Chemical state of a supported iron-cobalt catalyst during co disproportionation - II. Experimental study

Article abstract of DOI:10.1016/S0022-3697(00)00270-5

X-ray diffraction and M?ssbauer spectroscopy have been used to investigate the chemical state of a supported iron-cobalt catalyst during the disproportionation of carbon monoxide at 800 K. The results of this study have demonstrated a strong implication of cementite Fe₃C (may be in association with a cobalt carbide) in the alloy deactivation process. The deactivation due to the carburization is apparently reversible. The catalyst was, indeed, shown to partially recover its activity when submitted to conditions under which the carbide formation was not thermodynamically possible. The results of this study seem to indicate, however, that the alloy deactivation is not the consequence of a single phenomenon but results from the conjunction of two distinct processes. Additional deactivation due to the formation of a carbon layer at the particles surface appears as the most plausible hypothesis to explain this result.

Full text of DOI:10.1016/S0022-3697(00)00270-5

Journal of Physics and Chemistry of Solids 62 (2001) 1023±1037
www.elsevier.nl/locate/jpcs
Chemical state of a supported iron±cobalt catalyst during
co disproportionation
II. Experimental study
a
J.P. Pinheiro , P. Gadelle , C. Jeandey , J.L. Oddou
a,
b
b
*
a
 Â
Laboratoire de Thermodynamique et Physico-Chimie Metallurgiques, CNRS-INPG-UJF, Ecole Nationale Superieure d'Electrochimie et
 Á
d'Electrometallurgie de Grenoble, BP 7538402, Saint Martin d'Herss, France
b
 Â
Laboratoire de Chimie de CoordinationUnite de Recherche Associee UJF-CNRS no. 1194 Service de Chimie Inorganique et Biologique,
Â
 Á
Departement de Recherche Fondamentale sur la Matiere Condensee CEA38054 Grenoble Cedex 9, France
Received 28 January 2000; accepted 10 October 2000
Abstract
È
X-ray diffraction and Mossbauer spectroscopy have been used to investigate the chemical state of a supported iron±cobalt
catalyst during the disproportionation of carbon monoxide at 800 K. The results of this study have demonstrated a strong
implication of cementite Fe3C (may be in association with a cobalt carbide) in the alloy deactivation process. The deactivation
due to the carburization is apparently reversible. The catalyst was, indeed, shown to partially recover its activity when
submitted to conditions under which the carbide formation was not thermodynamically possible. The results of this study
seem to indicate, however, that the alloy deactivation is not the consequence of a single phenomenon but results from the
conjunction of two distinct processes. Additional deactivation due to the formation of a carbon layer at the particles surface
appears as the most plausible hypothesis to explain this result. q 2001 Elsevier Science Ltd. All rights reserved.
È
Keywords: A. Alloys; B. Chemical synthesis; B. Vapor deposition; C. X-ray diffraction; C. Mossbauer spectroscopy
1. Introduction
adequate conformation. A direct way consists in acting on
the composition of the gas mixture in contact with the cata-
lyst. Nolan et al. [7] have demonstrated, for instance, that the
addition of hydrogen to a CO±CO2 mixture resulted in a
radical change of the microstructure of the carbon deposit
produced by reaction over a nickel catalyst. Nolan observed,
indeed, that only nanotubes formed in the absence of hydro-
gen while, on the other hand, only carbon ®laments were
produced when hydrogen was introduced in the gas mixture.
We observed the same phenomena over a cobalt catalyst [8].
Alloying constitutes also an easy way to modify both the
catalyst activity and the morphology of the carbon deposit.
Chambers et al. [9] have observed, for instance, that the
addition of small amounts of copper to a cobalt catalyst
was suf®cient to produce a great increase of its activity
towards ethylene decomposition. The structure of the carbon
deposit was shown besides to differ greatly whether the reac-
tion was carried out with a cobalt or a cobalt±copper catalyst.
The Fe50 Co50 (wt%) alloy is an ef®cient catalyst for the
carbon ®lament growth when reacted with CO±CO2
The great variety of carbon materials that may be
produced by catalytic decomposition of hydrocarbons
constitutes probably one of the great interests of this
method. Structures as different as ®laments [1,2], nanotubes
[3±5] or platelets [6] can, indeed, be produced by applying
convenient reaction conditions. Although the fundamental
reasons for this diversity are not still understood, it seems
conceivable that, in the future, the morphology of the carbon
deposit could be designed in function of the required appli-
cation.
Independently of the nature of the catalyst itself, many
parameters have been shown to have an in¯uence on the
morphology of the carbon deposit. Many strategies can there-
fore be imagined to produce carbon materials with an
* Corresponding author. Tel.: 133-76-82-65-42; fax: 133-76-82-
66-77.
E-mail address: p.gadelle@ltpcm.inpg.fr (P. Gadelle).
0022-3697/01/$ - see front matter q 2001 Elsevier Science Ltd. All rights reserved.
PII: S0022-3697(00)00270-5

