La(1-y)Co0.4Fe0.6O3-δ perovskite oxides as catalysts for Fischer-Tropsch synthesis

Article abstract of DOI:10.1016/j.jcat.2005.07.025

A previous study of LaCo_xFe_(1-x)O₃ perovskites demonstrated the potential of such materials to produce light olefins from syngas by Fischer-Tropsch synthesis. The partial reduction of LaCo_0.4Fe_0.6O₃ generates small Co⁰ particles in interaction with a deficient perovskite. However, the stability of the perovskite limits the amount of metal to 2% in weight. To increase the amount of active phase (metal) that can be extracted from the oxide precursor by reduction, we attempted to synthesize A-site deficient perovskites. La _(1-y)Co_0.4Fe_0.6O_3-δ series prepared by a sol-gel-like route revealed to be, instead of expected A-site deficient oxides, γ-Fe₂O₃/LaCo_zFe _(1-z)O₃ composites. Magnetic nanocores of γ-Fe ₂O₃ are surrounded by a perovskite-type shell. The epitaxial growth of the perovskite phase on the cubic iron oxides drives its crystallization toward the same cubic system. The characterization of La _(1-y)Co_0.4Fe_0.6O_3-δ oxides are presented, and the structural hypotheses are discussed. The reducibility of such oxides is studied by thermoprogrammed reduction and Moessbauer experiments, and the formulae of the stable partially reduced materials are determined. These materials are composed of an equimolar Co-Fe alloy on a cubic perovskite enriched in iron with respect to that of the fresh catalyst. The γ-Fe₂O₃ disappear; one part of the Fe³⁺ cations is reduced into Fe⁰ in the alloy, and the other part of the Fe³⁺ replaces a part of the Co³⁺ cations of the perovskite, which also reduce into Co⁰ in the alloy. After partial reduction, the reactivity of these oxides in Fischer-Tropsch synthesis is evaluated. Their efficiency to produce light olefins and their stability over 300 h of test is demonstrated.

Full text of DOI:10.1016/j.jcat.2005.07.025

Journal of Catalysis 235 (2005) 279–294
www.elsevier.com/locate/jcat
La(1−y)Co0.4Fe0.6O3−δ perovskite oxides as catalysts
for Fischer–Tropsch synthesis
Laurent Bedel a, Anne-Cécile Roger a,∗, Jean-Luc Rehspringer b, Yvan Zimmermann a,
Alain Kiennemann a
a
Laboratoire des Matériaux, Surfaces et Procédés pour la Catalyse, ECPM, UMR 7515, 25 rue Becquerel, 67087 Strasbourg cedex 2, France
b
Groupe des Matériaux Inorganiques, Equipe Matériaux et Applications, IPCMS, UMR 7504, 23 rue du Loess, BP 43, 67034 Strasbourg cedex 2, France
Received 17 June 2005; revised 27 July 2005; accepted 29 July 2005
Abstract
A previous study of LaCo Fe
O perovskites demonstrated the potential of such materials to produce light olefins from syngas by Fischer–
3
x
(1−x)
0
Tropsch synthesis. The partial reduction of LaCo Fe O generates small Co particles in interaction with a deficient perovskite. However, the
0.4 0.6
3
stability of the perovskite limits the amount of metal to 2% in weight. To increase the amount of active phase (metal) that can be extracted from
the oxide precursor by reduction, we attempted to synthesize A-site deficient perovskites. La
Co Fe
O
series prepared by a sol–gel-
(1−y) 0.4 0.6 3−δ
like route revealed to be, instead of expected A-site deficient oxides, γ -Fe O /LaCo Fe
O composites. Magnetic nanocores of γ -Fe O
3 2 3
z
2
3
(1−z)
are surrounded by a perovskite-type shell. The epitaxial growth of the perovskite phase on the cubic iron oxides drives its crystallization toward
the same cubic system. The characterization of La Co Fe oxides are presented, and the structural hypotheses are discussed. The
O
(1−y) 0.4 0.6 3−δ
reducibility of such oxides is studied by thermoprogrammed reduction and Mössbauer experiments, and the formulae of the stable partially
reduced materials are determined. These materials are composed of an equimolar Co–Fe alloy on a cubic perovskite enriched in iron with respect
3+
0
3+
to that of the fresh catalyst. The γ -Fe O disappear; one part of the Fe cations is reduced into Fe in the alloy, and the other part of the Fe
2
3
3+
0
“replaces” a part of the Co cations of the perovskite, which also reduce into Co in the alloy. After partial reduction, the reactivity of these
oxides in Fischer–Tropsch synthesis is evaluated. Their efficiency to produce light olefins and their stability over 300 h of test is demonstrated.
2005 Elsevier Inc. All rights reserved.
Keywords: Perovskite; Composite; Fischer–Tropsch; Light olefins
1. Introduction
they can be partially reduced to generate metal particles within
stable cation-deficient perovskite [11,12].
Because Co and Fe are the most widely adapted metal cat-
alysts for Fischer–Tropsch synthesis, we integrated both met-
als in a La-based perovskite to obtain, after partial reduction,
efficient catalysts for light olefin formation from syngas. Ear-
lier studies on the LaCoxFe(1−x)O3 series [12] have pointed
out that the amount of extractable metal is closely related to
both the perovskite crystalline system (only orthorhombic per-
ovskites [x < 0.5] enable metal formation after partial reduc-
tion) and the calcination temperature. When calcined at 750 ◦C,
the maximum extractable metal amount reaches 2.1 wt% for
LaCo0.4Fe0.6O3. Metal formation is limited by the very high
stability of the perovskite. To improve it, the perovskite crys-
talline structure must be weakened. The challenge was to in-
troduce A-site vacancies in the ABO3 perovskite structure to
Perovskites have been of major interest over the past decades
because of their numerous applications [1,2]. For example, the
oxygen-permeating property of LaSrBFeO3 (B = Co, Ga) per-
ovskite makes it suitable as membrane [3], whereas interest in
the La1−xCaxMnO3 series is because of its magnetic proper-
ties [4]. Perovskites are also widely studied in the heteroge-
neous catalysis field, particularly for oxidation reactions, such
as total [5–8] or partial [9,10] methane oxidation. In addition,
*
Corresponding author. Fax: +33 3 90 24 27 68.
E-mail addresses: rogerac@ecpm.u-strasbg.fr (A.-C. Roger),
kiennemann@ecpm.u-strasbg.fr (A. Kiennemann).
0021-9517/$ – see front matter
doi:10.1016/j.jcat.2005.07.025
2005 Elsevier Inc. All rights reserved.

