Active site structure of a lithium phosphate catalyst for the isomerization of 2,3-epoxybutane to 3-buten-2-ol

Article abstract of DOI:10.1016/j.mcat.2017.11.018

Basic lithium phosphate (B-LPO) catalyst selectively produces unsaturated alcohols from epoxides. The catalytic activity of B-LPO is known to originate from appropriate acidic-basic properties, but no details were available on the structure of the active site. In this study, experimental methods and DFT calculations were performed in an attempt to identify the active surface structure of B-LPO for the isomerization of 2,3-epoxybutane to 3-buten-2-ol. The experimental results showed that exchanged Na ions in B-LPO suppressed the formation of an acid-catalyzed by-product (methyl ethyl ketone). In addition, H₂O had a negative effect on the formation of 3-buten-2-ol due to the preoccupation of the active site. DFT calculations in conjunction with these experimental observations showed that the most plausible active surface for the formation of 3-buten-2-ol is the (001) surface of LPO whose acidic proton is exchanged with Na atom. On this surface, the under-coordinated Li atoms and the surface P[dbnd]O groups are exposed, and these play a role in activating the C–O bond of an epoxide ring, and in receiving a proton from the terminal carbon, respectively.

Full text of DOI:10.1016/j.mcat.2017.11.018

Molecular Catalysis 445 (2018) 133–141
Contents lists available at ScienceDirect
Molecular Catalysis
journal homepage: www.elsevier.com/locate/mcat
Research Paper
Active site structure of a lithium phosphate catalyst for the
isomerization of 2,3-epoxybutane to 3-buten-2-ol
a
a
a
a
b
Tae Yong Kim , Chyan Kyung Song , Yang Sik Yun , Danim Yun , Jeong Woo Han ,
Jongheop Yi
a
a,∗
World Class University Program of Chemical Convergence for Energy & Environment, Institute of Chemical Processes School of Chemical and Biological
Engineering, Seoul National University, Seoul 151-742, Republic of Korea
b
Department of Chemical Engineering, University of Seoul, Seoul 130-743, Republic of Korea
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 15 September 2017
Received in revised form
Basic lithium phosphate (B-LPO) catalyst selectively produces unsaturated alcohols from epoxides. The
catalytic activity of B-LPO is known to originate from appropriate acidic-basic properties, but no details
were available on the structure of the active site. In this study, experimental methods and DFT calculations
were performed in an attempt to identify the active surface structure of B-LPO for the isomerization
of 2,3-epoxybutane to 3-buten-2-ol. The experimental results showed that exchanged Na ions in B-
LPO suppressed the formation of an acid-catalyzed by-product (methyl ethyl ketone). In addition, H2O
had a negative effect on the formation of 3-buten-2-ol due to the preoccupation of the active site. DFT
calculations in conjunction with these experimental observations showed that the most plausible active
surface for the formation of 3-buten-2-ol is the (001) surface of LPO whose acidic proton is exchanged
with Na atom. On this surface, the under-coordinated Li atoms and the surface P O groups are exposed,
and these play a role in activating the C O bond of an epoxide ring, and in receiving a proton from the
terminal carbon, respectively.
1
0 November 2017
Accepted 11 November 2017
Keywords:
Epoxide
Unsaturated alcohol
Isomerization
Heterogeneous catalysis
Density functional theory
©
2017 Elsevier B.V. All rights reserved.
1
. Introduction
best catalyst that showed 88% yield of 1,3-butadiene. They also
found that BO was an intermediate chemical during the forma-
tion of 1,3-butadiene. The same group also reported that ZrO2
based catalyst were effective to the production of BO from 2,3-
butanediol [4,5]. Filimonov et al. also reported the direct formation
of 1,3-butadiene using silica-supported sodium phosphates [8].
They observed that contact time and Na/P ratio considerably influ-
enced product distribution. As similar to the report by Sato et al., BO
was detected as an intermediate for the formation of 1,3-butadiene.
Dehydrative epoxidation of 2,3-butanediol using Cs/SiO2 cat-
alyst can be another alternative way for the production of
1,3-butadiene [6]. This reaction produces EB from 2,3-butanediol,
and the product can be isomerized into BO. There are several
reported catalysts for the selective isomerization of epoxides into
With depletion of fossil resources and a growing demand
for environmentally benign production technologies, the cat-
alytic conversion of biomass is expected to replace conventional
petrochemical processes [1]. 2,3-butanediol is one of important
biomass-derived feedstocks, and is a potential source for pro-
duction of valuable C4 chemicals [2]. Catalytic dehydration of
2
,3-butanediol is an effective way in that various chemicals such as
methyl ethyl ketone [3], 3-buten-2-ol (BO) [4,5], 2,3-epoxybutane
EB) [6], and 1,3-butadiene [7,8] can be produced. Among these
(
products, 1,3-butadiene is practically important chemical in the
synthesis of polymers, including ABS, NBR, and SBR polymers. In
this regard, production of 1,3-butadiene from 2,3-butanediol has
been extensively studied [7,8].
unsaturated alcohols, such as Au/TiO (liquid-phase isomerization)
2
Sato et al. reported the direct production of 1,3-butadiene from
[9,10] and lithium phosphates (gas-phase isomerization) [11–13].
