J O U R N A L O F C H E M I S T R Y Materials Synthesis of carbon nanotube–Fe-Al2O3 nanocomposite powders by selective reduction of diVerent Al1.8Fe0.2O3 solid solutions Ch. Laurent, A. Peigney and A. Rousset L aboratoire de Chimie desMate�riaux Inorganiques, ESA CNRS 5070, Universite� Paul-Sabatier, F 31062 T oulouse cedex 4, France Al1.8Fe0.2O3 solid solutions have been prepared as amorphous, g (cubic) and a (corundum) phases.The oxides have been reduced in a H2–CH4 gas mixture at 900 or 1000 °C, giving rise to composite powders containing alumina, a- and c-Fe, Fe3C and diVerent forms of carbon including nanotubes, thick tubes and spheroidal particles. The powders have been investigated using a combination of chemical analysis, X-ray diVraction, Mo�ssbauer spectroscopy, scanning and transmission electron microscopy, thermogravimetric analysis and specific surface area measurements.Using the stable form (corundum) of Al1.8Fe0.2O3 as starting material favours the formation of carbon nanotubes compared to the other forms of carbon. This could partly result from the fact that the metal nanoparticles formed upon reduction of the a solid solution, which act as a catalyst for CH4 decomposition and possibly nanotube nucleation, are smaller than when using amorphous or g solid solutions.Moreover, the crystallization of these latter compounds during the reduction in some way provokes the entrapment of carbon within the oxide grains. The nanotubes, most of which are less than 10 nm in diameter, are arranged in bundles several tens of micrometers long.shown19 that when using a hydrogen–hydrocarbon gas mixture Introduction instead of pure hydrogen for the reduction of an a-Al1.9Fe0.1O3 The discovery of carbon nanotubes1 triggered a worldwide solid solution, the pristine Fe nanoparticles formed in situ research eVort devoted to improving their synthesis,2–19 to upon reduction of the very homogeneously dispersed surface determining their structure20–24 and to calculating and measur- Fe3+ ions are active at a size adequate for the catalytic growth ing their physical properties.25–33 Particularly, regarding the of nanotubes.The resulting carbon nanotube–Fe-Al2O3 commechanical properties, the Young’s modulus of multi-walled posite powder contains a huge amount of single- and multicarbon nanotubes has been calculated to be up to 1.4 times walled carbon nanotubes, which have diameters in the that of a graphite whisker, i.e.about 1 TPa,34 and values 1.5–15 nm range. The nanotubes are arranged in bundles derived from thermal vibration experiments performed on smaller than 100 nm in diameter which may be up to 100 mm several multi-walled carbon nanotubes in a transmission elec- long.It has been calculated that the total bundle length in 1 g tron microscope are in the 0.4–3.7 TPa range.35 Moreover, of this powder is equal to about 100 000 km. Indeed, the metaltheir flexibility is remarkable36 and the bending may be fully oxide grains are uniformly covered by a web-like network of reversible up to a critical angle value as large as 110° for bundles and the powder is so densely agglomerated that it single-walled nanotubes.37 retains the shape of the reduction vessel when transferred to a Carbon nanotubes are most often prepared by arc-discharge storage box.between carbon electrodes in an inert gas atmosphere.1,2 Since the only metal particles active for the catalytic Catalytic species such as iron or cobalt can be used during the decomposition of the hydrocarbon are those located at the arc-discharge to improve both the quantity and the length of surface of the matrix grains, the active-to-inactive particle ratio tubes and to favour the formation of single-shell nanotubes.3–8 could be increased using powders with a high specific surface However, with this process, carbon nanotubes are obtained as area such as poorly crystallized or uncrystallized oxide solid a mixture with several other carbon forms, including much solutions. The aim of this work is to determine the influence amorphous carbon.Thus, purification has to be carried out of the crystallization level of an Al1.8Fe0.2O3 solid solution on and the yield of nanotubes is no more than 2%.38 Recently, the formation of carbon nanotubes, in order to increase the ‘ropes’ of single-walled carbon nanotubes were obtained by an tube yield, to maximize their length and to minimize their adapted laser-ablation technique.9,10 An alternative method is diameter, with the ultimate objective of obtaining a composite the catalytic decomposition of hydrocarbons on small metal powder suitable for the preparation of dense composite mateparticles (Fe, Co, Ni, Cu), which produces carbon filaments rials that may benefit from the exceptional properties of the including some authentic nanotubes.11–19 The minimal tube carbon nanotubes. diameter that can be achieved in this way is that of the catalytic particles. In order to maximize the nanotube yield, several Experimental authors investigated the influence of the temperature and of the nature of both the catalyst and the treatment atmos- The appropriate amounts of (NH4)2(C2O4) 2H2O, phere.13,39,40 Using a zeolite-supported Co catalyst, Ivanov Al(NO3)3 9H2O, and Fe(NO3)3 9H2O were mixed in an et al.14 and Hernadi et al.41 reported carbon tubes only 4 nm aqueous solution heated at 60 °C.The obtained clear solution in diameter and tubes of length 60 mm, but they point out that was cooled to room temperature and rapidly added to an the longest tubes are also the thickest. alcoholic medium, in which precipitation of the mixed In previous works, we have prepared metal-oxide nanocom- ammonium oxalate (NH4)3[Al0.9Fe0.1(C2O4)3] nH2O occurrposite powders by selective reduction in hydrogen of oxide ed immediately. After filtering, washing and oven drying, the solid solutions.42–47 In these materials, the metal particles (Cr, oxalate was finely ground and decomposed at 400 °C for 2 h.Fe, Co, Ni and their alloys) are generally smaller than 10 nm This powder will be denoted as C400 hereafter. Samples of