Evaluation of crystalline structure and SO2 storage capacity of a series of composition-sensitive De-SO2 catalysts

Article abstract of DOI:10.1016/S1381-1169(02)00515-0

A series of non-stoichiometric magnesium-aluminate solid solution spinels, used as sulfur-transfer catalysts in the fluid catalytic cracking units for SO_x emissions abatement, were prepared by using coprecipitation method and characterized with BET, TGA, AES, XRD and in situ IR techniques. It was found that both the crystalline structures and De-SO_x activities of magnesium-aluminate spinels are very sensitive to catalyst compositions. Several phase domains were produced by changing the mole ratio of alumina to magnesia in the preparation process. Owing to substitution of magnesium or aluminum ions in the spinel structure, that leads to a polarization of Al³⁺ or Mg²⁺ ions to adjacent oxygen lattices, the lattice cell parameter of spinel structure regularly varied with chemical compositions, producing a contraction or expansion effect in the lattice cell, that strongly affects De-SO_x activity. TG analysis showed that during SO₂ oxidative adsorption, most of sulfur species were captured on the surface and some sulfur species were stored in bulk of the solids. The IR and AES results confirmed that both surface and bulk-like sulfates with different H₂-reducibilities were formed on the catalysts, which is in good agreement with results of SO₂ monolayer adsorption measurement. Ten-cycle tests of SO₂ oxidative adsorption and reductive decomposition of the formed sulfate showed that the sample with X_Al = 0.33 is the optimum catalyst, its SO₂ capturing capacity reached 124.4 mg/g.

Full text of DOI:10.1016/S1381-1169(02)00515-0

Journal of Molecular Catalysis A: Chemical 194 (2003) 181–193
Evaluation of crystalline structure and SO storage capacity of a
2
series of composition-sensitive De-SO catalysts
2
a,∗
a
a
b
b
J.A. Wang , L.F. Chen , R. Limas-Ballesteros , A. Montoya , J.M. Dominguez
a
Superior School of Chemical Engineering and Industrial Extraction (ESIQIE), National Polytechnic Institute (IPN),
Col. Zacatenco, 07738 Mexico City, Mexico
Instituto Mexicano del Petróleo, Grupo de Simulación Molecular, Eje Central Lázaro Cárdenas No. 152,
b
Delegación G.A. Madero, 07730 México DF, Mexico
Received 9 May 2002; accepted 22 August 2002
Abstract
A series of non-stoichiometric magnesium-aluminate solid solution spinels, used as sulfur-transfer catalysts in the fluid
catalytic cracking units for SOx emissions abatement, were prepared by using coprecipitation method and characterized with
BET, TGA, AES, XRD and in situ IR techniques. It was found that both the crystalline structures and De-SOx activities of
magnesium-aluminate spinels are very sensitive to catalyst compositions. Several phase domains were produced by changing
the mole ratio of alumina to magnesia in the preparation process. Owing to substitution of magnesium or aluminum ions in
the spinel structure, that leads to a polarization of Al3 or Mg ions to adjacent oxygen lattices, the lattice cell parameter of
spinel structure regularly varied with chemical compositions, producing a contraction or expansion effect in the lattice cell,
that strongly affects De-SOx activity. TG analysis showed that during SO2 oxidative adsorption, most of sulfur species were
captured on the surface and some sulfur species were stored in bulk of the solids. The IR and AES results confirmed that both
surface and bulk-like sulfates with different H2-reducibilities were formed on the catalysts, which is in good agreement with
results of SO2 monolayer adsorption measurement. Ten-cycle tests of SO2 oxidative adsorption and reductive decomposition
of the formed sulfate showed that the sample with XAl = 0.33 is the optimum catalyst, its SO2 capturing capacity reached
+
2+
1
©
24.4 mg/g.
2002 Elsevier Science B.V. All rights reserved.
Keywords: Magnesium-aluminate spinel; Sulfur-transfer catalyst; Sulfur-storage capacity; SO2 emission control; FCC unit
1
. Introduction
cracking in FCC operation had become quite appar-
ent, indicating a shortage of high quality crude oil with
low content of sulfur on a global scale [1,2]. When
the feedstock containing organic sulfur is fed into the
FCC units without pretreatment, about 45–55% of the
total sulfur compounds present in the feed are con-
verted into hydrogen sulfur during the FCC procedure
and 35–45% are still remained in the liquid products,
the rest is deposited in the coke formed on the FCC
catalysts [3]. The sulfur deposits in the coke are trans-
ferred to SOx (more than 90% SO2 and less than10%
In the feedstock of fluid catalytic cracking (FCC)
process, there exist many different types of sulfur-
containing organic compounds, those may poison FCC
catalysts, reduce product quality and cause environ-
mental problems. Since 1996, the importance of resid
∗
Corresponding author. Tel.: +52-5-7296000x55124;
fax: +52-5-5862728.
E-mail address: jwang@ipn.mx (J.A. Wang).
1
381-1169/02/$ – see front matter © 2002 Elsevier Science B.V. All rights reserved.
PII: S1381-1169(02)00515-0

