Promoted chemoselective crotonaldehyde hydrogenation on zirconia-doped SiO2 supported Ag catalysts: Interfacial catalysis over ternary Ag–ZrO2–SiO2 interfaces

Article abstract of DOI:10.1016/j.jcat.2019.02.004

In gas-phase chemoselective hydrogenation of crotonaldehyde on Ag-based catalysts, zirconia doping on silica supports was found to improve catalytic performance in terms of unsaturated alcohol selectivity, hydrogenation activity, and stability. The surface modification of silica by zirconia doping favors the fine dispersion of Ag species due to the enhanced quantity and strength of surface acid sites, which enable construction of abundant catalytic sites effective for C[dbnd]O bond hydrogenation. High crotyl alcohol selectivity, exceeding 80%, and significant inhibition of monohydrogenation on the C[dbnd]C bond were observed on the optimal Ag/Zr–SiO₂ catalyst. Dynamic O₂ chemisorption measurement revealed that the pure Ag powders did not chemisorb O₂ irreversibly under 323 K, but SiO₂ or Zr–SiO₂ supported Ag catalysts did. The amounts of Ag active for O₂ chemisorption, which are at least one order of magnitude lower than that of surface Ag derived from TEM and XRD characterizations, match well with the perimeter interface Ag of hemispherical particles. A strong correlation between hydrogenation activity and O₂ uptake on those Ag/SiO₂ and Ag/Zr–SiO₂ catalysts with different Ag dispersions and deactivation degrees was observed, implying that the effective catalytic sites for crotonaldehyde chemoselective hydrogenation may originate from accessible interface sites with unique redox properties. Catalyst induction and deactivation were observed on both Ag/SiO₂ and Ag/Zr–SiO₂ catalysts in real catalytic operation. Changes in metal stable interface structure, rather than metal aggregation and coagulation, are assumed to be the main cause of irreversible catalyst deactivation, because the apparent Ag particle sizes changed slightly, but the oxygen chemisorption ability deteriorated considerably. Electropositive Ag sites interacting with neighboring oxygen from oxide supports at the ternary Ag–ZrO₂–SiO₂ interface are proposed to account for highly selective C[dbnd]O bond hydrogenation to produce the desired unsaturated alcohol.

Full text of DOI:10.1016/j.jcat.2019.02.004

Journal of Catalysis 372 (2019) 19–32
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Promoted chemoselective crotonaldehyde hydrogenation
on zirconia-doped SiO2 supported Ag catalysts: Interfacial
catalysis over ternary Ag–ZrO2–SiO2 interfaces
⇑
Haiqiang Lin , Hongyan Qu, Weikun Chen, Kang Xu, Jianwei Zheng, Xinping Duan, Hesheng Zhai,
⇑
Youzhu Yuan
State Key Laboratory of Physical Chemistry of Solid Surfaces, National Engineering Laboratory for Green Chemical Productions of Alcohols-Ethers-Esters, iChEM, College of
Chemistry and Chemical Engineering, Xiamen University, Xiamen 361005, China
a r t i c l e i n f o
a b s t r a c t
Article history:
In gas-phase chemoselective hydrogenation of crotonaldehyde on Ag-based catalysts, zirconia doping on
silica supports was found to improve catalytic performance in terms of unsaturated alcohol selectivity,
hydrogenation activity, and stability. The surface modification of silica by zirconia doping favors the fine
dispersion of Ag species due to the enhanced quantity and strength of surface acid sites, which enable
construction of abundant catalytic sites effective for C@O bond hydrogenation. High crotyl alcohol selec-
tivity, exceeding 80%, and significant inhibition of monohydrogenation on the C@C bond were observed
on the optimal Ag/Zr–SiO2 catalyst. Dynamic O2 chemisorption measurement revealed that the pure Ag
powders did not chemisorb O2 irreversibly under 323 K, but SiO2 or Zr–SiO2 supported Ag catalysts did.
The amounts of Ag active for O2 chemisorption, which are at least one order of magnitude lower than that
of surface Ag derived from TEM and XRD characterizations, match well with the perimeter interface Ag of
hemispherical particles. A strong correlation between hydrogenation activity and O2 uptake on those
Ag/SiO2 and Ag/Zr–SiO2 catalysts with different Ag dispersions and deactivation degrees was observed,
implying that the effective catalytic sites for crotonaldehyde chemoselective hydrogenation may origi-
nate from accessible interface sites with unique redox properties. Catalyst induction and deactivation
were observed on both Ag/SiO2 and Ag/Zr–SiO2 catalysts in real catalytic operation. Changes in metal
stable interface structure, rather than metal aggregation and coagulation, are assumed to be the main
cause of irreversible catalyst deactivation, because the apparent Ag particle sizes changed slightly, but
the oxygen chemisorption ability deteriorated considerably. Electropositive Ag sites interacting with
neighboring oxygen from oxide supports at the ternary Ag–ZrO2–SiO2 interface are proposed to account
for highly selective C@O bond hydrogenation to produce the desired unsaturated alcohol.
Received 24 August 2018
Revised 31 January 2019
Accepted 3 February 2019
Keywords:
Silver catalyst
Selective hydrogenation
a,b-Unsaturated aldehyde
Catalyst characterization
Zirconia doping
Interfacial site
Ó 2019 Elsevier Inc. All rights reserved.
1. Introduction
nature of metal–support interactions controlling activity, selectiv-
ity, and stability of supported metal catalysts remains unclear. The
In the field of heterogeneous catalysis, catalyst supports not
only allow the efficient dispersion of catalytically active phases
and the significant enrichment of surface active sites, but also
enable the construction of diverse low-coordinated or interfacial
sites to exhibit distinct catalytic behavior governed by specific
chemical and physical states. The support effects are usually
demonstrated to arise from metal cluster or nanoparticle stabiliza-
tion, mutual charge transfer, and synergetic participation of cat-
alytic sites located on support in catalysis. However, the exact
challenging chemoselective hydrogenation of a,b-unsaturated
aldehydes to produce unsaturated alcohols is of great fundamental
and industrial significance. Using proper catalyst supports (TiO2,
SiO2, CeO2, ZnO, etc.) to construct metal (Pt, Ag, Au, Ir, etc.) cata-
lysts has proven to be the key to driving C@O bond hydrogenation
to surpass the thermodynamically favored hydrogenation of the
C@C bond [1–16]. Among the metals catalytically active for hydro-
genation, Ag, although usually displaying relatively poor hydro-
genation activity due to its low affinity and activation force for
dihydrogen, displays high intrinsic capability to produce unsatu-
rated alcohols with remarkably high intramolecular selectivity in
⇑
Corresponding authors.
E-mail addresses: hqlin@xmu.edu.cn (H. Lin), yzyuan@xmu.edu.cn (Y. Yuan).
https://doi.org/10.1016/j.jcat.2019.02.004
0021-9517/Ó 2019 Elsevier Inc. All rights reserved.

20
H. Lin et al. / Journal of Catalysis 372 (2019) 19–32
heterogeneous chemoselective hydrogenation of
aldehydes (acrolein and crotonaldehyde) [9,17–20].
a
,b-unsaturated
the clean Ag(1 1 0) surface formed undesired saturated aldehyde
as the main product, but the surface Osub/Ag(1 1 1) with subsur-
face oxygen centers favored the formation of unsaturated alcohol
[26]. Experimentally, a profound influence of oxygen treatment
on activity and selectivity was observed on the Ag/ZnO and Ag/
SiO2 catalysts for acrolein hydrogenation [2]. Oxygen treatment
significantly enhanced the allyl alcohol selectivity of Ag/ZnO,
but subsequent hydrogen treatment only slightly decreased the
allyl alcohol selectivity, suggesting the importance of subsurface
oxygen. Introducing a small amount of air into the reaction mix-
ture was observed to promote allyl alcohol selectivity to a high
level of 42% on Ag/SiO2 catalyst. These facts stress that construct-
ing electropositive or polar Ag sites favoring the coordination of
the C@O bond is crucial for obtaining high production efficiency
for unsaturated alcohols. For real catalysts, constructing effective
Ag catalytic sites through the relatively stable interaction
between Ag atoms and neighboring oxygens from oxide supports
or additives may be more practical than simply inserting mobile
oxygen into Ag ensembles by high-temperature calcination.
An interesting pressure dependence of product distribution in
Differently from effective Pt-based catalysts supported on redu-
cible oxides such as TiO2 and CeO2 [2,21,22], Ag catalysts display
better unsaturated alcohol selectivity on nonreducible oxide
(SiO2, Al2O3, and ZnO). The O-vacancy sites on the Pt/TiO2 surface
help to activate the C@O bond of crotonaldehyde and form an
intermediate to react with the spillover H provided by the Pt to
produce crotyl alcohol. In contrast, Pt/SiO2 displayed poor perfor-
mance as a result of the inactive role of SiO2 [22]. Claus et al. com-
pared the performance of monometallic Ag catalysts supported on
TiO2 and SiO2 in crotonaldehyde hydrogenation [18]. Ag/TiO2-LTR
reduced at a low temperature of 473 K displayed a selectivity for
crotyl alcohol of 53%, which was much higher than that obtained
on Ag/TiO2-HTR reduced at a high temperature of 773 K (28%).
An antipathetic Ag particle size effect was observed, as Ag/TiO2-
HTR with a smaller Ag particle size of 1.4 0.5 nm possessed lower
hydrogenation activity than Ag/TiO2-LTR with a larger Ag particle
size of 2.8 1.9 nm. In contrast, Ag/SiO2 catalysts from different
preparation methods including sol–gel, impregnation, and
precipitation–deposition showed comparable crotyl alcohol selec-
tivity of 59 3% in spite of their differences in Ag particle size, pre-
senting structure-insensitive catalysis. For the difficult
chemoselective hydrogenation of acrolein, which is the simplest
gas-phase hydrogenation of
a,b-unsaturated aldehydes was
observed, and the change of adsorption geometry induced by dif-
ferent substrate surface coverage was assumed to be the main
cause [27,28]. Claus et al. found that the threshold pressure for
allyl alcohol formation was about 100 mbar, and the increase of
reaction pressure from 2 mbar to 20 bar led to a significant
increase of allyl alcohol selectivity from 0% to 42% on Ag/SiO2
[18,23]. Lambert et al. reported that on the Ag(1 1 1) surface, high
acrolein coverage forces the C@C bonds to tilt markedly and ren-
ders them less vulnerable to hydrogenation [10]. In addition, com-
petitive adsorption among various surface species including
hydrogen, intermediates, and deposits also influences the product
distribution. A study of acrolein hydrogenation on Ag/SiO2 revealed
that increasing partial pressures of acrolein and/or hydrogen from
50 mbar to 20 bar all led to increasing selectivity for allyl alcohol
[23]. Similarly, our previous investigation of bimetallic AuAg/
SBA-15 catalysts showed that under a constant reaction pressure
of 3.0 MPa, increasing the ratio of hydrogen to crotonaldehyde
brought about an increase in crotyl alcohol selectivity [29]. The
a,b-unsaturated aldehyde without a substituent at the C@C group,
monometallic Ag/SiO2, Ag/Al2O3, and Ag/ZnO showed an order of
maximum allyl alcohol selectivity of Ag/SiO2 (39%) < Ag/Al2O3
(42%) and Ag/ZnO (50%), reflecting the marked influence of oxide
supports [23]. A small Ag particle size effect was also revealed in
acrolein hydrogenation. Ag/SiO2 from incipient wetness with a
small average particle size of 2.5 nm displayed a turnover fre-
quency comparable to that of Ag/SiO2 from precipitation–deposi
tion with bigger Ag particles (14.2 nm), while the former showed
somewhat higher allyl alcohol selectivity [23].
An adsorption geometry of
a,b-unsaturated aldehydes on sur-
face catalytic sites is generally proposed for interpreting the
selectivity patterns. To date, the exact location and structure of
effective catalytic sites on Ag-based catalysts for the chemisorp-
tion and activation of the C@C and C@O bonds of
a,b-
unsaturated aldehydes and dihydrogen remain unclear and highly
controversial. Facet sites, low-coordination sites such as kinks or
edges, and interfacial sites on the boundaries of Ag particles and
supports have all been suggested as possible catalytic sites. Meyer
and co-workers found that the selectivity to allyl alcohol in Ag/
SiO2-catalyzed acrolein hydrogenation went up with increasing
Ag particle size from 1 to 9 nm and supposed that the Ag(1 1 1)
sites were likely responsible for the selective hydrogenation of
the C@O bond according to the variation of different surface sites
with metal particle size [15]. Theoretical study also predicted that
high-coordination-number sites over Ag(1 1 1) possess higher
selectivity for propanol than low-coordination-number sites
[24]. However, periodic density-functional calculations by Lim
et al. showed that on clear and regular Ag(1 1 1) surfaces, acrolein
is very weakly adsorbed, and the probability of Ag(1 1 1) being
responsible for the preferential activation of C@O bonds is very
low. Further, the flat Ag(1 1 0) and stepped Ag(2 2 1) surfaces also
interact very weakly with acrolein and provide very limited bond
activation. Therefore, the authors concluded that morphological
defect sites composing low-coordinated Ag atoms are responsible
for the formation of allyl alcohol [25]. Claus et al. ascribed the
high selectivity to crotyl alcohol over the Ag–Mn/SiO2 and Ag–
La/SiO2 catalysts to the formation of interfacial sites consisting
of accessible lower-valent metal cations or oxygen vacancies on
the matrix adjacent to Ag sites, which can induce interaction with
the free electron pair of the oxygen atom of the C@O bond.
Another density-functional study by Lim et al. suggested that
observed sensitivity of selective hydrogenation of
a,b-
unsaturated aldehydes mainly arises from the activation of polar
C@O bonds and nonpolar C@C bonds and the subsequent hydro-
genation with surface hydrogen species occurring on different cat-
alytic sites.
Rapid evolution of catalytic behavior in the initial stage of cat-
alytic operation and catalyst deactivation or instability are fre-
quently observed in the gas-phase hydrogenation of
a,b-
unsaturated aldehydes. Pt- or Rh-based catalysts commonly
undergo rapid deactivation during the first several hours on stream
and then step into the steady state with a relatively low deactiva-
tion rate [1,30]. However, few reports on catalytic performance
evolution of Ag-based catalysts in gas-phase hydrogenation of
a,
b-unsaturated aldehydes have been available so far. Strong time
dependence of activity and selectivity was reported in low-
pressure acrolein hydrogenation on the Ag/SiO2 [23]. At a reaction
pressure of 266 mbar, an induction period of 9 min for allyl alcohol
formation was instantly followed by a decrease in allyl alcohol
selectivity within a short reaction time of 20 min. Empirically, it
should be realized that the morphology and size of Ag aggregates,
the chemical state of Ag species, and the interaction and interfacial
structure between Ag crystallites and supports or additives may
change dynamically under real reaction conditions, resulting in
the induction, deactivation, or variation of Ag catalysts. With the
goal of developing practical Ag-based catalysts, catalytic stability
is a feature as important as activity and unsaturated alcohol
selectivity.

