Carbon nanotubes by a CVD method. Part I: Synthesis and characterization of the (Mg, Fe)O catalysts

Article abstract of DOI:10.1021/jp026631s

The controlled synthesis of carbon nanotubes by chemical vapor deposition requires tailored and well-characterized catalyst materials. We attempted to synthesize Mg_1-xFe_xO oxide solutions by the combustion route, with the aim of performing a detailed investigation of the influence of the synthesis conditions (nitrate/ urea ratio and the iron content) on the valency and distribution of the iron ions and phases. Notably, characterization of the catalyst materials is performed using ⁵⁷Fe Moì?ssbauer spectroscopy, X-ray diffraction, and electron microscopy. Several iron species are detected including Fe²⁺ ions substituting for Mg²⁺ in the MgO lattice, Fe³⁺ ions dispersed in the octahedral sites of MgO, different clusters of Fe³⁺ ions, and MgFe₂O₄-like nanoparticles. The dispersion of these species and the microstructure of the oxides are discussed. Powders markedly different from one another that may serve as model systems for further study are identified. The formation of carbon nanotubes upon reduction in a H₂/CH₄ gas atmosphere of the selected powders is reported in a companion paper.

Full text of DOI:10.1021/jp026631s

13186
J. Phys. Chem. B 2002, 106, 13186-13198
Carbon Nanotubes by a CVD Method. Part I: Synthesis and Characterization of the (Mg,
Fe)O Catalysts
†
†
‡
†
Pierre Coquay, Eddy De Grave, Alain Peigney, Robert E. Vandenberghe, and
Christophe Laurent*,‡
NUMAT, Department of Subatomic and Radiation Physics, UniVersity of Ghent, Proeftuinstraat 86,
B-9000 Gent, Belgium, and CIRIMAT, UMR CNRS 5085/LCMIE, Centre InteruniVersitaire de Recherche et
d’Ing e´ nierie des Mat e´ riaux, UniVersit e´ Paul-Sabatier, 31062 Toulouse Cedex 4, France
ReceiVed: July 29, 2002
The controlled synthesis of carbon nanotubes by chemical vapor deposition requires tailored and well-
characterized catalyst materials. We attempted to synthesize Mg1-xFe O oxide solid solutions by the combustion
x
route, with the aim of performing a detailed investigation of the influence of the synthesis conditions (nitrate/
urea ratio and the iron content) on the valency and distribution of the iron ions and phases. Notably,
5
7
characterization of the catalyst materials is performed using Fe M o¨ ssbauer spectroscopy, X-ray diffraction,
2
+
2+
and electron microscopy. Several iron species are detected including Fe ions substituting for Mg in the
MgO lattice, Fe3 ions dispersed in the octahedral sites of MgO, different clusters of Fe ions, and MgFe
+
3+
O -
2 4
like nanoparticles. The dispersion of these species and the microstructure of the oxides are discussed. Powders
markedly different from one another that may serve as model systems for further study are identified. The
2 4
formation of carbon nanotubes upon reduction in a H /CH gas atmosphere of the selected powders is reported
in a companion paper.
Introduction
reduction in H2/CH4.2,3 However, undesirable Co3O4 particles,
leading to thick carbon nanofibers rather than to CNTs, may
Catalytic chemical vapor deposition (CCVD) is widely used
for the synthesis of carbon nanotubes (CNTs). This method,
similar to those used for several decades for the synthesis of
various filamentous forms of carbon, involves the catalytic
decomposition of hydrocarbons or carbon monoxide on transi-
tion-metal nanometric particles. Theoretical and experimental
also be formed during the synthesis of the oxide using the
16,17
combustion route.
This method involves the exothermic
redox reaction between a mixture of metal nitrates (oxidizer)
and urea (fuel). The so-called stoichiometric composition of the
redox mixture is usually calculated using the total oxidizing
and reducing valencies of the metal nitrates and urea, which
serve as numerical coefficients for the stoichiometric balance,
1
-4
results
show that CNTs are produced by catalyst particles
below ca. 3 nm in diameter. One of the key points is to tailor
the starting material so that the catalyst particles retain a small
size and indeed are active for CNT formation under given
experimental conditions (nature of carbon source, temperature,
etc). However, relatively few articles report detailed studies on
18
so that the equivalency ratio is equal to unity. However, these
calculations are made using several hypotheses. One is that the
oxidizing power of the oxygen in the reaction atmosphere (air)
is not taken into account. Another is that N2 is the only nitrogen-
containing product of the reaction so that zero is taken as the
valency for nitrogen. In fact, ammonia and nitrogen oxides are
also released, and thus the actual valency to be used is not
5
-7
the starting material.
We have proposed an original CCVD
8
method, showing that the reduction of R-Al1.9Fe0.1O3 in a H2/
CH4 atmosphere produces pristine Fe nanoparticles at a tem-
perature high enough for them to catalyze the decomposition
of CH4 and for the in-situ formation of CNTs including single-
walled CNTs (SWNTs). Characteristics of the starting alumina-
19
precisely known. Zhang and Stangle have proposed that there
is a range of ratios defining a stoichiometric region for which
the desired product may be prepared, whereas lower ratios are
too oxidizing and higher ratios, too reducing. It was indeed
9
based oxide such as the iron content, the crystallographic
20
shown that by varying the urea/nitrate ratio the formation of
1
0
11
form, and the specific surface area have been investigated
with the aim of increasing the proportion of single-walled CNTs
Co3O4 particles was avoided. However, more details on the
valency and the precise distribution of the cobalt ions in the
MgO-based powders were not known. Thus, we attempt in the
present study to synthesize Mg1-xFexO oxide solid solutions
(SWNTs). Using Mg1-xMxAl2O4-based solid solutions, the
composition of the starting oxide and thus the composition of
the in situ-formed catalytic particles were also investigated for
57
and to characterize them using Fe M o¨ ssbauer spectroscopy.
M ) Fe, Co, and Ni1 and M ) Fe/Co, Fe/Ni, and Co/Ni.
2,13
14,15
The aim of this work is to study the influence of both the urea/
nitrate ratio (also called the urea ratio) and the iron content on
the valency and distribution of the iron ions and phases. The
formation of CNTs by reduction in a H2/CH4 atmosphere of
selected specimens is reported in a companion paper.21
Mg1-xCoxO powders were also studied and are notably
interesting because a simple soaking in HCl allows us to separate
the CNTs from the CNT-Co-MgO powders formed upon
*
To whom correspondence should be addressed. E-mail: laurent@
chimie.ups-tlse.fr. Tel.: 00-33-(0)5-61-55-61-22. Fax: 00-33-(0)5-61-55-
Experimental Section
6
1-63.
†
‡
Synthesis. The appropriate amounts of magnesium and iron
nitrates were mixed with the desired quantity of urea (NH2-
University of Ghent.
Universit e´ Paul-Sabatier.
1
0.1021/jp026631s CCC: $22.00 © 2002 American Chemical Society
Published on Web 11/21/2002