1
82
J.A. Wang et al. / Journal of Molecular Catalysis A: Chemical 194 (2003) 181–193
SO3) when these coke is combusted in the regenera-
tor of FCC units. Nowadays, these SOx emissions are
released out into the air without further treatment in
many countries, causing big environmental problems
such as acid rain.
Refiners presently have three strategies for con-
trolling FCC SOx emissions: feedstock pretreatment,
flue gas scrubbing and catalyst technology. In the last
decade, although many effects were devoted to hy-
drodesulfurization (HDS) before FCC process or to
flue gas treatment (FGT) after the FCC operation,
those would remove most of organic sulfur compounds
from FCC feedstock or SOx from flue emissions of the
regenerator, however, because of big cost mainly due
to expensive equipment and operations, many refiner-
ies so far have not applied HDS or FGT technologies
yet.
An interesting technique for SOx emission control
in the FCC units, so-called sulfur-transfer technique,
has attracted much attention. A number of patents and
scientific articles focusing SOx abatement in the re-
generator in the FCC units by using sulfur-transfer
method have been reported [4–8]. This is to use a co-
catalyst like magnesium-aluminate spinel material to
mix with FCC catalysts for picking-up SOx produced
in the regenerator and then release those sulfur species
from the De-SOx additives in a form of hydrogen sul-
fide in the reactor and stripper. These H2S are dis-
charged from the FCC units and separated from the
products for further treatment in a modified Clause
process, producing elemental sulfur. Because this tech-
nique realizes sulfur transferring from SOx to a useful
elemental sulfur, it is therefore called as sulfur-transfer
technique, and because the De-SOx procedure is car-
ried out within the FCC units, this technique is also
regarded as “in situ” SOx control. It has several attrac-
tive characteristics: (1): little or no capital investment
is required; (2) operation cost is very low, which is
only one-seventh or one-thirtieth of the cost used in
HDS or FGT; (3) reliability is high, equal to that of
the FCC itself; and (4) waste disposal problems are
minimal [9,10].
cation by introducing guest ions like iron, copper and
cerium, into a stoichiometric magnesium-aluminate
spinel to improve it De-SOx activity [3,7,11–15].
In fact, the crystalline structure of stoichiometric
magnesium-aluminate spinel can also be modified by
changing Al/Mg mole ratio in the preparation stage.
In the stoichiometric MgAl2O4 spinel structure, there
are eight molecules in its unit cell with 64 tetrahedral
symmetry sites and 32 octahedral ones. In the perfect
case, magnesium ions occupy eight tetrahedral posi-
tions and aluminum ions occupy 16 octahedral po-
sitions. In the case of magnesium or aluminum-rich,
various magnesium-aluminate solid solution spinels
with a formula such as MgAl2O4·nMgO or MgAl2O4·
mAl2O3 can be formed. The excess aluminum ions,
for instance, may occupy the positions of magne-
sium ions, causing spinel structural distortion. The
investigation towards relationship between crystalline
structure of non-stoichiometric magnesium-aluminate
spinel and its De-SOx catalytic activity is an inter-
esting topic. It was reported that De-SOx activity of
the Mg–Al–O catalysts varied with the Mg/Al mole
ratio, and magnesium-rich samples are superior to the
aluminum-rich ones [16]. However, whether the bulk
structure of the sample involves in the SO2 adsorption
and what kinds of sulfur species are formed on the
catalysts is still a matter of debate.
With an emphasis of effects of ratio of alumina
to magnesia upon the crystalline structure and thus
De-SOx catalytic activities as well as distribution of
the sulfur species from the surface to inner layer of
the samples, herein, we report the experiments con-
cerning SO2 pickup capacity of magnesium-aluminate
solid solution spinels, aiming at establishing rela-
tionship between De-SOx activity, reducibility of
the formed sulfate and crystalline structures of these
catalysts.
2. Experimental
2.1. Catalysts preparation
Magnesium-aluminate spinel materials have been
commercially used as a sulfur-transfer catalyst in
FCC units. However, to develop new generation of the
sulfur-transfer catalyst, studies toward improvement
of catalytic activity of magnesium-aluminate spinel
never cease, most of those focus on structural modifi-
By using coprecipitation method, a series of cat-
alysts with different chemical compositions were
prepared by altering mole ratio of aluminum to
magnesium with Mg(NO3)2·6H2O and NaAlO2
as precursors. As exemplified with stoichiometric

J.A. Wang et al. / Journal of Molecular Catalysis A: Chemical 194 (2003) 181–193
183
magnesium-aluminate spinel, two solutions were re-
spectively prepared by adding 76.8 g Mg(NO3)2·6H2O
and 42.0 g NaAlO2. At a same rate, these solutions
were slowly added into a 2000 ml-container contain-
ing 500 ml of deionized water. During the addition,
the formed slurry was stirred and pH value was ad-
justed at about 9 by using a concentrated nitric acid
solution or 2N NaOH solution. Afterwards, the slurry
was continuously stirred for an hour and then aged
at room temperature over night. The aged slurry was
filtered and washed with deionized water. The filtered
and a two-stage oil diffuse pump. Before adsorption
operation, the sample was heated in the adsorption
cell for 2 h at 300 C with evacuation. In the case
of SO2 oxidative adsorption, a mixture of 500 Pa of
SO2 and 200 Pa of O2 was fed into the IR adsorption
system for 30 min. In the half cycle of reduction pro-
cedure, the sulfated sample was reduced by 500 Pa
H2 at a desired temperature for 10 min.
◦
2.5. SO adsorption and sulfate reduction
2
measurements
◦
cake was dried at 120 C for 10 h and then calcined
◦
at 800 C for 4 h in a furnace in air. The product was
De-SO2 activity measurements were carried out on
a MR–GC–MS system. The reactor was a 1.5 cm o.d.
quartz tube which was placed in a single-zone fur-
nace controlled by a temperature-programmed system
(Model YCC-16). Two chromel–alumel thermocou-
ples encapsulated in a 3 mm o.d. stainless steel sheath,
were positioned 1 cm below the top of catalyst bed
to measure the temperatures. The 1 g of FCC cata-
lysts (supplied by China Jinling Petrochemical Co-
operation) mixing with 3 wt.% De-SO2 catalyst were
ground and the fine materials with 80–160 mesh were
used in the SO2 picking-up experiments.
2
.2. Specific surface area measurements
The specific surface areas of the catalyst samples
were measured by using liquid nitrogen adsorption
method (BET) on an ASAP-2400 apparatus. Before
the adsorption the samples were thermally treated at
◦
3
00 C for 1 h to eliminate adsorbed species.
◦
heated at 700 C under a flue of N2 for 20 min and
then 5 vol.% O2 in N2 was introduced into the reac-
tor for 5 min. A gaseous mixture containing 1.5 vol.%
SO2, 5 vol.% O2 in N2 was fed at 700 C for adsorbing
2
.3. X-ray diffraction analysis (XRD)
◦
X-ray diffraction analysis of the samples were
3
0 min at the rate of 100 ml/min. In the reduction half
performed on a Riguku D-max/IIIB with Cu K␣ ra-
diation. Intensity data were obtained in the 2θ range
between 10 and 70 . All the XRD data were corrected
by using silica as standard sample and therefore the
height displacement effect was eliminated. The lattice
cell parameter was calculated in virtue of the diffrac-
tion line appearing at about 44.0–46.0 which corre-
sponds to (4 0 0) plane of spinel crystal, according to
the following expression:
◦
cycle, the temperature was 500 C and the reduction
time was 10 min; 30 vol.% H2 in nitrogen was used as
reduction gas with a flue rate of 50 ml/min. The cata-
lyst De-SO2 activity was defined as the conversion of
SO2. The error of the experimental data reported in
Table 3 is 0.1%. Since we want to focus our interest
on study of SO2 adsorption and reduction capacity,
the adsorption and reduction time used in the exper-
iments are set longer in comparison with that of the
industrial FCC process.
◦
◦
2
2
2 1/2
a = (h + k + l )
× dh k l
(1)
where a is lattice constant (Å), and dh k l is the dis-
tance between the (h k l) planes.
2.6. Auger electron spectroscopy analysis (AES)
Auger electron spectroscopy analysis was perfor-
2
.4. Infrared (IR) characterization
med in a commercial system (Model PHI550-ESCA/
SAE) equipped with a 5 keV integral electron gun and
a 1 keV sputter-ion gun. All the measurements were
made under the following conditions: beam current,
Ia = 1 ␮A; primary energy, Ip = 3 keV; modulation
IR in situ measurements were carried out on a
HITACHI-270-30 IR spectrometer at different tem-
peratures under vacuum conditions. The adsorption
cell is an H-type quartz tube with KBr windows. This
adsorption–desorption system was on line with a vac-
uum set, which was combined two mechanic pumps
voltage, V
= 3 V; time constant: Rc = 0.03 s and
7
mod
−
the base pressure was about 1.30 × 10 Pa.

