Propane Dehydrogenation over In2O3-Ga2O3-Al2O3 Mixed Oxides

Article abstract of DOI:10.1002/cctc.201500916

A series of ternary mixed metal oxides containing Group III A elements (In, Ga, Al) is prepared by means of an alcoholic co-precipitation method. Specifically, oxide catalysts with a molar composition of In/Ga/Al=5:15:80, 10:10:80, and 15:5:80 are reported. The chemical composition, redox properties, and catalyst structures are fully characterized, with the results suggesting that the indium, gallium, and aluminum moieties are well-dispersed in the catalysts. The catalysts are evaluated for propane dehydrogenation (PDH) at 570 and 600 °C under 1 atm total pressure. The most effective catalyst with a composition of In/Ga/Al=5:15:80 provides 17 % conversion and approximately 86 % C₃H₆ selectivity with an initial activity of 4.6 mmol h^-1 g_cat^-1 and 24.1 μmol h^-1 m^-2. The intrinsic activity on an active metal (i.e. indium and gallium) basis is approximately 3 times that of the In₂O₃-Ga₂O₃ family and approximately 3-9 times that of the In₂O₃-Al₂O₃ family. The catalyst deactivates with time on stream, and regeneration tests show that removal of surface coke and recovery of an In₂O₃ state helps to regain the initial activity, whereas reducing In₂O₃ domains into In⁰ does not allow for recovery of the performance. Raman analysis of the carbonaceous species deposited on the catalyst indicates catalysts with higher gallium content give more graphitic carbon, which correlates with higher C₃H₆ selectivity, whereas catalysts with more disordered coke are associated with lower selectivity. However, higher gallium content causes more coke formation, which leads to faster deactivation. This initial study of this family of mixed oxides suggests that an ideal In/Ga ratio may exist whereby catalyst properties may be optimized.

Full text of DOI:10.1002/cctc.201500916

DOI: 10.1002/cctc.201500916
Full Papers
Propane Dehydrogenation over In O –Ga O –Al O Mixed
2
3
2
3
2
3
Oxides
[a]
[a]
[b]
[b]
[b]
Shuai Tan, Seok-Jhin Kim, Jason S. Moore, Yujun Liu, Ravindra S. Dixit,
John G. Pendergast, David S. Sholl, Sankar Nair,* and Christopher W. Jones*
[b]
[a]
[a]
[a]
A series of ternary mixed metal oxides containing Group IIIA
elements (In, Ga, Al) is prepared by means of an alcoholic co-
precipitation method. Specifically, oxide catalysts with a molar
composition of In/Ga/Al=5:15:80, 10:10:80, and 15:5:80 are
reported. The chemical composition, redox properties, and cat-
alyst structures are fully characterized, with the results suggest-
ing that the indium, gallium, and aluminum moieties are well-
dispersed in the catalysts. The catalysts are evaluated for pro-
pane dehydrogenation (PDH) at 570 and 6008C under 1 atm
total pressure. The most effective catalyst with a composition
of In/Ga/Al=5:15:80 provides 17% conversion and approxi-
mately 86% C H selectivity with an initial activity of
mately 3–9 times that of the In O –Al O family. The catalyst
2 3 2 3
deactivates with time on stream, and regeneration tests show
that removal of surface coke and recovery of an In O state
2
3
helps to regain the initial activity, whereas reducing In O do-
2
3
0
mains into In does not allow for recovery of the performance.
Raman analysis of the carbonaceous species deposited on the
catalyst indicates catalysts with higher gallium content give
more graphitic carbon, which correlates with higher C H selec-
3
6
tivity, whereas catalysts with more disordered coke are associ-
ated with lower selectivity. However, higher gallium content
causes more coke formation, which leads to faster deactiva-
tion. This initial study of this family of mixed oxides suggests
that an ideal In/Ga ratio may exist whereby catalyst properties
may be optimized.
3
À1
6
À1
À1
À2
4
.6 mmolh gcat and 24.1 mmolh m . The intrinsic activity
on an active metal (i.e. indium and gallium) basis is approxi-
mately 3 times that of the In O –Ga O family and approxi-
2
3
2
3
Introduction
Propylene is a key building block in the synthesis of a large
number of chemicals such as poly(propylene), acrylonitrile, and
CO , N O, or O ) is co-fed to oxidize propane to propylene
2
2
2
[
4]
while reducing the oxidant gas. However, a key problem with
this approach is combustion of the hydrocarbons. Several re-
views provide a detailed discussion of ODH routes for light-
[
1]
propylene oxide. The global consumption of propylene
reached 90 million tons in 2015 and its market is still growing.
With the influx of new supplies of light gases in North America
owing to hydraulic fracturing and because catalytic dehydro-
genation has been demonstrated to be an effective method to
produce unsaturated short-chain hydrocarbons on a commer-
cial scale, catalytic propane dehydrogenation (PDH) is attract-
ing significant attention as a means to fill the supply-demand
gap.
[
5]
olefin synthesis.
In parallel, non-oxidative dehydrogenation has been widely
investigated and is more commercially successful. This ap-
proach is an endothermic pathway with equilibrium limita-
tions. Hence, these reactions are often performed under high-
temperature (i.e., >4508C) and low-pressure conditions (i.e.,
atmospheric) to obtain a relatively high yield of propylene.
Currently, there are two well-known commercial processes ap-
plied in industry, namely the Catofin process from CB&I Lum-
Among the means of propylene production from propane,
oxidative dehydrogenation (ODH) has been widely studied be-
[2]
[1b,6]
[7]
cause of its thermodynamically favorable pathway. For this
route, the most studied catalysts are vanadium- and cerium-
mus
and the Oleflex process from UOP. The former is
based on a Cr O /Al O catalyst in a fixed-bed reactor, whereas
2
3
2
3
[
3]
based oxides. In such a process, an oxidative gas (usually
the latter uses a Pt-based catalyst with a moving-bed reactor.
