Influence of basicity on 1,3-butadiene formation from catalytic 2,3-butanediol dehydration over γ-alumina

Article abstract of DOI:10.1016/j.jcat.2016.09.003

The direct catalytic conversion of 2,3-butanediol (BDO) to 1,3-butadiene (BD) was studied over two commercial forms of alumina (denoted as F200 and SCFa) at temperatures between 240?°C and 450?°C. Even though these two catalysts are both high surface area forms of γ-alumina, they gave remarkably different results, with SCFa giving higher BD selectivities at all experimental conditions. The difference is attributed to the higher surface area of F200, which means a greater number of acid sites that can convert BDO to methyl ethyl ketone (MEK). NH₃ and CO₂-TPD results supported this conclusion by showing that the two forms of alumina had different acid/base properties. Experimental results also showed that BD selectivity was improved by increasing temperature, increasing residence time and co-feeding water. The residence time study combined with density functional theory (DFT) calculations proved that 3-buten-2-ol (3B2OL) is an important intermediate in the conversion of BDO to BD. BD selectivity decreases over sodium modified alumina SCFa. It is hypothesized that on sodium-modified alumina, 3B2OL is dehydrogenated to form methyl vinyl ketone (MVK) as opposed to dehydration to BD. Basic sites catalyzed the retro-aldol condensation of MVK, which produces acetone and formaldehyde via cleavage of the C[dbnd]C bond. This is in agreement with DFT calculations showing that the proposed pathway for acetone formation is more energetically favored on Na-modified γ-Al₂O₃ (1?1?0) surface compared to the pristine (1?1?0) surface.

Full text of DOI:10.1016/j.jcat.2016.09.003

Journal of Catalysis 344 (2016) 77–89
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Influence of basicity on 1,3-butadiene formation from catalytic
2,3-butanediol dehydration over
c-alumina
a
b
b
a
a
a,
⇑
Fan Zeng , William J. Tenn III , Sudhir N.V.K. Aki , Jiayi Xu , Bin Liu , Keith L. Hohn
a
Department of Chemical Engineering, Kansas State University, Durland Hall, Manhattan, KS 66506, United States
Intermediates R&D, INVISTA S.à r.l., 3055 A FM 1006 Road, Orange, TX 77631, United States
b
a r t i c l e i n f o
a b s t r a c t
Article history:
The direct catalytic conversion of 2,3-butanediol (BDO) to 1,3-butadiene (BD) was studied over two com-
Received 3 February 2016
Revised 25 August 2016
Accepted 2 September 2016
mercial forms of alumina (denoted as F200 and SCFa) at temperatures between 240 °C and 450 °C. Even
though these two catalysts are both high surface area forms of c-alumina, they gave remarkably different
results, with SCFa giving higher BD selectivities at all experimental conditions. The difference is attribu-
ted to the higher surface area of F200, which means a greater number of acid sites that can convert BDO
3 2
to methyl ethyl ketone (MEK). NH and CO -TPD results supported this conclusion by showing that the
Keywords:
two forms of alumina had different acid/base properties. Experimental results also showed that BD selec-
tivity was improved by increasing temperature, increasing residence time and co-feeding water. The res-
idence time study combined with density functional theory (DFT) calculations proved that 3-buten-2-ol
(3B2OL) is an important intermediate in the conversion of BDO to BD. BD selectivity decreases over
sodium modified alumina SCFa. It is hypothesized that on sodium-modified alumina, 3B2OL is dehydro-
genated to form methyl vinyl ketone (MVK) as opposed to dehydration to BD. Basic sites catalyzed the
retro-aldol condensation of MVK, which produces acetone and formaldehyde via cleavage of the C@C
bond. This is in agreement with DFT calculations showing that the proposed pathway for acetone forma-
2
,3-Butanediol
Biomass conversion
Dehydration
Density functional theory
Retro-aldol condensation
tion is more energetically favored on Na-modified
surface.
c-Al
O
2 3
(110) surface compared to the pristine (110)
Ó 2016 Elsevier Inc. All rights reserved.
1
. Introduction
Dehydration of BDO can readily occur on zeolites [7], and with
mineral acids such as sulfuric acid [8]. However, methyl ethyl
ketone (MEK), rather than BD, is the dominant product due to
the keto-enol tautomerization and pinacol rearrangement. Only a
few studies resulting in successful selective production of BD have
been reported. Among them, Winfield, who studied the catalytic
dehydration of BDO to BD over thoria, reported a single pass con-
version of 60% to BD at 500 °C [1]. Shlechter obtained BD from
the esterification of BDO with acetic acid, followed by pyrolysis
of the diacetate [9,10]. BD production started at temperatures
above 475 °C. The optimal condition appeared to be at a tempera-
ture of about 585 °C and contact time of 7.1 s, where the BD yield
was 84.9% under such conditions in his study. Recently, BDO dehy-
dration over rare earth metal oxides has been studied and basic
Nearly 70 years ago, Winfield demonstrated that 2,3-butanediol
(
(
BDO) could be dehydrated over thoria to produce 1,3-butadiene
BD) for synthetic rubber manufacturing [1]. However, due to the
abundance of petroleum resources, BD has been almost exclusively
produced from hydrocarbon feeds, such as naphtha [2]. Recently,
there has been renewed interest in producing industrial chemicals
from biomass resources [3]. In this regard, catalytic BD production
from renewable resources is an appealing strategy, which ensures
ample supply and reduces price volatility of the final product [4].
Fermentation of biomass-based glucose and xylose by Klebsiella
pneumoniae yields BDO at high yields [5]. Alternatively, BDO can
be produced by nonpathogenic bacteria utilizing industry waste
gas [6]. These routes suggest that BDO can be obtained on a large
scale economically. This also suggests that the dehydration of
BDO to BD may be a plausible route for production of renewable
BD.
2
sites generated by introducing CaO into ZrO can enhance the 3-
buten-2-ol (3B2OL) selectivity [11–13]. The maximum BD selectiv-
ity was 94% with 100% BDO conversion on a two-bed catalyst sys-
2 3 2 3
tem (Sc O + Al O ) [14].
As noted above, thoria and scandium oxide both are successful
catalysts for BD production, while neither of which are commonly
used as catalysts. Thoria is a radioactive compound, while scan-
dium oxide is rare and expensive. For this reason, we have been
⇑
Corresponding author.
E-mail address: hohn@ksu.edu (K.L. Hohn).
http://dx.doi.org/10.1016/j.jcat.2016.09.003
021-9517/Ó 2016 Elsevier Inc. All rights reserved.
0