Carbon Nanotubes by a CVD Method
J. Phys. Chem. B, Vol. 106, No. 51, 2002 13187
CO-NH2) to produce 2 g of a Mg1-xFexO solid solution (x )
influence on the other parameters. In most cases, good results
are obtained by considering a linear correlation between the
different distributions. The following linear correlation between
the isomer shift (δ) and/or quadrupole shift (2ꢀQ) and the
hyperfine field (Hhf) of a distributed sextet can be used: δk )
δ0 + Dδ(Hhf,max - Hhf,k) and 2ꢀQ,k ) 2ꢀQ,0 + Dꢀ(Hhf,max - Hhf,k)
where δ0 and 2ꢀQ,0 are the isomer-shift and quadrupole-shift
values correponding to Hhf,max. Similarly, in the case of a
distributed doublet, the following linear correlation between the
isomer shift (δ) and quadrupole splitting (∆EQ) can be used:
δk ) δ0 + Dδ(∆EQ,max - ∆EQ,k) where δ0 is the isomer-shift
value correponding to ∆EQ,max.
The oxide powders were observed with a JEOL JSM 6400
scanning electron microscope (SEM) and with a JEOL 2010
transmission electron microscope (TEM). A Tracor Voyager
energy-dispersive X-ray (EDX) analyzer (Ge detector) able to
work with a probe size as small as 4 nm was used for
quantitative analyses of very small areas.
0
.025; 0.05, 0.075; 0.10; 0.15; 0.20; 0.40). To increase the
substitution ratio of the iron cations in the MgO lattice, the urea
proportion was multiplied by 2 to 8 times the value calculated
for the so-called stoichiometric ratio. It was thus expected to
determine the stoichiometric region.19 The reagents were mixed
with 3.3 mL of water. The dissolution of the reagents is
endothermic and leads to an increase of the salt weights. The
mixture had to be continuously stirred and slightly warmed to
get a clear, homogeneous solution. The dissolution time
increased with increasing urea ratio, reaching 20 to 30 min for
a urea ratio of 8. After total dissolution, the dish containing the
solution had to be rapidly introduced into the furnace; otherwise,
a red precipitate formed at the bottom. With a urea ratio of 8,
the solution completely froze after a few seconds. However, in
all cases, rewarming the mixture led back to total dissolution.
The Pyrex dish containing the clear, homogeneous solution was
placed in a furnace preheated to 600 °C. The solution im-
mediately started to boil and dehydrate. Then, the combustion
process itself took place, as described in a later section. The
combustion product was slightly manually ground to produce
a powder.
Results and Discussion
Combustion Process. The nitrate-urea solution that was
placed in a furnace preheated to 600 °C immediately started to
boil and dehydrate. The decomposition of the reagents was
accompanied by a large release of gases (nitrogen oxides and
ammonia), increasing in proportion to the urea content. The
resulting paste then frothed and formed a foam that swelled in
proportion to the gas release. For a urea ratio of 1, a thin layer
was formed at the bottom of the dish, whereas for a urea ratio
of 4, the full dish was filled by the foam. For a ratio of 8, part
of the foam even expanded over the rim of the dish. Once the
gas release was complete, the foam blazed, and a white flame
appeared over the full dish. In most cases, the reaction was so
violent that the flame spread outside the oven, the door of which
was left open during the process. The violence of the reaction
increased with the urea ratio, reaching a maximum for urea ratios
between 3 and 5. The temperature of the oven then went up to
700 °C, and the local temperature at the flame was evaluated
to be 1200 °C. However, for higher ratios, the intensity of the
flame decreased, and the reaction was less violent. For a urea
ratio of 8, there was no blazing but just a glimmer appearing at
one point of the foam and slowly spreading over the foam. In
this case, the temperature of the oven did not exceed 600 °C.
The local flame temperature could not be evaluated, but it is
clear that the thermal conditions of the combustion process were
very different from those obtained with lower urea ratios. The
total combustion process was longer with increasing urea ratios,
from 5 min for low ratios to 10 min for a ratio of 8. This is
mainly due to the increasing gas release at the start of the
process.
For the sake of brevity, the Fe/MgO oxide powders will be
referred to according to the following example: Fe10U4 will
be used for a supposed Mg0.90Fe0.10O solid solution prepared
with a urea ratio of 4 (i.e., 4 times the so-called stoichiometric
ratio).
Characterization. Elemental analysis of carbon and nitrogen
was performed by flash combustion with an accuracy of (2%.
Specific surface-area measurements were performed using N2
adsorption at liquid N2 temperature in a Micromeritics Flow
Sorb II 2300 apparatus (BET method). This instrument gives a
specific surface-area value from one point (i.e., one adsorbate
pressure) and requires calibration. The reproducibility of the
results was determined to be in the (3% range.
X-ray diffraction (XRD) patterns were recorded with a
Siemens D501 diffractometer using Cu KR radiation and were
computer analyzed with the GUFI 5.0 program.2 For lattice-
parameter measurements, the powders were mixed with NaCl
as an internal standard, and the calculations were performed
with the UnitCell program.23 The crystallite sizes were evaluated
from the widths at half-maximum of the diffraction peaks using
the well-known Scherrer formula, with an accuracy on the order
of 20%.
2e
M o¨ ssbauer spectra (MS) were recorded with a 57Co (Rh)
source using a conventional time-mode spectrometer with a
constant-acceleration drive and a triangular reference signal.
Accumulation of the data was performed in 1024 channels until
6
a background of at least 10 counts per channel was reached.
The spectrometer was calibrated by collecting at room temper-
ature (RT) the MS of a standard R-Fe foil, and the isomer-shift
values quoted hereafter are with reference to this standard. The
measured absorbers were prepared with an amount of powder
corresponding to 10 mg of iron atoms per square centimeter.
Measurements between 14 K and RT were performed by
mounting the sample in a closed-cycle refrigerator. The spectra
were generally analyzed assuming symmetrical components with
Lorentzian line shapes. Asymmetrical M o¨ ssbauer patterns were
fitted with a model-independent hyperfine-field or quadrupole-
Up to a urea ratio of 3, the product that came out of the oven
appeared to be a dark-red crust stuck at the bottom of the dish.
For higher urea ratios, the product had a somewhat solid
consistency and could be taken out of the dish in one piece. Its
volume was comparable to the global volume obtained after
the gas release during the combustion process. All of the
products appeared to be homogeneous but with a color varying
from dark red for a urea ratio of 1 to greenish for a ratio of 4
and back to light red for a ratio of 8. It is interesting that red is
3
+
commonly associated with Fe ions and green usually corre-
24
2+
splitting distribution with Lorentzian-shaped elemental spectra.
sponds to Fe ions.
The quoted isomer-shift and quadrupole-shift values are in that
case average values for the distributed parameters. However,
the environment fluctuations of the M o¨ ssbauer atom responsible
of the distribution of a hyperfine parameter have generally some
For practical reasons, we also attempted to produce batches
of 6 g of oxide at a time, the size of the dish being fixed at 900
mL by the size of the oven. For a urea ratio of 1, no difference
was observed compared to the 2-g preparation, and the same

