OAH bond to the Lewis base site of TiO2, and the involvement of
a deuterium atom in this process is expected to reduce the reaction
enthalpy due to the larger difference between zero point energies
in the stronger OAH/D bond [55]. Both possible interpretations of
the observed kinetic isotope effects for aldol addition rates are con-
sistent only with a catalytic sequence in which nucleophilic attack
to form the CAC bond (step 1.3) is kinetically relevant, which sup-
ports further the interpretation of the observed dependence of pro-
duct formation rates on reactant pressures (Fig. 2) and the
predominance of ethanol on TiO2 surfaces during catalysis
(Fig. 3).
The mechanistic conclusions proposed here differ from those
proposed in several recent investigations of aldol condensation
on similar TiO2 catalysts. Wang et al. concluded that conversion
rates were determined by kinetically relevant deprotonation of
aldehydes to form enolates based on a combination of DFT-
calculations conducted on a Ti5O19H18 cluster that predicted this
to be the most unstable transition state and measured rates that
were proportional to the aldehyde (i.e., acetaldehyde or propanal)
pressure and independent of the alcohol (i.e., ethanol or propanol)
pressure, and which exhibited a normal kinetic isotope effect on
TiO2 at conditions (ꢁ0.1 kPa aldehyde, 0.7 kPa alcohol, 523 K) sim-
ilar to those used in this study but in the presence of a Cu-SiO2 co-
catalyst [25,28]. This conclusion involved an assumption that
active sites for the aldol condensation reaction were primarily
unoccupied during catalysis (i.e., * as MARI). The rate expression
for aldol condensation derived from the assumption that enolate
formation is kinetically relevant (see section S5.3. of SI for the
derivation) does not account for a second order dependence on
Fig. 5a shows selected regions of in situ FTIR spectra obtained
during steady-state aldol condensation in the presence of increas-
ing amounts of co-fed C5H5N (0.9 kPa C2H4O, 2.0 kPa C2H5OH, 0–
8.7 kPa C5H5N, 15 kPa H2, 443 K), and these spectra show clearly
bending modes of the methyl and methylene function of ethanol
(i.e.,
(Fig. S12 shows the full spectra from 1300 to 3100 cmꢀ1). The
intensity of the (C@O) feature from adsorbed acetaldehyde at
1720 cmꢀ1 has an intensity more than 20-fold less than
s(CH3)
of ethanol in the absence of C5H5N (Fig. S12). As the pressure of
C5H5N increases (0–8.7 kPa C5H5N), the intensity of the (C@O)
mode of acetaldehyde decreases (Fig. 5a), and vibrational features
indicative of C5H5N bound at Lewis acidic ( 19b(CCN) at 1443
cmꢀ1) and Bronsted acidic sites ( 19b(CCN) at 1540 cmꢀ1) [56]
m
(CAH), d(CAH)) between 1350 and 1500 cmꢀ1, respectively
m
m
m
m
m
appear in addition to the modes of ethanol within the same range.
Figs. S12 and S14 show in situ FTIR spectra that indicate the intro-
duction of C5H5N results in similar changes in the coverage of
acetaldehyde even at significantly different conditions (0.3 kPa
C2H4O, 2.7 kPa C2H5OH, 0–15 kPa C5H5N, 15 kPa H2, 503 K). The
ratio of the area of the feature at 1720 cmꢀ1 to that at 1443 cmꢀ1
(after subtracting the contributions of the das(CH3) mode of ethanol
centered at 1449 cmꢀ1) is denoted as
v (i.e., v = A1720 / (A1443-

H. Zhang et al. / Journal of Catalysis 361 (2018) 290–302
301
(b)
(a)
100
100
10-1
10-2
10-3
10-4
10-1
10-2
10-3
10-4
X
5
0.001 0.01 0.1
1
10
100
1760
1600 1520 1440 1360
Wavenumber (cm-1)
Pyridine Pressure (kPa)
Fig. 5. (a) In situ FTIR spectra of surface species formed on TiO2, 673 as a function of the C5H5N partial pressure at values ranging from 0 kPa (
), 1.7 kPa ( ), 3.2 kPa ( ) and 8.7 kPa (
C2H4O, 2.7 kPa C2H5OH, 15 kPa H2, 503 K) and the square of
(0.9 kPa C2H4O, 2.0 kPa C2H5OH, 0–8.7 kPa C5H5N, 15 kPa H2, 443 K). The dashed lined are included to guide the eye. A greater C2H4O pressure (0.9 kPa) were needed in FTIR
measurements in order to quantify the absorbance at 1720 cmꢀ1
), 0.16 kPa (
), 0.77 kPa
(
) (0.9 kPa C2H4O, 2.0 kPa C2H5OH, 0–8.7 kPa C5H5N, 15 kPa H2, 443 K), and (b) Aldol condensation rates (d) (0.15 kPa
v
(
) (the ratio of FTIR peak areas at 1720 cmꢀ1 and at 1443 cmꢀ1) as a function of C5H5N pressure on TiO2, 673
.