1
84
J.A. Wang et al. / Journal of Molecular Catalysis A: Chemical 194 (2003) 181–193
By using the 5 keV integral electron gun and 1 keV
Table 1
Standard experimental condition in TGA for SO2 adsorption and
reduction
sputter-ion gun, it is possible to analyze elemen-
tal compositions of the samples from surface to the
subsurface. The etching rate of argon ion beam is
SO2 oxidative
adsorption
Sulfur species
reduction
7
0–80 Å/min. Elemental compositions of the sam-
◦
Temperature ( C)
700
500
ples, in at.%, were determined by using the elemental
sensitivity factor method with the following formula:
Gases mixture
1.2 vol.% SO2,
30 vol.% H2
20 vol.% O2
Balance gas
Time (min)
Flow rate (ml/min)
N2
30
100
N2
10
50
Ix/Sx
Cx = ꢀ
(2)
Ix/Sx
where Cx is atomic fraction of the element x in the
sample, Ix the normalized AES peak height for ele-
ment x; Sx the sensitivity factor of element x and ꢀ is
the sum of all the elements present in the sample. The
data in Figs. 6 and 7 obtained by this technique are the
average of three measurements by choosing different
particles.
reported in Table 1. The standard error measured by
WRT-2 thermogravimetric analyzer is 0.1 mg.
3. Results and discussion
3.1. Crystalline structures and phase domains
2
.7. Thermogravimetric analysis (TGA) of SO2
of the samples
adsorption–desorption
The XRD diffraction patterns of the samples with
various alumina mole fractions between 0.75 and 0.33
show a single phase corresponding to magnesium-
aluminate spinel. All the XRD spectra are very simi-
lar but with a shift of the perk position as the alumina
mole fraction is changed. Fig. 2 shows the XRD
pattern of the sample with alumina mole fraction
A WRT-2 thermogravimetric analyzer, connected
with a computer, temperature-programmed control
and vacuum treatment system, was used to measure
the weight changes of the sample during the ad-
sorption and reduction processes (Fig. 1). Generally,
1
0–15 mg sample was placed in a quartz sample pan.
The standard experimental conditions of SO2 adsorp-
tion and H2-reduction of the sulfated samples are
X
= 0.7. The three strong diffraction peaks corre-
Al
sponding to the (3 1 1), (4 0 0) and (4 4 0) planes of
Fig. 1. A microbalance analysis system for SO2 adsorption and reduction measurements (1) needle valve; (2) flow meter; (3) stop valve;
4) six-way valve; (5) microbalance house; (6) sample holder; (7) furnace; (8) thermocouple; (9) vacuum system.
(