13188 J. Phys. Chem. B, Vol. 106, No. 51, 2002
Coquay et al.
TABLE 1: Specific Surface Area Sssaof the Fe/MgO Oxide
Powders Synthesized by Combustion
2
urea ratio
1
cat. % iron
oxide
S
ss (m /g)
5
Fe5U1
13.4 ( 3.1
22.1 ( 1.0
27.1 ( 4.4
30.6 ( 1.4
29.7 ( 2.7
25.0 ( 4.1
9.0 ( 0.3
5.3 ( 0.2
8.0 ( 0.3
5.6 ( 0.2
6.8 ( 0.8
5.7 ( 1.0
7.1 ( 0.2
5.0 ( 0.2
6.5 ( 0.2
19.2 ( 0.6
43.6 ( 1.3
43.5 ( 2.1
49.0 ( 4.5
50.1 ( 1.5
49.1 ( 8.9
7
0
5
0
10
10
10
5
0
5
0
5
0
10
10
10
5
.5
Fe7.5U1
Fe10U1
Fe15U1
Fe20U1
Fe10U2
Fe10U3
Fe10U3.5
Fe5U3.75
Fe10U3.75
Fe5U4
Fe10U4
Fe5U4.25
Fe10U4.25
Fe10U4.5
Fe10U5
Fe10U6
Fe5U8
1
1
2
2
3
b
3
3
.5
.75
b
b
1
1
1
4
4
Figure 1. Specific surface area of the Fe10 oxide powders versus the
urea ratio used for the combustion. The dashed line is a guide for the
eye.
b
b
.25
.5
4
5
6
8
b
b
the oxide. For the powders that were synthesized only once,
the Sss values are reported with an (underestimated) accuracy
of (3%. Figure 1 shows the Sss values for the Fe10 oxides
versus the urea ratio used in the preparation. Sss initially
decreases with increasing urea ratio and reaches a flat minimum
for urea ratios between 3.5 and 4.5. Subsequently, Sss increases
to a maximum for a urea ratio of 8. A similar evolution can be
extrapolated from the Sss values of the Fe5 oxides (Table 1).
Moreover, for a urea ratio of 1, an increase of the iron content
in the oxide gives rise to a proportional increase of Sss with
saturation for 15 cat. % iron. On the contrary, for urea ratios of
1
1
2
0
5
0
Fe10U8
Fe15U8
Fe20U8
a
The oven was preheated to 600 °C, and batches were 2 g each.
Average values are given for different combustion batches except for
those denoted by indicating powders that were synthesized only once.
b
The inaccuracy values are discussed in the text.
dark-red crust appeared at the bottom of the dish after the
combustion process. For a urea ratio of 4, the volume expansion
of the oxide did not increase considerably compared to the
volume of the 2-g preparation, but an inhomogeneous product
was obtained: green inside and red outside. This could reveal
3.75, 4, and 4.25, increasing the iron content in the oxide from
5 to 10 cat. % tends to decrease Sss slightly. For a urea ratio of
8, Sss increases from 5 to 10 cat. % iron and then remains
constant for higher iron contents, thus following the same trend
as in the case of oxides prepared with a urea ratio of 1 but with
a smaller influence of the iron content on Sss.
an inhomogeneous repartitioning of Fe2+ and Fe3+ ions in the
oxide. For a urea ratio of 8, most of the foam overflowed in
the oven, and the resulting product contained in the dish after
the combustion process was quantitatively and qualitatively
comparable to the preparation for 2 g of oxide.
The quantity of 2 g per batch was thus considered to be the
optimum choice for comparative studies with powders contain-
ing different iron contents (from 2.5 to 40 cat. %). Changing
the iron content had little influence on the combustion reaction
and on the aspect of the final product as described above. The
most noticeable difference is that higher iron contents yielded
a more pronounced color (either red or green) in the final
product.
Finally, chemical analysis indicates the presence of only traces
of carbon and nitrogen on the order of 0.1 wt % in most oxide
powders, regardless of the urea ratio used for the combustion.
The decomposition of urea can hence be considered to be
complete for all combustion processes, and the presence of
compounds involving carbon or nitrogen in the resulting oxide
powders can be ruled out.
Specific Surface Area. The specific surface area Sss of several
powders taken at different places in a same combustion batch
are within the instrumental accuracy of (3%, showing the good
homogeneity of the product. In most cases, however, the Sss
values of samples of the same oxide taken from different
combustion batches are outside this error bar. This is because
the combustion is a very fast and violent process that is not
fully controlled. When more than 2 g of an oxide was required
for further studies, batches containing powders with Sss values
within the error bar of (3% were mixed together. Sss indeed
appears to be the most sensitive parameter distinguishing
combustion batches for the same composition.
X-ray Diffraction. Figure 2a shows the XRD patterns of
Fe10U1, Fe10U4, and Fe10U8 where the five characteristic
peaks of MgO are well resolved. A minor phase corresponding
to MgFe2O4 is also detected. An enlargement of the area
corresponding to the main peak of MgFe2O4 (Figure 2b) reveals
an intensity minimum of this reflection line for Fe10U4 and a
strong increase of the intensity of the MgO(111) peak. Note
that the MgFe2O4 peak is more intense and narrow for Fe10U8.
One could propose that the presence of discrete MgFe2O4
particles of different sizes, but nevertheless relatively small, at
the surface of the MgO-based grains for Fe10U1 and Fe10U8
could account for the evolution of Sss as described above.
However, this hypothesis will be ruled out, as discussed later,
in light of SEM images.
For ideal Mg1-xFexO solid solutions, all iron should be in
2
+
the divalent state and should substitute for Mg . As the ionic
2
+
2+
radius of Fe is larger than the ionic radius of Mg , the unit-
cell parameter should increase with an increasing iron-substitu-
tion ratio following Vegard’s law. However, several authors2
have shown that Mg1-xFexO solid solutions present a positive
shift from ideality. Figure 3 shows the MgO unit-cell parameter
of the Fe10 oxides versus the urea ratio used for the combustion.
The unit-cell parameter increases with the urea ratio, reaches a
flat maximum for ratios between 3.5 and 4.5, and subsequently
decreases for higher ratios so that the unit-cell parameter is
equivalent for urea ratios of 1 and 8. Referring to Berthet,25 the
unit-cell parameters of Fe10U1 and Fe10U8 correspond to
poorly substituted MgO whereas the unit-cell parameter of
Fe10U4 corresponds to a substitution ratio of about 5 cat. %,
which is half of the desired value. This is in agreement with
the XRD patterns of Figure 2 showing an increased partitioning
of the MgFe2O4 phase for urea ratios of 1 and 8 compared to a
5-28
The Sss values (Table 1) are average values with an error
range covering all of the measured Sss values representative of

Carbon Nanotubes by a CVD Method
J. Phys. Chem. B, Vol. 106, No. 51, 2002 13189
Figure 2. XRD patterns of Fe10U1, Fe10U4, and Fe10U8 (a) and enlargement of the area (cf. dashed square) corresponding to the most intense
peak of MgFe (b).
2 4
O
of iron in solution. These Fe3 ions are associated with vacancies
in the MgO lattice, so more iron in solution corresponds to an
increase of the vacancy concentration, which favors the diffusion
processes during thermal treatments and gives rise to an increase
+
2
9,31,32
of the crystallite size.
For higher urea ratios, the reaction
temperature is lower, and the solubility of iron in MgO
decreases, leading to smaller crystallites. Moreover, the higher
expansion due to the gas release, notably for a urea ratio of 8,
should also lead to smaller MgO crystallites. Indeed, the
crystallite size seems to be smaller in Fe10U8 than in Fe10U1,
but this cannot be asserted because of the large error bars on
these values (Figure 4).
Figure 3. MgO unit-cell parameter a of the Fe10 oxide powders versus
the urea ratio used for the combustion. The dashed line is a guide for
the eye.
Finally, the average MgFe2O4 crystallite size could not be
determined because of the weak signals in the XRD patterns
(Figure 2). However, it appears that the MgFe2O4 peak is
qualitatively broader for Fe10U1 than for Fe10U8 (Figure 2b),
revealing smaller MgFe2O4 particles in the former specimen.
57
Fe M o1 ssbauer Spectroscopy. The literature reports numer-
ous M o¨ ssbauer studies of iron-substituted MgO prepared by
various methods. A relevant fact is that most of the resulting
2
+
3+
2+
oxide solid solutions involve Fe and Fe ions. The Fe ions
2
+
substitute for Mg on octahedral (Oh) sites and tend to form
28,31,33-37
3+
2+
clusters.
The Fe ions substitute for Mg on Oh sites
and are associated with vacancies. Clustering of Fe may
induce considerable deformation of the fcc lattice of MgO,
favoring the transfer of part of the Fe ions to tetrahedral (Td)
sites. These clusters can then be considered to be nuclei for
spinel formation since the oxygen arrangement is the same in
both structures. Many authors have indeed emphasized a narrow
association between Mg1-xFexO solid solutions and spinel-
3+
3
+
Figure 4. Average MgO crystallite size L of the Fe10 oxide powders
versus the urea ratio used for the combustion. The dashed line is a
guide for the eye.
3
4,38-40
urea ratio of 4. It thus appears that urea ratios between 3.5 and
structured MgFe2O4.
A prolonged annealing treatment
4.5 correspond to the nearest approximation of the desired
in air allows for the complete transformation of a Mg1-xFexO
solid solution in MgO with total phase partitioning of MgFe2O4.
However, it is difficult to characterize accurately the intermedi-
ate steps consisting of more or less agglomerated Fe3 ions in
a distorted undefined structure. Principally, the presence of large
clusters considered to be MgFe2O4-like particles can be
stoichiometry for the synthesis of Fe/MgO solid solutions
prepared according to the considered combustion process.
Figure 4 shows the average MgO crystallite size of the Fe10
oxides versus the urea ratio used for the combustion. Taking
into account the large error bars, a flat maximum can be
suggested for ratios between 3 and 4.5. As described in the
Experimental Section, the increase of the urea ratio from 1 to
+
3
+
confirmed by the detection of Td Fe . Moreover, the study of
superparamagnetic relaxation at low temperature can produce
3
+
4
corresponds to an increase of the reaction temperature, which
information on the degree of clustering of the Fe ions.
Considering the results of the previous sections, Fe10U1,
Fe10U4, and Fe10U8 were selected to be measured between
14 K and RT. The M o¨ ssbauer parameters are given in Tables
2, 3, and 4, respectively. Typical MS are presented in Figures
5, 6, and 7, respectively. For the measurements at RT, no
favors sintering of the material and gives rise to larger
crystallites. Moreover, the more violent reaction also leads to
an increase of the iron in solution in MgO, as seen in Figure 3.
2
9,30
Several authors
have shown that Mg1-xFexO solid solutions
contain some proportion of Fe3 that increases with the amount
+