280
L. Bedel et al. / Journal of Catalysis 235 (2005) 279–294
decrease its stability, as well as increase the amount of reducible
B-cations.
The literature contains numerous reports of deficient per-
ovskites [13–18]. For instance, Stevenson et al. [13] stud-
ied (La1−xSrx)1−ε(Ga0.8Mg0.2)O3−δ materials, and Waller et
al. [14] reported the existence of a 27% A-site deficient per-
The magnetic properties were studied on a Föner VSM mag-
netometer in the 300 to 4 K temperature range and under a
maximum magnetic field of 18,000 Oe.
Materials morphology was studied using a Topcon EM-
002B transmission microscope coupled to a Kevex Deltapro
Quantum EDX spectrometer.
Thermoprogrammed reductions (TPRs) were carried out on
50 mg of catalyst, heated from 25 to 900 ◦C (15 ◦C min−1) un-
der 10% H2/He (total gas flow, 50 mL min−1). The hydrogen
consumption was followed and quantified by an external stan-
dard.
ovskite phase, (La0.6Sr0.13)(Co0.2Fe0.8)O3−δ
.
This paper presents the materials of the general formula
La(1−y)Co0.4Fe0.6O3−δ and discusses the advanced character-
ics of the structure. It also describes the reducibility of the
lanthanum-deficient perovskites and discusses their efficiency
as catalysts for Fischer–Tropsch reactions to produce light
olefins.
2.3. Catalytic tests
For the reaction of CO dissociation, 100 mg of catalyst was
2. Experimental
first reduced in situ by heating from 20 to 450 ◦C (2 ◦C min−1
)
under a 10% H2/He flow (total gas flow, 50 mL min−1). The
flow was then replaced by only He at 30 mL min−1. At 350 ◦C,
pulses of CO (0.5 mL) were admitted onto the catalyst at steady
8-min intervals. The exit gases (CO and CO2) were analyzed by
gas chromatography.
Catalytic tests under CO/H2 were performed in a fixed-bed
reactor at 1 MPa pressure. First, 300 mg of catalyst was reduced
in situ under 10% H2/He. The reduced catalyst was then heated
under N2, which was then replaced by the CO/H2 (1:1) mixture
with a GHSV of 3000 h−1. The catalytic tests were carried out
between 210 and 280 ◦C, for 50 h at each temperature. The re-
action products were analyzed on-line by gas chromatography.
2.1. Preparation of fresh catalysts
The La(1−y)Co0.4Fe0.6O3−δ oxides (for y between 0 and
0.4) were prepared by a soft chemistry method based on the
thermal decomposition of mixed La–Co–Fe propionates [19].
Starting materials [powdered Fe0, La acetate La(OOCCH3)3 ·
1.5H2O, and Co acetate Co(OOCCH3)2 · 4H2O] were sepa-
rately dissolved in boiling propionic acid at a concentration of
3 × 10−3 molL−1 in cations. Each starting salt leads to the ex-
clusive formation of the corresponding propionates (controlled
by IR of the species crystallized from the propionate solution)
[19,20]. The three boiling solutions were then mixed to gener-
ate mixed La–Co–Fe propionates in solution, and, after 30 min
of stirring, propionic acid was evaporated until a dark-brown
pasty resin was obtained. The final oxide resulted from 6 h of
calcination of the resin at either 750 or 900 ◦C (2 ◦C min−1).
3. Results and discussion
3.1. Characterization of fresh catalysts
3.1.1. Bulk analysis and specific surface areas
The La(1−y)Co0.4Fe0.6O3−δ oxides (y = 0, 0.1, 0.2, 0.3, and
0.4) are black powders. Their exact composition, reported in
Table 1, was determined by ICP. Because of greater than ex-
pected hydration of lanthanum acetate, lanthanum deficiency
was generally greater than expected (e.g., y = 0.43 instead of
0.40). Despite this difference, the samples are labeled based on
their theoretical lanthanum deficiency.
2.2. Characterization
X-Ray diffraction (XRD) was done recorded using a Sie-
mens D5000 diffractometer equipped with a Co-Kα source.
Patterns were recorded between 20, and 85◦ (2θ scale) with
0.04◦ steps and a 5-s counting time.
57Fe Mössbauer spectra were recorded using a spectrome-
ter equipped with a 50 mCi 57Co in Rh source. A bath cryostat
allowed measurements for all samples at a liquid He tempera-
ture of 4 K. The spectrometer was calibrated with reference to
a sheet of α-Fe, and isomer shift values are given relative to
this.
For the series at 750 ◦C, the specific surface area increased
with y (from 6.8 m2 g−1 for y = 0 to 10.9 m2 g−1 for y = 0.4).
For the series at 900 ◦C, the specific surface area decreased with
y (from 4.3 m2 g−1 for y = 0 to 2.8 m2 g−1 for y = 0.4). In all
cases the surfaces were very low, and the variations were not
considered significant.
Table 1
ꢀ
θ
(dimension of the A-site vacancies) and a (theoretical cubic cell parameter) calculated parameters according to hypotheses 1, 2 and 3
1
ꢀ
ꢀ
y
Exact composition
a (Å)
θ
(Å)
a (Å)
a (Å)
1
◦
◦
hypothesis 1
hypothesis 2
hypothesis 3
750 C
900 C
0
LaCo Fe
O
Fe
Fe
Fe
3.912(4)
3.904(4)
3.910(3)
3.902(3)
3.909(3)
3.896(4)
3.882(4)
3.886(3)
3.893(3)
3.900(2)
–
3.911
3.894
3.878
3.879
3.871
3.911
3.895
3.876
3.877
3.866
0.4 0.6
3
0.1
0.2
0.3
0.4
La
La
La
La
Co
O
O
O
O
5.35
3.90
3.88
3.52
0.82 0.38 0.62 3−δ
Co
0.66 0.48 0.52 3−δ
Co
0.66 0.40 0.60 3−δ
Co
Fe
0.57 0.40 0.60 3−δ

L. Bedel et al. / Journal of Catalysis 235 (2005) 279–294
281
◦
. Calcined at 750 and 900 C.
Fig. 1. Diffractograms of La
Co Fe
O
(1−y) 0.4 0.6 3−δ
3.1.2. XRD
assigned to its maximum intensity. The cubic cell parameter
“a” of the perovskite phase was calculated using a least squares
method [21]; results are given in Table 1.
Despite an average ionic radius decreasing with the lan-
thanum content, the cell parameter remained the same within
experimental limits (3.902–3.912 Å for calcination at 750 ◦C
and 3.882–3.900 Å for calcination at 900 ◦C). Furthermore, this
very small variation is not linear with the La deficiency. No ob-
vious relation between the cell parameter of the perovskite and
its stoichiometry is seen.
Three structural hypotheses can be proposed to describe
such La-deficient perovskite single phases. First, considering
the work of Chakhmouradian et al. [18] on the Th0.2510.75NbO3
system, we can infer that A-site cationic vacancies may be
envisaged. Second, B-site cationic vacancies, with fewer con-
straints, have also been reported [22] and can be considered;
third, a combination of the two assumptions is possible. For-
mulations according to each hypothesis are as follows:
The XRD diffraction patterns of samples calcined at 750 and
900 ◦C are presented in Fig. 1. For both calcination tempera-
tures, a given sample led to the same observed reflections in the
diffractogram. Only the crystallinity was affected by varying
the calcination temperature; those at 900 ◦C exhibited a better
signal on noise ratio and narrower lines than those at 750 ◦C.
For La-deficiency < y = 0.4, all reflections can be assigned to
a single cubic perovskite-type phase, whereas non-La-deficient
LaCoxFe(1−x)O3 oxides crystallized in either the orthorhombic
(when x < 0.5) or the rhombohedral system (when x ꢀ 0.5)
[12]. An additional weak reflection was detected at 2θ = 41.4◦
(d = 2.53 Å) for the y = 0.4 samples (750 and 900 ◦C cal-
cined) only. This line may be due to a trace of iron oxide,
α- or γ -Fe2O3. The nonsymmetric shape of the reflections (es-
pecially in calcination at 750 ◦C) may be due to either the Kα
radiation or a gradient of the perovskite composition. To more
easily compare the different patterns, a reflection position was
2

