Você está na página 1de 12

Chemical Engineering Journal 175 (2011) 396407

Contents lists available at SciVerse ScienceDirect


Chemical Engineering Journal
j our nal homepage: www. el sevi er . com/ l ocat e/ cej
Kinetic study of double-walled carbon nanotube synthesis by catalytic chemical
vapour deposition over an Fe-Mo/MgO catalyst using methane as the carbon
source
Sigrid Douven
a,
, Sophie L. Pirard
a
, Georges Heyen
b
, Dominique Toye
a
, Jean-Paul Pirard
a
a
Laboratoire de Gnie Chimique, B6a, Universit de Lige, B-4000 Lige, Belgium
b
Laboratoire dAnalyse et de Synthse des Systmes Chimiques, B6a, Universit de Lige, B-4000 Lige, Belgium
a r t i c l e i n f o
Article history:
Received 20 June 2011
Received in revised form19 August 2011
Accepted 24 August 2011
Keywords:
Carbon nanotubes
CCVD process
Kinetic study
Methane
Mass transfer
Deactivation
a b s t r a c t
A kinetic study was performed to describe experimental reaction rates of double-walled carbon nanotube
(DWNT) synthesis by catalytic chemical vapour deposition (CCVD) over an Fe-Mo/MgO catalyst using
methane as the carbon source. Initial reaction rates were determined by mass spectrometry for methane
partial pressures ranging from 0.01 atm to 0.9 atm and for three temperatures: 900

C, 950

C and 1000

C.
Mass transfers from the bulk of the uid to the external surface of the catalytic bed and through the
catalytic bed were negligible as determined experimentally and conrmed by the methane mass balance.
A detailed kinetic study was carried out to discriminate between phenomenological kinetic models and
to validate the good agreement between the chosen model and the experimental data. The best model
was found to involve the irreversible dissociative adsorption of methane followed by the irreversible
decomposition of the adsorbed methyl group, which is the rate-determining centre. Activation energy
of adsorbed methyl decomposition was found to be equal to 58kJ mol
1
. The catalytic deactivation by
coking was expressed by a decreasing function of coke content, which leads to a sigmoid equation.
2011 Elsevier B.V. All rights reserved.
1. Introduction
Hollowcarbon bres have been observed for many decades [1]
but it was the ground-breaking report by Iijima [2] that made car-
bon nanotubes one of the most actively investigated materials. The
considerable interest in carbon nanotubes is due to their unique
properties and potential applications [3,4]. Double-walled carbon
nanotubes (DWNTs) consist of two concentric graphene sheets
rolled together in a nanometric cylinder. These are the simplest
example of multi-walled carbon nanotubes (MWNTs), in terms of
their structural aspect. Nevertheless, DWNTs have a thin diameter,
which confers to themproperties similar to those of single-walled
carbon nanotubes (SWNTs). So, the structural stability of MWNTs
[5,6] associated with the electronic properties of SWNTs make
DWNTs an interesting and promising material. Studies [79] have
shown that, in the future, it would be theoretically possible to
produce semiconductormetallic tubes or metallicsemiconductor
tubes; the latter represent a molecular wire covered by aninsulator
for use as connectors in nanoelectronic systems or as a molecular
capacitor in memory devices [10].

Corresponding author. Fax: +32 4366 3545.


E-mail address: S.Douven@ulg.ac.be (S. Douven).
Different methods have been used to produce carbon nan-
otubes: arc discharge [11], laser ablation [12], solar energy furnace
[13] and catalytic chemical vapour deposition (CCVD) [14]. The
CCVD method consists of the decomposition of a hydrocarbon
vapour ona catalyst into solidcarbonandgaseous hydrogenat high
temperature. The CCVD process appears to be the easiest method
to scale-up production capacity at a viable cost [15,16]. Indeed, the
CCVDprocess can operate continuously and offers the potential for
high yield production. Although other methods can produce high-
quality nanotubes, the yields are limited and these techniques are
not easily transferred to industrial production. Large-scale reactors
using the CCVD process to produce carbon nanotubes by met-
ric tons are already running, using either a uidized bed reactor
[1719] or an inclined mobile-bed rotating reactor [16,2022]. The
inclined mobile-bed rotating reactor seems to be one of the most
appropriate technologies because the kinetics of carbon nanotube
synthesis byhydrocarbondecompositionis quiteslow, andbecause
the ratio between the volume of the product and the volume of
the catalyst can be very large [16]. Successful development of a
large-scale CCVD reactor requires taking into account all the fac-
tors governingits performance: geometric, hydrodynamic, physical
and physico-chemical factors [20,23]. The present study deals with
the physico-chemical factor, involving the kinetics of reaction. The
ultimate aim is to use kinetic expression in the modelling of the
reactor, so as to be able to simulate reactor performance.
1385-8947/$ see front matter 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.cej.2011.08.066
S. Douven et al. / Chemical Engineering Journal 175 (2011) 396407 397
Nomenclature
a rst parameter of catalytic deactivation ()
b second parameter of catalytic deactivation ()
c third parameter of catalytic deactivation (s)
C dimensionless constant ()
C
b
methane concentration in the bulk of the uid
(kmol
CH
4
m
3
)
C
S
methaneconcentrationontheexternal catalytic bed
surface (kmol
CH
4
m
3
)
D
CH
4
He
binary diffusion coefcient (m
2
s
1
)
D
e
effective diffusivity (m
2
s
1
)
D
m
molecular diffusivity (m
2
s
1
)
d
b
thickness of the catalytic bed (m)
E
a1
activation energy of irreversible dissociative
adsorption of methane (kJ mol
1
)
E
a2
activation energy of irreversible decomposition of
adsorbed methyl group (kJ mol
1
)
F
0.95,n,
d
Fisher variable with
n
numerator and
d
denom-
inator degrees of freedomfor 95% condence ()
k kinetic rate constant (kmol
CH
4
kg
1
catalyst
s
1
atm
x
)
k
1
kinetic rate constant of irreversible dissociative ad-
sorption of methane (kmol
CH
4
kg
1
catalyst
s
1
atm
1
)
k
2
kinetic rate constant of irreversible decomposition
of adsorbed methyl group (kmol
CH
4
kg
1
catalyst
s
1
)
k
1ref
pre-exponential factor of kinetic rate constant k
1
(kmol
CH
4
kg
1
catalyst
s
1
atm
1
)
k
2ref
pre-exponential factor of kinetic rate constant k
2
(kmol
CH
4
kg
1
catalyst
s
1
)
k
M
mass transfer coefcient (ms
1
)
l characteristic length (m)
L length of the catalytic bed (m)
m
a
dimensionless mass of active catalyst ()
m
catalyst
mass of catalyst (kg)
M
CH
4
molecular weight of CH
4
(kgkmol
1
)
M
He
molecular weight of He (kgkmol
1
)
P
CH
4
methane partial pressure (atm)
P
He
heliumpartial pressure (atm)
P
t
total pressure (atm)
r specic reaction rate (kmol
CH
4
kg
1
catalyst
s
1
)
r
0
initial specic reaction rate (kmol
CH
4
kg
1
catalyst
s
1
)
r
0
initial specic reaction rate corresponding to the
theoretical model (kmol
CH
4
kg
1
catalyst
s
1
)
R perfect gas constant (kJ mol
1
K
1
)
Re Reynolds number ()
s
2
e
experimental variance ( kmol
2
CH
4
kg
2
catalyst
s
2
)
s
2
i
residual variance of Model i (kmol
2
CH
4
kg
2
catalyst
s
2
)
Sc Schmidt number ()
Sh Sherwood number ()
t time (s)
t
0
parameter of catalytic deactivation (s)
T temperature (K or

C)
T
ref
reference temperature (K or

C)
u owvelocity (ms
1
)
x methane partial order ()
x
CH
4
methane mole fraction (%)
x
H
2
hydrogen mole fraction (%)
x
He
heliummole fraction (%)
Y
ij
value of initial specic reaction rate of experiment j
at temperature i (kmol
CH
4
kg
1
catalyst
s
1
)
Y
ij
mean value of initial specic reaction rate of exper-
iment j at temperature i (kmol
CH
4
kg
1
catalyst
s
1
)

Y
ij
calculated value of initial specic reaction rate
by Model 2 of experiment j at temperature i
(kmol
CH
4
kg
1
catalyst
s
1
)
Z axial variable fromthe boat wall to the external sur-
face of the bed (m)
Greek symbols
relative loss of mass of the catalyst due to moisture
and solvent elimination at high temperature ()

2
(0.95,)

2
variable corresponding to degrees of freedom
for 95% condence ()
porosity of the catalytic bed ()
Weisz modulus ()
internal effectiveness factor ()