A1449,EtOH); deconvoluted FTIR spectra are included in Supporting
Information, Fig. S13 and Table S2, and consequently, is propor-
both the pressure of C2H5OH and the pressure of C5H5N (when >
1 kPa C5H5N). The isotopic substitution of C2H4O with C2D4O leads
to an unexpected increase in aldol condensation rates, which can
be consistent only with an associated increase in the coverage of
the enolate intermediate and which are not consistent with kinet-
ically relevant deprotonation to form the enolate. The use of in situ
FTIR demonstrates that reactive acetaldehyde species exist at very
low (but detectable) coverages throughout the range of conditions
tested here, and the surfaces of TiO2 and the reactive sites are pre-
dominantly covered by strongly bound alcohols and ethanol,
respectively. Comparisons to recent publications that have imple-
mented somewhat different reaction conditions strongly suggest
that the kinetic regime and the identity of the kinetically relevant
step depend on the prevalent surface intermediates, and that in the
presence of an ethanol as the MARI the kinetics of this system
change to reflect the difficulty of forming the required pairs of
reactive intermediate (i.e., enolate-acetaldehyde) in close
proximity.
v
tional to the ratio of the coverage of C2H4O*A to that of C5H5N*A.
Fig. 5b shows that value of v2 decreases significantly as the pres-
sure of C5H5N increases (v2 ꢁ [C5H5N]ꢀ2.0 0.2), which demon-
strates that C5H5N displaces acetaldehyde from Lewis acid sites
on the TiO2 surface. Fig. 5b shows also that aldol condensation
rates are influenced by the presence of C5H5N. Rates remain con-
stant at ꢁ0.15 (mmol of Cꢆ4 product)(mol Tis s)ꢀ1 with the addition
of C5H5N (0.002–1.0 kPa) to the reactant feed, however, aldol rates
decrease with a near inverse second order dependence on the
C5H5N pressure (r ꢁ [C5H5N]ꢀ2.1 0.1) at C5H5N pressures greater
than 1 kPa under these conditions (0.15 kPa C2H4O, 2.7 kPa
C2H5OH, 15 kPa H2, and 503 K). The combined observations of a
negative second order dependence of aldol rates (i.e., r ꢁ
[C5H5N]ꢀ2) and the similar pressure dependence of v2, which
reflects the displacement of C2H4O from active sites by C5H5N,
strongly suggest that two Lewis acid sites are involved in the kinet-
ically relevant step for aldol condensation on TiO2. Moreover, these
results indicate that the kinetically relevant step involves two
C2H4O*A species. This conclusion is consistent with our assumption
that the nucleophilic attack between the in situ generated enolate
and the molecularly adsorbed acetaldehyde is the kinetically rele-
vant for CAC bond formation.
Acknowledgements
We thank Ms. Megan E. Witzke, Mr. Daniel T. Bregante and Dr.
Takahiko Moteki for valuable discussions, and Mr. Kenan Al-
Bardan for assistance with analysis of rate measurements. MYI
acknowledges departmental student fellowships supported by
Dow Chemical and 3M. This material is based upon work sup-
ported by the National Science Foundation (NSF) under grant num-
ber CBET-1511819.