the in diameter and are located both inside and at the surface of obtained C400 batch were calcined in air for 2 h, either at 850 °C (specimen C850) or at 1100 °C (powder C1100).The the matrix grains (Al2O3, Cr2 O3, MgO, MgAl2O4). We have J. Mater. Chem., 1998, 8(5), 1263–1272 1263three diVerent oxide powders were reduced in a H2–CH4 gas [Fig. 1(c)]. These results are in agreement with an earlier work48 showing that the thermal decomposition at 400 °C of mixture (6 mol% CH4) during 4 h at 900 °C or at 1000 °C (specimens denoted C400/R900, C400/R1000 etc.), giving rise a mixed oxalate (NH4)3[Al0.9Fe0.1(C2O4)3] gives rise to an Xray amorphous Al1.8Fe0.2O3 solid solution, which upon calci- to the carbon–metal-oxide composite powders.The materials were studied using scanning and transmission nation at the appropriate temperature crystallizes into the g (cubic) or a (corundum) structure.electron microscopy (SEM and TEM), X-ray diVraction (XRD) using Co-Ka radiation (l=0.17902 nm) and 57Fe Mo� ssbauer SEM observations show that the powders are made up of 10–20 mm agglomerates of submicronic or nanometric primary spectroscopy. The Mo�ssbauer spectra were recorded at room temperature with a constant acceleration spectrometer using a grains. The specific surface area measured for each oxide powder (Sss; Table 1) is consistent with the crystallization level 50 mCi 57Co (Rh) source; the spectrometer was calibrated by collecting at room temperature the spectrum of a standard Fe revealed by XRD analysis: 80 m2 g-1 for the amorphous Al1.8Fe0.2O3 solid solution (C400), 30 m2 g-1 for the poorly foil and the center shift (CS) values quoted hereafter are with reference to this standard.The composite powders were oxid- crystallized g-Al1.8Fe0.2O3 powder (C850) and 3.1 m2 g-1 for the a-Al1.8Fe0.2O3 powder (C1100). ized in air at 850 °C in order to eliminate all or part of the carbon, as required for the specific surface area study. The specific surfacethe oxide powders (Sss), of the nano- Nanocomposite powders composite powders obtained after reduction (Sn) and of the X-Ray diVraction. The XRD patterns of the nanocomposite powders oxidized at 850 °C (Son) were measured by the BET powders prepared by reduction of the solid solutions at 900 method using N2 adsorption at liquid N2 temperature.We or 1000 °C are shown in Fig. 2. In C400/R900, we observe used a Micromeritics FlowSorb II 2300 apparatus that gives wide peaks accounting for the presence of diVerent forms of a specific surface area value from one point (i.e. one adsorbant transition alumina (g, c, h) and only traces of a-Al2O3. In pressure) and requires calibration. For a given powder, we addition, a-Fe and Fe3C (cementite) are detected. A wide peak successively measured Sss, Sn and Son using the same caliwhich could correspond to the distance between graphene bration.The reproducibility of the results is in the±5% range, layers (d002=0.34 nm) is also detected. Since neither the (hk0) corresponding to ±10% for DS=Sn-Son, which accounts for nor the other (hkl) reflections, which would have much smaller the quantity of carbon nanotubes as detailed later in this intensities for nanotubes as well as for graphite,33 are detected, paper.Furthermore, we checked that the specific surface area it is not possible from this XRD pattern to discriminate measurements performed by the present method are in good between graphite and other graphenic forms of carbon such agreement with those performed using a multipoint as nanotubes.Thus we have labelled this peak Cg (Cgraphene). Micromeritics Accusorb 2100 E apparatus. The carbon content The same phases are present in C400/R1000, but the intensities was determined by flash combustion. The oxidation of the of the peaks representing a-Al2O3, a-Fe and carbon are much nanocomposite powders was investigated by thermogravihigher. The XRD patterns of C850/R900 and C850/R1000 are metric analysis (TGA) in flowing air (heating rate 1 °C min-1).similar to those of C400/R900 and C400/R1000, respectively. The main diVerence is the variation in the proportions of the Results and Discussion diVerent forms of alumina on the one hand and in the proportions of a-Fe and Fe3C on the other hand. Also, the Oxide powders (110) peak of a-Fe is distinctly wider in the C850-derived The XRD patterns of the C400, C850 and C1100 powders are composites, denoting that the decrease in specific surface area shown in Fig. 1. The pattern of specimen C400 is blank whereas of the oxide solid solution favours the formation of smaller that of the C850 powder presents wide peaks characteristics metal particles, as shown previously.42 Analysis of the XRD of an g-Al1.8Fe0.2O3 solid solution as well as traces of the patterns of C1100/R900 and C1100/R1000 clearly reveals the (104), (113) and (116) peaks of an a-Al2O3-type (corundum) presence of a-Al2O3, a-Fe and Fe3C.The proportions of phase [Fig. 1(a), (b)]. The g-phase peaks are very wide (more carbide and, notably, carbon are much lower than in the other than 10° 2h), reflecting a small crystallite size and a poor powders.c-Fe may be present in all or some powders, but is crystallization level. For the C1100 powder, all peaks are in extremely diYcult to detect on the XRD patterns because the good agreement with an a-Al1.8Fe0.2O3 solid solution c-Fe (111) diVraction peak (d111=0.208 nm) is probably masked by the base of the corundum (113) peak (d113= 0.2085 nm), and more so if Fe3C (d210=0.206 nm) is present as well.A FeAl2O4 spinel phase, which is known to form upon reduction of Al1.8Fe0.2O3 in pure H2 at temperatures lower than 1000 °C,44–47 is not detected in the R900 powders. However, one can not rule out its presence in small quantities, particularly in the C400/R900 and C850/R900 specimens, because the FeAl2O4 main diVraction peaks (d113=0.245 nm, d004=0.202 nm and d044=0.143 nm) may be masked by transition- alumina and a-Fe peaks.Carbon content. The carbon content measured in the nanocomposite powders (Cn; Table 1) is in the 1–20 wt.