H. Lin et al. / Journal of Catalysis 372 (2019) 19–32
21
Here, we report an efficient approach involving surface modifi-
cation by zirconia doping on silica to construct a kind of supported
Ag catalyst displaying enhanced hydrogenation ability for the alde-
hyde group and improved long-term catalytic stability in the gas
phase hydrogenation of crotonaldehyde. Zirconia has moderate
acidity and basicity as well as reducing and oxidizing properties
and can easily combine with silica to form an oxide composite with
enhanced surface polarity [31,32], which may favor the stabiliza-
tion of Ag aggregates and the creation of polar Ag catalytic sites.
With the aim of studying the correlation of catalytic behavior
and catalyst microstructure, the performance of a series of Ag/
SiO2 and Ag/Zr-SiO2 catalysts with different Ag loading and zirco-
nia content was evaluated in the fixed-bed mode and kinetics
under steady-state conditions was studied. The physicochemical
properties of catalysts were characterized using transmission elec-
tron microscopy (TEM), X-ray photoelectron spectroscopy (XPS),
nitrogen physisorption, oxygen chemisorption, H2 temperature-
programmed reduction (H2 TPR), and powder X-ray diffraction
(XRD). Furthermore, the evolution of hydrogenation behavior on
Ag/SiO2 and Ag/Zr-SiO2 in a long catalytic operation was monitored
and the possible reasons for catalyst induction and deactivation are
discussed.
XRD patterns were recorded on a PANalytical X’pert Pro Super
X-ray diffractometer using CuK radiation (35 kV, 20 mA,
k = 0.15418 nm). The full width at half maximum of the XRD peak
at 2h = 38.2° for Ag(1 1 1) was adopted to calculate the average size
of Ag crystallites using Scherrer’s equation.
a
The H2 TPR profiles were recorded on a Micromeritics Auto-
chem II 2920 instrument. After a 0.1-g sample was pretreated in
5% O2/Ar at 373 K for 1 h, the temperature was ramped from 300
to 1073 K at 10 K/min in a flow of 5% H2/Ar. A thermal conductivity
detector (TCD) was used to monitor the variation of H2 concentra-
tion in the effluent gas from a cold trap for removing moisture. The
quantitative analysis of reducible species on catalyst precursors
was also performed by calibrating the TPR peak area with 15 mg
high-purity Cu2O powders. The measurement of the ammonia
temperature-programmed desorption (NH3 TPD) was performed
on the same instrument. A Hiden Qic-20 mass spectrometer was
used to detect the desorbed ammonia with the signals of m/
e = 16 and 17 in multiple ion detection mode. Prior to the temper-
ature ramping at a rate of 10 K/min from 373 K to 1073 K in a flow
of He, a 0.1-g sample was reduced by hydrogen at 473 K for 2 h and
then saturated with ammonia in a flow of 10% NH3/Ar at 373 K for
1 h. The dynamic measurement of O2 chemisorption on catalysts
was performed with the pulse-injection technique on
a
Micromeritics Autochem II 2920 instrument with a TCD. An ade-
quate sample (0.5–2.0 g) was reduced by hydrogen at 473 K for
2 h and then the temperature was ramped to 323 K in a stream
of high-purity He. A loop with a volume of 1.0 ml was used to
introduce 5% O2/He to the reduced sample until the adsorption
was saturated.
2. Experimental
2.1. Chemicals and reagents
Silver nitrate (AgNO3), zirconyl nitrate (ZrO(NO3)2), crotonalde-
hyde, SiO2, and ZrO2 were purchased from Sinopharm Chemical
Reagents (Shanghai, China). High-purity Ag and Ag2O powders
were provided by Aladdin Industrial Corporation (Shanghai,
China).
Characterization by XPS and Auger electron spectroscopy (AES)
was conducted on a Quantum 2000 Scanning ESCA Microprob
spectrometer equipped with an AlK
a X-ray radiation source
(hm = 1486.6 eV) operated at 20 mA and 15 kV. The binding energy
was calibrated with the peak of Si2p at 103.4 eV characteristic of
SiO2. Reduction treatment with hydrogen on Ag catalysts was con-
ducted in a glove box and the reduced sample was transferred to
an XPS analysis warehouse without exposure to air.
2.2. Catalyst preparation
Zirconia-doped silica supports (Zr–SiO2) were synthesized by
impregnating commercial silica spheres (40–60 mesh, specific sur-
face area 278 m2/g) with an aqueous solution of ZrO(NO3)2. The
mixture was subjected to aging at room temperature for 24 h.
The obtained powders were dried at 353 K for 12 h and then cal-
cined at 623 K for 3 h to yield the Zr–SiO2 supports. Ag/SiO2 and
Ag/Zr–SiO2 catalysts were prepared using the method of incipient
wetness impregnation. The dry support was mixed with an aque-
ous solution of AgNO3 and the mixture was aged at room temper-
ature for 24 h in the dark. Calcination at 573 K for 3 h was
performed after drying at 353 K for 12 h. The yielded sample was
denoted as XAg/SiO2 or XAg/YZr-SiO2 (X = Ag loading in weight per-
centage and Y = zirconia content in weight percentage).
2.4. Catalytic evaluation and kinetic measurements
Gas-phase hydrogenation of crotonaldehyde was evaluated in a
fixed-bed microreactor in continuous downflow mode. For each
catalytic evaluation, 250 mg of catalyst was loaded into a quartz
tubular reactor (inner diameter 7 mm) and the top side of the cat-
alyst bed was packed with 5.0 g of quartz grains. Prior to the cat-
alytic hydrogenation of crotonaldehyde, the catalyst sample was
prereduced in a flow of H2 at a rate of 50 ml/min at 473 K for 1 h
and then cooled to the reaction temperature. Crotonaldehyde
vapor supplied from a bubbling generator at 273 K or higher tem-
peratures was carried by the H2 stream into the catalyst bed. The
outlet gas was collected by an online six-way sampling valve (Vici
Valco) and analyzed by an online gas chromatograph (GC-2010,
Shimadzu) equipped with a flame ionization detector (FID) and a
2.3. Catalyst characterization
N2 adsorption–desorption isotherms were acquired at 77.3 K on
a Micromeritics TriStar II 3020 porosimetry analyzer. All samples
were degassed at 373 K for 4 h prior to the N2 physisorption mea-
surements. The Brunauer–Emmett–Teller (BET) method was used
to calculate the specific surface area. The pore volumes and pore
distributions were derived using the Barrett–Joyner–Halenda
(BJH) method.
TEM images were recorded on an FEI Tecnai 30 electron micro-
scope operated at 300 kV. Prior to the TEM measurement, the
reduced catalyst powders were ultrasonically dispersed in ethanol
for 30 min and then dripped onto copper grids. At least 200 Ag par-
ticles were measured to calculate the average Ag particle size for
each sample.
capillary column (DB-WAX, 30 m  0.32 mm  0.5
lm). Steady-
state kinetic studies were carried out in the same equipment on
catalysts with quasi-steady activity after 12 h of catalytic hydro-
genation. The crotonaldehyde conversion was kept below 15% to
analyze the data in a differential reactor mode. The total space
velocity was increased to 16,800 ml/h g-cat and the absence of
heat and mass transfer limitations was confirmed. Helium was
used as a balance gas and the partial pressure of crotonaldehyde
was kept at 0.87 kPa in studying the effect of hydrogen partial
pressure. To minimize the influence of catalyst deactivation, all
catalytic data were collected in no longer than 7 min. Therefore,
five-point kinetic measurement for the influence of factors