J.A. Wang et al. / Journal of Molecular Catalysis A: Chemical 194 (2003) 181–193
185
◦
Fig. 2. XRD pattern of the sample (XAl = 0.7) calcined at 800 C
Fig. 3. XRD pattern of a magnesium-rich sample (XAl = 0.2)
calcined at 800 for 4 h. This sample contains a MgO phase and
magnesium-aluminate spinel phase.
for 4 h. Only magnesium-aluminate spinel phase is formed.
a magnesium-aluminate spinel crystal, were clearly
observed, indicating the sample was well crystallized
when they were calcined at 800 C. However, the
where “||” is the boundaries between magnesium-
aluminate spinel crystals and precursor cationic ions.
During synthesis process, many fine crystal cores
of magnesium-aluminate spinel surrounded by many
magnesium and aluminum ions can be formed in the
◦
diffraction line shifts with changing of alumina mole
fraction. Table 2 gives two-theta data of the diffrac-
◦
tion line registered between 44.0 and 46.0 , which
corresponds to a value of d-spacing varying from
◦
solid mixture. Above 600 C, growth of these fine
1
.995 to 2.034 Å. When the alumina mole fraction
crystal cores is rapid since the reactions of spinel for-
mation occur on the boundaries or interfaces of these
crystals [17–21].
was very small (XAl < 0.33), two phases MgAl2O4
and MgO coexisted (Fig. 3).
The formation of magnesium-aluminate spinel with
various compositions can be generally illustrated as
follows:
It is assumed that the formation of solid spinel is
a result of Mg2 or Al ions diffusing into the de-
ficient spinel-type structure of ␥-alumina (Al8/3O4).
Therefore, the formula of solid solution can be written
+
3+
Mg2 ||MgAl O4||Al
+
3+
(3)
2
as MgxAl(
O4 [18]. The stoichiometric spinel
8−2x)/3
corresponds to x = 1, i.e. MgAl2O4. The lowest pos-
sible limit is x = 0 and that corresponds to ␥-Al2O3;
the highest possible limit is x = 4, corresponding to
pure MgO (Mg O ). In the above formula, x is the
Table 2
Data of alumina mole fraction (XAl), two-theta, d-spacing, lattice
cell parameter a and SO2 conversion of magnesium-aluminate
spinels
4
4
◦
2 3
MgO mole number, while, the Al O mole number (y)
XAl
2θ( )
d400 (Å)
Cell
SO2
parameter a (Å)
conversion (%)
is equal to (8 − 2x)/6 that is half of aluminum mole
in the formula of MgxAl O4. Thus, the mole
0
0
0
0
0
0
0
0
0
0
0
0
.91
45.52
45.40
45.24
45.22
45.18
45.20
45.06
44.94
44.76
44.62
44.88
44.92
1.995
2.000
2.007
2.008
2.009
2.009
2.015
2.021
2.028
2.034
2.023
2.020
7.981
8.000
8.030
8.032
8.036
8.035
8.059
8.080
8.112
8.134
8.089
8.082
58.6
(8−2x)/3
.89
.86
.80
.75
.70
.60
.50
.40
.33
.25
.20
2 3
Al
fraction of alumina (Al O ), X can be expressed
as
60.5
61.3
y
(8 − 2x)/6
((8 − 2x)/6) + x
8 − 2x
8 + 4x
X
Al
=
=
=
(4)
65.2
68.6
71.7
73.1
74.3
x + y
The lattice cell parameter (a) obtained from XRD
analysis, as a function of alumina mole fraction XAl,
is reported in Table 2 and plotted in Fig. 4. It is found
that the samples have different limited domains of the
composition on both sides of stoichiometric spinel.
70.0

1
86
J.A. Wang et al. / Journal of Molecular Catalysis A: Chemical 194 (2003) 181–193
Fig. 4. Lattice constants of the samples as function of alumina mole fraction.
When XAl > 0.5, the domains corresponded to a de-
ficient spinel since the total number of cations was
lower than 3. When XAl < 0.5, the domain was re-
lated to a magnesium-aluminate solid solution spinel
in which magnesium ion was excess because the total
number of cations was bigger than 3.
cent oxygen is different, this results in the lattice cell
parameter regularly changing with composition. For
3+
Al ions, the value of positive charge is high and
its radius is small (rAl3+ = 0.0675 nm), its polarizing
power to the adjacent oxygen ion is therefore strong.
In the case of aluminum-rich, the excess Al3 ions
+
In the domains II and III, where XAl was between
.33 and 0.70, the cell parameter a lineally increased
would diffuse into the lattice cell of MgAl2O4 and
occupied some Mg2 vacancies. The strong polariza-
tion to adjacent oxygen ion shortened the distance be-
tween cationic and adjacent oxygen ions, leading to a
contraction of lattice cell.
+
0
from 8.032 to 8.134 Å as alumina mole fraction de-
creased. However, in domain IV (0.70 < XAl < 0.86),
the cell parameter a remained almost unchanged.
When alumina content is very high (XAl > 0.86),
the cell parameter closely pointed to the value of
the lattice parameter of ␥-Al2O3 (domain V). In the
case of magnesium very rich (XAl < 0.33), however,
the cell parameter tended toward the value of the
stoichiometric spinel (domain I).
3+
On the other hand, compared with Al ion, the
positively charged Mg2 ion shows lower chemical
+
valence and larger ion radius (r 2+ = 0.071 nm),
Mg
hence, the ability of Mg2 polarizing to oxygen ions
+
is relatively weaker in comparison with aluminum
ion. In the magnesium-rich case, these excess Mg2
+
As it is well known that in crystals there exists ionic
polarization effect, which is mainly determined by two
factors: (i) ion radius and (ii) ion charges. This polar-
ization can alter the distance between a cationic ion
and its neighbor anion in the crystalline structure by
producing interaction between them and thus the lat-
tice cell parameter. In the magnesia–alumina system,
ions migrated into the lattice cell and occupied alu-
minum vacancies or lattice positions, expanding the
cell volume. The maximum value of lattice parameter
a corresponding to MgAl2O4·MgO is probably a re-
sult of the largest expansion of the lattice cell caused
by the strongest polarization due to the maximum
substitution of aluminum by magnesium ions. This
reasonably interprets the tendency of cell parameter
the polarizing power of Al3 and Mg to its adja-
+
2+

