Full text of DOI:10.1557/JMR.2002.0283

ARTICLES
Catalytic effects on carbon/carbon composites fabricated by a
film boiling chemical vapor infiltration process
H. Okuno, M. Trinquecoste, and A. Derre´
CRPP/CNRS, Avenue Dr Albert Schweitzer, 33600 Pessac, France
M. Monthioux
CEMES/CNRS, 29, rue Jeanne Marvig, 31055 Toulouse Cedex 4, France
P. Delhae`s
CRPP/CNRS, Avenue Dr Albert Schweitzer, 33600 Pessac, France
(Received 1 October 2001; accepted 11 April 2002)
Chemical vapor infiltration (CVI) has been widely studied under several conditions to
obtain C/C composites. A “film boiling technique” (so-called Kalamazoo), by the use of
liquid precursor, based on thermal gradient CVI has been recently developed as one
of the very effective techniques to increase the carbon yield and the densification rate.
A small cold wall type laboratory reactor has been realized to analyze the kinetics of
reactions and the deposited pyrocarbon matrix. In this study, ferrocene, as the source
of catalyst, is mixed to the liquid precursor to induce a catalytic effect on the film
boiling technique since the transition metals are known to increase the carbon
deposition rate. In addition to an important increase of the densification rate, it is
revealed that the deposition mechanism and microtextures are completely modified by
the presence of catalyst, with the presence of multiwall nanotubes within the matrix. A
model has been adapted from Allendorff and Hunt’s work to interpret this peculiar
deposition mechanism.
I. INTRODUCTION
improved carbon yield, a very low residual porosity, and
a “good” matrix quality. The usual matrix microtextures
are observed in specific anisotropic pyrocarbons, called
smooth laminar and rough laminar; the second one,
known for being graphitizable, is the most useful for
applications.3
The C/C composites have been studied as very useful
structural materials because of their thermal shock resis-
tance due to high thermal conductivity and low thermal
expansion, in addition to their mechanical properties
such as high strength, toughness, and stiffness. To com-
bine a porous fibrous preform with a carbon matrix act-
ing as a binder between the fibres, the most widespread
method to obtain the C/C composite is the isothermal
chemical vapor infiltration (CVI) process. Gaseous hy-
drocarbons are used as precursors, which lead to the
deposition of carbon matrix, in a competition between
pyrolysis chemical reactions and diffusion of the chemi-
cal species inside the porous preform.
Thermal gradient CVI processes using cold wall reac-
tors overcome diffusion limitations and increase the in-
filtration rate, giving shorter processing times.1 More
recently, a fast densification process, based on a film
boiling technique and so-called “Kalamazoo,” has been
developed.2 It is based on a very strong thermal gradient
created—by inductive or resistive Joule effect—between
the central part of the heated fibrous preform and its
peripheral part immersed in the liquid hydrocarbon pre-
cursor. The well-known results for the deposited matrix
are respectively a very high rate of densification with an
One effective factor to modify the deposition way and
the infiltration kinetics of CVI is the catalytic effect
which gives birth to a new process family: the catalytic
chemical vapor infiltration, CCVI.1 Many previous stud-
ies showed that small particles of transition metals such
as Fe, Ni, or Co have significant effect on the control of
the nucleation and growth characteristics of carbon4,5
and thus enhance appreciably the carbon deposition rate
from hydrocarbons.6 The principle of this effect is attrib-
uted to carbon diffusion through the catalytic particles,
which promotes gas decomposition. It has already been
revealed that precursor gas decomposition occurs on
these metal particles at lower temperatures than on regu-
lar carbon substrates.7 In addition, the catalyst particles
reaching the heated surface increase the density of avail-
able nucleation sites, which promotes the rapid matrix
deposition.8 CCVI is sensibly influenced by several ex-
perimental parameters such as catalyst particle size and
nature of the transition metal, which modify the process
for the carbon formation.
1904
J. Mater. Res., Vol. 17, No. 8, Aug 2002
© 2002 Materials Research Society
http://journals.cambridge.org
Downloaded: 16 Mar 2015
IP address: 111.68.111.42

H. Okuno et al.: Catalytic effects on carbon/carbon composites fabricated by a film boiling chemical vapor infiltration process
In a previous comparative study with small laboratory
reactors,9 in situ and ex situ experiments have been car-
ried out with various liquid carbon precursors to analyze
both the kinetics of deposition and the carbon yield with
the film boiling technique.
rod during the CVI experiment. As the densification pro-
ceeds, the hot volume of highest temperature extends
toward the external part, following the densification
front. This result corresponds well to that measured by
Bruneton et al.,2 with an inductive heating: this means
that the same principle governs both reactors.9
The ferrocene is diluted in cyclohexane in the range
0.1 to 5 wt%. The quantity of liquid mixture to com-
pletely immerse the carbon preform is poured into the
reactor. The central graphite rod is heated by Joule effect
at constant temperatures of 900, 1000, 1100, and
1200 °C, respectively.