282
L. Bedel et al. / Journal of Catalysis 235 (2005) 279–294
Fig. 2. La Co Fe
O
experimental Mössbauer spectrum at 4 K and its deconvoluted components (A) or their sum (B).
0.7 0.4 0.6 3−δ
1. A-Site cationic vacancies. The detailed formulation should
be:
Table 2
Characteristics of the La Co Fe
nents
O
deconvoluted Mössbauer compo-
0.7 0.4 0.6 3−δ
(La(1−y)1y)(Co0.4Fe0.6)O3−δ
.
−1
−1
)
Component δ (mm s
)
∆ (mm s
H (kG) Attribution
3+
(%)
58
Sextet
Sextet
0.358
0.349
0
0
538
512
Fe (B-site) 42
2. B-Site cationic vacancies and presence of Fe in A-sites.
Compounds can be written as:
Fe O
2
3
(La(1−y)Fey)(Co0.4Fe0.6−y1y)O3−δ
.
3.1.3. Mössbauer
To precisely determine the degree and location of Fe oxida-
tion, 57Fe Mössbauer spectroscopy was performed. A typical
spectrum of the 750 ◦C calcined La0.7Co0.4Fe0.6O3−δ sample
recorded at 4 K is shown in Fig. 2.
3. No cationic vacancies and presence of Fe in A-sites. The
formulation could be:
(La(1−y)Fey/2)(Co0.4Fe0.6−y/2)O3−δ
.
The experimental spectrum was deconvoluted into two sex-
tets, the parameters of which are reported in Table 2. To fit the
experimental spectrum, external bands of the first sextet must
be broadened by considering a gradient in the hyperfine field.
Such enlargement results from variation within the Fe3+ en-
vironment from changes in either the composition gradient of
the phase or the size distribution of the grains. The mean para-
meters of this sextet were δ = 0.358 mms−1 and H = 538 kG.
This sextet was assigned to Fe3+ located in the perovskite phase
B-sites [24,25]. Despite the gradient of the hyperfine field, a
According to hypotheses 2ꢀ and 3, we can calculate a theo-
retical cubic cell parameter, a , using the method of Poix [23]
based on the invariance of the cationꢀ–oxygen distance. For cu-
bic perovskite, the cell parameter a can be estimated by the
relation
ꢀ
ꢀ
1.207aꢀ =
νiθi +
νj βj ,
i
j
where νi,j are the stoichiometric coefficients and θi and βj are
the cation–oxygen distance with 12-fold and 6-fold coordina-
tion, respectively. Fe in A-sites is supposed to be Fe2+; 12-fold
coordinated Fe3+ in oxide is not known.
second sextet was needed to achieve the fit (δ = 0.349 mms−1
,
H = 512 kG). This sextet was assigned to Fe3+ in an Fe2O3
phase [26].
Thus, neither Fe2+ nor Fe3+ in perovskite A-sites was de-
tected; however, Fe in perovskite B-sites represents only 42%
of the total Fe contained in the sample. The crystallized Fe2O3
phase present in large amounts in the material (58% of the Fe)
was then observable by Mössbauer spectroscopy, but not by
XRD (except for y = 0.4). The fact that α- and γ -Fe2O3 have
similar Mössbauer spectra makes exact identification difficult;
however, we can speculate that instead of having an A-site defi-
cient perovskite-type phase, the excess B cations were rejected
via Fe into the observed Fe2O3 phase.
It is not possible to calculate a theoretical cell parameter in
accordance with the first hypothesis, because the dimension of
the A-site vacancies θ1 is unknown. However, we can estimate
this parameter (θ1) from the experimental cell parameter. For
hypothesis 1 to be true, the calculated θ1 must be constant for
all samples. Calculated parameters aꢀ and θ1 are reported in
Table 1.
The hypothetical A-site vacancies exhibited size fluctuations
from 5.35 to 3.52 Å; this clearly is not feasible. All of the
calculated aꢀ parameters are closely related to the La content
of the sample and decrease linearly with it, whereas we ob-
serve that the experimental cell parameter was almost constant
along the series and obviously independent of the La content.
None of the above assumptions are acceptable to describe the
La(1−y)Co0.4Fe0.6O3−δ materials. Further characterizations are
therefore required.
3.1.4. Magnetic measurements
The magnetic behaviors of α- and γ -Fe2O3 differ. α-Fe2O3
is an antiferromagnet below the Morin transition of 243 K and
a weak ferromagnet above this transition (maximum magneti-
zation of 0.4 emu g−1) [27]. γ -Fe2O3 is ferrimagnetic with a

L. Bedel et al. / Journal of Catalysis 235 (2005) 279–294
283
Table 3
Developed formulation of La
Co Fe
O
(1−y) 0.4 0.6 3−δ
ꢀ
◦
y
Raw formula
(ICP determination)
LaCo Fe
Developed formula
Calculated a
(Å)
a
(750 C)
exp
(Å)
0
O
[La][Co
[La][Co
[La][Co
[La][Co
[La][Co
Fe
]O
3
3.911
3.910
3.906
3.908
3.906
3.912(4)
3.904(4)
3.910(3)
3.902(3)
3.909(3)
0.40 0.60
3
0.40 0.60
0.1
0.2
0.3
0.4
La
La
La
La
Co
Fe
Fe
Fe
Fe
O
O
O
O
Fe
]O + 0.11Fe O
0.82 0.38 0.62 3−δ
0.46 0.54
3
2
3
3
3
3
Co
Fe
]O + 0.26Fe O
0.66 0.48 0.52 3−δ
0.73 0.27
3
2
Co
Fe
]O + 0.26Fe O
0.66 0.40 0.60 3−δ
0.61 0.39
3
2
Co
Fe
]O + 0.38Fe O
0.57 0.40 0.60 3−δ
0.70 0.30
3
2
small particles), then these particles should represent at least
6.5/59.2 = 11 wt% of the sample.
If we assume that the perovskite phase has no cation defi-
ciency and that the excess B cations form the γ -Fe2O3 phase,
then the formulation of the La(1−y)Co0.4Fe0.6O3−δ oxides can
be broken down to y Fe2O3 + La(1−y)Co0.4Fe(0.6−y)O3−δ. To
2
recover an ABO3 perovskite, all of the stoichiometric coeffi-
cients must be divided by (1 − y). The formulation of these
materials is summarized in Table 3.
The exact composition of the y = 0.3 sample is La0.66Co0.40
-
Fe0.60O3−δ. Taking into account no cation vacancy and pres-
ence of an iron oxide phase, this can be broken down into its
components, LaCo0.61Fe0.39O3 + 0.26Fe2O3. In this composite
material, iron oxide represents 10.6 wt% of the sample. This
formulation is then in close agreement with the 11 wt% of
γ -Fe2O3 estimated from magnetization data (Fig. 3). In this
formulation, the Fe in Fe2O3 represents 57% of the total Fe of
the sample, corresponding to the phase repartition calculated
from the Mössbauer results (58%, see Table 2).
Fig. 3. Virgin magnetization (after ZFC) and hysteresis effect at 4 K for
◦
La Co Fe
O
calcined at 750 C.
0.7 0.4 0.6 3−δ
maximum magnetization of 74 emu g−1 but only 59.2 emu g−1
when the particles are small [28]. The magnetic behavior
of the La(1−y)Co0.4Fe0.6O3−δ series was studied to obtain
more information about the structure. As shown in Fig. 3 for
La0.7Co0.4Fe0.6O3−δ calcined at 750 ◦C, a hysteresis cycle was
obtained when varying the magnetic field between −18 and
+18 kOe. The existence of a strong coercive field (8500 Oe) is
consistent with a hard magnetic material.
Thus, like the experimental cell parameters aꢀ(see Table 1),
the variation of the calculated cell parameters a along the se-
ries was very weak and independent of the La deficiency. This
also confirms the assumptions that we made to describe the
La(1−y)Co0.4Fe0.6O3−δ materials.
The observed magnetization of the sample of 7.8 emu g−1
cannot be assigned only to the perovskite contribution. Indeed,
previous work on pure LaCoxFe1−xO3 indicated that such ox-
ides present a maximal magnetization of about 1.3 emu g−1
[29]. A 6.5 emu g−1 contribution of Fe2O3 then must be consid-
In our systems, three structures drawing these two-phase ox-
ides can be envisaged, as schematized in Scheme 1. A core-
shell phenomenon can be imagined with either a unique
γ -Fe2O3 core (Scheme 1A) or multiple nanosized-cores dis-
persed in the perovskite phase (Scheme 1B). The last possibility
(Scheme 1C) is that the two phases coexist as individual parti-
cles, but with a γ -Fe2O3 size too small to diffract X-rays.
ered to explain the experimental magnetization of 7.8 emu g−1
.
Such magnetization (6.5 emu g−1) is higher than the theoret-
ical maximum for α-Fe2O3 (0.4 emu g−1 for pure α-Fe2O3)
The iron oxide phase cannot be considered α-Fe2O3, but if
we assign the experimental magnetization to γ -small Fe2O3
small particles (59.2 emu g−1 theoretical magnetization when
3.1.5. TEM
Samples were dispersed in a resin, then cut and polished.
The cut of 750 ◦C calcined La0.7Co0.4Fe0.6O3−δ material ob-
Scheme 1.