22
H. Lin et al. / Journal of Catalysis 372 (2019) 19–32
including temperature, H2, and crotonaldehyde partial pressure
can be finished in 24 min.
on silica promoted both the hydrogenation activity and the selec-
tivity for the desired crotyl alcohol. A gradual increase in zirconia
doping on silica from 0.5 to 5.0 wt% led to a gradual upsurge in cro-
tyl alcohol selectivity from 68.4 to 80.1% and crotonaldehyde con-
version from 48.9 to 73.9%. Further increasing the zirconia doping
to 10.0 wt% decreased the catalytic activity but retained the crotyl
alcohol selectivity at a high level (>80%). The modification of the
silica surface by zirconia doping significantly altered the product
distribution of the gas-phase crotonaldehyde hydrogenation. Along
with the enhancement in C@O monohydrogenation induced by zir-
conia doping, the dual hydrogenation to produce butanol signifi-
cantly surpassed the monohydrogenation on the C@C bond when
the zirconia content exceeded 1.0 wt%. The 30Ag/SiO2 catalyst
without zirconia modification yielded butanal as the second main
product at a selectivity of 28.4%. In contrast, on the optimal
30Ag/5Zr–SiO2 catalyst, butanol was the second main product,
with a selectivity of 17.2%, and simultaneously the selectivity for
butanal was as low as 1.9%. Additionally, the hydrogenolysis to
produce side products of hydrocarbons was inhibited to some
extent by the zirconia doping on silica. This significant changeover
in product distribution indicates that the modification of the silica
surface by zirconia doping markedly promotes the C@O bond
hydrogenation. Significant boosting of the butanol selectivity
may be ascribed to the butanal from the monohydrogenation of
the C@C bond rapidly undergoing subsequent hydrogenation on
the C@O bond. Furthermore, extremely low selectivity for butanal
(1.9–2.6%) on the 30Ag/Zr–SiO2 catalysts with 3–10 wt% zirconia
content indicates that the rate of butanal hydrogenation is proba-
bly higher than that of crotonaldehyde. Pure zirconia was also used
to prepare Ag/ZrO2 catalyst for comparison. Very poor hydrogena-
tion performance with a conversion of 2.5% and a selectivity for
crotyl alcohol of 37.6% was observed on the 30Ag/ZrO2 with
30.0 wt% Ag loading, demonstrating that using only ZrO2 as a sup-
port for Ag is disadvantageous for the selective hydrogenation of
C@O bonds.
3. Results
3.1. Catalytic behavior of crotonaldehyde hydrogenation
3.1.1. Catalytic behavior of Ag/SiO2 and Ag/Zr–SiO2
The catalytic behaviors of silica-supported monometallic Ag
catalysts from different preparation methods in the vapor-phase
hydrogenation of crotonaldehyde have been investigated under a
relatively high reaction pressure (2.0 MPa) [33], but a detailed
report on the influence of Ag loading on catalytic performance is
still unavailable for monometallic Ag/SiO2. Table 1 lists the cat-
alytic data of a series of Ag/SiO2 prepared by incipient wetness
impregnation with Ag loading ranging from 0.5 to 30 wt%. Appar-
ently, the conversion of crotonaldehyde increases with the increase
of Ag loading due to the upsurge of Ag active sites at a fixed cata-
lyst dosage (0.25 g). Under atmospheric pressure and at 373 K, the
Ag/SiO2 catalysts with Ag loading ꢀ5 wt% produced crotyl alcohol
from C@O bond hydrogenation as the main product with a selectiv-
ity range of 64–71%, which is in agreement with the results in the
literature [33]. Decreasing the Ag loading to a low level brought
about a negative effect on C@O bond hydrogenation. The highest
selectivity to undesired butanal from C@C bond hydrogenation
and increased production of hydrocarbons from hydrogenolysis
were observed on the 0.5Ag/SiO2 catalyst with the lowest Ag load-
ing of 0.5 wt%.
It was found that high-purity Ag powders with a low specific
surface area of 0.37 m2/g displayed very poor hydrogenation activ-
ity (Conv. = 0.5% at 1.2 g loading and 373 K). On Ag powders, no
crotyl alcohol product was detected and hydrocarbons derived
from hydrogenolysis were the main products with a selectivity of
41%; butanol and butanal were also formed with selectivity of
26.7% and 32.4%, respectively. It is interesting that the desired cro-
tyl alcohol can be formed on fresh Ag2O powders with a selectivity
of 39.4%. However, crotyl alcohol rapidly disappeared in products
because most of the oxygen is consumed by hydrogen (Ag2O
+ H2 = Ag + H2O) under the reaction conditions.
3.1.2. Influence of temperature and hydrogen/crotonaldehyde ratio
For chemoselective hydrogenation of intramolecular C@O and
C@C bonds of
a,b-unsaturated aldehydes, catalytic behavior is
Previous studies on the Ag-catalyzed hydrogenation of
a
,b-
markedly influenced by reaction conditions including temperature,
H2/crotonaldehyde ratio, and pressure. Fig. 1 shows the catalytic
performance of 30Ag/SiO2 and 30Ag/5Zr–SiO2 in the quasi-steady
state as a function of reaction temperature. A similar tendency
was observed on the two catalysts that with the increase of tem-
perature the catalytic activity increased significantly at the
expense of the selectivity of the desired crotyl alcohol. Contrary
to the decrease in crotyl alcohol selectivity, the selectivity for
butanal, butanol, and hydrocarbons tended to increase in the
unsaturated aldehydes demonstrated that constructing electropos-
itive or polar Ag sites favoring the coordination of the C@O double
bond is the key to promoting the production of unsaturated alco-
hols [23,34]. To enhance the surface acidity of silica, zirconia was
chosen to decorate silica surfaces via the convenient impregnation
method. Table 2 lists the catalytic data for a series of Ag/Zr–SiO2
with a fixed 30 wt% Ag loading prepared by the stepwise impreg-
nation of ZrO(NO3)2 and AgNO3. Apparently, the doping of zirconia
Table 1
Catalytic performance of Ag/SiO2 catalysts with different levels of Ag loading.
Catalyst
Conversion (%)
Selectivity (%)
Crotyl alcohol
Butanal
Butanol
Othersa
Ag powderb
Ag2Oc
0.5
3.7
2.3
2.2
6.0
9.0
17.1
21.2
—
26.7
9.6
32.4
43.9
2.0
8.8
5.0
3.9
3.0
2.5
40.9
7.1
4.8
1.6
1.4
1.2
0.8
0.5
39.4
58.5
60.0
64.6
65.6
67.8
71.0
0.5Ag/SiO2
1Ag/SiO2
5Ag/SiO2
10Ag/SiO2
20Ag/SiO2
30Ag/SiO2
34.7
29.7
29.0
29.3
28.4
26.0
Notes: Reaction conditions: catalyst loading = 0.25 g, reaction temperature = 373 K, H2/crotonaldehyde = 100, atmospheric pressure, H2 flow rate = 10 ml/min. The catalytic
data were collected at a time on stream of 125 min after the induction period.
a
Others are by-products including C3–C4 hydrocarbons and 2-ethylhexanal.
The loading was 1.2 g and the catalytic data are taken at a time on stream of 4 min.;
The loading was 0.5 g and the catalytic data are taken at a time on stream of 4 min.
b
c

H. Lin et al. / Journal of Catalysis 372 (2019) 19–32
23
Table 2
Catalytic performances of Ag/Zr–SiO2 catalysts with different zirconia contents.
Catalyst
Conversion (%)
Selectivity (%)
Crotyl alcohol
Butanal
Butanol
Others
30Ag/SiO2
40.4
48.9
61.0
66.3
73.9
59.7
2.5
64.5
68.4
72.0
79.5
80.1
80.7
37.6
28.4
21.9
6.8
2.6
1.9
3.1
4.1
4.0
0.6
1.1
1.0
0.8
0.8
3.5
30Ag/0.5Zr–SiO2
30Ag/1Zr–SiO2
30Ag/3Zr–SiO2
30Ag/5Zr–SiO2
30Ag/10Zr–SiO2
30Ag/ZrO2
20.1
16.9
17.2
15.9
22.7
2.6
36.2
Notes: Reaction conditions: catalyst loading = 0.25 g, reaction temperature = 413 K, H2/crotonaldehyde = 100, atmospheric pressure, H2 flow rate = 10 ml/min. The catalytic
data were collected at a time on stream of 125 min after the induction period.
Apparently, high reaction temperatures (>453 K) are not beneficial
(a)
crotyl alcohol
butanol
butanal
conversion
for efficient production of the desired crotyl alcohol because the
dual hydrogenation to butanol, hydrogenolysis, and the monohy-
drogenation to butanal on the C@C bond are markedly enhanced.
Taking into account that the product distribution of crotonalde-
hyde not only is determined by the properties of the catalyst but
also changes with the conversion level, a series of catalytic data
at three temperatures (373, 413, and 453 K) at a constant H2/cro-
tonaldehyde ratio and conversion (ca. 40%) are also compared
(Table S1 in the Supplementary Data). For both 30Ag/SiO2 and
30Ag/5Zr–SiO2, increasing the reaction temperature tends to
reduce the selectivity for unsaturated alcohol and promote the
C@C hydrogenation. A distinctive feature of 30Ag/5Zr–SiO2 is its
high selectivity for the dual-hydrogenation product (butanol). In
addition, hydrogenolysis to form hydrocarbon product becomes
serious under high temperature.
others
100
80
60
40
20
0
373
453
473
413
393
433
Pressure dependence of the selectivity for unsaturated alcohol
has been discovered on the silica-supported Ag catalysts for acro-
lein hydrogenation, in which high pressure favors the formation
of allyl alcohol from monohydrogenation on the C@O bond induced
by the promotion of beneficial adsorption geometry [20]. In this
study, the product distribution of crotonaldehyde hydrogenation
was found to be highly dependent on the ratio of hydrogen to cro-
tonaldehyde under atmospheric pressure and constant hydrogen
space velocity. For 30Ag/SiO2 and 30Ag/5Zr–SiO2, decreasing the
H2/crotonaldehyde ratio from 100 to 8 led to a decrease in crotyl
alcohol selectivity and a simultaneous increase in butanal selectiv-
ity (Fig. S1). This variation in chemoselectivity strongly suggests
that the existence of adsorption competition among hydrogen, cro-
tonaldehyde, and other products or intermediates for catalytic sites
affects the adsorption geometry of crotonaldehyde. In principle,
the C@C and HAH bonds are all nonpolar and tend to compete
for metallic Ag surface sites, but the competition between the polar
C@O bond and the nonpolar HAH bond may be weaker. Thus, the
sticking probability of the C@C bond may decrease at high hydro-
gen surface coverage, and thus an increase in crotyl alcohol selec-
tivity is obtained.
Temperature / K
(b)
crotyl alcohol
butanol
butanal
others
conversion
100
80
60
40
20
0
413
4 5 3 4 7 3
393
433
373
Temperature / K
Fig. 1. Influence of reaction temperature on catalytic behavior of (a) 30Ag/SiO2 and
(b) 30Ag/5Zr–SiO2. Reaction conditions: catalyst loading = 0.25 g, H2/crotonalde-
hyde = 100, atmospheric pressure, H2 flow rate = 10 ml/min.
3.1.3. Induction, deactivation, and regeneration of Ag/SiO2 and Ag/Zr–
SiO2 catalysts
Similarly to the phenomenon reported in literature, the cat-
alytic behavior of the Ag/SiO2 and Ag/Zr–SiO2 catalysts was
observed to be rapidly changing in the initial stage of the gas-
phase hydrogenation of crotonaldehyde. Fig. 2 shows the evolution
in catalytic behavior of the 30Ag/SiO2 and 30Ag/5Zr–SiO2 at 413 K
under atmospheric pressure. It should be noted that, after the cro-
tonaldehyde vapor was switched to pass through the freshly
reduced 30Ag/SiO2 and 30Ag/5Zr–SiO2 catalysts, the carbon bal-
ances were less than 100% before the time on stream of 45 min,
indicating that strong adsorption and/or deposit of substrates
and/or products occurred in the initial stage of catalytic operation.
temperature range 373–473 K on the 30Ag/SiO2. On the optimal
30Ag/5Zr–SiO2 with a zirconia content of 5 wt%, the crotyl alcohol
selectivity remained at a high level (80.1–82.6%) as long as the
temperature did not exceed 413 K. Increasing temperature from
373 to 413 K led to a sharp increase in butanol selectivity from
6.1% to 19.6% and a simultaneous decrease in butanal selectivity
from 9.9% to 1.9%. Further increasing the temperature to 433 K or
above significantly reduced the crotyl alcohol selectivity and
enhanced the formation of butanal, butanol and hydrocarbons.