However, both of the above-mentioned processes suffer from
rapid catalyst deactivation owing to coke formation and depo-
sition on active sites, with many efforts being made to improve
the stability, activity, and selectivity. Generally, a regeneration
step is a necessary and crucial part of the process for maintain-
ing good propylene yield in industrial operation. Owing to the
importance of this approach to propylene synthesis, continu-
ing efforts have focused on new catalyst development, with an
eye towards lower-cost catalysts that might provide improved
performance. Recent studies have included catalysts involving
[
a] Dr. S. Tan, Dr. S.-J. Kim, Prof. D. S. Sholl, Prof. S. Nair, Prof. C. W. Jones
School of Chemical & Biomolecular Engineering
Georgia Institute of Technology
Atlanta, GA 30332 (USA)
E-mail: sankar.nair@chbe.gatech.edu
cjones@chbe.gatech.edu
[
b] Dr. J. S. Moore, Dr. Y. Liu, Dr. R. S. Dixit, J. G. Pendergast
Engineering & Process Sciences
The Dow Chemical Company
Freeport, TX 77541 (USA)
Supporting Information for this article is available on the WWW under
http://dx.doi.org/10.1002/cctc.201500916.
[
7d,8]
[9]
metal alloy
or single-site oxide catalysts. Ideally, alterna-
ChemCatChem 2016, 8, 214 – 221
214
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
tive catalysts that do not contain precious metals (i.e., Pt) or
toxic elements (i.e., Cr) that still exhibit comparable catalytic
performance are strongly desired.
Group IIIA oxides, namely In O and Ga O , were recently
2
3
2
3
demonstrated to be active for some hydrogenation reac-
[
10]
tions. Since 2000, researchers have focused on investigating
the dehydrogenation activity of these oxides. Fan and co-work-
[4c,11]
[12]
ers
and Gao and co-workers
have extensively studied
a variety of catalysts in this family. They proposed a heterolytic
dissociation reaction mechanism over a Ga O catalyst with the
2
3
[11a]
aid of DFT calculations.
Many of these studies focused on
bulk Ga O catalysts or zeolite-supported gallium domains, and
2
3
most studies have been conducted in the presence of oxida-
tive feed gases (i.e., CO , N O). For In O catalysts, a metallic
Figure 1. XRD patterns of fresh In–Ga–Al catalysts.
2
2
2
3
indium phase has been proposed as the active site, with evi-
dence of an induction period at the beginning of the reaction
ous study suggested that bulk In O in mixed In–Ga oxide cata-
2
3
[11d]
[13]
suggested to be associated with indium reduction.
er, in our previous work exploring the use of In O –Ga O
3
Howev-
lysts typically exhibits a highly ordered, crystalline structure.
However, for the current series of mixed In–Ga–Al oxides, there
are no observed diffractions belonging to In O . This may be
2
3
2
mixed oxides for PDH, such kinetic trends were not ob-
2
3
[
13]
served. Overall, the nature of the active site(s) of Group IIIA
metal oxides is still not fully understood, and it may depend
on the metal oxide composition and operating conditions. The
complexity is associated with several factors, including the co-
existence of different polymorphs of Ga O and reports that
due to the low loading of indium, its amorphous state, or its
existence in very small crystalline domains.
The textural properties of the ternary oxides catalysts are
summarized in Table 1. The physisorption isotherms can be
2
3
the type of support and its interaction with Ga O strongly
2
3
affect the redox properties of the mixed oxide. Moreover, the
Table 1. Textural properties of In–Ga–Al mixed oxide catalysts and In-free
and Ga-free alumina.
easy reducibility of In O and the low melting point (1578C) of
2
3
indium metal also complicates material characterization.
Al O is a very common support used in catalysis because of
[a]
[b]
[b]
Catalyst
S
total
2
V
total
Pore size
[nm]
À1
3
À1
2
3
[m g
]
[cm g ]
its thermal and chemical stability. Hereby, we hypothesize that
the incorporation of Al O could improve the dispersion of the
IGA5-15-80
IGA10-10-80
IGA15-5-80
190
174
160
205
0.12
0.13
0.15
0.27
3.1
3.3
3.3
3.6
2
3
active phase (such as In O and Ga O in our previous report)
2
3
2
3
2 3
Al O
and substantially enhance the catalyst stability and activity
i.e., prevent agglomeration of active phase). With our contin-
(
[a] Based on BET method. [b] Based on BJH method.
ued interest in Group IIIA mixed oxides, we report a detailed
exploration of ternary In–Ga–Al mixed oxides as catalysts for
non-oxidative propane dehydrogenation. The array of mixed
oxide catalysts is characterized by a variety of techniques to
elucidate the structure, including XRD, X-ray photoelectron
found in the Supporting Information (Figure S1). As can be
seen, the Al O synthesized using the same method (as a con-
2
3
2
À1
trol experiment) has the largest surface area (>200 m g ) and
pore size (ꢀ3.6 nm). However, the materials of the In–Ga–Al
series all exhibit lower surface area and smaller pore size. In
addition, these values decrease as the indium amount increas-
es. This is consistent with our previous report of pure In O
spectroscopy (XPS), N -physisorption, H2 temperature-pro-
2
grammed reduction (H -TPR), NH temperature-programmed
2
3
desorption (NH -TPD), elemental analysis, TEM, SEM, energy-
3
dispersive X-ray spectroscopy (EDX), and Raman spectroscopy.
These catalysts show promising performance in terms of activi-
ty, selectivity, and stability.
2
3
2
À1
2
À1 [13]
(11 m g ) and Ga O (84 m g ).