4. Conclusions
Multiple forms of experimental evidence including kinetic mea-
surements, in situ infrared spectroscopy, kinetic isotope effects,
and the influence of the competitive adsorption of pyridine suggest
that the aldol condensation of acetaldehyde proceeds in the pres-
ence of ethanol via kinetically relevant nucleophilic attack of an
enolate surface intermediate to a nearby acetaldehyde molecules,
which are both bound to Lewis acidic Ti-cations on the anatase sur-
face of TiO2. The involvement of these two species and these active
sites is consistent with rates that increase as the square of the
C2H4O pressure and decrease with a second order dependence on
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at https://doi.org/10.1016/j.jcat.2018.02.030.
References
[1] M. Balat, H. Balat, C. Oz, Prog. Energy Combust. Sci. 34 (2008) 551–573.
[2] A. Demirbas, Prog. Energy Combust. Sci. 33 (2007) 1–18.

302
[3] R.W. Bentley, Energy Policy 35 (2007) 4763–4764.
H. Zhang et al. / Journal of Catalysis 361 (2018) 290–302
[30] S.C. Luo, J.L. Falconer, J. Catal. 185 (1999) 393–407.
[4] Service, R. F, Science 339 (2013) 1374–1379.
[31] H. Idriss, K.S. Kim, M.A. Barteau, J. Catal. 139 (1993) 119–133.
[32] J.D. Lewis, S. Van de Vyver, Y. Roman-Leshkov, Angew. Chem., Int. Ed. 54
(2015) 9835–9838.
[33] T. Moteki, A.T. Rowley, D.W. Flaherty, ACS Catal. 6 (2016) 7278–7282.
[34] H. Idriss, M.A. Barteau, Catal. Lett. 40 (1996) 147–153.
[35] I.M. Hill, S. Hanspal, Z.D. Young, R.J. Davis, J. Phys. Chem. C 119 (2015) 9186–
9197.
[5] J.T. Kozlowski, R.J. Davis, ACS Catal. 3 (2013) 1588–1600.
[6] L. Wu, T. Moteki, A.A. Gokhale, D.W. Flaherty, F.D. Toste, Chem 1 (2016) 32–58.
[7] P. Anbarasan, Z.C. Baer, S. Sreekumar, E. Gross, J.B. Binder, H.W. Blanch, D.S.
Clark, F.D. Toste, Nature 491 (2012) 235–239.
[8] D.C. Rakopoulos, C.D. Rakopoulos, D.T. Hountalas, E.C. Kakaras, E.G. Giakoumis,
R.G. Papagiannakis, Fuel 89 (2010) 2781–2790.
[9] D.M. Alonso, J.Q. Bond, J.A. Dumesic, Green Chem. 12 (2010) 1493–1513.
[10] D. Gabriels, W.Y. Hernandez, B. Sels, P. Van Der Voort, A. Verberckmoes, Catal,
Sci. Technol. 5 (2015) 3876–3902.
[11] A. Chieregato, J. Velasquez Ochoa, C. Bandinelli, G. Fornasari, F. Cavani, M.
Mella, ChemSusChem 8 (2015) 377–388.
[12] J.I. Di Cosimo, V.K. Diez, C.R. Apesteguia, Appl. Catal., A 137 (1996) 149–166.
[13] W.Q. Shen, G.A. Tompsett, R. Xing, W.C. Conner, G.W. Huber, J. Catal. 286
(2012) 248–259.
[14] J.I. Di Cosimo, C.R. Apesteguia, M.J.L. Gines, E. Iglesia, J. Catal. 190 (2000) 261–
275.
[15] S. Shylesh, D. Kim, A.A. Gokhale, C.G. Canlas, J.O. Struppe, C.R. Ho, D. Jadhav, A.
Yeh, A.T. Bell, Ind. Eng. Chem. Res. 55 (2016) 10635–10644.
[16] M.J.L. Gines, E. Iglesia, J. Catal. 176 (1998) 155–172.
[17] S. Hanspal, Z.D. Young, J.T. Prillaman, R.J. Davis, J. Catal. 352 (2017) 182–190.
[18] J. Sun, K. Zhu, F. Gao, C. Wang, J. Liu, C.H.F. Peden, Y. Wang, J. Am. Chem. Soc.
133 (2011) 11096–11099.
[19] Z.D. Young, S. Hanspal, R.J. Davis, ACS Catal. 6 (2016) 3193–3202.
[20] T. Moteki, D.W. Flaherty, ACS Catal. 6 (2016) 4170–4183.