24
H. Lin et al. / Journal of Catalysis 372 (2019) 19–32
selectivity for butanal slowly increased from 1.9% to 10.3% at the
expense of the butanol selectivity, which decreased from 17.2%
to 7.6%.
(a)
100
Conv.
butanal
others
crotyl alcohol
butanol
It should be noted that catalyst deactivation occurred to some
extent on both 30Ag/SiO2 and 30Ag/5Zr–SiO2 in the quasi-steady
stage. Although zirconia doping on the Ag/SiO2 catalyst promoted
hydrogenation activity and selectivity for crotyl alcohol, the dura-
bility of modified Ag catalysts improved undistinctively. Thermal
gravimetric analysis (TGA) of the used 30Ag/SiO2 and 30Ag/Zr–
SiO2 catalysts (Fig. S2) reveals that obvious weight loss can be
observed in the temperature range 450–650 K, strongly suggesting
the formation of strong adsorbates or carbonaceous deposits dur-
ing the catalytic reaction, especially in the early stage in which a
low carbon balance was observed. The formation of strong adsor-
bates or carbonaceous deposits on the catalyst surface should be
responsible for the rapid change of catalytic behavior in the early
stage and slow deactivation in the quasi-steady stage. In addition,
these carbonaceous species may also be involved in modification of
active sites and some catalytic steps such as hydrogen storage or
exchange. Regeneration of the partially deactivated catalyst was
attempted by a 120-min calcination in air at 573 K and a hydrogen
reduction at 473 K for 30 min. The catalytic activity could be partly
restored; however, the quick catalyst deactivation reappeared,
similarly to that which happened in the transient stage observed
on the fresh catalyst. After the regeneration treatment, the crotyl
alcohol selectivity slightly increased on the 30Ag/SiO2 but
decreased to some extent on the 30Ag/5Zr–SiO2. The partially
reproducible activity implies that the complicated causes for the
deactivation of Ag/SiO2 and Ag/Zr–SiO2 catalysts may include
reversible blocking of active sites by strong adsorbates or deposits
and irreversible changes in morphology of Ag crystallites, interfa-
cial structure, and metal–support interaction.
80
60
40
20
0
carbon balance
0
200 400 600 800 1000 1200 1400
Time on stream / min
(b)
100
80
60
40
20
0
Conv.
butanal
others
crotyl alcohol
butanol
carbon balance
3.2. Characterization of Ag/SiO2 and Ag/Zr–SiO2 catalysts
0
200 400 600 800 1000 1200 1400
Time on stream / min
3.2.1. Catalyst textures and morphology of Ag particles
The N2 physisorption isotherms in Figs. S3 and S4 and the tex-
ture parameters listed in Table 3 for Ag/SiO2 and Ag/Zr–SiO2 cata-
lysts with different Ag loading and zirconia doping reveal that the
mesoporous structure of the silica support, with an average pore
size of ꢁ10 nm, was retained after the doping with zirconia and
uploading of Ag onto supports by impregnation. Uploading Ag onto
the silica support at a high Ag loading of 30.0 wt% resulted in the
markedly reduced specific surface area and pore volume of the
yielded catalysts as compared with that of the pure silica. These
reductions mainly arose from the lowered actual amount of silica
in the 30Ag/SiO2 catalyst with a silica content of 70.0 wt% and sur-
face coverage and pore blocking by Ag particles. The zirconia dop-
ing slightly altered the mesoporous system, and the apparent
reduction in specific surface area was only observed on the
30Ag/10Zr–SiO2 catalyst with a high zirconia content of 10.0 wt%.
Fig. S5 displays the XRD patterns of the Ag/SiO2 catalysts with
Ag loading ranging from 1.0 to 30.0 wt% after reduction by hydro-
gen. With increased Ag loading, the reflection intensity of Ag crys-
tallites in the cubic phase constantly increased, while the broad
peak centered at 22° ascribed to amorphous silica gradually weak-
ened due to enrichment in surface-covering Ag particles. The Ag
particle size of the 30Ag/SiO2 with a high Ag loading of 30.0 wt%
derived by Scherrer’s equation increased to 14.9 nm, while that
on the 1Ag/SiO2 with a low Ag loading of 1.0 wt% was only
6.6 nm (Table 3). TEM observation of 30Ag/SiO2 revealed that the
Ag crystallites dispersed on silica surface had a statistical mean
size of 14.8 nm (Fig. S6a), which was in good agreement with the
XRD characterization. Relatively wide size distributions of Ag par-
ticles on the surface of silica were observed. Ag crystallites smaller
Fig. 2. Evolution of the time-dependent catalytic behavior of (a) 30Ag/SiO2 and (b)
30Ag/5Zr–SiO2. Reaction conditions: catalyst loading = 0.25 g, atmospheric pres-
sure,
H2
flow
rate = 10 ml/min,
H2/crotonaldehyde = 100,
reaction
temperature = 413 K.
After a time on stream of 65 min, the carbon balances increased to
approximately 100%. On the 30Ag/SiO2, a high crotyl alcohol selec-
tivity of approximately 75% was observed at 25–45 min with med-
ium catalytic activity (Conv. < 60%). However, high crotyl alcohol
selectivity could not be retained steadily and declined sharply to
66.8% at a time on stream of 125 min, along with a rapid recession
in catalytic activity. Thereafter, the variation in crotonaldehyde
conversion and product distribution became smoother. The
30Ag/5Zr–SiO2 with zirconia doping displayed much higher cat-
alytic activity than the 30Ag/SiO2 in the initial stage. At a time
on stream of 25–45 min, high crotonaldehyde conversion exceed-
ing 90%, unusually high selectivity for butanol from dual hydro-
genation (43.5–63.9%), and extremely low butanal selectivity
(<2%) were observed. From a time on stream of 25 min to
165 min, the crotyl alcohol selectivity gradually improved from
34.1% to 81.4%, along with a decline in crotonaldehyde conversion.
This continuous improvement in crotyl alcohol production proba-
bly arose from the marked recession of the subsequent hydrogena-
tion on the C@C bond of crotyl alcohol product due to the
diminution of catalytic sites responsible for the C@C bond hydro-
genation. In a catalytic run lasting 1205 min, the selectivity for cro-
tyl alcohol fluctuated only in a small range (79.1–82.6%), but the

H. Lin et al. / Journal of Catalysis 372 (2019) 19–32
25
Table 3
Textural properties of the Ag/SiO2 and Ag/Zr–SiO2 catalysts.
Sample
SBET (m2/g)
Pore volume (ml/g)
Pore diameter(nm)
Average crystallite size (nm)
Fresh
Used
SiO2
1Ag/SiO2
5Ag/SiO2
10Ag/SiO2
20Ag/SiO2
30Ag/SiO2
30Ag/0.5Zr-SiO2
30Ag/1Zr-SiO2
30Ag/3Zr-SiO2
30Ag/5Zr-SiO2
30Ag/10Zr-SiO2
278
242
233
205
166
148
151
135
143
137
127
0.91
0.82
0.78
0.69
0.56
0.51
0.52
0.47
0.49
0.49
0.48
9.9
—
6.6
8.5
9.4
10.2
14.9
14.1
13.9
13.1
9.2
—
7.7
9.2
9.6
11.9
15.3
14.6
14.3
13.9
11.0
7.8
1.02
10.0
10.1
10.2
10.3
10.2
9.7
10.1
9.8
10.1
6.5
than 5 nm and>20 nm simultaneously existed on the surface of the
30Ag/SiO2 catalyst. This feature is consistent with the Ag/SiO2 pre-
pared from incipient wetness in the literature [19].
is mainly ascribed to the gas-phase crotonaldehyde hydrogenation
having been conducted at a relatively low temperature (ꢂ413 K),
which cannot seriously induce thermal migration and sintering
of Ag particles. Although the apparent size distribution of Ag parti-
cles changed only slightly, the irreversible deactivation that has
substantially occurred during catalytic operation reminds us that
some key factors other than the Ag particle size play important
roles in determining catalytic behavior.
The morphology and orientation of Ag particles on the SiO2 and
Zr–SiO2 supports were investigated by HR-TEM (Fig. S8). On both
30Ag/SiO2 and 30Ag/5Zr–SiO2 catalysts reduced by hydrogen at
473 K, most of the Ag particles exposed the most densely packed
(1 1 1) plane with a lattice spacing of 0.237 nm. Single-crystalline
and multiply twinned Ag particles are present on the surfaces of
SiO2 and Zr–-SiO2 supports. It should be noted that, on some
boundaries of round Ag crystallites close to the support, the mea-
sured lattice spacing is ꢁ 0.251 nm, somewhat greater than that
for Ag(1 1 1), 0.237 nm. This lattice expansion can probably be
ascribed to the interaction of interfacial Ag and the support or
the subsurface oxygen dissolved in Ag crystallites.
The different zirconia doping levels on the silica support led to
changes in the XRD patterns of the reduced 30Ag/Zr-SiO2 catalysts
(Fig. 3). The surface dispersion of Ag improved significantly, while
the zirconia content was above 3.0 wt%, as evidenced by the obvi-
ous widening of XRD peaks characteristic of metallic Ag. The aver-
age Ag particle size of the 30Ag/5Zr–SiO2 with a zirconia content of
5.0 wt% was 9.2 nm, and further increasing the zirconia content to
10.0 wt% reduced the size to 6.5 nm (Table 3). TEM observation
showed that the Ag crystallites dispersed on the surface of
30Ag/5Zr–SiO2 possessed a statistical mean size of 10.5 nm, which
was slightly larger than the result from XRD characterization
(Fig. S6c). In addition, even at a high zirconia content of 10.0 wt
%, no XRD peaks characteristic of zirconia phases were detected,
indicating that zirconia species on the silica surface were in extre-
mely high dispersion or an amorphous state.
Partially deactivated catalysts were characterized by XRD and
TEM. The XRD patterns of the used 30Ag/SiO2 and 30Ag/5Zr–SiO2
after catalytic operation at <413 K for approximately 1000 min
were similar to those of their fresh counterparts (Fig. S7), suggest-
ing that the crystallinities and apparent mean sizes of Ag crystal-
lites did not change obviously. This result was also confirmed by
TEM observation (Fig. S6). A slight difference between fresh and
used Ag catalysts in the apparent size distribution of Ag crystallites
3.2.2. Redox properties and surface acidity
The TPR profile of the 30Ag/SiO2 in Fig. 4a shows that the
reduction of Ag species on the silica surface by hydrogen began
at a relatively low temperature (<373 K) and the reduction peak
was centered at 432 K. Zirconia doping on silica supports did not
Fig. 3. Wide-angle XRD patterns of as-reduced catalysts: (a) SiO2, (b) 30Ag/10Zr–
SiO2 (c) 30Ag/5Zr–SiO2, (d) 30Ag/3Zr–SiO2, (e) 30Ag/1Zr–SiO2, (f) 30Ag/0.5Zr–SiO2
and (g) 30Ag/SiO2.
Fig. 4. H2 TPR profiles of 30Ag/Zr–SiO2 catalysts with different zirconia content: (a)
30Ag/SiO2, (b) 30Ag/0.5Zr–SiO2, (c) 30Ag/1Zr–SiO2, (d) 30Ag/3Zr–SiO2, (e) 30Ag/
5Zr–SiO2, and (f) 30Ag/10Zr–SiO2.