13190 J. Phys. Chem. B, Vol. 106, No. 51, 2002
Coquay et al.
TABLE 2: M o1 ssbauer Parameters of Fe10U1 Measured between 14 K and RTa
super)para Fe3
+
(
2 4
magnetic MgFe O -like
c
d
T(K)
δ
∆E
Q
D
δ
Γ
P
δ
H
hf
2ꢀ
Q
D
δ
D
ꢀ
Γ
P
1
3
4
0
0.42b
0.42
0.48
0.77
1.07
0.63
7.5
0.41
0.41
508
500
0.01
0.01
-0.0009
-0.0014
-0.0017
-0.0013
0.49
0.44
92.5
74.5
25.5
33.5
38.0
50.5
65.0
80.5
1.43
5
7
1
1
1
0
0.42
0.69
1.20
0.65
1.10
0.68
1.25
0.67
1.35
0.57
0.95
1.55
0.49
0.90
1.80
0.42
0.85
1.55
0.37
0.80
1.30
0.0461
0.0246
0.0322
0.0435
0.0120
0.56
0.51
0.56
0.61
0.57
0.41
0.41
0.40
0.40b
0.40b
491
478
460
244
106
-0.02
0.01
-0.0006
-0.0001
0.0002
-0.0015
0.0004
0.0004
0.46
0.61
0.53
0.90
0.32
66.5
62.0
49.5
35.0
19.5
0
.39
0.42
.41
0.41
.39
0.40
.37
0.39
5
0
00
25
50
0.01
0
0.01b
0.01b
0
0
0
.38
.37
1
2
75
20
0.37
0.0108
0.0107
0.0160
0.53
0.50
0.41
88.0
0.40b
91
0.01b
0.25
12.0
0
0
.37
.36
0.34
100
0
0
.34
.33
RT
0.31
100
0
0
.30
.29
RT
bis
0.30
0.30
0.49
0.91
1.38
-
0.50
0.43
0.43
60.0
29.0
11.0
0
.27
a
(
Super)para: (super)paramagnetic; Hhf: hyperfine field at the maximum of the distribution (kOe); δ: (average) isomer shift (mm/s); ∆E
Q
:
quadrupole splitting at the maxima of the distribution (mm/s); 2ꢀ
Q
: (average) quadrupole shift (mm/s); D : linear correlation coefficient between
δ
the isomer shift and quadrupole splitting or hyperfine field; D : linear correlation coefficient between the quadrupole shift and hyperfine field; Γ:
ꢀ
Lorentzian line width (mm/s); P: proportion (%). Fixed parameter. Quadrupole-splitting distribution from 0.20 to 2.00 mm/s. d Hyperfine-field
b
c
distribution from 60 to 560 kOe.
TABLE 3: M o1 ssbauer Parameters of Fe10U4 Measured between 14 K and RTa
super)para Fe2
+
(super)para Fe3+
magnetic MgFe
2ꢀ
O
2 4
-like
(
c
d
Hhfe
T(K)
δ
∆E
Q
Γ
P
δ
∆E
Q
D
δ
Γ
P
δ
Q
D
δ
D
ꢀ
Γ
P
1
3
5
7
1
1
2
2
4
1.23
0.45
.31
0.45
.23
0.40
.16
0.41
.12
0.42
.15
0.35
.00
0.25
.70
0.22
.55
0.20
.48
0.75 30.5 0.47
0.44
-
0.97 13.5 0.41
492
0.02
0.00
-0.0003 -0.0005 0.46 56.0
-0.0002 -0.0001 0.40 40.5
-0.0000 -0.0010 0.24 21.0
-0.0010 -0.0004 0.27 14.0
1
0
1.17
1.18
1.15
1.14
1.10
1.08
1.06
1.04
0.50 38.5 0.46
0.47
0.54 41.5 0.46
0.47
0.59 42.0 0.46
0.47
0.52 40.0 0.46
0.47
0.46 36.5 0.45
0.46
0.45 37.0 0.42
0.44
0.48 38.5 0.37
0.41
0.27 38.0 0.31
0.40
0.37
1.00
0.45
1.00
0.40
0.95
0.46
0.95
0.53
1.00
0.45
0.93
0.48
1.02
0.46
1.10
-0.0223 0.45 21.0 0.41
-0.0322 0.61 37.5 0.42
-0.0101 0.49 44.0 0.41
489
1
0
488 -0.01
1
5
484
0.03
1
00
50
00
50
-0.0100 0.55 50.5 0.41b 463
-0.0067 0.62 58.0 0.40b 455
-0.0152 0.54 63.0
0.01b
0.01b
0.22
0.21
9.5
5.5
1
1
0
-0.0503 0.57 61.5
0
RT
-0.1385 0.53 62.0
0
RT
bis
1.04
0.20
0.48
0.34 37.0 0.32
0.41
0.49
1.12
-
0.57 36.0
0.67 27.0
a
(
Super)para: (super)paramagnetic; Hhf: hyperfine field at the maximum of the distribution (kOe); δ: (average) isomer shift (mm/s); ∆E
Q
:
quadrupole splitting at the maxima of the distribution (mm/s); 2ꢀ
Q
: (average) quadrupole shift (mm/s); D : linear correlation coefficient between
δ
the isomer shift and quadrupole splitting or hyperfine field; D : linear correlation coefficient between the quadrupole shift and hyperfine field; Γ:
ꢀ
Lorentzian line width (mm/s); P: proportion (%). Fixed parameter. Quadrupole-splitting distribution from 0.00 to 2.00 mm/s. d Quadrupole-
b
c
e
splitting distribution from 0.20 to 2.00 mm/s. Hyperfine-field distribution from 60 to 560 kOe.
in varying environments. However, no Fe2 phase is detected.
The spectrum was first fitted with a quadrupole-splitting
distribution where a linear correlation between the isomer shift
and quadrupole splitting allowed to refine the fit. Three maxima
are resolved in the distribution (Table 2 (RT) and Figure 5k).
Considering the parameters deduced from these maxima, a
second fit was then attempted with three independent doublets.
+
magnetic phases were detected, and all spectra have been
recorded on a small velocity scale ((4 mm/s), implying a higher
resolution. At lower temperature, however, most spectra showed
absorption lines at high velocities, and a higher scale ((11 mm/
s) had to be used.
Fe10U1 Oxide Powder. Fe10U1 at RT (Figure 5j) shows
broad lines, revealing several Fe3 (super)paramagnetic phases
+

