You are on page 1of 15

ARTICLE

pubs.acs.org/EF

Interaction of Iron Oxide with Alumina in a Composite Oxygen Carrier


during the Production of Hydrogen by Chemical Looping
Piran R. Kidambi, Jason P. E. Cleeton, Stuart A. Scott,*, John S. Dennis, and Christopher D. Bohn

Department of Engineering, University of Cambridge, Trumpington Street, Cambridge, CB2 1PZ, United Kingdom
Department of Chemical Engineering and Biotechnology, University of Cambridge, Pembroke Street, Cambridge, CB2 3RA, United Kingdom
ABSTRACT: Solid solutions of hematite (-Fe2O3) and corundum (-Al2O3) have been synthesized by coprecipitation. The
resulting particles have been used as oxygen carriers for the production of hydrogen by chemical looping and characterized using
X-ray diraction (XRD), temperature programmed reduction (TPR), specic surface area measurements (BET), scanning electron
microscopy (SEM), and energy dispersive X-ray spectroscopy (EDXS). The particles were repeatedly (i) reduced with, e.g., CO to,
nominally, Fe, (ii) then oxidized with steam to Fe3O4 to produce hydrogen, (iii) then returned to Fe2O3 by oxidizing with air. The
optimum loading of Al2O3 in the composite particles was found to be 25 wt % for the production of hydrogen over 50 cycles,
resulting in an average yield (mole H2 formed/theoretical amount possible on reduction to Fe and oxidation to Fe3O4) of hydrogen
of 48%. It was found that although Al2O3 is often thought of as inert, it participates in the oxidation and reduction reactions by
forming FeAl2O4 and various solid solutions with the iron oxides. This behavior has been explained with the help of phase diagrams,
and the applicability of these particles for the production of hydrogen by chemical looping is discussed.

Further oxidation of the iron oxide with air would generate


heat.

1. INTRODUCTION
Carbon capture and storage (CCS) is one option for mitigating climate change caused by anthropogenic emissions of CO2.1
Chemical looping combustion (CLC) oers the advantage of
combusting the fuel (e.g., coal, natural gas) and inherently
producing a pure steam of CO2, without the energy penalties
arising from having to separate the N2 from CO2 in the ue gas as
with conventional combustion in air.2 The oxygen needed for
combustion of the fuel in CLC is provided by a metal oxide,
known as the oxygen carrier, which undergoes repeated cycles of
reduction and oxidation. Oxygen carriers based on iron oxide are
attractive for commercial CLC applications owing to the low cost
of material, its wide availability, and the favorable thermodynamics when used in looping reactions.35 In a CLC process,
coal would be gasied with steam to generate synthesis gas, which
is then oxidized using iron oxide, to produce CO2 and H2O. The
water from this process could be condensed to produce a pure
stream of CO2 for sequestration.
Fe2 O3 CO h 2FeO CO2

2Fe3 O4 1=2O2 h 3Fe2 O3


 237:2kJ=mol

H850C  0:3kJ=mol

H850C  33:3kJ=mol

2
The reduced iron oxide could then be oxidized in steam to
produce a stream of H2 free from contamination by CO or CO2.
3FeO H2 Og h Fe3 O4 H2

H850C  47:1kJ=mol

3
3Fe 4H2 Og h Fe3 O4 4H2

Received: June 10, 2011


Revised:
October 25, 2011
Published: December 19, 2011

H850C  105:3kJ=mol

4
r 2011 American Chemical Society

It should also be noted that variants of this process are


sometimes referred to as steamiron processes58 patented as
early as 1910 by Messerschmitt.9 Then the emphasis was on the
production of hydrogen, rather than carbon sequestration.
Reaction 5 is essential for ecient carbon sequestration, since
only the reduction of Fe2O3 to Fe3O4 can achieve complete
conversion of the synthesis gas to CO2 and H2O.10 However,
further reduction to FeO and/or Fe is required to enable the
production of H2 via reactions 3 and 4. In practice, to achieve
both the goals of carbon sequestration and production of H2, a
multistage uidized bed or a moving bed reactor11 would be
required. Alternatively, the synthesis gas could be fed into a
packed bed reactor, whereby separate reaction fronts between
discrete phases of Fe2O3, Fe3O4, FeO, and Fe may exist.7 For
these chemical looping processes to be feasible on a large scale,
the metal oxide must be able to withstand many cycles of
oxidation and reduction. Pure iron oxide deactivates after only
a few cycles when reduced completely to Fe, but is reasonably
stable when reduced to FeO.5,7 Because the reduction to Fe
produces 4 times the H2 compared to reduction to FeO, it is
desirable to develop Fe-based oxygen carriers which are stable
over many cycles when reduced to Fe in CLC. One way of
achieving this is to mix the iron oxide with a supporting material
to prevent sintering and produce a higher yield of H2 over many
cycles. Several supports have been investigated to stabilize iron
oxide12,13 e.g., Al2O3, ZrO2, and TiO2 as well as SiO213 and

1
Fe2 O3 3CO h 2Fe 3CO2

H850C

603

dx.doi.org/10.1021/ef200859d | Energy Fuels 2012, 26, 603617

Energy & Fuels

ARTICLE

MgAl2O4.12 Doping pure iron oxide particles with materials such


as Al2O3 and Cr2O3 has also been found to be benecial.14
Furthermore, these supports may prevent the agglomeration
of particles, which may ultimately lead to deuidization of a
uidized bed.
Kierzkowska et al.15 and Li et al.5 have synthesized stable
particles containing 60 wt % Fe2O3 and 40 wt % Al2O3 using a
solgel method. In Kierzkowska et al.s15 experiments, the
particles were used for the steamiron reaction and gave stable
yields of H2 over 40 cycles when the reducing gas was a mixture of
CO and N2 derived from cylinders.15 However, the solgel
process is lengthy and uses expensive reagents; also the particles
can be friable. Hence, there is an incentive to develop a simple,
cheap, and scalable process for the synthesis of iron-based
particles to be used in chemical looping for the production of
hydrogen.
There is evidence to show that alumina in admixture with
Fe2O3 is not chemically inert. For example, Ishida et al.16
reported the formation of solid solutions of hematite and
corundum when powders of Fe2O3 (25 wt %) and Al2O3 (75
wt %) were mixed and ground together and then calcined at
various temperatures from 800 to 1370 C for 10 h to form
aggregates. The mechanical strength of the sieved aggregates
(average size 70 m) increased with an increase in the content
of corundum, which appeared to act as a binder. The particles
were found to be suitable as a solid looping material for CLC with
methane.16
The objective of this work is to optimize the performance of
composite particles consisting of mixtures of Fe2O3 and Al2O3
synthesized by coprecipitation and to study the interaction of the
iron oxide with alumina during oxidation and reduction.

crushed using a pestle and mortar and sieved to form particles in


the sieve size range 300425 m.
2.2. Particle Characterization. The specific surface area
(SSA) of the samples was calculated from N2 adsorption measurements (Micromeritics, Tristar 3000) using the BET model, and
the cumulative pore volume for pores between 1.7 and 300 nm in
diameter was estimated using the BJH model. The adsorption
average pore diameter (dpore) was calculated from 4V/S, where V
is the total volume absorbed and S is the BET surface area. Images
of the particles were taken with scanning electron microscopes
(JEOL JSM 5800LV and JEOL 6340 F) after the particles were
mounted on double-sided carbon tape and coated with platinum
(Emitech K575). Energy dispersive X-ray spectroscopy (EDXS)
(Leica Stereoscan 430) was used to map the elemental compositions in the particles. A Philips diffractometer (model PW1820,
Cu KR, 40 kV and 40 mA, step size 0.05 s1, dwell time of 4 s,
receiving and antiscatter slits 0.5 mm, divergence slit 0.5 mm) was
used to measure X-ray diffraction patterns in the range 10 < 2 < 80.
The patterns were analyzed with Topas 3, EVA, and Xpert High
score plus software. The compressive load to failure of the asprepared particles was measured by crushing a single particle at a
time using an Instron 5500 R 6025 machine.
2.3. Characterization of Reactivity, Extent of Reaction, and
Performance of the Particles during Cycling. All experiments
were performed at 1 atm pressure and all volumetric gas flows
are reported as normalised to 1 atm and 293 K.
A thermogravimetric analyzer (Mettler Toledo TGA/DSC 1)
was used to perform temperature programmed reduction on the
particles. The temperature programmed reduction in H2 was
performed using 50 mL/min of 5 vol% H2 in N2 (Spectrashield
BOC) in the sample chamber, with a balance purge of 10 mL/min
of N2. It was also possible to switch from the H2/N2 mix to
cylinder air (Air Liquide). During a TPR experiment, a sample of
10 mg was placed in a clean alumina crucible, and rst heated to
400 C (at 10 C/min) in air, where it was held for 10 min before
being cooled to 200 C (at 10 C/min). During this stage of the
experiment, water and CO2 that had been picked up by the
sample from the laboratory air were driven o. The air was then
swapped for the H2/N2 mix, and, after sucient time to remove
all the air from the sample chamber, the sample was heated to
1050 at 10 C/min, then held at 1050 C for 2 h. Prior to each
experiment, the crucible was heated to 1000 C in air for 1 h to
remove any contamination from the furnace or crucible; a blank
experiment (using the same crucible as in the experiment) was
then performed, followed by heating the crucible to 1000 C in
air for 1 h. Finally, the mass loss curves from the blank experiment
were subtracted from the experiment. The crucibles were cleaned
in hot, concentrated HCl between experiments.
Some experiments were also performed using mixtures of CO
and CO2 as the reducing gas within the TGA. The reducing gases
were mixed from cylinders containing 20 vol% CO2 in N2 (BOC,
accuracy (5%) and 10 vol% CO in N2 (BOC, accuracy (5%) via
rotameters, before entering the TGA.
To examine the performance of the particles over many cycles
of oxidation and reduction, experiments were also conducted
with the coprecipitated particles in a xed bed reactor. This reactor
(described elsewhere15) consisted of a vertically mounted,
recrystallized Al2O3 (Multi-Lab Ltd., >99 wt % purity) tube
(10 mm I.D.). The distributor plate in the middle of the reactor
consisted of an alumina plate with 4 equally spaced 1.5-mm
diameter holes The xed bed itself contained a layer of 5.5 g
Al2O3 (Boud Mineral, > 99 wt % purity, sieve size +1.4, 1.7 mm)