% range. Cn is much higher in the composites derived from the amorphous Al1.8Fe0.2O3 and g-Al1.8Fe0.2O3 phases than in those derived from the stable a-Al1.8Fe0.2O3 phase.This could result from the higher specific surface area of the former powders, which would allow more metal particles to nucleate and grow on the surface of the oxide grains and therefore to be active as catalysts for the decomposition of CH4. We also note that for a given starting powder, the carbon content is higher after reduction at 1000 °C than after reduction at 900 °C.According Fig. 1 XRD patterns of the oxide powders prepared at diVerent temperatures: (a) 400 °C; (b) 850 °C; (c) 1100 °C to Wagman et al.,49 the CH4 equilibrium content in H2–CH4 1264 J. Mater. Chem., 1998, 8(5), 1263–1272Table 1 Specific surface area (Sss: oxide solid solutions; Sn: nanocomposites; Son: nanocomposites oxidized at 850 °C; DS=Sn-Son) and carbon contents (Cn: nanocomposites; Con: nanocomposites oxidized at 850 °C) of the powders Sss Sn Son Cn Con DS DS/Cn oxide /m2 g-1 composite /m2 g-1 /m2 g-1 (wt.%) (wt.%) /m2 g-1 /m2 g-1 C400 79.5 C400/R900 23.5 20.0 13.4 0.65 3.5 26 C400/R1000 16.0 11.0 20.1 0.58 5.0 25 C850 30.5 C850/R900 44.1 39.3 8.9 0.30 4.8 54 C850/R1000 31.0 25.1 18.3 0.21 5.9 32 C1100 3.1 C1100/R900 6.7 4.1 1.65 0.00 2.6 155 C1100/R1000 8.4 3.5 6.2 0.00 4.9 79 that the reducibility of the Fe3+ ions substituting in the alumina lattice, and thus the formation of metal nanoparticles, are hampered by a decreasing specific surface area of the oxide powder44 and that consequently a 100 °C increase in reduction temperature has more impact with low specific surface area oxides. Mo�ssbauer spectroscopy.All composite powders except C400/R1000 were studied by Mo�ssbauer spectroscopy. Each spectrum was fitted assuming it is the sum of diVerent subspectra: a sextet accounting for ferromagnetic a-Fe, a sextet representing Fe3C and a singlet characteristic of non-ferromagnetic Fe (Fig. 3 and Table 2). It is noteworthy that neither Fe3+ nor Fe2+ ions were detected on the Mo�ssbauer spectra, even in the composites prepared from the C1100 oxides, i.e.the less easily reducible phase (a-Al1.8Fe0.2O3). This shows that the presence of CH4 in the reducing gas mixture tremendously favours the reduction of the iron(III ) ions substituting for aluminium in the corundum lattice: indeed, when using pure H2, Fe3+ contents of 35 and 65% are measured in C1000/R1000 and C1200/R1000 composite powders, respectively,44 from which one may estimate that the Fe3+ proportion would be of the order of 50% in a C1100/R1000 specimen. It has been shown50 that treatment at 1300 °C is necessary to fully reduce the Fe3+ ions substituting for aluminium in the corundum lattice to the metallic state.Fe2+ ions could account for the presence of a phase of the FeAl2O4 spinel type.44–47 As the detection limit in Mo�ssbauer spectroscopy is about 4%, a very small amount of spinel phase could indeed be present in the R900 composite powders while no corresponding pattern is seen in the spectra.44 However, considering the above observation that CH4 strongly favours the formation of metallic Fe, it is reasonable to assume that no residual Fe3+ and Fe2+ ions are present in the composite powders.Since Fe3C has two inequivalent crystallographic sites,51 the Mo�ssbauer parameters of the sextet accounting for Fe3C correspond to the average of the two Fe-site parameters one may obtain using two sextets for the fit. Our average values are in good agreement with those reported by Le Cae�r et al.52 for bulk Fe3C and by Bi et al.53 for Fe3C nanoparticles. The non-ferromagnetic phase of metallic Fe corresponding to the singlet could be either antiferromagnetic, paramagnetic or superparamagnetic a-Fe, or c-Fe.Paramagnetic a-Fe has Fig. 2 XRD patterns of the nanocomposite powders prepared by been observed in Fe/Ni-MgAl2O4 nanocomposite powders54 calcination (C) and reduction (R) treatments at diVerent temperatures: and superparamagnetic a-Fe was fresent in some (a) C400/R900; (b) C400/R1000; (c) C850/R900; (d) C850/R1000; Fe-Al2O3 specimens.44,45,47 However, the negative value of the (e) C1100/R900; (f ) C1100/R1000; a: a-alumina; O: transition alumina; +: Fe3C; *: a-Fe; Cg: corresponding to d002 in multi-walled nanotubes CS points towards the face-centered-cubic c-Fe phase.This and/or in graphite could reflect the formation of a c-Fe–C alloy, rather than pure Fe. It is interesting to note that Baker et al.55 have shown that gas mixtures is lower at 1000 °C (about 1%) than at 900 °C metallic Fe, and not cementite, is the active phase responsible (about 2%). Thus, the CH4 supersaturation level in the gas for the formation of carbon nanofilaments.In the present mixture used in the present study (6 mol% CH4) is much work, the proportions of the diVerent Fe species vary from higher at 1000 °C than at 900 °C and consequently more one powder to another, but analysis of the data does not allow carbon is deposited during the reaction at 1000 °C. Moreover, one to draw precise conclusions from their evolution. The very the increase in carbon content when the reduction temperature low value (2%) observed for Fe3C in the C400/R900 powder is changed from 900 to 1000 °C is higher when using solid solutions calcined at a higher temperature. This could reflect is probably a fit artefact and the relatively high proportion of J.Mater. Chem., 1998, 8(5), 1263–1272 1265in the composite powders reduced from the stable a solid solutions (C1100/R900 and R1000) and the latter one in the specimens prepared from the high specific surface area oxides, because this favours the formation of surface metal particles.Electron microscopy. Low-magnification SEM observations of each of the six composite powders show that the matrix grains (about 20 mm long) are covered by a web-like network of carbon filaments (Fig. 4), in line with earlier results.19 However, higher magnification SEM images [Fig. 5(a)–(f )] clearly reveal some diVerences depending on the calcination and/or the reduction temperature used for the preparation