G
global effectiveness factor ()
kinematic viscosity (m
2
s
1
)

i
number of degrees of freedomof Model i ()

m
number of degrees of freedom of empirical power
lawmodel ()

e
number of degrees of freedom corresponding to
experimental data ()
density of the catalytic bed (kg
catalyst
m
3
)

CH
4
He
characteristic length (LennardJones potential) (
o
A)
tortuosity of the catalytic bed ()

CH
4
He
diffusion collision integral (LennardJones poten-
tial) ()
The CCVD process requires the use of an adequate catalyst for
the reaction. The number of walls, and the quality and purity of
carbon nanotubes will depend on the catalyst, carbon source and
operating variables such as temperature, ow rates and feedstock
partial pressures [24]. Metals used to catalyse carbon nanotube for-
mation are most often transition metals, in particular, iron [25,26],
nickel [27,28] and cobalt [29,30]. Mixtures of transition metals are
often more efcient than one metal alone. For example, papers
have reported carbon nanotube formation on iron-nickel- [31],
iron-cobalt- [31] andnickel-cobalt-based[32] catalysts. Someother
metals have also been used as co-catalysts. Even if not necessar-
ily active alone, these metals help to improve the performance
of the usual transition-metal-based catalysts. The most important
co-catalyst is molybdenum. The addition of molybdenum to iron
[33,34], cobalt [24,35] and nickel [36,37] catalysts can lead to the
formation of thin carbon nanotubes.
Alot of research has been done to understand the growth mech-
anismof carbon nanotubes [38,39] but very fewstudies have been
carried out in order to determine the sequence of elementary steps
involved in carbon nanotube synthesis. One kinetic study of MWNT
synthesis over an Fe-Co/Al
2
O
3
catalyst using ethylene as the car-
bon source assumed that the rate-determining step could be the
elimination of the rst hydrogen atom of adsorbed ethylene. The
activation energy was found to be equal to around 130kJ mol
1
[40]. A study of SWNT synthesis by ferrocene vapour decomposi-
tion in carbon monoxide reports an activation energy of 1.39eV,
i.e. 134kJ mol
1
[41], and considers that the limiting step is the
carbon diffusion in Fe particles. One study of the growth rate of ver-
tically aligned small-diameter carbon nanotubes is reported using
acetylene as the carbonsource andanFe/Al
2
O
3
catalyst. The activa-
tion energy was found to be equal to around 1eV, i.e. 96kJ mol
1
,
and the authors interpret this as a diffusion-limited growth rate
[42]. Another study of MWNT synthesis with acetylene mentioned
398 S. Douven et al. / Chemical Engineering Journal 175 (2011) 396407
Flow controllers
Furnace
Hot zone Cold zone
Mass spectrometer
Exhaust gases
He
CH4
H2
Removable cap Bypass valve
Bypass
Reactor
Fig. 1. Experimental set-up for kinetic rate measurements.
an activation energy of 26kJ mol
1
over an Fe/Al
2
O
3
catalyst and
66kJ mol
1
over a Ni/Al
2
O
3
catalyst [43]. A kinetic study of SWNT
production over Co-Mo/MgO catalysts using methane as the car-
bon source reported that the rate-determining step would be the
dissociation of methane [24]. The activation energy was found to
be equal to 160kJ mol
1
over Co-Mo/MgOcatalysts and96kJ mol
1
over Co/MgO catalysts. The decomposition of methane into carbon
species has also been investigated by researchers working in the
eld of petrochemicals because of its importance in catalyst deacti-
vation. Indeed, the formation of lamentous carbon can even cause
the catalyst support structure to disintegrate. Two kinetic studies
have reportedcarbonlament formationby methane cracking over
a Ni catalyst but they each assumed different underlying mecha-
nisms. The rst study assumed a reversible molecular adsorption
of methane followed by the abstraction of the rst hydrogen atom
from molecularly adsorbed methane as the rate-determining step
and evaluated the activation energy as being equal to 59kJ mol
1
[44]. The second study assumed that methane adsorbed dissocia-
tively before the stepwise dehydrogenation of carbon species [45].
In the present paper, a kinetic study is presented to describe
experimental initial reaction rates of DWNT synthesis by the CCVD
process over an Fe-Mo/MgO catalyst using methane as the carbon
source. The determination of initial specic reaction rates is made
possible by the use of a mass spectrometer placed at the exit of the
device. This experimental device is probably unique in its determi-
nation of experimental reaction rates. Several phenomenological
models are tted onto experimental data. A statistical study is per-
formed in order to discriminate between these kinetic models. As
themodels correspondtodifferent mechanisms, thediscrimination
allows a better understanding of the elementary steps involved in
carbon nanotube production. In the reaction rate expression, par-
tial pressures are equal to partial pressures in the uid if there
are no diffusional limitations, i.e. in a chemical regime. Indeed,
the catalytic bed is a porous medium, through which the reac-
tants must diffuse to react. Some mass transfer limitations could
occur depending on the catalyst structure and experimental condi-
tions (temperature, pressure etc.). The assumption of the chemical
regime has been checked experimentally and quantitatively by the
resolution of the mass balance of methane. A sigmoid equation is
tted onto experimental data to describe the catalyst deactivation
by coking. The nal aimof the present study is the determination of
the specic reaction rate of carbon nanotube synthesis in order to
use it inthe modelling of a continuous inclinedmobile-bedrotating
reactor.
2. Experimental
The experimental set-up for kinetic measurements is shown in
Fig. 1. The reactor is a 1.5-m-long quartz tube (diameter of 0.05m)
placed in a 1-m-long thermolyne tubular furnace. The part of the
tube jutting out fromthe furnace is a coldzone. The reactor is closed
off at its downstreamend by a removable cap. This removable cap
is equipped with a system allowing the introduction of a quartz
boat containing the catalyst, under reaction gas or under inert gas,
from the cold zone to the hot zone without opening the reactor.
The cold zone allows: (i) the direct introduction of the catalyst into
the hot zone under a reactive atmosphere and (ii) after reaction,
the cooling down of the products under an inert atmosphere. The
carbon source is methane and the carrier gas is helium. The com-
position of the feed gas (CH
4
, H
2
and He) and the total ow are
adjusted by three Bronkhorst mass owcontrollers. By operating a
bypass valve, the feed gas can be directed towards either the tubu-
lar reactor or the bypass. The composition of the exhaust gas is
determined by a Balzer mass spectrometer placed at the exit of the
device. The total pressure in the reactor is equal to atmospheric
pressure. The catalyst used is an Fe-Mo/MgO catalyst, as described
elsewhere [46,47]. This experimental set-up has been chosen in
order to measure the reaction rate in a geometric conguration
close to industrial conditions. Indeed, this systemis well designed
considering that the ratio between the volume of product and the
volume of catalyst is very large.
Prior to any measurement, the mass spectrometer is calibrated
by using a mixture of methane, hydrogenandheliuminknownpro-
portions. These proportions are adjusted by using the three mass
owcontrollers and owing through the bypass to the mass spec-
trometer. Each kinetic measurement follows this procedure: (i) the
reactor is vented withheliumuntil the exhaust gas contains at least
99.5% helium; (ii) the reactor is opened; (iii) the catalyst powder
(110
3
kg) is deposited in a length of 0.2m onto a quartz boat,
this quartz boat containing the catalyst powder is placed in the
cold zone, the reactor is closed and the venting continues until
the reactor contains at least 99.5% helium; (iv) the desired feed
ow composition, for a total ow rate of 3.7210
6
kmol s
1
, is
set and introduced into the reactor; (v) once the feed is stabilized,
thequartz boat is introducedintothemiddleof thehot zoneunder a
reactive atmosphere; simultaneously, the recording of the exhaust
gas compositionby the mass spectrometer begins; (vi) after 15min,
the reaction is nished and the reactor is vented with helium; (vii)
once the reactor contains 99.5% helium, the quartz boat is cooled
down under an inert atmosphere in the cold zone, before weighing
the nanotubes produced. Associated to each kinetic experiment, a
blank measurement is performed by introducing the desired feed
ow composition into the reactor without any catalyst in the hot
zone. The reactor can be considered to be of the differential plug
owtype [48].
In the present study, kinetic rates were determined on an
Fe-Mo/MgO catalyst at three temperatures (900