Carbon Nanotubes by a CVD Method
J. Phys. Chem. B, Vol. 106, No. 51, 2002 13191
TABLE 4: M o1 ssbauer Parameters of Fe10U8 Measured between 14 K and RTa
super)para Fe3
+
(
2 4
magnetic MgFe O -like
c
d
T(K)
δ
∆E
Q
D
δ
Γ
P
δ
H
hf
2ꢀ
Q
D
δ
D
ꢀ
Γ
P
1
1
2
3
5
4
9
4
0
0
0.40
0.39
1.03
0.41
1.10
0.41
1.05
0.40
0.98
0.40
0.95
1.50
0.54
0.95
1.50
0.40
0.95
1.45
0.40
0.90
1.40
0.46
0.80
1.35
0.49
0.75
1.35
0.47
0.75
1.35
0.51
0.95
1.40
0.0194
0.0267
0.0253
0.0539
0.0322
0.56
14.0
0.39
0.42
0.42
430
509
511
-0.06
-0.02
0.01
-0.0004
-0.0001
-0.0001
-0.0005
-0.0010
-0.0004
-0.0001
-0.0012
-0.0015
-0.0007
0.53
86.0
0
.39
0.41
.40
0.41
.40
0.41
.39
0.41
0.66
0.61
0.55
0.54
25.0
46.5
60.0
73.0
0.47
0.47
0.52
0.60
75.0
53.5
40.0
27.0
0
0.42
0.42
0.42
509
506
503
-0.01
-0.03
-0.03
0
0
0
0
.39
.37
8
1
1
1
2
2
0
0.42
0.0744
0.0311
0.0301
0.0391
0.0481
0.0452
0.0687
0.52
0.52
0.52
0.50
0.49
0.50
0.39
75.0
79.0
80.0
80.0
81.0
81.5
0.41
0.42
0.42
0.39
0.35b
0.32b
491
489
487
479
455
423
-0.05
-0.04
-0.06
-0.07
0.01b
-0.0018
-0.0013
-0.0011
-0.0013
-0.0037
-0.0005
-0.0005
-0.0009
0.53
0.58
0.47
0.61
0.40
0.80
25.0
21.0
20.0
20.0
19.0
18.5
0
0
.39
.35
00
25
50
00
50
0.40
0
0
.38
.37
0.39
0
0
.38
.36
0.38
0
0
.37
.34
0.36
0
0
.34
.32
0.33
0.01b
0
0
.32
.29
RT
0.32
100
0
0
.29
.26
RT
bis
0.32
0.30
0.49
0.86
1.27
-
0.38
0.46
0.51
38.5
45.0
16.5
0
.25
a
(
Super)para: (super)paramagnetic; Hhf: hyperfine field at the maxima of the distribution (kOe); δ: (average) isomer shift (mm/s); ∆E
Q
: quadrupole
splitting at the maxima of the distribution (mm/s); 2ꢀ
Q
: (average) quadrupole shift (mm/s); D : linear correlation coefficient between the isomer
δ
shift and quadrupole splitting or hyperfine field; D : linear correlation coefficient between the quadrupole shift and hyperfine field; Γ: Lorentzian
ꢀ
line width (mm/s); P: proportion (%). Fixed parameter. Quadrupole-splitting distribution from 0.20 to 2.00 mm/s. d Hyperfine-field distribution
b
c
from 60 to 560 kOe.
The quality of the fit slightly decreases, but more-accurate results
are obtained for the M o¨ ssbauer parameters and for the propor-
tions of the three phases (Table 2 (RT bis) and Figure 9b). The
Indeed, MgFe2O4 is ferrimagnetic and produces two sextets
3
+
characteristic of Fe in the Oh and Td sites of the spinel
4
1
structure. The two sextets have slightly different M o¨ ssbauer
parameters in the temperature range considered here: Hhf(Oh)
- Hhf(Td) ) 20 to 30 kOe, δ(Oh) - δ(Td) ) 0.1 mm/s, and
2ꢀQ(Oh) ≈ 2ꢀQ(Td) ≈ 0 mm/s. The asymmetry in the pattern is
then due to the different isomer-shift values. However, the size
distribution and the varying composition of the MgFe O -like
3+
first doublet can be assigned to Fe ions in the Oh sites of
MgO or MgFe2O4 on the basis of the good agreement of its
M o¨ ssbauer parameters with the values reported by other
authors.3 The second doublet has a similar isomer shift, but
its quadrupole splitting is larger, suggesting a higher distortion
of the Oh sites or nearby structure that is probably due to
clustering of the concerned ions. Such a large quadrupole-
4,40
2
4
particles in Fe10U1 result in such a broad distribution of the
hyperfine field that subcomponents cannot be resolved. De
3
8
42
splitting value is also reported by Bhide and Tambe and by
Bakker was confronted with similar patterns obtained for
2
8
3+
Carles and Rousset for Fe ions in the Oh sites of MgO. The
chromium- and zinc-substituted MgFe O . He obtained satisfac-
2
4
third doublet has a smaller isomer shift and a larger quadrupole
splitting than the previous ones. It may arise from Fe ions in
tory fits with a hyperfine-field distribution including linear
correlation between the isomer shift and hyperfine field and
between the quadrupole shift and hyperfine field. On the basis
of the same procedure, the magnetic part of the present MS
was fitted with a hyperfine-field distribution with fields ranging
from 60 to 560 kOe, the line ratios being always fixed at 3:2:1
(Figure 5c, f, and i). The linear correlation allowed us to include
the asymmetry of the pattern in the fit. When the pattern lines
were too weak to produce consistent parameters, no correlation
was used, and the isomer-shift and quadrupole-shift values were
fixed at expected values. The average M o¨ ssbauer parameters
obtained at the maximum of the hyperfine-field distribution over
the full range of temperature (Table 2) are in good agreement
3
+
Td sites of MgFe2O4 precipitates in MgO, as suggested by Perez
3
4
40
et al. and Waychunas. This finding supports the presence
of MgFe2O4-like particles in Fe10U1, as concluded from the
XRD results.
The MS at low temperatures reveal that superparamagnetic
relaxation takes place below 220 K, where the central part of
the spectrum progressively decreases in favor of a broad six-
line pattern (Table 2 and Figure 5a, d, and g). At 14 K, the
pattern shows an asymmetry typical of a MgFe2O4-like phase:
in particular, the first, second, and third lines are deeper and
narrower than the fourth, fifth, and sixth lines (Figure 5a).

13192 J. Phys. Chem. B, Vol. 106, No. 51, 2002
Coquay et al.
Figure 5. MS of Fe10U1 measured between 14 K and RT (a, d, g, j) and corresponding quadrupole-splitting distributions of the (super)paramagnetic
3
+
2 4
Fe phases (b, e, h, k) and hyperfine-field distributions of the magnetic MgFe O -like phase (c, f, i).
with those expected for MgFe2O4-like particles undergoing
superparamagnetic relaxation.42 We acknowledge that this fitting
procedure is unrealistic, notably because of the non-null value
of the quadrupole shift, but it is nevertheless an adequate fitting
model that allows us to compare different samples.
in distorted Oh sites. The magnetic splitting at low temperature
indeed suggests that these ions form small superparamagnetic
3
+
clusters. The proportion of Fe that remains at 14 K is
essentially characteristic of isolated paramagnetic Fe3 ions
substituted in the MgO lattice.
+
Regarding the (super)paramagnetic Fe3 phases at low
temperatures, they were fitted with the same quadrupole-splitting
distribution as used at RT, giving rise to consistent parameter
values. Figure 8 shows the regular decrease of the (super)-
+
Fe10U4 Oxide Powder. The same fitting procedures were
considered for the MS of this sample (Table 3 and Figure 6).
3
+
At RT, the Fe quadrupole-splitting distribution exhibits two
maxima (Table 3 (RT) and Figure 6i and k), and a proper fit
could also be obtained with two independent doublets (Table 3
(RT bis) and Figure 9d). These two components were attributed
3
+
paramagnetic Fe proportion in Fe10U1 with decreasing
temperature in favor of the magnetic phase. From 125 K
downward, the third maximum in the distribution does not
appear anymore (Table 2 and Figure 5e and h). Moreover, the
proportion of the second maximum decreases faster than the
proportion of the first one, which appears alone in the distribu-
tion at 14 K (Table 2 and Figure 5b). The third maximum was
3
+
to Fe in the Oh sites of MgO (or possibly of MgFe2O4 for the
first one), the larger quadrupole splitting of the second one
3
+
suggesting distorted sites due to Fe clustering and thus
potential MgFe2O4-like particles. However, no Td Fe3 was
detected. Moreover, a second quadrupole-splitting distribution
had to be used to fit the spectrum properly (Figure 6i and j).
+
3
+
attributed to Fe ions in Td sites of MgFe2O4 and reflects the
presence of large Fe3 clusters. It is therefore consistent that
these clusters are the first to split magnetically with decreasing
+
2+
This extra pattern is typical of Fe ions, as revealed by its
isomer-shift value (Table 3). Two maxima were detected in the
distribution (Figure 6j).
3
+
temperature. The second maximum was attributed to Fe ions