2. EXPERIMENTAL SECTION
2.1. Synthesis of Particles. Oxygen carriers with varying mass
ratios of Fe2O3:Al2O3 (100:0, 90:10, 80:20, 75:25, 60:40, and
50:50) were prepared using a coprecipitation method (batch size
10 g). For each ratio, a solution containing the required ratio of
Fe:Al was prepared by mixing preprepared solutions of 1 M
aluminum nitrate nonahydrate (Fisher Scientific, purity >98 wt
%) and 1 M iron nitrate nonahydrate (Acros Organics purity >98
wt %) in RO water, in the proportions required to give a solution
with a total cation concentration (i.e., [Al3+] + [Fe3+]) of 1 M.
Dosing pumps (Cole Parmer Instrument Company, Masterflex
7518-00) were used to pump the resulting solution and a separate
solution of sodium carbonate (1 M Acros Organics purity >98 wt
%, in RO water) into a beaker (800 mL, agitated by a magnetic
stirrer at 800 rpm), where they were allowed to mix. The rate of
flow of each solution was controlled accurately to maintain a pH
of 7.8 ( 0.3 in the beaker. This took approximately 15 min. The
resulting precipitate was stirred at 800 rpm for 60 min, then
allowed to settle, carefully separated from the supernatant liquid,
and washed repeatedly by mixing with RO water until the
conductivity of the supernatant liquid was <40 S. Typically, 5
washes with 30:1 volume ratio of water/precipitate slurry were
needed to remove the soluble ions from the precipitates. The
precipitate was recovered as a cake by filtration. All of the above
processes were performed at an ambient temperature of 18 C.
The cake was dried at 110 C in air in an oven for 12 h. After drying,
the precipitate was heated at 10 C/min in a muffle furnace and
calcined for 3 h in air at 900 C. The resulting aggregates were
604

dx.doi.org/10.1021/ef200859d |Energy Fuels 2012, 26, 603617

Energy & Fuels

ARTICLE

Figure 1. XRD patterns for the as-prepared particles. Dotted line shows peaks of hematite and squares show peaks of corundum.

adjacent to the distributor plate, followed by 0.3 g of +300, 425


m carrier particles topped by a layer of 4.5 g of +1.4, 1.7 mm
Al2O3 to preheat the inlet gas, which owed downward through
the bed. The ow rates of the gases were measured using mass
ow meters (AWM5101VN, Honeywell) for pure N2 and air,
and a rotameter for CO. Solenoid valves (Burkert) were used to
perform the gas switching during the experiments. A syringe
pump was used to feed water at 30 mL/h through hypodermic
tubing (0.8 mm I.D.) into an electrically heated chamber at
150 C swept with N2 to produce steam. At the outlet of the
reactor, the gases were passed though two impinger tubes in
series immersed in an ice bath, followed by a tube lled with
CaCl2 to remove any moisture before entering the gas analyzers. The gas composition was determined using nondispersive
infrared analyzers for CO (range 020 vol % and 01 vol %,
ABB, EasyLine) and a thermal conductivity analyzer for H2
(range 030 vol %, ABB, Caldos27). The reactor was placed in
a tubular furnace and maintained at 850 C. A K-type thermocouple (1.5 mm O.D.) was used to measure the temperature of
the particles in the bed. A typical cycle in the xed bed reactor
consisted of the following sequence: 1 min purge with 2 L/min
N2, 10 min reduction in 2 L/min of 10 vol% CO in N2, 1 min
purge with 2 L/min N2, 5 min oxidation with steam in 2 L/min
N2, 1 min purge with 2 L/min N2, and 3 min oxidation with 2 L/
min air.
To allow particles to be quenched and recovered at dierent
stages of reaction, a dierent packed bed reactor operating at 1
bar was used, based on a single stage hot-rod reactor used
elsewhere.17 The modied system used an incoloy (800 HT,
TW Metals, inner diameter of 12 mm) reactor body, with gases
(10 vol% CO in N2, 15 vol% CO2 in N2, and N2) metered into it
using mass ow controllers (Brooks Smart Mass Flow (5850S)
for CO and N2 and Bronkhorst El Flow (F-201CV) for CO2).
The advantage of this system is that the reactor is heated
ohmically by passing an electrical current along the length of
the body of the reactor; however, the reactor is not insulated, so
that the system can be cooled quickly by switching o the
current. Using this system, the 75 wt % Fe2O325 wt % Al2O3

particles were reduced at 850 C using 2 L/min of total gas ow


with CO2/CO ratios of 30, 1.4, and 0.6 for 60 min. After
reduction, they were quenched in 2 L/min of N2 from 850 to
100 C by turning o the power source for heating. Typical
cooling times were of the order of 7 min.

3. RESULTS
The XRD pattern for the as-prepared particles is shown in
Figure 1. The 100 wt % Fe2O3 showed peaks corresponding to
hematite (-Fe2O3, ICSD 43465).18 The pattern for the 90 wt %
Fe2O3 also showed only peaks for hematite, which may be
explained by the solubility of Al2O3 in Fe2O3 being 8.7 wt % at
900 C.19 The as-prepared particles were cooled in a mue
furnace from 900 C, rather than rapidly quenched; thus,
additional free alumina might be expected to precipitate out as
the solubility is known to decrease with decreasing temperature.20
Nevertheless, for the 90 wt % particles it would seem that the
amount of free crystalline alumina in the particles is lower than
the detection limit of the XRD equipment. For samples with
>10 wt % Al2O3, distinct peaks corresponding to corundum
(-Al2O3, ICSD 31545)21 began to appear in the XRD pattern.
The intensity of these peaks increased with an increase in content
of Al2O3. Also, the peaks for hematite shifted to higher angles and
the peaks of corundum shifted to lower angles with the addition
of Al2O3. The main peaks (104) at 2 = 33.2 for hematite and
2 = 35.1 for corundum were used to calculate the crystallite
size using the Scherrer equation. The lattice parameters and
phase compositions calculated using Rietveld analysis and crystallite sizes have been summarized in Table 1, from which it can
be seen that the crystallite size for the 100 wt % Fe2O3 was
146 nm and decreased to 55 nm upon the addition of just 10 wt %
Al2O3. The crystallite size of Al2O3 was more or less constant for
all the samples at 42 nm, except for the sample with 40 wt %
Al2O3. The crystallite sizes for Fe2O3 and Al2O3 in the 60 wt %
Fe2O3 sample were almost the same, i.e., 36 and 35 nm,
respectively. The peak shift observed in the XRD patterns is also
reected in the lattice parameters a and c. The reported a and c
605

dx.doi.org/10.1021/ef200859d |Energy Fuels 2012, 26, 603617

Energy & Fuels

ARTICLE

found surface area and pore volume increased systematically with


the increased content of alumina, and achieved surface areas
similar to those of the particles produced here, at comparable
calcination temperatures.
Ishida et al.16 reported that the strength of particles increased
(as determined by the compressive load to failure) as the content
of alumina was increased. Here the crushing strength of a single
particle containing alumina was found to be higher than that for a
particle containing 100 wt % Fe2O3 ( 2.5 N). The crushing
strength generally increased with alumina to a maximum of
11 N for the 60 wt % material. There is, however, considerable
uncertainty in these values (with the error in excess of (50%),
both due to variability between replicates and the diculty in
identifying compressive load at failure in a single experiment.
The amount of hydrogen produced, in terms of mol hydrogen/g carrier in its oxidized form (i.e., Fe2O3 and Al2O3), was
measured in the packed bed reactor and is shown in Figure 2 for
20 cycles of reaction, with each cycle consisting of reduction in a
mixture of N2 + 10% CO, oxidation in N2 + steam, followed by air
at 850 C (due to irregularities in the steam feed, data for certain
cycles have been omitted). The number of moles of hydrogen
produced per g of oxygen carrier is an eective parameter with
which to compare the performance of particles with varying iron
oxide contents since it is an absolute measure. Evaluation of the
oxygen carriers based on molar yield (mol H2 produced/mol H2
theoretically possible) tends to favor the particles with low Fe2O3
contents, which might have a stable performance but produce
very little hydrogen. For the pure iron oxide, the amount of
hydrogen produced decreased markedly from 0.0148 mol/g in
cycle 1 to 0.0041 mol/g in cycle 10, equivalent to a drop in yield
from 90 to 25% (assuming that all of the iron oxide can be
reduced to Fe, and then subsequently oxidized to Fe3O4 in
steam). Hence, the pure Fe2O3 is not suitable for hydrogen
production over repeated cycles when reduced completely to Fe,
as noted by others.7,15 For the materials containing 90 and 80 wt %
Fe2O3, the quantity of hydrogen produced also drops from cycle
1 to cycle 10. However, the performance does not stabilize after
10 cycles and continues to deteriorate. The particles with 75 wt %
Fe2O3 show the best performance after 20 cycles, with a constant
production of 0.0060.007 mol H2/g of oxygen carrier,
corresponding to a molar yield of H2 of 4856%. The performance of the 60 and 50 wt % Fe2O3 stabilizes after 10 cycles and
reaches a constant value. Hence, the minimum number of cycles
needed for the oxygen carriers to approach a level of constant
performance from cycle to cycle seems to be around 20. During
the oxidation with steam, almost no CO was measured for any of
the samples, indicating that minimal deposition of carbon via the
Boudouard reaction had occurred.
The performance of the carrier containing 75 wt % Fe2O3 and
25 wt % Al2O3 is examined in greater detail in Figure 3. The open
squares are the results from Figure 2 for comparison. The lled
circles show the performance of the particles over 50 cycles. It
can be seen that the performance remains approximately constant after 20 cycles producing an average of 0.006 mols of H2
per g of oxygen carrier (48% H2 yield) between cycle 20 and 50.
Also shown in Figure 3 is the performance of the particles in the
absence of the air oxidation step; it can be clearly seen that the
performance of the particles drops gradually with the number of
cycles and is clearly inferior to the performance with the air
oxidation. This indicates that oxidation by air is necessary to
maintain the performance at a constant value over repeated cycles
as has been reported by others.14,15 The surface morphology of the

