Flame and Furnace Synthesis of Single-Walled and Multi-Walled

ShreVeport, Louisiana 71134. ReceiVed: July 24, 2001. Results are presented for flame synthesis of metal-catalyzed carbon nanotubes. A thermal evapora...
0 downloads 0 Views 799KB Size
J. Phys. Chem. B 2001, 105, 10249-10256

10249

Flame and Furnace Synthesis of Single-Walled and Multi-Walled Carbon Nanotubes and Nanofibers Randall L. Vander Wal* NCMR c/o NASA-Glenn, M.S. 110-3, 21000 Brookpark Road, CleVeland, Ohio 44135

Thomas M. Ticich Centenary College of Louisiana, Department of Chemistry, 2911 Centenary BouleVard, ShreVeport, Louisiana 71134 ReceiVed: July 24, 2001

Results are presented for flame synthesis of metal-catalyzed carbon nanotubes. A thermal evaporation technique is used to create the catalyst nanoparticles of Fe or Ni through gas condensation followed by entrainment into the flame. Results are compared with those using a high-temperature tube furnace to provide the reactive environment. Each system yields consistent results, with CO/H2 mixtures generally yielding single-walled nanotubes (SWNTs) with Fe while C2H2/H2 mixtures usually produce multiwalled nanotubes (MWNTs) with Ni. A ternary gas mixture of CO/C2/H2 produces a better yield of nanofibers than either a CO/H2 or C2H2/H2 mixture at 700 °C with Ni catalyst. Our results reflect a combination or possibly a synergy between thermal- plus adsorbate-induced restructuring and adsorbate-particle steric factors affecting particle structure and reactivity.

Introduction Metal-catalyzed carbon nanotubes are highly sought for a diverse range of applications that include nanoelectronics,1-3 battery electrode material,4 catalysis,5 hydrogen storage media,6,7 and reinforcing agents in polymer composites.8-11 These latter applications will require vast quantities of nanotubes at competitive prices to be economically feasible. Moreover, reinforcing applications may not require ultrahigh purity nanotubes. Indeed, functionalization of nanotubes to facilitate interfacial bonding within composites will naturally introduce defects into the tube walls, lessening their tensile strength.12 Current methods of aerosol synthesis of carbon nanotubes include laser ablation of bicomposite targets of carbon and catalyst metal within high-temperature furnaces13-15 and decomposition of a metal-containing gaseous precursor within a tube furnace.16-18 Common to each approach is the generation of particles in the presence of the reactive hydrocarbon species at elevated temperatures. In the laser-ablation approach, the situation is even more dynamic in that particles and nanotubes are borne during the transient cooling phase of the laser-induced plasma for which the temperature far exceeds that of the surrounding hot gases within the furnace process tube.19 A shared limitation is that more efficient methods of nanoparticle synthesis are not readily incorporated into these approaches. In constrast, combustion can quite naturally create nanomaterials such as TiO220 and carbon black.21 Flame synthesis is well-known for its commercial scalability and energy efficiency.22,23 However, flames do present a complex chemical environment with potentially steep gradients in temperature and species concentrations.24,25 Moreover, reaction times are limited within buoyantly driven flows to tens of milliseconds.26 A * Author to whom correspondence should be addressed. Phone: (216) 433-9065. Fax: (216) 433-3793. E-mail: [email protected].

further complication is that several variables can be intertwined. For example, increasing the inert concentration not only dilutes the reactive hydrocarbon but also depresses the global flame temperature.27 Given the viability of flame synthesis for other materials, we conducted preliminary investigations of nanotube synthesis using a pyrolysis flame. This flame system afforded the advantage that a reactant gas concentration could be tested that may not support an independent diffusion flame. Uniformity of temperature is also maintained along the greater length of the flame or equivalently during a larger portion of the residence time within the flame. To better understand the process variables within a flame environment, we report results from a process in which the nanoparticle synthesis is separated from that of nanotube growth. This separation affords the opportunity to discern fundamentals of nanotube growth related to particle size, crystallinity, and composition. In addition, the separation of processes was deemed necessary given the chemically complex environment of the flame. This allowed for tailoring the flame environment with respect to temperature and reactant gases to optimize nanotube growth, apart from catalyst synthesis. To further guide the choice of flame conditions for nanotube growth, we conducted preliminary experiments to characterize the catalyst synthesis method and subsequent nanotube syntheses by introducing the preformed aerosol catalyst into a high-temperature tube furnace. That the nanotube morphology depends on the catalyst composition is well-documented. Far less studied is the effect of the reactant gas identity upon nanotube growth and structure. In the case of hydrocarbons, time-dependent pyrolysis processes result in a range of species that may be more or less reactive toward the dissociative adsorption necessary for nanotube growth. In fact, polycyclic aromatic hydrocarbons (PAHs) may