284
L. Bedel et al. / Journal of Catalysis 235 (2005) 279–294
◦
Fig. 4. TEM pictures of cut of La Co Fe
O
(750 C): (a) general view (×3000); (b) zone of large grains (×20,000); (c) of zone of small grains (×29,000).
0.7 0.4 0.6 3−δ
Average molar distribution (%) of cations in the whole sample (a), the large grains (b) and the small grains (c).
served by TEM is shown in Fig. 4. The existence of two types
of grains is clearly shoen on low magnification (Fig. 4a). Zones
of small grains (20–50 nm; Fig. 4c) surround zones of larger
grains (120–190 nm; Fig. 4b). Analyzed by EDX, the ele-
mental composition over a large area covering both the small-
and large-grain regions is in agreement with that determined
by ICP. If separately analyzed, small-grain regions exhibit a
higher Fe content (53.4 mol%) than the large grain regions
(17.9 mol%). Because the small grains are not pure γ -Fe2O3
particles and have too large a diameter to not be detected by
XRD, Scheme 1C must be discarded. The different perovskite
compositions in the small and large grains (LaCo0.50Fe0.50O3 +
0.61Fe2O3 and LaCo0.67Fe0.33O3 + 0.015Fe2O3, respectively)
explains the nonsymmetric shape of the reflections observed on
the XRD patterns and is consistent with the composition distri-
bution taking the Mössbauer spectra into account.
The La content from one edge to the other of both small and
large grains was studied by analyzing consecutive 8-nm-wide
areas by EDX; results are plotted in Fig. 5. Whatever the size
of the grain, 190 or 26 nm, the La content remained almost
constant all along the grain. According to the unique central
γ -Fe2O3 core hypothesis, the La content should decrease when
analyzing the core; thus, this hypothesis also must be discarded.
The only structure that can describe our compounds is that in
Scheme 1B. Materials are then made of nanocores dispersed in
a La–Co–Fe perovskite matrix.

L. Bedel et al. / Journal of Catalysis 235 (2005) 279–294
285
◦
(750 C) grains.
Fig. 5. Distribution of the La-content along the La Co Fe
O
0.7 0.4 0.6 3−δ
Because of the rejection of the excess B-cations via Fe into
the γ -Fe2O3 phase, the perovskite phase average composition
is Co-rich, up to LaCo0.73Fe0.27O3 for y = 0.2. In a previous
study [12] we showed that when the La–Co–Fe perovskites con-
tain at least as much Co as Fe, they crystallize in the rhombo-
hedral system. The observed cubic symmetry of the perovskite
phase is probably caused by an epitaxial growth on the cubic
γ -Fe2O3 cores. This interpretation could also explain the higher
Fe content in the smaller grains than in the larger grains seen on
TEM.
0.90 to 4.85 mL for the series at 900 ◦C. Hydrogen consump-
tion is practically independent on calcination temperature. The
amounts of hydrogen consumed in the second zone (volumes
V2) do not vary much with varying y. The slight variations are
due to the difficulty in achieving total reduction at 900 ◦C.
The perovskite phases of the samples y = 0.2 and y = 0.4
are noticeably identical (see Table 4). However, V1 is largely
higher for y = 0.4 (4.79 mL) than for y = 0.2 (3.22 mL). These
two catalysts differ only in terms of their iron oxide content
(0.26 Fe2O3 per perovskite for y = 0.2 and 0.38 for y = 0.4).
A part of the hydrogen consumed in the first reduction zone
can then be attributed to the reduction of γ -Fe2O3. The same
conclusion can be drawn when comparing y = 0 and 0.1.
A partial reduction of the materials is then possible for the
two series of perovskites. In a previous work [12] we showed
that the partial reduction at 450 ◦C of LaCo0.4Fe0.6O3 generates
Co0 particles on partially reduced perovskite. Here the pres-
ence of an additional phase of γ -Fe2O3, which seems to be
reducible, would modify the nature of the metal phase gener-
ated by partial reduction, as can be seen by XRD, magnetic
measurements and Mössbauer.
3.2. Reducibility and reduced catalysts
3.2.1. TPR
Because the metal phase is the active phase in Fischer–
Tropsch synthesis, attention was given to reducibility. The re-
duction was first followed by TPR. The hydrogen consumption
profiles are presented in Fig. 6 for the series of perovskites cal-
cined at 750 and 900 ◦C.
The two series of catalysts present two zones of reduction,
the maxima of which appear around 450 and 900 ◦C for calci-
nation at 750 ◦C and a little higher for calcination at 900 ◦C.
The lanthanum deficiency does not strongly influence the re-
duction temperature, but it does affect hydrogen consumption
in the first zone of reduction; the volume of hydrogen consumed
at low temperature increases linearly with y.
The other effect of the higher calcination temperature is
to increase the separation between the two reduction zones.
Whereas for the series at 750 ◦C, separation involves only a
short return to the base line around 500 ◦C, the stability without
any hydrogen consumption lasts between 50 ◦C for y = 0 and
120 ◦C for 0.1 ꢁ y ꢁ 0.4 for the series at 900 ◦C. Consequently,
all of the partially reduced perovskites are stable under reduc-
ing conditions around 500 ◦C. This stability of partially reduced
oxides would allow their use as stable catalysts for Fischer–
Tropsch synthesis.
The volumes of hydrogen needed for the reduction of 50 mg
of fresh oxide at the two calcination temperatures are reported
in Table 4. The amount of hydrogen consumed in the first
zone (volumes V1) increases with the lanthanum deficiency y:
from 1.02 to 4.79 mL for the series at 750 ◦C, and from
3.2.2. XRD
The diffractograms of the partially reduced oxides (1.5 h at
450 ◦C under 10% H2/He) are presented in Fig. 7 for the two se-
ries. The crystalline structure of the perovskite phases was pre-
served whatever the calcination temperature and the lanthanum
deficiency. The peak of γ -Fe2O3 detected on fresh samples
for y = 0.4 disappeared to the benefit of a new peak around
2θ = 53◦, attributed to a bcc Co–Fe alloy [30]. The lattice pa-
rameters of the two phases of the partially reduces samples are
given in Table 5. The cubic lattice parameter of the perovskite
phase does not vary much with y; from 3.908(3) to 3.920(3) Å
for the series at 750 ◦C, and from 3.899(3) to 3.924(2) Å for the
series at 900 ◦C. The cubic lattice parameter of the metal phase
lies between 2.856(7) and 2.849(7) Å for the two series, cor-
responding to alloy composition between (Co0.50Fe0.50)0 and
(Co0.60Fe0.40)0 [31]. The alloy crystallite medium size was es-
timated by the Debye–Scherrer formula as 10 nm for the series
at 750 ◦C and 28 nm for the series at 900 ◦C.