7
8
F. Zeng et al. / Journal of Catalysis 344 (2016) 77–89
considering catalysts based on alumina to convert BDO to BD.
Specifically, we have been studying ways to modify alumina to
enhance BD selectivity. The literature suggests that acid-base prop-
erties are important in BDO dehydration. In particular, BD selectiv-
ity can be influenced in the presence of alkaline species. For
example, Kim and Lee reported that cesium doped silica can pro-
mote BD selectivity which increased from 51% to 62% as the
use. The compositions of the alumina samples as supplied by man-
ufacturers are listed in Table 1.
Sodium is present in most commercial alumina at approxi-
2
mately 0.3–1.5 wt% Na O. F200 is typical of aluminas derived from
the Bayer process which is typically low in cost. Water and acid
washing steps are sometimes employed to lower the sodium con-
tent in aluminas formed via the Bayer process. SCFa is a higher-
purity alumina derived from the Ziegler process for the manufac-
ture of linear fatty-alcohols, which uses an alkyl aluminum cata-
lyst. In the Ziegler process, ethylene is oligomerized by the
aluminum catalyst followed by oxidation to form an aluminum tri-
alkoxide, which upon hydrolysis gives rise to the fatty-alcohol
products and boehmite. After boehmite calcination, very high-
purity gamma-alumina is obtained, which has less than 20 ppm
sodium content.
Cs
2
O/Al
2
O
3
mass ratio increased from 11% to 40% [15]. Tsukamoto
-supported CsH PO catalyst shows
and coworkers found that SiO
2
2
4
the highest BD selectivity above 90% in a single-bed catalyst sys-
tem and researchers speculated that the high BD selectivity is
probably due to combination of the proper acid-base activity of
+
Cs phosphate and large ionic radius of Cs [16]. However, other
2
researchers have found that 10 wt% Cs O and alkali metal oxides
(
Na O and K O) loaded on silica gave 2,3-epoxybutane as the main
2
2
product, while BD was not observed [17]. Researchers also studied
BDO dehydration over silica-supported sodium phosphates and
found the yield of the elimination products (BD + 3B2OL) was over
To understand the role of sodium in the BDO dehydration, SCFa
was impregnated with different amounts of sodium. Samples with
different amounts of Na
deionized water and placing SCFa powder in the resulting solution.
0.041 g NaNO was loaded onto 3.75 g of SCFa to prepare a sample
containing 0.3% Na O (mimicking the amount of sodium in F200),
while 0.85 g NaNO was loaded onto 3.75 g SCFa to determine
2 3
O were prepared by dissolving NaNO in
6
0% at Na/P = 1.8–1.9 which was the optimum combination of
acidic and basic sites [18]. Díez and coworkers studied the effect
of acid/base properties on the product distribution in the dehydra-
tion of 1,3-butanediol and concluded that acidic oxides promote
3
2
3
1
,3-butanediol dehydration to unsaturated alcohols and basic oxi-
the impact of high levels of sodium on catalyst performance. The
samples were dried at 120 °C in an oven for 12 h and finally cal-
cined at 600 °C for 6 h.
des dehydrogenate-dehydrate the diol to the unsaturated ketone,
which further decomposes by retro-aldol condensation, giving
C1AC3 alcohols, aldehydes and ketones [19]. All these experiments
indicated that the acid/base property of the catalyst had a crucial
impact on the product distribution and will ultimately affect the
BD selectivity.
Insights on the reaction mechanisms governing the main and
side reactions will be valuable to advance BDO dehydration catal-
ysis. To this end, we report a combined experimental/theoretical
study on the dehydration of BDO using two commercial forms of
alumina. The surface of alumina contains hydroxyl groups, Lewis
acid sites (aluminum atoms) and basic sites (oxygen atoms) [20].
Since both acid and base sites are likely required to produce BD
from BDO [21], alumina is attractive as a potential catalyst for
BDO dehydration to BD. In this study, BDO dehydration has been
carried out over alumina samples where, the acid-base properties
are modified by using these two different forms of alumina, by cal-
cining the samples at different temperatures, and by doping alu-
mina with sodium. To complement the experimental studies, DFT
calculations were carried out to investigate the energetics and
The impact of calcination on the catalysts was also studied
because calcination will affect both acid/base properties and phys-
ical properties of the catalysts. Calcination is particularly of inter-
est for this study since F200 has a much higher surface area than
SCFa, so calcination could convert F200 to a form more similar to
SCFa in terms of surface area and acid/base site density. SCFa and
F200 calcined in the air at 400 °C, 600 °C, 1000 °C and 1100 °C for
24, 24, 5 and 24 h, respectively, are denoted as SCFa-X and F200-
X, where X is the calcination temperature.
2.2. Catalytic reaction
The dehydration of BDO was carried out in a fixed-bed Hastel-
loyÒ tube reactor of 0.305 inner diameter. Since Hastelloy is a
potential catalyst, blank tests were performed in an empty tube
with the same conditions as the actual catalytic activity tests. For
all catalyst activity experiments, 0.5 g of catalyst was placed in
the reactor between two plugs of quartz wool. Liquid phase BDO
(2 g/100 mL) aqueous solution was fed at a flowrate of 0.1 mL/
min through a micro pump (Eldex) to the top of the reactor
00
kinetics of key BDO dehydration steps on model
c-alumina sur-
faces, where sodium-doped surface was also introduced for com-
parison. A consistent experimental/theoretical view was obtained
for the main BD formation pathways and relevant competing reac-
tions, and the catalytic routes for considered catalysts were pro-
posed based on the experimental and DFT results.
through a nebulizer, where it was mixed with 100 mL/min of N
2
(regulated by a Brooks 5850E mass flow controller), which is used
as an internal standard for product analysis. The approximate res-
idence time is 0.14 s. The reactor was heated by heating tape. The
temperature was measured by a K-type thermocouple, and a con-
troller was used to ensure that temperature was held constant at
the desired value.
2
. Experimental
Product analysis was carried out on-line by a SRI 8610C gas
chromatograph using TCD and FID detectors. The gas chro-
2.1. Catalyst preparation
matograph was equipped with molecular sieve (to separate N
2
SCFa and F200 were obtained from Sasol and BASF, respectively.
from organic products) and MXT-1 columns (60 m, ID 0.53 mm).
The oven temperature was held at 40 °C for five minutes, then
raised to 230 °C at a rate of 40 °C/min, and the temperature
remained constant for 15 min. Argon gas was used as the carrier
gas. The following compounds were identified using the gas chro-
matograph: BD, acetone, 2-methylpropanal (MPA), 3-buten-2-ol
(3B2OL), MEK and isobutanol. In addition to the above species,
heavy products were detected. Their response factor was assumed
to be the same as BDO, and their total amount is lumped together
as ‘‘heavy species”. Condensed species were characterized using
GC-MS (Shimadzu GCMS-QP2010 SE). Species identified include
The catalyst pellets were crushed and sieved to >60 mesh before
Table 1
Comparison of compositions of SCFa and F200.
Composition and chemical/physical properties
Al [%]
Na O [%]
SiO [%]
Fe [%]
L.O.I (loss on ignition) [%]
SCFa
F200
2
O
3
98
0.002
0
0
2
92.7
0.3
0.02
0.02
7
2
2
2 3
O