In the isomerization of epoxides, it was commonly reported that
the acidic-basic properties of a catalyst determine catalytic perfor-
mance [9–14].
2
,3-butanediol using rare-earth metal oxides [7]. Sc O3 was the
2
Teranishi et al. studied the effects of acidic-basic properties on
the isomerization of propylene oxide [11,14]. When they utilized
acidic zeolites, only carbonyl compounds including acetone and
propionaldehyde were produced [14]. They also found that the
Abbreviations: B-LPO, basic lithium phosphate; N-LPO, neutral lithium phos-
phate; EB, trans-2,3-epoxybutane; BO, 3-buten-2-ol.
∗ Corresponding author.
E-mail address: jyi@snu.ac.kr (J. Yi).
http://dx.doi.org/10.1016/j.mcat.2017.11.018
2
468-8231/© 2017 Elsevier B.V. All rights reserved.

1
34
T.Y. Kim et al. / Molecular Catalysis 445 (2018) 133–141
ratio of acetone to propionaldehyde was proportional to the acidic
strength of a catalyst.
ing speed) for 6 h. The resultant powder was ground and sieved to
retain aggregates below 200 ␮m prior to a use in a catalytic reaction.
The preparation of neutral lithium phosphate (N-LPO) followed
a procedure similar to that of B-LPO, except for the source of phos-
phate. First, 1.28 g of lithium hydroxide monohydrate (0.030 mol,
In contrast, lithium phosphate (Li PO , hereafter LPO)-based
3
4
catalysts were capable of producing allyl alcohol [11]. Stoichio-
metric LPO exhibited rather poor selectivity toward allyl alcohol
(
17%), but, when LPO was synthesized with the addition of other
LiOH·H O, Sigma-Aldrich) was dissolved in 10 ml of deionized
2
◦
alkali-metals (Na or K), it showed higher selectivity (∼86%) as well
water, and then the solution was heated to 60 C with stirring. The
as higher activity. In particular, Li NaPO , so-called basic lithium
solution was neutralized via the drop-wise injection of 0.7 M H PO4
2
4
3
phosphate (B-LPO), showed the best performance. The LPO cata-
lysts that contained alkali metals possessed stronger basic sites
and weaker acidic sites than the stoichiometric LPO, which indi-
cated that basic sites are essential for the formation of allyl alcohol.
The authors also performed reaction tests by titrating acidic or basic
sites with pyridine or dichloroacetic acid, respectively, and thereby
found that both acidity and basicity were required for the formation
of allyl alcohol.
aqueous solution with vigorous stirring. The neutralized solution
was aged for 24 h at 60 C. The remaining procedures were iden-
tical to those for B-LPO preparation: filtration, washing, drying,
calcination, and grinding.
◦
2.2. Catalytic activity tests
Catalytic reactions for the isomerization of trans-2,3-
Zhong et al. reported a similar tendency whereby basic LPO
showed a catalytic performance that was higher than that of stoi-
chiometric LPO [12]. They also studied the effect of precursors for
lithium, sodium, and phosphate on the morphology and catalytic
properties of basic LPO [13]. Lithium hydroxide and sodium phos-
phate were more appropriate precursors, as the synthesized LPO
showed higher performance thanks to a crystallinity and a surface
area that was better and higher, respectively, than that of the others
tested.
Previous studies helped explain the isomerization of epoxides in
terms of acid-base catalysis. However, no studies regarding surface
structures and the roles of the surface moieties have been per-
formed, which limits a fundamental understanding of the catalysis
in the reaction. The identification of active surface structures would
be helpful to design improved catalyst for this reaction.
Herein, we rationally identified the active surface of an LPO cat-
alyst for the isomerization of EB to BO, and proposed the roles of the
surface moieties during catalysis. Basic and neutral lithium phos-
phates (B-LPO and N-LPO, respectively) were compared to identify
characteristics of the active catalytic materials. We performed
several experiments including catalytic activity tests, structural
characterization, and acidic-basic characterization. Based on the
experimental evidences, theoretical DFT calculations were carried
out to identify the active surface of LPO. The presence of the iden-
tified active surface was also confirmed via FT-IR analyses. Finally,
the roles of the surface moieties on the active surface structure were
proposed, based on the optimized geometry for adsorption of EB.
These results will be helpful in the design and synthesis of more
active catalysts for these types of reactions.