2
3
Surface acidity is an important property that could deter-
mine the catalytic activity, especially for a reaction that in-
volves CÀH and CÀC bond-breaking steps. Accordingly, the In–
Results and Discussion
Ga–Al series of catalysts were examined by NH -TPD and the
3
results are listed in Table 2. The traces for the In–Ga–Al metal
oxides contain two TPD peaks (see the Supporting Information,
Figure S2), one below 4508C and another at a higher tempera-
ture. These two regions refer to weak and strong acid sites, re-
spectively. It can be seen that the total acidity decreases with
increased indium loading, suggesting a stronger acidity associ-
ated with Ga O than with In O . The current study is consis-
Structural properties
The XRD patterns of fresh In–Ga–Al samples are shown in
Figure 1. It can be seen that all the In–Ga–Al samples show
broad diffraction lines at approximately 32, 35.5, and 658.
These peaks are assigned to (220), (311), and (440) planes of
[14]
g-Ga O (JCPDS 20-0426). Moreover, such broad lines indi-
2
3
2
3
2
3
cate low crystallinity of the synthesized Ga O domains, which
tent with previous NH -TPD studies over bulk In O and
3 2 3
2
3
[11b]
[4a,12c,13]
usually exist as a mixture of a/b/g polymorphs.
Our previ-
Ga O .
2 3
ChemCatChem 2016, 8, 214 – 221
www.chemcatchem.org
215
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Table 2. NH
3
-TPD results of In–Ga–Al mixed oxide catalysts.
Catalyst
a-peak [<4508C] b-peak [>4508C] Total
À1
À1
catÀ1
]
NH
3
[mmolgcat
]
NH
3
[mmolgcat
]
NH
3
[mmolg
IGA5-15-80
IGA10-10-80
IGA15-5-80
0.64
0.62
0.4
0.37
0.29
0.24
1.01
0.91
0.64
Chemical compositions were determined by means of XPS,
EDX, and inductively coupled plasma optical emission spec-
trometry (ICP-OES). The comparison of key results in the form
of indium/gallium ratios is shown in Table 3. It can be seen
2
Figure 2. H -TPR profiles of In–Ga–Al catalysts.
Table 3. Comparison of metal composition over In-Ga-Al mixed oxide
catalysts.
peared at approximately 6108C. As the indium amount de-
creases, the peak intensity in this region also decreases. It
should be mentioned that Ga O could not be reduced with
2
3
Catalyst
Indium/Gallium atomic ratio
[
16]
[a]
XPS
EDX
ICP-OES
H , whereas bulk In O exhibited a reduction temperature at
2 2 3
[
17]
approximately 7108C in previous work. It is accepted that
the reduction-peak temperature shift is associated with
changes in the In O domain size. Peaks at temperatures below
IGA5-15-80
IGA10-10-80
IGA15-5-80
0.5
1.33
3.67
0.45
1.28
3.52
0.42 (4:9:87)
1.19 (7:6:87)
3.94 (12:3:85)
2
3
3
008C are believed to be due to highly dispersed In O do-
2 3
[
a] Numbers in parenthesis are In/Ga/Al molar ratio obtained from ICP-
OES measurements.
mains within the mixed metal oxides.
Reactivity measurements
that despite the fact that the above-mentioned measurements
have different penetration depths or compositional coverages
The In–Ga–Al catalysts were evaluated in the propane dehy-
drogenation reaction. The main product is propylene, with side
products including methane, ethane, ethylene, and hydrogen.
Three types of experiments were carried out in this study. The
first type aimed at directly comparing the conversion and se-
lectivity under identical reaction conditions. The results are
shown in Figure 3a,b. It can be seen that the IGA15-5-80 and
IGA10-10-80 catalysts exhibit quite steady conversions around
10–15%, with a C H selectivity of 40–60%. In contrast, the
(
i.e., XPS is <10 nm, EDX probes ꢀ1–2 mm, and ICP examines
the entire bulk phase composition), the final results are quite
consistent with each other within the margin of error. The data
suggest that during the catalyst preparation, the indium, galli-
um, and aluminum salts are well-mixed and the corresponding
oxide did not have significant phase segregation of one type
of species. Elemental mapping also confirms such conclusions
(
see the Supporting Information, Figure S3).
3
6
The homogeneity of the mixed oxides and their morphology
IGA5-15-80 catalysts experienced rapid deactivation through-
out the entire experiment, although initial conversion is the
highest among the three catalysts. The overall activities are in
the order of IGA5-15-80>IGA10-10-80>IGA15-5-80.
were also confirmed by means of SEM and TEM (see the Sup-
porting Information, Figure S4). All the In–Ga–Al samples con-
tain similarly irregular particle shapes, suggesting that chang-
ing metal composition did not affect the particle morphology.
TEM imaging (Figure S4d–f) suggests that all the ternary
oxides exhibit a similar morphology. Unlike the work of Shi
and co-workers in the preparation of nanofibers containing
In O –Al O composites by means of a one-step electrospin-
Our previous study indicated that mixed In O –Ga O cata-
2
3
2
3
[
13]
lysts exhibit deactivation within the first 2 h on stream. The
current study suggests that incorporation of aluminum helps
to improve the catalyst stability.
A second series of experiments was carried out whereby
a steady, comparable conversion was achieved by adjusting
the catalyst amount while under fixed reaction conditions.
Under these conditions, the reaction rate (activity) and corre-
sponding selectivity can be directly compared. In Figure 4a,
the steady conversion values were controlled within 7–10%, al-
though the initial conversions varied. It should also be noted
that the IGA15-5-80 and IGA10-10-80 catalysts exhibited deac-
tivation only during the first hour on stream, whereas the
IGA5-15-80 catalyst suffered deactivation throughout the
2
3
2
3
[
15]
ning approach, we are unable to distinguish domains be-
longing to pure In O , Ga O , or Al O . Such observations again
2
3
2
3
2
3
suggest a good dispersion and mixing of species in the mixed
metal oxides.
Redox properties
The redox properties of the oxides are of importance for the
catalytic function of the In–Ga–Al mixed oxide catalysts. Thus,
H -TPR was applied to characterize the redox behavior of the
2
entire reaction. It can be observed in Figure 4b that the C H6
3
In–Ga–Al mixed oxides, and the results are shown in Figure 2.