In this study, we use the same experimental setup as
previously described. Mixed with cyclohexane as the liq-
uid carbon precursor, ferrocene (dicyclopentadienyliron:
C5H5–Fe–C5H5) is introduced as a source of iron to in-
vestigate the catalytic effect on the carbon formation. We
then present the physical and structural characteristics of
the deposited matrices under various conditions. All
polyaromatic carbons are multiscale materials. They are
described by different levels of organization. In order to
complete the microscale observation performed by opti-
cal microscopy and to be able to discuss about the cata-
lytic CVI mechanism, it is necessary to get a deeper
insight at the nanometer scale.10 In a detailed transmis-
sion electron microscopic (TEM) study, we will show
that under these catalytic conditions different nanopar-
ticles are also synthesized. This fact led us to propose a
chemical mechanism for this process, based on a moving
reaction front.
Because the temperature of the liquid phase could be
kept during the experiments at its boiling point (TB ס
80 °C for pure cyclohexane), the thermal gradient is
II. EXPERIMENTAL SETUP
In the boiling film technique, the heated porous sub-
strate or preform, is immersed directly into a liquid hy-
drocarbon as the source of carbon. The hydrocarbon is
vaporized by the calefaction effect and then cracked so as
to produce the carbon deposit at the contact with the
inner hot substrate. A mobile deposition front starts to
move from the core toward the preform surface. In this
method, a very strong thermal gradient, as opposed to the
gaseous conventional processes, is produced in the sys-
tem because the liquid phase surrounding the preform
keeps the outside temperature at its own boiling point.
A schematic illustration of the experimental setup is
showed in Fig. 1. The cylindrical carbon preform (⌽ ס
2 cm, L ס 4.5 cm) is made of ex-cellulose carbon felt
disks (RVC 2000 from Le Carbone-Lorraine Co., Cour-
bevoie, France9) of an apparent density of 0.1, stacked
around a central graphite rod (⌽ ס 3 mm) resistively
heated by Joule effect. This preform is placed vertically
in the axis of the reactor.
FIG. 1. Schematic experimental setup showing the axial temperature
profile.
The reference temperature of the sample is measured
in the axial channel of the heating rod with an S-type
thermocouple, fixed at the middle height of the preform
and protected against carbide formation. The axial ther-
mal gradient, measured by moving the thermocouple
along the axis of the sample,11 is lower than 5 °C/mm.
But the key thermal parameter is the radial gradient. As
shown in Fig. 2, the radial temperature distribution inside
the preform, measured once with a set of sleeved ther-
mocouples (placed at 1 and 3 mm, respectively, from the
graphite rod) is a function of the distance from the axial
FIG. 2. Radial temperature profiles in the sample for increasing times
in absence of catalyst.
J. Mater. Res., Vol. 17, No. 8, Aug 2002
1905
http://journals.cambridge.org
Downloaded: 16 Mar 2015
IP address: 111.68.111.42

H. Okuno et al.: Catalytic effects on carbon/carbon composites fabricated by a film boiling chemical vapor infiltration process
reaching about 300 °C/mm along a radius of the sample.
A deposition time of 30 min is kept constant for all the
experiments. A comparative study in the presence or not
of ferrocene in cyclohexane is carried out and presented
in the following.
The iron/carbon atomic ratio (Fe/C) in the deposited
matrix depends on the same ratio in the liquid phase as
shown in Fig. 4. This ratio is at least ten times larger in
the deposit than in the pristine solution. In addition, the
concentration of iron included in the deposits obtained at
1100 °C is about twice as much as at 1000 °C. It is thus
demonstrated that pyrolysis of ferrocene is faster than
cyclohexane leading to a depletion of catalyst in the mo-
bile phase.
III. EXPERIMENTAL RESULTS
As in previous studies, we examine the quantity of
deposited matrix, the deposition rate, and then the related
textural and structural features including graphitizability
and also the final iron content inside the matrix.
Indeed we must note that the transition metal whose
catalytic effect is studied here is not recovered but
trapped in the material it helps to produce and that
its vector (ferrocene) is destroyed, in contradiction with
a fundamental characteristic of regular catalysis (see
Sec. III.D).