26
H. Lin et al. / Journal of Catalysis 372 (2019) 19–32
obviously change the intensity and shape of the reduction patterns,
but interestingly shifted the position of the reduction peak to a
higher temperature. The increase in zirconia content from 0.5 to
10.0 wt% gradually elevated the peak temperature for the reduc-
tion of Ag species from 432 to 478 K. On the 30Ag/10Zr–SiO2 with
the highest zirconia content of 10.0 wt%, the summit of the reduc-
tion peak appeared at 478 K, which was 46 K higher than that of
the unmodified 30Ag/SiO2. The change in redox properties induced
by zirconia doping possibly arises from (1) the formation of some
kind of composite oxides composed of zirconia and Ag oxides with
relatively poor reactivity to hydrogen and (2) covering of some Ag
species with zirconia to reduce the accessibility of hydrogen. Quan-
titative analysis of reducible Ag in a series Ag catalyst precursor is
also performed (Table S2). It is found that the amount of reducible
Ag species on the 30Ag/SiO2 precursor is less than that on catalysts
with zirconia modification, possibly arising from the different
extent in the decomposition of AgNO3 to metal Ag during the cal-
cination. When the zirconia content ꢀ3%, the amounts of reducible
Ag species were close to but still less than the nominal value.
The acidic properties of 30Ag/Zr–SiO2 catalysts with different
zirconia doping levels were investigated using the NH3 TPD tech-
nique (Fig. 5 and Table S3). For pure SiO2, only a weak NH3 desorp-
tion peak, which corresponds to an amount of acidic sites of
10.9 lmol/g-cat, appeared near 493 K and a calibrated amount.
The uploading of Ag onto the SiO2 surface at a high loading of
30.0 wt% significantly enhanced the intensity of the NH3 desorp-
tion peak (26.0 lmol/g-cat) and simultaneously shifted desorption
to a higher temperature at 512 K. This enhancement of NH3
adsorption on moderate acidic sites is ascribed to the Ag species
dispersed on the SiO2 surface as Lewis acid sites. Importantly, the
quantity and strength of surface acid sites were markedly
enhanced by the zirconia doping on silica. The 5Zr–SiO2, with an
optimized zirconia content of 5.0 wt%, presented a larger ammonia
ammonia bonded to the 30Ag/Zr–SiO2 catalysts desorbed at a
lower temperature than that bonded to the 30Ag/SiO2 without zir-
conia doping. The ammonia desorption peak on the 30Ag/Zr–SiO2
with zirconia content ranging from 0.5 to 5.0 wt% was located at
493–499 K, whereas that for the 30Ag/10Zr–SiO2 with the highest
zirconia content of 10.0 wt% was 512 K. In addition, under an iden-
tical Ag loading of 30 wt%, the amount of NH3 bonded to the sur-
face Ag sites tended to increase with the increase in the zirconia
content (Table S3). This mainly arose from the above-mentioned
promotional effect of zirconia doping on the Ag dispersion, as
revealed by the XRD and TEM characterizations.
3.2.3. O2 chemisorption and chemical state of Ag
Measurement of the metallic dispersion of Ag/SiO2 and Ag/Zr-
SiO2 catalysts was attempted by dynamic O2 chemisorption using
a pulse-injection technique, in which a low temperature (323 K)
was adopted with the purpose of avoiding bulk diffusion of oxygen
(Table 4). However, the unexpected result that the micrometer-
sized Ag powders barely chemisorbed oxygen cast doubt on the
feasibility of calculating the number of surface Ag atoms based
on a stoichiometric reaction between surface Ag and O2. The possi-
bility of inadequate dosage could be excluded because 2.0 g Ag
powders with a specific surface area of 0.37 m2/g were sufficient,
given that the surface Ag atoms could bond oxygen irreversibly
under 323 K. In contrast, all the reduced Ag/SiO2 and Ag/ Zr–SiO2
catalysts were able to chemisorb O2 even when the Ag loading
was as low as 1.0 wt%. The quantity of chemisorbed O2 was
strongly dependent on the Ag loading and the content of doped zir-
conia. For the Ag catalysts supported on pure SiO2, low Ag loading
appeared to be beneficial for promoting O2 uptake (per gram Ag). A
high O2 uptake of 93.8 lmol/g-Ag was observed on the 1Ag/SiO2
with a low Ag loading, and further increasing the Ag loading led
to the decline of the O2 uptake based on the Ag mass. Zirconia dop-
ing on silica brought about substantial promotion in O2 chemisorp-
tion. At a high Ag loading of 30.0 wt%, the O2 uptake on the freshly
desorption peak (24.0
pure SiO2 at 493 K (Fig. 5b). In addition, the appearance of an extra
desorption peak (45.2 mol/g-cat) at a high temperature of 799 K
lmol/g-cat) near 513 K to replace that on
l
reduced 30Ag/5Zr–SiO2 was 45.4
lmol/g-Ag, much higher than
suggests the formation of strong acid sites on zirconia–silica com-
posites. It should be noted that the uploading of Ag onto the Zr–
SiO2 composites diminished the NH3 desorption peak correspond-
ing to the strong acid sites, indicating the complete coverage of
these sites by Ag species at a high Ag loading. Particularly, the
that on the 30Ag/SiO2 without zirconia doping (28.0
l
mol/g-Ag).
It is noteworthy that the O2 chemisorption abilities of these Ag cat-
alysts correlated closely to their catalytic activities (Table 4). The
30Ag/5Zr–SiO2 with a higher O2 uptake displayed much higher cat-
alytic activity than 30Ag/SiO2. This strong correlation between cat-
alytic activity and O2 chemisorption ability was also observed on
the partially deactivated 30Ag/SiO2 and 30Ag/5Zr–SiO2. The O2
uptake of the 30Ag/SiO2 and 30Ag/5Zr–SiO2 decreased consider-
ably just as their catalytic activity did after catalytic operation at
413 K lasted for 1200 min.
The attempt to discriminate the chemisorbed oxygen species on
Ag/SiO2 and Ag/Zr–SiO2 via in situ Raman spectrum analysis gave
no valuable information, since no changes could be observed after
O2 was introduced into the reduced Ag catalysts. Based on the sto-
ichiometry of Ag/O2 = 2 used in the literature [35], the calculated
amount of active Ag capable of chemisorbing O2 (Agactive) on the
30Ag/SiO2 is 16.8 lmol/g-cat and the Agactive/Agtotal ratio (total
Ag amount in catalyst) is as low as 0.006. This extremely low value
does not match the value of Agsurface/Agtotal (0.079) derived from
the average Ag particle size determined by TEM and XRD using a
hemispherical model (diameter 14.9 nm), strongly indicating that
only a small portion of the surface Ag has a unique redox property
at a relatively low temperature of 323 K. The periphery hypothesis
for CO oxidation on supported Au catalysts is that oxygen mole-
cules are activated at the perimeter interfaces [36]. For a hemi-
spherical Ag crystallite with diameter 14.9 nm, the ratio of Ag in
the perimeter interface to Agtotal is ꢁ0.004, which is close to the
measured Agactive/Agtotal. Therefore, it is most likely that those Ag
in the accessible interface tightly contacting the support
surface are the active Ag sites for O2 chemisorption. The measured
Fig. 5. NH3 TPD profiles of oxide supports and as-reduced Ag catalysts: (a) SiO2, (b)
5Zr–SiO2, (c) 30Ag/SiO2, (d) 30Ag/0.5Zr–SiO2, (e) 30Ag/1Zr–SiO2, (f) 30Ag/3Zr–SiO2,
(g) 30Ag/5Zr–SiO2, and (h) 30Ag/10Zr–SiO2.

H. Lin et al. / Journal of Catalysis 372 (2019) 19–32
27
Table 4
O2 uptake and catalytic activity of the Ag/ SiO2 and Ag/Zr–SiO2 catalysts.
Catalyst
O2 uptake
Catalytic activitya
Based on Ag (
l
mol/g-Ag)
Based on catalyst (
l
mol/g.cat)
Based on Ag (
l
mol/g-Ag.h)
Based on catalyst (l
mol/g-cat.h)
Ag powder
0
0
5.6
5.6
1Ag/SiO2
5Ag/SiO2
10Ag/SiO2
20Ag/SiO2
93.8
72.0
42.1
35.9
28.0
45.4
16.9
24.1
0.94
3.60
4.21
7.18
8.40
13.61
5.07
7.22
2455.2
1339.2
1004.4
954.2
788.6
1718.6
465.0
751.4
24.6
67.0
100.4
190.8
236.6
515.6
139.5
225.4
30Ag/SiO2
30Ag/5Zr–SiO2
30Ag/SiO2-used
30Ag/5Zr–SiO2-used
a
Catalytic activity refers to the amount of converted crotonaldehyde per unit time and per gram catalyst or Ag. Reaction conditions: catalyst loading = 0.25 g, reaction
temperature = 373 K, H2/crotonaldehyde = 100, atmospheric pressure, H2 flow rate = 10 ml/min. The catalytic data were collected at a time on stream of 125 min after the
induction period.
Agactive/Agtotal ratio for 30Ag/5Zr–SiO2 (0.010) is higher than that
for 30Ag/SiO2. This enhancement is mainly ascribed to the promo-
tion of Ag dispersion and the increase of the perimeter interface
induced by zirconia doping.
because the BE values for them are all close to 182.8 and
185.0 eV, which are characteristic of Zr4+, that is, ZrO2. In addition,
5Zr–SiO2 calcined in air at 773 K was found to display a Zr3d XPS
spectrum similar to those of the reduced 30Ag/5Zr–SiO2 catalysts
(Fig. S9a), further indicating that the presence of Ag and the hydro-
gen treatment do not obviously change the chemical state of ZrO2
dispersed on the SiO2 surface.
The surface Si/Ag, Ag/Zr, and Si/Zr ratios of a series of 30Ag cat-
alysts determined by XPS characterization are also listed in
Table S2 with the bulk ratios. The surface Si/Ag ratios of a series
of Ag catalysts with 30 wt% Ag loading are all higher than the bulk
values, probably arising from many Ag species being dispersed on
the internal surfaces of support mesopores. Low Zr content (ꢂ3 wt
%) slightly increases the surface Si/Ag ratio in comparison with the
30Ag/SiO2 catalyst, but further increasing the Zr content reduces
the surface Si/Ag ratio. It should be noted that the surface Si/Ag
ratio of a deactivated 30Ag/5Zr–SiO2 catalyst is higher than that
of the fresh one. It is assumed that the carbonaceous deposits cov-
ering the surface Ag sites during the catalytic reaction is responsi-
ble. Similarly, the surface Si/Zr ratio on the 30Ag/5Zr–SiO2 catalyst
also increase to some extent after long-term catalytic operation.
XPS and Auger characterization were performed to investigate
the chemical state of Ag species on the 30Ag/SiO2 and 30Ag/Zr–
SiO2 catalysts. The charging was calibrated with the binding energy
(BE) of Si2p (103.4 eV) characteristic of SiO2 and then the O1s peak
was centered at 532.8 eV, in good agreement with literature data.
The peaks of Ag3d5/2 for the 30Ag/SiO2 and 30Ag/Zr–SiO2 catalysts
reduced by hydrogen at 623 K appeared in a BE range of 367.7–
368.1 eV (Fig. 6), resulting in ambiguous determination of the
chemical state of Ag species (the characteristic BEs of metallic
Ag0 and Ag2O are 368.1 and 367.7 eV, respectively). The modified
Auger parameter (the sum of the kinetic energy of the Auger elec-
tron (Ag M4N45N45) and the BE of the Ag3d5/2 core-level peak),
which is independent of charging and sensitive to chemical shifts
[37], was calculated for the 30Ag/SiO2 and 30Ag/Zr–SiO2 catalysts
(Fig. 6). The modified Auger parameters for these catalysts were
close to the Auger parameter for Ag0 at 726.1 eV [38], demonstrat-
ing that most of the Ag species were in the metallic state, which
was in good agreement with the results of XRD characterization.
The Zr3d XPS spectra for the reduced Ag/Zr–SiO2 catalysts were
also measured in order to monitor the oxidation state of zirconium
(Fig. S9). Increasing the zirconia content from 1 to 10 wt% gradually
increases the intensity of XPS peaks corresponding to Zr3d5/2 and
Zr3d3/2. No obvious difference in the chemical states of Zr species
on the Ag/Zr–SiO2 catalysts with different zirconia content exists,
3.2.4. Kinetic behavior
Difficulties in the kinetic study of the gas-phase hydrogenation
of crotonaldehyde arise because serious deactivation in the transit
stage and gradual deactivation in the quasi-steady stage always
happen. It seems impossible to validly study the kinetics in the
transit stage because all catalytic data are obtained on an Ag
Fig. 6. Ag3d XPS and Ag MNN spectra of as-reduced catalysts: (a) 30Ag/SiO2, (b) 30Ag/0.5Zr–SiO2, (c) 30Ag/1Zr–SiO2, (d) 30Ag/3Zr–SiO2, (e) 30Ag/5Zr–SiO2, and (f) 30Ag/
10Zr–SiO2.