It can be seen that for the IGA15-5-80 sample, a large peak ap-
selectivities all decrease gradually, with the best value being
ꢀ88% during the first hour over the IGA5-15-80 catalyst.
ChemCatChem 2016, 8, 214 – 221
www.chemcatchem.org
216
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Figure 3. a) C
/N
3
H
8
conversion and b) C
3
H
6
selectivity over various In
2
O
3
–Ga
2
O
3
–Al
2
O
3
mixed oxides. Reaction conditions: 0.2 g catalyst, 6008C, 20 sccm 5 vol%
C
3
H
8
2
.
3 8 3 6 2 3 2 3 2 3 3 8 2
Figure 4. a) C H conversion and b) C H selectivity over various In O –Ga O –Al O mixed oxides. Reaction conditions: 6008C, 20 sccm 5 vol% C H /N . Cata-
lyst amount: IGA15-5-80, 0.2 g; IGA10-10-80, 0.15 g; IGA5-15-80, 0.1 g.
Table 4. Catalytic performance of In–Ga–Al mixed oxide catalysts.
[
a]
[a]
[b]
Catalyst
Conversion [%]
Atomic Selectivity [%]
Activity
]
À1 catÀ1
À1
À2
CH
4
C
2
H
6
C
2
H
4
C
3
H
6
[mmolh
g
[mmolh
m ]
IGA5-15-80
IGA10-10-80
IGA15-5-80
17.1 (12.5)
12.7 (9)
14.6 (7.5)
4.5
26.6 (11.8)
1
2.7 (4.9)
5.3 (6.1)
4.7 (4.1)
5.4
1.7 (2.5)
9.1
1.9 (3.6)
2.5 (3.7)
1.5 (2.6)
4.6
1.7 (3.4)
8.7
6.1 (8.5)
11.1 (11.4)
8.8 (8.6)
11.8
2.0 (4.1)
18.2
85.5 (56.8)
77.8 (68.5)
66.5 (49.5)
65.8
88.1 (85)
63
4.6 (3.3)
2.3 (1.6)
1.9 (1.0)
0.4
8.9 (4)
–
24.1 (17.6)
13 (9.2)
12.2 (6.3)
2
61.4 (27.2)
–
Al
Na
blank test
2
O
3
[c]
3
2
O-Cr
2
O
3
/Al
2
O
[
a] Initial conversion/selectivity data obtained at 10 min. Numbers in parenthesis are data obtained at 6 h on stream. [b] Instantaneous activities are calcu-
lated by using the first data point. Intrinsic activities (in parenthesis) are estimated by extrapolating the linear region of conversion to zero time. [c] Catalyst
tested has a composition of 1 wt% Na
2
O, 20 wt% Cr
2
O
3
deposited on Al
2
O
3
(provided by The Dow Chemical Company). Catalyst amount is 0.08 g.
Taking the catalyst loadings in these runs into account, the ac-
tivity still follows the order of IGA5-15-80>IGA10-10-80>
IGA15-5-80. It should be mentioned that no significant induc-
tion period was observed in the present study, which is differ-
ent from the work of Zhuang and co-workers on In O –
are summarized in Table 4, along with other key results. In this
study, we report two types of activities. The intrinsic activity is
obtained by extrapolating the linear region of conversion back
to time zero. The instantaneous activity is estimated by using
the first data point at 10 min reaction time. The data suggest
that the activity of In–Ga–Al oxides increases as the indium
amount decreases, which is consistent with our study on the
2
3
[
11d]
Al O ,
suggesting the possibility of different active site(s) in
2
3
the two catalysts.
[
13]
The reaction rates (activity) on both total mass and surface
area bases (according to data in Figure 4) were calculated and
In–Ga binary system. With that, the highest activities among
À1
À1
the series are from IGA5-15-80: 4.6 mmolh gcat
, and
ChemCatChem 2016, 8, 214 – 221
www.chemcatchem.org
217
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
À1
À2
2
4.1 mmolh m . The corresponding highest conversion is ap-
clude an air-oxidation regeneration treatment. It can be seen
from the data in Figure 5a that except for the first point, the
conversion was quite stable with only a slight decrease in each
cycle. Furthermore, catalyst regeneration by reduction in hy-
drogen did not yield much improvement. The observed stabili-
ty of the IGA10-10-80 catalyst is consistent with the data in
Figure 3 and 4. Meanwhile, the C H selectivity was approxi-
proximately 17%, with 86% C H selectivity. If only considering
3
6
In O and Ga O moieties as the active phase, the active metal
2
3
2
3
basis activity would be tripled, as the Al O composition
2
3
amounts to approximately two thirds of the total mass. It
should be mentioned that the In–Ga binary system gave activi-
À1
À1
ties ranging from approximately 1.9–4.9 mmolh gcat
.
3
6
Hereby, the current In–Ga–Al catalysts show superior activity
by a factor of three in terms of an active metal (In O and
mately 80%, with no significant decrease over time.
In contrast, after short air regeneration (Figure 5b), the
IGA10-10-80 almost recovered all of its initial activity (conver-
sion). However, this was followed by a drop to a stable value
of approximately 8% conversion. It should be noted that the
steady level of C H selectivity was also approximately 80%.
2
3
Ga O ) basis.