10.1021/jp012838u CCC: $20.00 © 2001 American Chemical Society Published on Web 10/03/2001

10250 J. Phys. Chem. B, Vol. 105, No. 42, 2001

Vander Wal and Ticich

impede or prohibit nanotube growth by virtue of their thermodynamic stability and associated resistance toward catalytic dissociation. Therefore, the work presented here also investigates the effect of reactant gas upon nanotube morphology. Experimental Section Catalyst nanoparticles were generated by means of a thermal evaporation technique.28 An electric current was passed through a 3.2-mm diameter carbon rod in which a trough had been made so as to resemble a canoe. The desired metal powder was then deposited in the trough. During resistive heating, an inert gas flow of argon was directed across the carbon rod to entrain evaporated metal. Scatter from a laser beam intersecting the gas flow downstream of the evaporator provided a visual guide of the vaporization rate and thus a means of feedback contol. The resulting catalyst aerosol was then directed to either a hightemperature tube furnace or to the fuel inlet tube of the burner supporting the pyrolysis flame. A high-temperature tube furnace was used both to characterize the sintering and annealing behavior of the metal nanoparticles and to identify conditions of temperature and gas species favorable for nanotube growth. Metal catalyst particles were mixed with the reactant gas mixture upon entrance into a 91.4cm long alumina process tube with a 2.5-cm outer diameter. The pyrolysis flame was established on a 1.1-cm outer diameter brass fuel tube running through the center of a McKenna burner. Metal nanoparticles and reactant gases were mixed prior to introduction to this fuel tube so as to establish a homogeneous mixture. Upon emerging from the fuel tube, the aerosol mixture was heated by the surrounding post-flame gases from a rich, premixed flame supported on the sintered metal top of the McKenna burner. The fuel-rich premixed flame was fueled by 11.0 slm air and 1.5 slm C2H2. Previous investigations found that equivalence ratios ranging from 1.4 to 1.62 had no demonstrable effect on the pyrolysis flame.29 A 7.5-cm long, 2.5-cm outer diameter steel chimney placed 1 cm above the burner served to stabilize the pyrolysis flame. Material samples from the flame were obtained by thermophoretic sampling 1 cm above the chimney. In this technique, aerosol material within a flow is driven to collect on a cold surface by a temperature gradient between the flow and surface.30 A double-action air-driven piston served to insert and retract a rod to which the sampling probe was attached. Transmission electron microscopy (TEM) grids were attached to the probe by a sandwich grid holder consisting of a 0.075mm thick brass shim with a 2 mm diameter hole exposing both sides of a holey TEM grid, as described previously.31 This grid holder was attached to the insertion rod of the probe by a small set screw. The dwell time of the probe within the flame was controlled by custom electronics that actuated a dual-valve solenoid to govern the pressurized air flow. Probe dwell times within the flow were kept short (250 ms) so as to minimize probe heating. For the dwell times used here, previous measurements have registered a probe temperature elevation of less than 200 °C.32 As a consequence, secondary reactions of sampled material with the probe surface are quenched. All samples were collected from the center axial region of the flames where radial gradients in species concentration and temperature are small.24,25 Both TEM and high-resolution transmission electron microscopy (HRTEM) were performed on a Phillips CM200 instrument fitted with Gatan image filter and real-time digital Fourier transform imaging.

Figure 1. A low-magnification TEM image of Fe nanoparticles produced by the thermal evaporation/condensation of Fe within Ar as an inert carrier gas.

Results and Discussion A. Particle Characterization. As a first step toward using the electric thermal evaporator as a particle source for nanotube growth, we characterized the size and composition of the generated particles. The synthesis of nanoparticles from a condensing plume of metal atoms is critically dependent upon the atom density. Both the rate of metal vaporization and carrier gas (Ar) flow can be used to govern this parameter. The range of carrier gas flow rates was set so as to achieve residence times within the heated zone of the furnace of roughly 20 s, a value commensurate with residence times reported for other nanotube synthesis methods.13-18 The current through the carbon rod boat, which determines the ambient temperature of the metal, determined the metal vaporization rate in our apparatus. High currents, which correlated with a large amount of scattered laser light and hence a high vaporization rate, led to the production of metal particles. Figure 1 provides an overview of the metal “soot” generated by condensation of the metal vapor. The metal particles, observed individually and joined together in aggregates, had diameters ranging from less than 1 nm to greater than 10 nm. Lower vaporization rates (achieved by lower heating currents) led to an increasing relative abundance of individual nanoparticles of decreasing size. Conditions that produced the desired small unaggregated nanoparticles were identified by TEM analysis of sampled metal aerosol. B. Tube FurnacesFe and Ni. To test the viability of the metal aerosol for nanotube growth, we connected the electric thermal evaporator to the gas input side of an alumina tube running through a high-temperature tube furnace. Reactant gases were introduced separately and then combined through a tee connection with the inert carrier (Ar) for the metal nanoparticles evolved. Figures 2 and 3 show results for Ni and Fe, respectively, using a CO/H2 mixture as the reactant gases. Both SWNTs and encapsulated metal nanoparticles were abundant at 1000 °C, with much lesser amounts of each produced at 700 °C. Figure 4a shows a HRTEM image of a double-walled nanotube obtained using Fe within a CO/H2/Ar mixture at 1000 °C. The inner nanotube displayed in Figure 4a possesses a similar diameter to that of the SWNT shown in Figure 4b for comparison. The figures also illustrate the range of catalyst