Table 1. Summary of Crystallite Sizes and Lattice Parameters


Calculated from XRD Patterns Using Rietveld Analysis
lattice

composition

crystallite

parameters ()

from XRD wt%

size (nm)

Fe2O3 wt%

100
90

5.0351
5.0042

13.746
13.644

80

5.0133

13.667

84.2

52.6

75

5.0158

13.676

76.4

54.1

60

5.0010

13.634

55.6

36.4

50

5.0102

13.665

52.4

33.8

20
25

4.785
4.7850

13.050
13.064

15.8
23.6

42.4
41.8

40

4.7777

13.030

44.4

34.6

50

4.7886

13.062

47.6

42.8

100

146
54.9

Al2O3 wt%
0
10

Table 2. Summary of Specic Surface Area, BJH Cumulative


Volume, Pore Width, and Compressive Strength of the
Freshly-Prepared Particles and Particles Cycled in the Fixed
Bed Reactor
Fe2O3

BET specic
surface area

BJH
volume

dpore

approximate
compressive load

wt%

(m2/g)

(cm3/g)

(nm)

at failure (N)

100

2.9

0.005

2.5

90

14.7

0.09

11

80

10.9

0.08

27

75

10.5

0.11

29

60
50

19.0
25.5

0.18
0.24

21
22

11
5

Fe2O3 wt%

BET specic surface

after hydrogen production

area (m2/g)

cycles in the xed bed reactor

100

0.60 ( 0.1

20 cycles

75

1.21 ( 0.1

20 cycles

75

1.39 ( 0.3

50 cycles

75

0.94 ( 0.3

20 cycles without the air oxidation

values for hematite and corundum are (5.04, 13.75 )18 and
(4.764, 13.01 ),21 respectively. Table 1 shows that the a and c
values for hematite have decreased and those for corundum have
increased, when alumina is added to the particles, commensurate
with the observed shifts in peak. The BET specic surface area,
BJH cumulative pore volume, and absorption average pore width
of the freshly prepared particles are summarized in Table 2. The
specic surface area generally increases with an increase in the
content of Al2O3 in the samples. The cumulative volume (BJH)
also shows a trend similar to the specic surface area. The
adsorption average pore width (dpore) increases from 5 nm for
the 100 wt % Fe2O3 and reaches a maximum of 27 and 29 nm for
the 80 and 75 wt % Fe2O3. Sharkawy et al.22 produced catalyst
particles consisting of Fe2O3Al2O3 by coprecipitation from
nitrates under conditions similar to those used here; they also
606

dx.doi.org/10.1021/ef200859d |Energy Fuels 2012, 26, 603617

Energy & Fuels

ARTICLE

Figure 2. Moles of hydrogen produced per g of oxygen carrier (in fully oxidized state) as a function of the number of cycles for coprecipitated
Fe2O3Al2O3 oxygen carriers.

Figure 3. Moles of hydrogen produced per g of oxygen carrier as a function of the number of cycles for coprecipitated 75 wt % Fe2O3 and 25 wt % Al2O3
oxygen carriers with and without air oxidation.

particles in their fully oxidized state can be seen from the SEM
images in Figure 4: (a) as-prepared 100 wt % Fe2O3, (b) asprepared 90 wt % Fe2O3, (c) as-prepared 75 wt % Fe2O3, and (d) 75
wt % Fe2O3 after 50 cycles. The grain size for the 100 wt % Fe2O3 is
around 150200 nm but drops to around 60 nm for the 90 wt %
Fe2O3 and 50 nm for the 75 wt % Fe2O3 in agreement with the
crystallite sizes calculated in Table 1. The morphology for the asprepared particles appears to be similar, whereas the 75 wt % Fe2O3
particles after 50 cycles exhibit a dierent morphology with loss in
surface area due to sintering of the grains. The sintering of the grains
is also reected in the BET surface area seen in Table 2 for the pure
iron oxide particles falling from 2.9 m2/g initially to 0.6 m2/g after 20
cycles. Some caution is required when interpreting results from N2
absorption experiments which yield such low BET areas (<2 m2/g),

since the instrument used cannot accurately measure such small


volumes absorbed; large errors in the BET area and other measurements derived from the absorption isotherm would be expected.
This is reected by the error bars shown in Table 2, which have been
derived from the range of results from a limited number of replicates.
Given the relative error, it is impossible to say if there is signicant
dierence in the surface area among the samples which have been
cycled in the xed bed. It is, however, clear that in all cases repeatedly
reducing and oxidizing the particles has caused the surface areas to
fall dramatically, to the point where there is little microporosity in
the samples. It should also be noted that measurements of surface
area are heavily biased toward smaller pores, and the BJH model is
unable to account for the volume in pores with a diameter larger
than 300 nm. Optical images of the cross sections of the particles
607

dx.doi.org/10.1021/ef200859d |Energy Fuels 2012, 26, 603617

Energy & Fuels

ARTICLE

Figure 4. SEM images for particles: (a) as-prepared 100 wt % Fe2O3,


(b) as-prepared 90 wt % Fe2O3, (c) as-prepared 75 wt % Fe2O3, and (d)
75 wt % Fe2O3 after 50 cycles.

Figure 6. EDXS maps of (a) Fe, (b) Al, (c) O, and (d) all three
elements for a sectioned 75 wt % Fe2O3 particle after 20 cycles with air
oxidation.

Figure 5. Optical images of cross sections of the 75 wt % Fe2O3


particles (a) after 20 cycles with air oxidation, (b) without air oxidation,
(c) and (d) after 50 cycles with air oxidation. An etch was used to show
the grain structure more clearly in (d).

Thus, it appears as if the oxidation in air is responsible for the


formation of pores and large voids within the particles. At a
higher magnication (Figure 5d), the grain structure within the
particles can be seen, along with pores ranging in size from a few
micrometers through to large voids. In Figure 5c, away from the
surfaces, at the grain level there appear to be two phases,
presumably corresponding to an iron oxide phase and an alumina
phase. XRD patterns for the as-prepared 75 wt % Fe2O3 and
75 wt % Fe2O3 reduced in gas mixtures with PCO2/PCO = 30, 1.4,
and 0.6 are shown in Figure 7. In this case the hot-rod reactor
was used and the particles were recovered after cooling in N2; this
reactor can be cooled quickly to minimize changes in the particles
that might occur during cooling. For the reduction with PCO2/
PCO = 30, the XRD pattern shows a hercynite (FeAl2O4, ICSD
74611)23 and magnetite phase (Fe3O4, ICSD 26410).24 It can be
seen that the peaks have shifted to higher angles with respect to
the pure magnetite phase but are shifted to lower angles with
respect to the pure hercynite phase. Reduction conditions using a
PCO2/PCO = 1.4 also show XRD patterns of a hercynite phase
along with small peaks of the wustite (FeO, ICSD 82233).25 The
peaks for the wustite phase increase in intensity and a mixture of
the wustite phase and the hercynite phase is observed for the
reduction under PCO2/PCO = 0.6. Also shown in Figure 7 are the
XRD patterns of the material containing 75 wt % Fe2O3 after it
had been subjected to cyclic hydrogen production experiments in
the xed bed reactor. The particles after 20 and 50 cycles show
a similar XRD pattern with peaks corresponding to the hematite
and spinel (magnetite and hercynite) phase. The particles recovered
after 20 cycles without air oxidation also show peaks corresponding
to phases of spinel (magnetite and hercynite) and wustite.
Figure 8 shows the mass lost during the temperature programmed reduction of fresh samples of the oxygen carriers, along
with the rate of mass loss (n.b., for clarity, only the results for the
100, 90, 75, and 50 wt % are shown in Figure 8a). During the
degassing phase (not shown), there was a signicant mass loss,
comparable to that expected for the Fe2O3 to Fe3O4 transition,
associated with the loss of water and CO2 (conrmed by FTIR
analysis of the gases leaving the TGA) from the initial decomposition of hydroxides and carbonates formed by exposure to lab