A. Matrix formation
The variation of the mass of deposits versus the initial
ferrocene concentration in the liquid precursor is shown
in Fig. 3. A maximum rate has been obtained at 1000 and
1100 °C with 0.5 wt% ferrocene. It is interesting to note
that in absence of ferrocene there is no carbon deposit
under 1000 °C, whereas at 1200 °C the catalytic effect
does not appear to be so crucial. With an increased con-
centration of ferrocene, the mass of deposit tends to de-
crease slightly in the temperature range 1000 to 1100 °C.
The density of the composite is around 2.5 (obtained
from He-pycnometry using a Micromeritics AccuPyc-
1330, Norcross, GA), exceeding the usual values for car-
bon composites. This, combined with the magnetic
behaviour of the samples, shows off the composition of
the deposited matrix which must contain a certain
amount of iron-containing particles. To determine this
average concentration, composite samples have been
heated up to 600 °C for 2 h in an open furnace. The
quantity of iron is calculated from the resulting mass of
iron oxide (i.e., Fe2O3 in principle).
B. Deposition rate and densification front
The observation of a cross section of a composite
sample by optical microscopy (Fig. 5) shows what we
have called a densification profile,9 related to the evolu-
tion of the temperature distribution as presented in Fig. 2.
It is composed schematically of three concentric zones
from the center: (i) a (more or less) fully densified
zone, with a residual closed porosity; (ii) a densification
front with an open porosity; (iii) a nondensified zone,
working only as a thermal and mass exchanger, heating
up the precursor flow (calefaction phenomenon).
The two factors necessary to understand the experi-
mental densification profile, are respectively the deposi-
tion and the densification rates. The deposition is the
basic process that leads to densification of a porous ma-
trix; this deposition rate is measured by the increase of
the deposited matrix thickness around each individual
fiber of the preform.11
FIG. 3. Mass of deposited matrix as a function of the initial ferrocene
concentration in the precursor.
FIG. 4. Relation between the iron/carbon atomic ratios in the solution
and the deposit, at two deposition temperatures (TD).
1906
J. Mater. Res., Vol. 17, No. 8, Aug 2002
http://journals.cambridge.org
Downloaded: 16 Mar 2015
IP address: 111.68.111.42

H. Okuno et al.: Catalytic effects on carbon/carbon composites fabricated by a film boiling chemical vapor infiltration process
The densification rate is measured from the radial
thickness of sample that looks completely densified
across a section of the sample. Figure 5 shows how this
densification rate is determined by optically measuring
the thickness of the pyrocarbon deposited around the
single filaments. As a function of the initial ferrocene
concentration in the liquid precursor, it is largely in-
creased by the presence of catalyst, reaching a value of
about 10 mm/h at 1100 °C.
Qualitatively, the aspect of the composite (Fig. 5) is
very different from what we observe in the absence of
catalyst.9 Two points are thus evidenced: (i) With cata-
lyst, there is practically no residual porosity at a micronic
scale, and we can only calculate an equivalent deposit
thickness (curve A) from the average half distance be-
tween two carbon fibers. Without catalyst, we can actu-
ally draw the deposition profile (curve B) from the
individually measured deposit thickness. (ii) The depo-
sition front, with catalyst, is very sharp, indicating a com-
pletely different chemical mechanism.
two other thermocouples inserted at two different dis-
tances inside the preform, respectively 1 mm (curves B
and D) and 3 mm (curves C and E) from the central
heating rod, we measured the radial temperature with
catalyst (curves B and C) and without (curves D and E).
In the presence of catalyst, the temperature rise starts
earlier than in absence of catalyst for both thermo-
couples. Two reasons are to be considered to explain this
change of temperature profile by the catalyst: a modifi-
cation of the thermal conductivity of the matrix due to the
deposition chemical mechanism and the presence of iron
particles.
C. Microtexture observation by
optical microscopy
Polished surfaces of the samples are observed by op-
tical microscopy under polarized light between crossed
polarizers. An optical activity, due to the birefringence of
graphitic carbon domains, is relevant to the presence
of large (in the micronic range) oriented sp2 domains as
in SL or RL type microtextures usually found under both
isothermal and “Kalamazoo” conditions.9,12,13
The obtained microtextures are summarized in Table I
for deposition temperatures TD ס 900, 1000, and 1100 °C
and initial catalyst concentration between 0 and 5% in
weight.
The densification in the presence of catalyst thus ap-
pears to be faster and more complete. A change in tem-
perature profile and temperature evolution with time
is also observed, due to the presence of catalyst (Fig. 6).
To understand this, we measured with a thermocouple
the axial reference temperature (curve A), and then with
FIG. 5. Densification profile observed by optical micrography of a
cross section of the composite (TD ס 1000 °C; ferrocene in cyclohex-
ane ס 0.1 wt%): (A) associated densification profile in presence of
catalyst; (B) for comparison, without ferrocene.