F. Zeng et al. / Journal of Catalysis 344 (2016) 77–89
79
2
-methyl-2-cyclopenten-1-one,
3,4-dimethyl-2-cyclopenten-1-
and it was found that the total energy has converged to within
50 meV. The break condition for self-consistent iteration is
one, 3,4,4-trimethyl-2-cyclopenten-1-one, 2-methyl-phenol, 3,4-
dimethylphenol and 2,4-dimethylphenol.
Each experimental condition was repeated at least three times
for each condition. All mass balances closed within 20%, with most
closing within 10%.
ꢀ6
1 ꢂ 10 . Ionic relaxation is stopped when the forces on all atoms
are smaller than ꢀ0.05 eV/Å. All the initial structures of the molec-
ular models were created using the Materials Studio Visualizer. A
vacuum equivalent to 4 layers of the slab was used with the bot-
tom two layers of the slab fixed at its bulk value. Dipole corrections
along the direction perpendicular to the surface were used to
decouple the electrostatic interaction between periodic images.
The energies and geometries of the gas phase species were cal-
culated by placing the molecule in a box with dimensions of
The conversion of BDO and selectivity of the main products
were calculated as below:
; 3-butanediol conversion ¼ mole
BDO; in ꢀ moleBDO; out
ꢁ 100% ð1Þ
ð2Þ
2
moleBDO; in
molei;out
moleproducts
2
0 ꢂ 20 ꢂ 25 Å. A
C-point k point was used for these calculations.
Selectivity ¼
ꢁ 100%
Gaussian smearing was used to assist convergence. All the calcula-
tions were non-spin polarized (except for the gas phase species
with unpaired electrons).
where moleproducts means the moles of products of reaction, includ-
ing MEK, MPA, BD and heavy products. The BDO conversion was
Transition states (TSs) and energy barriers were calculated
using the climbing-image Nudged Elastic Band (CI-NEB) method
1
2
2
00% except where stated.
[
26], coupled with the dimer method [27]. For the NEB calcula-
.3. Catalyst characterization
tions, seven images were used. The dimer method was then used
to further refine the preliminary TS structures. Each TS was con-
firmed to have only one imaginary vibrational mode. Finally, the
total energies are also corrected with Grimme’s DFT-D3 method
for the dispersion interactions between adsorbate molecules and
surface [28].
.3.1. XRD
X-ray diffraction measurements were performed on the catalyst
powders with a Rigaku MiniFlex II desktop X-ray diffractometer
with Cu K radiation (k = 0.15406 nm) in an operation mode of
0 kV and 15 mA. Data were collected from 10° to 80° with a scan-
a
3
ning rate of 1°/min. Crystallite sizes of
c-alumina were calculated
using the Scherrer formula:
3
. Results and discussion
d ¼ 0:9k=b cos h
ð3Þ
3.1. Characterization
where k = 0.15406 nm, b is the full width at half-maximum (FWHM)
in radians of the
Bragg angle.
c-alumina peak centered at 2h = 67°, and h is the
Fig. 1a and b shows the XRD patterns of alumina samples SCFa
and F200, respectively. As can be seen in this figure, spectra of
unmodified SCFa mainly correspond to -Al (JCPDS 10-0425)
with the dominant peaks detected at 45.65° and 67.14°. For F200,
peaks detected at 37.82° and 67.49° were attributed to -Al
c
2 3
O
2.3.2. TPD
Temperature programmed desorption of ammonia (NH
3
-TPD)
c
2 3
O .
was performed on an AMI-200. 200 mg of sample was treated
under Helium for 4 h at 600 °C. Next, 1% ammonia balanced in
Helium was flowed over the catalyst for 2 h. After saturation with
ammonia, the sample was flushed in a He flow at 120 °C for 1 h to
remove physically adsorbed ammonia. The temperature was then
raised to 600 °C at a heating rate of 5 °C/min. The amount of
ammonia desorption from the sample was measured by a thermal
conductivity detector. The peak area was converted to moles of gas
F200 displays fewer peaks and lower peak intensities as compared
to SCFa because it is a less crystalline, higher surface area form.
SCFa and F200 calcined at lower temperature such as 400 °C,
6
00 °C and 1000 °C did not have a noticeable phase change, but
as the temperature increased to 1100 °C, SCFa-1100 was converted
to the phase of alumina (JCPDS 42-1468) with peaks detected at
5.49°, 35.05°, 43.25°, 52.43°, 57.37°, 66.38° and 68.08°. It is easy
a
2
to see that calcination leads to an increased number of peaks and
increasing peak intensity for F200 similar in both peak distribution
and intensity to SCFa at 1100 °C. In Fig. 1a, no obvious difference in
the XRD pattern is noted after 0.03% Na was loaded onto SCFa.
2
by using a calibration file created for the species of interest. CO -
TPD was performed using the same procedure with 10% CO
2
in
Helium as the adsorbate.
2
However, as the Na O loading increased to 7.6%, diffraction peaks
2
.3.3. N
adsorption-desorption analyses were carried out using
Quantachrome Autosorb-1 instrument. The surface area was calcu-
lated by the BET method with P/P = 0.05–0.3. The total pore vol-
2
adsorption-desorption
attributed to sodium and support interactions were observed at
2h = 18.28 and 20.28. These can be assigned to sodium aluminum
oxide (Al Na O ) [29]. These two peaks are not observed in the
N
2
2
2 4
0
Na O XRD pattern, obtained as a reference by calcination of NaNO3
2
ume, average pore radius and pore size distribution were
calculated by the BJH method with the desorption branch of nitro-
gen isotherm. The samples were degassed under vacuum at 573 K
for 4 h prior to adsorption analysis.
at 600 °C for 6 h.
To explore the differences in acid-base properties between cal-
cined SCFa samples and calcined F200 samples, NH and CO TPD
3
2
experiments were performed. Fig. 2 shows the NH -TPD performed
3
2
on both calcined SCFa and F200 samples while CO -TPD is shown
2
.3.4. Density Functional Theory (DFT)
All DFT calculations were performed using the Vienna Ab initio
in Fig. 3. The total acidity and basicity in terms of moles of each
gas adsorbed are tabulated in Table 2.
Simulation Package (VASP) code [22]. The GGA-PBE functional was
used to account for the electron exchange-correlation effects [23].
The projector augmented wave (PAW) method was used to treat
the ion-electron interactions [24]. The energy cutoff for the plane
wave-based wave function was 380 eV. For the periodically
bounded slabs, the first Brillouin zone was sampled using a
3
As seen in Fig. 2, the broad NH desorption peak over SCFa
extends from 120 to 500 °C. For F200, there is a sharp peak at
200 °C with a broad shoulder extending out to 500 °C. For both
SCFa and F200 samples, the acid site density decreases as the cal-
cination temperature increases.
2
Fig. 3 shows the CO -TPD of the SCFa and F200 samples. In gen-
2
ꢂ 2 ꢂ 1 k-point mesh based on the Monkhorst-Pack scheme
eral, all spectra show a primary peak around 200 °C with a long tail
that extends to higher temperatures. At high temperature region,
[
25]. A higher number of k-points (i.e. 4 ꢂ 4 ꢂ 1) were also used

8
0
F. Zeng et al. / Journal of Catalysis 344 (2016) 77–89
Fig. 1. XRD patterns of alumina SCFa samples (a); and XRD patterns of alumina F200 samples (b).
3
Fig. 2. NH -TPD of SCFa samples (a); and F200 samples (b).
2
Fig. 3. CO -TPD of SCFa samples (a); and F200 samples (b).
SCFa exhibits a large desorption peak at 400 °C while this peak dis-
appears in sample SCFa-400 and SCFa-600. SCFa-1000 and SCFa-
The calculated acid site density (Acidity/BET surface area) of
2
SCFa and F200 are 1.8 and 1.6
lmol/m , respectively. For compar-
1
100 both exhibit a CO
but they both shift to lower temperature. Each F200 sample exhi-
bits a small CO desorption peak at high temperature region, while
it is hardly seen in sample F200-1100.
The CO -TPD data indicate that the base site density of both
SCFa and F200 decreases as the calcination temperature increases.
2
desorption peak at high temperature,
ison, Flego and coworker [30] reported an alumina acid site density
2
of 1.8
reported acid site densities of 1.6 and 1.9
The calculated base site density (Basicity/BET surface area) of SCFa
3
lmol/m by NH -TPD while Colorio [31] and Curtin [32]
2
2
lmol/m respectively.
2
2
and F200 are 0.45 and 0.37
lmol/m respectively, this is consistent
2
with the -Al base site density 0.48 lmol/m reported by Seki
c
2
O
3