286
L. Bedel et al. / Journal of Catalysis 235 (2005) 279–294
◦
. Calcined at 750 and 900 C.
Fig. 6. TPR profiles of La
Co Fe
O
(1−y) 0.4 0.6 3−δ
Table 4
netization increased with the lanthanum deficiency, from 3.3
to 25.3 emu g−1 for the series at 750 ◦C and from 1.3 to
33.4 emu g−1 for the series at 900 ◦C. Saturation magnetiza-
tions at ambient temperature for various iron-based alloys de-
pending on their composition have been given previously [32].
The content in alloy (whose composition is determined from
the lattice parameter) can then be deduced from the MS values.
For the two samples y = 0.1, the lattice parameter could not be
determined by XRD, and so the alloy composition was taken to
be equal to that of the sample y = 0.2 in the same series.
Volume of hydrogen consumed during the two zones of reduction. V , first
reduction around 450 C; V , second reduction around 900 C
1
◦
◦
2
◦
◦
900 C
y
Formula
750 C
V
(mL) V (mL) V (mL) V (mL)
1
2
1
2
0
LaCo
Fe
O
1.02
6.17
6.07
5.93
6.01
5.31
0.90
2.16
3.01
3.75
4.85
6.16
6.14
5.58
5.36
5.46
0,40 0,60
3
0.1 LaCo
0.2 LaCo
0.3 LaCo
0.4 LaCo
Fe
O + 0.11Fe O 2.21
0.46 0.54
3
2 3
Fe
O + 0.26Fe O 3.22
0.73 0.27
3
2 3
Fe
O + 0.26Fe O 3.83
0.61 0.39
3
2 3
Fe
O + 0.38Fe O 4.79
0.70 0.30
3
2 3
As reported previously, for stoichiometric perovskites (y =
0), the metal phase generated by reduction is only cobalt [12].
In contrast to the partially reduced stoichiometric perovskites,
in which the amount of metal (Co0) decreases with the initial
calcination temperature (by 2.1% for y = 0 at 750 ◦C and 0.8%
for y = 0 at 900 ◦C), the amount of metal formed after partial
reduction of deficient perovskites increased with the initial cal-
3.2.3. Magnetic measurements
In view of quantifying the metal generated by partial re-
duction, magnetization cycles were performed on partially re-
duced samples. The cycles for the two series of partially re-
duced catalysts are shown in Fig. 8, and the corresponding
saturation magnetizations are given in Table 5. Saturation mag-

L. Bedel et al. / Journal of Catalysis 235 (2005) 279–294
287
◦
◦
Fig. 7. Diffractograms of La
Co Fe
O
. Calcined at 750 and 900 C, after partial reduction at 450 C.
(1−y) 0.4 0.6 3−δ
Table 5
Lattice parameters of oxide and metal phases after partial reduction of the two La
Amount of metal (wt%) and formula of the metal phase
Co Fe
(1−y) 0.4 0.6
O series. Saturation magnetization (M ) after partial reduction.
3 S
◦
◦
y
750 C
900 C
−1
−1
)
a(oxide)
a(metal) (Å)
M
(emu g
)
a(oxide)
a(metal) (Å)
M
(emu g
S
S
(Å)
[composition]
[metal amount]
(Å)
[composition]
[metal amount]
0
3.920(3)
3.921(3)
3.914(3)
3.908(3)
3.914(3)
–
[Co]
–
[Co
2.856(7)
[Co Fe
2.856(7)
[Co Fe
2.854(7)
[Co Fe
3.3
[2.1 wt%]
11.0
[4.8 wt%]
15.2
[6.6 wt%]
21.8
[9.5 wt%]
25.3
[10.9 wt%]
3.899(3)
3.921(3)
3.919(3)
3.911(3)
3.924(2)
–
[Co]
–
1.3
[0.8 wt%]
9.5
[4.0 wt%]
15.9
[6.8 wt%]
23.4
[10.1 wt%]
33.4
[14.1 wt%]
0.1
0.2
0.3
0.4
Fe
]
]
]
]
0.50 0.50
2.853(7)
[Co
2.854(7)
[Co Fe
2.849(7)
[Co Fe
Fe
]
]
]
0.50 0.50
0.55 0.45
0.50 0.50
0.53 0.47
0.53 0.47
0.60 0.40

288
L. Bedel et al. / Journal of Catalysis 235 (2005) 279–294
◦
◦
Fig. 8. Magnetization cycles of La
Co Fe
O
. Calcined at 750 and 900 C, after partial reduction at 450 C.
(1−y) 0.4 0.6 3−δ
cination temperature (10.9% for y = 0.4 at 750 ◦C, compared
with 14.1% at 900 ◦C).
The amount of metal generated by partial reduction in-
creased significantly with increasing lanthanum deficiency. The
maximum was obtained for the sample y = 0.4 calcined at
900 ◦C.
tics of which are given in Table 6. The isomeric shift of the
first sextet (δ = 0.345 mms−1) and the value of the hyperfine
field (H = 537 kG) allowed its attribution to Fe3+ in the B
sites of the perovskite phase. The sextet of Fe3+ in γ -Fe2O3
encountered in the fresh oxide (see Table 2) disappeared to
the benefit of another sextet, whose isomeric shift close to 0
(δ = −0.003 mms−1) is characteristic of metal iron. The value
of the hyperfine field (H = 339 kG) is in accordance with an
alloy phase of composition Fe0.5Co0.5 (342 kG) [33]. Partial
reduction of the perovskite involved the total reduction of the
γ -Fe2O3 cores into metal iron, which combined with the metal
cobalt coming from perovskite reduction to form a Co0.5Fe0.5
alloy.
Note that for the series at 750 ◦C, the saturation magneti-
zation of reduced samples was difficult to reach (see Fig. 8),
accounting for a superparamagnetic behavior, in accordance
with the small particle size calculated from XRD (10 nm). This
was not the case for the series at 900 ◦C.
3.2.4. Mössbauer experiments
Combining the various characterization results allowed us
to determine the formulae of the partially reduced perovskites.
These are presented in Table 7.
The amount of metal generated by partial reduction in-
creased with increasing La deficiency y and with increasing
To understand the origin of the Co–Fe alloy, Mössbauer ex-
periments were carried out on partially reduced deficient per-
ovskites. The Mössbauer spectrum of y = 0.3 (calcination at
750 ◦C) after partial reduction is shown in Fig. 9. The exper-
imental spectrum is the sum of two sextets, the characteris-