2
3
Table 4 also includes results from 1 wt% Na O, 20 wt%
2
Cr O /Al O (a benchmark catalyst). It was found that the best
2
3
2
3
In–Ga–Al catalyst (IGA5-15-80) exhibits similar initial C H selec-
3
6
3
6
tivity (85.5 vs. 88.1%) with a slightly lower initial conversion
17.1 vs. 26.6%). The instantaneous activity was approximately
Considering that the coke species could be combusted in the
presence of air, these experiments suggest that the deactiva-
tion behavior is associated with deposition of coke and the
subsequent blocking of active sites. Taking into account that
(
À1
À1
half that of Cr O /Al O (4.6 vs. 8.9 mmolh gcat ). After 6 h of
2
3
2
3
reaction, the conversion of IGA5-15-80 was higher than that of
Cr O /Al O (12.5 vs. 11.8 %), and the intrinsic activity was simi-
0
In could be oxidized into In O , and In O moieties can also be
2
3
2
3
2
3
2
3
À1
À1
lar to each other (3.3 vs. 4 mmolh gcat ). However, it should
be mentioned that the C H selectivity was lower than that of
reduced by H under such temperatures (5708C), the observed
2
higher initial conversion in Figure 5b (even after two regenera-
tion cycles) than that in Figure 5a suggests that In O domains
3
6
the Cr-based catalyst (56.8 vs. 85 %). Thus, the chromium cata-
lyst provides a high level of propylene selectivity, but suffers
rapid deactivation and loss of conversion. In contrast, the In–
Ga–Al catalyst seems to be more stable in terms of maintain-
ing the conversion, but with inferior C H selectivity.
2
3
0
are more likely to be responsible for reactivity than In do-
mains. However, additional studies confirming the active site
hypothesis are needed.
3
6
Ex situ XRD measurements were carried out to reveal poten-
tial changes of crystallinity in In–Ga–Al samples after reaction.
The patterns are shown in the Supporting Information (Fig-
ure S5). In comparison with Figure 1, no change in terms of
crystallinity was observed. This observation is in contrast with
the previous study of In O –Ga O mixed oxides, in which
Coke analysis
In PDH, it is known that coke generation and deposition is the
main cause of catalyst deactivation. Furthermore, the above re-
activity data suggest that coke formation plays a key role in
the deactivation of the current series of mixed oxide catalysts.
To this end, Raman spectroscopy was applied for analysis of
the coke formed over the In–Ga–Al catalysts. The raw spectra
as well as curve fits are shown in the Supporting Information
(Figure S6). The deconvolution procedure has been described
2
3
2
3
0
sharp peaks owing to formation crystalline In domains and ag-
glomerated In O were present after reaction. The addition of
2
3
Al O to the In O and Ga O clearly improves the performance
2
3
2
3
2
3
of the catalysts.
[
13,18]
Finally, the stability of the ternary oxide catalyst was investi-
gated by using IGA10-10-80 as a representative catalyst. Fig-
ure 5a displays conversion and hydrocarbon selectivities with
in previous work.
In short, four bands were found to be as-
sociated with carbonaceous species on the catalyst surface in
À1
À1
the 950–1850 cm region. The band at ꢀ1200 cm is as-
signed to a disordered graphitic lattice, whereas the peak at
ꢀ1500 cmÀ1 is due to amorphous carbon.[19] Two intense
H -reduction treatment versus time-on-stream over three
2
cycles at 5708C. Figure 5b shows the reaction results that in-
2 3 8 2
Figure 5. Catalyst stability and regeneration test with a) H reduction and b) air oxidation. Reaction conditions: 5708C, 10 sccm 5 vol% C H /N , 0.2 g IGA10-
1
0-80.
&
=Conversion; * =methane; ~ =ethane; ! =ethylene; ^ =propene.
ChemCatChem 2016, 8, 214 – 221
www.chemcatchem.org
218
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
1
Table 5. D /G ratio from Raman spectroscopy and coke content of spent catalysts.
[
a]
[b]
[c]
Catalyst
D
1
/G
Activity
[
C
[%]
3
H
6
selectivity
Coke
Carbon balance
[%]
À1
mmolh /gcat
]
[wt%]
[d]
[e]
Raw
Calibrated
IGA5-15-80
IGA10-10-80
IGA15-5-80
0.45
0.58
0.59
0.78
4.6
2.3
1.9
8.9
85.5
77.8
66.5
88.1
8.5
3
0.9
0.4
96
97
97
98
99
99
98
99
2 2 3 2 3
Na O-Cr O /Al O
[
a] Instantaneous activity from Table 4. [b] Initial selectivity from Figure 4. [c] Data obtained by ICP-OES. [d] Based on average gas-phase compositions de-
termined by GC analysis (coke is not considered). [e] After addition of coke amount obtained from ICP-OES. Catalysts from reactions described in Figure 3.
Higher coke level with IGA5-15-80 catalyst due, at least in part, to significantly higher propane conversion.
À1
peaks were found at ꢀ1370 and 1600 cm . These peaks are
an intermediate In/Ga composition is associated with a balance
of good activity, selectivity, and stability.
assigned as the D band due to in-plane defects and the G
1
band from carbon species containing a highly crystalline struc-
[
20]
ture, respectively. The R ratio (D /G intensity ratio) is com-
1
1
Conclusions
monly used to evaluate the degree of organization of the de-
posited carbonaceous species. The values for each catalyst, as
well as the coke content obtained from ICP-OES analysis is
summarized in Table 5. It can be seen that as the indium/galli-
This work has focused on the development of a series of terna-
ry In–Ga–Al mixed oxides as catalysts for propane dehydrogen-
ation. XRD measurements show no significant crystallinity in
the samples, as well as no change of crystal structure after the
reaction, indicating good structural stability of the current ter-
nary oxide system. Elemental mapping, together with XPS,
EDX, and ICP-OES analysis show good homogeneity of the re-
sulting catalysts. The best catalytic performance was associated
with the IGA5-15-80 catalyst, giving approximately 86% initial
C H selectivity at approximately 17% conversion. The active
um ratio increases, the D /G ratio also increases, suggesting
1
a higher percentage of disordered carbonaceous species being
produced at higher indium loadings. Meanwhile, the initial ac-
tivity and selectivity decrease. Such observations suggest a pos-
sible linkage between the coke properties and the C H selec-
3
6
tivity; carbon materials with disordered features are associated
with lower C H selectivity. A similar trend was observed in the
3
6
3
6
[
13]
In O –Ga O series, as well as in reports by Weckhuysen and
metal basis intrinsic activity is approximately 3 times that of
the In O –Ga O mixed oxide family and approximately 3–
2
3
2
3
[
19,21]
co-workers using a Pt-Sn/Al O and Pt-Ga/Al O catalyst.
