The catalyst in the CCVD of carbon nanotubes—a review

Progress in Materials Science 50 (2005) 929–961 www.elsevier.com/locate/pmatsci

The catalyst in the CCVD of carbon nanotubes—a review Anne-Claire Dupuis Commissariat a` l
Energie Atomique, DRFMC, 17 rue des Martyrs, F-38054 Grenoble Cedex 9, France

Abstract Since their discovery [Ijima, 1991, Nature, 354, 56], carbon nanotubes (CNTs) have been widely studied due to their large potential applications. First produced in arc-discharge process or by laser-ablation, the CNTs grown by catalytic chemical vapor deposition (CCVD) have been showing however a large expansion for the past decade. A fundamental question remains after this 10-year experience: What is actually the role played by the catalyst in the CCVD of CNTs? This review intends to synthesize the data published in the scientific literature on this topic in order to better understand the parameters governing the catalytic properties of the metal nanoparticles. In particular, we will discuss the influence of the composition of the catalyst material, of the morphology of the catalyst nanoparticles, of the support, of the preparation method of the nanoparticles and of the reduction pretreatment. Ó 2005 Elsevier Ltd. All rights reserved.

Contents 1. 2.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Preparation methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. Sol–gel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2. Coreduction of precursors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3. Impregnation, incubation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4. Ion-exchange–precipitation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

E-mail address: [email protected] 0079-6425/$ - see front matter Ó 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.pmatsci.2005.04.003

930 932 933 933 933 933

930

3. 4. 5.

6.

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

2.5. Ion-adsorption–precipitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.6. Reverse micelle method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.7. Thermal decomposition of carbonyl complexes . . . . . . . . . . . . . . . . . . . . 2.8. Metalorganic chemical vapor deposition (MOCVD) . . . . . . . . . . . . . . . . . 2.9. Physical deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Models of catalytic growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Catalysts used for CNT growth. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Parameters determining the catalytic properties . . . . . . . . . . . . . . . . . . . . . . . . 5.1. Composition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1.1. Experimental comparison of different catalysts . . . . . . . . . . . . . .. 5.1.2. Electronic structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5.1.3. Carbon solubility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5.1.4. Stabilization of the catalyst . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5.1.5. Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5.2. Morphology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.1. Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5.2.2. Crystallographic orientation . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5.2.3. Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5.3. Preparation method. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.1. Morphology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.2. Electronic structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5.3.3. Other aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5.3.4. Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5.4. Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4.1. Experimental comparison of different supports . . . . . . . . . . . . . .. 5.4.2. Formation of intercompounds . . . . . . . . . . . . . . . . . . . . . . . . . .. 5.4.3. Support–catalyst interactions . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5.4.4. Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5. Pretreatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5.1. Experimental results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5.5.2. Role of pretreatment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5.5.3. Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

933 944 944 944 944 935 937 938 938 939 940 941 942 943 943 943 945 945 945 946 948 948 949 949 949 950 950 952 952 952 952 953 953 954 955

1. Introduction Carbon products in tubular form, called carbon filaments, were first observed when electron microscopes came into wide use around 1950 [1]. In the early 90s, such filaments were observed with a diameter in the range order of the nanometer [2] and have then been called carbon nanotubes (CNTs). These first observed multi-walled CNTs (MWNTs) were grown in an arc-discharge process. A transmission electron microscope image of MWNT is shown in Fig. 1. Two years later, single-walled carbon nanotubes (SWNTs) could be grown by laser-ablation. At the same time,

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

931

Fig. 1. Transmission electron microscope image of a multi-walled carbon nanotube [11].

the catalytic chemical vapor deposition (CCVD) method was first used to grow CNTs. Because of the peculiar electronic transport properties of the single-walled carbon nanotubes (SWNTs) on the one hand [3,4] and the high conducting properties of the multi-walled carbon nanotubes (MWNTs) on the other hand [5], CNTs arouse great interest in the microelectronics community and hope for new functional electronic devices [6–9]. The attraction and development of CCVD for growing CNTs by the scientific community can be partly explained by the fact that it is very familiar with CVD techniques. However, the ability to grow CNTs directly on a substrate at a desired position is a great challenge from a technological point of view. This control over the CCVD growth of CNTs would permit the integration of the CNT growth into fabrication processes of microelectronic circuits since the CCVD process requires moreover much lower temperatures than the arc-discharge and laser-ablation processes. The CCVD methods thus seems to be more adapted for large-scale production at lower cost [10]. This explains why many research groups in the whole world are working on the growth of CNTs by CCVD. The CVD process on a supported catalyst consists of several steps. The first is to prepare metal nanoparticles on a substrate. The substrate is then placed in a furnace and the nanoparticles are then generally submitted to a reduction treatment upon heating under typically H2 or NH3. And finally, hydrocarbon gas or CO is let into the furnace and carbon deposition occurs by catalytic decomposition of the hydrocarbon molecules on the metal nanoparticles by temperatures ranging roughly from 500 to 1200 °C (Fig. 2). A fundamental question remains, even if the CCVD of CNTs works properly: What is actually the role played by the catalyst in the catalytic chemical vapor

932

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

Fig. 2. Schematic illustration of the CVD process.

deposition of carbon nanotubes? One could simply answer, the first role of the catalyst is to decompose the hydrocarbon molecules. Surely it is. But its activity for hydrocarbon decomposition cannot be accounted for its ability of CNT formation [12,13]. In many cases, hydrocarbon molecules will decompose without forming CNT, as has been found for example by Sen et al. [14] who studied the decomposition by pyrolysis of benzene and/or metallocene. One may easily concede that the role of the catalyst is more complex than only hydrocarbon decomposition. But what happens exactly at the catalyst during the growth process? Why are transition metals appropriate for CNT growth? Why do different metals and combinations of them lead to sometimes completely different results? This review does not intend to answer fully all these questions, but to synthesize what is known today about the catalyst in the growth of CNTs. Because of the very large number of papers about carbon nanotubes, the review cannot be too exhaustive, but may help to explain how to better understand and thus control the parameters governing the catalytic properties of the metal nanoparticles. Since the catalyst is one essential key in the CCVD process, improving it should improve the quality of the obtained CNTs. In Section 2, we will see what are the principal methods used to prepare catalyst nanoparticles. As will be discussed in Section 5, there is certainly a correlation between the size of the nanoparticles and the diameter of the resulting CNTs. The preparation method therefore must fulfill strong requirements on the geometry of the nanoparticles. In Section 3, a synthesis is given of what is known about CNT growth mechanisms. In Section 4, we review the results obtained on most of the catalysts used in the last decade for CNT growth. They are principally based on iron, cobalt and nickel. We then try in Section 5 to discuss the results found in the literature in order to determine what are the important parameters of the catalyst on which the results of CNT growth depend.

2. Preparation methods It has been found that transition metals such as iron, cobalt and nickel are catalysts for the growth of CNT. However, bulk iron as itself, for example, is not able to catalyze the decomposition of methane to form carbon filaments: it has to be first

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

933

dispersed [15]. In order to obtain nanotubes, one therefore has to find methods to prepare catalyst nanoparticles. 2.1. Sol–gel In the heterogeneous sol–gel method, a porous precursor of the active component is impregnated with the precursor of a textural promoter taken in the estimated amount to have a given weight ratio between active component and textural promoter in the end product. The role of the textural promoter is to stabilize the active component structure and to prevent its sintering in the course of posttreatments [15]. One generally uses hard-to-reduce oxides (HRO) such as silica or alumina as textural promoters. To get iron/silica nanocomposite particles, one can mix e.g. tetraethoxysilane (precursor of textural promoter) with iron nitrate (precursor of the active component) aqueous solution and ethanol. The mixture is then gelated, dried to remove the excess water and solvent and finally calcinated [16]. 2.2. Coreduction of precursors Nitrates of a catalyst and of a metal oxide support, e.g. Co- or Ni(NO3)2 Æ 6H2O and Mg(NO3)2 Æ 6H2O, are mixed with an organic compounds like urea or citric acid and water [17–19]. Heating the mixture leads to reduction of the precursors to form intimately mixed oxide particles. 2.3. Impregnation, incubation The impregnation method consists in first dissolving a catalyst precursor (e.g. ironoxalate [20]) and then contacting a support with this solution. In this method, the whole precursor deposits onto or into (in case of a porous material) the substrate. The solvent is then evaporated and the catalyst dried [21]. In a case of organic supports and molecules, one speaks about incubation [22], but the method remains the same. 2.4. Ion-exchange–precipitation In this preparation method, a solution of a catalyst precursor (e.g. cobalt-acetate [23] or cobalt-nitrate [20]) is also used and brought in contact with a zeolite support. But in this case, the anion of the precursor is exchanged with an anion of the zeolite, giving a new precursor molecule. Calcination allows the thermal decomposition of the catalyst precursor and gives the catalyst element in an oxidized form. 2.5. Ion-adsorption–precipitation In the ion-adsorption–precipitation method, the support is put in a catalyst precursor solution (e.g. Co(H3C–CO2)2 Æ 4H2O [24]). An acid–base reaction takes place at the surface of the support, leading to precipitation of the catalyst precursor

934

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

(cobalt–acetate in the example above). The reaction can be controlled by the pH of the initial catalyst precursor solution. The sample is then calcinated to get only the catalyst element in an oxidized form. 2.6. Reverse micelle method In the reverse micelle method, a cationic surfactant is dissolved, e.g. in toluene. First metal salt is added to the solution and afterwards a reductor agent in order to reduce the oxidized metal to its neutral form. It results in a colloidal dispersion of metal nanoparticles [25]. The nanoparticles are then purified and the final dispersion can be cast onto a substrate and dried at room temperature. 2.7. Thermal decomposition of carbonyl complexes Metal can be synthesized in form of nanoclusters by thermal decomposition of its carbonyl complex. The procedure consists in mixing the carbonyl complex in an ether solution and adding protective agents to prevent the nanoparticles from aggregation (e.g. octanoic acid [26]). The solution is then refluxed at a temperature allowing the decomposition of the precursor (e.g. 286 °C [27]). In [27], Cheung et al. reported the obtention of nanocluster solutions with distinct and nearly monodisperse diameters of 3.2, 9 and 12.6 nm for three different protective agents used, respectively. 2.8. Metalorganic chemical vapor deposition (MOCVD) A metalorganic precursor (e.g. iron pentacarbonyl [21]) is vaporized and carried to the reactor zone by a carrier gas. Heating of the reactor zone allows the precursor to decompose and to depose onto the substrate. 2.9. Physical deposition Metal can be evaporated or sputtered to be deposited onto a substrate. The metal, if deposited at room temperature, will generally be amorphous and form a more or less smooth film on the surface of the substrate. Upon annealing, the equilibrium shape may be reached. To know this shape, the Young
s equation describing a contact between two phases A and B and the ambient atmosphere has to be considered: cA ¼ cAB þ cB cos h

ð1Þ

with c the corresponding interface energies, see Fig. 3. The physical deposition of B on a substrate A is largely used for epitaxy of B on A when the following condition yielding layer growth (h = 0) is fulfilled [28]: cB þ cAB < cA

ð2Þ

When Eq. (2) is not satisfied, B deposits on A in the form of islands (Volmer–Weber growth) as shown in I and II of Fig. 3.

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

935

Fig. 3. Conditions on the surface energies of substrate (A), deposit (B) and interface A–B in determining the type of growth: island or Volmer–Weber growth for I (non-wetting) and II (wetting) and layer-growth for III.

The property of island growth is required for obtaining nanoparticles on a substrate by physical deposition. After deposition usually performed at room temperature, annealing allows the atoms to move and reach the energetical most favored configuration. This method has been used to obtain nanoparticles of nickel, cobalt, iron or alloys of them to grow CNT. It has been found that the size of the particles (typically 10–100 nm) directly depends on the thickness of the deposited material [29–32]. For breaking up the thin film obtained after deposition, many authors reported the use of NH3 [29,33–35] or H2 [36] during annealing. The surface energies cA and cB used above are defined relative to the atmosphere gas and thus a change in the gas can dramatically change the final shape of B on A. 3. Models of catalytic growth This review focuses on the catalytic growth of CNTs by CVD. The growth process may be quite different in the CVD process compared with the arc discharge and laser-ablation methods and we shall not discuss the possible differences here. Because of the analogy between CNTs grown by CVD on a supported catalyst and carbon filaments grown also by CVD on a supported catalyst, we will first discuss the growth model for carbon filaments. The carbon filament growth model generally adopted is based on the concepts of the VLS (vapor–liquid–solid) theory developed by Wagner and Ellis [37]. In this model, molecular decomposition and carbon solution are assumed to occur at one

936

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

side of the catalytic particle, which then becomes supersaturated [1]. Carbon diffuses from the side where it has been decomposed to another side where it is precipitated from solution [38]. It is not clear at this point of the model which is the driving force for carbon diffusion within the catalytic particle. For many authors, this force originates from the temperature gradient created in the particle by exothermic decomposition of the hydrocarbon at the exposed front faces and endothermic deposition of carbon at the rear faces, which are initially in contact with the support face [38,39]. Others ascribed the driving force to a concentration gradient [1,40]. However, because of a much lower surface energy of the basal planes of the graphite compared with the prismatic planes, it is energetically favorable for the filament to precipitate with the basal planes as the cylindrical planes [1]. The metal–support interactions are found to play a determinant role for the growth mechanism [38,41]. Weak interactions yield tip-growth mode whereas strong interactions lead to base-growth. Both growth modes are schematically shown in Fig. 4. This growth model for carbon filaments has been widely used for carbon nanotubes. However, the specificity of the growth of nanotubes on nanoparticles with regard to the growth of carbon filaments is their nanometer dimensions. The energetic argument of the precipitation of carbon on its low-energy basal planes is for example no longer valid since the curving of the graphite layers introduces an extra elastic term into the free-energy equation of nucleation and growth, leading to a lower limit of about 10 nm [1]. Other mechanisms are therefore necessary to explain the growth of CNTs, whose diameter can be much smaller than this lower size limit [42]. In the catalytic growth of CNTs, not the ‘‘fluid nature’’ of the metal particle as in the VLS model has to be considered but the chemical interactions between the transition metal 3d electrons and the p carbon electrons [43]. Nanoparticles exhibit a very high surface energy per atom. The carbon in excess present during CVD process could

Fig. 4. The two growth modes of filamentous carbon.

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

937

solve this energetic problem by assembling a graphene cap on the particle surface with its edges strongly chemisorbed to the metal. Since the basal planes of graphite has an extremely low surface energy, the total surface energy diminishes. This is the Yarmulke mechanism proposed by Dai et al. [44]. The hemifullerene cap so formed on the partially carbon-coated particle lifts off and additional carbon atoms are continuously added to the edge of the cap, forming a hollow tube with constant diameter which grows away from the particle [45]. The driving force for the lifting process is believed to originate from the free energy release due to relaxation of the strain built up in the carbon cap around the spherical surface of the catalyst nanoparticle when carbon fragments assemble to form a CNT [46,47]. Growth of SWNTs in the arc-discharge and laser-ablation methods has been already widely discussed from a theoretical point of view. Molecular dynamics and total energy calculations revealed the basic atomic process by which singleshelled nanotubes can grow out of metal-carbide particles by the root growth mechanism [48]. In contrast, almost no theoretical work could be found on the CCVD growth of CNTs until the recent publication of Shibuta and Maruyama [49]. The authors performed molecular dynamics simulation of SWNT growth by CCVD and showed that carbon atoms are absorbed into the metal cluster until saturation to make a hexagonal network of carbon atoms inside the surface of the cluster. Further supply of carbon then leads to a separation of the carbon network from the metal surface itself. Furthermore, carbon atoms inside the supersaturated nickel cluster gradually lift up the carbon-shell. This described process can be regarded as an initial stage of the growth process of SWNT and at any rate confirms the Yarmulke mechanism proposed by Dai et al. [44]. A recent paper from Helveg et al. [50] deals with a high-resolution in situ transmission electron microscope observation of catalytic tip-growth of CNTs by decomposition of methane over a nickel-based catalyst. The nickel particles are observed first to elongate with simultaneous formation of graphene sheets with their basal planes oriented parallel to the nickel surface. A contraction follows, explained by the fact that the increase in the nickel surface energy can no longer be compensated for by the energy gained when binding the graphitic tube to the nickel surface. This study furthermore reveals that monoatomic steps are present at the nickel surface and that a graphene sheet terminates at each of these steps. Between a pair of such step edges, an additional graphene layer grows as the nickel steps move towards the ends of the nickel cluster and vanish, involving transport of carbon atoms towards and nickel atoms away from the graphene–nickel interface. Note that this proposed growth process, confirmed with density-functional theory calculations performed by the authors, concerns a tip-growth mechanism. It therefore is not really surprising that it clearly differs from the Yarmulke mechanism presented above, valid for base-growth processes.

4. Catalysts used for CNT growth Metals used to catalyze CNT formation are most often transition metals, in particular iron, cobalt and nickel. A very large number of papers can be found in the

938

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

literature reporting CNT growth. We will cite a few of them in the following, according to the catalyst used, so that the reader can have a short and quick overview of the general results obtained, especially as regards MWNTs and SWNTs. Most of the papers available in the literature about growth of CNTs on ironbased catalysts report growth of MWNTs [12–16,20,21,23,24,33,36,51–62]. Many groups also obtained SWNTs [10,22,27,45,63–68]. Most of the reported results obtained with cobalt-based catalysts concern formation of MWNTs [20,23– 25,34,56,57,69–74], a few others SWNTs [68,75–77]. Nickel-based catalysts generally lead to MWNTs [20,30–32,57,78–88], seldom to SWNTs [47,76,77]. Mixtures of transition metals are often observed to be more efficient for CNT production than one metal alone. Iron-nickel alloys are found to produce MWNTs [89–91], iron– cobalt alloys seem to produce rather SWNTs [92–96] and Dai et al. obtained SWNTs with a nickel–cobalt alloy [44]. Other authors studied several alloy catalysts and mainly obtained SWNTs [12,76,77,97,98]. Palladium seems to be an interesting metal for connecting carbon nanotubes since the contact between this metal and metallic CNTs is ohmic in nature [99]. The catalyst nanoparticle would therefore not disturb the electronic transport properties of the nanotube by forming a Schottky barrier. Nevertheless, palladium has been used a little as a catalyst for CNT formation, as described in [100–102]. These three papers report growth of MWNTs. Other metals than iron, cobalt, nickel or palladium have been used as cocatalyst. These metals are generally but not necessarily non-active catalysts alone but at any rate are used to improve the performance of the ‘‘classical’’ catalysts when added to them in different quantities. Molybdenum is the most important, added to iron [26,41,103–111] or to cobalt [29,112–121]. Except for in [119], all these papers report formation of SWNTs with iron–molybdenum or cobalt–molybdenum catalysts. Magnesium–nickel oxides have been used as catalysts for MWNT growth [17,122,123] and magnesium–cobalt oxides for SWNT growth [124,125,121].

5. Parameters determining the catalytic properties The peculiar ability of transition metals such as iron, cobalt and nickel to catalyze CNT formation is mostly linked to their catalytic activity for the decomposition of carbon compounds, their ability to form carbides and the possibility for carbon to diffuse through and over the metals extremely rapidly ([126] and references therein). Is this sufficient to explain their performances in CNT growth? In this section, we want to analyze the experimental results found in the literature in order to track down the influence of five aspects of the catalyst nanoparticles on the results obtained by CNT growth, namely: their composition, their morphology, their preparation method, their support and their pretreatment. 5.1. Composition As seen in Section 4, the three usual catalysts for CNT synthesis are iron, cobalt and nickel, all of them 3d-metals. If they all catalyze CNT growth reaction, they do

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

939

not yield exactly the same results. In this section, we will first summarize some of the results found in the literature where different catalysts were directly compared. We will then discuss the influence of the composition nature of the catalyst on three important properties for the catalyst activity: electronic structure, carbon solubility and stabilization of the catalyst material. 5.1.1. Experimental comparison of different catalysts Klinke et al. tested iron-, cobalt- and nickel-based catalyst on silica for CNT growth with acetylene. Iron produced the highest density of carbon structures at any considered temperature between 580 and 1000 °C [57]. Hernadi et al. tested ironand cobalt-based catalysts with different hydrocarbons on different supports and observed that iron/silica presents the highest activity in the decomposition of different unsaturated compounds [56]. Kong et al. came to the same conclusion [97]. However, an improvement of the quality of the CNT was obtained on silica by using cobalt [56]. The observations of Fonseca et al. [24] summarize the aforementioned results: iron is more active than cobalt but the quality of the resulting CNTs, in terms of graphitization and structure, is less good with iron. Ivanov et al. also observed a better graphitization on cobalt than on iron, but on nickel most of the filaments were amorphous [20]. They also tried with copper but found only amorphous carbon. The large majority of groups working with nickel-based catalyst interestingly reports growth of MWNTs, many of them with bamboo structure, or fibers. In [13], iron-based catalysts prepared by the sol–gel method with different iron contents were tested. The authors found out that MWNTs were obtained with acetylene for all tested catalysts but in different quantities and above all the type of the carbonaceous products strongly depends on the iron quantity in the catalyst. A catalyst with only 2.3 wt% iron exhibits MWNTs in only very small amounts. Increasing the iron amount to 4.4 wt% increases the amount of CNTs observed. For both catalysts, no fibers or carbon encapsulation are observed. Conversely, the catalyst with 8 wt% iron yields a large amount of amorphous carbon besides CNT. This is even worse for the 28.5 wt% iron containing catalyst with which coiled structures, carbon nanofibers and carbon encapsulated particles were also observed. Mixtures of two metals as catalyst had already been observed to dramatically change the catalyst performances for carbon filament growth ([38] and references therein). The same is observed for CNT growth when two of the transition metals iron, cobalt and nickel are mixed [76,127] or when another metal is added to one of them [29,41,107,115,116,128]. Mixing of two or more metals yields improvement of the catalyst performance in terms of quality of obtained product or of lowering of reaction temperature. Platinum, palladium and chromium were e.g. added to a cobalt–nickel catalyst in order to lower the growth temperature from 700–1000 °C to 500–550 °C (with different success rates) [127]. Harutyunyan et al. [107] found that unreduced iron was active for CNT growth only from 900 °C whereas it becomes active even at 680 °C with addition of 20% molybdenum. Terbium, in combination with iron, is shown to improve the uniformity and purity of the MWNTs obtained by pyrolysis of xylene [128].

940

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

5.1.2. Electronic structure The first action of the catalyst is to bond the hydrocarbon molecules at its surface. The hydrocarbon, now in an adsorbate state, interacts with the catalyst by transferring an amount of its electron density to the catalyst. This electronic interaction between a ‘‘donor’’ adsorbate and an ‘‘acceptor’’ catalyst metal is schematically illustrated in Fig. 5. Generally, simultaneous back-donation takes place, i.e. electron transfer from the catalyst to the non-occupied, antibonding orbitals of the adsorbate molecule. The electronic structure of the adsorbate is then changed in such a manner that dissociation of the molecule can occur. The transition metals have non-filled d shells and are for that reason able to interact with hydrocarbons and show catalytic activity. More precisely, properties contributing to the ability to make and break adsorbate bonds are: (i) the center of the d-bands, (ii) the degree of filling of the d-bands and (iii) the coupling matrix element between the adsorbate states and the metal d-states [129]. Therefore, the ability of a metal to catalyze dissociation of a hydrocarbon molecule is intimely linked to its electronic structure. This can explain why iron is found to be more efficient than nickel and cobalt by hydrocarbon decomposition [24,56,57,97]. Some papers also exemplify that an added component can lower the activation energy for dissociation, and thus the growth temperature, by changing the electronic structure of the catalyst [107,127]. Nevertheless, it is not clear whether differences in electronic structure between different catalyst materials can account for the observed differences in the quality of MWNT in terms of graphitization. Copper, a non-transition metal with its 3d shell completely filled, was

Fig. 5. Schematic illustration of electronic interactions in chemisorption. A filled orbital on the adsorbate overlaps with an empty one on the metal. For the sake of clarity, discrete levels of the metal are represented rather than electron bands.

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

941

observed to yield only amorphous carbon [20]. In addition, the growth model presented in Section 3 supposes formation of a carbon cap on the metal particle with its edges strongly chemisorbed to the metal particle [44]. One may therefore speculate that overlap of metal 3d empty orbitals with carbon valence orbitals plays not only a role in the hydrocarbon dissociation process, but also an essential part in the initial stage of the CNT growth process, since it allows chemisorption of the carbon cap edges to the catalyst nanoparticle. Besides, Kim et al. studied the formation of carbon filament structures on copper–nickel catalysts and found that the wetting behavior of the metal on graphite is extremely sensitive to the chemical nature of the particles [130]. Kim et al. namely supposed that metals that readily wet graphite will produce highly ordered carbon filament structures. As seen in Section 2.9, wetting behavior depends on the relevant interfacial energies. Roughly, two materials will wet if their interfacial energy is low, in other words, if they exhibit high interactions. These interactions can originate e.g. from Van der Waals forces but also from overlap of orbitals (chemisorption). One sees that both aspects just discussed may combine. However, it should be further investigated if and how the unfilled 3d-metal orbitals have an effect on the growth process and thus on the graphitization of the whole CNT. 5.1.3. Carbon solubility The finite solubility for carbon of 3d-metals in certain temperature ranges [131] may play an important part in the growth process, since, once the hydrocarbon molecules are broken, carbon atoms are believed to diffuse into the catalyst particle, leading to supersaturation of carbon in the metal, as explained in Section 3. This process can eventually yield formation of carbides. The role of solubility of carbon in the metal catalyst is subject to controversy in the literature. Kock et al. e.g. claimed that high carbide contents are required before nucleation of carbon filaments proceeds [132]. Fe3C has been detected during CNT growth process on iron nanoparticles and it was suggested that it may be the real catalyst [13,15,133]. However, Herreyre et al. [134] studied the evolution of iron catalysts during disproportionation of CO. They indeed detected formation of Fe3C but identified this phase as the only one present when the catalyst was entirely deactivated. They therefore attributed this phase to the cause of catalyst poisoning. Yoshida et al. [135] and Arie et al. [136] also identified the catalyst particle at the end of nanotubes to be Fe3C and Ni3C, starting from an iron and nickel catalyst, respectively. Hernadi et al. moreover tested Fe3C nanoparticles and could not observe any catalytic activity [55]. This last result does not seem to us a paradox since the formation of a metal carbide probably is an intermediate step in the whole growth process. In the VLS growth model, hydrocarbon dissociates to lead to formation of a metal carbide. This dissociation process requires a catalyst that is the metal. The next step of the growth process is carbon diffusion within the nanoparticle, followed by precipitation and CNT growth. If the driving force for carbon diffusion is a concentration gradient within the nanoparticle, the composition of the nanoparticle cannot be homogeneous, otherwise the catalyst would not be active anymore. This corresponds to the observations of catalyst poisoning mentioned above and reported in [134–136]. Note that

942

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

the yarmulke mechanism proposed for SWNT growth does not inevitably imply carbon diffusion into the metal nanoparticle and thus carbide formation. 5.1.4. Stabilization of the catalyst Whereas molybdenum alone is completely inactive for CNT growth and cobalt alone is unselective (i.e. yields formation of both SWNTs and MWNTs and eventually other carbonaceous deposits), molybdenum–cobalt alloys are found to be good catalysts for SWNT formation [114]. Liao et al. [119] also studied the composition of cobalt–molybdenum nanoparticles and its influence on the carbon morphology. They found that particles with 5–15 at.% cobalt tend to produce long CNTs, those with 40–45 at.% cobalt tend to produce short CNTs, and those with 85–98 at.% cobalt tend to produce onionated morphology. The functioning of cobalt–molybdenum catalysts for SWNT formation has been quite well understood: cobalt is the real catalyst and molybdenum stabilizes Co2+ ions. Indeed, when cobalt is not interacting with molybdenum, it sinters in the reduced state and forms large metal aggregates, which generate defective MWNTs, filaments and graphite nanofibers. In the presence of molybdenum, Co2+ ions are embedded in a molybdate-like environment. The cobalt ions, the actual catalysts, are stabilized in that geometry since the so formed complex remains unchanged during the reduction step at high temperature. When CO is introduced into the reaction chamber, the molybdate dissociates to form molybdenum carbide, thus liberating the cobalt ions which can then catalyze SWNT formation because of the still small size of the newly formed cobalt particles [114,116]. Liao et al. [119] moreover explained the observed trends with different compositions of cobalt–molybdenum nanoparticles by directly relating the CO decomposition activity to the cobalt content. The cobalt content therefore determines which step between CO decomposition and carbon diffusion is rate-limiting and thus the type of carbon deposit. Note, however, that Lan et al. succeeded in synthesizing bundles of SWNTs by disproportionation of CO on pure cobalt catalysts [137]. This result shows how important is the role played by the process conditions as a whole. Not only molybdenum has been used as non-transition metal additive to a usual transition metal catalyst (iron, cobalt, nickel) but also e.g. magnesium. The few authors who studied magnesium- and cobalt- or nickel-based catalyst all reported formation of a solid solution of type Ni(or Co)xMg1xO. The Ni2+ in the solid solution are found to be highly dispersed (host–dopant type) and thus difficult to reduce completely (valence stabilization by MgO crystal-field) to Ni0, which constitute the active sites. Deep reduction of nickel is therefore inhibited in NixMg1xO and thus the aggregation of the Ni0 to form large particles [17,122]. Flahaut et al. [125] used a CoxMg1xO catalyst with small amount of cobalt and no reduction treatment before growth and obtained SWNTs. They put forward kinetic arguments to explain their results, assuming that the low cobalt content is favorable for SWNT growth as it is equivalent to slow carbon decomposition rate. Concerning the mixture of two catalysts, Pinheiro and Gadelle [138] performed a thermodynamic study of the chemical state of a supported iron–cobalt catalyst during CO disproportionation. They showed that the stability of the catalyst and thus

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

943

the progress of the disproportionation reaction directly depends on the composition of the catalyst alloy. The most stable alloy is found to be that with roughly 50 wt% Co. More iron leads to formation of Fe3C or Fe3O4 and more cobalt to Co2C when contacted with CO–CO2 mixtures. 5.1.5. Conclusion The ability of a catalyst metal to dissociate hydrocarbon or CO molecules is linked to its electronic structure. Overlap of non-filled 3d-orbitals with carbon orbitals can favor the dissociation process. It should be further investigated if interactions between metal 3d and carbon orbitals could also have an effect on the further growth process. The carbon solubility of the catalyst material might be an important factor since carbon is believed to diffuse through the catalyst nanoparticle during the growth process (see Section 3). From the experimental results, it seems obvious that Fe3C forms during CNT growth on iron particles but its actual role in the catalysis process remains unclear. Addition of other components to or mixture of transition metals generally exhibits better results in terms of catalytic activity and CNT quality. This can be explained first by changes in the electronic structure, which can e.g. yield a lower activation energy and thus a lower growth temperature. Second, catalysts composed of several components can form either stable complexes or solid solutions which remain in the form of small nanoparticles until the beginning of the growth process or stable alloys, preventing then formation of species leading to a change of the catalytic behavior. The stabilization of the catalyst in small-scale particles seems to be crucial since large catalyst particles are found to be inactive for CNT growth. The relation between CNT growth and size of the catalyst nanoparticle will be treated in the next section. 5.2. Morphology As mentioned in the introduction of Section 2, the catalyst is able to catalyze the formation of CNTs only if it is in the form of particles. We want here to discuss the effect of the size and crystallographic orientation of the nanoparticles on the growth process. 5.2.1. Size There is a consensus in the literature concerning the correlation between the size of the catalyst nanoparticles and the CNT diameter. Indeed, many groups observed a direct dependence of the two quantities [20,22,27,29,44,67,77,88,96]. When the nanoparticles are prepared in holes or pores, the diameter of the former is subjected to the size of the latter and thus the resulting CNTs have a diameter roughly equal to the diameter of hole or pore [52,71]. Nikolaev et al. stated that the particle size after CNT growth is larger than the CNT diameter, suggesting that particles continue to grow even after nucleating a tube [45]. The relevant size of the nanoparticles for the resulting diameter of the CNTs is thus their size at the time of nucleation.

944

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

Fig. 6. Transmission electron microscopy image of SWNTs produced by decomposition of CO on a molybdenum–cobalt catalyst. The size correlation between catalyst nanoparticle and CNT is clearly shown. Reprinted with permission from [118]. Copyright (2002) American Institute of Physics.

Fig. 6 shows a transmission electron microscopy image obtained by Lan et al. [118] who produced SWNTs by decomposition of CO at 700 °C on an opal matrix embedded with molybdenum–cobalt catalyst. The correlation between the size of the catalyst nanoparticle and the nanotube is here clearly seen. Dai et al. noticed that larger particles always appear to be onionated and so are inactive for catalysis of CNTs [44]. Formation of non-selective forms of carbon was also observed by Pe´rez-Cabero et al. with the iron–silica catalysts with higher iron content [13]. They explained this result by the fact that the degree to which metallic iron aggregates upon reduction into metallic particles depends on the iron content. Thus catalysts with higher iron content exhibit larger metallic iron particles, leading to other forms of carbon than nanotubes. Large particles therefore appear to be unable to catalyze CNT growth. What are the arguments found in the literature to explain this result? First of all, since the catalyst particles used in CNT formation are of the nanometer range, there may be a ‘‘size effect’’ contributing to their catalytic properties. For a particle size smaller than 5 nm, the number of atoms in low-coordinated positions is greater than 10% of the total [139]. This may modificate the electron density of the nanoparticle material or stabilize unusual faces or sites on the surface of the aggregates [140]. Both will change the surface electronic structure, what can in turn affect the catalysis process, as discussed in Section 5.1.2. Furthermore, as seen in Section 3, the Yarmulke mechanism is valid only for very small nanoparticles, which exhibit a very high surface energy per atom. On the other hand, Hafner et al. supposed that

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

945

supply-limited growth allows more time to anneal to the lowest-energy structure so that smaller particles produce CNTs while larger particles are encapsulated [106]. This kinetic argument is closely akin to the one invoked by Flahaut et al. [125] to explain formation of SWNTs with CoxMg1xO catalyst containing small amount of Co (see Section 5.1.4). 5.2.2. Crystallographic orientation Beyond the size of the catalyst nanoparticles, another morphologic parameter might play a part in the catalytic growth process: the crystallographic orientation of the supported nanoparticles. Audier et al. studied the crystallographic characteristics of the catalyst particles at the end of carbon filaments (top-growth) [141]. Their study yields a correlation between the particle orientation and the tube axis. More recently, Ermakova et al. claimed that hydrocarbon decomposition on nickel occurs on different edges of the nanoparticle due to anisotropy of nickel and that the filament axis is found to be parallel to the nickel (1 1 1) planes [15]. The theoretical work of Shibuta et al. indicates strong interaction between the hexagonal carbon network formed in the first stage of the growth process and the structure of catalytic metal atoms [49]. The crystal orientation may thus play a critical role in the determination of chirality. Similarly, Vinciguerra et al. showed that the (1, 1, 0) plane of iron (bbc) and the (1, 1, 1) planes of cobalt and nickel (fcc) have the appropriate symmetry and distances to overlap with the lattice of graphene sheet [46]. It should be emphasized that these considerations about the effect of crystallographic orientation of catalyst nanoparticles on growth of CNTs make sense only if the nanoparticle remains solid during the growth process. 5.2.3. Conclusion The size of the catalyst nanoparticles seems to be the determining factor for the diameter of the CNT grown on it. Beyond this size correlation, only small nanoparticles are able to catalyze formation of CNT. This can be explained on the one hand by the fact that such very small nanoparticles can exhibit peculiar electronic properties (and thus catalytic properties as shown in Section 5.1.2) due to the unusual high ratio surface atom/bulk atom. On the other hand with a growth mechanism implying formation of a carbon cap on the nanoparticle surface to reduce its unusual high surface energy. Kinetic arguments are also invoked. Finally, the crystallographic orientation of the catalyst nanoparticle can be crucial for CNT growth. 5.3. Preparation method As seen in Section 5.2, it is essential, in order to get CNTs of a given diameter, to be first able to produce nanoparticles with control over their sizes. The chosen method to get catalyst nanoparticles therefore has a first influence on the catalytic growth process via the morphology of the obtained nanoparticles. Various preparation methods of nanoparticles have been presented in Section 2 and as we shall see in the following on the basis of few examples, they may yield different catalytic properties.

946

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

5.3.1. Morphology 5.3.1.1. Size. Li et al. [26] produced iron–molybdenum nanoparticles by thermal decomposition of their carbonyl complexes in octyl ether solution under a nitrogen atmosphere. They investigated the dependence of the particle size on the reactant concentration, reaction time and molar ratio of metal carbonyl and protective agents. By varying e.g. the reactant concentration and nature, the particle size can decrease from 16 nm to about 4 nm, as shown in Fig. 7. In such a complex system, the size of the produced nanoparticles depends on many factors, including the number of nuclei created, the total concentration of reactants and the effect of protective agents. All these factors influence simultaneously the size of the obtained products. Vander Wal et al. compared the decomposition of a metal precursor (nitrate salt) on a metal oxide surface with ferrofluid, i.e. maghemite particles stabilized as a water based suspension [142]. The morphology and size of the resulting nanotubes or nanofibers is much more uniform with the ferrofluid. Keeping in mind that the size of the nanotubes is directly dependent on the size of the catalyst particle, the authors explained this result by considering that the morphology of the preformed catalyst particle does not depend on the support. The size distribution of the particles will thus roughly remain the same on the support as in the solution. In contrast,

Fig. 7. Iron–molybdenum nanoparticles synthesized with different protective agents. A: 1 mmol of octanoic acid. B: 2.5 mmol of octanoic acid. C: 1 mmol of octanoic acid and 1 mmol of bis(2ethylhexyl)amine. D: 1 mmol of bis(2-ethylhexyl)amine. E: 2.5 mmol of bis(2-ethylhexyl)amine. The scale bars in all figures are 100 nm. Reprinted with permission from [26]. Copyright (2001) American Chemical Society.

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

947

the size of in situ prepared catalyst particles by decomposition of a precursor on a support is determined by the physical structure of the support, in particular by its void volume. Therefore, with in situ prepared catalyst particles, a further difficulty still remains, since the growth process implies high temperatures (500–1200 °C) and therefore can yield sintering of the nanoparticles. Ago et al. [25] indeed obtained well separated cobalt nanoparticles using the reverse micelle method. They found a mean diameter of 4 nm for their nanoparticles. Nevertheless, the inner diameters of the obtained MWNTs were ranging from 8 to 20 nm and the outer ones from 20 to 40 nm, suggesting that the as-prepared cobalt particles aggregate to form bigger clusters. 5.3.1.2. Dispersion. Ivanov et al. noted a better dispersion of metal on SiO2 with precipitation–ion-exchange than with impregnation [20]. Cassel et al. [41] prepared iron nanoparticles by impregnation of support materials in salt solutions. They compared the use of Fe2(SO4)3 vs. Fe(NO3)3 on Al2O3 supports and claimed that catalysts obtained with Fe2(SO4)3 in aqueous solution was superior in the growth of SWNT. They could probe the structures of catalyst materials and found different textural properties between the Fe2(SO4)3 and Fe(NO3)3 derived catalysts, namely a larger pore volume for the former. They explained this result first by the strong interactions existing in aqueous solution between sulfate ions and surface sites of alumina, which allows the metal species to be uniformly dispersed and anchored on the alumina surface. And second, the decomposition of these highly dispersed and strongly anchored sulfate species to Fe2O3 occurs at relative high temperature, leading to the generation of the observed textural pores. 5.3.1.3. Trapping with porous support. These few examples illustrate the challenging difficulty associated with each preparation method to produce nanoparticles of small enough size and well dispersed on the support surface in order to get CNTs with the desired diameters. In the face of this difficulty, a widely used way is to resort to porous supports. Hayashi et al., e.g. did not use zeolite as a template but to limit the size and the structure of catalytic particles [64]. Fan et al. observed iron oxide nanoparticles to form with a narrow size distribution on porous silicon due to their strong interaction with the support [54]. The idea is to trap the particles into the pores of the substrate (e.g. zeolite, which is a highly porous material) and thus prevent them agglomerating. This trapping by a zeolite support is schematically illustrated on Fig. 8 for both growth modes. There are many ways to obtain porous supports. One possibility is to prepare anodic aluminum oxide (AAO) by anodizing Al in various electrolyte solutions using dc current [71]. By varying the anodizing conditions, one can control the density, diameter and length of the pores of the substrate. One can also lithographically define nanoholes [52]. The sol–gel method is based on the use of aerogel materials as textural promoter, which exhibit high surface area and high porosity [109]. This method ensures a highly homogeneous distribution of transition metal ions in the gel. On this extremely porous support, the CNTs grow outward perpendicularly and form an aligned CNT array [12,16,143].

948

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

Fig. 8. Schematic image of growth of carbon nanotubes in which the catalyst nanoparticles (red balls) are trapped into the pore of the zeolite support (brown base). Reprinted with permission from [64]. Copyright (2003) American Chemical Society. (For interpretation of the references in color in this figure legend, the reader is referred to the web version of this article.)

5.3.2. Electronic structure Kukovecz et al. performed UV–VIS measurements on cobalt, iron and nickel incorporated into sol–gel SiO2–TiO2 matrices [144]. They could show that the composition of the matrix and the calcination treatment have an effect on the detected electronic transitions, i.e. on the valence shell electron structure of the bimetallic catalysts they tested. They moreover observed non-identifiable bands which may result from metal–metal electronic interactions. These interactions could be the explanation for different catalytic activity in producing CNTs between the different samples. 5.3.3. Other aspects Besides the morphologic aspect, the preparation method of the nanoparticles may play an important role in their catalytic properties. Hernadi et al. [23,55] found striking differences in the properties of the catalyst depending on the preparation method. So cobalt supported on Y prepared by ion exchange is found inactive for CNT formation whereas it exhibits high activity in the same growth conditions when prepared by impregnation [23]. Inversely, iron/silica is found to have better carbon yield and to lead to more homogeneous CNT diameter distribution and less amorphous carbon when prepared by ion-adsorption–precipitation compared to impregnation [55].

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

949

Su et al. found out that a supercritical drying process is a crucial step in preparing a catalyst by the sol–gel method [109]. They observed a weight gain varying from roughly 5% to 200% as function of the drying process. Wang and coauthors studied growth of CNTs with a nickel-iron catalyst prepared whether by plasma of by conventional (chemical) way [91]. They observed the presence of encapsulated metal within the obtained CNTs when the catalyst had been prepared by plasma and not otherwise. Fonseca et al. studied the effect of the pH of the catalyst precursor solution on the catalytic properties of the nanoparticles obtained by ion-adsorption–precipitation [24]. The authors reported quite different qualities of CNTs as function of the pH, suggesting that the pH has an important influence on the obtained nanoparticles by the ion-adsorption–precipitation method. 5.3.4. Conclusion The preparation method of the catalyst nanoparticles is shown to have an influence on their size and dispersion on the support surface and thus to their catalytic properties. Porous supports are often used to get small nanoparticles with narrow size distribution. The composition of the matrix into which the catalyst metal is incorporated and the calcination treatment can have an effect on the electronic transition of the catalyst. Many authors observed different CNT growth results with the same growth parameters and the same catalyst, but prepared by different ways. The reported effects have not been understood yet. One can suppose that different preparation methods lead to different crystallographic orientations of the nanoparticles, or that surface groups may form on the nanoparticle when prepared by a particular method, yielding particular catalytic properties. 5.4. Support Many of the publications cited in Section 4 reported results of CVD of CNTs on non-supported catalysts. In these cases, the catalyst in the form of a powder is placed in a quartz boat and then introduced into the reactor. In the case of supported catalysts, the interactions between the support and the catalyst nanoparticles have to be taken into account. They might play an essential part in the whole growth process since these interactions can alter all the other parameters treated in Section 5. Many authors compared the results of CNT growth with the same catalyst but on different supports. We shall in the following first review some of these comparison results and then try to elucidate the role of the support in the growth process. 5.4.1. Experimental comparison of different supports In [55], different iron-based catalysts prepared by impregnation were tested. Iron on graphite showed low activity and bad quality of CNT (fibers) whereas iron on silica yielded the best results (Y and ZSM-5 were also tested). Su et al. observed better results on Al2O3 than on SiO2 [109] and also a higher rate of SWNT with iron– molybdenum nanoparticles supported on an Al2O3 monolayer than with pure iron nanoparticles [110]. Colomer et al. also got better results for catalyst nanoparticles

950

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

supported on Al2O3 than on SiO2 [76]. Kukovecz et al. found nickel on alumina very active (but not for CNT formation!) whereas nickel on silica is poorly active [12]. Willems et al. observed that cobalt–molybdenum exhibits better results on alumina than on zeolite [96]. Kong et al. tested different catalysts on different substrates and reported that CoO/alumina and NiO–CoO/silica yield SWNTs whereas CoO/silica and NiO–CoO/alumina yield no SWNTs [97]. Vander Wal et al. tested copper, iron and nickel on CaO, Al2O3, SiO2 and TiO2 and observed that nickel was most active on TiO2, less active on SiO2 and Al2O3 and almost unreactive ib CaO. Copper was found most active when supported on CaO and SiO2 and less on TiO2 and Al2O3. Finally, CaO was the best support for iron and Al2O3 the worst. These few results demonstrate that the couple catalyst/support has to be considered. The support has to be carefully chosen since, first, intermediate indesirable compounds can form and, second, the interactions between support and metal nanoparticles are found to be crucial for the catalytic properties of the latter. 5.4.2. Formation of intercompounds The drawback posed by silicon substrates is the high propensity to interdiffusion of silicon and metal. Indeed, diffusion of nickel catalyst into silicon substrate is observed to occur at temperatures above 300 °C [84]. Likewise, iron on silicon yields formation of an iron silicide upon heating up to 850 °C (temperature of CNT growth) [145]. The problem can be circumvented by using a diffusion barrier, generally SiO2 or TiN [84,86,102]. Vajtai et al. stated that when deposited using a substrate voltage of 200 V, TiN is denser and more efficient as diffusion barrier [102]. They could not observe any chemical interaction between iron and TiN upon heating and obtained much better results of CNT growth with TiN than with silicon. Use of silicon substrates for CNT growth can moreover lead to formation of SiC [102]. 5.4.3. Support–catalyst interactions Interactions between a catalyst metal and its support can be physical or chemical. A physical interaction is e.g. the size determination of a metal particle by its porous support [142]. In contrast, chemical interactions involve charge transfer between the support and the catalyst. This charge transfer can take place via different pathways, e.g. via oxidation/reduction or acid/base (donor/acceptor) reactions. When the substrate is a metal oxide, oxidation/reduction reactions can occur by addition of neutral oxygen from the substrate to the metal particle. Acid/base interactions are linked to the corresponding Lewis acid and base characters of the materials involved. In a metal oxide substrate, surface anions act as Lewis base sites (electron pair donor) and cations as Lewis acid sites (electron pair acceptor). Support–catalyst interactions can affect the electronic structure of the catalyst and have an effect on the dispersion of the metal on the support. 5.4.3.1. Effect on electronic structure. Vander Wal et al. showed the importance of electronic interactions between metal nanoparticle and support on the ability of the metal to decompose adsorbate (hydrocarbons or CO in the case of CNT formation) [142]. Indeed, acquisition of negative charge by the metal catalyst from the sub-

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

951

strate can enhance its catalytic activity by strengthening back-donation of electron density into antibonding orbitals of the adsorbate. This electron sharing between catalyst and adsorbate sufficiently weakens the bonding within the adsorbate, resulting in its dissociation. As seen above, the propensity to electron donation of a surface is linked to its Lewis base character. The great catalytic activity of copper observed on support media with strong Lewis base sites (CaO and SiO2) [142] exemplifies this phenomenon. Indeed, as mentioned in Section 5.1, the catalytic activity of a metal is in a first approximation inversely dependent on the filling of its d-band. Bulk copper should therefore be catalytically inactive since its d-shell is completely filled. Nevertheless, upon acquisition of partial negative charge from the support (that here acts as a Lewis base), copper can donate negative charge to the adsorbate. In this view, the catalyst particle acts as a conduit for transferring negative charge to the adsorbate. 5.4.3.2. Effect on catalyst morphology. As shown in Eq. (1) of Section 2.9, the interactions between support surface and metal particle, expressed in terms of interface energies, will determine the shape of the particle. Wright et al. compared the shape of nickel nanoparticles on TiN and on MoSiO2 and found that nickel particles are more spherical on MoSiO2 because of a weaker wetting on MoSiO2 than on TiN [86]. Generally, one can assume that strong metal–support interactions lead to a good dispersion of the metal on the support surface. Many authors de facto ascribed their better results obtained on Al2O3 compared to SiO2 to the stronger Lewis acidity of the former support, yielding a better dispersion of the metal [109,76,41]. However, too strong an interaction between catalyst and support yields strong wetting and prevents the formation of appropriately shaped catalyst particles for CNT growth [47]. Seidel et al. emphasized that the surface diffusion rate of a specific atom on a substrate depends not only on the substrate nature but also on its roughness [47]. Delzeit et al. also studied the role of the underlayer in the growth of CNTs [104]. They suggested that ion-beam sputtering of the underlayer increases its surface roughness, thus providing more active nucleation sites. Surface roughness can therefore hinder surface diffusion of the catalyst and thus its coalescence into (too) large particles. Interactions between metal particles and surface of course not only depend on the support material but also on its surface orientation and on the metal as well. So Hongo et al. observed a dependence of the SWNT yield with the face of their sapphire support whereas they could not obtain any SWNTs with silicon or SiO2 [65]. Su et al. obtained better results with iron–molybdenum nanoparticles on Al2O3 than with pure iron ones, because the former more interact with the the Al2O3 surface than the latter. One can moreover imagine that the support will also have an impact on the crystallographic structure of nanoparticles, in particular on their orientation, and thus influence the morphology of the obtained nanotubes, as seen in Section 5.2. 5.4.3.3. Effect during growth mechanism. The majority of authors working with nickel-based catalysts observed a tip-growth mechanism where the catalyst particle is detached from the support during CNT growth. This shows that nickel probably

952

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

exhibits weak interactions with the most supports. This can have a bad influence on the quality of the nanotubes since bamboo structures are often observed with this growth type. As already mentioned in the previous section, many groups use porous supports [13,53,54,100]. The porosity of the support may also have an effect on the CNT growth itself, besides its function as particle trap. Indeed, CNTs were observed to grow at a rate of about 50% faster on porous silicon than on plain silicon [54]. This result might be explained by the porous structure of the support which makes it permeable to the ethylene molecules so that carbon feedstock reaches the catalyst at a higher rate. 5.4.4. Conclusion Interactions between catalyst and support are essential since they can affect the electronic structure of the nanoparticles and their morphology and in turn their catalytic properties. Chemical interactions involve charge transfer between support and catalyst. This charge transfer is roughly correlated to the Lewis base or acid character of the support. These catalyst-support interactions have to be strong to ensure good dispersion of the catalyst on the support but not too strong since it then yields strong wetting and will prevent formation of appropriately shaped catalyst particles for CNT growth. 5.5. Pretreatment The influence of pretreatment of the catalyst nanoparticles on the following CNT growth process has not been thoroughly studied until now. A few experimental data can however be found in the literature. We will further discuss the role played by the pretreatment. 5.5.1. Experimental results Ago et al. observed that pretreatment with H2S gas straightens the MWNTs and yields a wider range of outer diameter [25]. They wondered whether catalytic activity of cobalt is reduced by sulfur, forming S–Co. Ren et al. observed no CNT growth if N2 is used instead of NH3, although the surface of the nickel layer after NH3 or N2 plasma etching seems to be essentially the same in terms of morphology [32]. They further observed that when NH3 and C2H2 are introduced simultaneously, the obtained CNTs exhibit larger diameters, suggesting that the etching process and thus the resulting particle size is dependent on the etching gas. Harutyunyan et al. tested iron and iron–molybdenum nanoparticles for CNT growth with and without prereduction [107]. They claimed that both catalysts are active at 680 °C in the reduced state. In contrast, iron unreduced is found to be unactive at this temperature (becomes active only at 900 °C) and iron–molybdenum unreduced is active at 680 °C. 5.5.2. Role of pretreatment 5.5.2.1. Effect on electronic structure. Intuitively, one may think that pretreatment will first change the electronic state of the catalyst. Oxides are reduced to the metallic

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

953

state which is a priori the actual catalyst. Nonetheless, Baker et al. found out that FeO appears to be a much better catalyst than metallic iron for the formation of filamentary carbon and thus claimed that there was no reduction before to growth [146]. Others, like Hernadi et al. [55], considered that reduction pretreatment of iron oxide is not necessary since the hydrocarbon atmosphere is able to reduce the catalyst to the required extent under reaction conditions. 5.5.2.2. Effect on catalyst morphology. The second effect of pretreatment is on the nanoparticle morphology. Heating the substrate allows atoms to diffuse. The deposited particles then incur the risk of sintering and thus change their morphological characteristics, such as their diameter and height. In the case of physical deposited metal, an annealing step is essential to obtain nanoparticles. The right annealing time must then be found in order to get nanoparticles of the right size: too small catalyst clusters and isolated catalyst atoms poison the catalyst support surface and too large particles are unable to catalyze CNT growth [47]. As mentioned in Section 2.9, the atmosphere gas used by breaking up the film may have a great influence on the shape of the obtained nanoparticles. Kanzow and Ding claimed that a hydrogen atmosphere generally leads to a much better wetting behavior. The contact angle of the metal on the carbon substrate is shown to be strongly affected [39]. 5.5.3. Conclusion The pretreatment of the catalyst nanoparticles has the role of reducing the mostly oxidized catalyst metal. For catalysts prepared by PVD, the pretreatment ensures formation of nanoparticles and parameters such as type of gas, time and temperature will then be essential to get nanoparticles with the right shape.

6. Conclusion This review has shown the complexity of the role of the catalyst in the CCVD of CNTs. We showed in Section 5.1 that mostly 3d-metals are used to catalyze formation of CNTs because of their ability to decompose hydrocarbons. However, they are not all equivalent and, in particular, alloying them with each other or with other non-transition metals dramatically changes (generally improves) their catalytic properties. For few cases discussed in Section 5.1, the role of the non-transition metal is to disperse and to stabilize the catalyst metal—so an effect on the geometric aspect. On the other hand, effects of alloying two transition metals on e.g. the electronic structure or the carbon solubility and diffusion have still to be studied to explain the observed improvement in the catalyst performances. Furthermore, overlap of catalyst 3d-orbitals with carbon orbitals after hydrocarbon dissociation could be a relevant parameter for the CNT growth process and should be investigated. The link between the size of the catalyst nanoparticles and the diameter of the CNTs and the consequently absolute necessity to prepare stable nanoparticles of controlled size have been emphasized (Section 5.2). To meet this severe constraint high-technological efforts have to be made. Though, other parameters concerning

954

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

the catalyst morphology have an influence on the growth process like the crystallographic orientation of the nanoparticles. The preparation method of the catalyst particles also plays its part in the knotty problem of the factors influencing the catalytic properties of the metal nanoparticles. Section 5.3 deals with this question and we show there that the choice of the preparation method determines the morphology of the obtained nanoparticles. We also discussed the use of porous support to limit the size of the nanoparticles. But the influence of the preparation method is not limited to morphological effects and can directly affect catalytic properties, even if this influence has been little studied and is not yet clear. The problem is complex since, as pointed in Section 5.4, a given catalyst can yield completely different results upon CCVD when supported on different materials. One actually has to consider not the catalyst alone but the couple catalyst/support. The interactions between catalyst and support are found to be essential. First, they partly determine the morphology of the nanoparticles and, second, they also alter the electronic structure of the nanoparticles, thus altering their catalytic properties. These interactions depend on both support and catalyst materials but also on their crystallographic orientations, on the surface roughness and porosity of the support. Finally, pretreatment can be essential for activating the catalyst by reducing it to its actual catalytic form (pure metal) or by forming nanoparticles in the case of catalysts prepared by PVD, as seen in Section 5.5. One therefore realizes how challenging it is to control these numerous parameters since they are all interdependent and have to be within the specific process window. Moreover, we have considered here only parameters concerning the catalyst nanoparticles, which cannot alone explain whether CNTs will grow or not and how they will grow. The catalytic process requires a hydrocarbon gas (or CO) coupled with a supported catalyst in charge of its dissociation and the subsequent growth of the CNT. Lee et al. compared different source gases with various catalysts and indicated that the results of CNT growth depend on the gas used [147]. The couple catalyst/gas or rather the trio support/catalyst/gas should be considered for a complete understanding of the growth process. Other important factors influence the kinetic and thermodynamic aspects of the growth process, e.g. temperature, pressure, flow rate and reaction time. Many authors explain their success in producing SWNTs by the very small size of their catalyst nanoparticles. However, the differentiation between SWNT and MWNT growth processes is not really clear and kinetic arguments are often invoked. Production of SWNTs is then explained by limitation of the carbon supply, which is believed to be the limiting growth step (and not carbon diffusion through the catalyst particle) [106].

Acknowledgements We thank E. Rouvie`re, L. Jodin, M. Delaunay and R. Baptist for helping gather relevant literature and for subsequent discussions, G. Maheut for her explanations

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

955

concerning the chemical preparation methods and P. Gadelle and F. Triozon for proofreading this document and helpful comments.

References [1] Tibbetts GG. Why are carbon filaments tubular? J Cryst Growth 1984;66:632–8. [2] Iijima S. Helical microtubules of graphitic carbon. Nature 1991;354:56. [3] Hamada N, Sawada S, Oshiyama A. New one-dimensional conductors: graphitic microtubules. Phys Rev Lett 1992;68:1579–81. [4] Odom T, Huang J, Kim P, Lieber CM. Atomic structure and electronic properties of single-walled carbon nanotubes. Nature 1998;391:62–4. [5] Poncharal P, Berger C, Yi Y, Wang ZL, de Heer WA. Room temperature ballistic conduction in carbon nanotubes. J Phys Chem B 2002;106:12104–18. [6] Tans SJ, Verschueren ARM, Dekker C. Room-temperature transistor based on a single carbon nanotube. Nature 1998;393:49. [7] Javey A, Guo J, Wang Q, Lundstrom M, Dai H. Ballistic carbon nanotube field-effect transistors. Nature 2003;424:654. [8] Luyken RJ, Hofmann F. Concepts for hybrid CMOS-molecular non-volatile memories. Nanotechnology 2003;14:273–6. [9] Tseng Y-C, Xuan P, Javey A, Malloy R, Wang Q, Bokor J, et al. Monolithic integration of carbon nanotube devices with silicon MOS technology. Nano Lett 2004;4:123–7. [10] Cheng HM, Li SF, Pan G, Pan HY, He LL, Sun X, et al. Large-scale and low-cost synthesis of single-walled carbon nanotubes by the catalytic pyrolysis of hydrocarbons. Appl Phys Lett 1998;72:3282–4. [11] Jodin L, private communication. [12] Kukovecz A, Ko´nya Z, Nagaraju N, Willems I, Tama´si A, Fonseca A, et al. Catalytic synthesis of carbon nanotubes over Co, Fe and Ni containing conventional and sol–gel silica–aluminas. Phys Chem Chem Phys 2000;2:3071. [13] Pe´rez-Cabero M, Rodrı´guez-Ramos I, Guerrero-Ruı´z A. Characterization of carbon nanotubes and carbon nanofibers prepared by catalytic decomposition of acetylene in a fluidized bed reactor. J Catal 2003;215:305–16. [14] Sen R, Govindaraj A, Rao CNR. Carbon nanotubes by the metallocene route. Chem Phys Lett 1997;267:276–80. [15] Ermakova MA, Ermakov DY, Chuvilin AL, Kuvshinov GG. Decomposition of methane over iron catalysts at the range of moderate temperatures: the influence of structure of the catalytic systems and the reaction conditions on the yield of carbon and morphology of carbon filaments. J Catal 2001;201:183–97. [16] Pan SS, Xie ZW, Chang BH, Sun LF, Zhou WY, Wang G. Direct growth of aligned open carbon nanotubes by chemical vapor deposition. Chem Phys Lett 1999;299:97–102. [17] Chen P, Zhang H-B, Lin G-D, Hong Q, Tsai KR. Growth of carbon nanotubes by catalytic decomposition of CH4 or CO on a Ni–MgO catalyst. Carbon 1997;35:1495. [18] Bacsa RR, Laurent C, Peigney A, Bacsa WS, Vaugien T, Rousset A. High specific surface area carbon nanotubes from catalytic chemical vapor deposition process. Chem Phys Lett 2000;323:566–71. [19] Pinheiro JP, Schouler MC, Gadelle P. Nanotubes and nanofilaments from carbon monoxide disproportionation over Co/MgO catalysts I. Growth versus catalyst state. Carbon 2003;41:2949– 59. [20] Ivanov V, Nagy JB, Lambin P, Lucas A, Zhang XF, Bernaerts D, et al. The study of carbon nanotubules produced by catalytic method. Chem Phys Lett 1994;223:329–35. [21] Venegoni D, Serp P, Feurer R, Kihn Y, Vahlas C, Kalck P. Parametric study for the growth of carbon nanotubes by catalytic chemical vapor deposition in a fluidized bed reactor. Carbon 2002;40:1799.

956

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

[22] Li Y, Kim W, Zhang Y, Rolandi M, Wang D, Dai H. Growth of single-walled carbon nanotubes from discrete catalytic nanoparticles of various sizes. J Phys Chem B 2001;105:11424–31. [23] Hernadi K, Fonseca A, Nagy JB, Bernaerts D, Fudala A, Lucas AA. Catalytic synthesis of carbon nanotubes using zeolite support. Zeolites 1996;17:416–23. [24] Fonseca A, Hernadi K, Nagy JB, Bernaerts D, Lucas AA. Optimization of catalytic production and purification of buckytubes. J Mol Catal A: Chem 1996;107:159–68. [25] Ago H, Komatsu T, Ohshima S, Kuriki Y, Yumura M. Dispersion of metal nanoparticles for aligned carbon nanotube arrays. Appl Phys Lett 2000;77:79. [26] Li Y, Liu J, Wang Y, Wang ZL. Preparation of monodispersed Fe–Mo nanoparticles as the catalyst for CVD synthesis of carbon nanotubes. Chem Mater 2001;13:1008–14. [27] Cheung CL, Kurtz A, Park H, Lieber CM. Diameter-controlled synthesis of carbon nanotubes. J Phys Chem B 2002;106:2429–33. [28] Venables JA. The chemical physics of solid surfaces. In: King DA, Woodruff DP, editors. Growth and properties of ultrathin epitaxial layers, 8. Elsevier; 1997. p. 1–45. [29] Yoon YJ, Bae JC, Baik HK, Cho SJ, Lee S-J, Song KM, et al. Growth control of single and multiwalled carbon nanotubes by thin film catalyst. Chem Phys Lett 2002;366:109–14. [30] Yudasaka M, Kikuchi R, Matsui T, Yoshimasa O, Yoshimura S. Specific conditions for Ni catalyzed carbon nanotube growth by chemical vapor deposition. Appl Phys Lett 1995;67:2477. [31] Teo KBK, Chhowalla M, Amaratunga GAJ, Milne WI, Hasko DG, Pirio G, et al. Uniform patterned growth of carbon nanotubes without surface carbon. Appl Phys Lett 2001;79:1534. [32] Ren ZF, Huang ZP, Xu JW, Wang JH, Bush P, Siegal MP, et al. Synthesis of large arrays of wellaligned carbon nanotubes on glass. Science 1998;282:1105–7. [33] Cui H, Zhou O, Stoner BR. Deposition of aligned bamboo-like carbon nanotubes via microwave plasma enhanced chemical vapor deposition. J Appl Phys 2000;88:6072. [34] Hsu CM, Lin HL, ad Cgabg CH, Kuo CT. Growth of the large area horizontally-aligned carbon nanotubes by ECR-CVD. Thin solid films 2002;420–421:225–9. [35] Lee TY, Han J-H, Yoo CSHJ-B, Park C-Y, Jung T, Yu S, et al. Comparison of source gases and catalyst metals for growth of carbon nanotube. Surf Coat Technol 2003;169–170:348–52. [36] Wang YH, Lin J, Huan CHA, Chen GS. Synthesis of large area aligned carbon nanotube arrays from C2H2–H2 mixture by RF plasma-enhanced chemical vapor deposition. Appl Phys Lett 2001;79:680. [37] Wagner RS, Ellis WC. Trans AIME 1965;233:1053. [38] Baker RTK. Catalytic growth of carbon filaments. Carbon 1989;27:315–23. [39] Kanzow H, Ding A. Formation mechanism of single-wall carbon nanotubes on liquid–metal particles. Phys Rev B 1999;60:11180. [40] Rostrup-Nielsen JR, Trimm DL. J Catal 1977;48:155. [41] Cassel AM, Raymakers A, Kong J, Dai H. Large scale CVD synthesis of single-walled carbon nanotubes. J Phys Chem B 1999;103:6484–92. [42] Ajayan PM, Ebbesen TW. Nanometre-size tubes of carbon. Rep Prog Phys 1997;60:1025. [43] Charlier J-C, Iijima S. Growth mechanisms of carbon nanotubes. Topics Appl Phys 2001;80:55– 81. [44] Dai H, Rinzler AG, Nikolaev P, Thess A, Colbert DT, Smalley RE. Single-wall nanotubes produced by metal-catalyzed disproportionation of carbon monoxide. Chem Phys Lett 1996;260:471–5. [45] Nikolaev P, Bronikowski M, Bradley R, Rohmund F, Colbert D, Smith K, et al. Gas-phase catalytic growth of single-walled carbon nanotubes from carbon monoxide. Chem Phys Lett 1999;313:91–7. [46] Vinciguerra V, Buonocore F, Panzera G, Occhipinti L. Growth mechanisms in chemical vapour deposited carbon nanotubes. Nanotechnology 2003;14:655–60. [47] Seidel R, Duesberg GS, Unger E, Graham AP, Liebau M, Kreupl F. Chemical vapor deposition growth of single-walled carbon nanotubes at 600 °C and a simple growth model. J Phys Chem B 2004;108:1888–93. [48] Maiti A, Brabec CJ, Bernholc J. Kinetics of metal-catalyzed growth of single-walled carbon nanotubes. Phys Rev B Rapid Commun 1997;55:6079–100.

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

957

[49] Shibuta Y, Maruyama S. Molecular dynamics simulation formation process of single-walled carbon nanotubes by CCVD method. Chem Phys Lett 2003;382:381–6. [50] Helveg S, Lo´pez-Cartes C, Sehested J, Hansen PL, Clausen BS, Rostrup-Nielsen JR, et al. Atomicscale imaging of carbon nanofibre growth. Nature 2004;427:426–9. [51] Cho Y-S, Choi GS, Hong S-Y, Kim D. Carbon nanotube synthesis using a magnetic fluid via thermal chemical vapor deposition. J Cryst Growth 2002;243:224–9. [52] Duesberg GS, Graham AP, Liebau M, Seidel R, Unger E, Kreupl F, et al. Growth of isolated carbon nanotubes with lithographically defined diameter and location. Nano Lett 2003;3:257–9. [53] Fan S, Chapline MG, Franklin NR, Tombler TW, Cassell AM, Dai H. Self-oriented regular arrays of carbon nanotubes and their field emission properties. Science 1999;283:512–4. [54] Fan S, Liang W, Dang H, Franklin NR, Tombler MG, Chapline TW, et al. Carbon nanotube arrays on silicon substrates and their possible application. Physica E 2000;8:179–83. [55] Hernadi A, anFonseca K, Nagy JB, Bernaerts D, Lucas AA. Fe-catalyzed carbon nanotube formation. Carbon 1996;34:1249. [56] Hernadi K, Fonseca A, Nagy JB, Siska A, Kiricsi I. Production of nanotubes by the catalytic decomposition of different carbon-containing compounds. Appl Catal A 2000;199:245–55. [57] Klinke C, Bonard J-M, Kern K. Comparative study of the catalytic growth of patterned carbon nanotube films. Surf Sci 2001;492:195–201. [58] Mayne M, Grobert N, Terrones M, Kamalakaran R, Ru¨hle M, Kroto HW, et al. Pyrolytic production of aligned carbon nanotubes from homogeneously dispersed benzene-based aerosols. Chem Phys Lett 2001;338:101–7. [59] Maroto Valiente A, Navarro Lo´pez P, Rodrı´guez Ramos I, Guerrero Ruiz A, Li C, Xin Q. In situ study of carbon nanotube formation by C2H2 decomposition on an iron-based catalyst. Carbon 2000;38:2003. [60] Pan ZW, Xie SS, Chang BH, Wang CY, Lu L, Liu W, et al. Very long carbon nanotubes. Nature 1998;394:631–2. [61] Sacco A, Thacker P, Chang TN, Chiang ATS. The initiation and growth of filamentous carbon from a-iron in H2, CH4, H2O, CO2 and CO gas mixtures. J Catal 1984;85:224–36. [62] Zhang ZJ, Wei BQ, Ramanath G, Ajayan PM. Substrate-site selective growth of aligned carbon nanotubes. Appl Phys Lett 2000;77:3764. [63] Zhu S, Su C-H, Cochrane JC, Lehoczky S, Cui Y, Burger A. Growth orientation of carbon nanotubes by thermal chemical vapor deposition. J Cryst Growth 2002;234:584–8. [64] Hayashi T, Ahm Kim Y, Matoba T, Esaka M, Nishimura K, Tsukada T, et al. Smallest freestanding single-walled carbon nanotube. Nano Lett 2003;3:887–9. [65] Hongo H, Yudasaka M, Ichihashi T, Nihey F, Iijima S. Chemical vapor deposition of single-wall carbon nanotubes on iron-fil-coated sapphire substrates. Chem Phys Lett 2002;361:349–54. [66] Kim W, Choi HC, Shim M, Li Y, Wang D, Dai H. Synthesis of ultralong and high percentage of semiconducting single-walled carbon nanotubes. Nano Lett 2002;2:703–8. [67] Zhang Y, Li Y, Kim W, Wang D, Dai H. Imaging as-grown single-walled carbon nanotubes originated from isolated catalytic nanoparticles. Appl Phys A 2002;74:325–8. [68] Satishkumar BC, Govindaraj A, Sen R, Rao CNR. Single-walled nanotubes by the pyrolysis of acetylene–organometallic mixtures. Chem Phys Lett 1998;293:47–52. [69] Ago H, Murata K, Yumura M, Yotani J, Uemura S. Ink-jet printing of nanoparticle catalyst for site-selective carbon nanotube growth. Appl Phys Lett 2003;82:811–9. [70] Bower C, Zhu W, Jin S, Zhou O. Plasma-induced alignment of carbon nanotubes. Appl Phys Lett 2000;77:830. [71] Jeong S-H, Hwang H-Y, Lee K-H, Jeong Y. Template-based carbon nanotubes and their application to a field emitter. Appl Phys Lett 2001;78:2052. [72] Terrones M, Grobert N, Olivares J, Zhang JP, Terrones H, Kordatos K, et al. Controlled production of aligned-nanotube bundles. Nature 1997;388:52. [73] Terrones M, Grobert N, Zhang JP, Terrones H, Olivares J, Hsu WK, et al. Preparation of aligned carbon nanotubes catalysed by laser-etched cobalt thin films. Chem Phys Lett 1998;285:299– 305.

958

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

[74] Chhowalla M, Teo KBK, Ducati C, Rupesinghe NL, Amaratunga GAJ, Ferrari AC, et al. Growth process conditions of vertically aligned carbon nanotubes using plasma enhanced chemical vapor deposition. J Appl Phys 2001;90:5308. [75] Marty L, Bouchiat V, Bonnot A-M, Chaumont M, Fournier T, Decossas S, et al. Batch processing of nanometer-scale electrical circuitry based on in-situ grown single-walled carbon nanotubes. Microelectron Eng 2002;61–62:485–9. [76] Colomer JF, Bister JF, Willems I, Konya Z, Fonseca A, Van Tendeloo G, et al. Synthesis of singlewall carbon nanotubes by catalytic decomposition of hydrocarbons. Chem Commun 1999;14: 1343. [77] Colomer J-F, Stephan C, Lefrant S, Van Tendeloo G, Willems I, Ko´nya Z, et al. Large-scale synthesis of single-wall carbon nanotubes by catalytic chemical vapor deposition (CCVD) method. Chem Phys Lett 2000;317:83–9. [78] Choi YC, Shin YM, Lee YH, Lee BS, Park G-S, Choi WB, et al. Controlling the diameter, growth rate, and density of vertically aligned carbon nanotubes synthesized by microwave plasma-enhanced chemical vapor deposition. Appl Phys Lett 2000;76:2367. [79] Huang ZP, Xu JW, Ren ZF, Wang JH, Siegal MP, Provencio PN. Growth of highly oriented carbon nanotubes by plasma-enhanced hot filament chemical vapor deposition. Appl Phys lett 1998;73:3845. [80] Jiao J, Seraphin S. Single-walled tubes and encapsulated nanoparticles: comparison of structural properties of carbon nanoclusters prepared by three different methods. J Phys Chem Solids 2000;61:1055–67. [81] Kukovitsky EF, L
vov SG, Sainov NA, Shustov VA, Chernozatonskii LA. Correlation between metal catalyst particle size and carbon nanotube growth. Chem Phys Lett 2002;355:497–503. [82] Nolan PE, Lynch DC, Cutler AH. Carbon deposition and hydrocarbon formation on group VIII metal catalysts. J Phys Chem B 1998;102:4165–75. [83] Siegal MP, Overmyer DL, Provencio PP. Precise control of multiwall carbon nanotube diameters using thermal chemical vapor deposition. Appl Phys Lett 2002;80:2171. [84] Teo KBK, Chhowalla M, Amaratunga GAJ, Milne WI, Pirio G, Legagneux P, et al. Characterisation of plasma-enhanced chemical vapor deposition carbon nanotubes by auger electron spectroscopy. J Vac Sci Technol B 2002;20:116. [85] Valentini L, Kenny JM, Lozzi L, Santucci S. Formation of carbon nanotubes by plasma enhanced chemical vapor deposition: role of nitrogen and catalyst layer thickness. J Appl Phys 2002;92: 6188. [86] Wright AC, Xiong Y, Maung N, Eichhorn SJ, Young RJ. The influence of the substrate on the growth of carbon nanotubes from nickel clusters—an investigation using STM, FE-SEM, TEM and Raman spectroscopy. Mater Sci Eng C 2003;23:279–83. [87] Yudasaka M, Kikuchi R, Ohki Y, Ota E, Yoshimura S. Behavior of Ni in carbon nanotube nucleation. Appl Phys Lett 1997;70:1817. [88] Zhang WD, Wen Y, Tjiu WC, Xu GQ, Gan LM. Growth of vertically aligned carbon-nanotube array on large area of quartz plates by chemical vapor deposition. Appl Phys A 2002;74:419–22. [89] Baek Y-G, Honda S-I, Ikuno T, Ohkura S, Katayama M, Hirao T, et al. Formation of graphite layers during carbon nanotubes growth. Jpn J Appl Phys 2003;42:579–81. [90] Yang Y, Hu Z, Tian YJ, Lu¨ YN, Wang XZ, Chen Y. High-yield production of quasi-aligned carbon nanotubes by catalytic decomposition of benzene. Nanotechnology 2003;14:733–7. [91] Wang J-G, Liu C-J, Zhang Y-P, Zhu X-L, Yu K-L, Zou J. Effect of catalyst preparation on carbon nanotube growth. Mol Simult 2003;29:667–70. [92] Cassel AM, McCool GC, Tee Ng H, Koehne JE, Chen B, Li J, et al. Carbon nanotube networks by chemical vapor deposition. Appl Phys Lett 2003;82:817–9. [93] Chen B, Parker II G, Han J, Meyyappan M, Cassell AM. Heterogeneous single-walled carbon nanotube catalyst discovery and optimization. Chem Mater 2002;14:1891. [94] Maruyama S, Kojima R, Miyauchi Y, Chiashi S, Kohno M. Low-temperature synthesis of highpurity single-walled carbon nanotubes from alcohol. Chem Phys Lett 2002;360:229. [95] Murakami Y, Miyauchi Y, Chiashi S, Maruyama S. Characterization of single-walled carbon nanotubes catalytically synthesized from alcohol. Chem Phys Lett 2003;374:53–8.