[21] L. Silvester, J.-F. Lamonier, J. Faye, M. Capron, R.-N. Vannier, C. Lamonier, J.-L.
Dubois, J.-L. Couturier, C. Calais, F. Dumeignil, Catal, Sci. Technol. 5 (2015)
2994–3006.
[36] M.Y.S. Ibrahim, Master thesis from University of Illinois, Urbana-Champaign,
2015.
[37] D.S. Wang, L.B. Xiao, Q.Z. Luo, X.Y. Li, J. An, Y.D. Duan, J. Hazard. Mater. 192
(2011) 150–159.
[38] J. Zhang, M.J. Li, Z.C. Feng, J. Chen, C. Li, J. Phys. Chem. B 110 (2006) 927–935.
[39] A. Weibel, R. Bouchet, F. Boulc’h, P. Knauth, Chem. Mater. 17 (2005) 2378–
2385.
[40] J. Wang, V.F. Kispersky, W.N. Delgass, F.H. Ribeiro, J. Catal. 289 (2012) 171–
178.
[41] S.L. Shannon, J.G. Goodwin, Chem. Rev. 95 (1995) 677–695.
[42] T. Moteki, A.T. Rowley, D.T. Bregante, D.W. Flaherty, ChemCatChem 9 (2017)
1921–1929.
[43] H. Zhang, X. Pan, J. Liu, W. Qian, F. Wei, Y. Huang, X. Bao, ChemSusChem 4
(2011) 975–980.
[44] P.R.H. Griffiths, A. James, John Wiley & Sons, Inc.: Hoboken, NJ 2007.
[45] J. Rasko, A. Hancz, A. Erdohelyi, Appl. Catal., A 269 (2004) 13–25.
[46] J. Quesada, L. Faba, E. Diaz, S. Bennici, A. Auroux, S. Ordonez, J. Catal. 329
(2015) 1–9.
[47] H. Hollenst, H.H. Gunthard, Spectrochim. Acta, Part A A 27 (1971) 2027–2060.
[48] W.C. Wu, C.C. Chuang, J.L. Lin, J. Phys. Chem. B 104 (2000) 8719–8724.
[49] M. Falk, E. Whalley, J. Chem. Phys. 34 (1961) 1554–1568.
[50] M. Kang, J.F. DeWilde, A. Bhan, ACS Catal. 5 (2014) 602–612.
[51] K.J. Laidler, Chemical Kinetics, 3rd ed., Harper & Row, 1987, pp. 428–433.
[52] C.G. Swain, E.C. Stivers, J.F. Reuwer, L.J. Schaad, J. Am. Chem. Soc. 80 (1958)
5885–5893.
[22] S. Ogo, A. Onda, Y. Iwasa, K. Hara, A. Fukuoka, K. Yanagisawa, J. Catal. 296
(2012) 24–30.
[23] T. Tsuchida, J. Kubo, T. Yoshioka, S. Sakuma, T. Takeguchi, W. Ueda, J. Catal. 259
(2008) 183–189.
[24] C.R. Ho, S. Shylesh, A.T. Bell, ACS Catal. 6 (2016) 939–948.
[25] S. Wang, K. Goulas, E. Iglesia, J. Catal. 340 (2016) 302–320.
[26] D.G. Hanna, S. Shylesh, Y.-P. Li, S. Krishna, M. Head-Gordon, A.T. Bell, ACS Catal.
4 (2014) 2908–2916.
[27] J.E. Rekoske, M.A. Barteau, Ind. Eng. Chem. Res. 50 (2011) 41–51.
[28] S. Wang, E. Iglesia, J. Phys. Chem. C 120 (2016) 21589–21616.
[29] U. Diebold, Surf. Sci. Rep. 48 (2003) 53–229.
[53] R.P. Bell, Chem. Soc. Rev. 3 (1974) 513–544.
[54] E.V. Anslyn, D.A. Dougherty, Modern Physical Organic Chemistry, University
Science, 2005.
[55] S.J. Blanksby, G.B. Ellison, Acc. Chem. Res. 36 (2003) 255–263.
[56] M.I. Zaki, M.A. Hasan, F.A. Al-Sagheer, L. Pasupulety, Colloids Surf. A 190 (2001)
261–274.