1024
J.P. Pinheiro et al. / Journal of Physics and Chemistry of Solids 62 (2001) 1023±1037
permits particularly to obtain information on the chemical
mixtures at temperature ranging from 723 to 873 K. In a
previous paper [10], the results of a kinetic study performed
at 800 K were brie¯y presented. This study demonstrated
that the evolution of the reaction rate vs time was strongly
depending on the composition of the gas mixture. In order to
understand the origin of this phenomenon, thermodynamics
was used to predict the possible transformations of this
catalyst when contacted with CO±CO2 mixtures. The aim
of the present study is to compare these thermodynamical
predictions with the experimental results. A special effort
was also made to understand the deactivation mechanism of
this alloy.
57
state of the iron atoms. Fe Mossbauer measurements were
È
generally performed at room temperature. The source was
57Co in a rhodium matrix, moved in a sinusoidal mode with
a maximum velocity calibrated at 12 mm s21. The 14.4 keV
g-rays were detected by means of a proportional counter and
È
Mossbauer spectra recorded on a 512 multichannel analyzer
[11]. The spectra were analyzed in a least-square procedure
by full diagonalization of the Hamiltonian describing the
quadrupolar and magnetic interactions.
A blank analysis of the Torr Sealw resin demonstrated
È
that its contribution to the global Mossbauer spectrum was
Experiments were carried out at 800 K on a Fe50 Co50
(wt%) catalyst with CO percentages ranging from 25 to
100%. Characterization of the catalyst after reaction was
not nil. As no iron compound was mentioned in the resin
formula, we attributed this absorption signal to an iron-bear-
ing impurity. The contribution of the resin to the whole
spectrum was, however, shown to be generally minor in
comparison with the contribution of the other phases and
did not perturb their identi®cation.
È
performed by a combination of Mossbauer spectroscopy
and X-ray diffraction. Special care was taken to protect
post-reaction samples from any contact with air.
2. Experimental process
3. State of the catalyst before the reaction
The Fe50 Co50 (wt%) catalyst was prepared by incipient
wetness impregnation of alumino-silicate ®bers (Nextelw)
with a solution of ethanol containing a mixture of iron and
cobalt nitrates. The nitrate amounts to be dissolved in etha-
nol were adjusted in order to give, after appropriate treat-
ment, a nominal metal content of about 10%. The
impregnated samples were then dried in air at 353 K for
20 h before further treatment. Experiments were carried
out in a small-sized ¯ow reactor system operated at atmo-
spheric pressure. Samples were pyrolized for 1 h under an
helium ¯ow at 523 K. After heating of the furnace to 800 K,
helium was replaced by an H2/Ar (5/95) mixture and the
catalysts were reduced for 16 h. After the reduction step,
the reactor was ¯ushed with helium for at least 4 h at the
same temperature in order to remove any residual hydrogen.
The catalysts were, then, reacted with CO±CO2. On-line
chromatographic analyses of the gas mixture during the
reaction, both at the inlet and outlet of the reactor, allowed
computation of the reaction rate. It must be noticed that, in
any case, an ultra pure CO was used in order to avoid impu-
rities that could affect the results.
Magnetic measurements on the supported catalyst after
reduction were carried out by S. Herreyre [12]. This study
demonstrated that the saturation magnetization of the cata-
lyst was lying between 193 and 210 uem cgs g21
(1 uem cgs g21 ˆ 1 A m2 kg21). This value is higher than
half the sum of the saturation magnetization of the pure
metals, which seems to demonstrate that iron and cobalt
are not only coexisting as separated phases but are really
alloyed.
This hypothesis is corroborated by other results.
The diffraction diagram of the reduced catalyst exhibits
two lines whose position is close to those expected for the
Fe50, Co50 (wt%) alloy (the iron±cobalt alloy is known to
exhibit peaks whose position is close to that of fcc-iron). It is
noteworthy that no line attributable either to hcp or fccco-
balt is observed. For information, the position and the rela-
tive intensity of the different peaks related to iron or cobalt
phases are indicated in the experimental diffraction diagram
(Fig. 1).
Ê
The experimentally-determined cell parameter (2.8506 A)
is lower than the value which can be extrapolated from Ellis
Ê
and Greiner's data (2.8570 A) [13,14] and is closer to the
The reaction was stopped by ¯ushing the reactor with
helium and quenching it to room temperature. Owing to
its small dimensions, the reactor could be introduced in a
value expected for an alloy with approximate mass compo-
sition Fe41, Co59 (wt%). This would imply that 30.5% of
the iron atoms are not alloyed with cobalt. The hypothesis of
two separated phases is, however, contradicted by the
diffraction diagram. This one does not exhibit any peak
besides those due to the Nextelw and those we attributed
to the alloy. The latter are symmetrical. It is therefore
implausible that they result from the superposition of two
peaks, one due to the alloy and the other to isolated iron
particles. Consistently, an average crystallite diameter of
20±30 nm was estimated from the broadening of the X-
ray lines, values which are in good agreement with the direct
È
glove box. The X-ray diffraction and Mossbauer samples
were, by this way, prepared without any contact with air.
The whole post reaction sample was crushed and a part of
the resulting powder was characterized without additional
treatment using X-ray diffraction under controlled atmo-
Ê
sphere (l ˆ 1.5406 A). The rest of the crushed sample
È
was immersed in a Torr Seal w resin since Mossbauer
studies were performed under ambient atmosphere.
È
The Mossbauer spectroscopy is an analysis technique
specially well suited for compounds containing iron. It