of the powders. In the C400/R900 powder [Fig. 5(a)], we observe long, tight bundles with a diameter smaller than 50 nm, as well as curved filaments both thicker (up to 100 nm in diameter) and shorter (about 1 mm or less), and some clusters of spheroidal grains in the 20–50 nm diameter range.The C400/R1000 powder [Fig. 5(b)] presents less carbon bundles and thick filaments but more spheroidal grains which seem to be larger than in the C400/R900 powder. Both the C850/R900 and C850/R1000 powders [Fig. 5(c), (d)] contain tight bundles and small spheroidal clusters in various quantities from one powder grain to another.Thus, the diVerence in spheroidal cluster quantities between the two images does not reflects a diVerence between the two powders. Images of the C1100/R900 and the C1100/R1000 powders [Fig. 5(e)–(f )] show more bundles but interestingly neither thick filaments nor spheroidal clusters have been detected on any observed matrix grain.We also observe homogeneously dispersed nanometric particles which could be Fe or Fe3C particles. TEM [Fig. 6(a)–(c)] and HREM [Fig. 6(d)–(f )] images of nanocomposite powders show details of the diVerent forms of carbon previously revealed by SEM observations. A thick tube,12 irregularly shaped, is shown in the C400/R900 powder Fig. 3 Mo�ssbauer spectra of the nanocomposite powders prepared by calcination (C) and reduction (R) treatments at diVerent temperatures: (a) C400/R900; (b) C850/R900; (c) C850/R1000; (d) C1100/R900; (e) C1100/R1000; I: ferromagnetic a-Fe; II: non-ferromagnetic c-Fe; III: Fe3C the same phase in the C850/R1000 powder is in line with the XRD results.The sextet accounting for a-Fe is the major component in all specimens and the proportion of the singlet representing c-Fe is in the 20–35% range.The intragranular Fe particles, which number extremely high in such materials, 42,45 are probably protected from alloying with carbon and crystallize in the stable a form. Moreover, surface particles that form an Fe–C alloy but grow to a large enough size undergo the c–a transformation upon cooling from the Fig. 4 SEM image showing the network of filaments on a grain of the C400/R1000 powder reduction temperature. The former cause could be preeminent Table 2 Room temperature Mo�ssbauer parameters of some nanocomposite powders [ferro: ferromagnetic; non-ferro: non-ferromagnetic; H: hyperfine field/kG; CS: center shift/mm s-1; C: half-line width/mm s-1; 2eQ: quadrupole shift/mm s-1, P: proportion (%) ferro a-Fe non-ferro Fe Fe3C specimen CS H C P CS C P CS H 2eQ C P C400/R900 0.00 328 0.16 73 -0.10 0.16 25 0.10 187 0.00 0.10 2 C850/R900 -0.01 333 0.20 49 -0.09 0.20 35 0.20 205 0.06 0.27 16 C850/R1000 -0.01 332 0.18 38 -0.09 0.19 20 0.19 205 0.03 0.21 42 C1100/R900 -0.02 328 0.22 42 -0.11 0.22 35 0.21 204 0.06 0.20 23 C1100/R1000 -0.01 329 0.21 55 -0.11 0.18 32 0.19 201 0.06 0.25 13 1266 J.Mater. Chem., 1998, 8(5), 1263–1272Fig. 5 High magnification SEM images of the nanocomposite powders prepared by calcination (C) and reduction (R) treatments at diVerent temperatures: (a) C400/R900; (b) C400/R1000; (c) C850/R900; (d) C850/R1000; (e) C1100/R900; (f ) C1100/R1000 [Fig. 6(a)]. Its inner and outer diameter are about 10 nm and observations of the same powder show several Iijima-type1 nanotubes [Fig. 6(d)], between 3 and 4 nm in diameter, made 50 nm respectively and the inside cavity is partly filled with elongated Fe and/or Fe3C particles. The short, thick filament up of two or three concentric layers (d002=0.34 nm) as well as Fe and/or Fe-carbide particles covered with three well crys- in the C400/R1000 powder [Fig. 6(b)] appears to be made up of several hollow forms, one of them containing an Fe and/or tallized graphene sheets [Fig. 6(e)]. An image of the C1100/R1000 powder [Fig. 6(f )] reveals a two shell carbon Fe3C particle. This filament is superimposed with a large Fe and/or Fe-carbide particle, and is connected to a matrix grain nanotube, 2.8 nm in diameter, superimposed with an alumina grain and Fe and/or Fe3C particles, the smaller ones (4–8 nm) by what seems to be a carbon thin film.In the C1100/R900 powder [Fig. 6(c)], we observe several bundles composed of being probably dispersed inside the oxide grain. A previous study42 on the selective reduction in H2 of Al2-2xFe2xO3 filaments less than 10 nm in diameter, a few nanometric irregularly shaped carbon species, and a lot of nanometric Fe (0<x0.2) has shown that the size and size distribution of the Fe particles formed upon reduction mainly depend on the and/or Fe-carbide particles, appearing black on the image.We have also observed that the spheroidal clusters (20–50 nm in mono- or biphasic nature (0<x0.1 or x>0.1, respectively) of the starting oxide and also from their specific surface area.diameter) detected by SEM, particularly in the C400/R1000 powder [Fig. 5(b)], are graphitic carbon nanoparticles,56 most Starting with a monophasic oxide such as the present Al1.8Fe0.2O3 compound favours a more monomodal size of them containing Fe and/or Fe-carbide particles. HREM J. Mater. Chem., 1998, 8(5), 1263–1272 1267Fig. 6 TEM (a–c) and HREM (d–f ) images showing diVerent forms of carbon revealed in the nanocomposite powders: thick tube in the C400/900 powder (a); thick, short filament in the C400/R1000 powder (b); bundles of nanotubes in the C400/900 powder (c); four nanotubes pointed by arrows in the C1100/R900 powder (d); Fe and/or Fe-carbide particles covered with three well crystallized graphene layers in the C1100/R900 powder (e); a nanotube partially superimposed with an alumina grain containing nanometric Fe and/or Fe3C particles in the C1100/R1000 powder (f ) distribution and a low specific surface area favours a smaller Al1.8Fe0.2O3 (C850) and a-Al1.8Fe0.2O3 (C1100) powders.The variations of S could result from the superimposition of two average size of the metal particles.phenomena: on the one hand, crystallization of the oxide and sintering of the primarains inside the agglomerates lead to Specific surface area measurements. Comparison