It
2
3
2
3
2
3
2
3
should be noted that the absolute amount of coke monotoni-
cally decreases as the gallium percentage decreases (IGA5-15-
9 times higher than the In O –Al O family. Regeneration tests
2 3 2 3
show that removal of surface coke and oxidizing reduced
indium species to the In O state helps to regain the initial ac-
8
0>IGA10-10-80>IGA15-5-80). This trend correlates with the
2
3
0
deactivation behavior shown in Figure 3 and 4, consistent with
the notion that increased coking is associated with a faster
loss of activity. Two types of carbon balances are summarized
in Table 5. The raw carbon balance is calculated based on all
gas-phase compositions obtained by gas chromatography
tivity, whereas reducing the In O domains to In does not im-
2 3
prove the performance. Raman analysis of carbonaceous spe-
cies deposited on the catalyst surface indicates catalysts with
more gallium give more graphitic carbon species, which is as-
sociated with a higher level of C H selectivity than catalysts
3
6
(
(
GC). The value is an average over the entire reaction time
6 h). The calibrated carbon balance is reported by taking the
with more disordered coke. However, higher gallium loading
also causes a larger amount of coke formation, which leads to
faster deactivation.
coke amount (determined by ICP-OES) into consideration. It
can be seen that the mass balance is well-matched.
In summary, the current study introduces a new family of
ternary mixed oxides containing Group IIIA elements. The
system shows superior performance compared to the binary
mixture (In O –Ga O , or In O –Al O ), in terms of stability and
The Raman analysis was also carried out over the Na O-
2
Cr O /Al O control catalyst, and the results are summarized in
2
3
2
3
Table 5. Significant differences between the two families of cat-
alysts can be observed from the data in the table. The heavily
optimized Cr system offers higher reaction rates, selectivities,
and reduced coke levels, compared to the less developed In–
Ga–Al catalysts. In a recent study using Ga O /Al O -doped Pt
2
3
2
3
2
3
2
3
C H selectivity. There remains an opportunity to further opti-
3
6
mize the catalyst composition to achieve further improved cat-
alytic performance for propane dehydrogenation.
2
3
2
3
[21]
and K as the catalyst, the D/G ratio ranged from 0.73–1.00,
with the coke amount varying from 0.23–0.55 wt%. The cata-
lyst had a selectivity that ranged from 65.8–96.9%. However,
these values were obtained under higher reaction tempera-
tures (6208C) with an even harsher regeneration treatment
Experimental Section
Catalyst preparation
A family of ternary In O –Ga O –Al O mixed oxides was synthe-
2
3
2
3
2
3
sized through an alcoholic co-precipitation method. A detailed de-
scription was published in our previous study. In a typical syn-
(
7508C), making direct comparison problematic.
[13]
The overall implication here is that an optimal In/Ga ratio
thesis, a given amount of indium, gallium, and aluminum salts (ni-
trate hydrate, 99.9%, Sigma–Aldrich) at different molar ratios was
may exist in this catalyst family, the present work shows that
ChemCatChem 2016, 8, 214 – 221
www.chemcatchem.org
219
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
dissolved in ethanol solvent (95%, Alfa-Aesar) with a stirring rate
of 150 rpm. Then, a pre-mixed solution containing aqueous ammo-
nia (28 wt%, Sigma–Aldrich) and ethanol in 1:1 volume ratio was
added dropwise until no further precipitation occurred (pHꢀ9).
The resulting gel-like mixture was kept stirring (150 rpm) in the
flask for 1 h at RT. Next, the mixture was transferred into a vial and
centrifuged for 30 min at 6000 rpm. Afterwards, the precipitated
gel was dried in an oven at 708C overnight. Finally, the obtained
sample was transferred into a crucible and calcined in air at 6008C
for 6 h. The resulting In–Ga–Al ternary oxides are labelled as
IGAx-y-z, where x, y, and z denote the molar ratio of In/Ga/Al. For
example, IGA10–10–80 refers to sample with In/Ga/Al in a ratio of
flow, followed by cooling to 508C. Then, 30 sccm of a 10 vol% H2/
Ar was introduced while heating to 9008C with a ramp of
58Cmin . The signal was recorded by a thermal conductivity de-
À1
tector .
SEM images and EDX spectra were obtained on a JEOL LEO-1530
at a landing energy of 15 kV by using the “InLens” mode detector.
TEM images were taken by using a Hitachi HF-3300 TEM-STEM at
300 kV in TEM mode at Oak Ridge National Laboratory (ORNL). In
preparation for TEM imaging, samples were prepared by crushing,
dispersing the particles in methanol, followed by sonication and
dropping the slurry on a lacey carbon film/Cu grid.
1
0:10:80.
Catalyst evaluation
Catalyst characterization
The propane dehydrogenation reactions were performed in a U-
shape fixed-bed quartz reactor (ꢀ80 cm long with 3.5 mm internal
diameter). The catalyst with a 150–212 mm pellet size was loaded
inside the u-tube on a quartz wool support. The reaction was con-
ducted under atmospheric pressure. The reactor was heated to the
desired temperature (i.e., 6008C) in a Techne FB-08 Al O fluidized
N2 adsorption–desorption isotherms were carried out by using
a Tristar II 3020 from Micromeritics. Approximately 150 mg of
sample was degassed under vacuum and preheated at 1208C for
1
2 h on a Schlenk line before the measurements were taken. The
2
3
specific surface area, pore volume, and the average pore size were
estimated by using the Brunauer–Emmett–Teller (BET) method and
Barrett–Joyner–Halenda (BJH) method by using the software asso-
ciated with the apparatus.
bath under 20 sccm N flow. The feed gas was a mixture of fixed
2
volume percentage of C H balanced with N . The corresponding
3
8
2
hydrocarbon products as well as the H produced were determined
2
online by a Shimadzu GC-2014 gas chromatograph equipped with
a Restek Column (Rt-Alumina BOND/Na SO , 30 m”0.25 mm”
XRD measurements were carried out with an X’pert Pro X-ray dif-
2
4
fractometer from PANalytical with nickel-filtered CuKa radiation
4 mm) and a thermal conductivity detector, respectively. The data
were collected after flowing the feed gas for 10 min. Calculated
carbon balance ranges varied from 95–105%. The conversion and
atomic selectivity were calculated according to the following equa-
tions:
(
0.154187 nm), with generator settings of 45 kV and 40 mA. The
scanning regions of 5–908 with a scanning step size of 0.0088 were
used.