F. Zeng et al. / Journal of Catalysis 344 (2016) 77–89
81
Table 2
Physical properties and acid/base site density of calcined SCFa and F200.
2
Total pore volume (cc/g)
BET surface area (m /g)
Average pore radius (Å)
Acidity (mmol/g)
Basicity (mmol/g)
Crystallite size (nm)
SCFa
0.461
0.461
0.463
0.441
0.262
0.430
0.453
0.459
0.409
0.125
131.0
128.6
126.8
101.8
57.4
341.0
264.9
184.6
77.1
70.3
71.8
73.1
86.6
91.3
25.2
34.2
49.8
106.2
89.0
0.24
0.23
0.21
0.19
0.10
0.55
0.50
0.37
0.09
0.04
0.059
0.051
0.038
0.023
0.017
0.127
0.117
0.072
0.034
0.033
5.6
6.2
6.0
6.3
10.4
2.5
4.0
4.1
7.0
9.8
SCFa-400 24 h
SCFa-600 24 h
SCFa-1000 5 h
SCFa-1100 24 h
F200
F200-400 24 h
F200-600 24 h
F200-1000 5 h
F200-1100 24 h
28.0
Table 3
Physical properties and base site density of sodium contained alumina.
2
Total pore volume (cc/g)
BET surface area (m /g)
Average pore radius (Å)
Basicity (mmol/g)
SCFa
0.461
0.464
0.434
0.430
131.0
129.7
114.0
341.0
70.3
71.5
76.2
25.2
0.059
0.059
0.322
0.127
SCFa-Na-0.3%
SCFa-Na-7.6%
F200
[
33]. However, others report much smaller base site densities of
low pressure, the smooth adsorption indicates that few micropores
exist in SCFa. As the calcination temperature increases, the hys-
teresis moves toward a higher relative pressure due to the agglom-
eration of alumina particles.
2
0
.12
To study the effect of sodium on the density of base sites on alu-
mina, CO -TPD was performed on SCFa, Na-modified SCFa and
F200. The results are shown in Fig. 4. All four samples exhibit a
CO desorption peak at 200 °C. Only SCFa exhibits a second desorp-
2 3
lmol/m for c-Al O [34].
2
Fig. 6a shows a type IV isotherm for F200, which relates to a
mesoporous structure. The sample also has some micropores, as
indicated by adsorption at low relative pressure. F200 has a H2
hysteresis loop which indicates that the ink-bottle pore may be
present in the mesoporous alumina [35]. As the calcination tem-
perature increases, the adsorption at low relative pressure disap-
pears and the hysteresis loop moves toward higher relative
pressure values, indicating that capillary condensation occurs in
larger mesoporous. F200 calcined at 1000 and 1100 °C exhibits a
H3 hysteresis loop, indicating slit-like pores [36].
The pore size distributions shown in Figs. 5b and 6b indicate
that as the calcination temperature increases, the SCFa and F200
pore size distribution broadens. The pore size of SCFa series are
narrowly distributed between 5 and 17 nm while SCFa calcined
at 1000 °C and 1100 °C had a broad pore size distribution which
ranges from 6 to 30 nm. F200 alumina has a narrow pore size dis-
tribution around 3.84 nm. As the calcination temperature
increases, the pore size distribution widens and F200 calcined at
1000 and 1100 °C exhibits an even wider pore size distribution.
This can be explained by noting that mesopores in the alumina will
be collapsed during the long duration calcination at high tempera-
ture. These results are consistent with the differences in SCFa and
F200 isotherms shown in Figs. 5a and 6a.
2
tion peak from 300 to 500 °C. The addition of 0.3% sodium to SCFa
only slightly modifies the TPD profile, increasing the low-
temperature peak while decreasing the peak area in the middle
region. Addition of 7.6% sodium dramatically increases the peak
area in the low-temperature region, while there is no discernible
peak in the middle region. Clearly, sodium increases the number
of weakly basic sites, but appears to lower the number of sites with
medium base strengths. To investigate this further, SCFa was cal-
cined at 600 °C for 6 h (labeled as ‘‘SCFa comparison”) as a refer-
ence and characterized with CO
the high temperature CO desorption peak was also not present
in the calcined sample, while basic site densities in other regions
were close to SCFa. This suggests that the second CO desorption
peak in SCFa was removed by the calcination step when sodium
was added, and not from sodium modification.
The nitrogen adsorption isotherm for calcined SCFa is shown in
Fig. 5a. All of them exhibit a type IV isotherm (defined by IUPAC)
which is characteristic of mesoporous material. SCFa has a H1 hys-
teresis loop, indicating that SCFa has a cylindrical pore structure. At
2
-TPD. The results showed that
2
2
Fig. 7a shows that as the sodium loading in SCFa increased, the
H1 hysteresis also moves toward higher relative pressure due to
particles agglomeration, similar to what happened as the calcina-
tion temperature increased. The XRD suggested that Na atoms
2 3
are intercalated with Al O during the calcination process. How-
ever, the sodium content in SCFa did not modify the SCFa H1 hys-
teresis loop to a H3 hysteresis loop as seen for F200. Fig. 7b shows
that the pore size distribution slightly narrowed as the sodium
content increases to 7.6%. Again, sodium did not change the SCFa
pore size distribution to a distribution more like F200.
Table 2 summarizes the physical properties of alumina SCFa
and F200 including the total pore volume, BET surface area and
the average pore radius and crystallite size. The specific surface
area decreases as the calcination temperature increases due to
agglomeration. The increase in average pore size with calcination
time can be explained by the growth of crystallites as the
2
Fig. 4. CO -TPD of sodium contained alumina.

8
2
F. Zeng et al. / Journal of Catalysis 344 (2016) 77–89
Fig. 5. N
2
adsorption/desorption isotherms of alumina SCFa and SCFa calcined at 400, 600, 1000 and 1100 °C (a) and the corresponding pore diameter distribution (b).
2
Fig. 6. N adsorption/desorption isotherms of alumina F200 calcined at 400, 600, 1000 and 1100 °C (a) and the corresponding pore diameter distribution (b).
2
Fig. 7. N adsorption/desorption isotherms of sodium contained alumina (a) and the corresponding pore diameter distribution (b).
calcination time increases, which can be seen in the XRD results
Fig. 1). The total pore volume of both SCFa and F200 initially
the sodium modified SCFa, the BET surface decreases and the aver-
age pore radius increases as the sodium loading increases. The
(
increases as the calcination temperature increases to 600 °C but
crystallite size of c-alumina was calculated from Scherrer equa-
further decreases when the calcination increases from 600 to
tion. The crystallite size generally increases as the calcination tem-
1
100 °C because larger crystallites occupy more space [37]. For
perature increases.