epoxybutane (hereafter EB, purchased from Alfa Aesar (97%))
were performed on a fixed-bed quartz reactor with a thermocou-
ple well. A K-type thermocouple was placed on the thermocouple
well to measure the reaction temperature, and the reactor was
externally heated using an electric furnace. A varied amount of
catalyst (3–50 mg) was used in the catalytic reaction to main-
tain a conversion level lower than 20%. The reactor was heated
◦
to the desired temperatures with a ramping rate of 10 C/min
and was maintained for 30 min with a flow of dry N (99.999%,
2
−
1
3
0 cm3 min ). An aqueous solution of the reactant was then
injected using a liquid syringe pump (Cole-Palmer 74900 series)
◦
into a pre-heating zone, which was maintained at 200 C. The
values for the concentration of the reactant solution and dry
N2 flow rate were varied to control the partial pressure of H O
2
(
0–62 kPa) while maintaining other conditions in a constant
state: total flow rate (51 sccm) and the partial pressure of EB
0.9 kPa). In the reaction condition, there were no mass transfer
(
limitations (Mears criterion for external mass transfer limitation,
and Weisz-Prater criterion for internal mass transfer limitation;
see Supporting information). Product gases were cooled using
a helix-type condenser, and liquefied contents were collected
hourly in a sample tube containing deionized water. Acetonitrile
was used as an external standard for quantification. The products
were analyzed using gas chromatography (Younglin ACME 6100
®
instrument) equipped with a FID detector and an Rtx -VRX capil-
lary column (Restek, cat. # 19316). The data acquired after 2 h of
reaction time were used to compare the activity of the catalysts.
2.3. Characterization
High-resolution transmission electron micrograph (HR-TEM)
2
. Experimental and theoretical methods
images were obtained using a JEOL JEM-3010 microscope with an
acceleration voltage of 300 kV. The X-ray diffraction (XRD) pat-
terns were obtained using a Rigaku D-MAX2500-PC powder X-ray
diffractometer with Cu K␣ radiation (1.5406 Å) in an operating
mode of 50 kV and 100 mA. The crystallite sizes were calculated
based on peaks at 16.8 ((010) plane) and 36.9 ((002) plane) using
the Scherrer equation. N2 adsorption-desorption analyses were
carried out using a Micromeritics ASAP-2010 instrument. The spe-
cific BET surface area was calculated at P/Po = 0.1–0.2.
2.1. Preparation of catalysts
Basic lithium phosphate (B-LPO) was prepared via precipi-
◦
◦
tation. First, 1.82 g of sodium phosphate monobasic (0.015 mol,
NaH PO , Sigma-Aldrich) and 1.28 g of lithium hydroxide mono-
2
4
hydrate (0.030 mol, LiOH·H O, Sigma-Aldrich) were separately
2
dissolved in 15 ml and 10 ml of deionized water, respectively. The
former solution (NaH PO ) was transferred to a 100 ml round-
X-ray photoelectron spectroscopy (XPS) was performed on a
KRATOS AXIS electron spectrometer with Mg K␣ radiation. The
binding energies were corrected using C 1s as an internal standard
(284.5 eV). The peaks were fitted by mixed Gaussian-Lorentzian
functions (10% Lorentzian) with subtraction of the Shirley-type
background using a XPS peak-fitting program (XPSPEAK 4.1). The
2
4
◦
bottom flask, and was then heated to 40 C with stirring. The
aqueous solution of LiOH·H O was then added drop-wise with vig-
2
orous stirring. A white precipitate was immediately formed when
the solution was added. After complete injection of the solution,
◦
the mixed solution was aged for 24 h at 40 C. The white precip-
7
31
itates were isolated on a filter, washed three times with 200 ml
of de-ionized water, and dried overnight at 343 K. Calcination of
the dried precipitates was conducted at 673 K (5 K/min of ramp-
Li and P magic angle spinning (MAS) nuclear magnetic resonance
(NMR) spectra of the samples were recorded on a Bruker AVANCE
400 WB (400 MHz) spectrometer operated at frequencies of 156

T.Y. Kim et al. / Molecular Catalysis 445 (2018) 133–141
135
Fig. 1. TEM images for B-LPO (A, C) and N-LPO (B), and XRD patterns (D) for B-LPO and N-LPO.
and 162 MHz, respectively. A magic angle spin speed of 7 kHz was
used for the spectral recording.
free energy, identical bottom layers were used for terminations per-
pendicular to the same plane. The slab thicknesses and number of
atoms for the slab models are provided in Table S2. Vacuum gaps
thicker than 15 Å were included, to prevent non-physical electronic
interactions. More detailed description for the calculation of surface
free energies is provided in Supporting information.
2.4. Computational details
Periodic density functional theory (DFT) calculations were per-
The charge density difference for EB adsorption was obtained by
subtracting the charge densities of isolated EB and the bare surface
from that of the adsorbed geometry (VESTA ver. 3.3.1). A charge
formed using the Vienna ab initio simulation package (VASP) [15].
We used a generalized gradient approximation (GGA) parameter-
ized by the Perdew–Burke–Ernzerhof (PBE) exchange-correlation
functional [16]. Ionic cores were described using the projector aug-
mented wave (PAW) method [17]. Dispersion forces were included
with the DFT-D2 Grimme’s empirical correction [18]. The wave
functions were represented by the expansion of plane waves with
an energy cutoff of 400 eV. A 1 × 1 × 1 Monkhorst-Pack k-point
mesh was used to sample the Brillioun zone. For bulk optimiza-
tion, the energy cutoff and k-point mesh were increased to 520 eV
and 6 × 6 × 6, respectively. All calculations were converged until
the forces on all atoms were less than 0.03 eV/Å. The electronic
optimization steps were converged self-consistently to fewer than
3
density of 0.0025 electrons/Å was used to represent the isosurface.