J.P. Pinheiro et al. / Journal of Physics and Chemistry of Solids 62 (2001) 1023±1037
1025
Ê
Fig. 1. Diffraction diagram of the reduced iron±cobalt catalyst (l ˆ 1.5406 A).
È
Fig. 2. Mossbauer spectrum of the reduced iron±cobalt catalyst.
measurements performed on M.E.T. The hypothesis of
superposed peaks can therefore be ruled out.
the precautions taken to avoid any contact of the samples
È
with air. The Mossbauer spectrum of the reduced catalyst is
The difference with Ellis and Greiner's results may, in our
opinion, rather be attributed to the small size of the catalyst
particles. Several authors [15,16] have, indeed, reported
variations of the cell unit size occurring when the dimension
of the particles is lowered to nanometric scale. These varia-
tions are generally assumed to result from modi®cations of
the cohesive forces when the number of atoms decays under
100 in any direction of space.
characterized by relatively large absorption lines around
0.38 mm s21. This feature generally indicates the existence
of a distribution of hyper®ne ®elds inside the sample which
would be the sign, in the present case, of local variations of
the alloy composition. This distribution leads to a broad-
ening of the alloy spectrum lines which is more perceptible
for higher velocities (i.e. at both extremities of the alloy
spectrum). In order to improve the whole spectrum ®tting,
the alloy six-line pattern was analyzed, in the following, as a
combination of three two-line patterns characterized, each
of them, by different line widths. For the other compounds
which appeared in the studied samples, all the lines corre-
sponding to a de®ne spectrum were, on the contrary,
constrained to have the same width.
È
The Mossbauer analysis of the reduced catalyst provided
additional information (Fig. 2). The hyper®ne ®eld of the
Fe50, Co50 (wt%) alloy is not known accurately: according
to Johnson et al. [17], it would be between 346 and 355 kOe
(1 Oe ˆ 0.1 T). The value determined in the present work
(344 kOe) is roughly in agreement with these values. The
absence of peak corresponding to oxides, as previously in
the diffraction diagram, seems to con®rm the ef®ciency of
Experiments were also carried out to compare the reac-
tivity of the pure metals with those of the iron±cobalt

1026
J.P. Pinheiro et al. / Journal of Physics and Chemistry of Solids 62 (2001) 1023±1037
Fig. 3. Comparison of the alloy reactivity with that of its pure components.
mixture. The pure metals catalysts were prepared using the
same method as previously described for the iron±cobalt
catalyst. The reaction kinetic was shown to differ consider-
ably whether pure metals or an iron±cobalt mixture were
used. As can be seen in Fig. 3, lower reaction rates were
systematically measured in the case of the pure metals. In
addition, these catalysts were shown to be more sensitive to
deactivation than the iron cobalt catalyst.
proposed by Snoeck et al. [18,19]. Contrary to these
authors, we do not take into account the segregation
behavior of the carbon, which induces some differences
in the carbon concentration pro®le inside the particle,
principally near the surface. We believe, however, that
these differences do not question the validity of our
reasoning.
The carbon concentration in the particle is expected to
considerably increase during the period preceding the
carbon nucleation. Since no ®lament growth is observed,
it is, indeed, probable that the carbon issued from the CO
decomposition at the surface diffuses through the catalyst
particle and accumulates leading to a carbon supersatura-
tion. According to this model, the nucleation occurs when
the carbon concentration at the support side exceed a critical
value, noted C*. As the ®lament growth is initiated, the
carbon concentration at the support side drops to a value
equal to the carbon ®lament solubility in the alloy.
In conclusion, all these results seem to corroborate that
iron and cobalt are really alloyed.
4. Evolution of the reaction rate at the beginning of the
reaction
The results of our kinetic study demonstrate that the time
required to reach the maximal rate decreases when the CO
percentage is increased. As an indication, maximal rates
were measured at 800 K after times varying from 30 to
60 min for CO percentages lower than 30% and after only
6 min for higher CO percentages.
The time required to reach the critical concentration C*
depends, in this model, on the concentration gradient in the
particle. If an equilibrium is established at the gas±metal
interface, the carbon concentration at this interface is deter-
mined by the composition of the gas mixture and increases
as the CO percentage is augmented. It follows that the
nucleation occurs more quickly when the CO fraction in
the mixture is higher. This is in good agreement with the
evolution evidenced experimentally (i.e. the time required to
reach the maximal reaction rate decreases as the CO percen-
tage in the gas mixture is increased).
We will refer in this work to the classical model for the
carbon ®lament growth. According to this model, the
carbon-bearing gas adsorbs at the surface of the particle
(gas±metal interface) and decomposes. The resulting
species dissolve then into the metal particle and diffuse
through the particle to the rear faces (metal±carbon inter-
face) where they precipitate to form a ®lamentous carbon
structure. The ®lament growth is probably initiated at
distinct times for each particle depending on parameters
like the particle diameter, the crystallographic structure of
the surface, etc. The initial increase of the reaction rate
would be therefore characteristic of a transient regime.
A model for the evolution of the carbon concentration
in a catalyst particle during this initial period is proposed
in Fig. 4. This model is deliberately simplistic and has to
be considered as a simpli®ed version of the model
Another phenomenon which frequently occurs at the
beginning of the reaction is the fragmentation of the largest
catalyst particles into smaller ones. This phenomenon plays
a predominant role when the catalyst is used in the form of
foils or ®lings since it is generally assumed that only the
smallest particles are active towards ®lament formation. The
fragmentation, as it enlarges the active surface of the cata-
lyst, induces an increase of the reaction rate. The catalyst is,