of the specific surface areas (S) of the solid solutions (Sss; Table 1) a decrease of S when the reduction temperature is higher than the calcination temperature, i.e. when C400 and C850 powders with those of the nanocomposite powders (Sn; Table 1) shows that the reduction treatment induces a decrease of S when are reduced.On the other hand, the formation of carbon species, particularly carbon nanotubes and, to a much lesser amorphous starting powders (C400) are used, but that, in contrast, it induces an increase of S when starting from the g- degree, the formation of metal nanoparticles at the surface of 1268 J.Mater. Chem., 1998, 8(5), 1263–1272the matrix grains lead to an increase of S upon reduction of Ni-MgAl2O4 nanocomposite powders,43 it is inferred that the low-temperature weight gain (DTG peak in the 225–380 °C all the powders (C400, C850 or C1100). The former eVect outweighs the latter for the C400/R900 and C400/R1000 range) corresponds to the oxidation of the Fe and Fe-carbide particles located at the surface and in the open porosity of the powders; this is reversed for the C850/R900, C1100/R900 and C1100/R1000 composites.The two processes balance each oxide matrix. For the C1100/R900 and C1100/R1000 powders [Fig. 7(e), (f )], the high-temperature weight gain occurs in two other in the case of the C850/R1000 powder.This makes comparisons between the diVerent specimens steps (DTG peaks at 695 and 1165 °C, and 775 and 1180 °C respectively) as in Fe/Cr-Al2O357 nanocomposite powders. It diYcult and does not allow the derivation of precise data on the amount of nanotube formed upon reduction. Therefore, could similarly correspond to the oxidation of the intragranular metal particles. In contrast, for the other powders we oxidized the composite powders in air at 850 °C, in order to eliminate all carbon species, and we again measured the [Fig. 7(a)–(d)], the high-temperature weight gains are much smaller and only the first step is apparent (DTG peak in the specific surface area (Son; Table 1). In fact, as we will discuss below, TGA (Fig. 7) and carbon analyses of specimens oxidized 630–720 °C range).Two assertions can explain this point. Firstly, in the composite powders derived from the reduction at 850 °C (Con; Table 1) show that some carbon remains in some of these samples (C400/R900 or R1000, and C850/R900 of an amorphous Al1.8Fe0.2O3 solid solution (C400) or a poorly crystallized g-Al1.8Fe0.2O3 powder (C850), the Fe and Fe3C or R1000), but in an amount smaller than 10% of the initial content, which can be neglected.Since the oxidation tempera- particles are mainly located at the surface of the matrix grains and are thus oxidized at a low temperature. Secondly, a high- ture (850 °C) is lower than the reduction temperature (900 or 1000 °C), this treatment should not aVect the matrix and thus temperature weight loss takes place and probably masks the end of the first high-temperature weight gain.However it is Son is a good approximation of its specific surface area. We assume that the Fe2O3 particles formed upon oxidation of the obvious that above 1100 °C, the weight is constant [Fig. 7(a), (c)], which indicates that the second high-temperature weight surface Fe and Fe3C nanoparticles57 will be of a similar nanometric size and thus that the contribution of the oxide gain does not take place for these powders.The weight losses, which are supposed to represent the particles to Son will be roughly similar to that of the Fe and Fe3C nanoparticles to Sn. oxidation of carbon (free or combined with Fe), mainly occur between 350 and 650 °C and also between 850 and 1000 °C for As proposed elsewhere,19 the increase in specific surface area per gram of powder, DS=Sn-Son, essentially represents the some powders (but in a much smaller proportion) [Fig. 7(a)–(f )]. The DTG curves [Fig. 7( b), (d), (f )] show that quantity of nanotubes (more precisely of nanotube bundles) and the increase in specific surface area per gram of carbon, the first weight loss is generally made up of two or three steps which could represent the oxidation of diVerent forms of DS/Cn, can be considered as ‘quality’ data, a higher figure for DS/Cn denoting a smaller average tube diameter and/or more carbon. However, the steps are not well enough separated on the TGA and DTG curves to draw any conclusions.Further carbon in tubular form. Walker et al.58,59 have also reported specific surface areas in the 35–170 m2 g-1 range for carbon investigations are in progress to clarify this point.Comparisons can be made between the total weight loss filaments, which they correlate to a high degree of internal porosity because the geometrical surface area of their filaments measured in TGA and the carbon content (Table 1) on the one hand and between the total weight gain measured in TGA (100 nm in diameter and 1 mm long) does not exceed 15 m2 g-1.In contrast, the present electron microscopy observations have and the theoretical weight gain corresponding to the oxidation of all the Fe-containing species as Fe2O3 on the other hand. revealed important quantities of very long tubes of nanometric diameter, the geometrical surface area of which is of the order Weight losses are always smaller than the corresponding carbon contents, except for the C1100/R900 powder, because of several hundred m2 g-1.Other researchers60,61 have reported increases in specific surface area upon the catalytic gains and losses are generally superimposed. For the same reason, weight gains are always smaller than the theoretical formation of carbon nanofibers, which are in qualitative agreement with the present results. values, which have been calculated assuming that no residual Fe2+ or Fe3+ were present in the powders as indicated by On the one hand, whatever the initial solid solution (C400, C850 or C1100), DS is higher when the reduction is performed Mo�ssbauer spectroscopy.Note that for the C1100/R900 powder, a weight loss higher than the measured carbon content at 1000 °C than when it is performed at 900 °C, indicating