C, 950

C and
1000

C), and at each temperature the methane mole fraction was


varied between 1% and 90% balanced by helium.
3. Results
3.1. Use of mass spectrometer
Fig. 2 compares the exhaust gas composition during a blank
measurement and during a synthesis.
As can be seen in Fig. 2, the heliummole fraction decreases dur-
ing the reaction, which means that the ongoing reaction increases
S. Douven et al. / Chemical Engineering Journal 175 (2011) 396407 399
0
10
20
30
40
50
60
70
80
90
100
900 800 700 600 500 400 300 200 100 0
Time (s)
M
o
l
e

f
r
a
c
t
i
o
n

(
%
)
Helium
Hydrogen
Methane
Fig. 2. Mass spectrometer curves. Hollowsymbols refer to blank measurements and
full ones refer to reaction curves. For both curves: and correspond to helium,
and correspond to methane and, and correspond to hydrogen. Operating
variables: x
CH
4
= 30%, xH
2
= 0%, xHe =70% and T =950

C.
the total number of gaseous molecules. This qualitative observation
is in agreement with the stoichiometry of the chemical reaction
CH
4
C+2H
2
; the number of molecules of hydrogen produced is
twice the number of molecules of methane decomposed. Methane
consumption and hydrogen production can be assessed by the dif-
ference between the blank curve and the reaction curve. The initial
specic reaction rate r
0
is determined fromthose curves provided
by the mass spectrometer. Following the reaction stoichiometry,
the relationship between hydrogen production and methane con-
sumption is given by:
r
0
=
1
(1 )m
catalyst
d[CH
4
]
dt
=
1
2(1 )m
catalyst
d[H
2
]
dt
(1)
Therefore, the initial specic reaction rate, r
0
, can be calculated
from either the slope at the initial time of the hydrogen produc-
tion curve (Fig. 3) or the slope at the initial time of the methane
consumption curve. A comparison of the results conrms that ini-
tial specic reactionrates calculatedfromthe hydrogenproduction
curves are equal to those calculated from the methane consump-
tion curves, proving that the mass balance is respected. So, there is
no side reaction and no leak in the system. Hydrogen production
curves were used in this study. In Eq. (1), m
catalyst
is the initial mass
of catalyst and is the relative loss of mass of the catalyst due
to moisture and solvent elimination when the catalyst is placed
2 (1- )m
catalyst
r
0
0.1
0.2
900 800 700 600 500 400 300 200 100 0
Time (s)
1
0
3
x

H
y
d
r
o
g
e
n

p
r
o
d
u
c
t
i
o
n

(
k
m
o
l
)
0
Fig. 3. Example of hydrogen production curve as a function of time with the slope at
initial timecorrespondingtoinitial reactionrate, r
0
, for afeedgas compositionof 30%
CH
4
, 70% He at 950

C. The grey line and the black line correspond to experimental


and modelled hydrogen production curves respectively (Section 4.7).
at high temperature. As experimentally determined, is equal to
10.6%. The relative loss of mass was determined by weighing the
catalyst beforehand. Then the catalyst powder is placed onto the
quartz boat, introduced in the reactor at the reaction tempera-
ture under a heliumowfor 15min. The catalyst is then weighted
again. The mass difference is the loss of mass. In Fig. 3, the delay
of around 100s prior to the reaction beginning is the time neces-
sary for the gases to pass through the reactor and to reach the mass
spectrometer, placed at the exit (Fig. 1).
The rst part of this study focuses on the initial specic reac-
tion rate, r
0
. As illustrated in Fig. 3, a large part (0.0810
3
kmol)
of the total hydrogen production (0.1610
3
kmol) and there-
fore of carbon deposits refers to initial specic reaction rate, r
0
.
After approximately 250s, the slope of the hydrogen production
curve decreases whereas the operating conditions do not vary. This
decrease in the hydrogen production rate is due to catalytic deac-
tivation, which is studied in Section 4.7.
A transmission electron micrograph (TEM) image and a scan-
ning electron micrograph (SEM) image of a puried sample, i.e. a
sample whose catalyst is removed, in Fig. 4 showthat carbon nan-
otubes arethemainproduct of thereaction. Calculatedmasses from
spectrometer curves are in agreement with the weighted masses
of the nanotubes produced. It can be concluded that methane
is decomposed into carbon species such as carbon nanotubes,
graphite, amorphous carbon or carbon black without the formation
of gaseous or vaporizedliquidhydrocarbons heavier thanmethane.
3.2. Mass transfer
The catalytic bed is a porous medium where active sites are
homogeneously dispersed. So, reactants must diffuse fromthe bulk
of the gas to the external surface of the bed and then from the
external surface of the bed to the active sites, where the reaction
takes place. As a consequence, the observed reaction rate could
differ from the true reaction rate in chemical regime, i.e. without
mass transfer limitations. The global effectiveness factor,
G
, which
includes both external and internal resistances, is dened as the
ratio of the reaction rate with pore diffusion resistance and lm
resistance to the reaction rate with concentration in the bulk of the
uid [48].
3.2.1. External mass transfer
In order to check that there are no external transfer limitations,
the external mass transfer resistance can be estimated by [23]:
f
e
= 1
C
S
C
b
=
r
0
d
b
C
b
k
M
(2)
where C
b
is the methane concentrationinthe bulk of the uidandis
calculated by dividing the methane mole fraction by the tempera-
ture andthe perfect gas constant R, C
S
is the methane concentration
on the external catalytic bed surface, r
0
is the initial specic reac-
tion rate, is the density of the catalytic bed, d
b
is the thickness of
the catalytic bed and k
M
is the mass transfer coefcient. f
e
varies
from 1, for a mass transfer limited situation, to 0, for a reaction
limited situation.
The mass transfer coefcient k
M
can be correlated to the local
ow conditions and is typically obtained froma correlation of the
form:
Sh =
k
M
l
D
e
= C Re
1/2
Sc
1/3
(3)
where Sh is the Sherwood number, Re =ul/ is the Reynolds num-
ber, Sc =/D
e
is the Schmidt number, C is a dimensionless constant,
l is a characteristic length, D
e
is the effective diffusivity, u is the ow
velocity and is the kinematic viscosity.
400 S. Douven et al. / Chemical Engineering Journal 175 (2011) 396407
Fig. 4. (a) TEMimage; (b) SEMimage of a puried sample.
Unfortunately, the reactor geometry here is very complex and
unusual, so no suitable correlation can be found in the literature;
the reactive surface is very small compared to the total surface of
the tube and the ow of the uid in the reactor is perturbed by
the quartz boat and the growing nanotubes which, at the end of
the reaction, occupy a large part of the reactor cross-section. So the
mass transfer coefcient k
M
cannot be calculated a priori.
Therefore, adiagnostic test for external mass transfer limitations
has been carried out. In this experimental test, the initial specic
reaction rate r
0
is measured at constant space-time, but the ow
velocity is varied by adapting simultaneously the volume of the
catalytic bed and the total ow rate [49]. That way, the only oper-
ating variable that changes is the gas ow velocity, which directly
impacts the mass transfer coefcient, according to Eq. (3).
As the initial specic reaction rate, r
0
, is not affected by the vari-
ation of the total ow rate (Fig. 5), the conclusion is drawn that
there are no external transport limitations.
Furthermore, in an external mass transfer situation, i.e. when
the methane partial pressure at the external catalytic bed surface
is equal to zero, the kinetics can be written as:
r
0
=
k
M
P
CH
4
LRT
(4)
The methane partial order is equal to1andthe activationenergy
is equal to a fewJ mol
1
.
But in this study, the methane partial order is closed to 0 for
0.3 <P
CH
4
<0.9 and the activation energy is equal to 58kJ mol
1
which is a characteristic value for a chemical reaction (Section 4.5).
In conclusion, there is no external mass transfer resistance.
3.2.2. Internal mass transfer
In the absence of external diffusional limitations, the effect of
internal mass transfer on the reaction rate may be evaluated by
0
1
2
3
4
6 4 2 0
10
6
Total flow rate (kmol s
-1
)
1
0
4

r




(
k
m
o
l






k
g









s



)

C
H
4
c
a
t
a
l
y
s
t

-
1

-
1

0

Fig. 5. Initial specic reaction rate, r


0
, as a function of the total owrate for P
CH
4
=
0.6 atm, PH
2
= 0atm, PHe =0.4atmat 900