L. Bedel et al. / Journal of Catalysis 235 (2005) 279–294
289
Fig. 9. Partially reduced La Co Fe
O
experimental Mössbauer spectrum at 4 K and its deconvoluted components (A) or their sum (B).
0.7 0.4 0.6 3−δ
Table 6
and Mössbauer would then be located at the surface of the cat-
alyst grains.
Characteristics of the La Co Fe
luted Mössbauer components
O
(after partial reduction) deconvo-
0.7 0.4 0.6 3−δ
−1
−1
)
Component δ (mm s
)
∆ (mm s
H (kG) Attribution
3+
(%)
26
3.3. Catalytic results
Sextet
Sextet
0.345
−0.003
0
0
537
339
Fe (B-site) 74
0
Fe
Before testing the reactivity of the partially reduced systems
in Fischer–Tropsch synthesis, we evaluated their ability to initi-
ate the reaction by studying the CO dissociation reaction under
atmospheric pressure. The CO dissociation in to Csurface and
Osurface is considered the limiting step of the Fischer–Tropsch
reaction [34].
initial calcination temperature of the oxide, except for y = 0
and y = 0.1. For the system y = 0.2, the γ -Fe2O3 phase was
difficult to reduce, possibly because of its higher Co content
compared with other systems. The proportion of iron in the per-
ovskite was increased after reduction.
3.3.1. CO dissociation
3.2.5. TEM
CO conversion into CO2 and Csurface for the partially re-
duced series of perovskites calcined at 750 and 900 ◦C is shown
in Fig. 11. Among the catalysts of a single series, the activity
in CO dissociation increased with increasing lanthanum defi-
ciency, due to the increased metal generated by reduction. For
the series at 750 ◦C, CO conversion increased almost linearly
from 12% for y = 0 (2.1 wt% of metal) to 19% for y = 0.4
(10.9 wt% of metal).
Comparing the two series, it appears that despite their higher
metal content, the catalysts of the series at 900 ◦C had weaker
activity in CO conversion than those of the series at 750 ◦C.
For example, for y = 0.4 (14.1 wt% of metal), CO conversion
was only 12%. In these systems, the larger metallic particle size
(28 nm compared to 10 nm for the series at 750 ◦C) led to a
decrease in surface/volume ratio, and the higher metal content
did not have a beneficial effect on catalytic activity.
The cut of 750 ◦C calcined La0.7Co0.4Fe0.6O3−δ material af-
ter partial reduction observed by TEM is shown in Fig. 10. As in
the fresh catalyst, the grains are heterogeneous in size and com-
position. Mean grain size was the same in the reduced sample
and the fresh sample, for both large (100–200 nm) and small
(20–50 nm) grains. The morphology was not modified by re-
duction. EDX analysis found no grains free of lanthanum and
no isolated particles of alloy. Fig. 10 shows the mean molar
composition in cations in the whole sample, as well as in the
small and large grains. The grain composition of the reduced
sample was similar to that determined before reduction, small
grains are enriched in iron compared with large grains. The lan-
thanum content profile along the grains (Fig. 10) reveals that it
was always higher in the center of the grain than at the edge.
The metallic phase detected by XRD, magnetic measurements,
Table 7
◦
Developed formulation of partially reduced La
Co Fe
O
oxides calcined at 750 and 900 C
(1−y) 0.4 0.6 3−δ
◦
◦
Calcination at 900 C and partial reduction
y
Calcination at 750 C and partial reduction
2+
3+
2+
3+
0
0
0
Co
/LaCo
Co
Fe
O
Co
/LaCo
Co Fe
O
0.60 2.80
0.60 2.78
0.085
0.19 0.13
0
0.03
0.31 0.06
0
2+
2+
0.1
0.2
0.3
0.4
(Co
(Co
(Co
(Co
Fe
)
/LaCo
/LaCo
/LaCo
/LaCo
Fe
Fe
Fe
Fe
O
(Co
(Co
(Co
(Co
Fe
)
)
)
)
/LaCo
/LaCo
/LaCo
/LaCo
Fe
Fe
Fe
Fe
O
+ 0.02 FeO
+ 0.09FeO
x
0.50 0.50
0.65 2.82
0.55 0.45
0.64 2.82
0.22
0
0.35
2+
0.18
0
0.36
2+
Fe
)
)
)
O
+ 0.09FeO
+ 0.02FeO
+ 0.09FeO
Fe
O
x
x
x
x
0.50 0.50
0.44 2.72
0.55 0.45
0.46 2.73
0.33
0
0.56
2+
0.34
0
0.54
2+
Fe
O
Fe
O
+ 0.01FeO
x
0.50 0.50
0.63 2.82
0.53 0.47
0.66 2.83
0.47
0
0.37
2+
0.50
0
0.75
0.34
2+
Fe
O
Fe
O
0.50 0.50
0.61 2.80
0.60 0.40
0.75 2.84
0.58
0.39
0.25

290
L. Bedel et al. / Journal of Catalysis 235 (2005) 279–294
◦
Fig. 10. TEM pictures of cut of La Co Fe
Distribution of the La-content along the grains (bottom).
O
(750 C) after partial reduction (top). Average molar distribution (%) of cations in the sample (middle).
0.7 0.4 0.6 3−δ
3.3.2. Fischer–Tropsch synthesis
After the initial evaluation of the catalytic activity by CO
dissociation at atmospheric pressure, the catalysts were tested
in real Fischer–Tropsch synthesis conditions under 1 MPa pres-
sure (see Section 2). The CO conversion obtained for partially
reduced catalysts of the series at 750 ◦C, as well as the most ac-
tive catalyst of the second series (y = 0.4 at 900 ◦C), are shown
in Fig. 12. The increased CO conversion with y is in accordance
with the results obtained by CO dissociation. A temperature of
230 ◦C was sufficient to obtain a CO conversion of 4.5% with
the catalyst La0.6Co0.4Fe0.6O3−δ (y = 0.4), whereas the system
in which y = 0 required an increase in reaction temperature to
280 ◦C to reach a conversion of the same order (3.7%). At the
same La deficiency (y = 0.4), as foreseen by CO dissociation,
Fig. 11. CO conversion (%) versus lanthanum deficiency y for the two series of
catalysts (calcined at 750 and 900 C), reaction of CO dissociation.
◦