J.P. Pinheiro et al. / Journal of Physics and Chemistry of Solids 62 (2001) 1023±1037
1027
Fig. 4. Evolution of the carbon concentration inside the catalyst particle at the beginning of the reaction. (1), (2), (3): carbon accumulation
inside the particle, (4): beginning of the carbon ®lament nucleation, (5): steady state growth of the carbon ®lament.
in the present case, supported by a substrate. The catalyst
particles are therefore smaller than in the case of catalysts
foils or ®lings and direct reaction of the particles with
the gases is possible without preliminary fragmentation.
The contribution of fragmentation to the initial increase of
the reaction rate can therefore be considered as secondary.
Several authors [20,21] mention the existence of an initial
induction period during which no reaction is observed. No
such phenomenon was noticed in our case. It seems, on the
contrary, that carbon deposition rapidly occurs after the
catalyst is contacted with the gas mixture. This fact is
supported by T.E.M. observations which clearly establish
the ®lamentous nature of the carbon deposit after a reaction
time equal to 15 min (Fig. 5).
5.1. Stability of the compounds formed during the reaction
A preliminary experiment was carried out to estimate the
stability of the compounds formed during the reaction. A
catalyst was, ®rst, completely deactivated by reaction with
pure CO at 800 K. After the reactor was ¯ushed with
helium, the temperature was then rapidly decreased to
293 K. The catalyst was maintained at this temperature for
20 h. After this period, the temperature of the reactor was
increased to 800 K and the catalyst was anew contacted with
pure CO. No recovery of the catalyst activity was observed
after this treatment. This experiment demonstrates that the
compounds responsible for the deactivation of the catalyst
are stable (or metastable) under inert atmosphere and that
they can therefore be isolated using classical analysis tech-
niques.
5. Evolution of the reaction rate after the maximum
5.2. State of the catalyst after the reaction
The evolution of the reaction rate after the maximum was
shown to depend strongly on the composition of the gas
mixture. The reaction rate was, indeed, observed either to
level off or to decrease steadily depending on the CO
percentage. The following study was carried out to identify
the processes responsible for the catalyst deactivation.
The kinetic curves corresponding to a ®rst set of experi-
ments are plotted in Fig. 6. These experiments were
conducted until complete deactivation of the catalyst. In
some cases (namely with 89.7 and 100%CO), the experi-
ments were pursued even after the reaction had stopped in

1028
J.P. Pinheiro et al. / Journal of Physics and Chemistry of Solids 62 (2001) 1023±1037
Fig. 5. Carbon ®lament formed after a 15-min reaction.
order to increase the catalyst transformation. The only
exception concerned the reaction with 31.3%CO. The cata-
lyst was, in this case, still active when the experiment was
stopped, 23 h after the beginning of the reaction.
this oxide. As visible on Figs. 8 and 9, the lines correspond-
ing to Fe3O4 disappear, both on diffraction diagrams and
È
Mossbauer spectra, for higher CO percentages.
È
The Mossbauer spectra obtained for 89.7 and 100%CO
The evolution of the different compounds contributions is
plotted in Fig. 7 as a function of the gas mixture composi-
tion. We de®ne the contribution of a compound as the ratio
between the area of its pattern and that of the whole
can be described, neglecting the resin contribution, as a
combination of two six-line patterns. The ®rst of them is
attributable to the alloy. From the number and the position
of the absorption lines, the second pattern was ascribed to
the cementite uFe3C.
È
Mossbauer spectrum. As can be seen, the resin contribution
is relatively constant and accounts for approximately 6±7%
of the total spectrum area. Since it originates from an impur-
ity, the resin contribution constitutes a good gauge of the
measures uncertainty. We resolved, for this reason, to only
consider as really present the compounds whose contribu-
tion exceeded that of the resin.
The iron atoms are distributed in this carbide between two
kinds of sites submitted to close hyper®ne ®elds (206 and
208 kOe at 296 K) and having a similar isomeric shift
(0.18 mm s21/aFe at 296 K) [23]. The atoms occupying
these two sites are consequently undiscernible below the
Curie temperature (TC ˆ 485 K) and the six-line patterns
corresponding to each site are completely merged into an
unique sextuplet.
È
The presence of magnetite Fe3O4 was evidenced by Moss-
bauer spectroscopy for 31.3%CO (Fig. 8). This iron oxide
exhibits at room temperature a complex asymmetrical spec-
trum. The most recognizable absorption lines of this spec-
trum are located at about 7.9 mm.s21 and between 28 and
26.5 mm.s21 [22]. The formation of Fe3O4 was con®rmed
by X-ray diffraction (Fig. 9). The diffraction pattern of this
sample exhibits, indeed, lines for 2u values closer from
35.422, 56.942 and 62.5158 which are characteristic of
uFe3C is not the only iron carbide exhibiting such a spec-
trum. e0Fe2.2C is also known to only have one kind of site
[24]. However, the hyper®ne ®eld of this compound is too
different to make confusions possible with cementite
(H ˆ 173 kOe for e0`Fe2.2C).
The analysis of these samples by means of X-ray diffrac-
tion did not provide additional information on the catalyst