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

959

[96] Willems I, Ko´nya Z, Colomer J-F, Van Tendeloo G, Nagaraju N, Fonseca A, et al. Control of the outer diameter of thin carbon nanotubes synthesized by catalytic decomposition of hydrocarbons. Chem Phys Lett 2000;317:71–6. [97] Kong J, Cassel AM, Dai H. Chemical vapor deposition of methane for single-walled carbon nanotubes. Chem Phys Lett 1998;292:567–74. [98] Thess A, Lee R, Nikolaev P, Dai H, Petit P, Robert J, et al. Crystalline ropes of metallic carbon nanotubes. Science 1996;273:483–7. [99] Mann D, Javey A, Kong J, Wang Q, Dai H. Ballistic transport in metallic nanotubes with reliable PD ohmic contacts. Nano Lett 2003;3:1541–4. [100] Tsai SH, Chao CW, Lee CL, Shih HC. Bias-enhanced nucleation and growth of the aligned carbon nanotubes with open ends under microwave plasma synthesis. Appl Phys Lett 1999;74:3462. [101] Thomann AL, Salvetat JP, Breton Y, andreazza Vignolle C, Brault P. Thermal stability of metal nanoclusters formed by low-pressure plasma sputtering. Thin Solid Films 2003;428:242–7. [102] Vajtai R, Korda´s K, Wei BQ, Be´ke´si J, Leppa¨vuori S, George TF, et al. Carbon nanotube network growth on palladium seeds. Mater Sci Eng C 2002;19:271–4. [103] Cui H, Eres G, Howe JY, Puretkzy A, Varela M, Geohegan DB, et al. Growth behavior of carbon nanotubes on multilayered metal catalyst film in chemical vapor deposition. Chem Phys Lett 2003;374:222–8. [104] Delzeit L, Chen B, Cassel A, Stevens R, Nguyen C, Meyyappan M. Multi-layered metal catalysts for controlling the density of single-walled carbon nanotube growth. Chem Phys Lett 2001;348:368–74. [105] Franklin NR, Li Y, Chen RJ, Javey A, Dai H. Patterned growth of single-walled carbon nanotubes on full 4-in. wafers. Appl Phys Lett 2001;79:4571. [106] Hafner JH, Bronikowski MJ, Azamian BR, Nikolaev P, Rinzler AG, Colbert DT, et al. Catalytic growth of single-wall carbon nanotubes from metal particles. Chem Phys Lett 1998;296:195–202. [107] Harutyunyan AR, Pradhan BK, Kim UJ, Chen G, Eklund PC. CVD synthesis of single wall carbon nanotubes under ‘‘soft’’ conditions. Nano Lett 2002;2:525–30. [108] Hornyak GL, Grigorian L, Dillon AC, Parilla PA, Jones KM, Heben MJ. A temperature window for chemical vapor decomposition growth of single-wall carbon nanotubes. J Phys Chem B 2002;106:2821–5. [109] Su M, Zheng B, Liu J. A scalable CVD method for the synthesis of single-walled carbon nanotubes with high catalyst productivity. Chem Phys Lett 2000;322:321–6. [110] Su M, Li Y, Maynor B, Buldum A, Lu JP, Liu J. Lattice-oriented growth of single-walled carbon nanotubes. J Phys Chem B 2000;104:6505. [111] Zheng B, Li Y, Liu J. Cvd synthesis and purification of single-walled carbon nanotubes on aerogelsupported catalyst. Appl Phys A 2002;74:345–8. [112] Ago H, Ohshima S, Uchida K, Yumura M. Gas-phase synthesis of single-wall carbon nanotubes from colloidal solution of metal nanoparticles. J Phys Chem B 2001;105:10453. [113] Ago H, Ohshima S, Uchida K, Komatsu T, Yumura M. Carbon nanotube synthesis using colloidal solution of metal nanoparticles. Physica B 2002;323:306–7. [114] Alvarez W, Kitiyanan B, Borgna A, Resasco D. Synergism of Co and Mo in the catalytic production of single-wall carbon nanotubes by decomposition of CO. Carbon 2001;39:547–58. [115] Herrera JE, Balzano L, Borgna A, Alvarez WE, Resasco DE. Relationship between the structure/ composition of Co–Mo catalysts and their ability to produce single-walled carbon nanotubes by CO disproportion. J Catal 2001;204:129–45. [116] Herrera JE, Resasco DE. Loss of single-walled carbon nanotubes selectivity by disruption of the Co– Mo interaction in the catalyst. J Catal 2004;221:354–64. [117] Kitiyanan B, Alvarez WE, Harwell JH, Resasco DE. Controlled production of single-wall carbon nanotubes by catalytic decomposition of CO on bimetallic Co–Mo catalysts. Chem Phys Lett 2000;317:497–503. [118] Lan A, Iqbal Z, Aitouchen A, Libera M, Grebel H. Growth of single-wall carbon nanotubes within an ordered array of nanosize silica spheres. Appl Phys Lett 2002;81:433–5. [119] Liao XZ, Serquis A, Jia QX, Peterson DE, Zhu YT. Effect of catalyst composition on carbon nanotube growth. Appl Phys Lett 2003;82:2694.

960

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

[120] Murakami Y, Miyauchi Y, Chiashi S, Maruyama S. Direct synthesis of high-quality single-walled carbon nanotubes on silicon and quartz substrates. Chem Phys Lett 2003;377:49–54. [121] Tang S, Zhong Z, Xiong Z, Sunu L, Liu L, Lin J, et al. Controlled growth of single-walled carbon nanotubes by catalytic decomposition of CH4 over Mo/Co/MgO catalysts. Chem Phys Lett 2001;350:19–26. [122] Hu YH, Ruckenstein E. High-resolution transmission electron microscopy study of carbon deposited on the NiO/MgO solid solution catalysts. J Catal 1999;184:298–302. [123] Lueking A, Yang RT. Hydrogen spillover from a metal oxide catalyst onto carbon nanotubes— implications for hydrogen storage. J Catal 2002;206:165–8. [124] Flahaut E, Govindaraj A, Peigney A, Laurent C, Rousset A, Rao C. Synthesis of single-walled carbon nanotubes using binary (Fe, Co, Ni) alloy nanoparticles prepared in situ by the reduction of oxide solid solutions. Chem Phys Lett 1999;300:236–42. [125] Flahaut E, Peigney A, Laurent C, Rousset A. Synthesis of single-walled carbon nanotube–Co–MgO composite powders and extraction of the nanotubes. J Mater Chem 2000;10:249–52. [126] Sinnott SB, Andrews R, Qian D, Rao AM, Mao Z, Dickey EC, et al. Model of carbon nanotube growth through chemical vapor deposition. Chem Phys Lett 1999;315:25–30. [127] Lee CJ, Park J, Kim JM, Huh Y, Lee JY, No KS. Low-temperature growth of carbon nanotubes by thermal chemical vapor deposition using Pd, Cr, and Pt as co-catalyst. Chem Phys Lett 2000;327:277–83. [128] Kichambare PD, Qian D, Dickey EC, Grimes CA. Thin film metallic catalyst coatings for the growth of multiwalled carbon nanotubes by pyrolysis of xylene. Carbon 2002;40:1903–9. [129] Ruban A, Hammer B, Stoltze P, Skriver HL, Nørskov JK. Surface electronic structure and reactivity of transition and noble metals. J Mol Catal A Chem 1997;115:421–9. [130] Kim MS, Rodriguez NM, Baker RTK. The role of interfacial phenomena in the structure of carbon deposits. J Catal 1992;134:253–68. [131] Dai H. Carbon nanotubes: opportunities and challenges. Surf Sci 2002;500:218–41. [132] Kock AJHM, De Bokx PK, Boellaard E, Klop W, Geus JW. The formation of filamentous carbon on iron and nickel catalysts. J Catal 1985;96:468–80. [133] Nishimura K, Okazaki N, Pan L, Nakayama Y. In situ study of iron catalysts for carbon nanotube growth using X-ray diffraction analysis. Jpn J Appl Phys 2004;43:L471–4. [134] Herreyre S, Gadelle P, Moral P, Millet JMM. Study by Mo¨ssbauer spectroscopy and magnetization measurement of the evolution of iron catalysts used in the disproportionation of CO. J Phys Chem Solids 1997;58:1539–45. [135] Yoshida N, Yasutake M, Arie T, Akita S, Nakayama Y. Quantitative analysis of the magnetic properties of metal-capped carbon nanotube probe. Jpn J Appl Phys 2002;41:5013–6. [136] Arie T, Nishijima H, Akita S, Nakayama Y. Carbon-nanotube probe equipped magnetic force microscope. J Vac Sci Technol B 2002;18:104–6. [137] Lan A, Zhang Y, Zhang X, Iqbal Z, Grebel H. Is molybdenum necessary for the growth of singlewall carbon nanotubes from CO? Chem Phys Lett 2003;379:395–400. [138] Pinheiro JP, Gadelle P. Chemical state of a supported iron–cobalt catalyst during CO disproportionation. I. Thermodynamic study. J Phys Chem Solids 2001;62:1015–21. [139] Lambert RM, Pacchioni G. Chemisorption and reactivity on supported clusters and thin films. NATO ASI Series E, 331. Dordrecht, The Netherlands: Kluwer; 1997. [140] Lo´pez N, Illas F, Pacchioni G. Electronic effects in the activation of supported metal clusters: density functional theory study of H2 dissociation on Cu/SiO2. J Phys Chem B 1999;103:8552. [141] Audier M, Oberlin A, Coulon M. Crystallographic orientations of catalytic particles in filamentous carbon: case of simple conical particles. J Cryst Growth 1981;55:549–56. [142] Vander Wal RL, Ticich TM, Curtis VE. Substrate–support interactions in metal-catalyzed carbon nanofiber growth. Carbon 2001;39:2277–89. [143] Ida T, Tsuiki H, Ueno A, Tohji K, Udagawa Y, Iwai K, et al. Characterization of iron oxide in Fe2O3/SiO2 catalyst. J Catal 1987;106:428–39. [144] Kukovecz A, Ko´nya Z, Mo¨nter D, Reschetilowski W, Kiricsi I. UV–VIS investigations on Co, Fe and Ni incorporated into sol–gel SiO2–TiO2 matrices. J Mol Struct 2001;563–564:403–7.

A.-C. Dupuis / Progress in Materials Science 50 (2005) 929–961

961

[145] de los Arcos T, Vonau F, Garnier MG, Thommen V, Boyen H-G, Oelhafen P, et al. Influence of iron–silicon interaction on the growth of carbon nanotubes produced by chemical vapor deposition. Appl Phys Lett 2002;80:2383. [146] Baker RTK, Alonzo JR, Dumesic JA, Yates DJC. Effect of the surface state of iron on filamentous carbon formation. J Catal 1982;77:74–84. [147] Lee TY, Han J-H, Choi SH, Yoo J-B, Park C-Y, Jung T, et al. Comparison of source gases and catalyst metals for growth of carbon nanotube. Surf Coat Technol 2003;169–170:348–52.