L. Bedel et al. / Journal of Catalysis 235 (2005) 279–294
291
lar selectivity into oxygenated products (mainly methanol and
ethanol) ranged from 6.4% for y = 0 and 12.1% for y = 0.4.
More than a catalyst effect, the formation of such oxygenated
products seems to be favored by lower reaction temperatures.
In comparison with the literature data on FeCo alloys in
Fischer–Tropsch, our catalytic systems appeared to produce lit-
tle CO2. Previous studies of Co–Fe alloys deposited on zeolites
at a 3% conversion of CO found variable molar selectivity to
CO2 depending on the support, 41% with HZMS-5 and 51%
with HY zeolite [35], whereas at the same order of conversion,
the CO2 selectivity of our systems was only around 20%.
+
In the hydrocarbon fraction, the ratio [C2–C4]/C5 was in-
dependent of y and equals 2.4 (weight distribution). For lan-
thanum deficiency up to y = 0.3, methane was the main hy-
drocarbon fraction (accounting for 47–52% of the total weight
of the hydrocarbon fraction). For the most deficient catalyst
(y = 0.4), decreasing the reaction temperature (to 230 ◦C) in-
duced a significant decrease in methane formation, to only
32% of the total weight. For this catalyst, at CO conversion of
5%, the main fraction was the C2–C4 fraction (48% of the to-
tal weight). The decrease in reaction temperature changed the
olefin proportion in the C2–C4 fraction. Olefins were clearly
favored for y = 0.4 (230 ◦C), because the olefin/paraffin ratio
reached the value of 3 (75% of olefins in this fraction). This
is in accordance with a decrease in the secondary reactions of
olefin readsorption and their consecutive hydrogenation at low
reaction temperature [36,37].
Fig. 12. CO conversion (%) vs. Fischer–Tropsch reaction temperature for the
catalysts of the series at 750 C and y = 0.4 at 900 C.
◦
◦
The selectivity into olefins in the light fraction is much
higher than that reported by Chen et al. [38] and Ishihara et
al. [39] on FeCo3-based catalysts. Chen et al. reported [37] an
olefin/paraffin ratio of 0.2–0.5 (depending on the reducing pro-
cedure) over FeCo3/carbon catalyst, whereas Ishihara et al. [38]
reported olefin/paraffin ratios in the C2–C5 fraction of 1.5 over
TiO2 and 0.6 over SiO2. The FeCo-based catalysts studied by
Butt et al. [35] were also poorly selective into light olefins
(olefin/paraffin ratios in the C2–C3 fraction between 0.3 and
0.7).
To summarize this comparative study of the catalysts, it ap-
pears that the amount of metal phase generated by controlled re-
duction did not significantly affect the selectivity of the reaction
but did direct Fischer–Tropsch activity. La0.6Co0.4Fe0.6O3−δ
exhibited high CO conversion at moderate reaction tempera-
tures (21% conversion at 255 ◦C), with such temperatures al-
lowing high selectivity toward light olefins.
The aim of the present study—to increase the metal amount
to improve the catalytic performance of the La–Co–Fe perovs-
kites—thus was largely achieved. With the previous non-La-
deficient perovskite [12] (which corresponds to y = 0 in the
present work), Fischer–Tropsch activity was much lower (2%
of CO conversion at 270 ◦C).
Fig. 13. Fischer–Tropsch product distribution at CO isoconversion of 5% for
the catalysts of the series at 750 C.
◦
the activity of the catalyst calcined at 900 ◦C was lower than
that of the catalyst calcined at 750 ◦C.
The product distribution obtained at CO isoconversion of al-
most 5% is shown in Fig. 13. At isoconversion, the product dis-
tribution did not vary much among the series. Molar selectivity
to CO2 varied between 26% for y = 0.2 and y = 0.3 and 19%
for y = 0.4. Hydrocarbon fraction was always the major part
of the products (around 67–69% of molar selectivity). The mo-
3.3.3. Long-term activity
The perovskite La0.7Co0.4Fe0.6O3−δ of the series at 750 ◦C
corresponding to a La deficiency y = 0.3 was extensively
characterized before catalytic tests and after reducing treat-
ments. Long-term Fischer–Tropsch activity studies were then
performed on that catalyst, under the same experimental condi-

292
L. Bedel et al. / Journal of Catalysis 235 (2005) 279–294
tions as before, at 250 ◦C, for 300 h. The results, presented in
Fig. 14, show that CO conversion increased in the first 75 h of
catalytic testing from 4.5 to almost 7%, and then remained quite
constant over the next 225 h. No deactivation was observed.
During the test, the product distribution remained almost un-
changed. As seen for the CO conversion, an initial period of
75 h was necessary to reach the steady state. Hydrocarbons rep-
resented 70% (in mol) of the products; the molar selectivity into
CO2 and oxygenated products stabilized at around 25 and 5%,
respectively. Among the hydrocarbons, the C2–C4 fraction rep-
resented 46% of the total weight, 41% for CH4 and 13% for
heavier hydrocarbons. The olefin/paraffin ratio in the C2–C4
fraction decreased initially from 3 to 2.3, and remained con-
stant at this value.
3.4. Characterization after testing
Fig. 15 presents diffractograms of the La0.7Co0.4Fe0.6O3−δ
catalyst after partial reduction at 450 ◦C and compares them
with those obtained after CO dissociation reaction and after
Fischer–Tropsch synthesis. As explained earlier, after partial re-
duction, the perovskite crystalline structure was preserved, and
the reduction of cobalt cations in the perovskite along with the
reduction of the γ -Fe2O3 cores gave rise to a (Co–Fe)0 alloy
detected by XRD. After CO dissociation, the perovskite struc-
ture was still preserved, and the alloy was still detectable by
XRD. However, after Fischer–Tropsch synthesis, although the
perovskite phase was not modified, the alloy main diffraction
peak disappeared. No carbide phase was detected, neither χ-
Fe5C2 (thought to be the active phase by many authors [40–44])
nor cobalt carbide Co2C, unstable at ambient temperature but
detected on bimetallic Co–Fe Fischer–Tropsch catalysts after
testing [45,46]. One explanation for this finding could be the
redispersion of metal aggregates into smaller particles (metal
and/or carbide) that are not detectable by XRD. This assump-
tion could not have been confirmed by TEM observation of
the catalyst after Fischer–Tropsch reaction, because no isolated
particles of Co–Fe (without lanthanum) were detected by TEM.
Fig. 14. Fischer–Tropsch product distribution, CO conversion and O/P fraction,
◦ ◦
(750 C) at 250 C.
vs. time on stream for La Co Fe
O
0.7 0.4 0.6 3−δ
◦
(750 C): (a) after partial reduction, (b) after CO dissociation, (c) after SFT.
Fig. 15. Diffractograms of La Co Fe
O
0.7 0.4 0.6 3−δ