3. Results and discussion
3.1. Physicochemical properties
The crystallographic and textural properties were characterized
by TEM, XRD (Fig. 1), NMR (Fig. S2), and N2 adsorption-desorption
analyses. The specific BET surface areas for B-LPO and N-LPO were
2
23.5 and 3.4 m /g, respectively. These surface areas are similar to
−
5
2
× 10 eV.
the values in the previous study by Teranishi et al. [11].
An experimentally identified LPO structure with the Pmn21
In TEM images of both catalysts (Fig. 1A and B), agglomerates of
rod-shaped crystallites were observed. The crystallites were some-
what inhomogeneous, whose sizes ranges from tens of nanometers
to several hundred nanometers. An HR-TEM image of B-LPO clearly
space group [19] was used as an initial structure for the bulk
optimization. The geometry and structural parameters for the opti-
mized bulk are provided in the Supporting information (Fig. S1 and
Table S1) The optimized lattice constants were 6.16 Å (a), 5.28 Å
shows the lattice of the (010) plane of Li PO4 crystallite (space
3
(
b), 4.90 Å (c), which was in good agreement with the experimental
group Pmn21 [19]), and the (001) plane can also be recognized in
its FFT pattern (Fig. 1C and inset image). The (001) and (010) planes
constitute the longitudinal and transverse directions, respectively,
of the LPO crystallite.
values of 6.11 Å (a), 5.23 Å (b), 4.85 Å (c). The optimized parameters
were used to construct asymmetric slab models for (001), (010),
and (100) planes of LPO. The surface unit cells were obtained by
cleaving all possible terminations perpendicular to these planes,
and the p(2 × 2) supercells of the surface unit cells were used. All
slab models consisted of at least 4 atomic layers. The bottom two
layers were fixed while the remaining top layers were allowed to
relax. In order to compare the relative stability in terms of surface
The XRD patterns for the two catalysts matched well with pat-
terns of the Li PO4 with the Pmn21 space group (PDF#25-1030,
3
Fig. 1D). Neither peak shifts nor additional peaks were observed
for both catalysts. Crystallite sizes for B-LPO was smaller than that
for N-LPO, as evidenced by broader peaks. The calculated crystallite

1
36
T.Y. Kim et al. / Molecular Catalysis 445 (2018) 133–141
Fig. 2. Li 1s and P 2p XPS spectra for B-LPO (A, B) and N-LPO (C, D). Inset shows Na 1s spectrum of B-LPO. Raw data (open circle), fitted peak (red or blue line), and background
black line). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
(
Table 1
XPS results for B-LPO and N-LPO.
B-LPO
N-LPO
Binding energy (eV)
Atomic ratio
Li 1s
P 2p3/2
O 1s
54.3
54.4
133.5
531.1
–
133.4
531.1
1072.1
Na 1s
Li/P
O/P
Na/P
2.9
3.1
0.1
2.8
3.1
–
Scheme 1. Isomerization of EB to BO and MEK.
3.2. Reaction tests
◦
sizes based on (010) peak (16.8 ) are 25 and 37 nm, respectively for
B-LPO and N-LPO. When the size is calculated using the (002) peak
Reaction of EB over LPO catalysts produced two isomerization
products, i.e. BO and MEK. BO is the product when C C bond is
◦
(
36.9 ), it gives larger values for crystallite size (35 nm for B-LPO
and 40 nm for N-LPO) than those from (010) peak. It also implies
formed after C O bond cleavage, whereas MEK is produced via C
cleavage followed by C O formation (Scheme 1).
O
7
31
rod-like shape of LPO crystallites. The spectra of Li and P NMR
for the two catalysts were almost identical (Fig. S2). This indicates
that the local environment near Li and P atoms were quite simi-
lar for the two catalysts. The peak in NMR spectra of B-LPO was
slightly broader than that of N-LPO, which could be caused from
lower crystallinity of B-LPO.
Catalytic performance of two catalysts for isomerization of EB
is compared in Fig. 3. Reaction without a catalyst showed quite
low activity (conversion 3%, BO selectivity 21%, and MEK selectiv-
ity 68%). Compared to N-LPO, B-LPO showed a 6-fold higher specific
activity (49.3 vs. 8.3 mmolEB/gcat h), and even higher BO forma-
tion rate (46.2 vs. 4.4 mmolBO/gcat h) due to its higher selectivity
toward BO (94 vs. 53%). The higher specific activity of B-LPO could
be attributed to its higher surface area. This is supported by similar
values for the ratio of specific activity (B-LPO/N-LPO = 5.9) and the
ratio of surface area (B-LPO/N-LPO = 6.9).