Carbon Nanotubes by a CVD Method
J. Phys. Chem. B, Vol. 106, No. 51, 2002 13193
Figure 6. MS of Fe10U4 measured between 14 K and RT (a, e, i) and corresponding quadrupole-splitting distributions of the (super)paramagnetic
2
+
3+
2 4
Fe phases (b, f, j) and (super)paramagnetic Fe phases (c, g, k) and hyperfine-field distributions of the magnetic MgFe O -like phase (d, h).
Upon decreasing the temperature, the Fe3+ phases progres-
sively split into a magnetic pattern comparable to the one
described above (Figure 6a, d, e, and h). Figure 8 shows the
reported by other authors.33,35 Consequently, the hyperfine-field
distribution at 14 K (Figure 6d) includes the contribution of
the Fe clusters, which have become magnetic at that low
temperature.
2
+
(
super)paramagnetic Fe3+ proportion versus the measuring
temperature, with values normalized to 100% at RT. The
Fe10U8 Oxide Powder. The Fe10U8 MS at RT was fitted
with a quadrupole-splitting distribution characteristic of Fe
decrease of the (super)paramagnetic Fe3+ proportion starts at
3+
2
00 K and shows a more gentle slope than for Fe10U1.
phases (Table 4 (RT) and Figure 7j and k). Using a linear
correlation between the isomer shift and quadrupole splitting,
three maxima are detected in the distribution profile (Figure
7k). An adequate fit was also obtained with three independent
doublets (Table 4 (RT bis) and Figure 9f), giving more-accurate
information on these three phases that can be attributed, as in
3
+
Magnetic splitting at lower temperatures suggests smaller Fe
clusters in Fe10U4 compared to those in Fe10U1, which is
3+
compatible with the apparent absence of Td Fe . At 14 K, the
second maximum of the Fe3 quadrupole-splitting distribution
has completely disappeared, revealing the proportion of isolated
paramagnetic Fe ions substituted in the MgO lattice (Table 3
and Figure 6c). This proportion is comparatively higher in
Fe10U4 than in Fe10U1, as shown in Figure 8.
The two maxima of the Fe2 quadrupole-splitting distribution
are increasingly more separated with decreasing temperature,
the second one showing a drastic increase in quadrupole splitting
on lowering the temperature (Table 3 and Figure 6b, f, and j).
The quadrupole splitting of Fe2 ions in crystalline fields of
low symmetry is very sensitive to temperature variations, and
+
3
+
3+
the case of Fe10U1, to O Fe ions in MgO or MgFe O , O
h
h
2
4
3+
3+
Fe ions that form clusters in MgO, and T Fe ions revealing
d
2
+
the presence of MgFe O -like particles. No Fe phase is
2
4
+
detected.
At low temperature, superparamagnetic relaxation takes place
as in the two previous specimens, and a similar magnetic pattern
appears (Figure 7a, c, d, f, g, and i). However, the evolution of
+
3+
the (super)paramagnetic Fe proportion with decreasing tem-
perature is different from that obtained for Fe10U1 and Fe10U4
28,31,33-36
3+
by referring to other work,
these two maxima can be
(Figure 8). Indeed, the (super)paramagnetic Fe proportion in
attributed to Fe2 ions distributed in the Oh sites of MgO, the
+
Fe10U8 decreases rapidly below RT to reach a constant value
of about 80%, which further remains constant between 250 and
100 K. It subsequently slightly decreases between 100 and 50
K and then steeply decreases below 50 K. Moreover, below 50
2+
one with larger quadrupole splitting revealing Fe ions, which
form clusters. Moreover, the Fe2 proportion can be considered
to be constant down to T ) 30 K, but it decreases significantly
at 14 K (Table 3), which is accompanied by a decrease in the
+
3+
K, the third maximum in the Fe quadrupole-splitting distribu-
tion disappears (Figure 7b and e). This suggests a bimodal size
2+
second maximum in the Fe quadrupole-splitting distribution
Figure 6b). This finding suggests that the Fe2+ clusters can be
considered to be very small superparamagnetic particles, as
3+
(
distribution of Fe clusters that is considerably more established
than in Fe10U1. The magnetic splitting slightly below RT

13194 J. Phys. Chem. B, Vol. 106, No. 51, 2002
Coquay et al.
Figure 7. MS of Fe10U8 measured between 14 K and RT (a, d, g, j) and corresponding quadrupole-splitting distributions of the (super)paramagnetic
3
+
Fe phases (b, e, h, k) and hyperfine-field distributions of the magnetic MgFe
O
2 4
-like phase (c, f, i).
(
Figure 2). The decrease at temperatures below 50 K, however,
3
+
suggests very small Fe clusters in MgO that are probably not
detected in the XRD analysis. At 14 K, the second maximum
3
+
at ∆EQ ≈ 1.00 mm/s is still present in the Fe quadrupole-
splitting distribution of Fe10U8 (Figure 7b), which was not the
case for Fe10U1 and Fe10U4. Moreover, a second maximum
at lower field (Hhf ≈ 430 kOe) is clearly observed in the
hyperfine-field distribution at 14 K (Table 4 and Figure 7c),
denoting the superparamagnetic relaxation of these very small
clusters.
Comparison of Fe5 and Fe10 Oxide Powders. Table 5 gives
the RT M o¨ ssbauer parameters of Fe5U1, Fe5U4, and Fe5U8
together with the corresponding parameters for Fe10U1, Fe10U4,
and Fe10U8 discussed in the previous sections. Three indepen-
Figure 8. (Super)paramagnetic Fe3+ proportions in Fe10U1, Fe10U4,
and Fe10U8 as determined from the M o¨ ssbauer study (values from
Tables 2, 3, and 4, respectively) versus the measuring temperature.
The values for Fe10U4 were normalized to 100% at RT for the sake
of comparison. The lines are guides for the eye.
3+
2+
dent Fe doublets and one Fe quadrupole-splitting distribution
were considered for the fitting of the MS that are presented in
Figure 9.
2+
reveals MgFe2O4-like particles that are larger than those detected
in Fe10U1, confirming the qualitative result of the XRD analysis
Fe5U1 contains no Fe phase, and the M o¨ ssbauer parameters
3
+
of the three Fe doublets are similar to those of Fe10U1. The