that a higher CH4 supersaturation level enhances both the quantity could be considered as being a consequence of the cumulative errors on the two measurements.Thus, it appears that TGA of nanotubes and the carbon content (Cn; Table 1). On the other hand, for a given reduction temperature, there is no of the present composite powders can not give quantitative results for either the determination of the content of diVerent great diVerence in DS for the diVerent solid solutions (Table 1).Consequently, the ratio DS/Cn is much higher for C1100/R900 carbon forms or the reduction yield of the solid solutions. Interestingly, the high-temperature weight loss is detected or R1000 powders for which Cn is low; the C1100/R900 powder thus presents the higher quality figure (DS/Cn=155 m2 g-1).only for the C400- and C850-derived powders [Fig. 7(a)–(d)]. For the C1100-derived specimens [Fig. 7(e)–(f )], this weight Comparisons of DS and DS/Cn values between C850/R900 and R1000 or between C1100/R900 and R1000 powders (Table 1) loss either does not occur or is masked by the large weight gain arising from the oxidation of intragranular Fe particles. show that the increase in quantity obtained when the reduction temperature is increased from 900 to 1000 °C is detrimental to The carbon content was determined for each of the six powders after oxidation in air at 850 °C (Con; Table 1), the temperature the quality.at which the high-temperature weight loss phenomenon begins.A small quantity of carbon is measured for each of the four Thermogravimetric analyses. The oxidation of the nanocomposite powders was investigated by thermogravimetric analysis C400- and C850-derived powders, but it is much smaller than the observed high-temperature weight loss [Fig. 7(a), (c)]. For (TGA) in flowing air up to 1400 °C.The TGA curves [Fig. 7(a), (c), (e)] and the corresponding diVerential thermogravimetric the C1100/R900 and C11/R1000 powders, carbon was not detected, and therefore we infer that the weight loss is not (DTG) curves [Fig. 7( b), (d), (f )] show that the oxidation of the composites occurs in several steps, with both weight gains masked but rather does not occur. For the sake of comparison, TGA was performed on an and weight losses.For all powders, the first weight gain occurs at temperatures alumina powder previously calcined at 400 °C and then heattreated at 900 °C in the H2–CH4 gas mixture, as the C400/R900 lower than 500 °C and other gains at temperatures exceeding 600 °C. From the conclusions of previous works on the oxi- composite powder.Three steps are detected on the TGA and DGA curves (Fig. 8). At low temperature (DTG peak at dation of Fe/Cr-Al2O3 and Fe/Cr-Cr2O357 as well as Co- and J. Mater. Chem., 1998, 8(5), 1263–1272 1269Fig. 7 Thermogravimetry (TGA and DTG) curves measured in flowing air of the nanocomposite powders: C400/R900 and C850/R900 (a, b); C400/R1000 and C850/R1000 (c, d); C1100/R900 and C1100/R1000 (e, f ) 505 °C), we observe a small weight loss which probably of some carbon entrapped inside the alumina grains.It is noteworthy that the third step is more important for this accounts for the oxidation of crystallized form(s) of carbon located at the surface of the alumina grains. The second step powder than for the Fe-containing specimens, because no Fe particles can act as catalyst for CH4 decomposition and (DTG peak at 900 °C) is a weight gain, probably corresponding to an oxygen gain due to the establishment of alumina therefore preferentially interact with the resulting deposited carbon. In fact, the second and third phenomena may overlap stoichiometry during the beginning of the crystallization in the a-form.Indeed, before this oxidation, the thermal treatment in and thus the actual weight loss may be higher than determined by TGA.In spite of this, the total weight loss (5.93%) is higher the H2–CH4 gas mixture at 900 °C produced transition alumina, which is probably oxygen-deficient. The third step (DTG than the measured carbon content (5.49%). As pointed out above for the composite powders, the high-temperature weight peak at 1065 °C) is a weight loss, probably due to the oxidation 1270 J. Mater.Chem., 1998, 8(5), 1263–1272(C1100) is used, carbon seems to be obtained only as bundles of nanotubes, thin graphitic films at the surface of Fe and/or Fe3C particles, and combined with Fe in cementite. The consequences are a low carbon content (Cn; Table 1) and a surface of carbon which is rather high (DS; Table 1) and thus a rather high quality parameter (DS/Cn; Table 1). This could partly result from the fact that the metal nanoparticles formed upon reduction of the a solid solution, which act as catalyst for CH4 decomposition and possibly nanotube nucleation, are much smaller than when using amorphous or g solid solutions42 and therefore appear to strongly favour the catalytic deposition of carbon in the form of nanotubes, smaller than 10 nm in diameter and arranged in bundles several tens of micrometers long, corresponding to aspect ratios as high as 1000–10000.19 It should be noted that the macroscopic characteristics of the powders are repeatable provided the specific surface area of the starting solid solution is the same.Reducing a solid Fig. 8 Thermogravimetry (TGA and DTG) curves measured in flowing solution of a given structure but with a lower specific surface air of an alumina powder calcined at 400 °C and then heat-treated at area yields less carbon and thus lower values of Sn and DS. 900 °C in H2–CH4 gas mixture Influence of the reduction temperature. As shown by the above Mo�ssbauer spectroscopy results, 900 °C is a high enough loss is also higher than the carbon content measured, before this step occurs, in powders oxidized at 850 °C.A possible temperature, in the H2–CH4 atmosphere, to fully reduce to the metallic state the Fe3+ ions substituting in the diVerent explanation for these observations is that the flash combustion technique used to determine the carbon contents does not solid solutions.Thus the increase in both carbon content Cn