F. Zeng et al. / Journal of Catalysis 344 (2016) 77–89
83
3
.2. BDO dehydration over SCFa and F200
28%, versus less than 18% for F200. SCFa also produces more isobu-
tanol and 3B2OL than F200 (Table 4) at low temperatures and ace-
tone at high temperatures.
The product distributions from the reaction of BDO over SCFa
and F200 are summarized in Table 4. It is easy to see that SCFa
had a higher activity than F200 comparing the BDO conversion at
The catalysts appear to be relatively stable over time, at least on
the timescale of hours. After running SCFa for seven hours, for
example, the BD selectivity had changed from 24.7% to 22.6%,
while conversion was still 100% at a temperature of 400 °C.
MEK is the major product from BDO dehydration even in the
blank test. Higher temperature drives the reaction toward the for-
mation of BD, and inhibits the formation of MEK. Two mechanisms
are used to explain the formations of ketone structures. The enol
2
40 °C. Fig. 8 shows the species distributions for the main product
in Table 4.
As shown in Table 4, dehydration of BDO can occur even in the
blank reactor tube. Little reaction occurs below 400 °C (<4% con-
version), but at 400 °C over 30% of the BDO is converted to a mix-
ture of MEK, MPA, 3B2OL, and C3 hydrocarbons, with MEK
comprising almost 70% of the products.
With catalysts loaded in the reactor, both SCFa and F200 pro-
duce MEK as the major product. F200 always produces more
MEK than SCFa. For both catalysts, BD selectivity increases dramat-
ically with increasing temperature. Also, BD selectivity is always
higher on SCFa than on F200. The difference in BD selectivity is
most noticeable at 450 °C, where the BD selectivity on SCFa is
Scheme 1. Keto-enol tautomerism.
Table 4
Dehydration of BDO over different alumina samples at different temperatures.
Sample
T (°C)
Conversion (%)
TOFa (hꢀ1
)
Selectivityb (mol.%)
1
2
3
4
5
6
7
8
9
10
3
350
00
0.0
3.5
34.2
0
0
6.2
0
0
0
0
0
0
0
0
0
0
0
0
0
18.4
19.9
0
0
4.4
0
81.6
69.5
0
0
0
0
0
0
Blank test
400
2
3
350
4
4
40
00
31.7
3.5
4.8
3.4
2.3
0.0
1.4
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.4
0.0
0.0
0.0
1.9
1.5
4.0
7.3
9.9
8.1
3.3
3.6
0.0
0.0
0.0
0.0
83.1
72.8
64.4
57.7
56.9
1.2
1.4
0.0
0.0
0.0
3.3
0.0
0.8
7.5
8.4
100.0
100.0
100.0
100.0
P11.1
P11.1
P11.1
15.2
22.6
24.7
28.0
SCFa
F200
00
50
2
3
350
4
4
40
00
31.6
1.5
0.0
0.0
0.3
2.5
1.4
0.0
2.4
14.3
19.4
17.8
0.0
0.0
0.2
1.2
1.0
0.0
0.0
0.0
0.0
0.6
0.0
0.0
1.0
3.4
5.0
5.4
3.6
7.7
8.2
7.5
5.1
5.7
0.0
0.0
0.0
80.5
83.3
75.4
65.3
58.8
5.1
3.5
1.0
0.0
0.0
3.9
1.4
0.0
0.0
7.7
100.0
100.0
100.0
100.0
P4.8
P4.8
P4.8
P4.8
00
50
2
3
350
4
4
40
00
55.2
0.0
0.0
0.0
1.3
2.2
0.0
0.0
0.0
0.0
0.8
1.1
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
3.0
4.2
6.3
7.0
9.6
10.5
8.8
3.8
1.3
0.0
0.0
0.0
90.0
77.7
66.6
60.1
56.8
0.0
2.2
0.0
0.0
0.4
0.0
0.0
0.0
0.0
2.4
100.0
100.0
100.0
100.0
11.9
23.8
24.3
24.0
SCFa-Na-0.3%
SCFa-Na-7.6%
00
50
2
300
40
0.0
0.0
0.0
0.0
1.5
0.5
0.0
0.0
0.0
1.9
7.3
0.0
0.0
0.0
0.0
1.8
0.0
0.0
0.0
0.0
2.0
0.0
0.0
0.0
21.2
24.8
0.0
0.0
0.0
3.1
6.9
1.9
0.0
0.0
0.0
0.0
1.4
0.0
0.0
0.0
19.6
30.9
13.5
13.9
28.7
96.7
100.0
30.4
21.2
11.6
1.7
69.6
56.2
24.5
46.6
3
4
4
50
00
50
a
TOF: turnover frequency, conversion of BDO (mol/h) divided by acidic sites (mol).
b
1
: C3 (propane and propylene); 2: BD; 3: trans-2-butene; 4: cis-2-butene; 5: acetone; 6: MPA; 7: 3B2OL; 8: MEK; 9: isobutanol; 10: heavy species. Species identified
include 2-methyl-2-cyclopenten-1-one, 3,4-dimethyl-2-Cyclopenten-1-one, 3,4,4-trimethyl-2-Cyclopenten-1-one, 2-methyl-Phenol, 3,4-dimethylphenol and 2,4-
dimethylphenol.
Fig. 8. Comparisons of BD, MPA, MEK, and acetone selectivities from the reactions of BDO solution (2 g/100 mL) over (a) SCFa; and (b) F200. Points are the experimental data
and lines are the average value.

8
4
F. Zeng et al. / Journal of Catalysis 344 (2016) 77–89
structure can quickly form a keto structure via the keto-enol tau-
tomerization as shown in Scheme 1 [38].
becomes more like SCFa in terms of both surface area and acid
and base site densities as it is calcined. The F200 sample calcined
2
Another possible explanation is pinacol rearrangement, which
is a typical 1, 2-shift reaction of vicinal diols under acidic condition
and commonly believed to be the main pathway for MEK forma-
tion [39]. We have proposed a stepwise pinacol rearrangement
mechanism for BDO dehydration (Scheme 2). A migration of hydro-
gen with a pair of electrons is known as a hydride shift, while
migration with an alkyl group is known as a alkyl shift Hydride
migration leads to the formation of MPA, whereas methyl group
migration produces MEK.
at 600 °C, has a surface area of 184.6 m /g, which is close to the
2
SCFa surface area 131 m /g. The acid and base site densities of
F200-600 are 0.37 and 0.072 mmol/g respectively. When compared
to F200 samples calcined at other temperatures, they are close to
the SCFa acid and base densities 0.24 and 0.059 mmol/g. It is inter-
esting to note that these two samples gave similar catalytic results
in BDO dehydration. We hypothesize that the higher concentration
of acid sites present in F200 selectively produces MEK [7]. Another
possibility is that well-crystallized forms of alumina are superior to
less-crystallized forms for BD production, and that calcining F200
3.3. BDO dehydration over calcined alumina SCFa and F200
improves BD selectivity because it produces a more well-
crystallized surface. As shown in Table 2, the crystallite size of
F200 increases as it is calcined. Sato and coworkers proposed that
calcination could be important in dehydration of diols due to
changes in crystallinity [12].
Both SCFa and F200 alumina calcined at 1100 °C exhibited
much lower BD selectivity. This can be explained by the
Fig. 9 shows the BD selectivity for BDO dehydration over cal-
cined SCFa and F200. In Fig. 9a, the BD selectivity slightly increases
as the calcination temperature increases and then drops dramati-
cally when the calcination temperature is 1100 °C. In Fig. 9b, a sim-
ilar trend is found for F200. The BD selectivity increases as the
calcination temperature increases, reaching a maximum at F200-
transformation of
-Al which is generally considered as the most inert of all
the aluminum oxides [40].
2 3
c-Al O into the more thermodynamically stable
6
00. The BD selectivity further decreases as the calcination temper-
a
2 3
O
ature increases to 1000 °C and 1100 °C. As shown in Table 2, SCFa
surface area dropped only 3% as the calcination temperature
increased to 600 °C. A high surface area was maintained even at
3.4. Sodium effect on BD selectivity
1
000 °C. The small change in surface area in the SCFa, SCFa-400,
SCFa-600 and SCFa-1000 samples is likely the main reason behind
the constant BD selectivity. For F200, the surface area monotoni-
cally decreases as the temperature increases, and the surface area
losses are more noticeable than those in SCFa.
Different amounts of sodium were added to SCFa to determine
how it affected BD selectivity. The results are shown in Fig. 10,
which shows that as the sodium loading increases, the BD selectiv-
2
ity decreases. As the Na O increases by 0.3%, only slight decreases
As already described, SCFa consistently gives higher BD selectiv-
ity than F200, even though both samples are alumina. This is likely
due to differences in physical and chemical properties. Tables 2
and 3 compare the two catalyst samples, and show that there are
significant differences. However, these tables also show that F200
are seen at 300, 400, and 450 °C. This suggests that the small
amounts of sodium present in F200 are probably not responsible
for the differences in BD selectivity between F200 and the
higher-purity SCFa. As the sodium content increases to 7.6%, BD
selectivity drastically decreases. In addition, SCFa with high levels
of sodium gave much lower conversions than unmodified SCFa,
and produced large amounts of acetone (21.2% and 24.8%) at 400
and 450 °C respectively. This result is consistent with the previous
literature results that catalyst acid/base property had a critical
impact on the product distribution of the BDO dehydration reac-
tions [14–18].
The formation of acetone on sodium-modified SCFa may be
attributed to the enhancement of basic properties of the alumina
catalysts resulting from sodium impregnation. It has been reported
that the conversion rate of cinnamaldehyde to benzaldehyde and
acetaldehyde increased as the basicity increased [41], indicating
that the basic sites promote the retro-aldol condensation, which
breaks the C@C bond. In contrast, more acidic sites catalyze aldol
Scheme 2. MEK and MPA formation from pinacol rearrangement of BDO.
Fig. 9. BD selectivity over calcined alumina SCFa (a); and F200 (b).