J.P. Pinheiro et al. / Journal of Physics and Chemistry of Solids 62 (2001) 1023±1037
1029
Fig. 6. Time dependence of the reaction rate for different gas mixture compositions (Fe50 Co50 (wt%), T ˆ 800 K).
state. No peak attributable either to iron or cobalt carbide
was, indeed, clearly evidenced. This result does not ques-
È
tion, however, the validity of the Mossbauer analyses. The
complexity of the substrate diffraction pattern makes,
indeed, dif®cult the identi®cation of additional peaks. The
detection of such peaks is, moreover, complicated by the
fact that small amounts of catalysts are generally used (it
is to be reminded that the catalyst only accounts for 10% of
the weight of the whole sample). In the particular case of the
cementite, these dif®culties are increased by the fact that the
most intense peak corresponding to this carbide is hidden by
a peak originating from the alloy.
One of the most important result of this ®rst set concerns
È
the experiment carried out with 50.0%CO. The Mossbauer
analysis demonstrated, indeed, that the alloy did not undergo
in this case any transformation. This absence of evolution is
in accordance with the phase diagram which predicts the
existence of an alloy stability domain (Fig. 10). The implica-
tions of this result will be discussed further.
È
Complementary Mossbauer analysis were performed at
77 K. Since temperature variations generally induces great
alterations of the spectra appearance, this procedure is often
È
Fig. 7. Evolution of the compounds contributions to the Mossbauer
spectrum as a function of the gas mixture composition.

1030
J.P. Pinheiro et al. / Journal of Physics and Chemistry of Solids 62 (2001) 1023±1037
È
Fig. 8. Evolution of the Mossbauer spectra as a function of the gas mixture composition.
used to con®rm the identity of the analyzed compounds. The
results of these analyses were close to those obtained at
room temperature (Table 1). In particular, the evolution
with temperature of the patterns attributed to Fe3O4 and
Fe3C was shown to agree within the experimental error
with that reported in literature.
The position of these experiments on the phase diagram is
illustrated on Fig. 11 and the results are summarized in
Fig. 12. This second set of experiments demonstrated that the
formation of Fe3O4 occured for CO percentages much higher
than those predicted (the presence of this iron oxide was,
indeed, evidenced for CO percentages as high as 44.6%). On
the other hand, the experiment carried out with 57.2%CO
seemed to con®rm the existence of the alloy stability domain.
Many hypotheses can be proposed to explain the diver-
gences between the experimental results and the predictions
of the phase diagram. We favor, for our part, the hypothesis
of local variations of the gas mixture composition. The
nature of the compounds formed during the reaction
depends fundamentally on the CO fraction at the surface
of the catalyst particles. Its value may differ notably from
the bulk fraction if the reaction is controlled by mass
transfer phenomena in the gas phase. Such a phenomenon
Fig. 10 illustrates the positioning of the different experi-
ments in the Fe-Co-C-O phase diagram calculated at 800 K.
The evolution of the different compounds contributions is,
on the whole, in accordance with what was expected from
this diagram. Important differences are, however, percepti-
ble respecting the extent of the alloy stability domain. Fe3O4
was, for example, evidenced for 31.3%CO although the
formation of this oxide was not predicted for CO percen-
tages higher than 28%.
Complementary experiments were carried out in order to
de®ne more accurately the limits of the alloy stability domain.

J.P. Pinheiro et al. / Journal of Physics and Chemistry of Solids 62 (2001) 1023±1037
1031
Fig. 9. Evolution of the X-ray patterns as a function of the gas mixture composition.
could explain the displacement of the stability domain
observed experimentally. Because of the mass-transfer
limitations, the CO percentage of the gas mixture in contact
with the particles would, indeed, be lower than that in the
bulk and Fe3O4 could form in apparently unfavorable condi-
tions. It is also within the realm of possibility that the trans-
the chemical state of the iron atoms. The reconstitution of
the cobalt evolution during reaction was principally based,
for this reason, on the information provided by the X-ray
diffraction analyses and by the phase diagram. Many para-
meters made dif®cult such a study:
È
formations evidenced by Mossbauer spectroscopy and X-
² The phase diagrams did not take into account complex
compounds (i.e. compounds including both iron and
cobalt atoms). The thermodynamic data used to deter-
mine the stability domains of the cobalt carbides were
moreover doubtful [25].
ray diffraction for CO percentages higher than 28% do not
affect all the particles but only a small fraction of them
which would be submitted to local particular conditions.
² The information provided by the X-ray diffraction
analyses is not always satisfactory principally for the
higher CO percentages.
5.3. Evolution of the cobalt during the reaction
È
The Mossbauer spectroscopy only gives information on