28
H. Lin et al. / Journal of Catalysis 372 (2019) 19–32
catalyst under quite different status. However, the long-term cat-
alytic behavior displayed in Fig. 2 reveals that those Ag catalysts
in the quasi-steady stage are relatively stable and provide a chance
for valid kinetic study. Fig. S10 shows the Arrhenius plots for cro-
tonaldehyde hydrogenation over the 30Ag/SiO2 and 30Ag/Zr–SiO2
catalysts in the quasi-steady state. The apparent activation energy
(Ea) of 30Ag/SiO2 is 47.1 kJ/mol in the temperature range 373–
413 K, which is higher than that of 30Ag/Zr–SiO2 (39.7 kJ/mol).
The reduction in Ea induced by zirconia doping on the silica sup-
port clearly indicates, in addition to the changes in the size and dis-
persion of Ag particles, the surface adsorption and activation of
substrates and products desorption on the Ag/Zr–SiO2 are essen-
tially altered by the formation of unique catalytic structures co-
constructed from Ag, SiO2, and ZrO2. Since two competing reac-
tions (crotyl alcohol from C@O monohydrogenation and butanal
from C@C monohydrogenation) occur simultaneously in the cat-
alytic hydrogenation of crotonaldehyde, the Arrhenius plots of
the formation of crotyl alcohol and butanal are also shown in
Fig. S9. For both 30Ag/SiO2 and 30Ag/Zr–SiO2, the Ea(crotyl alcohol)
for the formation of crotyl alcohol is lower than Ea(butanal), sug-
gesting that low temperatures favor the selective hydrogen of
C@O bonds. Zirconia doping on silica reduces the apparent activa-
tion energies for both crotyl alcohol and butanal.
The effects of hydrogen and crotonaldehyde partial pressure on
the formation rates of hydrogenation products are displayed in
Figs. S11 and S12. On both 30Ag/SiO2 and 30Ag/5Zr–SiO2, the crotyl
alcohol and butanal formation rates increase linearly with the
increase of crotonaldehyde and hydrogen partial pressure. Over
30Ag/SiO2, the apparent reaction orders of hydrogen for crotyl
alcohol and butanal formation are 0.18 and 0.25, respectively
(Figs. 7 and 8). It is interesting that the apparent reaction orders
of crotonaldehyde for crotyl alcohol and butanal formation (0.60
and 0.62) are higher than those of hydrogen. This observation sug-
gests that the adsorption and activation of crotonaldehyde are
more rate-determining than the activation of hydrogen on the
Ag/SiO2 catalyst. The zirconia doping on SiO2 increases the H2 order
for crotyl alcohol formation from 0.18 to 0.27 and simultaneously
reduces that for butanal formation from 0.25 to 0.10. The close to
zero order of H2 for C@C bond hydrogenation on 30Ag/5Zr–SiO2
implies that the surface concentration of the active hydrogen spe-
cies responsible for the C@C hydrogenation is nearly saturated
under the reaction conditions. In contrast, the increase of hydrogen
partial pressure increases the surface concentration of the active
hydrogen species responsible for C@O hydrogenation, as evidenced
by the higher H2 order for crotyl alcohol formation. Importantly,
the apparent reaction orders of crotonaldehyde on 30Ag/Zr–SiO2
are much lower than those on 30Ag/SiO2. The zirconia doping on
SiO2 significantly reduces the crotonaldehyde order for crotyl alco-
hol formation from 0.60 to 0.34, suggesting the enhancement of
the capability for C@O activation.
4. Discussion
In this study of the influence of zirconia doping on the silica
support in Ag-catalyzed crotonaldehyde hydrogenation, promo-
tions including improved selectivity for unsaturated alcohol (crotyl
alcohol), enhanced hydrogenation activity, and promoted catalyst
stability were observed on the Ag/Zr–SiO2 catalysts. The enhance-
ment in hydrogenation activity can reasonably be ascribed to the
improved Ag surface dispersion induced by zirconia doping on sil-
ica. The XRD and TEM characterizations all clearly revealed that the
average Ag particle size of the reduced 30Ag/SiO2 was larger than
that of the reduced 30Ag/Zr–SiO2, especially when the zirconia
content exceeded 3.0 wt%. The NH3 TPD investigation illuminated
that the promotion of metallic dispersion induced by zirconia dop-
ing mainly originated from the significant change in surface acidic
properties. The quantity and strength of surface acid sites on the
Zr–SiO2 composites were markedly enhanced, as evidenced by
the significantly increased amount of desorbed ammonia, the
higher desorption temperature, and the emergence of strong acidic
sites. It is well known that silica surfaces are weakly acidic, with an
isoelectric point ranging from 1.7 to 3.5. Therefore, inadequate
electronegative adsorption sites capable of sticking to the Ag
cations in AgNO3 solution do not favor the fabrication of a catalyst
precursor with high Ag dispersion under relatively high metal
loading [39]. In fact, the 30Ag/SiO2 catalyst prepared by incipient
wetness impregnation had a large average Ag particle size close
to 15 nm. Upon zirconia doping on silica, the enhanced surface
acidity provided more adsorption sites for anchoring Ag cations
and subsequently enabled the stabilization of small Ag crystallites
-16.0
30Ag/SiO
2
30Ag/5Zr-SiO
2
-16.5
-17.0
-17.5
-18.0
-16.5
-17.0
-17.5
-18.0
-18.5
-19.0
Crotyl alcohol
Reaction order = 0.18
Crotyl alcohol
Reaction order = 0.27
Butanal
Reaction order = 0.25
Butanal
Reaction order = 0.10
3.4 3.6 3.8 4.0 4.2 4.4 4.6
3.4 3.6 3.8 4.0 4.2 4.4 4.6
Ln(PHydrogen
/ kPa)
Ln(PHydrogen/ kPa)
Fig. 7. H2 pressure dependence in the formation of crotyl alcohol and butanal on 30Ag/SiO2 and 30Ag/5Zr–SiO2. Conditions: 373 K, H2 + He = 50 ml/min,
Pcrotonaldehyde = 0.87 kPa.

H. Lin et al. / Journal of Catalysis 372 (2019) 19–32
29
-16.8
-17.0
-17.2
-17.4
-17.6
-17.8
-18.0
-18.2
-18.4
-18.6
-18.8
-19.0
-19.2
-16.6
30Ag/SiO
2
30Ag/5Zr-SiO
2
-16.8
-17.0
-17.2
-17.4
-17.6
-17.8
-18.0
-18.2
-18.4
-18.6
-18.8
-19.0
Crotyl alcohol
Reaction order = 0.34
Crotyl alcohol
Reaction order = 0.60
Butanal
Butanal
Reaction order = 0.46
Reaction order = 0.62
-0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2
Ln(Pcrotonyl aldehyde/ kPa)
-0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2
Ln(Pcrotonyl aldehyde/ kPa)
Fig. 8. Crotonaldehyde pressure dependence in the formation of crotyl alcohol and butanal on 30Ag/SiO2 and 30Ag/5Zr–SiO2. Conditions: 373 K, H2 = 50 ml/min,
Pcrotonaldehyde = 0.87–3.24 kPa.
during calcination and hydrogen reduction. The disappearance of
strong acid sites on the Ag/Zr–SiO2 catalysts revealed by the NH3
TPD characterization is a strong proof for the occupation and cov-
erage of acid sites on the Zr–SiO2 composites by Ag species. The
promotional effect of zirconia doping on Ag dispersion is signifi-
cant, so that the 30Ag/10Zr–SiO2 possessed a small Ag particle size
of 6.5 nm, even under an Ag loading as high as 30.0 wt%.
from dual hydrogenation was raised to a high level (17.2%). In con-
trast, the butanal selectivity (28.4%) of the 30Ag/SiO2 catalyst with-
out modification was much higher than the butanol selectivity
(3.1%) under identical conditions. The considerable boosting of
alcohol production on the Ag/Zr–SiO2 catalysts strongly demon-
strates that zirconia doping on silica substantially promotes the
chemoselective hydrogenation of the C@O bond. This marked pro-
motion is assumed to arise from the strong interaction between Ag
species and zirconia doped on the silica surface and the consequent
formation of electropositive Ag sites effective for bonding and acti-
vating C@O bonds.
It seems difficult to explain why the apparent Ag particle sizes
of the deactivated Ag/SiO2 and Ag/Zr–SiO2 catalysts determined by
XRD and TEM were almost unchanged, but in the meantime, the
catalytic activity deteriorated seriously. Fortunately, the measure-
ment of O2 chemisorption provided valuable information for prob-
ing the nature of active sites for chemoselective hydrogenation of
crotonaldehyde. Dynamic O2 chemisorption measurement at a rel-
atively low temperature of 323 K clearly revealed that the surfaces
of pure Ag powders mainly composing the Ag(1 1 1) facet did not
irreversibly chemisorb oxygen. An extremely low sticking proba-
bility of molecularly or atomically adsorbed oxygen species on
the Ag(1 1 1) facet had been theoretically disclosed in a previous
study [40]. Nevertheless, uploading Ag onto the surface of SiO2 or
Zr–SiO2 composites via impregnation unexpectedly activates the
ability of Ag crystallites to chemisorb oxygen at a low temperature.
This oxygen-bonding ability could be observed on Ag/SiO2 catalysts
with a wide range of Ag loading from 0.5 to 30 wt%. Furthermore,
the O2 uptake based on Ag mass for the fresh Ag/SiO2 catalysts
clearly displayed correlative variation with the Ag particle size
determined by XRD and TEM (Tables 3 and 4). This correlation
between O2 uptake and Ag particle size was also observed on the
Ag/Zr–SiO2 catalysts with zirconia doping on silica. The freshly
reduced 30Ag/5Zr–SiO2 with an average Ag particle size of
Previous investigation of silica-supported Ag catalysts from dif-
ferent preparation methods proposed that the chemoselective
hydrogenation of crotonaldehyde to crotyl alcohol was structure-
insensitive because the selectivity of unsaturated alcohol was inde-
pendent of the Ag particle size in the range 3.7–6.3 nm [18]. Our
catalytic results for a series of Ag/SiO2 catalysts with different Ag
loading and Ag particle size seem to disagree with this assumption
of size-independent catalysis. In this study, the Ag/SiO2 catalysts
with low Ag loading (0.5–1.0 wt%) and small Ag particle size
(<8.5 nm) displayed poorer crotyl alcohol selectivity (<60%) than
those with higher Ag loading and larger Ag particle size. However,
the important fact that pure SiO2 did catalyze the crotonaldehyde
hydrogenation to some extent to yield butanal as the main product
should not be ignored. Since the Ag/SiO2 catalysts with low Ag
loadings displayed poor catalytic activity, the undesired hydro-
genation products catalyzed by the silica surface uncovered by
Ag species could not be overlooked, resulting in the relatively
low crotyl alcohol selectivity. Anyway, high coverage of Ag species
on silica surface is beneficial for promoting the selectivity and
space time yield of desired crotyl alcohol.
Although the Ag particle size effect seems to be inconspicuous
in the product distribution of crotonaldehyde hydrogenation cat-
alyzed by Ag/SiO2 catalysts, the support effect is significant in
the monohydrogenation of the C@O bond to produce the desired
crotyl alcohol. Pure Ag powder is completely inactive in catalyzing
the monohydrogenation of the C@O bond. However, once silica is
adopted to support Ag particles, the monohydrogenation of the
C@O bond in crotonaldehyde is triggered and the desired crotyl
alcohol is produced as the main product with high selectivity. Fur-
thermore, the considerable change in product distribution induced
by zirconia doping on silica also stresses the crucial support effect
on Ag-catalyzed crotonaldehyde hydrogenation. On the most
active 30Ag/5Zr–SiO2 catalyst, very little butanal was formed from
the monohydrogenation of the C@C bond, with an extremely low
selectivity of 1.9%, but simultaneously, the selectivity ford butanol
9.2 nm exhibited an O2 uptake of 45.4
lmol/g-Ag, which was much
higher than that of the 30Ag/SiO2 (28.0
l
mol/g-Ag) with a larger
Ag particle size of 14.9 nm.
Importantly, the amounts of chemisorbed O2 on these sup-
ported Ag catalysts are much lower than that of surface Ag calcu-
lated from XRD and TEM characterization. Combining this finding
with the result of the HR-TEM observation that most of the surface
of Ag crystallites on SiO2 and Zr-SiO2 support is Ag(1 1 1) facets, it