On the other hand, selectivity is more closely related to acidic-
basic properties of a catalyst. It was reported that both acidic and
basic sites are required for the formation of unsaturated alcohols,
while strong acidic sites facilitate the formation of carbonyl com-
pounds [11,12]. It was also known that B-LPO contained weaker
acidic sites than N-LPO, resulting in lower selectivity toward car-
bonyl compounds [11]. The weaker acidic strength of B-LPO can
be related to remaining Na ions in B-LPO. In the characterization
results, we found that the Na atoms in B-LPO were located on the
surface, and the chemical state of the Na atoms was similar to that of
The atomic composition and chemical states of the elements
were analyzed via XPS (Table 1, Fig. 2). For both catalysts, the atomic
ratio of Li to P approximated a stoichiometric value of 3 (2.9 for
B-LPO and 2.8 for N-LPO). The measured oxygen content was some-
what lower than a stoichiometric value of 4 (3.1 for both catalysts)
due to the lower inelastic mean free path of photoelectrons from
the O 1s orbitals [20,21]. A small quantity of sodium (Na/P = 0.1) was
also detected for B-LPO. The binding energies (BEs) for Li 1s were
5
2
4.3 and 54.4 eV for B-LPO and N-LPO, respectively. The BEs for P
p3/2 were 133.4 and 133.5 eV for B-LPO and N-LPO, respectively.
These values are within the previously reported values of lithium
contained oxides and phosphates [22]. For B-LPO, the BE of Na 1s
was 1072.1 eV, which is higher than the reported value for Na PO
3
4
(
1070.9–1071.3 eV) [22–24]. This value more closely approximates
that for Na exchanged zeolites [25].

T.Y. Kim et al. / Molecular Catalysis 445 (2018) 133–141
137
bond. The bond length between P atom and O atom in P OH group
is 1.61 Å.
For the (010) surface, a geometry that exposes three-fold coordi-
nated Li and O atoms was calculated to be the most stable (Fig. 5B).
No distinct changes in the atomic positions were observed upon
surface relaxation for this surface. Further surface oxidation or
hydroxylation was an unfavorable process.
In contrast to the aforementioned planes, termination of the
(
100) plane exposes unstable polar surface due to an alternate
2−
2+
stacking of the [Li O ]
and [LiPO2] layers (Fig. S5). Surface
2
2
hydroxylation has been reported to stabilize the instability from
the polar stacking [26]. For the most stable (100) surface (Fig. 5C),
2
the hydroxyl density is 3.9 OH/nm , and three-fold coordinated Li
and O atoms are present.
Fig. 3. Catalytic reaction results for the conversion of EB over B-LPO and N-LPO.
Reaction conditions: T = 350 C, 0.05 g catalyst, 0.15 ml/h EB (1.35 ml/h H2O). Con-
version of EB for N-LPO and B-LPO was 25 and 60%, respectively.
◦
Considering the structural differences among the three surfaces,
the P-OH group on the (001) surface is important to note, as this
moiety can act as an acid site. The (100) surface also contains pro-
tons, but these protons are rather basic because they are attached
Na in Na-exchanged zeolites. In the reaction results, B-LPO showed
much less formation of MEK that is the product via acid catalyzed
reaction. Therefore, it is reasonable to conclude that the role of Na
atoms in B-LPO is eliminating Brønsted acidic site.
to O . Therefore, P OH group on (001) surface is plausible active
Li
site for the formation of MEK.
As we observed in characterization results, Na ions were dis-
persed on the surface of B-LPO. Therefore, plausible exchange site
for Na atom was also explored on (001) and (100) surfaces, and the
results are shown in Table 2 and Fig. 6. Exchange energies for (001)
and (100) surfaces are compared in Table 2. The negative value
of exchange energy for (001) surface indicates that Na-exchange
on (001) surface is energetically favored. In contrast, considerably
higher exchange energy for (100) surface means that Na-exchange
on this surface is unfavorable process. Therefore, probable location
of Na ions in B-LPO is on (001) surface. These results also sup-
port that the P OH group on (001) surface is the active site for
the formation of MEK.
Using B-LPO, we further explored the effect of H O on catalytic
2
activity. Fig. 4A shows the variations in the specific BO forma-
tion rate as a function of H O partial pressure. The presence of
2
H O considerably suppressed the specific formation rate of BO.
2
When the H O partial pressure was gradually increased from 0
2
to 62 kPa, the specific formation rate of BO was decreased from
2
0.4 to 11.2 mmolBO/gcat h, while the selectivity values were merely
influenced.
When we compared the apparent activation energy of BO forma-
tion with and without H O flow (Fig. 4B), almost identical values
2
were found in both cases. The apparent activation energies were
estimated to be 102 and 108 kJ/mol, respectively for 0 and 40 kPa of
H O partial pressure. Since H O is not involved in the isomerization
3.3.2. Adsorption of H2O on the stable surface geometries
2
2
of EB to BO, and the presence of H O did not alter the selectivity val-
We found the most stable surface geometries for three planes
considering surface oxidation, hydroxylation, and Na-exchange. On
2
ues and the activation energies, the effect of H O cannot be caused
2
by a change of neither the reaction mechanism nor the active site
structure. A preoccupation (or poisoning) of the active site by H O
2
is a more reasonable cause for the adverse effect of H O. Therefore,
2
an active site for the formation of BO strongly binds to H O.
2
Stability of B-LPO catalyst was also tested, and no deactivation
was observed for 10 h of reaction (Fig. S3).