FIG. 6. Temperature evolutions inside the sample (TD ס 1100 °C):
(A) reference axial temperature; (B) at 1 mm from the heating rod with
catalyst; (C) at 3 mm from the heating rod with catalyst; (D) at 1 mm
from the heating rod without catalyst; (E) at 3 mm from the heating rod
without catalyst.
TABLE I. Matrix microtextures under various experimental conditions (as seen by optical microscopy).
Ferrocene
Temperature
0%
0.1%
0.25%
0.5%
1%
5%
1100
1000
900
Anisotropic
No deposition
Isotropic and anisotropic
Isotropic
Isotropic and anisotropic
Isotropic
Isotropic and anisotropic
Isotropic
Isotropic
Isotropic
Isotropic
Isotropic
J. Mater. Res., Vol. 17, No. 8, Aug 2002
1907
http://journals.cambridge.org
Downloaded: 16 Mar 2015
IP address: 111.68.111.42

H. Okuno et al.: Catalytic effects on carbon/carbon composites fabricated by a film boiling chemical vapor infiltration process
In the presence of any amount of iron, the mostly
represented microtexture of the matrix looks completely
isotropic as defined by the absence of any bright area. It
is the only kind of matrix observed with ferrocene at 900
and 1000 °C and, in the case of initial ferrocene concen-
tration from 1% and beyond, at 1100 °C. At this high
temperature and for ferrocene concentration under 1%,
we find well-defined areas where anisotropic matrix is
deposited, looking like the RL microtexture normally ob-
served at 0%.9
In fact, at any temperature, deposition starts under
“catalytic conditions,” giving a matrix containing iron
particles with specific characteristics, particularly optical
isotropy. Then, as a consequence of the pyrolysis kinetics
of ferrocene, faster than that of cyclohexane, when tem-
perature is high enough (>1100 °C) and initial ferrocene
concentration low enough (<1%), there may be in some
restricted areas a lack of catalyst leading to carbon depo-
sition in “out of catalyst” conditions. Figure 7 shows
such an area where carbon fibers are covered first by an
isotropic matrix layer and then by an overlay of a clas-
sical anisotropic laminar pyrocarbon.
FIG. 8. Images of a composite sample (TD ס 1100 °C; ferrocene in
D. Iron concentration by electron probe
cyclohexane ס 0.5 wt%): (a) by optical microscopy; (b) by E.P.M.A.
of iron.
microanalysis (EPMA)
This method has been used to reveal the precise dis-
tribution of the iron included in each part of the compos-
ite samples. Giving at the same time for the same point
an image of the sample and the distribution of iron
(Fig. 8), this method is an accurate tool to determine the
relation between the microtexture and the concentra-
tion of iron.
Figure 9 shows this distribution inside the isotropic
phase obtained in the early deposition time at the contact
with the heating graphite rod at 1000 °C with an initial
FIG. 7. Optical micrographs of superimposed isotropic and aniso-
tropic microtextures (TD ס 1100 °C; ferrocene in cyclohexane ס
0.1 wt%).
FIG. 9. E.P.M.A. of iron in a sample (TD ס 1000 °C; ferrocene in
cyclohexane ס 0.1 wt%).
1908
J. Mater. Res., Vol. 17, No. 8, Aug 2002
http://journals.cambridge.org
Downloaded: 16 Mar 2015
IP address: 111.68.111.42

H. Okuno et al.: Catalytic effects on carbon/carbon composites fabricated by a film boiling chemical vapor infiltration process
ferrocene concentration of 0.1%. Iron particles appear
in white in this picture. A very high concentration of
these particles is detected at the boundary between the
deposit and the graphite heating rod, showing that depo-
sition begins with the pyrolysis of ferrocene on the hot
surface.
The presence of seed particles is reported to promote
carbon deposition8 by increasing the available density of
nucleation sites on the substrate. It helps carbon nuclea-
tion that cannot easily occur on the graphite substrate at
low temperature. It can be thought as one reason the
temperature close to the graphite heating rod increases
rapidly, as shown in Sec. III.A.