30
H. Lin et al. / Journal of Catalysis 372 (2019) 19–32
is reasonable to identify those Ag species on the perimeter inter-
face instead of the surface Ag on Ag(1 1 1) as the O2 chemisorption
sites. Using the hemispherical model, the calculated ratios of
Agperimeter/Agtotal on the 30Ag/SiO2 and 30Ag/5Zr–SiO2 are all close
to the measured values of Agactive/Agtotal, strongly supporting the
perimeter interface hypothesis. The low-temperature O2
chemisorption observed on supported Ag catalysts probably arises
from the interaction between interfacial Ag and oxide support
bringing about some physiochemical changes such as lattice
expansion, coordination number, and chemical valance. Mavrikakis
et al. indicated that metal surface reactivity increases with lattice
expansion, based on their self-consistent density functional calcu-
lations [41]. Similarly to the observation in the literature [42], lat-
tice expansion on the boundary area between Ag layers and oxide
support was found in our HR-TEM characterization on supported
Ag catalysts (Fig. S7). In addition, the interfacial Ag sites on the
perimeter of the hemispheric Ag particle have a low coordination
number, like the sites on a sharp-angle step. These unique features
may endow interfacial Ag sites with enhanced reactivity to O2.
In heterogeneous catalysis, reaction rate or catalytic activity is
generally governed by factors including numbers and abilities of
catalytic sites, temperature, concentrations of reactants, and reac-
tion orders. In this study, the catalytic activities of the Ag/SiO2 cat-
alysts with different Ag loadings apparently increased with the
increase of their O2 uptake (Table 4), implying close correlation
between the catalytic sites for crotonaldehyde hydrogenation and
the oxygen-bonding sites. O2 chemisorption measurement on the
deactivated 30Ag/SiO2 and 30Ag/5Zr–SiO2 catalysts further con-
firmed the presence of this correlation. Although the apparent Ag
particle size determined by XRD and TEM was almost unchanged
on both the deactivated 30Ag/SiO2 and 30Ag/5Zr–SiO2 catalysts,
their O2 uptake markedly declined, along with their catalytic activ-
ity. In Fig. 9, the catalytic activity as a function of the O2 uptake for
the fresh and deactivated Ag/SiO2 and Ag/Zr-SiO2 catalysts is plot-
ted to clearly display the non-negligible correlation of catalytic
sites to the chemisorption sites for oxygen.
demonstrated that the single-crystal surfaces of silver do not par-
ticipate in the chemoselective hydrogenation due to the extremely
low adsorption probability [24]. In our study, the pure Ag powders
mainly composed of Ag(1 1 1) and a small number of the corner
and edge sites were found to be completely inactive for producing
the desired crotyl alcohol and chemisorbing oxygen, even under
increased dosage. Conversely, the desired crotyl alcohol from the
monohydrogenation of the C@O bond appeared as the main pro-
duct on supported Ag catalysts. This significant support effect
strongly implies that the effective active sites for the C@O hydro-
genation are co-constructed by supports and Ag crystallites and
are located at the accessible interface. Generally, the proportion
of low-coordination sites (step, edge, and corner) increases with
the decrease of metal particle size [43]. In fact, our investigation
of the influence of Ag loading disclosed that those Ag/SiO2 catalysts
with low Ag loading and small Ag particles exhibited relatively
poor selectivity for the desired unsaturated alcohol. This result fur-
ther provides support for the viewpoint of perimeter interface.
Another important fact, that the zirconia doping on silica induced
a marked change in product distribution, stresses the crucial role
of zirconia species in influencing the chemoselective hydrogena-
tion of crotonaldehyde. This strong effect should not be a simple
size effect, but most probably is exerted by the interfacial zirconia
species modifying the adjacent Ag active sites and forming ternary
Ag–ZrO2–SiO2 interfaces. Second, the weak size dependence of
crotyl alcohol selectivity on Ag/SiO2 catalysts was disclosed in
literature and this study [18]. In contrast, the support effect was
found to be vital in the catalytic production of crotyl alcohol. If
the facet, edge, and corner Ag sites are responsible for chemoselec-
tive crotonaldehyde hydrogenation, the slight change in apparent
Ag particle size observed on the seriously deactivated Ag/SiO2
and Ag/Zr–SiO2 catalysts should be in contrast to their marked
deterioration in catalytic activity. However, the interfacial struc-
tures are not only influenced by the shape and size of metal parti-
cles, but also dependent on features including the support–metal
interaction, the fraction of interfacial atoms of metal crystallites,
and the thickness of metal layers. Conventional characterizations
by XRD and TEM are insensitive to these features. Third, our inves-
tigation reveals that the low-temperature O2 chemisorption capa-
bility clearly displays a close correlation with the crotonaldehyde
hydrogenation activity on supported Ag catalysts. The amounts
of Ag active for O2 chemisorption are much lower than that of
the surface Ag calculated from TEM and XRD characterizations,
but match well the perimeter interface Ag of hemispherical
particles. Below 323 K, the irreversible chemisorption of O2 only
occurred on Ag particles dispersed on the surface of SiO2 and
Zr–SiO2 supports, strongly indicating the unique redox properties
of interfacial metal. A previous study using X-ray absorption spec-
troscopy revealed that the metals at the interface were polarized
sufficiently via bonding with the oxygen of the support, and
hydrogen treatment on the supported metals induced a greater
metal–support oxygen distance [44]. This metastable interface
structure may be apt to bond oxygen while placed in an oxidizing
atmosphere.
With respect to the exact locations of catalytic sites for chemos-
elective hydrogenation of crotonaldehyde, Ag sites on the facets,
steps, edges, and corners of Ag crystallites and perimeter metal–
support interfaces are all possible candidates. Based on this study,
it seems more reasonable to identify the perimeter Ag–support
interface sites as the catalytic sites for chemoselective crotonalde-
hyde hydrogenation. First, previous theoretical study has
Tentatively, a mechanism for selective hydrogenation of croton-
aldehyde to crotyl alcohol at Ag–-SiO2 and Ag–ZrO2–SiO2 interface
is proposed in Scheme S1. Since the perimeter interface Ag atoms
inevitably interact with the oxygen from ZrO2 and SiO2, the
induced decrease in electron density generates a population of
interfacial Agd+ sites, which are partially electropositive. The Agd+
site and neighboring oxygen from oxide support can form a kind
of combinational site active for the polar adsorption and activation
of the C@O bond of crotonaldehyde. Simultaneously, dihydrogen is
dissociatively chemisorbed at stepped or cornered Ag sites with
low coordination number to form active hydrogen species. The
neighboring or spillover active hydrogen species can attack the
Fig. 9. Correlation of catalytic activity with O2 uptake on Ag/SiO2 and Ag/Zr–SiO2
with different Ag loadings and deactivation degrees.