3.3. Identification of the active surface
Through experimental analyses, we found that acid sites over
LPO surface were eliminated by Na atoms (B-LPO), and H O is
2
strongly bound by an active site for BO formation. With these
results, DFT calculations were performed to find an active surface
for isomerization into BO.
3.3.1. Stable geometries for B-LPO surface
Various geometries for (001), (010), and (100) surfaces were
explored with consideration for surface oxidation and hydroxyla-
tion. Fig. 5 shows the most stable surface geometries at a relevant
range of temperature (300–1000 K). Details in finding the ther-
modynamically stable geometries are discussed in the Supporting
information (Figs. S4–S11).
The most stable (001) surface exposes under-coordinated
(
three-fold coordination) Li atoms, apical O atoms (P O), and
hydroxyl group (P OH) (Fig. 5A). Surface hydroxyl density for the
2
most stable geometry is 0.77 OH/nm . The bonding between the
apical O atom and the P atom is contracted to some extent com-
pared with the bulk P O bond length (1.49 Å compared with 1.56 Å
of bulk LPO), which is attributed to the formation of a P O double
Fig. 4. The specific BO formation rate as a function of H2O partial pressure at 598 K
(
A) and the Arrhenius plot under H2O partial of 0 kPa and 40 kPa (B). Products other
than BO and MEK were not detected.

1
38
T.Y. Kim et al. / Molecular Catalysis 445 (2018) 133–141
Table 3
Vibrational frequencies of (001) surface and H2O adsorbed surfaces.
Model
001)Na
Wavenumber (cm−1)
Assignment
(
1144
ꢀ(P O)
ꢀ(PO4)
977
H2O/(001)Na
3536–3689
ꢀas(HOH)
ꢀs(HOH)
␦(HOH)
3168–3186
1336–1570
1115–1122
ꢀ(P O)HB
Adsorption of H O on (100) surface involves chemisorption on
2
under-coordinated Li atom and two hydrogen bonding with sur-
face hydroxyl group and under-coordinated O atom (Fig. 7C). The
adsorption strength is −83 kJ/mol.
The values of the free energy of H O adsorption are compared in
2
Fig. 7D. Partial pressure of H O (p
= 20–62 kPa) was also consid-
2
H2O
ered, and the reaction temperature in Fig. 4A (598 K) is indicated by
blue line. It clearly shows that the adsorption of H O on the (001)Na
2
surface is much stronger than that on (010) and (100) surfaces. Fur-
thermore, the values for free energy of adsorption for (010) and
(100) surface are ranging from 26 to 31 kJ/mol, which are too weak
to affect the catalytic activity. Consequently, the (001)Na surface is
the most plausible active surface for the isomerization of EB to BO
among the explored surfaces.
3.4. Vibrational frequencies and FTIR analysis
In order to confirm the presence of the surface moieties, vibra-
tional frequencies for the surfaces models were calculated, and
experimental FTIR analyses were carried out.
Table 3 summarizes the calculated characteristic frequencies for
(
001)Na as well as the H O adsorbed geometries. For a bare surface,
2
−
1
two frequencies, 1144 and 977 cm , appeared from the stretching
modes of the P O bond and the PO4 group, respectively.
Fig. 5. The most stable surface geometries for (001), (010), and (100) planes. Oxygen
red), lithium (green), phosphorus (grey), hydrogen (white), top layer (Ball-and-
stick), second layer (stick). Under-coordinated sites were indicated by arrows. (For
interpretation of the references to colour in this figure legend, the reader is referred
to the web version of this article.)
(
When H O is adsorbed onto (001)Na, a considerable variation in
frequency for the P O bond stretching mode occurs. The formation
2
Table 2
Vibrational frequencies of (001) surface and H2O adsorbed surfaces.
Model
No. of exchanged H
1
Exchange energy (eV)a
(
001)
100)
−0.56
(
1
2
3
4
1.16
2.48
5.57
10.04
a
Exchange energy is calculated by the following equation: Eexc = ENa−surface +
n
2
n
EH2O − Esurface − ENa O, where ENa−surface is total energy of Na-exchanged surface,
2
2
Esurface is total energy of bare surface, EH2O is total energy of H2O gas, ENa2O is bulk
energy of Na2O per O atoms, and n is number of exchange protons.
these surface models, the adsorption strength of H O was com-
2
pared to find the plausible active surface for the isomerization of EB
to BO. The most strongly adsorbed geometries are shown in Fig. 7.
On the (001)Na surface, the under-coordinated Li sites were
the preferential adsorption sites (Fig. 7A). H O is adsorbed on the
2
bridge position between two under-coordinated Li atoms, and two
hydrogen bonds are formed with the nearest two P O groups. The
adsorption strength is −105 kJ/mol.
With respect to the (010) surface (Fig. 7B), hydrogen bonding
is the more important driving force for adsorption rather than
interaction with the under-coordinated Li site. In the most stable
adsorbed geometry, no chemisorption occurs, but two hydrogen
bonds give rise to −81 kJ/mol of an adsorption strength.
Fig. 6. The optimized geometries for Na-exchanged (001) and (100) surfaces.