C.
measuring the dependence of the initial specic reaction rate, r
0
,
with the thickness of the catalytic bed, d
b
.
The experimental test consists of measuring the initial specic
reactionrate, r
0
, withvarious thicknesses of the catalytic bed, under
identical operating conditions. This means that the total ow rate
and the volume of catalyst are kept constant. With a constant mass
of catalyst, the variation of the bed thickness is obtained by varying
the length of the catalytic bed. In the plot of the observed initial
specic reaction rate r
0
vs. the catalyst bed thickness expressed in
terms of mass per unit length, constant values are expected for the
chemical regime as the total mass of catalyst does not vary. If the
observed initial specic reaction rate decreases when the catalytic
bed thickness increases, internal mass transfer limitations occur
[49].
At rst (Fig. 6), theinitial specic reactionrate, r
0
, is independent
of the mass of catalyst per unit length until around 510
3
kgm
1
and then it decreases as the mass per unit length increases. The
experiments used in this kinetic study were performed with a mass
per unit length of around 410
3
kgm
1
in order to avoid internal
diffusion problems.
Besides the experimental test, a quantitative approachwas used
to evaluate the internal diffusion, based on the calculation of the
Weisz modulus. The Weisz modulus compares the initial spe-
cic reaction rate to the diffusion ow of methane at the external
catalytic bed surface. If 1, methane diffusion is not the limiting
factor and the operating regime is a chemical regime. If 1, diffu-
sion limitations occur in the catalytic bed and the operating regime
is a diffusional regime. The Weisz modulus is dened by [48]:
=
r
0
d
2
b
D
e
C
S
(5)
0
1
2
3
4
10 8 6 4 2 0
10
3
m
catalyst
per unit length (kg m
-1
)
1
0
4

r




(
k
m
o
l






k
g









s



)

C
H
4
c
a
t
a
l
y
s
t

-
1

-
1

0

Fig. 6. Initial specic reaction rate, r
0
, as a function of the mass per unit length for
P
CH
4
= 0.6 atm, PH
2
= 0atm, PHe =0.4atmat 900

C.
S. Douven et al. / Chemical Engineering Journal 175 (2011) 396407 401
Table 1
Initial specic reaction rates r
0
and associated Weisz modulus with corresponding operating methane and heliumpartial pressures at 900

C, 950

C and 1000

C.
P
CH
4
(atm) PHe (atm) 10
4
r
0
(kmol
CH
4
kg
1
catalyst
s
1
) ()
900

C 950

C 1000

C 900

C 950

C 1000

C
1 0.01 0.99 0.27 0.28 0.28 0.32 0.31 0.31
2 0.03 0.97 0.74 0.80 0.81 0.29 0.30 0.30
31 0.05 0.95 1.14 1.22 1.24 0.27 0.28 0.28
32 0.05 0.95 1.42 1.39 1.56 0.33 0.32 0.35
41 0.1 0.9 1.72 2.18 2.34 0.20 0.25 0.26
42 0.1 0.9 1.76 0.21
51 0.2 0.8 2.52 3.12 3.39 0.15 0.18 0.19
52 0.2 0.8 2.39 3.18 0.14 0.18
61 0.3 0.7 3.26 3.63 4.15 0.13 0.14 0.15
62 0.3 0.7 4.23 0.16
71 0.4 0.6 3.49 4.10 4.82 0.10 0.12 0.13
72 0.4 0.6 4.82 0.13
81 0.5 0.5 3.30 4.26 5.18 0.08 0.10 0.11
82 0.5 0.5 3.30 0.08
91 0.6 0.4 3.17 4.27 5.41 0.06 0.08 0.10
92 0.6 0.4 3.22 0.06
10 0.7 0.3 3.16 4.01 5.55 0.05 0.07 0.09
11 0.8 0.2 3.09 3.98 5.07 0.05 0.06 0.07
121 0.9 0.1 2.78 3.76 4.94 0.04 0.05 0.06
122 0.9 0.1 2.82 0.04
123 0.9 0.1 3.28 0.04
124 0.9 0.1 3.06 0.04
125 0.9 0.1 2.98 0.04
where r
0
is the observed initial specic reaction rate, is the den-
sity of the catalytic bed (200kg
catalyst
m
3
), d
b
is the thickness of
the catalytic bed (110
3
m), C
S
is the methane concentration on
the external catalytic bed surface and is calculated by dividing the
methane mole fraction by the temperature and by the gas con-
stant R, and D
e
is the effective diffusivity through the catalytic bed
dened by:
D
e
=
D
m

(6)
where and are respectively, the porosity and tortuosity
of the catalytic bed. The porosity was estimated from mea-
surements by helium pycnometry, mercury porosimetry and
adsorptiondesorption isotherm and varied from 0.5 and 0.7 as a
function of the compression of the catalyst powder. Tortuosity has
been approximated by 1/ [50,51]. D
m
is the molecular diffusivity
and is calculated by Chapman and Enskogs relation [52]:
D
m
= D
CH
4
He
= 0.0018583
T
3/2
_
1/M
CH
4
+1/M
He
_
1/2
P
t
(
CH
4
He
)
2

CH
4
He
1.648 10
8
T
3/2
(7)
where D
CH
4
He
is the binary diffusion coefcient, T is the temper-
ature, M
CH
4
and M
He
are the molecular weights of CH
4
and He
respectively, and P
t
is the total pressure. The characteristic length,

CH
4
He
, and the diffusion collision integral,
CH
4
He
, are functions
of the LennardJones potential for the CH
4
He molecular system
and are determined fromtabulated values [52].
The Weisz modulus is calculated for each experiment and is
shown in Table 1. In the operating conditions (P
CH
4
= 0.6atm at
900

C) of Fig. 6, the value of the Weisz modulus is 0.06, which


conrms the absence of internal diffusion limitations determined
experimentally. The maximumvalue of the Weisz modulus, i.e. the
least favourable value, is equal to 0.35, which lets suspect some
possible internal diffusion limitations and so could introduce an
error in the estimation of the reaction rate.
4. Discussion
4.1. Empirical model
The simplest approach for expressing the rate of reaction is to
determine an empirical model corresponding to a power law:
r
0
= kP
x
CH
4
(8)
where r
0
is the initial specic reaction rate corresponding to the
model, P
CH
4
is the methane partial pressure, k is the kinetic rate
constant and x is the methane partial order of reaction. Parame-
ter estimation was performed for each temperature. Results of the
parameter estimation are presented in Table 2.
A statistical Fisher F-test, with 95% condence is performed
in order to validate the agreement between the empirical model
and the experimental data [53,54]. The value of Fisher variable
F
0.95,m,e
is compared to the ratio between the variance of the
model andtheexperimental variance. Thedegrees of freedomof the
model (
m
) for the Fisher F-test are equal to the number of experi-
ments (12 for each temperature) minus the number of parameters
(2). For the experimental data, the degrees of freedom (
e
) corre-
spond to the number of replicates minus the number of means. The
results of the F-tests lead us to conclude that the empirical model
is not able to reproduce the experimental data. For example, at
900

C,
m
=10,
e
=9, so F
0.95,m,e
= 3.1, while the ratio between
variances is equal to 14.9. As the ratio is higher than the Fisher vari-
able, the empirical model is not able to reproduce the experimental
data.
Table 2
Parameter estimation and associated standard deviations for the empirical power
lawmodel r
0
= kP
x
CH
4
.
Parameters 900

C 950

C 1000

C
10
4
k (kmol
CH
4
kg
1
catalyst
s
1
atm
x
) 3.450.15 4.800.30 6.300.30
x () 0.300.05 0.350.05 0.400.05
402 S. Douven et al. / Chemical Engineering Journal 175 (2011) 396407
4.2. Kinetic models
As asimpleempirical model is not suitabletoformulatethereac-
tion rate, the reaction mechanisms have to be studied in greater
detail. Several rate equations are derived fromvarious schemes of
nanotube synthesis. Infact, different sequences of elementarysteps
and rate-determining reactions can be envisaged to model experi-
mental data. The estimation of the reaction order towards methane
partial pressureallows arst discriminationbetweenthesemodels.
According to the results of the adjustment of the empirical power
law model in Section 4.1, the methane order lies between 0 and 1
andincreases withtemperature. The kinetic constant alsoincreases
with temperature.
Among the envisaged mechanisms, the following phenomeno-
logical models are retained; they are in agreement with the
previous conclusion, i.e. leading to a methane partial order of
between 0 and 1:
Model 1: CH
4
+s
1
+s
2
k
1
CH
3
s
1
+Hs
2
CH
3
s
1
+s
2
k
2
. . .
leading to the rate equation:
r
0
=
k
1
P
CH
4
1 +k
1
/k
2
P
CH
4
(9)
Model 2: CH
4
+2s
k
1
CH
3
s +Hs
CH
3
s +s
k
2
. . .
leading to the rate equation:
r
0
=
k
1
P
CH
4
_
1 +k
1
/k
2
P
CH
4
_
2
(10)
Model 3: CH
4
+2s
k
1
CH
3
s +Hs
CH
3
s +CH
3
s
k
2
. . .
leading to the rate equation:
r
0
=
k
1
P
CH
4
_
1 +
_
k
1
/k
2
_
0.5
P
0.5
CH
4
_
2
(11)
Models 1, 2 and 3 assume an irreversible dissociative adsorp-
tion of methane followed by the dehydrogenation of the adsorbed
methyl group, which is the rate-determining active centre, i.e. the
most abundant centre. They differ by the presence of one type (s) or
two types (s
1
and s
2
) of active sites and by the number of adsorbed
species involved in the kinetically signicant step.
Some phenomenological mechanisms leading to similar rate
equations can be identied. One can assume the reversible adsorp-
tion of hydrocarbon on the active site followed by the elimination
of the rst atomof hydrogen fromthe adsorbed hydrocarbon as the
rate-determiningstep. This typeof mechanismwas proposedbyLiu
et al. [55] using acetylene as the carbon source and by Pirard et al.
[40] using ethylene on iron-based catalysts. Associated rate equa-
tions are mathematically similar to the rate equations of Model 1
(Eq. (9)) and Model 2 (Eq. (10)) but the physico-chemical meaning
of the parameters is different [56]. Snoeck et al. [44] also proposed
this type of mechanismfor carbon lament formation by catalytic
methane cracking and justied their choice by the fact that exper-
iments were performed at pressures higher than 0.5MPa. Those
mechanisms wererejected, as themolecular adsorptionof methane
has a low probability of occurring compared to the dissociative
adsorption of methane. It is for this reason that Alstrup and Tavares
[45], studying carbon lament formation by methane cracking,
0.80
0.85
0.90
0.95
1.00
1 0.8 0.6 0.4 0.2 0
Z/d
b
e , CH
CH
P
P
= 0.14
= 0.18
= 0.25
= 0.28
= 0.30
= 0.31
= 0.05
= 0.06
= 0.07
= 0.08
= 0.10
= 0.12
Fig. 7. Proles of methanepartial pressure(dimensionless) fromthebottom(Z/d
b
=
0) to the top (Z/d
b
= 1) of the catalytic bed, for Model 2 at 950