1032
J.P. Pinheiro et al. / Journal of Physics and Chemistry of Solids 62 (2001) 1023±1037
Fig. 10. Position of the ®rst set of experiments at 800 K.
This explains that there was still some doubt on the real state
of cobalt after reaction in some cases. The possibility of
complex compounds formation, in particular, will be
discussed in the following.
are equally distributed in this oxide between two kinds of
sites with similar hyper®ne ®elds (516 kOe) [26]. Ultra®ne
CoFe2O4 powders are known besides to show superpara-
magnetic behavior [26,27]. The spectrum of this oxide (a
six-line pattern, sometimes superimposed on a two-line
pattern) differs greatly, for all these reasons, from the spec-
trum of magnetite. This result tended to prove the validity of
the phase diagram predictions for the lower CO percentages
(i.e. only the iron atoms reacted to form Fe3O4 while the
cobalt remained unchanged).
CoFe2O4 is, to our knowledge, the only complex oxide
reported in literature. As magnetite, CoFe2O4 crystallizes in
a face-centered cubic structure. The cell parameters of these
Ê
Ê
two oxides are very close (8.377 A for CoFe2O4 and 8.396 A
for Fe3O4). They can hardly, consequently, be differentiated
by means of X-ray diffraction.
È
The Mossbauer analyses enabled, however, in the present
case, to rule out the formation of this oxide. The iron atoms
In their study of the oxidation of a Ni 59, Fe 41 (wt%)
alloy, Greco et al. [28] observed that the interaction of
oxygen with a (100) surface induced a pronounced segrega-
tion of the alloy components. The compound formed at the
surface by reaction with oxygen was predominantly an iron
oxide including small amounts of nickel. The nickel ratio
could even be reduced by annealing the sample, suggesting
that the equilibrium state of this surface oxide corresponded
to the pure iron oxide. We can not rule out, in the present
case, the occurrence of a similar phenomenon. It is therefore
possible that the catalyst particles are covered by a magne-
tite layer.
Table 1
È
Comparison of the Mossbauer analyses performed at room tempera-
ture and 77 K
CO/CO2 ratio Contribution to the
Contribution to
È
Mossbauer spectrum
(room temperature)
È
the Mossbauer
spectrum (77 K)
31.3%CO
50.0%CO
100%CO
Alloy: 77.5%
Fe3O4: 13.3%
Fe3C: 2.4%
Resin: 6.8%
Alloy: 88.1%
Fe3O4: 3.1%
Fe3C: 1.9%
Resin: 6.9%
Alloy: 78.9%
Fe3O4: 0.0%
Fe3C: 15.4%
Resin: 5.7%
Alloy: 79.5%
Fe3O4: 14.2%
Fe3C: 0.0%
Resin: 6.3%
Alloy: 90.4%
Fe3O4: 3.0%
Fe3C: 0.0%
Resin: 6.6%
Alloy: 84.0%
Fe3O4: 0.0%
Fe3C: 10.2%
Resin: 5.8%
For the higher CO percentages, the phase diagram
predicts the coexistence of Fe3C and Co2C. No peak attri-
butable to theses compounds was, however, evidenced on
the diffraction patterns.
The formation of a complex carbide with a composition
close to (Fe0.5Co0.5)3C was reported by Pavel et al. [29]
under conditions, however, greatly different from the present
ones (these authors were, indeed, interested in synthesizing
diamond at a pressure of t5.5 GPa and a temperature of
È
t1800 K). On the basis of a Mossbauer study, they

J.P. Pinheiro et al. / Journal of Physics and Chemistry of Solids 62 (2001) 1023±1037
1033
Fig. 11. Position of the second set of experiments at 800 K.
attributed tothiscarbide an hyper®ne ®eld of 193.5 kOe, value
which is slightly lower than that of cementite (206,208 kOe).
This complex carbide would be structurally equivalent to
a solid solution of Fe3C and Co3C. These two compounds
have a similar crystallographic structure and very close cell
unit parameters (Table 2). The formation of a solid solution
involving these two compounds is therefore fully plausible.
(It must be noted that a similar reasoning with Fe3C and
Co2C brings us to the conclusion that these two phases
are, on the contrary, not miscible). As no thermodynamic
data concerning this complex carbide was available, the
determination of the conditions favoring its formation was
not possible. The positioning of its stability domain in the
phase diagram is, therefore, still a mystery.
È
Fig. 12. Evolution of the compounds contributions to the Mossbauer spectra as a function of the gas mixture composition (second set of
experiments).

1034
J.P. Pinheiro et al. / Journal of Physics and Chemistry of Solids 62 (2001) 1023±1037
Fig. 13. In¯uence of a modi®cation of the gas mixture composition on the reaction rate.
Failing more accurate information, no de®nitive conclu-
The results of this study also demonstrate a strong implica-
tion of cementite (may be in association with a cobalt
carbide) in the alloy deactivation process. A survey of the
kinetic results shows, indeed, that the catalyst deactivation
occurs all the more rapidly that the CO ratio in the gas
sion can be drawn concerning the state of cobalt for the
higher CO percentages. Cobalt could either exist as Co2C
(as predicted by the phase diagram) or form as Co3C a solid
solution with Fe3C. As previously the hypothesis of a segre-
gation of the alloy components cannot be ruled out. The
surface of the particles could therefore be covered by a
Fe3C layer.
È
mixture is higher. As demonstrated by the Mossbauer
analyses, this evolution can be correlated to an increase of
the Fe3C amount formed during the reaction.
In order to con®rm this hypothesis, we submitted, in
the same experiment, a reduced catalyst to gas mixtures
with different CO/CO2 ratios (Fig. 13). The catalyst was,
®rst, contacted with pure CO at 800 K. In accordance
with the phase diagram which predicts that such condi-
tions favor the formation of the carbide phases, a
complete deactivation of the catalyst was observed 3 h
after the beginning of the reaction. The composition of
the gas mixture was, at this time, rapidly modi®ed
(28.5%CO) in such a way that the catalyst was submitted
to conditions under which the carbide formation was not
thermodynamically possible. A recovery of the catalyst
activity was, then, rapidly observed. This experiment, as
well as it con®rmed the role played by the carbide phases
in the poisoning process, demonstrated that the alloy
deactivation is (at least partially) a reversible phenom-
enon. It is noteworthy that the evolution of the reaction
rate after the catalyst activity recovery was similar to
that previously described for the lower CO percentages.
The reaction could be, indeed, prolonged for 16 h after
the modi®cation of the gas mixture composition and no
6. Deactivation mechanism of the alloy
It appears from this study that the formation of Fe3O4 only
slightly affects the reaction rate. This result can be inter-
preted in two different ways:
1. The magnetite catalytic activity is similar to that of the
iron cobalt alloy. Bennet et al. [30]have managed to
obtain carbon ®laments by reacting Fe3O4 with acetone.
The few information we have about these experiments
does not permit, however, to conclude whether Fe3O4 is
really the active phase or it is reduced when contacted
with acetone (which would mean that the decomposition
of acetone is, in fact, catalyzed by iron).
2. The magnetite has no catalytic activity but its formation
leads to a fragmentation of the catalyst particles. The
resultant increase of the active area could therefore
compensate the fact that Fe3O4 does not catalyze the
CO disproportionation.
Table 2
Crystallographic structure of Fe3C and Co3C
Ê
Cell parameters (A)
Compound
Structure
Fe3C
Co3C
Orthorhombic
Orthorhombic
a ˆ 5.0910
b ˆ 4.993
b ˆ 6.7434
c ˆ 6.707
c ˆ 4.5260
a ˆ 4.444