J.A. Wang et al. / Journal of Molecular Catalysis A: Chemical 194 (2003) 181–193
187
changing with chemical composition on both sides of
3.2. De-SO2 activity and H2 reducibility
stoichiometric spinel within the domains II and III.
Free MgO phase can be segregated from the
magnesium-aluminate spinel in the case of magnesium
very rich. When XAl was less than 0.33 (domain I), a
new peak corresponding to magnesia phase appeared
An ideal sulfur-transfer catalyst is the one on
which the formed sulfates are thermally stable under
the regeneration condition but are easily reduced un-
der the reaction condition in FCC process. Thermal
non-stability of the sulfate formed on the De-SO2
catalysts usually results in occurrence of the re-
versible reaction of sulfate formation, as in the case
◦
at around 62.4 , showing that free MgO and spinel
phases coexisted in the sample (Fig. 3). MgO forma-
tion diminishes the replacement of aluminum ions by
magnesium ions in the structure of MgAl2O4, lead-
ing to the lattice parameter of magnesium-aluminate
spinel close to the value of the lattice parameter
corresponding to stoichiometric spinel.
It is noted that in domain IV, the cell parame-
ter remained almost unaffected when alumina mole
fraction changed. Though we could not find a sat-
isfactory explanation for this result, it is assumed
that magnesium-aluminate spinel coexisted with
of alumina used as additives, where SO is produced
2
from the decomposition of Al (SO ) at temperature
2
4 3
◦
700 C in the regeneration condition of FCC units.
The sulfate formed on magnesium-aluminate spinel
has been proved to be stable at 700 C [22]. How-
◦
ever, if sulfates can not be easily removed from the
De-SO additives in the riser reactor and stripper, the
2
active sites for capturing SO may be partially lost
x
and this must shorten lift-time of the catalysts. There-
␥
-alumina in this domain. Since the crystalline struc-
fore, it is important for a high quality sulfur-transfer
ture of magnesium-aluminate spinel and ␥-alumina
have the same atomic coordinates, their XRD patterns
are very similar, that produces difficulty to strictly
distinguish the difference between them. However,
change of the lattice cell parameter reflects its struc-
tural modification or distortion. In this region, a slight
increase in magnesium content might only cause an
increase in amount of magnesium-aluminate spinel.
This is because in the aluminum-rich case, magnesium
ions can be consumed by reacting with the excess
aluminum ions to form more magnesium-aluminate
spinel, which, of course, is deficient. It may be more
deficient than that in domain III but less than that in
domain V. The cell parameter is a result of contri-
butions of formation of both alumina and MgAl2O4
spinel as well as each weight fraction, that finally
leads to the parameter remaining almost constant.
In the domain V (0.86 < XAl < 1.0), aluminum
content is very high. It is probably in favor of the
formation of the spinel having a very deficient struc-
ture similar to ␥-alumina. Therefore, the cell param-
eter decreases with the increasing of alumina mole
fraction and it points toward the value of ␥-alumina.
Or, similar to the explanation for the domain IV, it
might be also interpreted by assuming coexistence of
magnesium-aluminate spinel and ␥-alumina in this do-
main. Since ␥-alumina is the dominant phase, it results
in lattice parameter linearly diminishing with increas-
ing of alumina mole fraction.
catalyst that its SO oxidative adsorption ability must
match to reducibility of the formed sulfate during the
FCC cycles. In order to measure such a compatibility
of different spinel-type catalysts, 10-time cycles of
SO2 oxidative adsorption and reductive desorption
2
were carried out. The SO conversions on different
2
catalysts are reported in Tables 2 and 3.
It is observed from the Tables 2 and 3 that the De-
SO activities of aluminum-rich samples are lower
2
than that of magnesium-rich samples. However, when
the samples were very rich magnesium (XAl < 0.33)
not only De-SO2 activities but also mechanical
Table 3
De-SO2 activities of 10 cycles of SO2 adsorption and reduction
of formed sulfur species on the different magnesium-aluminate
spinels measured by using microreactor system
No. of the
Captured SO2 (%)
reduction–oxidation
cycle
XAl = 0.5
XAl = 0.33
XAl = 0.8
1
2
3
4
5
6
7
8
9
71.7
68.2
68.7
69.1
67.4
66.0
68.2
65.1
63.1
62.9
74.3
70.8
70.2
69.3
70.8
68.8
68.4
70.5
70.4
71.5
61.3
57.4
60.2
56.3
55.9
57.6
58.0
56.4
56.1
55.8
1
0

1
88
J.A. Wang et al. / Journal of Molecular Catalysis A: Chemical 194 (2003) 181–193
strength became poor that would reduce the ability
of attrition resistance. Therefore, a suitable content
of magnesium in spinel is an important factor in the
design of De-SOx catalysts.
All the samples showed relatively high initial ac-
tivities. After two cycles of adsoprtion–reduction op-
eration, the De-SOx activities of the catalyst samples
decreased; then they almost remained at constant after
eight cycles. This is because many active sites exist
on the fresh surfaces, after several cycles of SO2 ad-
sorption, part of formed sulfate species still remained
on the surface or subsurface, poisoning some of ac-
tive sites and thus resulting in partial deactivation. For
which also causes weight loss, should be taken into
account when we evaluate the reducibility. Since the
values of Wr/W for all the samples are less than 1,
i
and after consideration of the weight loss resulting
from the lattice oxygen desorption, we conclude that
some adsorbed sulfur species are still retailed in the
samples after reduction with hydrogen.
3.3. Capacity of SO monolayer adsorption and
2
bulk-storage
The total amount of sulfur dioxide adsorbed on the
catalyst sample could be measured by microbalance
technique as shown above. However, it is interesting
to know how many SO2 were captured on the catalyst
surface. The cross sectional area of SO2 molecules is
the best catalyst, MgAl2O4·MgO (X = 0.33), there
Al
was only an approximate 2.8% loss of the activity in
the 10th oxidation–reduction cycle compared with the
fresh sample.
2
estimated to be 0.1 m /␮mol [23–25]. For calculation
The abilities of SO2 pickup under a fixed temper-
the amount of SO2 monolyer adsorption on the cat-
alyst surface, we assumed that each adsorbed sulfur
dioxide molecule close two-dimensional packing cov-
ers an area that is equal to the value of cross-sectional
area of single SO2 molecule (Ss). According to sur-
face area of the sample and value of Ss, the amount of
SO2 adsorbed in a monolayer on the sample surface
was calculated (Table 5). It is clearly shown for all the
tested samples that the total amount of adsorbed SO2
is more than that corresponding to a monolayer.
Three possible reasons are responsible for the extra-
adsorbed species: (i) the adsorbed oxygen species
since it is a oxidation adsorption process; (ii) multi-
layer adsorbed SO2 on the surface and (iii) SO2
storing in the inner layer of the sample.
◦
◦
ature 700 C and reducibility at 500 C of the differ-
ent catalysts were also measured by using microbal-
ance technique (Table 4). The data in Table 4 show
that the sample with XAl = 0.33 is the most active
sulfur-capturing catalyst, its SO2-capturing capacity
reaches 124.4 mg/g. This result is also in consistent
with the results obtained from microreactor test shown
in Table 3.
When being reduced by hydrogen, the sulfated
samples show some weight loss. The reducing factor,
Wr/Wi, ranges from 0.958 to 0.992. It seems from
Wr/W values that reducibilities of the samples are
i
quite good. However, the desorption of lattice oxy-
gen on the surface of the samples due to the reaction
between lattice oxygen and hydrogen to form water,
For testing the first possibility, we used pure oxygen
as adsorbent to measure weight gain of the samples
Table 4
Data of SO2 oxidative adsorption and reduction of the sulfur
species formed on the different catalysts obtained by using TGA
technique
Table 5
Amount of SO2 monolayer adsorption and storage capacity of SO2
in bulk of the different samplesa
a
2
XAl
Wi (mg)
Wa (mg)
Wr (mg)
Wr/Wi
ꢁW/Wi
XAl
S (m /g)
Wm
Wo
We
Ws
(mg/g)
(mg/g)
(mg/g)
(mg/g)
(mg/g)
0
0
0
0
0
.86
11.44
10.11
14.80
10.05
12.39
12.38
11.04
16.25
11.30
13.73
11.25
9.87
14.34
9.97
0.983
0.976
0.969
0.992
0.958
82.5
92.0
98.0
124.4
108.2
0.86
0.70
0.50
0.33
0.20
116.9
132.6
147.1
169.2
151.6
74.8
84.9
94.1
108.3
97.1
1.6
2.4
3.3
2.7
0.9
7.7
7.1
13.9
16.1
11.8
6.1
4.7
10.6
13.4
10.9
.70
.50
.33
.20
12.94
a
a
Wi: initial weight of the fresh sample; Wa: weight of the
S: Specific surface area; Wm: amount of monolayer adsorbed
◦
sulfated sample at 700 C; Wr: weight of the reduced samples at
SO2; Wo: amount of adsorbed oxygen; We: amount of the extra
weight gain, We = Wa − Wm; Ws: amount of SO2 storing in the
bulk of the sample, Ws = We − Wo.
◦
5
00 C; Wr/Wi: reducing factor, ꢁW = Wa −Wi; ꢁW/Wi: capacity
of SO2 pickup.