C: P
CH
4
,e
= 0.01atm
(), 0.03atm (+), 0.05atm ( ), 0.1atm (), 0.2atm (), 0.3atm (), 0.4atm (),
0.5atm (), 0.6atm (), 0.7atm (), 0.8atm (), 0.9atm (). Weisz modulus is
indicated on the right.
assumedthedissociativereversibleadsorptionof methane; but this
mechanismleads to a methane partial order of reaction equal to 1
and not one of between 0 and 1.
One can also assume the irreversible adsorption of hydrocarbon
on the active site and the bulk diffusion of carbon to be the rate-
determining step. This mechanismwas proposed by Lee et al. [57]
in a study of carbon nanotube synthesis using acetylene as the car-
bon source over an iron-based catalyst. The rate equation obtained
is mathematically equivalent to the rate equation of Model 1 (Eq.
(9)). The results of parameter adjustment are the same and a sta-
tistical study cannot distinguish between these models, since only
the meaning of the parameters differs [56].
4.3. Proles of methane partial pressure
The mass balance of methane has been established in order to
determine the methane gradient within the catalytic bed and to
estimate the possible error. In fact, doubt remains over possible
internal diffusion limitations at the lowest methane partial pres-
sures, and neglecting the diffusion problem could lead to an error
inthe values of the calculatedkinetic parameters. The mass balance
is detailed in the Appendix.
As an example, Fig. 7 illustrates, for Model 2 at the intermediate
temperature 950

C, methane partial pressure proles through the


catalytic bed. The corresponding Weisz moduli (Table 1) are also
indicated. We should mention that the y-axis is enlarged to distin-
guish the different proles. Z is the axial variable on the thickness
of the catalytic bed. According to the y-axis, the proles of methane
partial pressure are rather at. However, the higher the Weisz mod-
ulus is, themorepronouncedtheprole. Inthis case, thechemical
reaction is faster than the diffusion and the concentration gradient
of methane is higher.
The internal effectiveness factor, , is dened as the ratio
between the experimental reaction rate and the reaction rate if all
the catalytic bed was at the same concentration, equal to the con-
centration at the surface of the catalytic bed, which is identical to
the concentration in the bulk of the uid as there are no external
diffusion problems.
Experimental values of the internal effectiveness factor, , are
somewhat higher than the characteristic curve of the rst-order
reaction represented by the solid line on Fig. 8. In fact, the reac-
tion order is between 0 and 1 (Section 4.1). The minimal value
of the internal effectiveness factor is 0.873 for a methane partial
pressure of 0.01atm. It becomes higher than 0.950 for a methane
partial pressure of 0.1atm. The kinetic parameters were calcu-
lated through the methane mass balance. These values were then
S. Douven et al. / Chemical Engineering Journal 175 (2011) 396407 403
0.1
1
10 1 0.1 0.01

Fig. 8. Internal effectiveness factor as a function of Weisz modulus at 900

C
(), 950

C () and 1000

C (). The solid line is the theoretical equation for a rst-


order reaction and a at plate catalyst. The dotted line is for a zero-order reaction.
Experimental values are represented by hollow symbols as a function of the Weisz
modulus given in Table 1.
compared to those obtained neglecting the internal diffusion limi-
tations.
4.4. Model discrimination
In order to discriminate between the three different rate equa-
tions (Eqs. (9)(11)) on the basis of experimental data, kinetic
parameters are estimated for each model at each temperature. A
maximumlikelihood formulation is adopted, thus minimizing the
weighed sumof squares of the differences between calculated and
measured reaction rates. Parameters are shown in Table 3 with
associated standard deviations: (i) neglecting the internal diffu-
sion limitations and (ii) taking into account the internal diffusion
limitations.
A statistical Fisher F-test, with 95% condence, allows for dis-
crimination between the kinetic models. The value of the Fisher
variable F
0.95,
1
,
2
is compared to the ratio between variances (s
2
i
)
of the different models. If the ratio of variances is higher than the
Fisher variable, then the models are statistically different. Other-
wise, they are not distinguishable. The degrees of freedom(
i
) for
the Fisher F-test are equal to the number of experiments (12 for
each temperature) minus the number of parameters (2 for each
model). F-tests are presented in Table 4 in both cases, i.e. with and
without diffusion limitations.
As canbeseeninTable4, at 900

Cand950

C, theratios between
variances of Models 1 or 3 andthe variance corresponding to Model
2are higher thanthe Fisher variable. Sodiscriminationbetweenthe
models is possible and it can be seen that Model 2 is the best one.
Table 3
Kinetic parameters of Model 1, Model 2 and Model 3: (i) neglecting the internal
diffusion limitations, (ii) taking into account the internal diffusion limitations.
T (

C) Model Without internal


diffusion limitations
With internal diffusion
limitations
10
4
k
1
10
4
k
2
10
4
k
1
10
4
k
2
900 1 425 3.50.1 466 3.50.1
2 261 13.10.1 281 13.10.1
3 14040 4.70.3 16146 4.60.3
950 1 415 4.80.2 466 4.80.2
2 301 16.50.1 321 16.60.1
3 105 30 7.10.7 12637 6.90.7
1000 1 373 6.60.3 404 6.50.3
2 281 21.00.3 301 21.00.3
3 8515 111 10120 101
Table 4
Statistical Fisher F-tests neglecting the internal diffusion limitations and taking into
account these limitations.
T (

C) Model s
2
i
/s
2
2
F
0.95,
i
,
2
Without internal
diffusion limitations
With internal
diffusion limitations
900 1 4.00 3.73 2.98
2
3 6.59 6.07 2.98
950 1 7.33 6.62 2.98
2
3 14.87 13.50 2.98
1000 1 2.09 1.98 2.98
2
3 3.92 3.62 2.98
At 1000

C, Models 3 and 2 are signicantly different but Models 1


and 2 are not signicantly different, since the ratio is smaller than
the Fisher variable.
As Model 2 is the only one to satisfy the Fisher test for all tem-
peratures, it canbe concludedthat the models canbe discriminated
and Model 2 is the one best able to reproduce the experimental
kinetic data.
4.5. Parameter estimation
In order to estimate kinetic parameters, parameter estimation
was performed for the best model (Model 2) with all the kinetic
data at all temperatures together. So, kinetic parameters have to be
expressed as a function of temperature following Arrhenius law:
k
1
= k
1ref
exp
_

E
a1
R
_
1
T

1
T
ref
__
, (12)
k
2
= k
2ref
exp
_

E
a2
R
_
1
T

1
T
ref
__
, (13)
where E
a1
and E
a2
are the activation energies of the rst and the
second elementary steps respectively; k
1ref
and k
2ref
are the cor-
responding pre-exponential factors. T is the temperature and T
ref
is a reference temperature xed to an intermediate value in the
temperature domain studied (927