and DS (Table 1) when the reduction temperature is raised allow the release of all the carbon entrapped inside the alumina grains, thereby decreasing its amount. from 900 to 1000 °C are not a consequence of a more complete reduction producing a higher number of catalytic particles, Similarly, in the composite powders in which the alumina matrix is not completely crystallized in the a-form (C400/R900 but result from the higher CH4 sursaturation level in the reducing atmosphere.However, a lower quality parameter and R1000, C850/C900 and R1000), some carbon is probably entrapped inside the transition-alumina grains and it is oxid- (DS/Cn; Table 1) is observed in the R1000 powders, notably when starting with an a-solid solution (C1100).ized only at high temperature. Thus a compromise has to be found between a high value of DS denoting a high quantity (C1100/R1000) and a high Influence of the form of the initial Al1.8Fe0.2O3 solid solution. These characterization results have shown that, depending on value of DS/Cn denoting a high quality (C1100/R900). Since a huge number of carbon nanotubes in the composite powder is the crystallization level of the initial Al1.8Fe0.2O3 solid solution, the reduction process in a H2–CH4 gas mixture can lead to desirable to reach a suYciently high volume fraction, suitable for a possible enhancement of the properties of materials made the formation of several carbon species in the composite powder. The microscopic observations can be directly corre- from these powders, a high reduction temperature (1000 °C or more) is to be preferred.We have yet to study the influence of lated with macroscopic results, particularly with specific surface area measurements. other synthesis parameters to enhance the quality of the obtained nanotubes. When amorphous (C400) or transition (C850) solid solutions are used, carbon species other than bundles of carbon nanotubes appear, including thick tubes, hollow carbon forms Conclusions and/or clusters of graphitic carbon nanoparticles, most of them containing Fe and/or Fe3C particles. The consequence is a An amorphous Al1.8Fe0.2O3 solid solution has been prepared from the precipitation and thermal decomposition of a mixed high carbon content (Cn; Table 1), but a moderate carbon surface (DS; Table 1) and thus a low quality parameter (DS/Cn; ammonium oxalate.Calcination in air of the amorphous oxide at the appropriate temperatures produced the g (cubic) and Table 1). Since these solid solutions have a relatively high specific surface area (Sss; Table 1), most of the Fe3+ ions are the stable a (corundum) forms of Al1.8Fe0.2O3. The three oxides have been reduced in a H2–CH4 gas mixture at 900 or at located at or near the surface of the grains and are easily reduced to the metallic state, giving rise to a very high number 1000 °C, giving rise to composite powders containing alumina, a- and c-Fe, Fe3C, and several forms of carbon, including of metal particles which in turn can easily coalesce at the surface of the oxide grains to form much larger particles.42 carbon nanotubes, thick tubes and clusters of graphitic carbon nanoparticles.Thus, the diameter of most of these metal particles may be too high for the catalytic formation of nanotubes, in agreement The powders have been studied using a combination of chemical analysis, X-ray diVraction, Mo�ssbauer spectroscopy, with Baker and Rodriguez12 who claim that only suYciently small (less than 20 nm in diameter) metal particles can lead to scanning and transmission electron microscopy, thermogravimatric analysis and specific surface area measurements.In nanotubes. Moreover, we have shown that upon heating in a H2–CH4 atmosphere, the crystallization of the amorphous or particular, we have made use of the specific surface area of carbon (DS, see text) as a representation of the quantity of g-Al1.8Fe0.2O3 solid solutions in some way provokes the entrapment ofbon within the oxide grains.This entrapped carbon nanotubes in a powder and we have considered the ratio of this value to the carbon content (DS/Cn) as a quality carbon, as well as other non-tubular forms of carbon, would probably deteriorate the mechanical properties of massive value, a higher figure for DS/Cn denoting a smaller average tube diameter and/or more carbon in tubular form.composites prepared from these powders. Consequently, amorphous (C400) or transition (C850) solid solutions can not The presence of CH4 in the reducing gas mixture has been found to markedly favour the reduction of the iron(III ) ions be retained as precursors of composite materials including carbon nanotubes.substituting for aluminium in the corundum lattice. The composite powders prepared from the amorphous or g solid In contrast, when a stable a-Al1.8Fe0.2O3 solid solution J. Mater. Chem., 1998, 8(5), 1263–1272 127123 D. Ugarte,Microsc.Microanal.Microstruct., 1993, 4, 505.solutions contain important quantities of non-tubular carbon, 24 L. A. Bursill, J. L. Peng and X. D. Fan, Philos. Mag. A, 1995, resulting in a poor value of the quality parameter DS/Cn. 71, 1161. Moreover, some carbon is entrapped within the oxide grains 25 S. J. Tans, M. H. Devoret, H. Dai, A. Thess, R. E. Smalley, upon the crystallization of alumina during the reduction step.L. J. Geerligs and C. Dekker, Nature (L ondon), 1997, 386, 474. This entrapped carbon is more stable than carbon nanotubes 26 A. Y. Kasumov, I. I. Khodos, P. M. Ajayan and C. Colliex, Europhys. L ett., 1996, 34, 429. with respect to oxidation in air and probably, more generally, 27 Y. Nakayama, S. Akita and Y. Shimada, Jpn. J. Appl. Phys., 1995, with respect to most thermal treatments, which may prove 34, L10.extremely detrimental if those powders were to be used for the 28 H. Dai, E. W.Wong and C. M. Lieber, Science, 1996, 272, 523. production of massive materials by sintering. 29 L. Langer, L. Stockman, J. P. Heremans, V. Bayot, C. H. Olk, In contrast, the composite powders prepared from the stable C. Van Haesendonck, Y. Bruynserade and J. P. Issi, J.Mater.Res., a-solid solution contain carbon essentially in the form of 1994, 9, 927. 