F. Zeng et al. / Journal of Catalysis 344 (2016) 77–89
85
Fig. 10. Effects of temperature on BD (a); and acetone (b) selectivity over SCFa, SCFa-Na-0.3%, F200, and SCFa-Na-7.6%.
condensation reactions, where MVK has been found to be the sole
product from formaldehyde and acetone [42]. In this study, consid-
ering that the acetone selectivity increased as the catalyst basicity
was enhanced by sodium (shown in Fig. 2), we propose a mecha-
nism that acetone formation is catalyzed by sodium-modified alu-
mina through the retro-aldol condensation of MVK. Formaldehyde
was not directly observed in the products when acetone was
detected; however, larger amounts of heavy species (shown in
Table 4) were observed compared to BDO dehydration over SCFa.
These heavier species may have been formed from the reaction
of formaldehyde with other aldehydes and alcohols. To further val-
idate the proposed mechanism, 3B2OL (2 g/100 mL) and MVK
(see Table 4), BD was not observed over either form of alumina;
however, 3B2OL was observed instead, which indicates that high
temperature is required for complete dehydration of BDO to BD.
3.6. Effect of water vapor on dehydration performance over SCFa
BDO produced by fermentation is typically in a dilute aqueous
solution, and the production of pure BDO requires expensive sepa-
ration. Therefore, it is technologically significant to convert BDO to
BD in dilute solutions. For this reason, the impact of water on BDO
dehydration was studied in this work. Fig. 12 shows the effect of
water on dehydration of BDO over SCFa using pure BDO compared
with the previous diluted BDO (2 g/100 mL). The flow rate for pure
BDO was lower than that of diluted BDO (0.01 mL/min for pure
BDO and 0.1 mL/min for diluted BDO) because higher flow rates
of pure BDO led to problems in reactor operation. It can be seen
that acetone and MPA selectivities were not affected by the pres-
ence of water. Since the pure BDO flowrate is lower than diluted
BDO, it had a longer residence time (0.397 s for pure BDO and
0.139 s for diluted BDO) and it should give a higher BD selectivity
according to Fig. 11; however, the BD selectivity is higher in the
presence of water vapor. The main reason for this is that heavy spe-
cies are formed during the dehydration process due to BD polymer-
ization and aldol condensation; including water in the reaction
medium inhibits the formation of these heavy compounds by dilut-
ing BDO and its reaction products and slowing coupling reactions.
The diluted and pure BDO conversion in this figure is 100% for all
the temperatures.
(
2 g/100 mL) are separately fed into the reactor with 0.5 g SCFa-
Na-7.6% catalyst at 400 °C. Acetone is found in both experiments
and MVK is also found in the reaction of 3B2OL. These results are
consistent with the proposed mechanism.
3.5. Residence time study
Fig. 11 depicts BD, 3B2OL and BDO composition as functions of
residence time in the reaction of BDO over SCFa at 300 °C in a stain-
less steel reactor. By adjusting the catalyst weight loading in the
reactor and the feed rate of the aqueous BDO solution
(
0
2 g/100 mL), residence time was varied from 0.00786 s to
.109 s. Fig. 11 shows that the 3B2OL composition initially
increases to a maximum value at a residence time of 0.0133 s
and then drops as the residence time increases. The gap between
the data SCFa 0.1 g and 0.25 g may due to experimental error.
Meanwhile, the BD composition increases and the BDO composi-
tion decreases with increasing residence time. The trend for
3.7. Reaction mechanism
3
B2OL is what would be expected for a reaction intermediate in
a series reaction, which indicates that 3B2OL is the intermediate
in the route BDO dehydration to BD. At a temperature of 240 °C
Based on the results described above, we propose a reaction
mechanism show in Scheme 3. As seen in this scheme, BD is
Fig. 11. BD (a), 3B2OL (b) and BDO (c) composition as a function of residence time (points are the experimental data and lines are the trend line).

8
6
F. Zeng et al. / Journal of Catalysis 344 (2016) 77–89
Fig. 12. Product distributions from pure BDO (a); and BDO solution (2 g/100 mL) (b) dehydration over SCFa.
produced by the 1,2-elimination of water from BDO to produce
B2OL followed by a second dehydration step. This route to BD is
supported by the residence time studies described above, as well
as prior literature on conversion of BDO to BD [14,16]. Formation
of MEK readily occurs on either form of alumina by the pinacol
rearrangement [7].
One of the objectives of this work was to understand how basic
sites impact BDO conversion. For low loadings of Na, differences in
BD selectivity were not noticeable. However, we observed that
adding 7.6% Na to SCFa led to 3B2OL formation only at high tem-
peratures. This suggests that the strong base sites formed by dop-
ing with the alkali metal inhibited the formation of 3B2OL and
further dehydration of 3B2OL to BD, which explains why BD selec-
tivity decreased as the sodium loading increased. This conclusion is
inhibition of further dehydration of 3B2OL by base sites makes
the other reaction pathways for 3B2OL conversion possible over
SCFa-Na-7.6%, such as dehydrogenation to MVK. We propose that
3B2OL was converted into MVK by dehydrogenation on basic sites,
which then catalyzed the retro-aldol condensation by cleavage of
the C@C bond to form acetone and formaldehyde. The detection
of small amounts of hydrogen that accompany acetone formation
when using the SCFa-Na-7.6% catalyst supports the hypothesis that
dehydrogenation can play a role in the reaction mechanism for the
base-modified catalysts.
3
3.8. DFT calculations for BDO dehydration
2
in contrast to results by Kim and Lee who suggested that Cs O
The products from experimental BDO dehydration (shown in
loadings from 11 to 40% increased BD selectivity [15] and by
Tsukamoto and coworkers reported high BD selecitivites when
using cesium-doped hydrogen phosphates on silica [16]. One
explanation is that the basic sites on the catalysts synthesized in
this paper are different from those in the Kim and Lee and
Tsukamoto et al. papers. This is certainly plausible in the case of
the alkali metal hydrogen phosphate-modified silica catalysts.
Tsukamoto proposes that the active site for BDO dehydration is
the negative oxygen atom in the phosphate that is adjacent to
the alkali metal. They suggest that the alkali metal impacts the
basicity of this oxygen atom and that the ionic radius of the alkali
metal is important to ensure that the oxygen atom interacts with
the external hydrogen, leading to 3B2OL [16].
Fig. 8 and Table 4) indicate that the selectivity to MEK formation
is substantially higher than BD on both SCFa and F200. It is also
shown that calcination and sodium affect BD selectivity, and at
much higher sodium content (SCFa-Na-7.6%), acetone formation
becomes substantial. In order to gain further insights on these
experimental observations by elucidating the proposed reaction
pathways (Scheme 3), periodic DFT calculations were performed
to understand the selectivity of BDO dehydration producing MEK
and 3B2OL, and also to explore the relevant pathways for acetone
formation.
2 3
Periodically bounded c-Al O slabs with the (110) facet, which
is the preferentially exposed surface, are shown in Fig. 13 [43,44].
The 3-coordinated Al(III) site, which is catalytically active in alco-
hol dehydration [45,46], is highlighted with the gold color. The
sodium species were introduced as an adatom (equivalent to a
Na mass fraction of 2.7%, as shown in Fig. 13b). In this discussion,
we will primarily focus on the dry (dehydrated) surface. Calcula-
tions on hydroxylated surface were also performed, but the main
conclusions are not affected by the hydroxylation.
This work reports an additional function of base sites in BDO
conversion not discussed in most papers on the subject: produc-
tion of MVK and eventually acetone. It is hypothesized that the
The optimized structures of experimentally relevant reaction
species, such as BDO, CH -CH-CHOH-CH (INT, according to
3 3
Scheme 2), 3B2OL, MEK, and BD are shown in Fig. 14.
Molecular configurations and binding sites were explored to
identify the structures (Fig. 14) at global energy minima and then
were used to construct the potential energy surfaces. Except for
BD, all the species prefer to interact with the Al(III) site in both
pristine and sodium-modified (110) surface. BDO (a and f) and
3
B2OL (c and h) bind via their hydroxyl groups; INT (b and g) binds
with its unsaturated C atom; and MEK (d and i) binds with its
ketone oxygen (as in C@O) at the Al(III) sites, respectively. Notice-
able structural changes at the Al(III) site were observed upon BDO
and 3B2OL adsorptions, as shown in Fig. 14a and c, indicating that
surface restructuring of the Lewis acid site may occur [47]. For
Scheme 3. Proposed reaction pathways of BDO conversion over alumina.