L. Bedel et al. / Journal of Catalysis 235 (2005) 279–294
293
4. Conclusions
higher than that of the best catalyst of the previous series (y = 0
in the present work, CO conversion of 2% at 270 ◦C).
La(1−y)Co0.4Fe0.6O3−δ materials prepared by the thermal
decomposition of mixed La–Co–Fe propionates did not lead
to La-deficient perovskites, although single perovskite phases
were observed by XRD up to a La deficiency of 0.3. Mössbauer
spectroscopy revealed the presence of an Fe2O3 phase in addi-
tion to the perovskite. Magnetic measurements were in favor of
the iron oxide phase γ -Fe2O3 instead of α-Fe2O3. According to
both micrographs and EDX analysis, iron oxide particles were
too small to diffract X-rays. They were surrounded by a non-
cation-deficient perovskite phase. La(1−y)Co0.4Fe0.6O3−δ ma-
terials thus may be considered nano-composites. The very small
size of some γ -Fe2O3 particles made the saturation of their
magnetization difficult to reach. However, these composites al-
ready exhibited very strong coercive fields (around 8500 Oe at
4 K). Even when Co-rich (e.g., LaCo0.73Fe0.27O3 for y = 0.2),
the perovskite-type phase crystallized in the cubic system. This
unexpected feature of LaCoxFe(1−x)O3 perovskite series being
either orthorhombic for x < 0.5 or rhombohedral for x ꢀ 0.5
can be explained by an epitaxial growth of the perovskite phase
over nanosize cubic γ -Fe2O3 cores.
References
[1] L.J. Tejuca, J.L.G. Fierro, Properties and Applications of Perovskite Type
Oxides, Dekker, New York, 1993.
[2] M.A. Peña, J.L.G. Fierro, Chem. Rev. 101 (2001) 1981.
[3] S. Lee, K.S. Lee, S.K. Woo, J.W. Kim, T. Ishihara, D.K. Kim, Solid State
Ionics 158 (3–4) (2003) 287.
[4] B.T. Cong, P.N. Anh Huy, N.H. Long, J. Magn. Magn. Mater. 262 (3)
(2003) 437.
[5] H. Arai, T. Yamada, K. Eguchi, T. Seiyama, Appl. Catal. 26 (1986) 265.
[6] T. Nitadori, M. Misono, J. Catal. 93 (1985) 93.
[7] K.S. Chan, J. Ma, S. Jaenicke, G.K. Chuah, Appl. Catal. A 107 (1994)
337.
[8] R. Doshi, C.B. Alcock, J.J. Carberry, Catal. Lett. 18 (1993).
[9] C. Petit, A. Kiennemann, P. Chaumette, O. Clause, French Patent 92/
11638 (1992).
[10] J.O. Pentuchi, M.A. Ulla, J.A. Marcos, E.A. Lombardo, J. Catal. 70 (1981)
1787.
[11] L. Wachowski, S. Zielinski, A. Burewicz, Acta Chim. Sci. Hung. 106
(1981) 217.
[12] L. Bedel, A.C. Roger, C. Estournès, A. Kiennemann, Catal. Today 85
(2003) 207.
The study of the reducibility of these composite oxides
showed that the γ -Fe2O3 nanocores were reducible at 450 ◦C
under hydrogen. This partial reduction gives rise to a metal
phase, the composition of which depends on the initial calcina-
tion temperature of the material (i.e., Co0.5Fe0.5 for the series
at 750 ◦C and Co0.6Fe0.4 for y = 0.4 at 900 ◦C). The cobalt
cations extracted from the cubic perovskite by reduction are re-
placed by iron cations of the Fe2O3 phase, and the oxide struc-
ture then is not destroyed by reduction. The amount of metal
alloy was directly related to the initial lanthanum deficiency
(i.e., increases with increasing y) and increased with increas-
ing initial calcination temperature of the oxides (10.9 wt% for
y = 0.4 at 750 ◦C and 14.1 wt% for y = 0.4 at 900 ◦C). The
increase in calcination temperature came with an increase in
metal particle size (10 nm for y = 0.4 at 750 ◦C and 28 nm for
y = 0.4 at 900 ◦C).
[13] J.W. Stevenson, T.R. Armstrong, L.R. Pederson, J. Li, C.A. Lewinshon,
S. Baskaran, Solid State Ionics 113–115 (1998) 571.
[14] D. Waller, J.A. Lane, J.A. Kilner, B.C.H. Steele, Solid State Ionics 86–88
(1996) 767.
[15] F.W. Poulsen, Solid State Ionics 129 (2000) 145.
[16] G.Ch. Kostogloudis, Ch. Ftikos, Solid State Ionics 126 (1999) 143.
[17] V.V. Kharton, A.P. Viskup, E.N. Naumovich, A.A. Tonoyan, O.P. Reut,
Mater. Res. Bull. 33 (7) (1998) 1087.
[18] A.R. Chakhmouradian, R.H. Mitchell, P.C. Burns, J. Alloys Compd. 307
(2000) 149.
[19] A.C. Roger, C. Petit, A. Kiennemann, J. Catal. 167 (1997) 447.
[20] H. Provendier, C. Petit, J.L. Schmitt, A. Kiennemann, J. Mater. Sci. 34
(1999) 4121.
[21] M. Evain, U-Fit V 1.3 Institut des Matériaux de Nantes, 1992.
[22] G. Dezanneau, A. Sin, H. Roussel, M. Audier, H. Vincent, J. Solid State
Chem. 173 (1) (2003) 216.
[23] P. Poix, Bull. Soc. Ceram. 72 (1966) 1.
[24] V.C. Belessi, T.V. Bakas, C.N. Costa, A.M. Efstathiou, P.J. Pomonis, Appl.
Catal. B 28 (2000) 13.
[25] V.C. Belessi, P.N. Trikalitis, A.K. Lavados, T.V. Bakas, P.J. Pomonis,
Appl. Catal. A 177 (1999) 53.
[26] J.G. Stevens, A.M. Khasanov, J.W. Miller, H. Pollak, Z. Li (Eds.), Möss-
bauer Mineral Handbook, Mössbauer Effect Data Center, 1998.
[27] J. Kaczér, T. Shalnikova, in: International Conference on Magnetism, Not-
tingham, 1964.
[28] O. Jarjayes, P.H. Fries, G. Bidan, J. Magn. Magn. Mater. 137 (1994) 205.
[29] L. Bedel, A.C. Roger, C. Estournès, A. Kiennemann, in: Proceedings of
9th International Symposium on Heterogeneous Catalysis, Varna, 2000,
p. 519.
[30] Powder Diffraction File, International Center for Diffraction Data, File
48-1818.
[31] W.C. Ellis, E.S. Greiner, Trans. Am. Soc. Met. 29 (1941) 415.
[32] Landolt Börstein—New Series III/19a 189.
[33] Landolt Börstein—New Series III/19a 157.
[34] P. Chaumette, P. Courty, A. Kiennemann, B. Ernst, Top. Catal. 2 (1995)
117.
Fischer–Tropsch activity is directly related to the amount of
metal alloy generated by partial reduction. Thus, the activity
increases with increasing lanthanum deficiency, that is, with an
increasing amount of γ -Fe2O3 nanocores in the fresh oxide. For
a similar lanthanum deficiency, the increasing amount of metal
with increasing initial calcination temperature has no beneficial
effect on catalytic activity, because of the concomitant enlarge-
ment of the alloy particle size.
The most deficient catalyst (y = 0.4) exhibited excellent C2–
C4 olefin selectivity. The decreased reaction temperature with
an increasing amount of metal led to diminished secondary
reactions of olefin readsorption and their consecutive hydro-
genation. The catalysts exhibited high stability under testing for
300 h.
The aim of improving catalytic activity by increasing the
amount of metal extractable by reduction with lanthanum de-
ficiency, without generating any real A-site deficiency, was
achieved. This is because the activity of the best catalyst of this
series (y = 0.4, CO conversion of 21% at 255 ◦C) was much
[35] J.B. Butt, T.-A. Lin, L.H. Schwartz, J. Catal. 97 (1986) 261.
[36] H. Schulz, M. Claeys, Appl. Catal. A 186 (1999) 109.
[37] J. Patzlaff, Y. Liu, C. Graffmann, J. Gaube, Appl. Catal. A 186 (1999) 145.
[38] A.A. Chen, M. Kaminsky, G.L. Geoffroy, M.A. Vannice, J. Phys. Chem.
90 (1986) 4810.

294
L. Bedel et al. / Journal of Catalysis 235 (2005) 279–294
[39] T. Ishihara, K. Eguchi, H. Araï, Appl. Catal. A 30 (1987) 225.
[40] H. Storch, N.G. Golumbic, R.B. Anderson, The Fischer–Tropsch and Re-
lated Syntheses, Wiley, New York, 1951.
[41] C.D. Frohning, H. Kölbel, M. Ralch, W. Rottig, F. Schnur, H. Schulz,
in: J. Falbe (Ed.), Chemierohstoffe aus Kohle, Georg Thieme Verlag,
Stuttgart, 1977, p. 219.
[43] J.W. Niemantsverdriet, A.M. van der Kraan, W.L. van Dijk, H.S. van der
Baan, J. Phys. Chem. 54 (1980) 3363.
[44] C.N. Satterfield, R.T. Hanlon, S.E. Tung, Z. Zou, G.C. Papaefthymiou,
Ind. Eng. Chem. Prod. Res. Dev. 25 (1986) 401.
[45] F. Tihay, A.C. Roger, G. Pourroy, A. Kiennemann, Energy Fuels 16 (2002)
1271.
[46] O. Ducreux, J. Lynch, B. Rebours, M. Roy, P. Chaumette, Stud. Surf. Sci.
Catal. 119 (1998) 125.
[42] H. Schulz, G. Schaub, M. Claeys, T. Riedel, Appl. Catal. A 186 (1999)
215.