C). Since the results without


considering internal diffusion limitations are almost identical to
those including internal diffusion limitations, the estimation of the
parameters was performed neglecting the diffusion phenomenon.
From parameter estimation, the activation energy E
a2
of irre-
versible decompositionof the adsorbedmethyl groupis foundto be
equal to 58kJ mol
1
, while the activation energy E
a1
of irreversible
dissociative adsorption of methane is not signicantly different
from zero (86kJ mol
1
). Model 2 can thereby be adjusted with
the parameter E
a1
being equal to zero so that only three parameters
have to be adjusted. Parameter estimation gives exactly the same
value of activation energy E
a2
as when the value of E
a1
is allowed
to vary. This conrms that the assumption of activation energy E
a1
equal to 0kJ mol
1
is relevant. A comparison of the experimen-
tal rates of carbon nanotube synthesis and the rates calculated by
Model 2 is presented in Fig. 9, while the parity diagramis shown in
Fig. 10.
The activation energy of irreversible decomposition of the
methyl group is equal to 582kJ mol
1
. As a comparison, in the
kinetic study of the carbon lament formation by methane crack-
ing on a Ni catalyst performed by Snoeck et al. [44], the activation
energy was found to be equal to 59kJ mol
1
, whereas Alstrup and
Tavares [45] foundtheactivationenergytobeequal to120kJ mol
1
.
However, as mentioned previously, these two kinetic studies dif-
fer in the sequence of elementary steps involved in the lament
synthesis. A kinetic study of SWNT synthesis over Mo-Co/MgO
404 S. Douven et al. / Chemical Engineering Journal 175 (2011) 396407
0
1
2
3
4
5
6
1 0.8 0.6 0.4 0.2 0
P (atm)
CH
4
1
0
4

r




(
k
m
o
l






k
g









s



)

C
H
4
c
a
t
a
l
y
s
t

-
1

-
1

0

Fig. 9. Initial specic reaction rate, r
0
, as a function of methane partial pressure.
Symbols correspond to experimental rates at 900

C (), 950

C () and 1000

C ().
Solid lines represent rates calculated by Model 2.
catalysts using methane as the carbon source reported that acti-
vation energy is equal to around 160kJ mol
1
for Mo-Co/MgO
catalysts and around 95kJ mol
1
for Co/MgO catalysts [24].
4.6. Statistical study
4.6.1. Experimental validation
A
2
test was carried out in order to validate Model 2 as a
function of temperature. This test allows for the consideration
that experimental variance can vary with temperature. The
2
test
compares the
2
variable value to the ratio between the sum of
the squares of residuals and experimental variance. Experimental
variance s
2
e
is estimated fromreplicated experiments at each tem-
perature. No temperature dependence is observed and s
2
e
is equal
to 0.02 10
8
kmol
2
CH
4
kg
2
catalyst
s
2
. The
2
variable is tabulated
as a function of the degrees of freedom. If the calculated value of
the ratio exceeds the tabulated
2
variable, the model is rejected
because of a lack of t.
The number of degrees of freedomis obtained by deducing the
number of parameters (3) from the number of experiments (36).
0
1
2
3
4
5
6
6 5 4 3 2 1 0
C
a
l
c
u
l
a
t
e
d

1
0
4
x

r


(
k
m
o
l





k
g














)

C
H
4
c
a
t
a
l
y
s
t
0
-
1
s


)
Experimental 10
4
x r (kmol kg s )
CH
4
-1
catalyst
-1
0

-
1
^
Fig. 10. Parity diagramfor Model 2.
-3
-2
-1
0
1
2
3
875 900 925 950 975 1000 1025
T (C)
R
e
s
i
d
u
a
l
s
Fig. 11. Residuals as a function of temperature for Model 2.
So the
2
variable is determined for 33 degrees of freedom and is
equal to 47.4 for 95% condence.

2
(0.95,33)
= 47.4 >

j
_
Y
ij


Y
ij
_
2
s
2
e
= 45.2 (14)
According to Eq. (14), the ratio is smaller than the
2
variable
and therefore Model 2 is validated.
4.6.2. Analysis of residuals
The
2
test establishes whether the overall t of the model is
satisfactory. Important discrepancies canstill exist andcanoftenbe
detected through the analysis of residuals, i.e. by examining the set
of deviations between the experimental and the predicted values.
If the model ts well, the residuals should be randomly distributed.
Systematic deviations fromrandomness indicate that the model is
not totally satisfactory.
Generally, experiments are plannedinorder to avoidsystematic
errors. Partial pressures were chosen randomly but each temper-
ature was considered consecutively because the time necessary
to stabilize a given temperature is very long. Therefore, an anal-
ysis of residuals is necessary to check that there is no systematic
error as a function of temperature. The residual for experiment j at
temperature i is dened as [53]:
Residual =
Y
ij


Y
ij
_
s
2
2
(15)
where Y
ij
is the observed reaction rate,

Y
ij
is the reaction rate cal-
culated by Model 2 and s
2
2
is the residual variance of Model 2.
Analyses of residuals were performed for Model 2 as a function
of temperature and methane partial pressure. Figs. 11 and 12 high-
light the randomdistributionof residuals according to temperature
and methane partial pressure, respectively.
-3
-2
-1
0
1
2
3
1 0.8 0.6 0.4 0.2 0
P (atm)
R
e
s
i
d
u
a
l
s
CH
4
Fig. 12. Residuals as a function of methane partial pressure for Model 2.
S. Douven et al. / Chemical Engineering Journal 175 (2011) 396407 405
4.7. Catalyst deactivation
In order to design a reactor and allowits scaling-up, the knowl-
edge of the initial specic reaction rate, r
0
, is not sufcient and it is
necessary to model the evolution of the specic reaction rate, r, as
a function of time: i.e. to model the catalyst deactivation.
As seen in Fig. 3, reaction rate decreases as a function of time.
Catalysts frequently lose a fraction of activity while in operation
[48]. There are many causes of deactivation. Among the mech-
anisms that might be responsible for this phenomenon are: (i)
poisoning, i.e. the strong chemisorptionof species oncatalytic sites,
thereby blocking sites for the catalytic reaction, (ii) covering of the
catalyst with carbon, which hinders the accessibility of the active
sites to the reactive species and (iii) thermal degradation leading
to the sintering of active sites or the support or to chemical trans-
formations of active phases to non-active phases of the catalyst
[48,58].
As deactivation may involve several phenomena whose ori-
gins are often not well-known, the determination of a deactivation
phenomenological equation is complex. Generally only empirical
equations are determined. In this study, after the burning of the
nanotubes produced, the catalyst is recovered and used for a new
synthesis in the same conditions. The activity of the pristine cata-
lyst is identical to the activity of the recovered catalyst. Therefore,
the deactivation is reversible and probably due to the formation of
amorphous carbon, in agreement with previous literature [59,60].
A typical expression of catalyst deactivation by coking is expressed
by a decreasing function of coke content:
m
a
=
1
1 +
1
C
P
(16)
where m
a
is the dimensionless mass of active catalyst and C
P
is the
concentration in amorphous carbon. Considering a rst-order law
for the deposit of amorphous carbon, the concentration C
P
is given
by:
C
P
=
2
exp(
3
(t t
0
)) (17)
The nal expression combining both equations corresponds to
a sigmoid catalytic deactivation equation:
r = r
0
m
a
(18)
m
a
= a btanh
_
t t
0
c
_
(19)
where r
0
is the initial specic reaction rate, r is the specic reac-
tion rate and m
a
is the dimensionless mass of active catalyst. In
the sigmoid equation of m
a
, t is the time, whereas a, b, c and t
0
are
parameters.
Parameter a can be expressed as a function of other parameters.
In fact, at the initial time, the dimensionless mass of active catalyst
is equal to the initial dimensionless mass of catalyst placed in the
quartz boat introduced in the reactor. Therefore,
m
a
= 1 +btanh
_
t
0
c
_
btanh
_
t t
0
c
_
(20)
So three parameters remain to be tted to experimental data
(Fig. 13): (i) parameter b corresponds to half of the difference
betweenthemaximumdimensionless mass of activecatalyst, at the
beginning of the reaction, and the minimum dimensionless mass
of active catalyst, at the end of the reaction; (ii) parameter t
0
is
the time corresponding to the inection point of the dimensionless
mass of active catalyst curve; (iii) parameter c represents the slope
of the dimensionless mass of active catalyst curve at time t
0
.
In order to determine b, c and t
0
, parameter estimation was
performed for each experiment previously used for the determi-
nation of initial reaction rates. The approach is identical whatever
the temperature considered. Results are qualitatively identical at
0
0.5
1
800 600 400 200 0
Time (s)
m
a
Fig. 13. Dimensionless mass of active catalyst curve as a function of time and
methane partial pressure at 950