Carbon Nanotubes by a CVD Method
J. Phys. Chem. B, Vol. 106, No. 51, 2002 13195
Figure 9. MS of Fe5U1 (a) and Fe10U1 (b), Fe5U4 (c) and Fe10U4 (d), and Fe5U8 (e) and Fe10U8 (f) measured at RT. Three independent
doublets were considered to characterize the Fe3+ phases.
TABLE 5: M o1 ssbauer Parameters of Fe5U1 and Fe10U1, Fe5U4 and Fe10U4, and Fe5U8 and Fe10U8 Measured at RTa
Fe2
+
O
h
Fe3+ b
O
h
Fe3+ c
T
d
Fe3+
O
h
oxide
powder
d
δ
∆E
Q
Γ
P
δ
∆E
Q
Γ
P
δ
∆E
Q
Γ
P
δ
∆E
Q
Γ
P
Fe5U1
Fe10U1
0.30
0.30
0.50
0.49
0.45
0.50
56.5
60.0
0.30
0.30
0.93
0.91
0.45
0.43
37.5
29.0
0.26
0.27
1.49
1.38
0.40
0.43
6.0
11.0
Fe5U4
1.07
1.04
0.13
.45
0.20
.48
0.29
0.34
32.0
37.0
0.35
0.32
0.45
0.49
0.41
0.57
22.5
36.0
0.32
0.41
0.85
1.12
0.52
0.67
37.0
27.0
0.28
1.58
0.54
8.5
0
Fe10U4
0
Fe5U8
Fe10U8
0.32
0.32
0.51
0.49
0.37
0.38
46.0
38.5
0.29
0.30
0.87
0.86
0.37
0.46
34.5
45.0
0.26
0.25
1.26
1.27
0.41
0.51
19.5
16.5
a
Three independent doublets were considered to characterize the Fe3+ phases. O
: quadrupole splitting (at the maxima of the distribution) (mm/s); Γ: Lorentzian line width (mm/s); P: proportion (%) Isolated
h
: octahedral sites; T
d
: tetrahedral sites; δ: (average) isomer
b
shift (mm/s); ∆E
ions in MgO (or possibly ions in MgFe
Q
c
2
O
4
). Ions forming clusters in MgO. d Quadrupole-splitting distribution from 0.00 to 1.10 mm/s.
relatively small variations in the proportions are not significant
enough to infer differences between the two samples.
Fe5U4 exhibits an Fe2 quadrupole-splitting distribution with
Fe5U4, which favors the formation of MgFe2O4-like particles.
This can be associated with the slightly higher Sss value
measured for Fe5U4 than for Fe10U4, suggesting smaller
particles on the surface of the MgO grains for Fe5U4 (Table
1). It thus appears that the substitution of Fe and Fe ions
in the MgO lattice is considerably favored using 10 cat. % iron
rather than 5 cat. % iron when a urea ratio of 4 is used for the
combustion.
+
2
+
two clearly separated maxima (Figure 9c). The global Fe
proportion in Fe5U4 is slightly smaller than the one in Fe10U4
2
+
3+
2
+
(
Table 5), revealing that the number of substituted Fe ions is
more than twice as large in Fe10U4 as in Fe5U4. However,
contrary to the proportion in Fe10U4, the proportion of the first
maximum in the Fe2 quadrupole-splitting distribution of Fe5U4
is higher than the proportion of the second one (Figure 9c and
d), showing that the larger number of Fe2 ions substituted in
+
The results for Fe5U8 are similar to those obtained for
Fe10U8, in particular, showing a rather large proportion of Td
+
3+
Fe ions. It can thus be extrapolated that a similar bimodal
2+
the MgO lattice of Fe10U4 favors the formation of Fe clusters.
Moreover, the proportion of Oh Fe3 ions that form clusters is
higher in Fe5U4 than in Fe10U4, and the MS of Fe5U4 contains
size distribution, corresponding to large MgFe2O4-like particles
+
3+
and small Fe clusters, respectively, is also the case for this
sample.
3+
a doublet characteristic of Td Fe , which is not detected in
As suggested above, the greenish color observed for the
powders synthesized with a urea ratio of 4 points to the presence
Fe10U4. This reveals a higher degree of Fe3 clustering in
+

13196 J. Phys. Chem. B, Vol. 106, No. 51, 2002
Coquay et al.
Figure 10. SEM images of Fe10U1 (a, b), Fe10U4 (c, d), and Fe10U8 (e, f).
2
+
of Fe ions. For powders prepared with ratios of 1 and 8, no
the primary grain sizes evaluated on these images seem to be
globally larger than the crystallite sizes obtained from the widths
at half-maximum of the MgO peaks in the XRD patterns (Figure
4), showing that not all of the primary grains are monocrystal-
line. MgFe2O4-like particles could not be detected in the SEM
images.
2+
3+
Fe is detected, and the Fe ions are responsible for the red
color of the powders. Moreover, the green and red areas
observed in the FeU4 oxide powders prepared per batch of 6 g
were analyzed separately by M o¨ ssbauer spectroscopy. The green
area is characterized by a high proportion of Fe ions (50% of
all the iron) and by Fe3 ions that are well dispersed in the Oh
sites of MgO (20%), whereas the red area is characterized by a
high proportion of Td Fe3+ ions (35%) and by Oh Fe ions,
which form clusters in MgO (55%). Even if the present FeU4
oxide powders prepared per batch of 2 g appear to be
homogeneous because of a better expansion of the powder
2+
+
These observations can be correlated to the evolution of Sss
versus the urea ratio presented in Figure 1. For Fe10U1, Sss is
mainly influenced by the small MgFe2O4-like particles detected
by XRD and M o¨ ssbauer spectroscopy. Indeed, an increase of
the iron content in the oxide corresponds to an increase of Sss
(Table 1), whereas the microstructure remains compact (Figure
10a and b). When the urea ratio increases from 1 to 4, the
temperature during the reaction increases, favoring sintering of
the material and coalescence of the crystallites. Moreover, the
XRD and M o¨ ssbauer spectroscopy analyses have shown that
more iron is dissolved in the MgO lattice. This explains the
decrease of Sss observed between the urea ratios of 1 and 4
(Figure 1). The increase of the gas release explains the porosity
of the foam observed for Fe10U4 (Figure 10c and d), but the
pores are too large to increase the Sss values notably. When the
urea ratio increases from 4 to 8, the temperature of the reaction
decreases, leading to very little sintering and thus to much
smaller primary grains and crystallites. Moreover, the gas release
and the resulting expansion of the material considerably increase,
giving rise to the more porous foam observed for Fe10U8
(Figure 10e and f). Consequently, the maximal Sss value is
observed for Fe10U8. Indeed, the MgFe2O4-like particles
3+
3
+
during the combustion process, a local agglomeration of Fe
clusters can be expected.
Scanning Electron Microscopy. SEM images (Figure 10)
show the microstructural differences between Fe10U1, Fe10U4,
and Fe10U8. For Fe10U1, grains on the order of 10 µm are
observed (Figure 10a). At a higher magnification, it appears
that these grains consist of closely packed primary grains on
the order of 100 nm (Figure 10b). For Fe10U4, no such
individual grains can be observed, but the combustion product
appears to be a porous foam, showing large open cavities (Figure
1
0c). At higher magnification, however, it appears that this foam
also consists of closely packed primary grains, the size of which
seems to be slightly larger than in Fe10U1 (Figure 10d). For
Fe10U8, the product is even more porous (Figure 10e), and the
primary grains are very loosely packed. Moreover, the primary
grain size is smaller than in Fe10U4 and Fe10U1. Note that