XPS analysis was performed by using a Thermo K-Alpha spectrom-
eter equipped with a monochromatic AlKa X-ray source. Pressures
C3H À C H
À7
8in
3
8out
near 3”10 mbar were observed in the analytical chamber during
C
3
H
8
Conversionð%Þ ¼
 100 %
C H
3
8in
surface analysis. The binding energies (B.E.) of all elements were
calibrated with the C1s peak at ꢀ284.8 eV with an uncertainty of
Æ0.2 eV.
Y
3
Xout
C3H8in À C H
Atomic Selectivityð%Þ ¼
Â
 100 %
8out
3
Elemental analysis for In, Ga, Al, and C within the catalysts was per-
formed by using ICP-OES. This service was provided by ALS Envi-
ronmental.
Where Xout refers to products, such as CH , C H , C H , C H , and
4 2 6 2 4 3 6
others and Y is the number of carbon atoms in the product mole-
cules.
Raman spectroscopy was carried out with a Thermo Nicolet Disper-
sive Raman Spectrometer equipped with a CCD detector and con-
focal optics. A microscope was applied to focus the excitation laser
with a power of ꢀ15 mW on the sample. The signal was collected
in backscattering mode. The acquisition time for the background
and sample scans was 45 s. The number of scans for each sample
was 30.
Acknowledgements
This work was financially supported by the Dow Chemical Com-
pany. Microscopy research was carried out in part at ORNL’s
ShaRE User Facility. We acknowledge Dr. Karren L. More (ORNL)
for TEM microscopy and Kimberly Shawn Reeves (ORNL) for assis-
tance in TEM sample preparation. The authors would also like to
thank Seung-Won Choi for fruitful discussions.
The total surface acidity was determined by NH -TPD at ambient
3
pressure with an Autochem II Chemisorption Analyzer from Micro-
meritics. Approximately 150 mg of sample was preheated to 5008C
À1
with a ramp of 108Cmin under a 25 sccm He flow and held
under those conditions for 1 h, followed by cooling to 1208C with
Keywords: alkanes · dehydrogenation · gallium · indium ·
oxides
À1
a ramp of 508Cmin . Then 30 sccm, 5 vol% NH /He was injected
3
for 2 h to saturate the sample, followed by purging with He
(
25 sccm) for 1 h. The desorption of NH was detected with a ther-
3
[
[
1] a) M. Chen, J. Xu, Y. M. Liu, Y. Cao, H. Y. He, J. H. Zhuang, Appl. Catal. A
010, 377, 35–41; b) J. Baek, H. J. Yun, D. Yun, Y. Choi, J. Yi, ACS Catal.
2012, 2, 1893–1903.
2] a) P. Michorczyk, J. Ogonowski, React. Kinet. Catal. Lett. 2003, 78, 41–47;
b) P. Michorczyk, P. Kustrowski, L. Chmielarz, J. Ogonowski, React. Kinet.
Catal. Lett. 2004, 82, 121–130; c) M. Salamanca, Y. E. Licea, A. Echavarria,
A. C. Faro, L. A. Palacio, Phys. Chem. Chem. Phys. 2009, 11, 9583–9591;
d) B. Schimmoeller, Y. J. Jiang, S. E. Pratsinis, A. Baiker, J. Catal. 2010,
274, 64–75.
mal conductivity detector (TCD) and recorded during heating to
7
2
À1
008C at a ramp of 108Cmin .
The reducibility of the mixed In–Ga–Al oxide catalysts was investi-
gated by H -TPR with the above-mentioned equipment. Approxi-
mately 150 mg sample was loaded in a quartz tube. A water bake-
off pretreatment was applied before recording the TPR signal. Spe-
cifically, the sample was heated to 1508C for 1 h in a 50 sccm Ar
2
ChemCatChem 2016, 8, 214 – 221
www.chemcatchem.org
220
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
[
3] a) B. Y. Jibril, S. M. Al-Zahrani, A. E. Abasaeed, R. Hughes, Catal. Commun.
003, 4, 579–584; b) J. H. Blank, J. Beckers, P. F. Collignon, F. Clerc, G.
7817–7825; c) J. Ye, C. Liu, D. Mei, Q. Ge, ACS Catal. 2013, 3, 1296–
1306.
2
Rothenberg, Chem. Eur. J. 2007, 13, 5121–5128; c) J. Beckers, G. Rothen-
berg, Green Chem. 2010, 12, 939–948; d) C. A. Carrero, R. Schloegl, I. E.
Wachs, R. Schomaecker, ACS Catal. 2014, 4, 3357–3380; e) X. Rozanska,
R. Fortrie, J. Sauer, J. Am. Chem. Soc. 2014, 136, 7751–7761; f) Z. Zhai, X.
Wang, R. Licht, A. T. Bell, J. Catal. 2015, 325, 87–100.
[11] a) Y. Liu, Z. H. Li, J. Lu, K. N. Fan, J. Phys. Chem. C 2008, 112, 20382–
20392; b) M. Chen, J. Xu, F. Z. Su, Y. M. Liu, Y. Cao, H. Y. He, K. N. Fan, J.
Catal. 2008, 256, 293–300; c) M. Chen, J. Xu, Y. M. Liu, Y. Cao, H. Y. He,
J. H. Zhuang, K. N. Fan, Catal. Lett. 2008, 124, 369–375; d) M. Chen, J.
Xu, Y. Cao, H.-Y. He, K.-N. Fan, J.-H. Zhuang, J. Catal. 2010, 272, 101–
108.