Sodium (yellow), oxygen (red), lithium (green), phosphorus (grey), hydrogen
(
white), top layer (Ball-and-stick), second layer (stick). Under-coordinated sites
were indicated by arrows. (For interpretation of the references to colour in this
figure legend, the reader is referred to the web version of this article.)

T.Y. Kim et al. / Molecular Catalysis 445 (2018) 133–141
139
Fig. 7. The optimized geometries for H2O adsorption on (001)Na (A), (010) (B), and (100) surface (C), and free energy of H2O adsorption as a function of temperature (D).
gave rise to the absorption near 1650 cm 1. In addition, the peaks
−
of hydrogen bonding between P O and H O results in an elongation
2
−1
of the P O bond (1.49 Å → 1.52 Å), and leads to a decrease in the
at the lower frequencies (1500–1350 cm ) can be assigned to the
−
1
ꢀ
(P O) frequencies (1144 → 1115–1122 cm ).
perturbed bending modes of H O molecules. This frequency range
2
Fig. 8A shows the experimental FTIR spectrum for the B-LPO cat-
alyst. The broad peak at 3500–3000 cm can be attributed to the
symmetric and asymmetric stretching modes of highly hydrogen-
is similar to the bending frequency of H O in geometry B. Small
2
−
1
−1
peaks were observed at 1250 and 1228 cm , verifying the pres-
ence of a P O double bond [28]. The bulk phonon modes of the PO
group in LPO gave rise to the strong absorption at 1010 cm [28]
(Fig. S12), which made it difficult to identify the calculated absorp-
tion corresponding to the P O double bond that forms hydrogen
4
−
1
bonded H O molecules and hydroxyl groups on a surface [19]. The
2
−1
small and complex peaks at 3600–3750 cm can be attributed to
the stretching vibrations of surface hydroxyl groups that do not
−
1
form hydrogen bonds [27]. The bending modes of physisorbed H O
bonding (1115–1122 cm ).
2
Fig. 8. FTIR spectrum for B-LPO catalyst (A) and variations in the spectrum with increases in temperature (B, C); C bottom shows the changed spectrum after the cooling to
room temperature.

1
40
T.Y. Kim et al. / Molecular Catalysis 445 (2018) 133–141
Fig. 9. Geometry for the adsorption of EB on (001)Na (A, B) and the contour for difference in electron density (C, D). Gain and loss of electron are colored by blue and red,
respectively. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
As the temperature was ramped up, absorptions at 1220 and
260 cm gradually increased (Fig. 8C top) while the absorption
the terminal carbon of EB. This presumably receives a proton from
the EB molecule for the formation of a C C double bond, and then
transfer the proton to the epoxy oxygen.
−
1
1
related to adsorbed H O decreased (Fig. 8B). For a clear observation,
2
the sample was first cooled to room temperature to exclude the
thermal contribution in IR absorption, and then the spectrum was
acquired (Fig. 8C bottom). The spectrum clearly showed increased
levels of absorption at 1208 and 1260 cm . The increased absorp-
tion indicated that the surface model (001)Na (or at least a similar
structure) existed in the B-LPO catalyst (Table 3).
4
. Conclusions
−
1
The active surface of the basic lithium phosphate (B-LPO) cata-
lyst for the isomerization of trans-2,3-epoxybutane to 3-buten-2-ol
was identified based on experimental and theoretical results. The
experimental characterization and the reaction results indicated
3.5. The role of surface moieties
that the active surface strongly binds H O molecules, and dispersed
2
Na ions on B-LPO surface suppress the formation of acid sites. Based
on the DFT results, the (001) surface whose acidic proton (P OH) is
exchanged with Na atom was proposed as plausible active surface
for the reaction. The presence of the (001) surface was confirmed
via FT-IR analysis. The roles of the surface moieties on the (001)
surface were proposed on the basis of the adsorbed geometry of
trans-2,3-epoxybutane and according to the charge density con-
tours. The under-coordinated Li sites activate the C O bond of an
epoxide ring, and the surface P O moieties receive a proton from
the terminal carbon.
Through the vibrational frequency calculations and FTIR anal-
yses, it was confirmed that a structure resembling (001)Na was
present under the reaction temperatures. In order to understand
the role of the surface moieties on this surface, the adsorption of EB
and corresponding change in the electron density were calculated.
Fig. 9A and B show the most stable geometry whereby EB is
adsorbed on the (001)Na surface. This geometry was found via sys-
tematic exploration of adsorption sites on the surface and then
rotation of the EB at the identified site. The adsorption energy
is −121 kJ/mol. Oxygen in the epoxide group interacts with two
under-coordinated Li atoms, and the distances between the surface
Li atoms and Oepoxide are 2.25 and 2.38 Å. The C O bond lengths in
Acknowledgements
the epoxide ring is elongated to 1.50 and 1.47 Å from 1.45 Å, which
indicates that surface Li sites activates the C O bond upon adsorp-
tion. It is also noteworthy that the surface P O group is in close
proximity to the terminal hydrogen of EB (P O· · ·H C distance
This research was supported by C1 Gas Refinery Program
through the National Research Foundation of Korea (NRF)
funded by the Ministry of Science, ICT & Future Planning (NRF-
2
016M3D3A1A01913255). This research was also supported by the
2
.27 Å).