The latter corresponds to black parts in Fig. 10 with a
size distribution from 5 to 200 nm. The deposits present
different shapes and mainly three kinds of morphology
are observed; in all cases, the carbon growth occurred
from the particle surfaces. Figure 10(a) corresponds to a
bifilament, where the growth of carbon started from the
surface of a catalytic particle with a square shape, in two
opposite directions. Figure 10(b) shows a boundary be-
tween the particle and the initial deposit; it can be seen
that the aromatic layers are formed along the particle
surface so that graphenes are oriented parallel to the crys-
tal face. Figure 10(c) shows concentric multiwall shells
It also reveals that the concentration of iron included in
deposits depends on their structure. Figure 8 shows the
concentration of iron [Fig. 8(b)] in each texture revealed
by its optical microview [Fig. 8(a)]. It is confirmed that
iron is concentrated in the isotropic matrix areas and then
in a thin layer on the surface of the anisotropic pyrocar-
bon (probably as ferrocene adsorbed at low temperature
at the end of the experiment). Indeed iron concentration
inside the usual anisotropic pyrocarbon is very low.
E. Graphitization process
Graphitizability is a very interesting property for the
thermostructural applications of carbon/carbon compos-
ite materials. This is checked by x-ray diffraction analy-
sis of samples before and after a high temperature
treatment (HTT) for 90 min at 2500 °C. For example the
isotropic microtexture (TD ס 1100 °C with 5% of fer-
rocene) analyzed before HTT has no measurable value
for d002 but is 3.365 Å with a full width half-maximum
value of 1.15° after thermal treatment. This sharp 002
diffraction peak shows that this isotropic phase behaves
as a RL pyrocarbon microtexture which is the only
graphitizable one, i.e., evolving toward the thermody-
namically stable phase of hexagonal graphite.10 This
rather surprising result has incited us to carefully exam-
ine this “isotropic” matrix.
F. TEM observations
Since the pyrocarbon deposited by the “Kalamazoo”
method normally exhibits anisotropic microtexture,9,11
the catalytic effect of iron is to modify the matrix mi-
crotexture to an isotropic one. To reveal the real texture,
nanoscale observations have been carried out by TEM.
Two samples (TD ס 1100 °C with respectively 5 and
0.5% of ferrocene mixed with cyclohexane) are prepared
by gentle grinding then deposited on a lacey carbon grid.
Both were investigated using a Philips EM400 TEM
equipped with a high-resolution stage and operating at
120 kV (point resolution ס 3 Å).
FIG. 10. TEM bright-field images of various shapes found in the
sample composite (TD ס 1100 °C, ferrocene in cyclohexane ס
5 wt%): (a) carbon bifilament growing from a catalyst particle; (b)
carbon/catalyst particle boundary; (c) multiwall shells carbon growth
around round shaped catalyst particles; (d, e) carbon nanotube.
The morphology, texture, and size of the various py-
rocarbon deposits strongly depend on that of the catalytic
particles, as shown on the first sample (ferrocene 5%).
J. Mater. Res., Vol. 17, No. 8, Aug 2002
1909
http://journals.cambridge.org
Downloaded: 16 Mar 2015
IP address: 111.68.111.42

H. Okuno et al.: Catalytic effects on carbon/carbon composites fabricated by a film boiling chemical vapor infiltration process
deposited around the particles, especially when exhibit-
ing round morphology. In the latter case, tubes and fila-
ments are not formed since the growth does not follow
specific directions. Figure 10(d) exhibits a carbon nano-
tube, inside of which normally many small particles are
placed. The observed tubes have a diameter of 10 to
50 nm and a length around 1 ␮m. Figure 10(e) shows the
walls of such tubes constituted of long and straight par-
allel aromatic layers with a continuous length of more
than 100 nm.
On the second sample (ferrocene 0.5%), although the
densification remains uncomplete at low scale, with an
open porosity likely to induce a poor fiber/matrix adhe-
sion [Fig. 11(a)], a lower ferrocene amount gives rise to
a higher compactness at high scale [Fig. 11(b)]. Corre-
spondingly, nanofibrous morphologies such as that of
Figs. 10(a) and 10(d) are not evidenced. Local textures
are most often concentric, generally around “catalyst”
particles. Their porosity is rather low but increases lo-
cally when increased anisotropic display of graphenes
occurs [Fig. 11(c)]. It is worth noting that the bulk
textural aspect of the deposited matrix may somewhat
resemble that of a carbonized, heterogeneous petroleum
pitch like those used for the liquid impregnation of fi-
brous preforms and the subsequent preparation of
carbon/carbon composites.
Such features, good nanotexture and polyhedral graph-
ene stacks, are found for temperatures far beyond
2000 °C for heat-treated liquid carbon precursors without
catalyst.14 The structure is turbostratic (Fig. 12) although
diffraction rings (hk0) appear sharp, i.e., no hkl diffrac-
tion ring is found (d002 ꢀ 0.344 nm), and becomes com-
pletely graphitizable under further treatment as just
described (see Sec. III.E). As already shown,15,16 this is
related to the presence of catalyst inside the matrix.