J.P. Pinheiro et al. / Journal of Physics and Chemistry of Solids 62 (2001) 1023±1037
1035
Fig. 14. Alloy deactivation mechanism when the carbide formation is not thermodynamically possible.1. Fr . Fd: formation of a carbon layer
at the surface of the particle. 2. Fr ˆ Fd: the carbon coating stops.
sign of catalyst deactivation was perceptible when the
experiment was stopped.
decrease of the reaction rate (®gure). The rate of carbon
produced at the surface becomes, for this reason, progres-
sively closer to that diffusing through the alloy particle.
When the equality is reached, the carbon coating stops
and the reaction rate stabilizes at a constant value. In
absence of other deactivating phenomenon, no further
decrease of the reaction rate should be observed. Our results
seems to demonstrate, however, that the carbon coating goes
on, with a reduced rate, until complete deactivation of the
catalyst. This idealized model has the merit of explaining
the initial decrease of the reaction rate as well as its posterior
stabilization.
We can conclude on the basis of these results that the
cobalt and iron carbides have no catalytic activity towards
the carbon monoxide disproportionation. Concerning the
cementite, this conclusion is consistent with the ®ndings
of authors who have studied the interactions between iron
and carbon-bearing gases [31±33].
This study demonstrates, however, that the carbide
formation is not the only phenomenon responsible for the
È
catalyst poisoning. The Mossbauer analyses performed for
50.0 and 57.2%CO showed, indeed, that the deactivation did
not result from a chemical transformation of the catalyst. In
the absence of any chemical transformation of the catalyst,
the most probable hypothesis is that the deactivation results,
in this case, from the formation of a carbon layer at the
surface of the particles which will prevent any contact
with the gas mixture.
The previous model is slightly modi®ed when the carbide
formation becomes thermodynamically possible (Fig. 15).
The deactivation of the catalyst is, in that case, the conse-
quence of two phenomena acting conjointly. As previously,
the carbon coating and the carburization of the catalyst
particles leads to a strong decrease of the reaction rate in
the beginning of the reaction. When Fr becomes equal to Fd,
the carbon coating phenomenon is stopped (or, at least,
notably slowed down). The carbide formation proceeds,
on the contrary, until complete transformation of the parti-
cles surface. No stabilization of the reaction rate is, there-
fore, observed in this case and the carbide formation on the
carbon-free surface leads rapidly to complete deactivation
of the catalyst. A good example of such an evolution is
given by the experiment carried out under pure CO.
According to our experimental results, the carbon coating
would only occur at 800 K for CO percentages higher than
35±45%. At the same temperature, the carbide formation
was only evidenced for CO percentages higher than 57±
65%. The number of interfering phenomena would, there-
fore, depend on the gas mixture composition. This would
explain the different evolutions of the reaction kinetics with
the composition of the gas mixture.
It can be assumed that the carbon coating occurs when the
rate of carbon produced by the CO disproportionation at the
surface (noted Fr) is greater than the rate diffusing through
the particle (Fd). A model for the catalyst deactivation when
no carbide formation is thermodynamically possible is
proposed in Fig. 14. The formation at the beginning of the
reaction of a carbon layer partially isolating the surface of
the particles from the gas phase leads to a important
7. Conclusion
The formation of cementite Fe3C was evidenced at 800 K
È
by Mossbauer spectroscopy for CO percentages higher than
57±65%. The results of this study demonstrate that this