C; P
CH
4
= 0.1atm (), 0.2atm (), 0.3atm (),
0.4atm(), 0.5atm(), 0.6atm(), 0.7atm(), 0.8 atm( ), 0.9atm().
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1 0.8 0.6 0.4 0.2 0
P (atm)
b
CH4
Fig. 14. Evolutionof parameter b as a functionof methane partial pressure at 950

C.
each temperature. The results at 950

C are reported here. The t-


ting between the experimental hydrogen production curve and the
modelled hydrogen production curve is excellent as illustrated in
Fig. 3. The associated sigmoid catalytic deactivation curve can be
seen in Fig. 13. It is important to note that the model is tted onto
data fromthe initial time of hydrogen production curve, i.e. around
100s.
Parameters b, c and t
0
of the sigmoid equation (Eq. (20)) have
to be determined as a function of methane partial pressure. Results
of the parameter estimation at 950

C are presented in Figs. 1416


respectively.
Parameter b is independent of methane partial pressure for a
given temperature (Fig. 14) and is equal to 0.46. It is important to
note that b is not exactly equal to 0.5 because the mass of active
catalyst is not equal to zero after a reaction time of 900s. This is
illustrated in Fig. 3, where hydrogen production keeps on increas-
ingslightly. Parameters t
0
andc decreaseslightlyas methanepartial
pressure increases (Figs. 15 and16); so the higher the methane par-
tial pressure, the faster the deactivation, as illustrated in Fig. 13,
0
100
200
300
1 0.8 0.6 0.4 0.2 0
P (atm)
0.32
CH
0
4
130

= P t
t


(
s
)
CH4
0
Fig. 15. Evolutionof parameter t
0
as afunctionof methanepartial pressureat 950

C.
406 S. Douven et al. / Chemical Engineering Journal 175 (2011) 396407
0
50
100
150
1 0.8 0.6 0.4 0.2 0
P (atm)
c

(
s
)
0.24
CH
4
79

= P c
CH4
Fig. 16. Evolution of parameter c as a function of methane partial pressure at 950