are given in Figure 5. After 20 cycles including the air oxidation some
porosity is visible in Figure 5a; there also appears to be some bulk
segregation of material near the surface of the particle. EDXS
measurements, shown in Figure 6, conrm the outer layer of the
particle had become richer in Fe than the interior. When cycled
without the air oxidation (Figure 5b) no bulk segregation was
observed. With the exception of the particles in Figure 5b, all the
particles were recovered after oxidation in air, and cooled
relatively slowly to room temperature. Some change in morphology could have occurred during cooling; images of freshly
prepared particles which were fully oxidized then cooled much
more slowly in a mue furnace showed no such bulk segregation,
implying that the segregation is caused by the cyclic oxidation
and reduction. With further cycling (Figure 5c), large voids
were observed to form in the particles, with the boundaries
between particles becoming less distinct due to agglomeration.
608

dx.doi.org/10.1021/ef200859d |Energy Fuels 2012, 26, 603617

Energy & Fuels

ARTICLE

is of course a gross oversimplication, and does not take into


account, for example, the solubility of hercynite in magnetite, or
the eect this has on the free energies of each species. In fact,
using the simplied reaction leads to only one signicant error in
the points marked in Figure 8; the mass where wustite has fully
formed and should begin reducing to iron (+) is incorrect,
since, as will be discussed below, some magnetite remains within
the hercynite phase and is not reduced to wustite.
The overall mass loss for samples with 75 wt % Fe2O3 or less
was close to that expected if the complete reduction to Fe was
achieved. Any deviation for the expected mass loss is a reection
of the total experimental error associated with the production of
the particles, and any uncertainties in the thermogravimetric
experiments. If the mass loss curves were taken at face value, then
the iron contents of the nominally 100, 90, 80, and 75 wt %
Fe2O3 carriers would be 101.3, 88.5, 78.5, and 73.1 wt %,
respectively, indicating an error of up to 2%. The 50 and 60
wt % carriers were still a long way from being completely reduced
at the end of the experiment, so it is not possible to estimate the
iron content of these samples (although overall error similar to
the higher wt % carriers would be expected). Some of this error
may arise from the losses of water and CO2, which still remained
after the degassing step (though this was largely eliminated by the
degassing), and, in the case of the sample containing alumina, it is
possible that they were not quite fully reduced.
The curves of normalized mass against temperature for the
materials containing either 100 or 75 wt % Fe2O3 reduced with
gas supplied to the TGA such that PCO2/PCO = 1.2 are shown in
Figure 9. Also shown is the temperature for the transition from
Fe3O4 to Fe0.947O expected for PCO2/PCO = 1.2 according to
thermodynamic data from Barin and Knacke28 and NASA.29 The
concentrations of gases used prevented the reduction of FeO
further to Fe. It can be seen that the 100 wt % Fe2O3 undergoes
two distinct reductions in mass; the rst can be assigned to
reduction to Fe3O4, the second to the formation of wustite, with
a stoichiometry of Fe0.91O. The behavior of the 75 wt % Fe2O3 is
somewhat dierent. In this case, the rst loss in mass is
approximately equal to that for the formation of Fe3O4, with a
slight discrepancy from the theoretical mass loss of 2.5%
probably due to small amount of decomposition from absorbed
water and CO2 when the sample was exposed to the atmosphere.
In fact, this sample had only been degassed at 150 C compared
with 400 C for the TPR experiments, so may still have contained
some bound water and CO2. The loss of mass beyond the
formation of Fe3O4 is rather low and is consistent with the
formation of FeAl2O4 plus a small amount of FexO. The vertical
drop in mass at 950 C corresponds to mass lost when the sample
was held at 950 C after the temperature ramp; after 30 min the
reaction appeared to have nished, but still had not resulted in
complete reduction to FexO.

Figure 7. XRD patterns for the 75 wt % Fe2O3 particles: as-prepared,


reduced in dierent ratios of CO2 and CO, and after hydrogen
production cycles in a xed bed reactor. Dotted lines denote peaks
positions of hematite, lled rectangles denote corundum, open rectangles denote magnetite, lled circles are hercynite, and open circles are
wustite.

air after calcination. For the purposes of the mass balances, the
initial mass was taken to be that obtained after degassing. Points
are marked in Figure 8a where, based on the mass of oxygen lost,
the samples should have been reduced to the species indicated,
assuming that the reaction proceeds sequentially by the reactions
given below
3Fe2 O3 H2 f 2Fe3 O4 H2 O

Fe3 O4 3Al2 O3 H2 f 3FeAl2 O4 H2 O


Fe3 O4

4




3
3
3
H2 f Fex O 4  H2 O
x
x
x

Fex O H2 f xFe H2 O
FeAl2 O4 H2 f Fe Al2 O3 H2 O

8
9
10

4. DISCUSSION
It should also be noted that there is some uncertainty in mass
corresponding to wustite in these calculations since wustite is not
stoichiometric and can vary from x = 0.85 to 0.95,26 with the
equilibrium value being a function of the temperature and the
oxygen partial pressure. For temperatures below 900 C, x
should be 0.90.95, and closer to 0.95 when the wustite is in
equilibrium with Fe.26 A value of x = 0.92 was used in Figure 8
since initial experiments (reported elsewhere27) found this to be
a representative value for 100 wt % samples.27 The above scheme

4.1. Interactions between the Iron and Aluminum Oxides


and Phase Equilibria for the FeAlO System. Muan et al.30,31

have measured the phase equilibrium of the Fe2O3Al2O3


system in air at 1 atm and the stability of the phase of Fe2O3
Al2O3 at temperatures higher than 1000 C. The relationship
among the various phases of oxides of Fe and Al oxides below
1000 C in air consists predominantly of hematite and corundum
and solid solutions of the two.20,30 Furthermore, it has been
suggested that the presence of hematite (-Fe2O3) favors the
609

dx.doi.org/10.1021/ef200859d |Energy Fuels 2012, 26, 603617

Energy & Fuels

ARTICLE

Figure 8. Results from the temperature programmed reduction of the oxygen carriers in 5% H2. (a) Mass loss (only 4 results shown for clarity) and (b)
the derivative of the mass loss curve. Mass remaining after reduction to (O) Fe3O4, () Fe3O4 and FeAl2O4, (+) Fe0.92O and FeAl2O4, (0) Fe and
FeAl2O4, () Fe and Al2O3. A vertical line is marked at 570 C; below 570 C pure wustite will dissociate into magnetite and hematite.

formation of the corundum (-Al2O3) phase3234 at the temperatures considered in this work. The XRD patterns in Figure 1
for the as-prepared particles also indicate that solid solutions
have formed. The shift in the XRD peaks and the corresponding
change in the lattice parameters when alumina is added to the
iron particles can be explained for the formation of substitutional
solid solutions; since the radius of Fe3+ (0.645 )35 is larger than
that of Al3+ (0.535 ),35 substitution of Fe3+ by Al3+ in Fe2O3
would decrease the cell volume and cause the peaks in the XRD
pattern to shift to higher angles and vice versa.34 Tartaj and
Tartaj32 also found that alumina doped with hematite would
form a solid solution with Al3+ ions substituted with Fe3+ (within
the limits of solubility), and that addition of Fe3+ caused the
corundum lattice to expand; the lattice parameters varying
linearly with the amount of hematite present, as has been noted

by others.34 From the XRD results, a decrease in crystallite size


for the Fe2O3 is observed from 146 nm for the 100 wt % Fe2O3 to
34 nm for the 50 wt % Fe2O3, as seen in Table 1. This decrease is
also observed in the SEM images for the 90 wt % Fe2O3 and 75 wt
% Fe2O3 in Figure 4. As reported in Table 2, an increase in
specific surface area with an increase in the mass fraction of Al2O3
in the composite particles suggests that the Al2O3 prevents
sintering of the Fe2O3 during calcination. Although the BJH
cumulative volume of pores also increases with the content of
alumina, the average pore diameter calculated from BET is
largest for the 75 wt % Fe2O3 particles.
To interpret the results presented here, it is helpful to consider
the thermodynamics of the ironaluminumoxygen system.
Elrefaie and Smeltzer19 have constructed the detailed equilibrium
phase diagram for the FeAlO system as a function of partial
610

dx.doi.org/10.1021/ef200859d |Energy Fuels 2012, 26, 603617

Energy & Fuels

ARTICLE

Figure 9. Mass loss curve for 100 and 75 wt % Fe2O3 as a function of


temperature when reduced in a TGA at 10 C/min in CO2/CO = 1.2. At
the end of the experiment, the sample was held at 900 C for 30 min.

pressure of oxygen at temperatures of 1073, 1173, and 1273 K,


based to a large extent on the data and the phase diagrams by
Turnock and Eugster20 for the region involving spinel phases.
Whereas alumina is reasonably soluble in hematite, any wustite
phase which forms is relatively pure, containing only up to
0.55 mol % A12O3.19 Hercynite (FeA12O4) and magnetite are
mutually soluble, leading the formation of solid solutions of the
two.20 In fact, the hercynite phase is only stoichiometric with the
formula FeA12O4 when equilibrated with iron and alumina,
whereas an iron-rich spinel (eectively Fe3O4 dissolved in
FeA12O4) coexists with iron and wustite.19
A simplied schematic of an isothermal section of the phase
diagram for the FeAlO20 system as a function of the oxygen
fugacity is shown in Figure 10 for temperatures between 570
and 860 C. In the diagram, y is dened as the hypothetical mole
fraction of Fe2O3 in the material, if the material was fully
oxidized. This denition is equivalent to the cation fraction used
by Elrefaie and Smeltzer19 in their phase diagrams, and allows the
hematite/corundum region to be shown on the same gure as the
magnetite/hercynite region. Detailed isothermal sections showing the exact locations of the phase boundaries can be found in
Turnock and Eugster20 and Elrefaie and Smeltzer.19
Below 860 C the hercynite and magnetite solid solutions
are not completely soluble and can form dierent phases with the
phase boundaries indicated by Figure 10; the locations of the
boundaries, and the associated oxygen fugacities, are a function of
the temperature.20 Below 570 C wustite is no longer stable,20
while at temperatures higher than 860 ( 15 C, the main change
in the form of the phase diagram is that magnetite, which is an
inverse spinel, and hercynite, a normal spinel, become completely miscible and form a single phase.19,20,36 The nature of the
equilibrium among hematite, corundum, and magnetite shown in
the gure can be inferred from data given by Turnock and Eugster,20
and diers slightly from that given by Elrefaie and Smeltzer.19
The oxygen fugacity at which magnetite(ss) is in equilibrium
with alumina(ss) and hematite(ss) is shown as being equal to that
for the equilibrium between pure magnetite and pure hematite
equilibrium. Turnock and Eugster20 suggest that the dierence is
too small to measure experimentally, while the more sensitive
electrochemical measurements of Elrefaie and Smeltzer19 do indicate

Figure 10. Schematic of the FeAlO equilibrium oxygen pressure


diagram for temperatures between 570 and 860 C20. Adapted from
Figure 11 in Turnock, A.C. & Eugster, H.P., Fe-Al Oxides: Phase
Relationships below 1,000 C. Journal of Petrology, 1962, 3,3, p. 553
by permission of Oxford University Press. The positions of the phase
boundaries and compositions are a function of the temperature as given
by Table 8 in Turnock and Eugster.20 The composition of the hercynite(ss) at the Fe + FeO + hercynite(ss) equilibrium at between 750 and
850 C has been marked on the gure, i.e., y = 0.42.19,20 The dotted lines
show the location of the 50 and 75 wt % materials, and their trajectory as
the oxygen fugacity is reduced; tie lines emanating from these dotted
lines give the phase compositions.

that the oxygen fugacity may be lower for the magnetite(ss) +


alumina(ss) + hematite(ss) equilibria. Here, such a dierence
is unlikely to be large enough to be signicant.
Consider, for example, 75 wt % Fe2O3 (y = 0.66), which at the
start of a cycle has been exposed to air (PO2 = 0.21 atm); from the
phase diagram, the oxygen carrier should be a mixture of a
hematite phase (actually, a solid solution containing Al with
y 0.9), coexisting with corundum (which is actually a solid
solution containing Fe with y 0.05).19 In the diagram, there is a
direct link between PO2 (in atm) and PCO2/PCO at equilibrium, so
that decreasing the partial pressure of O2 is equivalent to
decreasing PCO2/PCO, giving a more reducing environment. The
conversion between PO2 and PCO2/PCO or PH2O/PH2 is given by
1=2

PO2 Kp1

PCO2
PH O
Kp2 2
PCO
P H2

11

where Kp1 and Kp2 are the equilibrium constants for the combustion of CO and H2, respectively, and the standard pressure is 1
atm. As shown in Figure 10, on decreasing PO2, the material
crosses into a region where the hematite(ss) is reduced to
magnetite(ss) with a separate corundum phase. The magnetite
is not pure, and does in fact contain some hercynite, so that the
average oxidation state of the iron in this magnetite is slightly
lower than for pure magnetite. Elrefaie and Smeltzer19 indicate
that this transition might occur under slightly more reducing
conditions than for pure hematite to pure magnetite. Further
611

dx.doi.org/10.1021/ef200859d |Energy Fuels 2012, 26, 603617

Energy & Fuels

ARTICLE

reduction of PO2 results in the formation of two phases denoted by


[A] and [B] in Figure 10; phase [B] corresponds to hercynite with
magnetite dissolved within it, referred to here are hercynite(ss),
while phase [A] is a solid solution of mainly magnetite with some
hercynite (i.e., magnetite(ss)). Hercynite (FeAl2O4 or FeO 3 Al2O3)
is a spinel in which the iron has eectively been reduced to the +2
oxidation state, compared to magnetite (Fe3O4 or FeO 3 Fe2O3),
which is a inverse spinel with the iron present in both the +2 and
+3 oxidation states giving an average oxidation state of +2.66.
When the value of PO2 is reduced further to point [F], nearly pure
wustite can form and can be in equilibrium with phases [A] and
[B]. At lower values of PO2, the magnetite phase reduces to nearly
pure wustite, in equilibrium with a separate phase of hercynite(ss)
(still containing some residual magnetite). The Fe3O4 within the
hercynite phase must have a lower chemical potential than the pure
Fe3O4 since it is in a mixture, indicating that Gmix < 0, and so is
stable as Fe3O4, in reducing atmospheres that would have reduced
pure Fe3O4 to FeO or even Fe. Only when the value of PO2 is
reduced by an order of magnitude lower than needed to form pure
Fe from FeO does the hercynite phase approach FeAl2O4, before
nally forming separate phases of pure Fe and Al2O3.
During the oxidation with steam to produce H2, the above
argument can be reversed. In fact, it can be seen that it is much
harder to oxidize the material in the hercynite phase to magnetite,
which will reduce the amount of hydrogen that can be produced.
For example, at 1073 K, a value of PO2 = 2.5  1013 atm
(equivalent PH2/PH2O 1  103) is needed to oxidize the Fe2+
in the hercynite phase to magnetite; in practice, recovering
hydrogen from such a dilute source is infeasible. Hercynite will
however, readily oxidize in air to produce a solid solution of
hematite in equilibrium with an alumina phase.
The XRD patterns observed in Figure 7 for the reduction of 75
wt % Fe2O3 at 850 C and dierent ratios of CO2 to CO conrm
the formation of the spinel and the wustite with increasing
reducing conditions, as would be expected from the phase
diagram. For PCO2/PCO = 30 the XRD pattern shows a single
spinel phase, the peaks of which are between those for pure
magnetite and pure hercynite. When PCO2/PCO is decreased
further to 1.4, a wustite peak forms, as expected from the phase
diagram, and the peaks in the XRD associated with the spinel
phase shift closer to those of pure hercynite, as some of the
magnetite is removed from the phase. This peak shift has been
used by others to measure the composition of the spinel phase,
and has been shown to vary linearly between the two end
members of the hercynitemagnetite solid solution; for example, Turnock and Eugster20 showed that the peak value of 2 for
the (440) plane varies from 62.6 for pure magnetite to 64.7
for pure hercynite, consistent with the variation seen in Figure 7. At
lower temperatures of 400 C, typical for iron-based catalysts, Jung
and Thompson37 used in situ XRD to show that, during reduction,
magnetite on alumina forms a spinel, which becomes progressively
richer in aluminum as the iron is reduced to metallic iron.
4.2. Accounting for the Features Observed during Temperature Programmed Reduction. Discussion of the results
from temperature programmed reduction (TPR) shown in
Figure 8 is complicated by the fact that the material is changing
temperature, and the behavior observed may well reflect both
kinetic and thermodynamic effects. Furthermore, the reducing
gas contained only H2 and N2, which should drive the reduction
fully to Fe, because PO2 = 0, i.e., the extent of reaction was not
controlled by the oxygen fugacity of the reducing gas. In the
following discussion it will be assumed that the reactions occur

sequentially, in order of decreasing equilibrium oxygen fugacity,


as the sample is reduced. The reaction would be expected to
proceed sequentially if the distance between pores in the particle
of oxygen carrier is small, and ions are able to move through the
crystals fast enough so that, locally at least, the local oxygen
fugacity is buffered by the reaction taking place. With the samples
containing alumina, matters are complicated by the fact that
different phases can form, and would require the particle to be
homogeneous both on the macro-scale, and very well mixed on
the micro scale, so that equilibrium mixtures could be reached
quickly. Of course, this is a limiting case, and given that different
phases may form during the reduction, may not be a particularly
good description. Turnock and Eugster20 provide isothermal
phase diagrams at temperatures over the range experienced by
the samples in the present work during TPR, which are represented schematically in Figure 10. The phase diagram gives some
insight into the processes taking place at each point in the
reduction. Furthermore, the mass lost during the reduction,
along with elemental balances, can be used to locate points on
the phase diagram at a given sample temperature.
In Figure 8, the 100 wt % sample showed three clear reduction
peaks, with changes in mass at each peak corresponding to the
successive transitions Fe2O3 f Fe3O4, Fe3O4 f FexO, and
FexO f Fe. The second peak in the derivative nishes roughly at
a mass which would indicate that the stoichiometry of the wustite
corresponds to Fe0.92O, although deriving an exact value is
dicult owing to the overlap between the reactions. After the rst
reduction to Fe3O4, a small amount of mass is lost below 570 C
(when wustite is unstable and magnetite reduces directly to
iron20); there is a small shoulder in the peak of dM/dt at this
temperature, indicating a switch to the reduction to wustite.
Thus, it would appear that, for the 100 wt % sample, at least, the
approximation of sequential reactions is reasonably accurate.
For all the samples containing alumina, the peak of dM/dt
occurring at the lowest temperature in Figure 8b corresponds to
the reduction of Fe2O3 to Fe3O4, with a loss of mass slightly more
than would be expected if Fe3O4 were the only product. Turnock
and Eugster20 showed that, when alumina is present, the equilibrium between magnetite and hematite is replaced by the equilibrium between a solid solution of hematite containing alumina,
in equilibrium with magnetite containing a small amount of hercynite
(a few mole % at 500600 C). However, this equilibrium
amount of hercynite is too small to account for a signicant loss
of mass in the experiments, and the rst loss in mass should
correspond approximately to that for Fe2O3 to Fe3O4. Further
formation of hercynite, and a corresponding loss in mass, would
be expected after completion of the reduction of the hematite to
magnetite, with the additional hercynite rst dissolving into the
magnetite, then forming an additional phase when the solubility
limit is exceeded. For the materials containing 80 and 90 wt %
iron oxide, the loss of mass in this rst peak corresponds to the
formation of Fe3O4 and all of the alumina present forming
hercynite; although it should be noted the small amount of alumina
present in these samples and uncertainties in the experiment make it
dicult to say for certain when hercynite formation took place. For
the samples with a lower content of iron oxide, the mass lost in this
rst peak represents the reduction to Fe3O4 plus a small fraction of
the alumina forming hercynite. Patrick et al.38 also concluded that,
during a TPR, the rst loss of mass can be attributed to the
formation of magnetite followed by the formation of hercynite.
The mass of the sample at the point where all of the alumina
forms hercynite and all of the iron oxide forms magnetite is
612

dx.doi.org/10.1021/ef200859d |Energy Fuels 2012, 26, 603617

Energy & Fuels

ARTICLE

indicated in Figure 8 by  ; the fact that this point was calculated


using the simple stepwise scheme given by eqs 610 is irrelevant,
since the mixing of the hercynite and magnetite has no eect on
the mass balance. This can be seen more clearly in Figure 12,
which shows the TPR results for the 75 wt % material, in which
the mass corresponding to the complete formation of hercynite is
indicated by Mt + Hc(ss).
From the plots of rate of loss of mass for the as-prepared particles in Figure 8b, it is observed that, as the content of alumina in
the sample is increased, the peak for the rst reduction shifts
initially to higher temperatures, then to lower temperatures as the
content of alumina is increased beyond 25 wt %. Kock et al.39
found that the onset of the reduction of hematite to magnetite
during TPR occurred at a lower temperature for the hematite
supported on alumina than for pure hematite. In contrast, Patrick
et al.38 found that adding alumina could make reduction to
magnetite during TPR more dicult. At lower calcination
temperatures they38 obtained larger surface areas, and found
that the peak corresponding to the rst reduction moved to lower
temperatures than that for pure iron oxide. Thus, the alumina
would seem to cause two eects: a reduction in reactivity per unit
area, but an increase in the surface area of the particle, with the
nal result being a balance between the two eects. In the current
work, the results largely follow this pattern. Thus, the material
containing 90 wt % iron oxide has a surface area of 14.7 m2/g
compared with 2.87 m2/g for the pure iron oxide, and the rst
peak occurs at a higher temperature than that of the pure iron
oxide, while the samples with the highest content of alumina have
the largest surface areas, with the occurrence of the rst loss of
mass being at lower temperatures than that for pure iron oxide.
Of course, it should also be noted that the surface area changes after
several cycles of reduction and oxidation, as noted in Table 2.
Accordingly, TPR experiments (not shown here) on material which
had been cycled 20 times and had developed a surface area of
1.2 m2/g, showed the onset of the rst reduction occurring at
signicantly higher temperatures than for the newly prepared material.
After the hematite has been consumed, the magnetite phase
will react with the alumina to produce more hercynite dissolved
within it until reaching the composition which can be in
equilibrium with corundum (i.e., alumina) and a solid solution
of magnetite containing hercynite (corresponding to the point
marked [A] in Figure 10). If there is sucient alumina present,
further reduction will result in the formation of a hercynite phase
with some magnetite dissolved in it (corresponding to point [B]
in Figure 10), as oxygen is removed by the global reaction (eq 7)

Figure 11. Composition of the magnetite(ss) and hercynite(ss) phases in


equilibrium with alumina and with wustite.20 Adapted from Turnock, A.C. &
Eugster, H.P., Fe-Al Oxides: Phase Relationships below 1,000 C. Journal of
Petrology, 1962, 3,3, p.552 by permission of Oxford University Press. Also
shown are the trajectories followed by the 75 wt % and 50 wt % experiments
as computed from the mass loss curves obtained from the TPR experiments.

excess alumina; the mass loss and corresponding temperature


where this would be expected to occur is indicated in Figure 8 by
Hc(ss) + Al2O3, and corresponds to a kink in Figure 8a, and the
end of a second trough in the curve of rate of mass loss with time.
Thus, for the 50 wt % material, the second peak in the curve of the
rate of loss of mass with time is largely due to hercynite formation
and nishes roughly when the distinct magnetite phase has
been completely reduced to the hercynite phase. To form more
hercynite, the residual magnetite in the hercynite phase must be
reduced. The reaction is no longer buered at the oxygen fugacity
described by the magnetite(ss) + hercynite(ss) + alumina
equilibrium and will instead follow the line in Figure 10 between
points [B] and [C] (noting that the position of this line also
changes with temperature). At point [C] all the alumina has been
consumed: this corresponds to the mass marked in Figure 8 with
, calculated from a crude mass balance and the assumption
that all of the alumina has reacted to form hercynite. A loss of
mass of 0.63% on going from [B] to [C] would be expected,
based on the composition of the phases and the mass balance.
Once there is no alumina present, the reaction is no longer
constrained by the equilibrium; further loss of oxygen must occur
by the reduction of the residual magnetite in the hercynite phase
to produce either wustite or iron. The value of y for equilibrium
of Fe + FeO + hercynite(ss) is y = 0.42 (corresponding to a mole
fraction of magnetite of 13%) between 750 and 850 C19,20 as
indicated in Figure 10; this equilibrium value of y is relatively
insensitive to temperature20 over the temperature range of
interest. For the 50 wt % sample, y = 0.39, i.e., less than 0.42,
so the reaction can move to point [D] and the residual magnetite
in the hercynite phase can then be reduced directly to iron. Thus
in Figure 8, the points marked on the curves with a + for the
50 wt % material, which correspond to when all of the Fe3O4
has reduced to FexO, are incorrect. Since the composition of
the hercynite phase at the equilibrium between Fe + FeO +
hercynite(ss) is relatively insensitive to temperature, the fact that
the 50 wt % sample may have reached this point at a temperature
diering slightly from 800 C is inconsequential. The loss of mass
for the 50 wt % sample expected as the residual magnetite
reduces from point [D] to [E] is 2.9%, which is close to the

Fe3 O4 3Al2 O3 H2 f 3FeAl2 O4 H2 O


If the reactions were proceeding sequentially, as outlined
above, this would cause the oxygen fugacity to be buered
at the equilibrium value at the point where magnetite(ss),
hercynite(ss), and alumina coexist; this point is a function of
temperature. Turnock and Eugster20 give the compositions at
points [A] and [B] as a function of temperature and the oxygen
fugacity, and the plot is shown in Figure 11. The trajectory
followed during the temperature programmed reduction of the
50 wt % sample in the present work is also shown in Figure 11.
The trajectory was calculated from the loss of mass observed in
the TGA reduction experiments as the temperature increased. It
can be seen that the trajectory crosses from the two-phase region,
where a magnetite phase exists, into a region where there is only a
single hercynite phase at sample temperature of 737 C. At this
point the sample should contain only the hercynite phase and
613

dx.doi.org/10.1021/ef200859d |Energy Fuels 2012, 26, 603617

Energy & Fuels

ARTICLE

Figure 12. Mass and the rst and second derivative of mass with respect to time for the 75 wt % sample, during temperature programmed reduction in
5% H2. Points have been marked on the mass curve to indicate the phases which should be present based on the mass of oxygen lost, and assuming a
sequential reaction. Mt(ss) = Magnetite solid solution, Hc(ss) = Hercynite solid solution, Hc(ss)* = hercynite solid solution with a composition in
equilibrium with iron, and wustite, Wus = wustite, Hc = stiochiometric hercynite, Fe = iron.

mass lost during the third large peak in the plot of dM/dt,
i.e., between the points marked by  and 0. At point [E] only
pure hercynite and Fe are present in the sample (denoted by 0
in Figure 8). Finally, the slow reduction of mass, which begins
shortly after the start of the isothermal phase of the experiment
begins at a mass which is reasonably close to the point where only
hercynite remains to be reduced, viz., the long slow mass loss at
the end of the experiments appears to be largely due to the
reduction of hercynite.
For the samples containing less than 50 wt % alumina, the
behavior diers. For example, for the sample containing 75 wt %
Fe2O3 (y = 0.66), once all of the hematite has been consumed,
distinct phases of magnetite and hercynite can form, but there is
no excess alumina present. The sample remains within the twophase region shown in Figure 11. The trajectory for the 75 wt %
sample in Figure 11 has been computed as far as the point at
which iron begins to form. The kink in the curve at 40 wt %
corresponds to the point where the alumina has been fully
converted to hercynite. Once the alumina has been fully consumed, further loss of mass is presumably from the reduction of
the Fe3O4 in the magnetite(ss) phase to wustite, so that the
reaction jumps to point [F] in Figure 10. The production of
wustite will proceed until the magnetite phase is fully consumed,
corresponding to the point in Figure 11 where the trajectory
crosses from the two-phase into the single-phase region. Some of
the residual magnetite should also then be reduced to wustite,
until the hercynite phase becomes suciently rich in hercynite
(i.e., when the composition of this phases reaches 13 mol %
magnetite in hercynite, equivalent to y = 0.42), for the residual
magnetite to start being reduced directly to iron. However, at this
point the reduction of magnetite to Fe should be buered by the
equilibrium between wustite and iron, so that it is only after all
the wustite has been reduced to iron that the residual magnetite
can be reduced to iron. The masses at which these transitions
should take place are marked in Figure 12 for the 75 wt % sample,
based on the mass balance and the composition of the solid solution
of hercynite phase given by Turnock and Eugster,20 at each point.
In Figure 8, the samples containing 7590 wt % alumina do
not show a distinct peak in the plot of dM/dt corresponding to

the formation of wustite: instead, there is a second broad peak


between the masses where magnetite has fully formed () which
nishes at a mass lower than that expected if all the magnetite
were to form wustite (+). The addition of alumina seems to have
suppressed the formation of wustite. This is seen more clearly in
Figure 12, where, after the rst peak in dM/dt which corresponds
to the formation of magnetite (ending at the point marked Mt),
there is a slow loss of mass until the point where all the alumina
has been fully converted to hercynite (indicated as Mt(ss) +
Hc(ss)). If the formation of hercynite can buer the oxygen
fugacity, this will prevent wustite forming until the alumina is
consumed, so that the wustite formation is delayed to a higher
temperature. Between 550 and 850 C there is a second, broad
peak in the plot of dM/dt which makes it dicult to separate the
possible stages of reduction.
Between the masses where wustite should rst begin forming
(i.e., Mt(ss) + Hc(ss) in Figure 12) and ending at a mass where
the wustite should rst begin to reduce to iron (i.e., Wus +
Hc(ss)*), there is a slight shoulder in this broad peak of dM/dt.
This is more clearly seen in the second derivative of the mass loss,
which shows a dip between 630 and 706 C, indicating rate of
mass loss has been accelerated by an additional reaction. Some of
the mass loss in this interval may come from the reduction of the
magnetite held within the hercynite phase (i.e., between Wus +
Hc(ss) and Wus + Hc(ss)* from 706 to 734 C), however, this
mass change is small, and would be dicult to separate from the
reduction of the magnetite phase to wustite. The second broad
peak in dM/dt then reaches a minimum at 790 C, at a mass
where iron should be forming from the wustite phase. The peak
associated with the reduction of the wustite to iron appears to be
quite broad, and overlaps somewhat with the next phase of
reduction, the reduction of the residual magnetite in the hercynite phase to iron (i.e., between Fe + Hc(ss)* and Fe + Hc). A
broad peak in dM/dt was also observed for the reduction of pure
iron oxide when wustite was being reduced to iron, so this is
not entirely unexpected. Despite the fact that the solubility of
magnetite in the hercynite phase which is in equilibrium
with wustite and iron (denoted by Hc(ss)* in Figure 12) is only
15 wt %, the reduction of magnetite held within the hercynite
614

dx.doi.org/10.1021/ef200859d |Energy Fuels 2012, 26, 603617

Energy & Fuels

ARTICLE

at this point would cause a loss of mass of 2.1%; this reaction


might account for the nal large shoulder in the plot of dM/dt in
Figure 11 (i.e., between 930 and 984 C) . Finally, in
Figure 11, there is a slow loss of mass above 984 C, over
masses which are consistent with the reduction of pure
hercynite to iron and alumina.
In summary, although the assumption that the reactions
proceed separately and sequentially is unlikely to be entirely
accurate, it would appear from the results that the reduction does,
indeed, proceed in the stages identied using the phase diagrams.
The addition of alumina is generally found to make the
reduction to wustite or metallic iron more dicult.38,39 In the
TPR results presented here, the addition of alumina does seem to
have made reduction more dicult, with wustite and iron
forming at higher temperatures than in the case of pure iron
oxide. The reason for the retardation in rate is likely to be a
combination of kinetics eects (viz., changes in activation energy
and rate constants), and the eects outlined above. From the
aforegoing discussion, hercynite in particular appears to be very
dicult to reduce. In fact, the oxygen fugacity at equilibrium
required to reduce pure hercynite to iron at 850 C is 1020,20
which corresponds PCO/PCO2 25, so only a relatively small
amount of CO2 produced in the gas phase during reduction with
CO would be required to cause the reaction to be equilibriumlimited. The results from the TGA suggest that hercynite reduces
slowly even when heated to 1050 C, at which point the
equilibrium oxygen fugacity is reduced to 1016. Thus, during
the cycling experiments, it seems unlikely that any of the hercynite
would have been reduced over the time scale of a cycle. The slow
reduction of hercynite suggests that the solid would require a long
residence time within a uidized or moving bed reactor (or else a
large inventory of solid within a packed bed reactor), to fully reduce
the available iron oxide to Fe, i.e., in practice the Fe present as
hercynite may not participate in hydrogen production.
It is interesting to note that, in Figure 9, with PCO2/PCO = 1.2
in the reducing gas rather than H2, the onset of the loss of mass
occurred at higher temperatures for the 75 wt % Fe2O3 particles
compared with the pure iron oxide. As with the reduction in H2,
the rst loss of mass corresponds to the formation of magnetite.
The ratio PCO2/PCO = 1.2 should be sucient to reduce the pure
magnetite to wustite at temperatures higher than 850 C. However, when alumina is present, the phase diagrams given by
Turnock and Eugster20 indicate that for PCO2/PCO = 1.2, the
oxygen fugacity is initially low enough to drive the equilibrium to
a solid solution of magnetite and hercynite, with further reduction to produce a separate wustite phase possible at a slightly
higher temperature than that required for pure iron oxide. In
contrast to the materials containing 100 wt % Fe2O3, where full
conversion to wustite was seen, for material containing 75 wt %
Fe2O3, a loss of mass corresponding only to the formation of a
mixture of Fe3O4 + FeAl2O4, and a small amount of FexO, is
observed. Additionally, even when the temperature was maintained at 950 C for 30 min, the 75 wt % Fe2O3 material was not
completely reduced to FexO. This is consistent with the phase
diagram, which suggests that the magnetite is more dicult to
reduce when alumina is present and that at 950 C and PCO2/PCO =
1.2, it should only be possible to reduce a fraction of the magnetite
to FexO.
Thus, during chemical looping with composite particles of
Fe2O3 and alumina, the alumina is not inert and can have a strong
inuence on the oxidation and reduction process by forming
hercynite, solid solutions, and altering the chemical potential of

the iron-oxides contained in these phases. The presence of


alumina alters the kinetics of the reactions, and also the nal
(i.e., equilibrium) extent of reaction. When alumina is present the
nal conversion of the iron oxide can depend on the oxygen
fugacity in the gas phase, in contrast with pure iron oxide, which
will react to completion.
4.3. Role of Oxidation with Air on Structure and Cyclic
Performance. Figure 3 shows that the composite particles with
75 wt % Fe2O3 attain a level of constant performance after 20
cycles, and that the air oxidation step is important to maintain
this stable performance. Oxidation in air should restore the
particles to their initial state of a mixture of hematite-rich and
corundum-rich solid solutions. This can be seen from the XRD
patterns shown in Figure 7 for 75 wt % Fe2O3 after 20 cycles,
which show peaks corresponding to hematite and spinel
(magnetite and hercynite), whereas, without air oxidation, peaks
corresponding to spinel (magnetite and hercynite) and wustite
are seen.
For the experiment without air oxidation, the yield of hydrogen is low after 20 cycles, and the performance drops in a similar
fashion to the 100 wt % Fe2O3. In contrast, when the sample is
reoxidized in air at the end of a cycle, the performance is
improved. There is little signicant dierence in the BET surface
areas of the cycled materials (Table 2), which indicates that the
performance of the particles does not rely on a large surface area.
Measurements of surface area are heavily biased toward smaller
pores, and the BJH model is unable to account for the volume in
pores with a diameter larger than 300 nm. The optical images
are persuasive, since they show clearly the development of large
macropores in the material when cycling includes oxidation in air.
There also appears to be some bulk segregation of material
during cycling, which can be seen more clearly in the EDXS maps
given in Figure 6. Images of the surface of the particles in Figure 4
show that the cycled material has developed features on the scale
of a few micrometers. In contrast, when the air oxidation was not
included, the particles appeared to be uniform and did not
contain large voids. Thus, although there has been signicant
sintering of the micro- and mesopores, there is still sucient
porosity for gas to be able to penetrate into the interior of the
particle. Also, the grains in the material are still suciently small
to prevent a signicant product layer resistance. It is worth noting
from other work15 that mechanically mixed particles with little
microporosity and grains as large as a few micrometers perform
well for hydrogen production provided they are not reduced fully
to Fe. From these results it seems likely that the air oxidation
step, which should produce separate phases of solid solutions of
corundum and hematite, is responsible for the formation of large
pores in the particles. However, it should also be noted from the
XRD patterns that the oxidation of the spinel to hematite and
alumina may not have been complete.
Paananen et al.40 studied the reduction of magnetite to wustite
while reducing briquetted mixtures of commercial magnetite and
Al2O3 under an atmosphere containing CO2 and CO in the ratio
1:0.12 at 950 C in a TGA. They reported that the Al cations
trapped in the magnetite induced swelling and cracking of the
briquettes during reduction, as the magnetite phase became
richer in Al and approached the composition of hercynite.40
Assuming the formula of wustite to be Fe0.88O, Paananen
et al.40 reported that the structure would expand by about 6 vol %
during the reduction from magnetite to wustite, compared to 16
vol % if it were hercynite being reduced to wustite. They
suggested that, at the boundaries between the magnetite and
615

dx.doi.org/10.1021/ef200859d |Energy Fuels 2012, 26, 603617

Energy & Fuels


growing wustite phase, the alumina content in the magnetite is
enriched (since the alumina is not soluble in the wustite),
resulting in a larger volume change near the boundary, and a
substantial stress at the interface with the wustite. These stresses
would, in turn, cause the fracturing of the structure during the
reduction. In the present work, fracturing during the reduction
from hercynite to wustite alone cannot account for the stabilization in performance, since it is observed that oxidation in air is
also required. However, since the materials are cycled repeatedly
between the fully oxidized and reduced state, it is likely that the
molar volume changes associated with the various phases that can
form are more severe when alumina is present, than when the
iron oxide is pure. Bulk segregation of iron and alumina was also
observed near the surfaces of grains in the particles cycled with air
oxidation. During reduction, and at the temperature used here,
i.e., 850 C, the magnetite and hercynite which form are mutually
soluble and exhibit complete miscibility at 860 ( 15 C,20 which
would tend to make the particle more homogeneous. On the
other hand, the formation of separate phases of hematite and
alumina on oxidation seems to result in macroscale segregation, possibly driven by dierences in surface energy between
the dierent phases. The nal morphology (and hence the
reactivity) of the particles will be a balance of these two eects. Thus,
in these experiments, oxidation in air seemed to be benecial,
resulting in increased macroporosity, with complete segregation of
the alumina and iron phases avoided because the particles were
repeatedly passed through a state where any mutual solubility of the
phases which form is high or even complete.

ARTICLE

AUTHOR INFORMATION
Corresponding Author

*E-mail: sas37@cam.ac.uk; phone: + 44 (0) 1223 332645; fax:


+ 44 (0)1223 332662.

ACKNOWLEDGMENT
This work was supported by the Engineering and Physical
Sciences Research Council (UK) [grant EP/G063265/1]. Financial support from the Cambridge Commonwealth Trust is
gratefully acknowledged. We also wish to acknowledge assistance
from Dr. M. Vickers with Rietveld analysis, Mr. S. Griggs with
SEM images, Mr. Z. Saracevic with BET measurements, and
access to the Chemical Database at Daresbury.
REFERENCES
(1) Intergovernmental Panel on Climate Change. Climate Change
2007 - The Physical Science Basis; Working Group I Contribution to
the Fourth Assessment Report of the IPCC (ISBN:9780521705967).
Cambridge University Press, October 2007 (Available online at http://
www.ipcc.ch/publications_and_data/publications_and_data_reports.
shtml).
(2) Lyngfelt, A.; Leckner, B.; Mattisson, T. Chem. Eng. Sci. 2001,
56, 31013113.
(3) Abad, A.; Mattisson, T.; Lyngfelt, A.; Johansson, M. Fuel 2007,
86, 10211035.
(4) Hossain, M.; Delasa, H. Chem. Eng. Sci. 2008, 63, 44334451.
(5) Li, F.; Kim, H. R.; Sridhar, D.; Wang, F.; Zeng, L.; Chen, J.; Fan,
L. Energy Fuels 2009, 23, 41824189.
(6) Galvita, V.; Sundmacher, K. Appl. Catal., A 2005, 289, 121127.
(7) Bohn, C. D.; Muller, C. R.; Cleeton, J. P.; Hayhurst, A. N.;
Davidson, J. F.; Scott, S. A.; Dennis, J. S. Ind. Eng. Chem. Res. 2008,
47, 76237630.
(8) Yang, J.; Cai, N.; Li, Z. Energy Fuels 2008, 22, 25702579.
(9) Messerschmitt, A. Process for producing hydrogen, U.S. Patent
971,206, 1910.
(10) Jerndal, E.; Mattisson, T.; Lyngfelt, A. Chem. Eng. Res. Des.
2006, 84, 795806.
(11) Fan, L.; Li, F.; Ramkumar, S. Particuology 2008, 6, 131142.
(12) Mattisson, T.; Johansson, M.; Lyngfelt, A. Energy Fuels 2004,
18, 628637.
(13) Adanez, J.; de Diego, L. F.; Garca-Labiano, F.; Gayan, P.; Abad,
A.; Palacios, J. M. Energy Fuels 2004, 18, 371377.
(14) Bohn, C. D.; Cleeton, J. P.; Muller, C. R.; Chuang, S. Y.; Scott,
S. A.; Dennis, J. S. Energy Fuels 2010, 24, 40254033.
(15) Kierzkowska, A. M.; Bohn, C. D.; Scott, S. A.; Cleeton, J. P.;
Dennis, J. S.; Muller, C. R. Ind. Eng. Chem. Res. 2010, 49, 53835391.
(16) Ishida, M.; Takeshita, K.; Suzuki, K.; Ohba, T. Energy Fuels
2005, 19, 25142518.
(17) Monteiro Nunes, S.; Paterson, N.; Dugwell, D. R.; Kandiyoti, R.
Energy Fuels 2007, 21, 30283035.
(18) Cox, D. E.; Takei, W. J.; Miller, C.; Shirane, G. J. Phys. chem.
Solids 1962, 23, 863874.
(19) Elrefaie, F.; Smeltzer, W. Metall. Mater. Trans. B 1983, 14, 8593.
(20) Turnock, A. C.; Eugster, H. P. J . Petrol. 1962, 3, 533.
(21) Cox, D. E.; Moodenbaugh, A. R.; Sleight, A.; Chen, H. Y. N.B.S.
(U.S.), Spec. Publ. 1980, 567, 189201.
(22) El-Sharkawy, E. A.; El-Hakam, S. A.; Samra, S. E. Mater. Lett.
2000, 42, 331338.
(23) Larsson, L.; ONeill, H. S. C.; Annersten, H. Eur. J. Mineral.
1994, 6, 3951.
(24) Fleet, M. E. Acta Crystallogr., B: Struct. Crystallogr. Cryst. Chem.
1981, 37, 917920.

(25) Fjellvag,
H.; Grnvold, F.; Stlen, S.; Hauback, B. J. Solid State
Chem. 1996, 124, 5257.

5. CONCLUSIONS
Solid solutions of hematite (-Fe2O3) and corundum (-Al2O3),
synthesized by the coprecipitation method, have been studied
for application as oxygen carriers in chemical looping for the
production of hydrogen. The particles were subjected to upto
50 cycles of reduction and oxidation in a xed bed reactor at
850 C and 1 atm pressure. It was found that 20 cycles was the
minimum number necessary for the particles to approach a
constant level of performance from cycle to cycle. The optimum
loading of Al2O3 in the composite particles was found to be
25 wt % for the production of hydrogen over 50 cycles with
48 % yield.
The supporting Al2O3 was not inert. Indeed, it has been found
to participate actively in the oxidation and reduction processes,
principally through the formation of hercynite and the formation
of various solid solutions. The thermodynamics of the FeAlO
system is more complicated than that of the pure FeO and the
behavior of the alumina-containing carriers was found to be
consistent with published phase diagrams for this system. The
presence of alumina alters the kinetics of the reactions, and also
the nal (i.e., equilibrium) extent of reaction. When alumina is
present the nal conversion of the iron oxide can depend on the
oxygen fugacity in the gas phase, in contrast with pure iron oxide,
which will react to completion. The presence of alumina retards
the rate of reduction, and in practice may limit the extent of
conversion of the iron oxide to Fe.
To maintain the reactivity of the alumina-containing carriers, a
nal oxidation in air was required. This caused separate phases of
alumina and hematite to form, and seemed to result in the
formation of macropores. The oxygen carriers did not need a
large surface area to be reactive, and the macroporosity and
resulting grain sizes were sucient to allow reaction to continue.
616

dx.doi.org/10.1021/ef200859d |Energy Fuels 2012, 26, 603617

Energy & Fuels

ARTICLE

(26) Giddings, R. A.; Gordon, R. S. J. Am. Ceram. Soc. 1973, 56, 111116.
(27) Cleeton, J. P. Chemical Looping Combustion with Simultaneous
Power Generation and Hydrogen Production using Iron Oxides; PhD
Thesis; University of Cambridge, 2011.
(28) Barin, I.; Knacke, O. Thermochemical Properties of Inorganic
Substances; Springer-Verlag: Berlin, 1973.
(29) McBride, B.; Zehe, M.; Gordon, S. NASA Glenn Coecients
for Calculating Thermodynamic Properties of Individual Species; NASA
TP-2002-211556; NASA: Washington, DC, 2002 (Data available online
at http://www.grc.nasa.gov/WWW/CEAWeb/ceaThermoBuild.htm).
(30) Muan, A. J. Am. Ceram. Soc. 1956, 39, 207214.
(31) Muan, A. Am. J. Sci. 1958, 256, 413422.
(32) Tartaj, P.; Tartaj, J. Chem. Mater. 2002, 14, 536541.
(33) Zhong, Z. Y.; Prozorov, T.; Felner, I.; Gedanken, A. J. Phys.
Chem. B 1999, 103, 947956.
(34) Liu, M.; Li, H.; Xiao, L.; Yu, W.; Lu, Y.; Zhao, Z. J. Magn. Magn.
Mater. 2005, 294, 294297.
(35) Shannon, R. D. Acta Crystallogr. 1976, A32, 751767.
(36) Petric, A.; Jacob, K. T.; Alcock, C. B. J. Am. Ceram. Soc. 1981,
64, 632639.
(37) Jung, H.; Thomson, W. J. J. Catal. 1991, 128, 218230.
(38) Patrick, V.; Gavalas, G. R.; Sharma, P. K. Ind. Eng. Chem. Res.
1993, 32, 519532.
(39) Kock, A. J. H. M.; Fortuin, H. M.; Geus, J. W. J. Catal. 1985,
96, 261275.
(40) Paananen, T.; Heinanen, K.; Harkki, J. ISIJ Int. 2003, 43, 597605.

617

dx.doi.org/10.1021/ef200859d |Energy Fuels 2012, 26, 603617

You might also like