J.A. Wang et al. / Journal of Molecular Catalysis A: Chemical 194 (2003) 181–193
189
◦
in the same condition as SO2 oxidative adsorption.
It was found that weight gain for all of the samples
is less than 3.5 mg/g. These results indicate that oxy-
gen adsorption may be responsible for part of the
extra weight gain when the samples adsorb SO2 in
the presence of oxygen. The data regarding oxygen
adsorption are also reported in Table 5. However, it is
noteworthy that SO2 adsorption procedure is an oxi-
dation adsorption process, the adsorbed oxygen may
oxidize the adsorbed SO2 to SO3 and form sulfate,
temperature is as high as 700 C. Existence of multi-
layer SO2 is only possible in the low temperature due
to very weak bond between the absorbents.
After getting rid of the first two possibilities, it is
postulated that the extra weight gain (We) was caused
by the storage of SO2 in some form in inner layer of
the samples. Capacity of SO2 storage in the bulk in
each sample is presented in Table 5.
The data in Tables 4 and 5 show that most sul-
fur species were adsorbed on the surface of samples
and less than 15 wt.% of sulfur adsorbed species were
“stored” in bulk. Also it was found that SO2 storage
capacity in magnesium-rich sample (XAl = 0.2) is
higher than that in aluminum-rich samples (XAl = 0.7
and 0.86). The sample with XAl = 0.33 had the largest
ability to store the SO2 in the bulk, which might be
related to its special structure, since this sample has
the biggest lattice cell volume and the largest surface
for example, when SO2 adsorbs on Mg2 , MgSO4
+
may be formed based on the Eqs. (5) and (6):
1
SO2 + O2 = SO3
(5)
(6)
2
SO3 + MgO = MgSO4
Therefore, the adsorbed oxygen had been taken into
account when we measured the weight gain during
SO2 oxidative adsorption. The amount of adsorbed
oxygen molecules, which are not involved in sulfate
2
area (169.2 m /g). Moreover, SO adsorption on base
2
magnesium-aluminate spinel may be an acidic–basic
reaction process, its capacity may also associate to ba-
sicity and surface area of the solids. A plot concerning
the relationship among the SO adsorption capacity,
catalyst basicity and surface area is shown in Fig. 5.
◦
formation at 700 C, is very small. Its existence does
not produce significant effect on weight gain.
The second possibility is generally exclusive since
it conflicts with our adsorption experiments where the
2
Fig. 5. Sorption capacity for SO2, basicity and surface area of the catalysts as function of alumina mole fraction. Sorption capacity of
2
−1
.
SO2, mg/g; surface area, m /g; basicity, CO2 ␮mol/g × 10