H. Lin et al. / Journal of Catalysis 372 (2019) 19–32
31
activated crotonaldehyde at the interface to yield unsaturated
5. Conclusions
alcohol product.
The surface adsorption and activation of H2 on Ag catalysts is a
key step in the catalytic cycle of crotonaldehyde hydrogenation.
Our measurement of hydrogen adsorption on Ag powder and sup-
ported Ag catalysts reveals that no irreversible hydrogen adsorp-
tion can be measured due to the very weak interaction between
hydrogen and silver. An interesting kinetic study reveals that the
apparent reaction orders of hydrogen for the 30Ag/SiO2 and
30Ag/5Zr–SiO2 catalysts are all lower than 0.3. The estimated zero
reaction order with respect to hydrogen concentration seems to be
contradictory to the generally accepted view that hydrogen activa-
tion is crucial and hydrogen species are involved in the rate-
determining step in Ag-catalyzed hydrogenation due to the poor
hydrogen affinity of metallic Ag. Note that the measured hydrogen
reaction orders for the monohydrogenation of C@C and C@O bonds
are different. The presence of homolytic and heterolytic hydrogen
activation probably accounts for the difference. The hydrogenation
of the polarized C@O bond prefers the HÀ and H+ species from the
heterolytic activation, while the addition of the atomic hydrogen
from the homolytic dissociation of the C@C bond follows the Hor-
iuti–Polanyi mechanism. The zirconia doping on SiO2 brings about
an increase in the hydrogen reaction order for the formation of the
desired unsaturated alcohol and a simultaneous reduction for the
butanal formation, possibly implying that the involvement of zir-
conia species brings about different hydrogen activation and
changes the dominance of the homolytic and heterolytic hydrogen
activation patterns to some extent.
Over Ag/SiO2, the estimated reaction orders with respect to the
crotonaldehyde pressure are higher than the hydrogen reaction
orders, suggesting that the effective activation of crotonaldehyde
is the rate-determining step and the Eley–Rideal reaction mecha-
nism can be excluded. This kinetic behavior is different from
another highly selective catalyst, Ir/SiO2 (crotyl alcohol selectiv-
ity = 86%), on which the reaction order of crotonaldehyde is lower
than that of hydrogen [13]. Considering that pure Ag particles
without interfacial sites are completely inactive for the monohy-
drogenation of the C@O bond to produce unsaturated alcohol, it
is reasonable to indicate the vital role of perimeter interface Ag
sites in activating the C@O bond. In fact, the number of perimeter
interface Ag sites is sparse as compared with the surface Ag. There-
fore, the formation and supply of the crotonaldehyde adspecies
activated at the C@O bond and their subsequent reaction with
the adjacent or spillover active hydrogen species determine the
hydrogenation rate. Zirconia doping on SiO2 is found to reduce
the reaction orders with respect to the crotonaldehyde pressure,
indicating that the adsorption of crotonaldehyde on Ag/Zr–SiO2 is
stronger than that on Ag/SiO2. The interpretation is that the Lewis
acidic sites enhanced by zirconia doping provide more effective
sites for binding the polar C@O bonds.
In the gas-phase hydrogenation of crotonaldehyde, via zirconia
doping on the surface of silica, the catalytic behavior of supported
Ag catalysts can be significantly tuned to favor the production of
the desired crotyl alcohol from the monohydrogenation of the
C@O bond. High crotyl alcohol selectivity exceeding 80% and
improved catalytic activity were obtained on the optimized
30Ag/5Zr–SiO2 catalysts. Zirconia doping on the silica surface sub-
stantially enhances the number and strength of acid sites, pro-
motes the surface dispersion of Ag species, and stabilizes Ag
crystallites in smaller sizes. These promotional effects bring about
boosted catalytic activity and prolonged lifetime of catalysts.
The measurement of O2 chemisorption at 323 K disclosed that
pure Ag powders without interfacial sites did not irreversibly che-
misorb oxygen, but all supported Ag catalysts did, strongly indicat-
ing synergism between metal and oxide supports in constructing
the oxygen-bonding sites, which are most likely to locate in the
accessible Ag–support interface. The amounts of chemisorbed oxy-
gen on the Ag/SiO2 and Ag/Zr–SiO2 catalysts with different Ag dis-
persions and deactivation degrees were found to correlate closely
with the hydrogenation activity. This investigation strongly
implies a close correlation between the catalytic sites and the
oxygen-bonding sites. It is assumed that the accessible interface
Ag sites, rather than the facet, step, edge, and corner sites, are
responsible for the chemoselective hydrogenation of crotonalde-
hyde to crotyl alcohol. The catalytic production of the desired cro-
tyl alcohol with high selectivity may be attributed to the interface
Ag sites with unique redox properties and polarizations, which are
beneficial for chemisorbing and activating the polar C@O bonds.
Greatly improved crotyl alcohol selectivity on the Ag/Zr–SiO2 cata-
lysts probably originates from the formation of the ternary Ag–
ZrO2–SiO2 interfacial sites. Catalyst induction and deactivation
observed on the Ag/SiO2 and Ag/Zr–SiO2 catalysts indicate that
the Ag–support interfacial structures are changeable under real
catalytic operation.
Acknowledgments
We acknowledge financial support from the National Key R&D
Program of China (2017YFA0206801), the National Natural Science
Foundation of China (91545115 and 21473145), and the Program
for Innovative Research Team in Chinese Universities (IRT_14R31).
Appendix A. Supplementary material
Supplementary data to this article can be found online at
https://doi.org/10.1016/j.jcat.2019.02.004.
In summary, the accessible interface Ag sites rather than the
facet, edge, step, and corner Ag sites are assumed to play crucial
roles in the chemoselective hydrogenation of crotonaldehyde to
crotyl alcohol. The interfacial Ag sites polarized by oxide supports
act as catalytic sites for chemisorbing and activating the polar C@O
bonds and then drive catalytic hydrogenation to form the desired
crotyl alcohol as the main product. The Zr–SiO2 composite with
enhanced acidity not only leads to the improved dispersion of Ag
crystallites beneficial for promoting catalytic activity, but also
helps to construct unique ternary Ag–ZrO2–SiO2 interfacial struc-
tures with boosted capability for activating C@O bonds and pro-
ducing the desired unsaturated alcohol. This study of Ag-
References
[1] M.A. Vannice, B. Sen, J. Catal. 115 (1989) 65–78.
[2] A. Dandekar, M.A. Vannice, J. Catal. 183 (1999) 344–354.
[3] Salnikov, K.V. Kovtunov, D.A. Barskiy, A.K. Khudorozhkov, E.A. Inozemtseva, I.
P. Prosvirin, V.I. Bukhtiyarov, I.V. Koptyug, ACS Catal. 4 (2014) 2022–2028.
[4] E.V. Ramos-Fernández, J. Ruiz-Martínez, J.C. Serrano-Ruiz, J. Silvestre-Albero, A.
Sepúlveda-Escribano, F. Rodríguez-Reinoso, Appl. Catal. A 402 (2011) 50–58.
[5] P. Claus, Appl. Catal. A 291 (2005) 222–229.
[6] D. Wang, F. Ammari, R. Touroude, D.S. Su, R. Schlögl, Catal. Today 147 (2009)
224–230.
[7] B. Campo, María Volpe, S. Ivanova, R. Touroude, J. Catal. 242 (2006) 162–171.
[8] P. Reyes, M.C. Aguirre, I. Melián-Cabrera, M.L. Granados, J. Fierro, J. Catal. 208
(2002) 229–237.
[9] W. Grünert, A. Brückner, H. Hofmeister, P. Claus, J. Phys. Chem. B 108 (2004)
5709–5717.
[10] K. Brandt, M.E. Chiu, D.J. Watson, M.S. Tikhov, R.M. Lambert, J. Am. Chem. Soc.
131 (2009) 17286–17290.
[11] P. Chen, J. Lu, G. Xie, L. Zhu, M. Luo, React. Kinet. Mech. Cat. 106 (2012) 419–
434.
catalyzed chemoselective hydrogenation of
a,b-unsaturated alde-
hydes strongly indicates that the complicated support effect in
heterogeneous metal catalysis may originate from the unique
interfacial structure constructed by the metal and support.

32
H. Lin et al. / Journal of Catalysis 372 (2019) 19–32
[12] Q. Yu, X. Zhang, B. Li, J. Lu, G. Hu, A. Jia, C. Luo, Q. Hong, Y. Song, M. Luo, J. Mol.
Catal. A 392 (2014) 89–96.
[26] K. Lim, A.B. Mohammad, I.V. Yudanov, K.M. Neyman, M. Bron, P. Claus, N.
Rösch, J. Phys. Chem. C. 113 (2009) 13231–13240.
[13] M. Tamura, K. Tokonami, Y. Nakagawa, K. Tomishige, A.C.S. Sust, Chem. Eng. 5
(2017) 3685–3697.
[14] K. Yuan, T. Song, D. Wang, X. Zhang, X. Gao, Y. Zou, H. Dong, Z. Tang, W. Hu,
Angew. Chem. Int. Ed. 57 (2018) 5708–5713.
[15] H. Wei, C. Gomez, J. Liu, N. Guo, T. Wu, R. Lobo-Lapidus, C.L. Marshall, J.T.
Miller, R.J. Meyer, J. Catal. 298 (2013) 18–26.
[16] C.M. Engelhardt, R.M. Kennedy, J.A. Enterkin, K.R. Poeppelmeier, D.E. Ellis, C.L.
Marshall, P.C. Stair, ChemCatChem 10 (2018) 632–641.
[17] T. Mitsudome, M. Matoba, T. Mizugaki, K. Jitsukawa, K. Kaneda, Chem. Eur. J.
19 (2013) 5255–5258.
[18] P. Claus, H. Hofmeister, J. Phys. Chem. B 103 (1999) 2766–2775.
[19] M. Bron, D. Teschner, A. Knop-Gericke, B. Steinhauer, A. Scheybal, M. Hävecker,
D. Wang, R. Födisch, D. Hönicke, A. Wootsch, R. Schlögl, P. Claus, J. Catal. 234
(2005) 37–47.
[20] C.E. Volckmar, M. Bron, U. Bentrup, A. Martin, P. Claus, J. Catal. 261 (2009) 01–
08.
[27] P. Gallezot, D. Richard, Catal. Rev. Sci. Eng. 40 (1998) 81–126.
[28] L.J. Durndell, C.M.A. Parlett, N.S. Hondow, M.A. Isaacs, K. Wilson, A.F. Lee, Sci.
Rep. 5 (2015) 9425.
[29] H. Lin, J. Zheng, X. Zheng, Z. Gu, Y. Yuan, Y. Yang, J. Catal. 330 (2015) 135–144.
[30] P. Claus, Top. Catal. 5 (1998) 51–62.
[31] S. Bhogeswararao, D. Srinivas, J. Catal. 285 (2012) 31–40.
[32] Z. Dang, B.G. Anderson, Y. Amenomiya, B.A. Morrow, J. Phys. Chem. 99 (1995)
14437–14443.
[33] P. Claus, P. Kraak, R. Schodel, Stud. Surf. Sci. Catal. 108 (1997) 281–288.
[34] S. Fujii, N. Osaka, M. Akita, K. Itoh, J. Phys. Chem. 99 (1995) 6994–7001.
[35] S.R. Seyedmonir, D.E. Strohmayer, G.L. Geoffroy, M.A. Vannice, H.W. Young, J.
W. Linowski, J. Catal. 87 (1984) 424–436.
[36] M. Haruta, Faraday Discuss. 152 (2011) 11–22.
[37] C.D. Wagner, Faraday Discuss. Chem. Soc. 160 (1975) 291–300.
[38] L.H. Tjeng, M.B.J. Meinders, J. van Elp, J. Ghijsen, G. Sawatzsky, R.L. Johnson,
Phys. Rev. B. 41 (1990) 3190–3199.
[21] M. Abid, V. Paul-Boncour, R. Touroude, Appl. Catal. A 297 (2006) 48–59.
[22] G. Kennedy, L.R. Baker, G.A. Somorjai, Angew. Chem. Int. Ed. 53 (2014) 3405–
3408.
[39] M. Kosmulski, Chemical properties of material surfaces, Dekker, New York,
2001.
[40] C.T. Campbell, Surf. Sci. 157 (1985) 43–60.
[23] M. Bron, D. Teschner, A. Knop-Gericke, F. Jentoft, J. Kröhnert, J. Hohmeyer, C.
Volckmar, B. Steinhauer, R. Schlöegl, P. Claus, Phys. Chem. Chem. Phys. 9
(2007) 3559–3569.
[41] M. Mavrikakis, B. Hammer, J. Nørskov, Phys. Rev. Lett. (1998) 2819–2822.
[42] D. Su, T. Jacob, T. Hansen, D. Wang, R. Schlögl, B. Freitag, S. Kujawa, Angew.
Chem. Int. Edit. 47 (2008) 5005–5008.
[24] K. Wang, B. Yang, Catal. Sci. Technol. 7 (2017) 4024–4033.
[25] K. Lim, Z. Chen, K.M. Neyman, N. Rösch, Chem. Phys. Lett. 420 (2006) 60–64.
[43] V. Hardeveld, F. Hartog, Surf. Sci. 15 (1969) 189–230.
[44] D.C. Koningsberger, B.C. Gates, Catal. Lett. 14 (1992) 271–277.