F. Zeng et al. / Journal of Catalysis 344 (2016) 77–89
87
compared to the parallel route from MEK, due to the steep poten-
tial well for MEK. Fig. 15a also shows that the production of BD
should be more favorable on the pristine (110) surface than the
sodium-modified (110) surface, which is also consistent with the
trend derived from experimental product distribution correspond-
ing to catalyst samples with lower sodium contents. The BDO
dehydration pathways are also tested on hydroxylated alumina
surface (both pristine and sodium-doped), where the Al(III) site
remains open and the rest of the surface is saturated with dissoci-
ated water molecules (Fig. S1). It can be seen that the free three-
coordinate Al(III) sites remain catalytically active as the preferred
binding site (Fig. S2). The overall reaction pathways shift to higher
potential energies, and the dehydration energetics becomes more
endothermic. Compared to the dry pristine surface, the MEK route
is still energetically favorable, but to a lesser extent, indicating that
hydroxylation could indeed influence BD selectivity. This behavior
is manifested by the catalyst samples calcined at higher tempera-
tures, corresponding to a less hydroxylated surface (Fig. 9). The
optimized surface and reaction species structures (Figs. S1 and
S2), and the dehydration potential energies (Figs. S3 and S4) are
shown in the Supporting Information.
The kinetics of the first dehydration step in the MEK and 3B2OL
routes is shown in Fig. 15b. It is assumed that the first dehydration
step proceeds via direct CAO bond activation, producing INT (as
shown in Scheme 2) and one hydroxyl group. The first dehydration
is completed with a dehydrogenation step via the CAH bond cleav-
age. The CAO activation energy barriers are 1.79 eV and 0.1 eV on
respective pristine and sodium-doped alumina. The much lower
energy barrier can be understood as Na being able to stabilize
the transition state structure. The formations of 3B2OL and MEK
result from INT dehydrogenation (via CAH bond cleavage). The
transition state structures are shown as inset figures in Fig. 15b.
On pristine (110), these energy barriers are 1.08 eV (TS1) and
0.91 eV (TS2), respectively. On sodium-doped (110), the respective
CAH bond cleavage barriers increase to 2.32 eV (TS3) and 1.68 eV
(TS4). In particular, the barrier for TS4, which is responsible for
MEK formation, is much lower (by approximately 0.64 eV) than
the barrier for 3B2OL formation on the sodium-doped (110) sur-
face. Since 3B2OL is the main intermediate forming BD, the higher
energy barrier corresponding to the 3B2OL pathway could explain
the lower BD selectivity observed experimentally.
2 3
Fig. 13. Top and side views of relaxed pristine (a) and Na-doped (b) c-Al O (110)
surfaces. The Al(III) and Al(IV) sites are represented with gold and pink spheres
respectively. Sodium is represented by the purple sphere and is labeled in the
figure.
sodium-modified alumina, however, its Al(III) site is found to be
stable without significant structural changes.
Using gas-phase BDO and clean alumina surface as the refer-
ence (0 eV potential energy), the potential energy surfaces of the
MEK and 3B2OL dehydration routes to BD formation (Scheme 3)
are shown in Fig. 15a. For each dehydration step, the water from
dehydration is treated as a gas phase species. The overall potential
energies for BD production are lower on the pristine (110) sur-
faces. The adsorption energies of BDO on pristine and sodium-
modified (110) surfaces (with the inclusion of van der Waals inter-
actions) are 2.19 eV and 1.66 eV, respectively. The MEK route is
much more energetically favorable than the 3B2OL route on both
surfaces, due to stronger binding of MEK, and is consistent evi-
dence supporting the observed higher MEK selectivity. The MEK
desorption energies of MEK from the pristine and sodium-
modified (110) surfaces are quite endothermic, at 2.63 eV and
DFT calculations were also performed to explore possible reac-
tion pathways for acetone production observed experimentally for
catalysts with high sodium content (e.g. SCFa-Na-7.6% in Table 4).
The proposed mechanism for acetone formation via retro-aldol
condensation [48], involving MVK from 3B2OL dehydrogenation,
was investigated. The optimized structures of MVK and acetone
1
.99 eV, respectively.
BD is produced from a second dehydration, which is an
endothermic step. As shown in Fig. 15a, BD formation from the
B2OL route will be much more thermodynamically favorable
3
Fig. 14. (a–e) Optimized structures for BDO, INT, 3B2OL, MEK, and BD on pristine
2 3 2 3
c-Al O (110) surfaces; and (f–j) sodium-modified c-Al O (110) surfaces. White, gray, red,
pink, gold, and purple spheres represent the H, C, O, Al (IV), Al(III), and Na atoms, respectively.

8
8
F. Zeng et al. / Journal of Catalysis 344 (2016) 77–89
Fig. 15. (a) Potential energies of the two-step BDO dehydration forming 3B2OL, MEK, and BD on pristine (blue) and sodium-modified (red)
c
2 3
-Al O (110) surfaces. The
asterisks represent surface species. (b) Energy barriers for the first step of BDO dehydration on pristine (blue) and sodium-modified (red)
are the transition state structures identified from DFT calculations.
c-Al O (110) surfaces. Inset figures
2 3
are shown in Fig. 16. Like MEK, both molecules bind via their
ketone group at the Al(III) sites of the pristine and sodium-
modified (110) surfaces. The adsorption energies of acetone on
pristine and sodium-modified surfaces are 2.69 eV and 2.27 eV,
respectively.
The potential energies for the proposed 3B2OL ? MVK ? ace-
tone pathways on two surfaces are shown in Fig. 17. On the
sodium-doped (110) surface, it is more energetically favorable
for 3B2OL to form MVK via dehydrogenation (ꢀ0.32 eV exother-
2
mic, forming H gas), versus dehydration forming BD (1.79 eV,
endothermic). The retro-aldol condensation enables MVK to break
the C@C bond, producing acetone and formaldehyde. This step is
0
.96 eV endothermic on sodium-doped (110). DFT calculations
suggest that the MVK pathway competes strongly with BD forma-
tion given sufficiently high sodium contents (i.e. 7.6% Na O), and
2
hence is able to explain the observed acetone formation (Table 4).
On the pristine (110) surface, both MVK and BD pathways are
endothermic, with reaction energies of 0.28 eV and 0.77 eV respec-
tively. Higher selectivity for BD production, based on experimental
Fig. 17. Potential energies for proposed reaction pathways responsible for acetone
formation on pristine and sodium-modified -Al (110) surfaces.
c
2 3
O
analysis, suggests that kinetics (with potential high CAH bond
cleavage barriers) might favor the BD pathway on the more acidic
pristine (110) surface. As shown in Fig. S5, similar trend can be
noted regarding the selectivity for the acetone formation pathway,
except that energetic favorability for MVK more pronounced. On
the hydroxylated pristine surface, the MVK and BD formation path-
ways remain competitive. Both trends are consistent with the cal-
culations on the non-hydroxylated surfaces.
4
. Conclusion
Dehydration of BDO was investigated over alumina catalysts at
temperatures between 240 °C and 450 °C. Two different forms of
alumina gave different product selectivities, with the higher purity
form (SCFa) producing BD selectivities up to 10% higher than the
lower purity form (F200). BD selectivity was enhanced by increas-
2
ing reaction temperature and using longer residence times. CO -
TPD revealed that enhancing the basic site of SCFa decreased the
BD selectivity. Diluting BDO with water inhibited the formation
of heavy species and increased BD selectivity. DFT calculations
indicated that MEK was the most thermodynamically stable pro-
duct on alumina surface which explains why MEK was the domi-
nant product of BDO dehydration over alumina. The difference
between the two forms of alumina was due to either the larger
number of acid sites or lower crystallinity in F200 that gave a lower
BD selectivity. It is hypothesized that sodium addition inhibits
Fig. 16. Optimized structures for MVK (a and c) and acetone (b and d) on respective
pristine and sodium-modified
2 3
c-Al O (110) surfaces. White, gray, red, pink, gold,
and purple spheres represent the H, C, O, Al (IV), Al(III), and Na atoms, respectively.