1
90
J.A. Wang et al. / Journal of Molecular Catalysis A: Chemical 194 (2003) 181–193
It is found that both sorption capacity for SO2 and
surface area first increases as alumina mole fraction
increases from 0.2 to 0.33, and then dramatically de-
crease as XAl declining to 0.86. However, the change
tendency of the basicity as function of XAl follows
some different way, indicating that SO2 oxidative ad-
sorption is not a strict base–acid reaction. These re-
sults show that the capacity of SO2 picking-up of the
catalysts is closely related to not only Al/Mg compo-
sition but also surface area.
are completely removed, while, as the sputter time in-
creased, the sulfur concentration gradually increased,
and at 200–210 Å (after 3 min of Ar etching), a max-
+
imum value was achieved, and then sulfur concentra-
tion diminished with the increment of depth. These
results show that sulfur species formed on surface or
subsurface can be easily removed, however, reduction
of the sulfur species in the bulk is more difficult.
The maximum value of the residual sulfur species
in the depth range of 200–210 Å is likely caused by
comprehensive reasons: on the one hand, the amount
of SO2 adsorption decreased as the depth increased,
and in the interior region, which may be near the cen-
ter of particle, it was much less; on the other hand,
reduction of the sulfur species on the surface was eas-
ier than that on the inner surface. Consequently, after
reduction by hydrogen, on the surface and the inner
layer at greater depth, the amount of residual sulfur
species was less than that presenting in the range of
200–210 Å depth.
3
.4. Distribution of sulfur species in depth along
the catalyst particles
AES technique was applied to measure the sulfur
species distribution in depths in the samples by using
argon ion as sputter gun. Sulfur atom concentrations
+
in the samples as a function of Ar sputter time that
is proportional to depth of the particle, are shown in
Figs. 6 and 7, respectively.
Sulfur atom concentration in the sulfated samples
reached a top value on the surface and then it decreased
from the outer to inner layer (Fig. 6). After being re-
3.5. Formation of surface sulfate and bulk-like
sulfate
◦
duced by H2 at 500 C, some residual sulfur species
located at different depths were still observed (Fig. 7).
On the surface and subsurface, all the sulfur species
Although both microbalance and AES analysis
confirm sulfur species being stored in the bulk or
+
Fig. 6. Sulfur content as a function of Ar sputter time for the various sulfated samples.

J.A. Wang et al. / Journal of Molecular Catalysis A: Chemical 194 (2003) 181–193
191
+
Fig. 7. Sulfur content as a function of Ar sputter time for the various reduced samples.
−
1
inner layer of the samples, however, some questions
naturally arise: in which form the sulfur species were
stored in the surface or bulk and how about their
thermal stability and H2-reducibility? To investigate
this, the sample with XAl = 0.33 was characterized
by in situ IR technique (Fig. 8). Five absorption
bands were observed when the sample was exposed
in the mixture of SO2 and O2 in the IR adsorption
and 1140 cm
1189 cm
(weakly chemisorbed); 1255 and
−
−
1
1
(chemisorbed on acidic sites) and
1060 cm (strong chemisorbed species) [26]. Waqif
et al. observed an intense IR absorption band in the
−
1
region 1100–1200 cm
on magnesium-aluminate
spinel sample and it was ascribed to bulk or ion-like
sulfate [22]. Similarly, bulk-like sulfate was found
on MgO and ␥-Al2O3 samples during SO2 oxidative
adsorption [25,27]. Referring to IR character of the
sample impregnated with ammonia sulfate [28–30],
the above bands were assigned to asymmetric (νa)
and symmetric (νs) O=S=O and O–S–O stretching
vibration of surface and bulk-like sulfates. It is rea-
sonable that during adsorption, SO2 weakly adsorbed
on some active sites as physical adsorption species
on the one hand and strongly bounded sulfur-species
reacted with oxygen lattices of the catalysts to form
surface sulfate or bulk-like sulfate on the other hand
[24]. The formation of bulk-like sulfate MgSO4 on
magnesium-aluminate spinel is also confirmed by
XRD analysis [3,7]. We here therefore, assigned the
◦
cell at 500 C for adsorption 30 min. These bands,
respectively appeared at 1400, 1300, 1190, 1075 and
−1
9
50 cm . After 60 min of adsorption, the IR spectra
remained unchanged, indicating that the adsorption on
this sample reached saturation under this condition.
−
1
The band at 1300 cm rapidly disappeared when the
adsorption system was evacuated up to 5 Pa, however,
the intensities of other bands remained more or less
unaffected (not shown here). This reveals that the
−
1
band at 1300 cm was produced by weakly adsorbed
sulfur species. With respect to thermal stability, the
−
1
bands at 1400, 1075, 1190 and 950 cm are assigned
to the strong bounded sulfur species on the samples.
Various sulfur bands formed on different oxides
exposed under SO2 atmosphere have been reported.
Datta et al. observed five different sulfur species
−1
more thermal stable bands at 1075 and 1190 cm to
bulk-like sulfates.
It was reported that when the sulfur complexes can
be formed through O– or S–metal bond, depending on
−
1
formed on the ␥-Al2O3: 1334 and 1148 cm
physically adsorbed species on OH groups); 1322
−
1
(
the value of (νa −νs): if νa −νs > 190 cm , it may be

1
92
J.A. Wang et al. / Journal of Molecular Catalysis A: Chemical 194 (2003) 181–193
lysts move into the riser reactor of FCC unit, they still
◦
remain some temperature between 700 and 500 C for
a few seconds, this is very important since the bulk-like
sulfates can be removed at relatively high temper-
ature during this short residence time. In compari-
son with stoichiometric magnesium-aluminate spinel,
MgAl2O4·MgO has larger SO2 picking-up capacity
and higher ability to reduce the formed sulfate [30].
4. Conclusions
The crystalline structures and De-SO2 activities of
a series of magnesium-aluminate solid solution spinel
catalysts were found to be very sensitive to chemical
compositions. A lattice cell expansion or contrac-
tion, occurring in the structures of magnesium- or
aluminum-rich samples, was determined, it was a
result of polarization effect of Al3 ions or Mg
+
2+
ions to its adjacent lattice oxygen ions in the crys-
talline structure of MgAl2O4. The catalytic properties
of magnesium- or aluminum-rich catalysts remark-
ably differed from that of stoichiometric spinel. The
similar trends of both of lattice cell parameters and
De-SO2 activity varying with the catalyst composi-
tion indicate that the crystalline structure of catalyst
is one of the key factors to affect catalytic activity.
SO2 was captured on all the samples in an amount
more than that of a monolayer. The concentration of
the sulfur species distributing in depth of the sample
decreased from surface to central region. Most of sul-
fur species were formed on surface, which could be
easily removed during hydrogen reduction procedure
Fig. 8. IR spectra of the sample (XAl = 0.33) treated with SO2
and O2 and then being reduced with H2 at different conditions.
(
a–b) After the sample was exposed in a mixture of 500 Torr SO2
◦
and 200 Torr O2 for 30 min at 500 C (a) and 60 min (b); (c–e)
after the adsorption system was evacuated to 20 Pa at 500 C, a
◦
steam of 500 Torr 100% H2 was fed into the adsorption cell, IR
◦
spectrum was recorded at 500 C after the sample being reduced
for 3 min (c), 30 min (d) then the temperature was increased up
◦
to 600 C (e).
O–M bond; however if νa − νs < 190 cm 1, the S–M
bond may be formed [26]. Based on these argument,
in our cases sulfate species was probably formed via
S–M bond since the value of νa − νs was less than
−
◦
at 500 C. However, some of sulfur species was stored
−1
1
90 cm
.
in sulfate in the bulk that resisted to H2-reduction up
◦
When the sulfated sample was treated with hydro-
to 600 C. The formation of both surface and bulk-like
◦
gen at 500 C for 3 min, all the bands were also re-
markably reduced in intensity. After 30 min reduced,
sulfates on the catalysts enhanced the capacity of SO2
pickup.
–
1
both the 1400 and 950 cm
bands disappeared and
−
1
the intensities of 1065 and 1200 cm bands were fur-
ther reduced. As the reduction temperature increased
Acknowledgements
◦
−1
to 600 C, the 1065 and 1200 cm bands were al-
most disappeared. These observations reveal that all
surface sulfates could be reductively decomposed at
moderate temperature but the bulk-like sulfates, which
are only partially removed at the reaction tempera-
J.A. Wang would like to thank the financial
support from CONACyT-31282-U (Mexico) and
FEIS-98-29-III (IMP-IPN-Mexico) and a scientific ad-
visor fellowship of the Proyect-IMP-D.1234.03.012.
L.F. Chen thanks the fellowship of visiting profes-
sor supported by the CONACyT and FIES projects
◦
ture (500 C). Fortunately, in the industrial practice,
when the De-SOx catalysts together with FCC cata-