From TEM observations, it has been revealed that a
variety of carbon deposits in the so-called “isotropic”
phase always starts from the surface of the catalytic par-
ticles. Although the temperature of 1000 °C is high
enough for the usual formation of carbon deposits in
catalyst-free conditions,9 no “isotropic” pyrocarbon has
been observed starting from the deposition on a carbon
fiber surface. The deposition mechanism starts from the
seed particles which are supposed to be transformed at a
lower temperature in cementite (Fe3C) by dissolution of
carbon within iron.17,18
To summarize, our experimental conditions have led
to the formation of different types of carbon nanopar-
ticles all mixed up. In particular, for a high content of
ferrocene, multiwall nanotubes are obtained inside this
confined volume following an already described catalytic
way for nanotubes formation.15,19
IV. PROCESS MECHANISMS
A. Discussion about the deposition mechanism
The catalytic effect is due to the iron introduced during
the formation of the matrix by decomposition of the fer-
rocene molecules. It allows much higher amount of de-
posit starting at a lower temperature (900 °C) compared
to classical CVI. This is confirmed by a kinetics analysis;
FIG. 11. TEM bright-field images of the sample (TD ס 1100 °C,
ferrocene in cyclohexane ס 0.5 wt%): (a) low scale of the deposited
matrix at the fiber contact; (b) concentric texture-rich area of the
matrix; (c) anisotropic textured area of the matrix.
FIG. 12. Electron diffraction pattern from the deposited matrix in
“catalytic” conditions. The most intense ring is 002, corresponding to
d ∼ 0.344 nm.
1910
J. Mater. Res., Vol. 17, No. 8, Aug 2002
http://journals.cambridge.org
Downloaded: 16 Mar 2015
IP address: 111.68.111.42

H. Okuno et al.: Catalytic effects on carbon/carbon composites fabricated by a film boiling chemical vapor infiltration process
from the Arrhenius plot, it is deduced that the densifica-
tion rate is practically exponential with the reaction tem-
perature. As shown in Table II, the apparent activation
energy (EA) values obtained in presence of catalyst are
around 70 kJ/mol, which is very low compared to the
value for the densification in the absence of catalyst as
determined by Bruneton et al.2 EA ס 222 kJ/mol. It
shows that the deposition mechanism in the presence of
catalyst is different with a faster kinetics constant which
overbalances the usual one. Indeed the formation of an
“isotropic” phase obtained by catalytic effect easily oc-
curs and is not strictly limited by lower temperature con-
ditions, as compared to the usual pyrocarbon deposition.
The presence of catalyst modifies the deposition
mechanism and therefore the carbon microtexture which
results in an improvement of the densification rate. Nor-
mally the pyrocarbon deposition occurs around each fi-
ber of the preform without catalyst.9 In the presence of
catalyst, the matrix deposition can be considered to occur
on the surface of seed particles in vapor phase as in
classical supported catalysis (Fig. 13), associated to a
modification of the temperature gradient.
(i) As we have seen from the observation of the ther-
mal gradient (Fig. 6), the catalytic effect is active in pro-
moting carbon deposition at a much lower temperature
(around 500 °C) than for classical CVD processes. We
can assume from our TEM experiments that, in a first
step, transition metal particles are formed and then, due
to the presence of a reductive atmosphere, cementite-like
grains may be formed from them during the pyrolysis of
the precursor mixture.21,22
In a second step, the growing of carbon nanoparticles
is occurring as described for example by Tibbets.23 From
our experiments we cannot describe the mechanism in
details, but as shown by our TEM and optical microscopy
observations of the matrix, we have found different types
of particles, like concentric shells and filaments, as al-
ready noticed by McAllister and Wolf24 in carbon/carbon
composites. Nevertheless all these composites are not
presenting high mechanical properties useful for struc-
tural uses, but due to their specific properties such as
graphitizability and ferromagnetism,24 they may be use-
ful in other applications as, for example, thermal
protections.
(ii) From the mechanism analysis we can infer the
increase of the deposition gates by a particle–vapor co-
deposition. Hurt and Allendorf20 have developed a CVD
model applied for deposition of silicon carbide which
leads to a growth rate enhancement characterized by the
definition of a volumic deposition rate (dV/dt) as
B. Model for the densification process
Our experimental results have shown that a catalytic
effect is present which allows an infiltration process at
lower deposit temperatures to be obtained with a weight
uptake multiplied by 5 at TD ס 1100 °C (Fig. 3) as dem-
onstrated by the lowering of the apparent activation en-
ergy EA (Table II). To explain this behavior we consider
both the mechanisms and kinetics of CCVI process by
taking into account a model for particle–vapor codepo-
sition.20 Its optimization, as evidenced Fig. 3, should be
related to the inlet of precursor gas inside the preform
and the codeposition mechanism.