Supercomputing Center/Korea Institute of Science and Technology
Information with supercomputing resources including technical
support (KSC-2015-C2-0007).
The difference in the electron density contour shows that a
significant change occurs in the electron density upon adsorption
Fig. 9C and D). In particular, electron density in the space between
(
epoxy oxygen and surface Li sites is increased considerably, which
means that the interaction between them is the major driving force
for the adsorption. The terminal hydrogen and the apical oxygen
Appendix A. Supplementary data
Supplementary data associated with this article can be found,
in the online version, at http://dx.doi.org/10.1016/j.mcat.2017.11.
(
Oap) also show a change in electron density. These redistribution
of electron density and the resultant electrostatic force cause the
strong adsorption of EB on the surface.
0
18.
Based on the adsorbed geometry and the change in the elec-
tron density, we proposed roles for the moieties on the surface of
lithium phosphate. The significant moieties on the surface of LPO
were under-coordinated Li atoms and the apical oxygen of phos-
phate. The former moiety plays a role of adsorbing EB by strongly
binding the epoxy oxygen, and of activating the C O bonds of an
epoxide ring. The latter interacts with a hydrogen atom attached to
References
[
[
[
1] J.C.S. Ruiz, J.A. Dumesic, Energy Environ. Sci. 4 (2011) 83–99.
2] X.J. Ji, H. Huang, P.K. Ouyang, Biotechnol. Adv. 29 (2011) 351–364.
3] W. Zhang, D. Yu, X. Ji, H. Huang, Green Chem. 14 (2010) 3441–3450.
[4] H. Duan, Y. Yamada, S. Kubo, S. Sato, Appl. Catal. A 530 (2017) 66–74.
[5] H. Duan, Y. Yamada, S. Sato, Appl. Catal. A 487 (2014) 226–233.

T.Y. Kim et al. / Molecular Catalysis 445 (2018) 133–141
141
[
6] T.Y. Kim, J. Baek, C.K. Song, Y.S. Yun, D.S. Park, W. Kim, J.W. Han, J. Yi, J. Catal.
[17] P.E. Blöchl, Phys. Rev. B 50 (1994) 17953–17979.
[18] S. Grimme, J. Comput. Chem. 27 (2006) 1787–1799.
3
23 (2015) 85–99.
[
[
7] H. Duan, Y. Yamada, S. Sato, Appl. Catal. A 491 (2015) 163–169.
8] W. Kim, W. Shin, K.J. Lee, H. Song, H.S. Kim, D. Seung, I.N. Filimonov, Appl.
Catal. A 511 (2016) 156–167.
[19] C. Keffer, A.D. Mighell, F. Mauer, H. Swanson, S. Block, Inorg. Chem. 6 (1967)
119–125.
[20] J. van den Brand, W.G. Sloof, H. Terryn, J.H.W. de Wit, Surf. Interface Anal. 36
(2004) 81–88.
[
9] M. Breuer, K. Ditrich, T. Habicher, B. Hauer, M. Keßeler, R. Stürmer, T. Zelinski,
Angew. Chem. Int. Ed. 43 (2004) 788–824.
[21] Y. Duval, J.A. Mielczarski, O.S. Pokrovsky, E. Mielczarski, J.J. Ehrhardt, J. Phys.
Chem. B 106 (2002) 2937–2945.
[22] W.E. Morgan, J.R. Van Wazer, W.J. Stec, J. Am. Chem. Soc. 95 (1973) 751–755.
[23] P. Swift, Sur. Interface Anal. 4 (1982) 47–51.
[24] R. Gresch, W. Müller-Warmuth, H. Dutz, J. Non-Cryst. Solids 34 (1979)
127–136.
[
10] Y.S. Demidova, O.V. Ardashov, O.A. Simakova, I.L. Simakova, K.P. Volcho, N.F.
Salakhutdinov, D.Y. Murzin, J. Mol. Catal. A 388–389 (2014) 162–166.
11] T. Imanaka, Y. Okamoto, S. Teranishi, Bull. Chem. Soc. Jan. 45 (1972)
[
1
353–1357.
[
[
12] W. Ma, W. Si, W. Wu, Q. Zhong, Catal. Lett. 141 (2011) 1032–1036.
13] H. Li, W. Jiang, W. Ma, H. Qu, Q. Zhong, Chem. Eng. Commun. 203 (2016)
[25] T.L. Barr, Appl. Surf. Sci. 15 (1983) 1–35.
3
39–344.
14] T. Imanaka, Y. Okamoto, S. Teranishi, Bull. Chem. Soc. Jpn. 45 (1972)
251–3254.
[26] M. Digne, P. Sautet, P. Raybaud, P. Euzen, H. Toulhoat, J. Catal. 226 (2004)
54–68.
[27] M.L. Hair, J. Non-Cryst. Solids 19 (1975) 299–309.
[28] J.J. Hudgens, S.W. Martin, J. Am. Ceram. Soc. 76 (1993) 1691–1696.
[
3
[
[
15] G. Kresse, J. Furthmuller, Phys. Rev. B 54 (1996) 11169–11186.
16] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996) 3865–3868.