Carbon Nanotubes by a CVD Method
J. Phys. Chem. B, Vol. 106, No. 51, 2002 13197
Stangle.19 Indeed, the so-obtained oxides contain about 40% of
2+
the total iron substituting as Fe in MgO, and a large proportion
3+
of the Fe ions are dispersed in the Oh sites of MgO. However,
3
+
local agglomeration of Fe clusters can be expected in these
samples. The powders appear as a foam with large cavities,
resulting from the sintering of small primary grains (∼60 nm).
For a urea ratio increased by a factor of 8, the considerable
expansion due to the gas release during the combustion results
in a very porous powder constituted of small loosely packed
2
+
MgO primary grains (∼25 nm). No Fe ions are formed, and
3
+
the Fe ions are present in iron agglomerates with a bimodal
3
+
size distribution: very small Fe clusters that are not detected
for other urea ratios and MgFe2O4-like particles that are larger,
but much better dispersed, than those detected for a urea ratio
of 1. The Fe/MgO oxides prepared with urea ratios of 1, 4, and
8
are clearly different from one another, presenting several well-
defined iron species. The formation of CNTs from these powders
(
with different iron contents) upon reduction in a H2/CH4 gas
2
1
atmosphere is studied in a companion paper.
Acknowledgment. We thank Mr. L. Datas for his assistance
with the TEM observations, which have been performed at the
Service Commun de Microscopie Electronique a` Transmissions
Universit e´ Paul-Sabatier. This research is supported by the
Belgian National Program of Inter-University Attraction Pole
on Reduced Dimensionality Systems (P4/10), by the Fund for
Scientific ResearchsFlanders, and by the Franco-Belgian
TOURNESOL program (T99/006-T99/045).
Figure 11. TEM images of Fe10U1 (a) and Fe10U8 (b).
detected by XRD and M o¨ ssbauer spectroscopy make a negligible
contribution to Sss, as reflected by the weak increase of Sss with
the iron content (Table 1).
Transmission Electron Microscopy. In the FeU1 and FeU8
oxides, a high proportion of iron forms Fe3 clusters and
MgFe2O4-like particles. To get more information about the
repartition of iron outside the MgO lattice in these samples,
Fe10U1 and Fe10U8 were examined by TEM. Many EDX
measurements were carried out on different areas of the samples.
The impact position and the surface area of the beam were
always adjusted to be characteristic of the surface of the studied
area. Magnesium and iron proportions (atom %) were compared.
Figure 11 shows typical TEM images of Fe10U1 and Fe10U8.
For Fe10U1, unstructured areas are observed among cubic
forms. The cubic forms are typical of the rock salt structure of
MgO crystallites. They contain very little iron (1-2 at %).
However, the unstructured areas are very rich in iron (40-50
atom %), revealing a large amount of agglomeration of the iron
in Fe10U1. For Fe10U8, no such difference could be observed,
and all measurements gave an iron proportion of 5 to 15 atom
+
References and Notes
(1) Hafner, J. H.; Bronikowski, M. J.; Azamian, B. K.; Nikolaev, P.;
Rinzler, A. G.; Colbert, D. T.; Smith, K. A.; Smalley, R. E. Chem. Phys.
Lett. 1998, 296, 195.
(2) Flahaut, E.; Peigney, A.; Laurent, Ch.; Rousset, A. J. Mater. Chem.
2
000, 10, 249.
3) Bacsa, R. R.; Laurent, Ch.; Peigney, A.; Bacsa, W. S.; Vaugien,
Th.; Rousset, A. Chem. Phys. Lett. 2000, 323, 566.
4) Peigney, A.; Coquay, P.; Flahaut, E.; Vandenberghe, R. E.; De
Grave, E.; Laurent, Ch. J. Phys. Chem. B 2001, 105, 9699.
5) Herrera, J. E.; Balzano, L.; Borgna, A.; Alvarez, W. E.; Resasco,
D. E. J. Catal. 2001, 204, 129.
6) Alvarez, W. E.; Kitiyanan, B.; Borgna, A.; Resasco, D. E. Carbon
2001, 39, 547.
(7) Li, Y.; Liu, J.; Wang, Y.; Wang, Z. L. Chem. Mater. 2001, 13,
008.
8) Peigney, A.; Laurent, Ch.; Dobigeon, F.; Rousset, A. J. Mater. Res.
(
(
(
(
1
(
1997, 12, 613.
%
, revealing that iron is much more dispersed in Fe10U8 than
(9) Peigney, A.; Laurent, Ch.; Dumortier, O.; Rousset, A. J. Eur.
Ceram. Soc. 1998, 18, 1995.
10) Laurent, Ch.; Peigney, A.; Rousset, A. J. Mater. Chem. 1998, 8,
263.
11) Laurent, Ch.; Peigney, A.; Flahaut, E.; Rousset, A. Mater. Res.
Bull. 2000, 35, 661.
12) Govindaraj, A.; Flahaut, E.; Laurent, Ch.; Peigney, A.; Rousset,
A.; Rao, C. N. R. J. Mater. Res. 1999, 14, 2567.
(13) Coquay, P.; De Grave, E.; Vandenberghe, R. E.; Dauwe, C.; Flahaut,
E.; Laurent, Ch.; Peigney, A.; Rousset, A. Acta Mater. 2000, 48, 3015.
14) Qu e´ nard, O.; De Grave, E.; Laurent, Ch.; Rousset, A. J. Mater.
Chem. 1997, 7, 2457.
(15) Flahaut, E.; Govindaraj, A.; Peigney, A.; Laurent, Ch.; Rousset,
A.; Rao, C. N. R. Chem. Phys. Lett. 1999, 300, 236.
(16) Kingsley, J. J.; Patil, K. C. Mater. Lett. 1988, 6, 427.
17) Patil, K. C. Bull. Mater. Sci. 1993, 16, 533.
18) Jain, S. R.; Adiga, K. C.; Pai Verneker, V. R. Combust. Flame
1981, 40, 71.
19) Zhang, Y.; Stangle, G. C. J. Mater. Res. 1994, 9, 1997.
20) Bacsa, R. R.; Laurent, Ch.; Peigney, A.; Vaugien, Th.; Flahaut,
E.; Bacsa, W. S.; Rousset, A. J. Am. Ceram. Soc. 2002, 85, 2666.
(21) Coquay, P.; Peigney, A.; De Grave, E.; Vandenberghe, R. E.;
Laurent, Ch. J. Phys. Chem. B 2002, 106, 13199.
22) Dinnebier, R. E.; Eysel, W. Abstract of Powder Diffraction Satellite
Meeting of the XVth Congress of the International Union of Crystallography,
Toulouse, France, 1990; p 279.
in Fe10U1.
(
1
Conclusions
(
We attempted to synthesize Mg1-xFexO oxide solid solutions
by the combustion route with the aim of studying the influences
of the nitrate/urea ratio and the iron content on the valency and
distribution of the iron ions and phases. The Fe/MgO oxides
synthesized with the so-called stoichiometric ratio contain large
grains (∼10 µm) that consist of highly agglomerated small
primary grains (∼35 nm), some of them being made up of
(
(
several crystallites. No Fe2 ions substitute for Mg in the MgO
+
2+
(
(
lattice, and most of the Fe3 ions form clusters and MgFe2O4-
like particles that are poorly dispersed in the powder, as revealed
by large fluctuations of the local iron content. The combustion
conditions are therefore not sufficiently reducing. To increase
the substitution ratio of the iron cations in the MgO lattice, the
urea proportion was gradually increased up to a factor 8.
Increasing it by a factor in the range of 3.5-4.5 leads to
combustion conditions that would correspond as closely as
possible to the stoichiometric region as defined by Zhang and
+
(
(
(
(23) Holland, T. J. B.; Redfern, S. A. T. Mineral. Mag. 1997, 61, 65.

13198 J. Phys. Chem. B, Vol. 106, No. 51, 2002
Coquay et al.
(
24) Vandenberghe, R. E.; De Grave, E.; de Bakker, P. M. A. Hyperfine
(34) Perez, A.; Marest, G.; Sawicka, B. D.; Sawicki, J. A.; Tyliszczak,
Interact. 1994, 83, 29.
T. Phys. ReV. B 1983, 28, 1227.
(35) Simkin, D. J.; Ficalora, P. J.; Bernheim, R. A. Phys. Lett. 1965,
(25) Berthet, A. Ph.D. Thesis, University of Nancy, Nancy, France, 1963.
(26) Berthet, A.; Perrot, P. Mem. Sci. ReV. Metall. 1970, LXVII, 747.
(27) Trinel-Dufour, M. C.; Perrot, P. Ann. Chim. 1977, 2, 309.
(28) Carles, V.; Rousset, A. Solid State Ionics 1996, 83, 309.
(29) Blank, S. L.; Pask, J. A. J. Am. Ceram. Soc. 1969, 52, 669.
(30) Brynestad, J.; Flood, H. Z. Elektrochem. 1958, 62, 953.
(31) Boudart, M.; Delbouille, A.; Dumesic, J. A.; Khammouma, S.;
1
9, 536.
(36) Topsoe, H.; Dumesic, J. A.; Derouane, E. G.; Clausen, B. S.; Morup,
S.; Villadsen, J.; Topsoe, N. Stud. Surf. Sci. Catal. 1979, 3, 365.
(37) Woods, K. N.; Fine, M. E. J. Appl. Phys. 1969, 40, 3425.
(38) Bhide, V. G.; Tambe, B. R. J. Mater. Sci. 1969, 4, 955.
(
(
(
39) Krawitz, A.; Cohen, J. B. J. Am. Ceram. Soc. 1974, 57, 186.
40) Waychunas, G. A. J. Mater. Sci. 1983, 18, 195.
41) De Grave, E.; Govaert, A.; Chambaere, D.; Robbrecht, G. Physica
Topsoe, H. J. Catal. 1975, 37, 486.
B 1979, 96, 103.
42) De Bakker, P. Ph.D. Thesis, University of Ghent, Ghent, Belgium,
1994.
(
(
32) Yamaguchi, G.; Tokuda, T. J. Ceram. Assoc. Jpn. 1968, 76, 350.
33) Jing, J.; Campbell, S. J. Hyperfine Interact. 1991, 68, 283.
(