dV/dt ס (dV/dt)CVD + (4/3 ␲ R3) F/G
,
(1)
where the first term is defined for the classical CVD
process characterized by a deposition rate G in a kinetics-
controlled regime in absence of catalyst. The second term
is related to seed particles which are supposed to be
spheres of radius R; they are deposited at the interface
with a rate F which characterizes a sticking coefficient of
products and a rather sharp reacting front. Indeed the
second term defines the enhancement factor related to
the fine particle deposition on the substrate that we sup-
pose here to occur following the described catalytic
mechanism. Indeed, this model accounts for seed par-
ticles which are injected inside a hot wall reactor, there-
fore corresponding to different chemical and thermal
conditions. This model of competing codeposition is
however useful to understand that through two chemical
ways with two different kinetics constants which work in
different parts of the reactor because of the thermal
gradient, a higher deposition rate is finally observed
(Fig. 13).
(iii) As we have already described,9 this fast densifi-
cation process of a porous preform is specific and char-
acterized by a competition of mass and heat transports in
the presence of a steep thermal gradient associated with
interfacial chemical reactions.
FIG. 13. Deposition model with catalytic seed particles and correlated
temperature profile.
J. Mater. Res., Vol. 17, No. 8, Aug 2002
1911
http://journals.cambridge.org
Downloaded: 16 Mar 2015
IP address: 111.68.111.42

H. Okuno et al.: Catalytic effects on carbon/carbon composites fabricated by a film boiling chemical vapor infiltration process
TABLE II. Apparent activation energies, EA, rate constants, k, and densification rate in function of the ferrocene concentration [Arrhenius law:
k ס A exp(−EA/RT)].
Densification rate (g/h)
Ferrocene (wt%)
EA (kJ/mol−1
)
k (103 g/h)
900 °C
1000 °C
1100 °C
1200 °C
0.0
0.5
1.0
2.0
2222
73.9
83.7
63.7
…
4.9
15.2
2.5
…
…
…
…
9.68
20.43
14.22
2.12
3.24
4.04
5.60
5.14
5.44
8.92
7.60
6.26
Even if a global modeling is out of the current ability,
two interesting points can be indicated. As we have
shown, the molecular diffusion mechanism is still effi-
cient in absence of catalyst.9 Using the Knudsen number
(Kn),25 which is characterized by the ratio between the
mean free path of involved molecules and the pore di-
ameters, two situations are recognized: (a) Kn < 2 is the
diffusion regime where the Fick coefficient is predomi-
nant as the usual situation for small molecules.9 (b) Kn >
1 is the ballistic regime where the Knudsen process is
important. The mean free path of molecules is inversely
proportional to the square of the molecular diameter.
A rough calculation shows immediately that the ob-
served catalytic nanoparticles are in a ballistic regime.
Indeed, after their chemical formation in the gas phase,
they stick directly to the surface and they form a straight
wall as already observed by optical microscopy (Fig. 5).
A complementary point concerns the coupling be-
tween the chemistry and the thermal gradient. As we
have observed, the thermal gradient is not so steep in the
presence of ferrocene (Fig. 7). This is related to the cata-
lytic effect which is starting around 500 °C; when the
nanoparticles are formed, the thermal exchange inside
the preform as well as the hot CVD region profile (i.e.
above 1000 °C) are dramatically lowered when the cata-
lytic effect is dominant. This is the reason, following the
Hurt and Allendorf’s model,20 we can rapidly change
from one regime to the other and, in consequence, from
isotropic to anisotropic microtexture.
We have also demonstrated that classical long-range
anisotropic pyrocarbons may be deposited around this
isotropic phase obtained for an early deposition time.
Thanks to the microprobe analysis we have clearly
shown that this morphological change is due to the pres-
ence and then the lack of iron-containing nanoparticles.
This observation is evidencing clearly that the nucleation
and growth mechanisms are modified in the presence of
iron with two major consequences.
(1) First, the type of matrix is clearly modified with the
presence of filaments and nanotubes of carbon as dem-
onstrated by TEM observations; this mechanism has
been explained qualitatively thanks to a particle–vapor
codeposition process. The obtained composites seem
however fragile with a smaller mechanical resistance
than classical ones. However it is known that submi-
cronic carbon filaments are useful fillers for composite
materials.26 The key point to progress in this direction
would be to control in situ the filament production for
reinforcing the matrix.
(2) Second, this catalytic mechanism appears very in-
teresting with the formation of particles in the gas phase
at a lower temperature which modifies the thermal gra-
dient and, therefore, their thermal history.