J.A. Wang et al. / Journal of Molecular Catalysis A: Chemical 194 (2003) 181–193
193
in the Instituto Plitecnico Nacional de Mexico. The
technical assistance from Professor H. L. Zhou for
the AES analysis and from Ms. Y. F. Chen for the IR
experiments is greatly appreciated.
[13] J.A. Wang, Z.L. Zhou, C.L. Li, J. Mol. Catal. 139 (1999)
15.
3
[
[
14] J.A. Wang, O. Novaro, C.L. Li, J. Mater. Sci. 33 (1998) 3671.
15] J.A. Wang, L.F. Chen, C.L. Li, React. Kinet. Catal. Letts. 1
(1998) 73.
[16] J.S. Yoo, A.A. Bhattacharyya, C.A. Radlowski, Ind. Eng.
Chem. Res. 30 (1991) 1444.
References
[17] O. Salas, H. Ni, V. Jayaram, K.C. Vlach, C.G. Levi, R.
Mehrabian, J. Mater. Res. 6 (1991) 1964.
[
1] J.P. Verlaan, P. O’Connor, H.W.M. Sonnemans, Y. Inoue, in:
Proceedings of the 15th World Petroleum Congress, Beijing,
China, 1997, p. 1.
[18] Y.M. Chang, W.D. Kingery, J. Am. Ceram. Soc. 72 (1989)
271.
[19] C.C. Wang, J. Appl. Phys. 40 (1969) 3433.
[20] G. Yamagchi, M. Nakano, M. Tosaki, Bull. Chem. Soc. Jpn.
42 (1969) 2801.
[2] B.A. Lerner, Hydrocarbon Eng. 3 (1998) 26.
3] A. Corma, A.E. Palomares, F. Rey, Appl. Catal. B 4 (1994)
[
1
69.
4] R.J. Bertolacinl, E.H. Hirschberg, F.S. Modica, US Patent
,432,019 (1983).
5] J.S. Yoo, F. Lossamoor, J.A. Jaecker, US Patent 4,492,678
1985).
[21] A. Navrotski, B.A. Wechsler, F. Siefert, J. Am. Ceram. Soc.
69 (1986) 418.
[22] M. Waqif, O. Saur, J.C. Lavalley, Y. Wang, B.C. Morrow,
Appl. Catal. 71 (1991) 319.
[23] P.H. Emmett, S. Brunauer, J. Am. Chem. Soc. 59 (1937)
1553.
[24] S.W. Nam, G. Gavalas, Appl. Catal. 55 (1989) 193.
[25] S.W. Nam, G. Gavalas, Appl. Catal. 74 (1991) 53.
[26] A. Datta, R.G. Gavell, R.W. Tower, Z.M. George, J. Phys.
Chem. 89 (1985) 443.
[
[
4
(
[
[
6] L.M. Magnabosco, E.J. Demmel, US Patent 5,108,979 (1992).
7] J.S. Yoo, A.A. Bhattachryya, C.A. Radlowski, J.A. Karch,
Ing. Eng. Chem. Res. 31 (1992) 1252.
[
8] E. Astorino, G. Busca, G. Ramis, R.J. Willey, Catal. Lett. 23
(1994) 353.
[9] F.A. Pettersen, W.A. Blanton, National AIChE Meeting,
Session No. 37, 26 August 1986, Boston, MA, USA.
[27] M. Bensitel, M. Waqif, O. Saur, J.C. Lavalley, J. Phys. Chem.
93 (1989) 6581.
[
[
[
10] D.P. MxArthur, H.D. Simpson, K. Baron, Oil Gas J. (1983)
[28] M. Waqif, O. Saur, J.C. Lavally, G. Centi, J. Phys. Chem. 95
(1991) 4051.
[29] T. Yamaguchi, T. Jin, K. Tanabe, J. Phys. Chem. 90 (1986)
3148.
7
0.
11] J.S. Yoo, A.A. Bhattacharyya, C.A. Radlowski, Appl. Catal.
B: Environ. 1 (1992) 169.
12] J.A. Wang, L.F. Chen, C.L. Li, J. Mol. Catal. 139 (1999) 31.
[30] J.A. Wang, C.L. Li, Appl. Surf. Sci. 161 (2000) 406.