C.
Fig. A.1. Differential volume element of the catalytic bed.
which represents the evolution of the dimensionless mass of active
catalyst as a function of methane partial pressure. The rate of amor-
phous carbondeposit is proportional tothemethaneconcentration.
So the higher the methane partial pressure is, the higher the rate
of amorphous carbon deposit.
5. Conclusion
The kinetic study performed allows us to determine the equa-
tion of the initial specic reaction rate of DWNT synthesis by CCVD
over an Fe-Mo/MgO catalyst using methane as the carbon source,
as well as the equation of catalytic deactivation.
The methane mass balance on the catalytic bed allows us to
determine the proles of methane partial pressure throughthe cat-
alytic bed taking into account the internal diffusion. The internal
effectiveness factor has been calculated and the kinetic parameters
obtainedare inagreement withthose calculatedwhenthe methane
partial pressure is kept equal to that in the bulk of the uid.
From a statistical point of view, the kinetic equation best able
to reproduce the whole set of experimental data corresponds
to a mechanism involving the irreversible dissociative adsorp-
tion of methane followed by the irreversible decomposition of
the adsorbed methyl group as the signicant step. The activation
energy of methyl decompositionis foundto be equal to 58kJ mol
1
.
The catalyst deactivation by coking during the reaction can be cor-
rectly ttedby a sigmoidequationwhose parameters are a function
of methane partial pressure and temperature.
The initial specic reaction rate equation and associated cat-
alytic deactivation equation are the essential physico-chemical
data required to model the performance of the continuous inclined
mobile-bed rotating reactor and to allowits scaling-up.
Acknowledgements
S. Douven is grateful to the Rgion Wallonne Direction
Gnrale des Technologies, de la Recherche et de lnergie for
a Ph.D. grant and also for a Postdoctoral Researcher position in
the framework of the Programme de Formation et dImpulsion
la Recherche Scientique et Technologique (FIRST PINSYNAC
and FIRST AGATHA). S.L. Pirard is grateful to the National Fund
for Scientic Research, Belgium (F.R.S.-FNRS) for a Postdoctoral
Researcher position. The authors also thank the Belgian Fonds pour
la Recherche Fondamentale Collective (FRFC), the Ministre de la
Communaut Franc aise Direction de la Recherche Scientique,
the Fonds de Bay and the Interuniversity Attraction Pole (IAP 6/17
INANOMAT) for their nancial support. The involvement of the Uni-
versity of Lige in the Project SOMABAT of the European Union
seventh framework programis also acknowledged.
Appendix A.
Consider the catalytic bed as a at plate of thickness d
b
, in con-
tact with reactant on one side but sealed on the other side as shown
in Fig. A.1.
Themass balanceof methaneonthedifferential volumeelement
of the catalytic bed dZ is:
D
e
dC
CH
4
dZ
= r
0
dZ D
e
_
dC
CH
4
dZ
+
d
2
C
CH
4
dZ
2
dZ
_
(A.1)
Leading to:
D
e
d
2
C
CH
4
dZ
2
= r
0
(A.2)
The dimensionless expression is:
1
d
2
b
R
D
e
T
d
2
P
CH
4
dz
2
= r
0
(A.3)
In this equation, 1/d
2
b
R is a constant and D
e
/T is only function
of temperature, r
0
corresponds to the models presented in Section
4.2.
For each model, the corresponding equation is then:
1
d
2
b
R
D
e
T
d
2
P
CH
4
dz
2
=
k
1
P
CH
4
1 +k
1
/k
2
P
CH
4
for Model 1; (A.4)
1
d
2
b
R
D
e
T
d
2
P
CH
4
dz
2
=
k
1
P
CH
4
_
1 +k
1
/k
2
P
CH
4
_
2
for Model 2; (A.5)
1
d
2
b
R
D
e
T
d
2
P
CH
4
dz
2
=
k
1
P
CH
4
_
1 +
_
k
1
/k
2
_
0.5
P
0.5
CH
4
_
2
for Model 3; (A.6)
This is to be solved with the boundary conditions: (i) at z =0,
dP
CH
4
/dz = 0, i.e. the gradient of methane partial pressure is zero
at the wall of the quartz boat and (ii) at z =1, P
CH
4
= P
CH
4
,b
, that is,
the methane partial pressure at the external surface of the catalytic
bed is equal to the methane partial pressure in the bulk of the uid,
as there is no external diffusion limitation.
For initial values of k
1
and k
2
, the resolution of equations
(A.4)(A.6) leads to P
CH
4
and dP
CH
4
/dz proles within the catalytic
bed. The integration, on the thickness of the bed, of these equations
corresponds to the measured initial specic reaction rate r
0
. New
values of kinetic parameters are obtained. The iterative resolution
leads to the values of kinetic parameters k
1
and k
2
presented in
Table 3.
In these equations:
S. Douven et al. / Chemical Engineering Journal 175 (2011) 396407 407
C
CH
4
Methane concentration (kmol
CH
4
m
3
)
d
b
Catalytic bed thickness (m)
De Effective diffusivity (m
2
s
1
)
P
CH
4
Methane partial pressure (atm)
P
CH
4
,b
Methane partial pressure in the bulk of the uid (atm)
r
0
Initial specic reaction rate corresponding to the
theoretical model (kmol
CH
4
kg
1
catalyst
s
1
)
R Perfect gas constant (kJ mol
1
K
1
)
T Temperature (K)
z =Z/d
b
Dimensionless axial variable ()
Z Axial variable fromthe boat wall to the external surface of
the bed (m)
Density of the catalytic bed (kgm
3
)
References
[1] M. Monthioux, V.L. Kuznetsov, Carbon 44 (2006) 16211623.
[2] S. Iijima, Nature 354 (1991) 5658.
[3] N. Grobert, Mater. Today 10 (2007) 2835.
[4] M. Paradise, T. Goswami, Mater. Des. 28 (2007) 14771489.
[5] Y.A. Kim, H. Muramatsu, T. Hayashi, M. Endo, M. Terrones, M.S. Dresselhaus,
Chem. Phys. Lett. 398 (2004) 8792.
[6] H. Muramatsu, T. Hayashi, Y.A. Kim, D. Shimamoto, Y.J. Kim, K. Tantrakarn,
M. Endo, M. Terrones, M.S. Dresselhaus, Chem. Phys. Lett. 414 (2005)
444448.
[7] R. Saito, R. Matsuo, T. Kimura, G. Dresselhaus, M.S. Dresselhaus, Chem. Phys.
Lett. 348 (2001) 187193.
[8] A. Charlier, E. McRae, R. Heyd, M.F. Charlier, D. Moretti, Carbon 37 (1999)
17791783.
[9] A. Hashimoto, K. Suenaga, K. Urita, T. Shimada, T. Sugai, S. Bandow, H. Shinohara,
S. Iijima, Phys. Rev. Lett. 94 (2005) 045504.
[10] R. Pfeiffer, T. Pichler, Y.A. Kim, H. Kuzmany, in: A. Jorio, M.S. Dresselhaus, G.
Dresselhaus (Eds.), Carbon Nanotubes, Topics in Applied Physics 111, Springer,
Heidelberg, 2008, pp. 495530.
[11] J.L. Hutchison, N.A. Kiselev, E.P. Krinichnaya, A.V. Krestinin, R.O. Loutfy, A.P.
Morawsky, V.E. Muradyan, E.D. Obraztsova, J. Sloan, S.V. Terekhov, D.N.
Zakharov, Carbon 39 (2001) 761770.
[12] W.K. Maser, E. Mu noz, A.M. Benito, M.T. Martnez, G.F. de la Fuente, Y. Maniette,
E. Anglaret, J.-L. Sauvajol, Chem. Phys. Lett. 292 (1998) 587593.
[13] D. Laplaze, P. Bernier, W.K. Maser, G. Flamant, T. Guillard, A. Loiseau, Carbon 36
(1998) 685688.
[14] B.C. Liu, B. Yu, M.X. Zhang, Chem. Phys. Lett. 407 (2005) 232235.
[15] A.M. Thayer, Chem. Eng. News 85 (2007) 2935.
[16] J.-P. Pirard, Chem. Eng. News 86 (2008) 5.
[17] P. Serp, R. Feurer, C. Vahlas, P. Kalck, WO03002456 (2003), US 2004234445 A1.
[18] R. Philippe, P. Serp, Ph. Kalck, Y. Kihn, S. Bordre, D. Plee, P. Gaillard, D. Bernard,
B. Caussat, AIChE J. 55 (2009) 450464.
[19] R. Philippe, A. Moranc ais, M. Corrias, B. Caussat, Y. Kihn, Ph. Kalck, D. Plee, P.
Gaillard, D. Bernard, P. Serp, Chem. Vap. Deposition 13 (2007) 447457.
[20] S.L. Pirard, C. Bossuot, J.-P. Pirard, AIChE J. 55 (2009) 675686.
[21] S.L. Pirard, G. Lumay, N. Vandewalle, J.-P. Pirard, Chem. Eng. J. 146 (2009)
143147.
[22] J.-P. Pirard, C. Bossuot, P. Kreit, WO 2004069742 (2004), EP 1 594 802 B1.
[23] J. Villermaux, Gnie de la raction chimique Conception et fonctionnement
des racteurs, Lavoisier, Paris, 1993.
[24] L. Ni, K. Kuroda, L.-P. Zhou, T. Kizuka, K. Ohta, K. Matsuishi, J. Nakamura, Carbon
44 (2006) 22652272.
[25] Q. Li, H. Yan, J. Zhang, Z. Liu, Carbon 42 (2004) 829835.
[26] M.A. Ermakova, D.Y. Ermakov, A.L. Chuvilin, G.G. Kuvshinov, J. Catal. 201 (2001)
183197.
[27] P. Chen, H.-B. Zhang, G.-D. Lin, Q. Hong, K.R. Tsai, Carbon 35 (1997) 14951501.
[28] L. Piao, Y. Li, J. Chen, L. Chang, J.Y.S. Lin, Catal. Today 74 (2002) 145155.
[29] M. Terrones, N. Grobert, J.P. Zhang, H. Terrones, J. Olivares, W.K. Hsu, J.P.
Hare, A.K. Cheetham, H.W. Kroto, D.R.M. Walton, Chem. Phys. Lett. 285 (1998)
299305.
[30] K.Y. Tran, B. Heinrichs, J.-F. Colomer, J.-P. Pirard, S. Lambert, Appl. Catal. A 318
(2007) 6369.
[31] T.V. Reshetenko, L.B. Avdeeva, V.A. Ushakov, E.M. Moroz, A.N. Shmakov, V.V.
Kriventsov, D.I. Kochubey, Y.T. Pavlyukhin, A.L. Chuvilin, Z.R. Ismagilov, Appl.
Catal. A 270 (2004) 8799.
[32] H. Dai, A.G. Rinzler, P. Nikolaev, A. Thess, D.T. Colbert, R.E. Smalley, Chem. Phys.
Lett. 260 (1996) 471475.
[33] Y. Li, X. Zhang, L. Shen, J. Luo, X. Tao, F. Liu, G. Xu, Y. Wang, H.J. Geise, G. Van
Tendeloo, Chem. Phys. Lett. 398 (2004) 276282.
[34] J.H. Hafner, M.J. Bronikowski, B.R. Azamian, P. Nikolaev, A.G. Rinzler,
D.T. Colbert, K.A. Smith, R.E. Smalley, Chem. Phys. Lett. 296 (1998)
195202.
[35] S. Tang, Z. Zhong, Z. Xiong, L. Sun, L. Liu, J. Lin, Z.X. Shen, K.L. Tan, Chem. Phys.
Lett. 350 (2001) 1926.
[36] Y. Li, X.B. Zhang, X.Y. Tao, J.M. Xu, W.Z. Huang, J.H. Luo, Z.Q. Luo, T. Li, F. Liu, Y.
Bao, H.J. Geise, Carbon 43 (2005) 295301.
[37] L.P. Zhou, K. Ohta, K. Kuroda, N. Lei, K. Matsuishi, L. Gao, T. Matsumoto, J.
Nakamura, J. Phys. Chem. B 109 (2005) 44394447.
[38] A. Moisala, A.G. Nasibulin, E.I. Kauppinen, J. Phys. Condens. Matter 15 (2003)
S3011S3035.
[39] Y. Yao, L.K.L. Falk, R.E. Morjan, O.A. Nerushev, E.E.B. Campbell, Carbon 45 (2007)
20652071.
[40] S.L. Pirard, S. Douven, C. Bossuot, G. Heyen, J.-P. Pirard, Carbon 45 (2007)
11671175.
[41] A.S. Anisimov, A.G. Nasibulin, H. Jiang, P. Launois, J. Cambedouzou, S.D. Shan-
dakov, E.I. Kauppinen, Carbon 48 (2010) 380388.
[42] C.T. Wirth, C. Zhang, G. Zhong, S. Hofmann, J. Robertson, ACS Nano 3 (2009)
35603566.
[43] C.-T. Hsieh, Y.-T. Lin, W.-Y. Chen, J.-L. Wei, Powder Technol. 192 (2009) 1622.
[44] J.-W. Snoeck, G.F. Froment, M. Fowles, J. Catal. 169 (1997) 250262.
[45] I. Alstrup, M.T. Tavares, J. Catal. 135 (1992) 147155.
[46] D. Mhn, A. Fonseca, G. Bister, J.B. Nagy, Chem. Phys. Lett. 393 (2004) 378384.
[47] P. Dubey, S.K. Choi, J.H. Choi, D.H. Shin, C.J. Lee, J. Nanosci. Nanotechnol. 10
(2010) 39984006.
[48] G.F. Froment, K.B. Bischoff, Chemical Reactor Analysis and Design, Wiley, New
York, 1979.
[49] G.A. LHomme, Ind. Chim. Belge 35 (1970) 291307.
[50] C.N. Sattereld, Mass Transfer in Heterogeneous Catalysis, M.I.T. Press, Cam-
bridge, 1970.
[51] C. Gommes, S. Blacher, C. Bossuot, P. Marchot, J.B. Nagy, J.-P. Pirard, Carbon 42
(2004) 14731482.
[52] R.C. Reid, J.M. Prausnitz, B.E. Poling, The Properties of Gases and Liquids,
McGraw-Hill, NewYork, 1987.
[53] D.M. Himmelblau, Process Analysis by Statistical Methods, Wiley, New York,
1970.
[54] D.C. Montgomery, Design and Analysis of Experiments, Wiley, NewYork, 1997.
[55] K. Liu, K. Jiang, C. Feng, Z. Chen, S. Fan, Carbon 43 (2005) 28502856.
[56] S.L. Pirard, S. Douven, J.-P. Pirard, Carbon 45 (2007) 30503052.
[57] Y.T. Lee, J. Park, Y.S. Choi, H. Ryu, H.J. Lee, J. Phys. Chem. B106(2002) 76147618.
[58] C.H. Bartholomew, Appl. Catal. A 212 (2001) 1760.
[59] K. Kuwana, H. Endo, K. Saito, D. Qian, R. Andrews, E.A. Grulke, Carbon 43 (2005)
253260.
[60] S.L. Pirard, G. Heyen, J.-P. Pirard, Appl. Catal. A 382 (2010) 19.

Você também pode gostar