We infer that the transport properties of heat and mass
in porous media, which are two orders of magnitude
larger than in isothermal CVI process, are the keys to
understanding these fast densification processes.
These experiments show clearly that there is a cou-
pling between the chemical and thermal processes inside
the densification front: their accurate control may lead to
various and useful composite materials.
V. CONCLUSIONS
We have developed a bench reactor for studying a fast
densification process in the presence of the catalytic ef-
fect given by the addition of ferrocene in the liquid pre-
cursor. This effect has been evidenced with the matrix
deposited below 1000 °C and an increase of the deposi-
tion rate in the usual temperature range for making C/C
composites. Indeed we have shown that the addition of
0.5% in weight of ferrocene to the carbon precursor (cy-
clohexane) gives rise to an optimized mass uptake at TD
ס 1000 and 1100 °C. This catalytic process transforms
also the texture of the deposited matrix from long-range
to short-range anisotropy, appearing as isotropic at the
optical microscopy resolution level.
REFERENCES
1. I. Golecki, Mater. Sci. Eng. R20, 37, 124 (1997).
2. E. Bruneton, B. Narcy, and A. Oberlin, Carbon 35, 1593 (1997).
3. P. Delhae`s, in EURO-CVD 11, edited by M.D. Allendorf and
C. Bernard (Electrochemical Society Proceedings 97-25, Penning-
ton, NJ, 1997), pp. 486, 495.
4. R.E. Zielinski and D.T. Grow, Carbon 30, 925 (1992).
5. N.M. Rodriguez, A. Chambers, and R.T.K. Baker, Langmuir 11,
3862 (1995).
6. P.A. Tesner, in Chemistry and Physics of Carbon, edited by
P.A. Thrower (Marcel Dekker, New York, 1984), Vol. 19,
Chap. 2.
1912
J. Mater. Res., Vol. 17, No. 8, Aug 2002
http://journals.cambridge.org
Downloaded: 16 Mar 2015
IP address: 111.68.111.42

H. Okuno et al.: Catalytic effects on carbon/carbon composites fabricated by a film boiling chemical vapor infiltration process
7. S.D. Robertson, Carbon 8, 365, 374 (1970).
8. M.D. Allendorf, R.H. Hurt, and N. Yang, J. Mater. Res. 8, 651
(1993).
9. D. Rovillain, M. Trinquecoste, E. Bruneton, A. Derre´, P. David,
and P. Delhae`s, Carbon 39, 1355 (2001).
19. H. Gaucher, Ph.D. Thesis, Orle´ans University, Orle´ans, France
(1997).
20. R.H. Hurt and M.D. Allendorf, AIChE J. 37, 1485 (1991).
21. R. Sen, A. Govindaraj, and C.N.R. Rao, Chem. Phys. Lett. 267,
276 (1997).
10. J. Goma and A. Oberlin, Carbon 24, 135 (1986).
11. D. Rovillain, Ph.D. Thesis, Bordeau I University, Bordeaux,
France (1999).
12. H.O. Pierson and M.L. Liberman, Carbon 13, 159 (1975).
13. X. Bourrat, B. Trouvat, G. Limousin, G. Vignoles, and F. Doux,
J. Mater. Res. 15, 42, 101 (2000).
22. N. Grobert, W.K. Hsu, Y.Q. Zhu, J.P. Hare, H.W. Kroto,
B.R.M. Walton, M. Terrones, H. Terrones, P. Redlich, M. Ru¨hle,
R. Escudero, and F. Morales, Appl. Phys. Lett. 75, 3363, 3365
(1999).
23. G.G. Tibbetts, M.G. Devour, and E.J. Rodda, Carbon 25, 367
(1987).
14. M. Monthioux (unpublished results).
15. R.T. Baker, Carbon 27, 315 (1989).
16. K.S. Kim, N.S. Rodriguez, and R.T.K. Baker, J. Catal. 134, 253
(1992).
17. W.R. Ruston, M. Warzee, J. Hennaut, and J. Waty, Carbon, 7, 47
(1969).
24. P. McAllister and E.E. Wolf, Carbon 30, 189 (1992).
25. S.K. Griffiths and R.H. Nilson, in EURO-CVD 11, edited by
M.D. Allendorf and C. Bernard (Electrochemical Society Pro-
ceedings 97-25, Pennington, NJ, 1997), pp. 544, 551.
26. D.D.L. Chung, Carbon 39, 1119 (2001).
18. A. Oya and S. Otani, Carbon, 17, 131 (1979).
J. Mater. Res., Vol. 17, No. 8, Aug 2002
1913
http://journals.cambridge.org
Downloaded: 16 Mar 2015
IP address: 111.68.111.42