30 X. K. Wang, R. P. H. Chang, A. Pataashinski and J. B. Ketterson, nanotubes, in line with a higher value of DS/Cn. This could J.Mater. Res., 1994, 9, 1578. partly result from the fact that the metal nanoparticles formed 31 D. H. Robertson, D. W. Brenner and J. W. Mintmire, Phys. Rev. upon reduction, which act as catalyst for CH4 decomposition B, 1992, 45, 12 592.and possibly nanotube nucleation, are smaller than when using 32 W. A. De Heer, W. S. Bacsa, A. Chatelain, T. Gerfin, R. Humfreyamorphous or g solid solutions. In these powders, most Baker, L. Forro and D. Ugarte, Science, 1995, 268, 845. nanotubes are less than 10 nm in diameter and they are 33 R. Seshadri, A. Govindaraj, H.N. Aiyer, R. Sen, G. N. Subbanna, A. R. Raju and C. N. R. Rao, Curr. Sci., 1994, 66, 839. arranged in bundles several tens of micrometers long. 34 S. B. Sinnott, C. T. White and D. W. Brenner, Mater. Res. Soc. Starting from an a-solid solution, the increase in reduction Symp. Proc., 1995, 359, 241. temperature from 900 to 1000 °C produces an increase in the 35 M. M. J.Treacy, T. W. Ebbesen and J. M. Gibson, Nature amount of nanotubes, probably owing to the higher CH4 (L ondon), 1996, 381, 678. supersaturation level in the reduction atmosphere. However, a 36 S. Iijima, C. Brabec, A. Maiti and J. Bernholc, J. Chem. Phys., 1996, concomitant decrease in tube quality is also observed, pointing 104, 2089. 37 J. F. Despres, E. Daguerre and K. Lafdi, Carbon, 1995, 33, 87.out the need for further studies. 38 K. Tohji, T. Goto, H. Takahashi, Y. Shinoda, N. Shimizu, B. Jeyadevan, I. Matsuoka, Y. Saito, A. Kasuhka, T. Oshuna, K. Hiraga and Y. Nishina, Nature (L ondon), 1996, 383, 679. References 39 R. T. K. Baker, P. S. Harris, R. B. Thomas and R. J. Waite, J. Catal., 1993, 30, 86. 1 S. Iijima, Nature (L ondon), 1991, 354, 56. 40 A.Oberlin, M. Endo and T. Koyama, J. Crystal Growth, 1976, 2 T.W. Ebbesen and P. M. Ajayan, Nature (L ondon), 1992, 358, 220. 32, 335. 3 S. Iijima and T. Ichihashi, Nature (L ondon), 1993, 363, 603. 41 K. Hernadi, A. Fonseca, J. B. Nagy, D. Bernaerts, A. Fudala and 4 D. S. Bethune, C. H. Kiang, M. S. de Vries, G. Gorman, R. Savoy, A. A. Lucas, Zeolites, 1996, 17, 416. J. Vazquez and R.Beyers, Nature (L ondon), 1993, 363, 605. 42 X. Devaux, Ch. Laurent and A. Rousset, Nanostruct. Mater., 1993, 5 C. H. Kiang, W. A. Goddard, R. Beyers, J. R. Salem and 2, 339. D. Bethune, J. Phys. Chem. Solids, 1996, 57, 35. 43 O. Que�nard, Ch. Laurent, M. Brieu and A. Rousset, Nanostruct. 6 S. Seraphin and D. Zhou, Appl. Phys. L ett., 1994, 64, 2087. Mater., 1996, 7, 497. 7 T. W. Ebbesen, H. J. Lezec, H. Hiura, J. W. Bennett, H. F. Ghaemi 44 Ch. Laurent, A. Rousset, M. Verelst, K. R. Kannan, A. R. Raju and and T. Thio, Nature (L ondon), 1996, 382, 54. C. N. R. Rao, J.Mater. Chem., 1993, 3, 513. 8 C. Guerret-Plecourt, Y. Le Bouar, A. Loiseau and H. Pascard, 45 M. Verelst, K. R. Kannan, G. N. Subbanna, C. N. R. Rao, Ch. Nature (L ondon), 1994, 372, 761. Laurent and A.Rousset, J.Mater. Res., 1992, 7, 3072. 9 A. Thess, R. Lee, P. Nikolaev, H. Dai, P. Petit, J. Robert, 46 X. Devaux, Ch. Laurent, M. Brieu and A. Rousset, J. Alloys D. T. Colbert, C. Xu, Y. H. Lee, S. G. Kim, A. G. Rinkler, Compd., 1992, 188, 179. D. T. Colbert, G. E. Scuseria, D. Tomanek, J. E. Fisher and 47 A. Marchand, X. Devaux, B. Barbara, P. Mollard, M. Brieu and R.E. Smalley, Science, 1996, 273, 483. A. Rousset, J.Mater. Sci., 1993, 28, 2217. 10 S. Witanachchi and P. Mukherjee, J. Vac. Sci. T echnol. A, 1995, 48 A. Rousset and J. Paris, Bull. Soc. Chim. Fr., 1967, 10, 388. 3, 1171. 49 D. Wagmann, J. E. Kilpatrick, W. J. Taylor, K. S. Pitzer and 11 M. J. Yacaman, M. M. Yoshida, L. Rendon and J. G. Santiesteban, F. D. Rossini, J. Res. Natl. Bur. Stand., 1945, 34, 143. Appl. Phys. L ett., 1993, 62, 657. 50 Ch. Laurent, J. J. Demai, A. Rousset, K. R. Kannan and 12 R. T. K. Baker and N. Rodriguez, Mater. Res. Soc. Symp. Proc., C. N. R. Rao, J.Mater. Res., 1994, 9, 229. 1994, 349, 251. 51 H. L. Yakel, Int. Met. Rev., 1985, 30, 17. 13 S. Herreyre and P. Gadelle, Carbon, 1995, 33, 234. 52 G. Le Cae�r, J. M. Dubois and J. P. Se�nateur, J. Solid State Chem., 14 V. Ivanov, A. Fonseca, J. B. Nagy, A. Lucas, P. Lambin, 1976, 19, 19. D. Bernaerts and X. B. Zhang, Carbon, 1995, 33, 1727. 53 X. X. Bi, B. Ganguly, G. P. HuVman, F. E. Huggins, M. Endo and 15 K. Hernadi, A. Fonseca, J. B. Nagy, D. Bernaerts, J. Riga and P. C. Eklund, J. Mater. Res., 1993, 8, 1666. A. Lucas, Synth.Met., 1996, 77, 31. 54 O. Que�nard, E. De Grave, Ch. Laurent and A. Rousset, J. Mater. 16 A. Fonseca, K. Hernadi, J. B. Nagy, Ph. Lambin and A. Lucas, Chem., 1997, 7, 2457. Carbon, 1995, 33, 1759. 55 R. T. K. Baker, J. R. Alonzo, J. A. Dumesic and D. J. C. Yates, 17 M. Endo, K. Takeuchi, K. Kobori, K. Takahashi, H. W. Kroto and J. Catal., 1977, 74, 82. A. Sarkar, Carbon, 1995, 33, 873. 56 Y. Saito, Carbon, 1995, 33, 979. 18 A. Thess, R. Lee, P. Nikolaev, H. Dai, P. Petit, J. Robert, 57 Ch. Laurent, Ch. Blaszczyk, M. Brieu and A. Rousset, Nanostruct. D. T. Colbert, C. Xu, Y. H. Lee, S. G. Kim, A. G. Rinkler, Mater., 1995, 6, 317. D. T. Colbert, G. E. Scuseria, D. Tomanek, J. E. Fisher and 58 P. L. Walker, Jr., J. F. Rakszawski and G. R. Imperial, J. Phys. Chem., 1959, 63, 133. R. E. Smalley, Science, 1996, 273, 483. 59 P. L. Walker, Jr., J. F. Rakszawski and G. R. Imperial, J. Phys. 19 A. Peigney, Ch. Laurent, F. Dobigeon and A. Rousset, J. Mater. Chem., 1959, 63, 140. Res., 1997, 12, 613. 60 N. M. Rodriguez, M.-S. Kim and R. T. K. Baker, J. Phys. Chem., 20 X. F. Zhang, X. B. Zhang, G. Van Tendeloo, S. Amelinckx, M. Op 1994, 98, 13 108. De Beeck and J. Van Landuyt, J. Crystal Growth, 1993, 130, 368. 61 W. B. Downs and R. T. K. Baker, J.Mater. Res., 1995, 10, 625. 21 P. J. Harris, Eur. Microsc. Anal., 1994, 9, 13. 22 G. Hu, X. F. Zhang, D. P. Yu, S. Q. Feng, W. Xu and Z. Zhang, Solid State Commun., 1996, 98, 547. Paper 7/06726G; Received 16th September, 1997 1272 J. Mem., 1998, 8(5), 1263–1272