[
[
4] a) M. Chen, J. L. Wu, Y. M. Liu, Y. Cao, K. N. Fan, Catal. Commun. 2011,
1
2, 1063–1066; b) T. Shishido, K. Shimamura, K. Teramura, T. Tanaka,
[12] a) B. Zheng, W. M. Hua, Y. H. Yue, Z. Gao, J. Catal. 2005, 232, 143–151;
b) B. J. Xu, B. Zheng, W. M. Hua, Y. H. Yue, Z. Gao, J. Catal. 2006, 239,
470–477; c) J. Wang, F. Zhang, W. M. Hua, Y. H. Yue, Z. Gao, Catal.
Commun. 2012, 18, 63–67.
[13] S. Tan, L. B. Gil, N. Subramanian, D. S. Sholl, S. Nair, C. W. Jones, J. S.
Moore, Y. Liu, R. S. Dixit, J. G. Pendergast, Appl. Catal. A 2015, 498, 167–
175.
[14] T. Wang, P. V. Radovanovic, Chem. Commun. 2011, 47, 7161–7163.
[15] J. Gao, L. Wang, K. Kan, S. Xu, L. jing, S. Liu, P. Shen, L. Li, K. Shi, J. Mater.
Chem. A 2014, 2, 949–956.
[16] M. Saito, S. Watanabe, I. Takahara, M. Inaba, K. Murata, Catal. Lett. 2003,
89, 213–217.
Catal. Today 2012, 185, 151–156; c) J.-L. Wu, M. Chen, Y.-M. Liu, Y. Cao,
H.-Y. He, K.-N. Fan, Catal. Commun. 2013, 30, 61–65.
5] a) R. Grabowski, Catal. Rev. 2006, 48, 199–268; b) F. Cavani, N. Ballarini,
A. Cericola, Catal. Today 2007, 127, 113–131; c) J. J. H. B. Sattler, J. Ruiz-
Martinez, E. Santillan-Jimenez, B. M. Weckhuysen, Chem. Rev. 2014, 114,
1
0613–10653.
6] a) B. M. Weckhuysen, R. A. Schoonheydt, Catal. Today 1999, 51, 223–
32; b) M. S. Kumar, N. Hammer, M. Ronning, A. Holmen, D. Chen, J. C.
[
[
2
Walmsley, G. Oye, J. Catal. 2009, 261, 116–128.
7] a) O. A. Bariås, A. Holmen, E. A. Blekkan, Catal. Today 1995, 24, 361–
3
64; b) M. van Sint Annaland, J. A. M. Kuipers, W. P. M. van Swaaij, Catal.
Today 2001, 66, 427–436; c) G. Siddiqi, P. P. Sun, V. Galvita, A. T. Bell, J.
Catal. 2010, 274, 200–206; d) L. Nykänen, K. Honkala, ACS Catal. 2013,
[17] A. Gervasini, J. A. Perdigon-Melon, C. Guimon, A. Auroux, J. Phys. Chem.
B 2006, 110, 240–249.
3
, 3026–3030; e) F. Jiang, L. Zeng, S. Li, G. Liu, S. Wang, J. Gong, ACS
[18] S. Tan, X. Sun, C. T. Williams, Phys. Chem. Chem. Phys. 2011, 13, 19573–
19579.
[19] J. J. H. B. Sattler, A. M. Beale, B. M. Weckhuysen, Phys. Chem. Chem. Phys.
2013, 15, 12095–12103.
Catal. 2015, 5, 438–447; f) S.-W. Choi, C. W. Jones, S. Nair, D. S. Sholl,
J. S. Moore, Y. Liu, R. S. Dixit, J. G. Pendergast, AIChE J. 2015, 61, 922–
9
35.
[
[
8] a) J. Schäferhans, S. Gómez-Quero, D. V. Andreeva, G. Rothenberg,
Chem. Eur. J. 2011, 17, 12254–12256; b) E. A. Redekop, V. V. Galvita, H.
Poelman, V. Bliznuk, C. Detavernier, G. B. Marin, ACS Catal. 2014, 4,
[20] O. Beyssac, B. Goffe, J. P. Petitet, E. Froigneux, M. Moreau, J. N. Rouzaud,
Spectrochim. Acta Part A 2003, 59, 2267–2276.
[21] J. J. H. B. Sattler, I. D. Gonzalez-Jimenez, L. Luo, B. A. Stears, A. Malek,
D. G. Barton, B. A. Kilos, M. P. Kaminsky, T. W. G. M. Verhoeven, E. J.
Koers, M. Baldus, B. M. Weckhuysen, Angew. Chem. Int. Ed. 2014, 53,
9251–9256; Angew. Chem. 2014, 126, 9405–9410.
1
812–1824.
9] a) N. M. Schweitzer, B. Hu, U. Das, H. Kim, J. Greeley, L. A. Curtiss, P. C.
Stair, J. T. Miller, A. S. Hock, ACS Catal. 2014, 4, 1091–1098; b) B. Hu,
A. B. Getsoian, N. M. Schweitzer, U. Das, H. Kim, J. Niklas, O. Poluektov,
L. A. Curtiss, P. C. Stair, J. T. Miller, A. S. Hock, J. Catal. 2015, 322, 24–37;
c) B. Hu, N. M. Schweitzer, G. Zhang, S. J. Kraft, D. J. Childers, M. P. Lanci,
J. T. Miller, A. S. Hock, ACS Catal. 2015, 5, 3494–3503.
[
10] a) T. Shishido, H. Kuno, K. Teramura, T. Tanaka, Chem. Phys. Lett. 2012,
39–540, 79–82; b) J. Ye, C. Liu, Q. Ge, J. Phys. Chem. C 2012, 116,
Received: August 16, 2015
5
Published online on November 3, 2015
ChemCatChem 2016, 8, 214 – 221
www.chemcatchem.org
221
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim