Sei sulla pagina 1di 13

The Eect of Addition of ZrO

2
to Fe
2
O
3
for Hydrogen Production by
Chemical Looping
Wen Liu,*
,
John S. Dennis,

and Stuart A. Scott

Department of Chemical Engineering and Biotechnology, University of Cambridge, Pembroke Street, Cambridge CB2 3RA, United
Kingdom

Department of Engineering, University of Cambridge, Trumpington Street, Cambridge CB2 1PZ, United Kingdom
*S Supporting Information
ABSTRACT: In this paper, a synthetic mixture of ZrO
2
and Fe
2
O
3
was prepared by coprecipitation for use in chemical looping
and hydrogen production. Cycling experiments in a uidized bed showed that a material composed of 30 mol % ZrO
2
and 70
mol % Fe
2
O
3
was capable of producing hydrogen with a consistent yield of 90 mol % of the stoichiometric amount over 20 cycles
of reduction and oxidation at 1123 K. Here, the iron oxide was subjected to cycles consisting of nearly 100% reduction to Fe
followed by reoxidation (with steam or CO
2
and then air) to Fe
2
O
3
. There was no contamination by CO of the hydrogen
produced, at a lower detection limit of 500 ppm, when the conversion of Fe
3
O
4
to Fe was kept below 90 mol %. A preliminary
investigation of the reaction kinetics conrmed that the ZrO
2
support does not inhibit rates of reaction compared with those
observed with iron oxide alone.
1. INTRODUCTION
Hydrogen is considered to be a clean energy substitute for fossil
fuels, because the combustion of hydrogen does not yield CO
2
;
however, in many cases, the hydrogen is made in the rst place
from a carbonaceous fuel. The steam-iron process was the rst
large-scale process for producing hydrogen, based on the
oxidation of iron by steam:
1
+ +
= H
3Fe 4H O Fe O 4H
105.3 kJ/mol
2 3 4 2
1123 K
o
(1)
However, this process was quickly replaced by the steam
reforming of methane:
+ + = H CH H O CO 3H 226.8 kJ/mol
4 2 2 1123 K
o
(2)
+ + = H CO H O CO H 33.5 kJ/mol
2 2 2 1123 K
o
(3)
which provides the majority of the hydrogen used commercially
because of its low cost, wide availability of natural gas, and
proven technology. There are limitations to the steam-methane
reforming process. Firstly, owing to equilibrium considerations
for the water-gas-shift reaction 3, the purity of the hydrogen
produced will not necessarily satisfy the requirement for
hydrogen fuel cells: as much as 50 ppmv CO will poison the Pt
anode in a proton exchange membrane fuel cell (PEMFC).
2
Secondly, at small scales of production, where the capital cost
of extensive heat integration is not justied, the hydrogen
produced is expensive.
In contrast, cyclic operation of the steam-iron process oers
the prospect of producing very pure hydrogen viably from the
small to large scale: it also allows simultaneous separation of a
pure stream of CO
2
, suitable for sequestration.
3
The process
involves subjecting Fe
2
O
3
, an oxygen carrier, to the following
reactions:
+ +
= H
3Fe O H 2Fe O H O
5.9 kJ/mol
2 3 2 3 4 2
1123 K
o
(4)
+ +
= + H
1.202Fe O H 3.807Fe O H O
56.6 kJ/mol
3 4 2 0.947 2
1123 K
o
(5)
+ +
= H
3Fe O CO 2Fe O CO
40.8 kJ/mol
2 3 3 4 2
1123 K
o
(6)
+ +
= + H
1.202Fe O CO 3.807Fe O CO
34.9 kJ/mol
3 4 0.947 2
1123 K
o
(7)
+
= H
2Fe O 1/2O 3Fe O
237.2 kJ/mol
3 4 2 2 3
1123 K
o
(8)
Reaction 5 is responsible for hydrogen production, whereas
reaction 8 supplies heat, which can be utilized within the system
or exported. The authors
3
limited the reduction of iron oxide to
wustite as they found substantial decay in hydrogen yield after
10 cycles, if the material (i.e., 99 wt % pure iron oxide) was
reduced further to Fe. Reed and Berg
4
proposed undertaking
reactions 4 to 8 using three, interconnected uidized beds, with
reduction from Fe
2
O
3
to Fe by a fuel gas taking place in the rst
reactor and conducting reaction 1 in the second reactor and
Received: September 26, 2012
Revised: November 30, 2012
Accepted: November 30, 2012
Published: November 30, 2012
Article
pubs.acs.org/IECR
2012 American Chemical Society 16597 dx.doi.org/10.1021/ie302626x | Ind. Eng. Chem. Res. 2012, 51, 1659716609
reaction 8 in the third. Cleeton et al.
5
showed that the inclusion
of reaction 8 was benecial for heat integration of the whole
process.
The deactivation of particles of pure iron oxide after repeated
redox cycles in the above reaction scheme has been reported by
several authors. Bohn et al.
6
tested pure iron oxides made by
granulating relatively pure (i.e., 99%+ purity), powdered
hematite with water and sintering the resulting particles at
900 C to give mechanical strength. They found that particles
of pure iron oxide could undergo the reduction from Fe
2
O
3
to
Fe
3
O
4
without signicant deactivation over 10 cycles; however,
the reduction from Fe
3
O
4
to Fe
0.947
O showed a progressive
decay in rate of reaction over successive cycles. Bohn et al.
6
found that the rate constant for the intrinsic rate of reaction, k
i
= A exp(E
A
/RT), decreased because the pre-exponential
factor, A, fell: the activation energy, on the other hand,
remained constant over many cycles. This suggested that
deactivation was caused by a loss of surface area or active sites
on the particles of iron oxide. Bleeker et al.
7
also found a rapid
diminution of the BrunauerEmmettTeller (BET) surface
area with an iron oxide catalyst for ammonia synthesis, when it
was reduced and oxidized between Fe
3
O
4
and Fe over repeated
cycles. The unused catalyst had a BET area of 31 m
2
/g, which
fell to 0.37 m
2
/g (with the iron oxide as Fe
3
O
4
) after 14 cycles.
Numerous research groups have attempted to stabilize the
production of hydrogen in the steam-iron cycle by the addition
of oxides of other metals. Otsuka et al.
8
studied the impact of
26 individual metal additives on the performance of iron oxide
in steam-iron cycles. They claimed that Al, Mo, and Ce were
eective agents against sintering for the oxidation of Fe to
Fe
3
O
4
below 873 K. Galvita and Sundmacher
9
developed a
mixed oxide containing Cr
2
O
3
, Fe
3
O
4
, CeO
2
, and ZrO
2
, which
showed stable performance over 100 cycles at 1013 K.
Subsequently, Galvita et al.
10
showed that mixed oxides of Fe,
Ce, and Zr were also suitable for long-term operation at 1098
K. However, in both studies (Galvita and Sundmacher,
9
Galvita
et al.
10
), the conversion from Fe
3
O
4
to Fe was kept below 60
mol % to avoid the deposition of carbon by the Boudouard
reaction. Bohn et al.,
11
Li et al.,
12
and Kierzkowska et al.
13
have
all found that the addition of Al
2
O
3
does indeed reduce the
extent of deactivation when cycling between Fe
2
O
3
and Fe at
1123 K. However Kierzkowska et al.
13
found that an inert
species, FeAl
2
O
4
, was formed and did not react noticeably with
steam at 1123 K. Hence, the stoichiometric yield of hydrogen
was reduced with the addition of Al
2
O
3
. A further investigation
into the Al
2
O
3
-modied Fe
2
O
3
materials showed the formation
of solid solutions of magnetite (Fe
3
O
4
) and hercynite
(FeAl
2
O
4
).
14
The formation of the solid solution substantially
reduced the thermodynamic driving forces for the redox
reactions, and thus, the kinetics of the reduction from the
nominal magnetite phase to wusite phase became much
slower.
15
As a result, the oxygen carrier was less reactive in
the presence of alumina, as much stronger reducing conditions
and oxidizing conditions were required to decompose the solid
solutions to free the active iron oxides for hydrogen production.
The objective of the present work was to examine the
resistance to deactivation and carbon deposition of iron oxide
when subjected to cycles consisting of nearly 100% reduction to
Fe followed by reoxidation (with steam and then air) to Fe
2
O
3
,
at 1123 K, using a Fe
2
O
3
/ZrO
2
oxygen carrier. Zirconia has
been chosen to avoid the chemical interaction between the
oxides of iron and the support material seen with an alumina
support. From the available literature, mutual cationic solubility
in the FeOZr system is very low, with an equilibrium
solubility of Fe
2
O
3
in ZrO
2
of 2 mol % and ZrO
2
in Fe
2
O
3
of 1
mol % at 1373 K.
16
However, a metastable phase of ZrO
2
containing up to 6.8 mol % Fe
3+
may exist at 1173 K.
17
When
the iron oxide is reduced to wustite, the equilibrium solubility
of wustite in ZrO
2
is about 2 wt % at 1473 K, and it shows a
decreasing trend with decreasing temperature, whereas the
wustite does not seem to dissolve ZrO
2
at all.
18,19
It is
anticipated that this level of interaction is insignicant in
hindering the reaction kinetics or the ultimate conversion of
iron oxides during chemical looping. The performance of the
carrier was investigated under various operating conditions to
verify its inertness, and comparisons are made with alumina-
supported Fe
2
O
3
carriers.
2. EXPERIMENTAL SECTION
2.1. Material Preparation. The ZrO
2
-supported Fe
2
O
3
oxygen carriers were prepared by coprecipitation. In a typical
synthesis, nitrates of the metals (viz., Fe(NO
3
)
3
9H
2
O and
Zr(NO
3
)
4
5H
2
O) were mixed in the desired molar composition
(viz., 10 mol % metal additive with 90 mol % Fe or 30 mol %
metal additives with 70 mol % Fe) in deionized water to give a
solution with metal ion concentration of approximately 0.75 M.
The solution of nitrates was then added into a batch of 1.0 M
aqueous Na
2
CO
3
solution. The molar ratio of total metal ions
to Na
2
CO
3
was controlled to be around 1:4, which gave a nal
pH value of 10.0 0.2 in the resulting precipitate. The
precipitate was stirred for a further 10 min and aged at room
temperature unstirred for 2 h. After aging, the precipitate was
washed with deionized water until its ionic conductivity fell
below 120 S cm
1
, ltered, and dried at 353 K for 18 h in air.
The dried cake was calcined at 1223 K for 3 h, followed by
crushing and sieving to obtain particles in the size range of
500600 m for experimentation in a uidized bed reactor. In
the case of packed bed experiments, the sieve size fraction of
300425 m was used instead. The slightly larger carrier
particles were used in the uidized bed to facilitate separation
from the inert bed material (white aluminum oxide, 99.76%,
Boud Materials Ltd.) of sieve size 300425 m, as described
below in Section 2.2.
In addition, iron oxide powders (Sigma-Aldrich, 99% trace
metals basis, d
p
< 5 m) were granulated using water as a
binder, and the resulting granules were sieved to the desired
size fraction of 300425 m, followed by calcination at 1223 K
for 3 h in air. The sintered particles were cooled to room
temperature and sieved again to ensure particles of the desired
size fraction of 300425 m were obtained. The unsupported
iron oxides particles were denoted as Fe100 and investigated in
selected experiments as a basis of comparison with the
supported oxygen carriers.
2.2. Experimental Apparatus. Three reactors were used
to study the synthesized materials: a packed bed reactor, a
uidized bed reactor, and a thermogravimetric analyzer (TGA).
The walls of both tubular reactors were made of recrystallized
Al
2
O
3
, with internal diameters of 9 and 20 mm, respectively. In
the packed bed arrangement, the gas was fed from the top. The
packing arrangement, from top to bottom, consisted of the
following layers of materials: (i) 10.0 g of -Al
2
O
3
(14001700
m), (ii) 0.1 g of samples (300425 m), (iii) 2.0 g of -Al
2
O
3
(300425 m), (iv) 2.0 g of -Al
2
O
3
(14001700 m), and
(v) a ceramic distributor plate with three holes of 1.5 mm
diameter in a triangular array. The uidized reactor consisted of
a perforated ceramic distributor with 5 holes of 1 mm diameter
Industrial & Engineering Chemistry Research Article
dx.doi.org/10.1021/ie302626x | Ind. Eng. Chem. Res. 2012, 51, 1659716609 16598
in a square array. A packing of 15 g of -Al
2
O
3
(14001700
m) was placed below the distributor of the uidized bed
reactor to enhance the rate of preheating of the inlet gas. The
bed material in a typical uidized bed experiment consisted of
20 g of -Al
2
O
3
(sieve fraction of 300425 m) and 0.5 g of
sample (500600 m).
Both the packed bed reactor and the uidized bed reactor
used the same control systems, as described below. Cylinders of
compressed gas were used to supply (i) 10 vol % CO, N
2
balance (BOC), (ii) pure N
2
(Air Liquide), (iii) air (Air
Liquide), (iv) CO
2
(Air Liquide), (v) 15% CO
2
in N
2
(BOC),
and (vi) 5% H
2
, N
2
balance (Spectrashield, BOC). Diluted
steam was generated by saturating N
2
with water vapor in a
bubbler. The degree of saturation of the N
2
by water was
veried in a preliminary experiment, in which a large batch of
fully-reduced, ZrO
2
-supported Fe
2
O
3
was oxidized by a small
ow of diluted steam passing through a packed bed, such that
the residence time of the gases was suciently long for
equilibrium to have been achieved between steam and
hydrogen at the exit from the bed. The measured partial
pressure of the product hydrogen was found to be very close to
the theoretical value of p
H
2
O,s
/(1 + K
p
), where p
H
2
O,s
= 0.026
bara is the saturated vapor pressure of water at the operating
temperature of the bubbler, 295 K. The ow rates of the gases
were adjusted and measured using calibrated rotameters and
mass ow sensors (Honeywell AWM5103VN). Solenoid valves
were used to switch the gases.
In uidized bed experiments, all gas ow rates were
nominally 2.5 L/min (1 atm, room temperature), which
corresponds to U/U
mf
7 at 1123 K, where U is the
supercial velocity of the uidizing gas, and U
mf
is the minimum
uidization velocity, calculated using the correlation of Wen
and Yu.
20
The ceramic reactors were heated by tubular
furnaces, and the temperatures were measured by type N
thermocouples. To measure the rate and extent of reaction, a
sample of the euent gas was withdrawn continuously through
a quartz tube, i.d. of 5.0 mm, by a diaphragm pump at a rate of
0.5 L/min at (1 atm, room temperature). The sampled gases
were ltered of particulates and dried by two condensing tubes
and a tube packed with CaCl
2
. The composition of the cleaned
gas was continuously determined by three analysers connected
in parallel: (i) ABB EL3020 Caldos 27 for H
2
, 020 vol %, (ii)
ABB EL3020 Uras26 with Magnos206 for O
2
, 0100 vol %,
CO, 020%, and CO
2
, 030 vol %, and (iii) ABB Easyline CO,
01 vol %. The third analyzer was used to measure the quality
of H
2
during the steam oxidation step.
The purpose of the experiments in the packed bed was to
allow exible changes to gas ow rates without having to be
concerned with either deuidisation, at low rates of ow, or
elutriation, at high rates of ow. The redox potential, i.e., p
CO
2
/
p
CO
or p
H
2
O
/p
H
2
, of the gas owing through the packed bed was
varied by adjusting the gas ow rates of N
2
, 5 vol % H
2
,
moisturized N
2
saturated with water vapor (295 K), 10 vol %
CO, and CO
2
.The details of the gas composition used in the
cycling experiments are described in the Results section, below.
The TGA experiments were important where kinetics were
so slow that the dierence between inlet and outlet gas
concentration of the uidized bed reactor was too small to be
distinguished by the gas analysers. This applied particularly to
the reduction of magnetite to wustite and from wustite to iron.
In a typical TGA experiment, a permanent ow of N
2
of 20
mL/min (1 atm, room temperature) was used as the purge gas.
In addition to the permanent purge gas, a reaction gas was also
fed to the reaction chamber during the reduction. The reaction
gas ow rate was kept at 50 mL/min (1 atm, room
temperature), which was composed by a ow of 10% CO in
N
2
and a ow of 20% CO
2
in N
2
. The ratio of p
CO
2
/p
CO
was
controlled by adjusting their corresponding gas ows using
rotameters. Prior to reacting with the solid sample, the reactive
gas was mixed with the N
2
purge; however the p
CO
2
/p
CO
ratio
was not aected by the dilution of the N
2
purge. In an addition
experiment, the TGA was used to check the weight percentage
of Fe
2
O
3
in the nal mixed oxides by reducing the oxygen
carriers in H
2
isothermally at 1173 K.
2.3. Material Characterization. The surface morphology
of the fully oxidized samples was studied using a scanning
electron microscope (JEOL 5800 LV SEM). Powder X-ray
diraction (XRD, Philips PW1820, Cu K, 40 kV and 40 mA,
0.025 per s, with receiving and antiscatter slits of 1,
divergence slit of 0.2 mm, in air at 298 K) was used to detect
any chemical interactions between the iron oxides and the
metal additives. Changes in specic surface area were measured
and calculated from N
2
adsorption and desorption isotherms at
77 K (Micrometrics instrument Tristar 3000) using the
BrunauerEmmettTeller (BET) model.
21
3. RESULTS
3.1. Cycling Experiments. Prior to the cycling experi-
ments, the compositions of the oxygen carriers were
experimentally determined using the TGA, by completely
reducing the samples in hydrogen at 1173 K. The results show
that the samples with 10 and 30 mol % Zr
4+
(cation fraction
only) contain 83.7 and 58.0 wt % Fe
2
O
3
, respectively. To
conrm the composition of the oxygen carriers, X-ray
uorescence (XRF) was performed on both samples with an
S4 Explorer XRF system (Bruker AXS GmbH), and the
contents of Fe
2
O
3
were found to be 58.2 and 82.2 wt %,
respectively, in close agreement with the TGA results. From
this point onward, for convenience, these two mixed oxides are
denoted as Zr10 (10 mol % Zr
4+
) and Zr30 (30 mol % Zr
4+
).
The synthesized materials were examined under gaseous
chemical looping environments at 1123 K in a uidized bed
for 20 cycles. Both materials were cycled either with or without
the nal oxidation step in air, which will oxidize the Fe
3
O
4
fully
back to Fe
2
O
3
. In a typical cycle, the material was exposed to
the following gaseous environments: (i) 2 min N
2
purge, (ii) 5
min reduction in 10% CO, (iii) 2 min N
2
purge, (iv) 3 min
oxidation in 14.6% CO
2
, (v) 1 min N
2
purge, and (vi) 2 min
oxidation in air. When the air oxidation was absent, steps (v)
and (vi) were omitted in each cycle. The CO
2
oxidation step
(iv) is the thermodynamic equivalent of steam oxidation at 850
C, which oxidizes Fe and FeO to Fe
3
O
4
.
11
This validity of the
substitution of steam by CO
2
was experimentally veried by
reproducing the same cyclic conversion in preliminary
experiments using diluted H
2
O over ten cycles at 850 C in
the uidized bed (shown later in Figure 2). The purpose for
substituting steam oxidation with CO
2
is to reduce back-mixing,
which would otherwise have a much larger mixing time when
condensation of steam is required. A typical concentration
prole, during the redox cycles using CO
2
and air as the
oxidants, is depicted in Figure 1. During CO reduction, the gas
mole fraction prole exhibited a sharp peak of CO
2
, produced
during the reduction of Fe
2
O
3
to FeO, followed by a CO
2
shoulder which nominally corresponds to the reduction of FeO
Industrial & Engineering Chemistry Research Article
dx.doi.org/10.1021/ie302626x | Ind. Eng. Chem. Res. 2012, 51, 1659716609 16599
to Fe; as complete reduction to Fe was approached, the CO
2
concentration decreased to zero, just before the start of the N
2
purge. During CO
2
oxidation, a sharp peak of CO was
produced as a product of the oxidation from Fe to Fe
3
O
4
. From
cycle 4 to cycle 7, it can be seen that the pattern of the gas mole
fraction proles in each cycle is highly repeatable. In each cycle,
the capacity of the iron oxides in the sample to produce
hydrogen was calculated from the o-gas concentration proles
using the following eqs 9 and 10, which apply, respectively, to
oxidation with steam or CO
2
:

N N
y
y
t
1
d
H N
H
H
2 2
2
2
(9)

=

+

N N N y t ( ) d
CO N CO
CO 2 2 (10)
where N
H
2
is the number of moles of hydrogen or CO
equivalent produced in each cycle in mol; N

H
2
and N

CO
2
are the
molar ow rates of nitrogen and CO
2
, respectively, during
oxidation, in mol s
1
; y
H
2
and y
CO
are the gas phase mole
fractions of H
2
and CO, as measured by the gas analysers, after
any water had been removed. The hydrogen yield of the iron
oxide, X
H
2
, was dened as the ratio of the total moles of
hydrogen, or CO equivalent, produced in a cycle to the
theoretical amount of hydrogen from complete steam or CO
2
oxidation of metallic iron per gram of Fe
2
O
3
present:
=

X
N
x
1000
0.5
1
16.67
100%
H
H ,CO
Fe O
2
2
2 3
(11)
where x
Fe
2
O
3
is the solid phase mass fraction of Fe
2
O
3
present in
the sample: the amount of hydrogen which can be theoretically
produced by 1 g of Fe
2
O
3
in a redox cycle is 16.67 mmol. The
same analysis was used to calculate the normalized yield of
CO
2
: this yield also corresponded to the conversion of the
iron(III) oxide to Fe, based on the amount of CO
2
leaving the
bed in the oxide-reducing part of a cycle. Thus:

=


+
N N y dt
CO CO N
CO 2 2
2 (12)
=

X
N
x
1000
0.5
1
18.75
100%
CO
CO
Fe O
2
2
2 3
(13)
where N
CO
2
is the number of moles of CO
2
produced in each
cycle in mol; N

CO+N
2
is the total molar ow rate of the gases
during reduction, in mol s
1
; y
CO
2
is the gas phase mole fraction
of CO
2
, measured by the gas analyzer; x
Fe
2
O
3
is the mass fraction
of Fe
2
O
3
present in the sample. The yield of the iron oxide to
iron, X
CO
2
, was dened as the ratio of the total moles of CO
2
produced in a cycle to that theoretically possible by reducing all
the Fe
2
O
3
to Fe, viz., 18.75 mmol/g Fe
2
O
3
. In the case of
cycling without the air oxidation stage, X
CO
2
was calculated by
assuming complete conversion from Fe
3
O
4
to Fe, which would
give 16.67 mmol CO
2
/g Fe
2
O
3
.
It can be seen that, at 1123 K, the coprecipitated material
showed very stable performance with high oxygen carrying
capacity and consequent yield of hydrogen of >90% of
stoichiometric, over 20 cycles. In the absence of the air
oxidation stage, there is a slight decay in the yield of hydrogen
as a function of cycle number after the fth cycle: nevertheless,
the yield of hydrogen at cycle 20 still exceeds 90 mol %. It
should be noted, however, that Figure 2 does not provide
information on any change in the kinetics of this material, as
slower kinetics could have resulted in the same hydrogen yield
per cycle given the reaction periods of 300 s reduction and 180
s oxidation with CO
2
. To explore the eect of cycling on the
kinetics, the apparent rates of conversion, dX/dt, as a function
of solid conversion, X, were compared for dierent cycles, as
depicted in Figure 3. Prior to plotting Figure 3, the observed
rate curves were deconvoluted to allow for mixing in the
sampling system using a rst-order mixing time of 13 s, which
was experimentally determined using a step change of tracer gas
in a bed without reaction under the same operating conditions.
The details of the deconvolution are given in the Appendix 1. It
can be seen that the rates do vary very slightly with number of
Figure 1. Concentration proles of the o gases from a uidized
experiment, during the 4th, 5th, 6th, and 7th cycles when the material
Zr30 was cycled at 850 C in a uidized bed. CO
2
was used to simulate
steam oxidation. The thick black line, thick gray line, and thin dashed
black line correspond to CO, CO
2
, and O
2
respectively. The cyclic
hydrogen yield as a function of cycle number, X
H
2
, is plotted in Figure
2 for the oxidation to produce H
2
(using steam) or CO (using CO
2
).
These yields were calculated assuming that ZrO
2
does not participate
in the redox reactions.
Figure 2. H
2
yield as a function of cycle number, based on the number
of Fe
2
O
3
available in the ZrO
2
/Fe
2
O
3
mixed oxides during chemical
looping operation at 850 C. 100% H
2
yield corresponds to all iron
species being fully reduced to Fe and oxidized by steam or CO
2
to
Fe
3
O
4
. The reduction period was 300 s, and the CO
2
or steam
oxidation period was 180 s in each cycle. The symbols denote dierent
compositions of metal oxides with dierent oxidation sequences in
each cycle.

: Zr10 by steam and air;

: Zr30 by steam and air; :
Zr30 by CO
2
and air; : Zr30 by CO
2
only.
Industrial & Engineering Chemistry Research Article
dx.doi.org/10.1021/ie302626x | Ind. Eng. Chem. Res. 2012, 51, 1659716609 16600
cycles, but more interestingly, there are two distinct peaks in
the proles of rate of reduction. Before further commenting
further on these rate curves, the rate of mass transfer from the
bulk gas phase to the particulate phase must be evaluated. An
estimate of the coecient of mass transfer was made using a
modied Fro ssling equation:
22
= + Sh Re Sc 2 0.6
mf
0.8 0.6
(14)
where Re = 2.5(U
mf
/
mf
)(d
p
/). In the experiments, the
minimum uidization velocity of alumina sand in N
2
was
calculated from the correlation of Wen and Yu
20
as U
mf
= 0.07
m s
1
,
mf
0.42 is the voidage of the bed at minimum
uidization, d
p
= 5.50 10
4
m is the diameter of a carrier
particle, is the kinematic viscosity of the gas phase, and Sc =
/D, with D being the molecular diusivity of CO in N
2
and
CO
2
in N
2
during CO reduction and CO
2
oxidation,
respectively. Because of the relatively dilute concentration of
the product stream, as seen in Figure 1, the eect the product
gases on the overall diusivity of the reactant gases was
neglected in this calculation. The external mass transfer also
involves equimolar counter-diusion, so that advective motion
of the reactant and product gases does not have to be included
in the calculation. The resulting coecient of mass transfer, k
g
= (DSh/d
p
), takes the values of 0.49 and 0.47 m s
1
for CO
reduction and CO
2
oxidation, respectively. In the case of
reducing or oxidizing Zr30 in cycling experiments, these values
of k
g
correspond to theoretically maximum rates of conversion
of 0.19 and 0.32 s
1
for reduction and oxidation, respectively,
both of which are 1 order of magnitude higher than the
measured rates of reaction. Hence, it has been veried that the
rates of conversion shown in Figure 3 were not signicantly
aected by external mass transfer. Other factors contributing to
the rates of reaction observed in the cycling experiments
include the rates of the intrinsic chemical reactions, the rates of
diusion of gaseous species through layers of product, and the
rates of diusion of gases within the porous structure of the
solid. Although these component rates were not investigated
quantitatively in this study, it is reasonable to expect the
chemical rate and the rate of diusion in the product layer to be
correlated with the specic surface area of the particles. On the
other hand, the rate of intraparticle diusion would drop
considerably over cycles, if the pore structures were gradually
destroyed as a result of sintering. A change in any of the
component rates might aect the overall observed rate.
As noted above, several interesting features are seen in Figure
3. Firstly, during CO reduction in each cycle, the rate curve
shows a dip between two distinct peaks. The dips appear
consistently when the conversion reaches approximately 0.33 or
0.25, depending, respectively, on whether air oxidation stages
were included or not. At the reaction temperature, it is
observed that the rates of reactions 5 and 7 were so fast, that
their rates formed a single peak and could not be distinguished.
Therefore, the two peaks can be assigned to the fast phase (i.e.,
reactions 5 and 7) and the slow phase (i.e., Fe
x
O + CO xFe
+ CO
2
), respectively, where x takes the value between 0.892
and 0.947 at 1123 K and varies with the redox potential, e.g.,
Figure 3. The deconvoluted rate of conversion of the solid as a function of total solid conversion, when the material was investigated in a uidized
bed at 850 C, undergoing (a) CO reduction, (b) CO
2
oxidation when air oxidation was included at the end of each cycle, or (c) CO reduction, (d)
CO
2
oxidation when air oxidation was not included. The results of cycles 2, 5, 10, and 20 are presented and compared.
Industrial & Engineering Chemistry Research Article
dx.doi.org/10.1021/ie302626x | Ind. Eng. Chem. Res. 2012, 51, 1659716609 16601
p
CO
2
/p
CO
. According to the correlation of Giddings and
Gordon
23
the transition from wustite to iron takes place
when x = 0.947 at 1173 K. Following air oxidation, the total
oxygen loss during the Fe
2
O
3
to Fe
0.947
O transition is about
33% of the total oxygen loss during the full transition from
Fe
2
O
3
to Fe. When the air oxidation step was not included, the
highest oxidation state possible is Fe
3
O
4
, and the oxygen loss
during the transition from Fe
3
O
4
to Fe
0.947
O is about 25% of
the total oxygen loss during the transition from Fe
3
O
4
to Fe.
The rate proles as a function of conversion for the transition
from wu stite to iron are consistent with a nucleation
mechanism.
Secondly, it can be seen that, during reduction, the height of
the rst rate peak decreased slightly with increasing cycle
number. The decrease in maximum reaction rate was found to
be accompanied by a decrease in BET surface area, although at
the magnitudes of surface area measured (e.g., decreasing from
2.9 to 1.9 m
2
g
1
after the rst 10 cycles in a uidized bed), a
quantitative link is dicult to establish, primarily owing to the
large errors in surface area measurement. The study of kinetics
by Bohn et al.
5
gave a similar conclusion, i.e., the rst maxima
are related to the surface area of the sample. The surface
morphology of the oxygen carriers, as photographed by SEM in
Figure 4, is roughly in agreement with the above speculation, as
grain sizes appear to be larger after 10 cycles. However, as seen
in Figure 4, the porosity of the material was preserved at the
micrometer scale, suggesting that intraparticle diusion was not
signicantly aected by cycling. In addition, this sintering eect
was found to be nondensifying, as the sieve size of the
recovered Zr30 particles from the uidized bed experiments
was found to be unchanged after 20 cycles of chemical looping.
The preservation of this porosity is likely to be an important
factor in maintaining the cyclic conversion.
The second maximum of the rate curves occurs during the
formation of Fe from wustite, and this feature is very similar to
that seen when wustite in unsupported iron oxide, or in a Fe
2
O
3
+ Al
2
O
3
mixture, is reduced.
24
The rates of formation of iron
from wustite were not aected by the possible loss of surface
area noted above and showed a very consistent prole over 20
cycles, regardless of the presence or absence of air oxidation.
A decrease in the maximum rate of reaction is also seen from
Figure 3b,d, whereas the characteristics of the rate of
conversion curves did not seem to vary in the absence of air
oxidation. The fact that there is a shoulder (i.e., at X = 0.4) in
the rate prole rather than a dip in reaction rate at the
conversion where wustite would be expected to oxidize to
Figure 4. SEM images of the freshly prepared Zr30 (left) and Zr30 10 redox cycles at 1123 K (right). Both materials are in their completely oxidized
form.
Figure 5. XRD patterns of the modied Fe
2
O
3
oxygen carriers of (a) freshly prepared 10 mol % Zr, (b) 10 mol % Zr after 10 cycles, (c) freshly
prepared 30 mol % Zr, and (d) 30 mol % Zr after 10 cycles. Peaks corresponding to crystalline phases were marked by the following symbols: ,
Fe
2
O
3
;

, ZrO
2
tetragonal;

, ZrO
2
monoclinic.
Industrial & Engineering Chemistry Research Article
dx.doi.org/10.1021/ie302626x | Ind. Eng. Chem. Res. 2012, 51, 1659716609 16602
magnetite suggests that the oxidation of wustite to magnetite is
not limited by nucleation (in contrast to the equivalent
reduction reaction). Since the shoulder appears after the
conversion expected for this transition, there must be
signicant overlap between the oxidation from Fe to wustite
and wustite to Fe
3
O
4
. Alternatively, there may be some eect of
the support providing an alternative reaction route, as outlined
below. To investigate the possibility of the alternative reaction
route via the support, XRD patterns of the materials were
examined to explore this peculiar feature exhibited by Zr30
during CO
2
oxidation. The results of this analysis for both fresh
and cycled material were inspected for the formation of any
unexpected phases and are shown in Figure 5. In both Zr10 and
Zr30 samples, the phases present were Fe
2
O
3
and two phases of
ZrO
2
: monoclinic (Baddelyite) and tetragonal zirconia. The
coexistence of the two zirconia phases is not uncommon. When
ZrO
2
was mixed with Fe
2
O
3
, the phase transition from the
tetragonal to the monoclinic phase could have taken place at
the reaction temperature of 1123 K.
25
It is not apparent how
the performance of the oxygen carrier was aected by the
presence of a mixture of two phases, and a possible explanation
is given below in the Discussion. In addition, the relative peak
intensities and peak broadening of the two ZrO
2
phases remain
unchanged, which veries the inertness of the ZrO
2
support
after the material preparation step. On the other hand, the peak
intensity of the Fe
2
O
3
phase decreases and peak widths
increased after cycling, both indicating that the average size of
the crystal grains of the iron oxides decreased over cycles.
Another possible contribution to the slight decrease of the
overall cyclic conversions and the maximum rates is attrition
and elutriation of the resulting ne carrier particles. It is dicult
to characterize the rate of attrition in the current study since the
experimental design was limited to the investigation of the
reactivity of the oxygen carriers only. However, the rate of
attrition and elutriation would need to be investigated, should
this material be considered for application in a continuous
system.
3.2. Eect of Redox Potential. In the presence of an
alumina support, previous work
15
has shown that the rate of
reaction, as well as the ultimate conversion of the active iron
oxide, decreased, owing to the formation of the spinel FeAl
2
O
4
and solid solutions of the active iron oxides in the spinel phases.
This hindering eect of the support material on the reactivity
became more pronounced when the thermodynamic driving
force during reduction was lowered, i.e., when p
CO
2
/p
CO
was
increased.
15
To verify the improvement obtained using ZrO
2
as
a support material, similar experiments to those of Cleeton
15
were performed, in which Zr30 was reduced by various
reducing gas mixtures of CO
2
+ CO or H
2
O + H
2
, with varying
p
CO
2
/p
CO
or p
H
2
O
/p
H
2
. These experiments were performed in
both the packed bed reactor and the TGA. The TGA was able
to detect very slow rates of reaction. However, the
conguration of the TGA only allowed for a gas mixture of
CO
2
and CO with a p
CO
2
/p
CO
ratio between 0.6 and 2.4,
because p
CO
2
/p
CO
beyond this range required a much larger
total gas ow rate, which would have caused disturbance in the
proposed gravimetric measurements. On the other hand, in a
packed bed experiment, it was easier to increase gas ow rates,
and injection of steam was possible; thus, it was possible to
employ a wider range of redox potentials, i.e., p
CO
2
/p
CO
or
p
H
2
O
/p
H
2
. The gas compositions and ow rates for the packed
bed and TGA experiments are summarized in Table 1. The
solid conversions during reduction were converted to
percentage hydrogen yield equivalents and plotted in Figure 6.
In the packed bed experiments, Zr30 was cycled 20 times at
850 C. In each cycle, the ratio of partial pressures of the
reducing gas to the oxidizing gas was varied in a random order,
to avoid any potential systematic error. When a mixture of H
2
O
and H
2
was used as the reducing gas, the reducing potential of
p
H
2
O
/p
H
2
was converted to a p
CO
2
/p
CO
equivalent, by means of
the equilibrium constant for the water-gas shift reaction:
+ + = K H O CO CO H 0.7760
2 2 2 p,1123 K (15)
where the value of K
p,1123 K
was calculated from the correlation
of Moe:
26
Table 1. Gas Flow Arrangements for Experiments to Investigate the Rate and Extent of Conversion of Zr30 under Various
Redox Environments at 850C
a
reducing gas oxidizing gas
experiment series composition ow rate composition ow rate p
oxidant
/p
reductant
reduction time (min)
PBR1 10% CO 0.634.1 L/min 100% CO
2
0.121 L/min 02.3 8
PBR2 10% CO 0.470.95 L/min 15% CO
2
00.70 L/min 01.8 8
PBR3 5% H
2
0.461.74 L/min 2.6% H
2
O 01.79 L/min 02.35 5
PBR4 10% CO 0.951.43 L/min 15% CO
2
00.25 L/min 01.42 5
TGA 10% CO 22.938.5 mL/min 20% CO
2
11.727.5 mL/min 0.62.4 300
a
PBR stands for packed bed reactor. All ow rates were measured at room temperature and 1 atm. The balancing gas is N
2
in all cases.
Figure 6. Potential yield of hydrogen from reduced Zr30, as a function
of reducing potential (represented by p
CO2
/p
CO
). The hydrogen yield
is calculated on the basis of the amount which could theoretically be
obtained during oxidation with steam. The reaction conditions are
summarized in Table 1. The symbols corresponds to dierent
experimental series:

, PBR1;

, PBR2; , TGA; , PBR3; ,
PBR4. The ideal H
2
yield, assuming complete solid conversion, is
indicated by the dotted line.
Industrial & Engineering Chemistry Research Article
dx.doi.org/10.1021/ie302626x | Ind. Eng. Chem. Res. 2012, 51, 1659716609 16603
= +

K
T
exp 4.33
4577.8
p
(16)
in which T is temperature in kelvin. The yield of hydrogen,
plotted in Figure 6, was estimated on the basis of the total
number of moles of gaseous product generated in each cycle
when the reduced sample was oxidized by either CO
2
or H
2
O,
as given by eq 13. In the case of TGA experiments, fresh Zr30
particles were reduced under dierent p
CO
2
/p
CO
, and the total
weight loss during reduction was converted to its correspond-
ing hydrogen yield, assuming that all reduced iron oxides could
be fully oxidized to Fe
3
O
4
by steam. Results are shown in
Figure 6, which suggests that, if reduction is only to wustite
(i.e., p
CO
2
/p
CO
> 0.58), the yield of hydrogen is much closer
to the theoretical value. Considering, now, the xed bed reactor
results for p
CO
2
/p
CO
< 0.58, for which, thermodynamically,
reduction to Fe is possible, it should be remembered that the
amount of hydrogen produced reects the degree of reduction
achieved during the reduction stage, which had a xed duration.
Figure 6 shows that, when reduction from wustite to Fe is
thermodynamically favored, the hydrogen yield appears to
decrease as the reducing potential decreases. The reason for
this apparent correlation lies in the kinetics of reduction. To see
this, in Figure 6, with p
CO
2
/p
CO
= 0, it took 5 min for the Fe
3
O
4
in Zr30 to be 90% converted to Fe when 10 vol % CO in N
2
was used as the reductant. Assuming chemical reaction control,
the driving force for the reaction is ([CO] [CO
2
]/K
p
), so
that as the [CO
2
] in the reducing gas is increased, for a xed
[CO], the observed rate should fall. For example, in
experimental series PBR4 in Figure 6, with p
CO
2
/p
CO
= 0.42,
the chemical driving force during the oxidation of Fe to wustite
was about 17.5% of that when p
CO
2
/p
CO
= 0. To estimate
roughly the eect of the change in the chemical driving force on
the rate of reaction, the following assumptions about the
kinetics were made: (i) rapid and complete conversion of
magnetite to wustite prior to the subsequent reduction of
wustite to iron, (ii) the apparent rate of reaction during the
oxidation of Fe to wustite is directly proportional to ([CO]
[CO
2
]/K
p
), and (iii) the apparent rate of reaction is
approximately independent of the conversion from wustite to
iron, as observed in Figure.3c, i.e., the conversion increases
linearly with time. Thus, the expected total hydrogen yield is
25% + 17.5% 75% = 38% when the reduction period is 5 min.
This compares favorably with the 38% yield observed
experimentally and plotted in Figure 6. From the results in
Figure 6 for reduction only to wustite (i.e., p
CO
2
/p
CO
> 0.58),
the hydrogen yield is directly related to the conversion from
magnetite to wustite during the partial reduction, a fact studied
in detail below.
When the reducing condition favors wustite as the product of
the reduction, the concentration of Fe
2+
in wustite is a strong
function of redox potential,
23
i.e., p
CO
2
/p
CO
or p
H
2
O
/p
H
2
. At
1123 K, the O/Fe ratio at equilibrium is given by the
correlation:
23
=
+ p p
O/Fe
ln( / ) 27.902
25.904
CO CO
2
(17)
The resulting yield of hydrogen based on the theoretical wustite
composition was compared with the experimental results in
Figure 7, where the yields of hydrogen were calculated on the
basis that 100% yield corresponds to the hypothetical transition
from Fe
3
O
4
to hypothetical FeO (i.e., Fe
x
O with x = 1) during
the reduction stage. Except for the experimental set PBR1, most
experimental results in Figure 7 are in agreement, showing that,
in a xed amount of time, the material was nearly reduced to
the equilibrium wustite composition given by Giddings and
Gordon,
23
albeit with a slight oset. To investigate this oset,
batches of Fe100 were investigated in a TGA using the identical
experimental protocol to that used in the packed bed, to
determine experimentally the equilibrium line for the pure iron
oxide system. The results are shown in Figure 7 and indicate
that, under the same experimental conditions, Zr30 gave almost
identical magnetite to wustite conversion as a function of redox
potential as Fe100, with both materials displaying about a 10%
oset from literature values.
23
In this study, the Zr30 samples
were therefore considered to have 100% conversion from
magnetite to wustite when p
CO
2
/p
CO
was between 0.6 and 2.4;
i.e., the presence of ZrO
2
did not change the phase boundaries
of the iron oxide system. In comparison, the FeAlO system
gives much lower equilibrium values of x and higher Fe/O ratio
under the same p
CO
2
/p
CO
, and the kinetics of reduction are
slower, owing to the much lower thermodynamic driving forces
arising from the presence of a solid solution of Fe
3
O
4

FeAl
2
O
4
.
15
Therefore, using ZrO
2
as a support material gives
much improved performance over the Al
2
O
3
support.
The rate of reduction as a function of time is further
examined, in Figure 8, where conversions of the solid are
normalized with respect to a target product of Fe
0.92
O. Prior to
discussing the rate of reaction for dierent experiments, the
eect of external mass transfer between the reactant gas owing
over the TGA pan and the upper surface of the bed of particles
in the pan was estimated using a Stefan-Maxwell Diusion
model, described in Appendix 2, Supporting Information. The
model suggested that the observed rates of reduction of Zr30
were suciently fast during the rst 400 s for this source of
external mass transfer to play a signicant role in controlling
the overall kinetics. After the rst 500 s, the concentration on
the surface of the solid can be approximated as the
concentration in the bulk gas phase and the rate is no longer
limited by external mass transfer. Thus, one can conclude from
Figure 8 that, the magnitudes of the rates of conversion of Zr30
Figure 7. The percentage conversion from magnetite to wustite, as a
function of reducing potential (represented by p
CO2
/p
CO
), when Zr30
was reduced in a packed bed reactor or a TGA. The reaction
conditions are summarized in Table 1. The symbols corresponds to
dierent experimental series:

, PBR2; , TGA; , PBR4; +, TGA
experiments with Fe100. The dashed line corresponds to the
maximum theoretical hydrogen yield based on literature.
23
Industrial & Engineering Chemistry Research Article
dx.doi.org/10.1021/ie302626x | Ind. Eng. Chem. Res. 2012, 51, 1659716609 16604
were comparable with those of the unsupported iron oxide in
the TGA, both being signicantly higher than that seen with
Al10 and considerably aected by external mass transfer
resistance. In contrast, owing to the slow chemical rate of
reduction of Al10, the external mass transfer limitation had little
eect on its apparent rate in the TGA experiments.
15
3.3. Extent of Carbon Deposition. One of the advantages
of generating hydrogen using the steam-iron reaction is that
contamination of the hydrogen product by CO can be
theoretically completely avoided. However, in practice, the
hydrogen could be contaminated if carbon were deposited on
the solid particles during reduction by carbonaceous species.
When CO is used, the Boudouard reaction may occur:
+ 2CO CO C
2
(18)
which is thermodynamically favored by low p
CO
2
/p
CO
ratios and
low temperatures. This reaction is also sensitive to the nature of
the solid surfaces, e.g., an oxidizing solid surface would shift this
reaction to the left and hence inhibit the formation of carbon.
During steam oxidation, the deposited carbon is oxidized by
H
2
O to give CO, which is inevitably mixed with the product
hydrogen. As CO is poisonous to Pt anodes in PEM fuel cells,
the contamination by CO would aect the value of the nal
product and ideally needs to be minimized. In a set of packed
bed experiments, the reduced Zr30 material was reacted with
steam to generate hydrogen, and the mole fraction of CO in the
o-gas was monitored: the limit of detection of the analyzer
used was 0.001 vol %. The reducing conditions, viz.,
temperature, reduction time, and reduction potential, were
varied orthogonally, to nd a suitable operating regime for the
production of hydrogen of high purity. The CO contamination
in the hydrogen during hydrogen production is presented as
N
CO
/p
H
2
in Figures 9 and 10. Both plots indicate that, under
the current operating conditions, viz., 10231223 K, 0 < p
CO
2
/
p
CO
< 0.40, no deposition of carbon could be detected when
the hydrogen yield was below 95%, which corresponds to a
95% conversion from Fe
3
O
4
to Fe. When the solid was reduced
further, the Boudouard reaction was manifested through the
production of a small amount of CO during steam oxidation.
One common way of suppressing carbon formation in chemical
looping operations is to inject steam with the reactant gases,
accordingly, when the ratio p
H
2
O
/p
CO
was set to 0.23 in the
reducing gas during the packed bed experiments and no CO
was detected during the subsequent hydrogen generation step
for any degree of reduction of the carrier to Fe. Nevertheless, it
should be noted that, since the mole fractions of the reactive
gases used in the experiments were very low, e.g., 2.6 vol %
steam during steam oxidation, the lower detection limit of the
overall CO contamination, N
CO
/N
H
2
, was limited to the order
of 500 ppm, which was an order of magnitude higher than the
upper CO limit of 50 ppm, for use in a PEM fuel cell. To verify
the purity of the hydrogen product with the current detection
limit, hydrogen of much higher concentrations, e.g., 50%, would
need to have been generated, which could only have been
achieved using 100 vol % steam, not practicable with the
current experimental apparatus.
4. DISCUSSION
4.1. Stability and Reactivity of the Iron Oxide. During
the cycling experiments, Zr30 produced on average 95 mol % of
the theoretical hydrogen yield, based on the amount of Fe
2
O
3
present. This corresponds to 9 mol of H
2
/kg oxygen carrier/
Figure 8. The conversion of Fe
3
O
4
to Fe
0.92
O as a function of time,
when three dierent samples of iron oxides were reduced by dierent
CO
2
and CO mixtures in a TGA at 850 C. Fe100 stands for
unsupported iron oxide prepared by wet granulation. Al10 corresponds
to a sample containing 10 wt % Al
2
O
3
and 90 wt % Fe
2
O
3
, results of
which were reported by Cleeton.
15
The numbers after each sample
denote the ratio of p
CO2
/p
CO
during the experiment. All samples were
particles of the size fraction 300425 m.
Figure 9. The degree of CO contamination in the hydrogen product,
as a function of cyclic hydrogen yield, when dierent reducing gases
were used at a constant temperature of 1123 K. The hydrogen yield
was experimentally controlled by adjusting the length of reduction.
The symbols correspond to dierent reducing potentials. , CO in N
2
;

, p
CO2
/p
CO
= 0.21; , p
CO2
/p
CO
= 0.31;

, p
CO2
/p
CO
= 0.40; , p
H2O
/
p
CO
= 0.23.
Figure 10. The degree of CO contamination in the hydrogen product,
as a function of cyclic hydrogen yield, when dierent reducing gases
were used at dierent temperatures using 10% CO in N
2
as the
reducing gas. The hydrogen yield was controlled by adjusting the
length of reduction.
Industrial & Engineering Chemistry Research Article
dx.doi.org/10.1021/ie302626x | Ind. Eng. Chem. Res. 2012, 51, 1659716609 16605
cycle, when the iron content of the material was oxidized from
Fe to Fe
3
O
4
. This yield is higher than the values reported by
previous literature, e.g., Kidambi et al. (6.9 mol/kg using 75 wt
% Fe
2
O
3
on Al
2
O
3
),
14
Kierzkowska et al. (7.5 mol/kg using 60
wt % Fe
2
O
3
on Al
2
O
3
),
15
Galvita and Sundmacher (1.7 mol/kg,
using a 5 wt % Cr
2
O
3
, 38 wt % Fe
2
O
3
, and 57 wt % Ce
0.5
Zr
0.5
O
2
mixture),
9
and Bohn et al. (3.5 mol H
2
/kg by limiting the
degree of reduction to wustite using Fe100).
2
In contrast to the
results of Kidambi et al.,
14
although there is a very slight
decrease in the yield of hydrogen as a function of cycle number
in the absence of the air oxidation stage, Zr30 still shows
superior performance, which gives more freedom in the process
design, if one wishes to maximize the hydrogen production. In
terms of performance, this material is an improvement over the
Fe
2
O
3
Al
2
O
3
oxygen carrier, which loses a large fraction of the
stoichiometric yield of hydrogen after cycling, owing to the
formation of the unreactive FeAl
2
O
4
Fe
3
O
4
solid solution.
14,15
The apparent kinetics obtained from the uidized bed
experiments suggest that, overall, the surface area of the
material decreased with cycling. The evolution of surface
morphology, as examined by electron microscope, suggests that
the loss of surface area was associated with sintering (Figure 4);
however, the pore structure of the solid was preserved at a
micrometer scale, (as seen in Figure 4), which may have helped
to maintain the reaction kinetics by preserving a high value of
the eective intraparticle gas diusivity. Of course, the presence
of ZrO
2
was a key to the structural stability, because of its high
melting point of 2715 C and its chemical inertness (as veried
by XRD in Figure 5), so the sintering was nondensifying. The
method of coprecipitation used in the preparation of the carrier
ensures good mixing of the support material with the active
Fe
2
O
3
, so that the sintering resistance provided by ZrO
2
was
maximized, and the overall thermal stability and cyclic stability
of the oxygen carrier were signicantly improved. It is probable
that the cost of manufacturing these oxygen carriers by
coprecipitation might be higher than those made by more
robust methods, such as mechanical mixing, impregnation, and
spray drying.
27
However, the method of coprecipitation
provides a good, reproducible base case for supporting iron
oxides by ZrO
2
. Optimization of the trade-o between cost of a
particular manufacturing route and performance would of
course need to be considered in any future large-scale
application.
4.2. Inertness of the Support Material. During chemical
looping operations, generally, the support material may
participate in the redox reaction, by forming tertiary metal
oxides with the active material, e.g., FeAl
2
O
4
in the FeAlO
system. In the case of Al
2
O
3
, the chemical interaction between
the support material and the active oxides is undesirable, as the
resulting spinel, i.e., FeAl
2
O
4
, is extremely unreactive and has a
negative impact on the overall materials performance.
15
This is
not the case for ZrO
2
-supported material, and the active iron
oxide on ZrO
2
has comparable reactivity to iron oxide on its
own, as shown in Figure 8.
The results in Figure 8 may be further interpreted by
considering the nature of the reduction of Fe
3
O
4
to wustite.
According to Bohn et al.,
5
for unsupported Fe
2
O
3
, when the
rate is reaction controlled, the reduction of magnetite to wustite
follows:

=
=

k
K
k
K
f X
rate [CO]
[CO ]
[CO]
[CO ]
( )
i
mag
2
p
0
mag
2
p
(19)
where k
i
is the intrinsic rate constant, k
0
is the initial rate
constant,
mag
is the density of magnetite, K
p
is the equilibrium
constant at the magnetitewustite boundary, and f(X) is a
function describing the contribution of solid conversion, X.
Bohn
10
suggested that, for unsupported Fe
2
O
3
with an average
particle size d
p
= 363 m, similar to that used in the TGA in the
present work, the rate was dominated by chemical reaction and
the eectiveness factor, = 1 at intermediate conversions (e.g.,
X > 0.2). If the same applies here, one can verify the order of
reaction with respect to the chemical driving force [CO]
[CO
2
]/K
p
, by dening a specic rate, K
s
as:
=

K
m t
y
d /d
y
K
s
CO
CO
2
p (20)
where dm /dt is the rate of fractional mass loss in the TGA
experiments. Values of K
s
were calculated for three dierent
experiments, where the mole fractions of CO and CO
2
at the
surface of the particle were evaluated using the Maxwell-Stefan
model. The solid conversion, X, was normalized on the basis of
the assumption that full conversion to the equilibrium state of
wustite (X = 1) is achieved when the rate of reaction drops to
zero. This assumption was veried by the control experiments
using Fe100, as see in Figures 7 and 8. If there is no solid
solution eect, i.e., the activity of the solid phases is 1, then the
specic rate can be correlated with the rate from eq 19, i.e., K
s
=
4.833 k
0
f(X), where the value of 4.833 is the ratio of the
molecular mass of Fe
3
O
4
to the atomic mass of oxygen and k
0

f(X) is independent of the concentrations of the gaseous
species.
The results of the use of this interpretation of K
s
(Figure 11)
shows a large discrepancy between values of f(X) at lower
conversions (X < 0.2), whereas for conversions between X =
0.2 and 1.0, all three experiments showed overlapping proles
of K
s
. The disagreement between the curves of K
s
in the early
Figure 11. The specic rate as a function of the fractional solid
conversion, X, during the reduction of magnetite to wustite for
dierent experiment sets in a TGA. The numbers denote the ratio of
CO
2
/CO used in each experiment.
Industrial & Engineering Chemistry Research Article
dx.doi.org/10.1021/ie302626x | Ind. Eng. Chem. Res. 2012, 51, 1659716609 16606
stage might be a result of large uncertainties in the estimation
of the gas concentration proles at the gassolid interface.
Also, because of the low K
p
in this reaction, the calculation of
the chemical driving force, y
CO
(y
CO
2
/K
p
) becomes highly
sensitive to small uncertainties in the concentration measure-
ment. Diculties in estimating accurately the rate of reaction at
small conversions (X < 0.2) were also found by Bohn et al.,
5
when the Fe
3
O
4
Fe
x
O transition of unsupported iron oxides
was studied in a uidized bed. Between X = 0.2 and 0.8, the
specic rates are comparable for each experiment, suggesting
that the rate of reaction is indeed rst order with respect to y
CO
(y
CO
2
/K
p
), and the ZrO
2
support imposes no solid solution
eects on the kinetics.
The other issue when ZrO
2
is used alone as the support
material is the monoclinic-tetragonal phase transition which,
thermodynamically, could occur at temperatures typical of
chemical looping, i.e., 8501000 C. This phase transition is
reversible during the cooling process for pure zirconia.
However, when ZrO
2
is doped interstitially with trivalent
metal ions, e.g., Fe
3+
by coprecipitation in this case, the ZrO
2
could exist in a metastable phase of tetragonal ZrO
2
.
25
The
coexistence of the two phases of ZrO
2
might be responsible for
the shoulder feature observed during the CO
2
oxidation of the
reduced Zr30 (Figure 3d). In general, when ZrO
2
is doped
interstitially with divalent or trivalent ions, high concentrations
of oxygen ion vacancies will be present in the crystal lattice of t-
ZrO
2
, and the ionic conductivity of O
2
in the tetragonal phase
is much greater than that of the monoclinic phase.
28
Since the
primary mechanism for the oxidation of pure Fe to magnetite is
via the diusion of Fe
2+
through p-type metal decit,
29
the
consequential high O
2
mobility through n-type decit in
tetragonal ZrO
2
might assist the rate of oxidation via a
secondary mechanism, which is illustrated in Figure 12. As a
result, when a fraction of unreacted Fe was trapped by the
support material, ionic conduction through the solid would be
the only route by which the oxidation reaction would proceed,
and the fraction of Fe trapped by t-ZrO
2
becomes more reactive
than that trapped by m-ZrO
2
. Of course, this reasoning is only
valid if the rate of the oxidation reaction is largely controlled by
ionic diusion, which is possible at intermediate to high solid
conversions, when a layer of solid product has been established
outside a solid grain within a reduced oxygen carrier particle.
This hypothesis is explored in a later paper.
Despite the possible presence of solid solutions of Fe
3+
and
Fe
2+
in ZrO
2
phases, XRD patterns (Figure 5) indicate that the
chemical composition (i.e., the relative peak intensities) and
physical structure (i.e., the peak broadening eect) of the ZrO
2
phases did not vary after cycling. Therefore, one can rely on the
inertness of the support to make a very stable iron oxide oxygen
carrier for extended chemical looping. If the incomplete phase
transformation of ZrO
2
during the preparation stage is to be
avoided, the addition of other dopants, e.g., Y
3+
or Ce
4+
, could
be used to assist the stabilization of ZrO
2
in its tetragonal form.
Existing literature has already shown that using stabilized ZrO
2
to support Fe
2
O
3
cycles gives promising reactivity for chemical
looping combustion.
30
5. CONCLUSION
Synthetic Fe
2
O
3
based oxygen carriers were made from a
solution precursor using coprecipitation. ZrO
2
was used as an
inert support, which has been shown to stabilize the cyclic
hydrogen yield of Fe
2
O
3
over 20 cycles, without signicant
deterioration of the reaction kinetics. The surface area of the
solid particle was found to decrease as a result of nondensifying
sintering, which had slightly negative impact on the maximum
rate of reaction over cycles, but did not aect the kinetics at
higher conversions. Despite the loss of surface area, the
porosity of the material was preserved over the number of
cycles used. This structural stability was an important factor in
maintaining the kinetics of the redox reactions of the active iron
oxide.
Experiments in a uidized bed reactor, a packed bed reactor,
and a TGA have conrmed that the ZrO
2
did not alter the
thermodynamics of the iron system, nor did it signicantly
aect the kinetics of reduction from Fe
2
O
3
stepwise to Fe.
Although two phases of ZrO
2
coexist in the as-prepared solid
sample, their relative quantity and crystal size were stable
during chemical looping. Carbon deposition was found only
when the conversion from Fe
3
O
4
to Fe exceeded 90 mol %,
although the limit of detection with our apparatus of CO was
500 ppmv CO in the product gas.

ASSOCIATED CONTENT
*S Supporting Information
1) Deconvolution of concentration proles from gas analysers.
2) StefanMaxwell diusion model. This material is available
free of charge via the Internet at http://pubs.acs.org.

AUTHOR INFORMATION
Corresponding Author
*E-mail: wl247@cam.ac.uk.
Notes
The authors declare no competing nancial interest.

ACKNOWLEDGMENTS
The authors would like to thank Mr. Simon Griggs with SEM,
Mr. Zlatko Saracevic with BET measurement, and the Chemical
Data Service at Daresbury for crystallographic data. Dr. Paul
Fennell from Imperial College, London, and members of his
research group, viz., Dr. John Blamey, Mr. Zili Zhang, and Mr.
Tomas Hills, are gratefully acknowledged for their assistance
with the XRF analysis. Financial support from the Engineering
and Physical Sciences Research Council (Grant number: EP/
G063265/1) is also acknowledged.
Figure 12. A schematic diagram of the possible mechanism in which a
grain of tetragonal ZrO
2
participates in the steam or CO
2
oxidation of
reduced iron oxide. The ZrO
2
phase contains interstitial Fe
3+
. The
parameters
1
and
2
represent, respectively, the degree of n-type and
p-type metal decits in the metal oxide lattice.
Industrial & Engineering Chemistry Research Article
dx.doi.org/10.1021/ie302626x | Ind. Eng. Chem. Res. 2012, 51, 1659716609 16607

NOTATION
Parameters
A = surface area of the TGA crucible, m
2
[CO] = concentration of CO, mol m
3
[CO
2
] = concentration of CO
2
, mol m
3
c
tot
= total concentration, mol m
3
D
ij
= gas molecular diusivity of species i in medium j, m
2
s
1
d
p
= particle diameter, m

1
= n-type metal oxide lattice decit,

2
= p-type metal oxide lattice decit,
E
A
= activation energy, kJ mol
1

mf
= voidage of the uidized bed at minimum uidization,
J
i
= ux of species I, mol m
2
s
1
k
g
= external mass transfer coecient, m s
1
k
i
= rst order intrinsic rate constant, s
1
k
0
= initial rate constant, s
1
K
p
= equilibrium constant,
K
s
= specic rate of reaction, mol
1
m
3
s
1
m = mass of samples in a TGA, g
N

i
= rate of change of number of moles of species i, mol s
1
N
i
= number of moles of species i, mol
= kinematic viscosity, m
2
s
1
p
i
= partial pressure of species i, bara
R = universal gas constant, 8.314 J mol
1
K
1
rate = rate of reaction, mol m
3
s
1
Re = Reynolds number,
Sc = Schimdt number,
Sh = Sherwood number,
T = temperature, K
t = time, s
t
d
= dead time, s
U = supercial velocity, m s
1
U
mf
= minimim uidization velocity, m s
1
X = fractional solid conversion,
x = ratio of Fe/O in wustite,
X
CO
2
= percentage CO
2
yield during CO reduction in each
cycle,
X
H
2
= percentage hydrogen yield during steam oxidation in
each cycle -
x
i
= mass fraction of solid species i,
y
i
= mole fraction of gas species i,
z = the vertical distance from the surface of solid samples, m
H
o
T
= standard enthalpy of reaction at temperature in
Kelvin, kJ mol
1
= eectiveness factor,

mag
= density of magnetite, kg m
3
= characteristic mixing time, s
Acronyms
BET = theory of Brunauer, Emmett, and Teller
CLC = chemical looping combustion
i.d. = internal diameter, m
PBR = packed bed reactor
PEMFC = proton exchange membrane fuel cell
SEM = scanning electron microscopy
TGA = thermal gravimetric analyzer
XRD = X-ray diraction
XRF = X-ray uorescence

REFERENCES
(1) Messerschmitt, A. Process of producing hydrogen. U.S. Patent
971,216, 1910.
(2) Choudhary, T.; Goodman, D. CO-free fuel processing for fuel
cell applications. Catal. Today 2002, 77, 6578.
(3) Bohn, C. D.; Muller, C. R.; Cleeton, J. P.; Hayhurst, A. N.;
Davidson, J. F.; Scott, S. A.; Dennis, J. S. Production of very pure
hydrogen with simultaneous capture of carbon dioxide using the redox
reactions of iron oxides in packed beds. Ind. Eng. Chem. Res. 2008, 47,
76237630.
(4) Reed, H.; Berg, H. Hydrogen process. U.S. Patent 2,635,947.
1953.
(5) Cleeton, J. P. E.; Bohn, C. D.; Muller, C. R.; Dennis, J. S.; Scott,
S. A. Clean hydrogen production and electricity from coal via chemical
looping: Identifying a suitable operation regime. Int. J. Hydrogen Energy
2009, 34, 112.
(6) Bohn, C. D.; Cleeton, J. P.; Muller, A. N.; Davidson, C. R.;
Hayhurst, J. F.; Scott, S. A.; Dennis, J. S. The kinetics of the reduction
of iron oxide by carbon monoxide mixed with carbon dioxide. AIChE J.
2009, 56, 10161029.
(7) Bleeker, M. F.; Kersten, S. R. A.; Veringa, H. J. Pure hydrogen
from pyrolysis oil using the steam-iron process. Catal. Today 2007,
127, 278290.
(8) Otsuka, K.; Kaburagi, T.; Yamada, C.; Takenaka, S. Chemical
storage of hydrogen by modified iron oxides. J. Power Sources 2003,
122, 111121.
(9) Galvita, V.; Sundmacher, K. Cyclic water gas shift reactor
(CWGS) for carbon monoxide removal from hydrogen feed gas for
PEM fuel cells. Chem. Eng. J. 2007, 134, 168174.
(10) Galvita, V.; Hempel, T.; Lorenz, H.; Rihko-Struckmann, L. K.;
Sundmacher, K. Deactivation of modified iron oxide materials in the
cyclic water gas shift process for CO-free hydrogen production. Ind.
Eng. Chem. Res. 2008, 47, 303310.
(11) Bohn, C. D. The production of pure hydrogen with simultaneous
capture of carbon dioxide. PhD Thesis, University of Cambridge,
Cambridge U.K., 2010
(12) Li, F.; Kim, H. R.; Shridhar, D.; Wang, F.; Zeng, L.; Chen, J.;
Fan, L.-S. Syngas chemical looping gasification process: Oxygen carrier
particle selection and performance. Energy Fuels 2009, 23, 41824189.
(13) Kierzkowska, A. M.; Bohn, C. D.; Cleeton, J. P.; Scott, S. A.;
Dennis, J. S.; Muller, C. R. Development of iron oxide carriers for
chemical looping combustion using sol-gel. Ind. Eng. Chem. Res. 2010,
49, 53835391.
(14) Kidambi, P. R.; Cleeton, J. P. E.; Scott, S. A.; Dennis, J. S.; Bohn,
C. D. Interaction of iron oxide with alumina in a composite oxygen
carrier during the production of hydrogen by chemical looping. Energy
Fuels 2012, 26, 603617.
(15) Cleeton, J. P. E. Chemical looping combustion with
simultaneous power generation and hydrogen production using iron
oxides. PhD Thesis, University of Cambridge, Cambridge U.K., 2011.
(16) Popovic , S.; Grz eta, B.; S

tefanic , G.; Czako -Nagy, I.; Music , S.


Structural properties of the system m-ZrO
2
--Fe
2
O
3
. J. Alloys Compd.
1996, 241, 1015.
(17) Jiang, J. Z.; Poulsen, F. W.; Mrup, S. Structure and thermal
stability of nanostructured iron-doped zirconia prepared by high-
energy ball milling. J. Mater. Res. 1999, 14, 13431352.
(18) Fischer, W. A.; Hoffman, A. Equilibrium studies in the system
FeO-ZrO
2
. Arch. Eisenhuettenwes 1957, 28, 739743.
(19) Bechta, S. V.; Krushinov, E. V.; Almjashev, V. I.; Vitol, S. A.;
Mezentseva, L. P.; Petrov, Yu.B.; Lopukh, D. B.; Khabensky, V. B.;
Barrachin, M.; Hellmann, S.; Froment, K.; Fischer, M.; Tromm, W.;
Bottomley, D.; Defoort, F.; Gusarov, V. V. Phase diagram of the ZrO
2
-
FeO system. J. Nucl. Mater. 2006, 348, 114121.
(20) Wen, C.; Yu, Y. A generalised method for predicting minimum
fluidisation velocity. AIChE J. 1966, 12, 610612.
(21) Brunauer, S.; Emmett, P. H.; Teller, E. Adsorption of gases in
multi molecular layers. J. Am. Chem. Soc. 1938, 60, 309319.
(22) Dennis, J. S.; Hayhurst, A. N. A simplified analytical model for
the rate of reaction of SO
2
with limestone particles. Chem. Eng. Sci.
1986, 41, 2536.
Industrial & Engineering Chemistry Research Article
dx.doi.org/10.1021/ie302626x | Ind. Eng. Chem. Res. 2012, 51, 1659716609 16608
(23) Giddings, R. A.; Gordon, R. S. Review of oxygen activities and
phase boundaries in wustite as determined by electromotive force and
gravimetric methods. J. Am. Ceram. Soc. 1973, 56, 111116.
(24) Bao, J.; Liu, W.; Cleeton, J. P. E.; Scott, S. A.; Dennis, J. S.; Li,
Z.; Cai, N. Interaction between Fe-based oxygen carriers and n-
heptane during chemical looping combustion. Proc. Combust. Inst.
2012. In press, available from: http://dx.doi.org/10.1016/j.proci.2012.
07.079.
(25) Narwankar, P. K.; Lange, F. F.; Levi, C. G. Microstructure
evolution of ZrO
2
-(Fe
2
O
3
, Al
2
O
3
) materials synthesized with solution
precursors. J. Am. Ceram. Soc. 1997, 80, 16841690.
(26) Moe, J. M. Design of water-gas shift reactors. Chem. Eng. Prog.
1962, 58, 3336.
(27) Adanez, J.; Abad, A.; Garcia-Labiano, F.; Gayan, P.; de Diego., L.
S. Progress in chemical-looping combustion and reforming tech-
nologies. Prog. Energy Combust. 2012, 38, 215282.
(28) Bonanos, N.; Butler, E. P. Ionic conductivity of monoclinic and
tetragonal yttria-zirconia single crystals. J. Mater. Sci. Lett. 1985, 4,
561564.
(29) Birks, N.; Meier, G. H.; Pettit, F. S. High-temperature oxidation of
metals, 2
nd
ed.; Cambridge University Press: Cambridge, 2006; pp 84
86.
(30) Ryde n, M.; Cleverstam, E.; Johansson, M.; Lyngfelt, A.;
Mattison, T. Fe
2
O
3
on Ce-, Ca-, or Mg-stabilized ZrO
2
as oxygen
carrier for chemical looping combustion using NiO as additive. AICheE
J. 2010, 56, 22112220.
Industrial & Engineering Chemistry Research Article
dx.doi.org/10.1021/ie302626x | Ind. Eng. Chem. Res. 2012, 51, 1659716609 16609

Potrebbero piacerti anche