1036
J.P. Pinheiro et al. / Journal of Physics and Chemistry of Solids 62 (2001) 1023±1037
Fig. 15. Alloy deactivation mechanism when the carbide formation is thermodynamically possible. 1. Fr . Fd:carbide formation 1 carbon
coating. 2. Fr ˆ Fd: the carbon coating stops but the carbide formation goes on. 3. The carbon-free surface is completely carburized.
A.G. Rinzler, D.T. Colbert, K.A. Smith, R.E. Smalley, Chem.
Phys. Letters 296 (1998) 195.
È
[5] A. Thaõb, G.A. Martin, J.P. Pinheiro, M.C. Schouler, P.
compound is strongly involved in the alloy deactivation
phenomenon. The catalyst was, indeed, shown to partially
recover its activity when submitted to conditions under
which the carbide formation was not thermodynamically
possible.
Gadelle, Catal. Lett. 63 (1999) 135.
[6] N.M. Rodriguez, A. Chambers, R.T.K. Baker, Langmuir 11
(1995) 3862.
The results of this study seem to indicate, however, that
the alloy deactivation is not the consequence of a single
phenomenon but results from the conjunction of two distinct
processes. The alloy deactivation for 50.0%CO was, indeed,
found not be due to any chemical transformation. Additional
deactivation due to the formation of a carbon layer at the
particles surface appears as the most plausible hypothesis to
explain this result.
[7] P.E. Nolan, D.C. Lynch, A.H. Cutler, J. Phys. Chem. B102
(1998) 4165.
[8] J.P. Pinheiro, M.C. Schouler, P. Gadelle, M. Mermoux, E.
Doorhyee, Carbon 38 (2000) 1469.
[9] A. Chambers, N.M. Rodriguez, T.K. Baker, J. Phys. Chem. 99
(1995) 10581.
[10] J.P. Pinheiro and P. Gadelle, J. Phys. Chem. Solids (Accepted
for publication).
[11] J.M. Alameda, D. Givord, C. Jeandey, H.S. Li, Q. Lu, J.L.
Oddou, J. Physique 46 (1985) 1581.
The evolution of the cobalt atoms during the reaction was
not clearly elucidated. Contrary to the predictions of the Fe-
Co-C-O diagrams, no cobalt carbide could be evidenced by
means of X-ray diffraction. Its formation is nevertheless
plausible. A segregation of the two alloy components during
the reaction can not be ruled out.
[12] S. Herreyre, Thesis, Grenoble, 1995.
[13] W.C. Ellis, E.S. Greiner, Tr. Amer. Soc. Met. 29 (1941) 415.
[14] W.B. Pearson, in Handbook of Lattice Spacings and Struc-
tures of Metals and Alloys. Edited by Pergamon Press Ltd,
London, Great Britain, 1964.
[15] F.W.C. Bosswell, Proc. Phys. Soc. A 64 (1951) 465.
[16] E.A. Owen, D. Madoc Jones, Proc. Phys. Soc. B 67 (1954)
456.
References
[17] C.E. Johnson, M.S. Ridout, T.E. Cranshaw, P.E. Madsen,
Phys. Rev. Letters 6 (9) (1961) 450.
[1] M. Audier, J. Guinot, M. Coulon, L. Bonnetain, Carbon 19
(1981) 99.
[18] J.-W. Snoeck, G.F. Froment, M. Fowles, J. Catal. 169 (1997)
240.
[2] S. Herreyre, P. Gadelle and J.L. Ginoux, in Extended
Abstracts Carbon `94, Granada, Spain, 5222 (1994).
[3] P.E. Nolan, M.J. Schabel, D.C. Lynch, A.H. Cutler, Carbon 33
(1995) 79.
[19] J.W. Snoeck, G.F. Froment, M. Fowles, J. Catal. 169 (1997)
250.
[20] F.W.A.H. Geurts, R.G. Cnossen, A. Sacco Jr., R.R. Bieder-
man, Carbon 32 (1994) 1151.
[4] J.H. Hafner, M.J. Bronikowski, B.R. Azamian, P. Nikolaev,

J.P. Pinheiro et al. / Journal of Physics and Chemistry of Solids 62 (2001) 1023±1037
1037
[21] C. Vanvoren, Thesis, Grenoble, 1981.
[22] L. HaggstroÈm, H. Annersten, T. Ericsson, R. WaÈppling,
W. Karner, S. Bjarman, Hyper®ne Interactions 5 (1978)
201.
[27] W.J. Schuele, S. Shtrikman, D. Treves, J. Appl. Phys. 36
(1965) 1010.
[28] S.E. Greco, J.P. Roux, J.M. Blakely, Surf. Sci. 120 (1982) 203.
[29] E. Pavel, G. Baluta, D. Barb, D.P. Lazar, M. Morariu, M.
Popescu, M. Sorescu, J. Mat. Sci. 28 (1993) 1645.
[30] M.J. Bennet, G.H. Chaffey, A.J. Langford, D.R.V. Silvester,
A.E.R.E. Report R7407 (quoted by Ref. [31]).
[31] R.T.K. Baker, J.R. Alonzo, J.A. Dumesic, D.J.C. Yates, J.
Catal. 77 (1982) 74.
È
[23] G. Le Caer, J.M. Dubois, J.P. Senateur, J. Solid State Chem.
19 (1976) 19.
[24] J.W. Niemantsverdriet, A.M. van der Kraan, W.L. van Dijk,
H.S. van der Baan, J. Phys. Chem. 84 (1980) 3363.
[25] Cobalt Monograph Edited by Centre d'Information du Cobalt,
1960.
[32] P.L. Walker Jr., J.L. Rakszewski, G.R. Imperial, J. Phys.
Chem. 63 (1959) 140.
[26] T.K. McNab, A.J.F. Boyle, Proceedings of the Conference on
Hyper®ne Structure and Nuclear Radiations, North Holland,
Amsterdam, 1968, p. 957.
[33] S. Herreyre, P. Gadelle, P. Moral, J.M.M. Millet, J. Phys.
Chem. Solids 58 (1997) 1539.