F. Zeng et al. / Journal of Catalysis 344 (2016) 77–89
89
dehydration of 3B2OL to BD while enhancing dehydrogenation of
B2OL to MVK. Retro-aldol condensation of MVK could then occur,
leading to significant amounts of acetone produced on sodium-
doped alumina catalysts. DFT calculation proved that retro-aldol
[11] H. Duan, D. Sun, Y. Yamada, S. Sato, Catal. Commun. 48 (2014) 1.
[12] H. Duan, Y. Yamada, S. Sato, Chem. Lett. 43 (2014) 1773.
[13] H. Duan, Y. Yamada, S. Sato, Appl. Catal. A 487 (2014) 226.
[14] H. Duan, Y. Yamada, S. Sato, Appl. Catal. A 491 (2015) 163.
3
[15] U. Kim, H. Lee, Patent 10-2012-0099818, 2012.
[
[
[
[
16] D. Tsukamoto, S. Sakami, M. Ito, K. Yamada, T. Yonehara, Chem. Lett. 45 (2016)
31.
17] T.Y. Kim, J. Baek, C.K. Song, Y.S. Yun, D.S. Park, W. Kim, J.W. Han, J. Yi, J. Catal.
23 (2015) 85.
18] W. Kim, W. Shin, K.J. Lee, H. Song, H.S. Kim, D. Seung, I.N. Filimonov, Appl.
Catal. A 511 (2016) 156.
condensation of MVK is more favored on Na-Al
Al
2 3
O than on pure
8
2 3
O .
3
Acknowledgment
19] V. Díez, P. Torresi, P. Luggren, C. Ferretti, J. Di Cosimo, Catal. Today 213 (2013)
18.
Financial support for this work by INVISTA is gratefully
acknowledged. The authors are thankful to Dr. Kyle Probst and
Dr. Praveen Vadlani for liquid phase product analyses. The authors
are also thankful to Michael Heidlage and Dr. Mary Rezac for the
[
[
20] X. Liu, R.E. Truitt, J. Am. Chem. Soc. 119 (1997) 9856.
21] Y. Ono, H. Hattori, Solid Base Catalysis, Springer Heidelberg and Tokyo Inst.
Tech, London, NY, 2011, p. 48.
[22] G. Kresse, J. Furthmüller, Phys. Rev. B 54 (1996) 11169.
[
[
[
23] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996) 3865.
24] P.E. Blöchl, Phys. Rev. B 50 (1994) 17953.
25] H.J. Monkhorst, J.D. Pack, Phys. Rev. B 13 (1976) 5188.
2
help in N adsorption/desorption measurement. J. Xu and Dr. B.
Liu acknowledge the financial support from start-up funds pro-
vided by Kansas State University. We also acknowledge computa-
tional supports provided by Beocat Research Cluster at Kansas
State University, which is funded in part by NSF grants CNS-
[26] G. Henkelman, B.P. Uberuaga, H. Jónsson, J. Chem. Phys. 113 (2000) 9901.
[27] G. Henkelman, H. Jónsson, J. Chem. Phys. 111 (1999) 7010.
[28] S. Grimme, J. Antony, S. Ehrlich, H. Krieg, J. Chem. Phys. 132 (2010) 154104.
[29] M. Auta, N.A. Darbis, A.M. Din, B. Hameed, Chem. Eng. J. 233 (2013) 80.
[30] C. Flego, W.O. Parker, Appl. Catal. A 185 (1999) 137.
1
006860 and the National Energy Research Scientific Computing
Center (NERSC) under the contract No. DE-AC02-05CH11231.
[31] G. Colorio, J. Védrine, A. Auroux, B. Bonnetot, Appl. Catal. A 137 (1996) 55.
[32] T. Curtin, J. McMonagle, B. Hodnett, Appl. Catal. A 93 (1992) 91.
33] T. Seki, M. Onaka, J. Mol. Catal. A 263 (2007) 115.
[
Appendix A. Supplementary material
[34] X. Zhao, Y. Yue, Y. Zhang, W. Hua, Z. Gao, Catal. Lett. 89 (2003) 41.
[
35] E.A. El-Katatny, S.A. Halawy, M.A. Mohamed, M.I. Zaki, Powder Technol. 132
2003) 137.
36] G. Leofanti, M. Padovan, G. Tozzola, B. Venturelli, Catal. Today 41 (1998) 207.
(
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.jcat.2016.09.003.
[
[37] J. Sanchez-Valente, X. Bokhimi, F. Hernández, Langmuir 19 (2003) 3583.
[
[
[
38] R.T. Morrison, R.N. Boyd, Organic Chemistry (1963), p. 174.
39] K. Nakamura, Y. Osamura, J. Am. Chem. Soc. 115 (1993) 9112.
40] B.C. Lippens, J.J. Steggerda, in: B.G. Linsen (Ed.), Physical and Chemical Aspects
of Adsorbents and Catalysts, Academic Press, London, New York, 1970, p. 189.
41] H. Chen, H. Ji, AIChE J. 56 (2010) 466.
References
[
[
[
[
[
[
1] M. Winfield, J. Counc. Sci. Ind. Res. 18 (1945) 412.
2] W. White, Chem. Biol. Interact. 166 (2007) 10.
3] A. Demirba sß , Energy Convers. Manage. 42 (2001) 1357.
4] M.M. Bomgardner, Chem. Eng. News 89 (2011) 18.
5] N.B. Jansen, G.T. Tsao, Adv. Biochem. Eng. Biotechnol. 27 (1983) 85.
6] M. Kopke, C. Mihalcea, F. Liew, J.H. Tizard, M.S. Ali, J.J. Conolly, B. Al-Sinawi, S.D.
Simpson, Appl. Environ. Microbiol. 77 (2011) 5467.
[
[
[
[
[
42] M. Ai, J. Catal. 106 (1987) 273.
43] P. Nortier, P. Fourre, A.M. Saad, O. Saur, J. Lavalley, Appl. Catal. 61 (1990) 141.
44] J. Beaufils, Y. Barbaux, J. Chim. Phys. Phys. – Chim. Biol. 78 (1981) 347.
45] S. Roy, G. Mpourmpakis, D. Hong, D.G. Vlachos, A. Bhan, R. Gorte, ACS Catal. 2
(
2012) 1846.
46] J.F. DeWilde, H. Chiang, D.A. Hickman, C.R. Ho, A. Bhan, ACS Catal. 3 (2013)
98.
47] J. Joubert, A. Salameh, V. Krakoviack, F. Delbecq, P. Sautet, C. Copéret, J.M.
Basset, J. Phys. Chem. B 110 (2006) 23944.
[
[
[
[
[
7] W. Zhang, D. Yu, X. Ji, H. Huang, Green Chem. 14 (2012) 3441.
8] R.R. Emerson, M.C. Flickinger, G.T. Tsao, Ind. Eng. Chem. Prod. Res. Dev. 21
7
(
1982) 473.
[
9] N. Shlechter, D. Othmer, S. Marshak, Ind. Eng. Chem. 37 (1945) 900.
48] S. Sato, F. Sato, H. Gotoh, Y. Yamada, ACS Catal. 3 (2013) 721.
[
10] N. Shlechter, D. Othmer, R. Brand, Ind. Eng. Chem. 37 (1945) 905.