Synthesis of Carbon Nanotubes and Nanofibers

J. Phys. Chem. B, Vol. 105, No. 42, 2001 10251

Figure 3. HRTEM images of Fe-catalyzed SWNTs using the same CO/H2 mixture and reaction conditions as for Ni. Figure 3b also illustrates the range of particle sizes produced by the thermal evaporator.

Figure 2. TEM images of encapsulated Ni nanoparticles and a Nicatalyzed SWNT produced by Ni aerosol entrained in a CO/H2 mixture flowing through the tube furnace at 1000 °C. Figure 2b shows the graphitic carbon encapsulation characteristic of the CO/H2 mixture. Figure 2c illustrates the high degree of graphitic encapsulation observed at 1000 °C. In each case, the CO and H2 flows were 200 and 100 sccm, respectively, with the nebulized nitrate particles entrained in 0.1 slm Ar.

particle sizes produced by the thermal evaporator as well as the range of SWNT diameters obtained. The relative yield of Fe-catalyzed MWNTs was comparable to that of Fe-catalyzed SWNTs with CO at 700 °C.

The metal particles produced by coalescence of condensing Fe atoms can be described as an amorphous cluster lacking any crystalline geometry or facets. While clusters are known for their high chemical reactivity, particularly toward insertion within hydrocarbon bonds,33,34 variations in particle structure or geometry with size may give rise to a size-dependent reactivity.35 For example, a lack of crystalline lattice planes and interstitial sites for carbon atom migration may prohibit dissociative adsorption and carbon diffusion,36 necessary steps for nanotube or nanofiber growth.37 Thus SWNT synthesis does not begin until the particles achieve a suitable structure. In contrast, small metal particles may be highly active toward dissociative adsorption, if not toward nanotube synthesis. Thus the relative variation of encapsulated metal nanoparticles versus SWNTs may reflect the relative rates of particle deactivation by encapsulation versus SWNT synthesis, as dictated by the change in particle structure with size. This variation in structure may be manifested as a size-dependent reactivity and/or preferential reactivity toward particular gases, as discussed next. In addition to temperature, the product (SWNTs or nanofibers) relative yield depended upon the reactant gas composition. Fe produced mainly SWNTs while Ni yielded mainly encapsulated particles using CO/H2 mixtures. Within C2H2 mixtures, the metal catalyst roles were nearly reversed; Fe was far less active toward C2H2 mixtures, producing primarily encapsulated metal nanoparticles while Ni yielded mainly nanofibers. Finally, the

10252 J. Phys. Chem. B, Vol. 105, No. 42, 2001

Figure 4. HRTEM images of a double-walled nanotube (4a) and SWNT (4b) catalyzed by Fe at 1000 °C. The same CO/H2 mixture and reaction conditions were used in the tube furnace.

relative nanotube yields produced by a given metal were significantly greater at 1000 °C than at 700 °C. Given that the particle size governs the nanotube dimensions, the variation of major product (MWNT vs SWNT) with different reactive gases points to a size-dependent reactivity. Differences in the extent of graphitization for the encapsulated particles were also apparent between the two reactive gases mixtures. CO/H2 mixtures led to highly graphitic encapsulation of the metal nanoparticles, particularly with Ni (Figure 2b). In contrast, C2H2/H2/Ar mixtures produced amorphous encapsulation as defined by an absence of extended carbon lamella as observed by lattice fringe imaging. Lower C2H2 concentrations produced less encapsulation and a reduction in amorphous material but without any significant increase in MWNT growth, for both metals. The product dependence upon the reactant gas composition and relative yield dependence on temperature suggest these as causitive factors underlying structural differences between the metal nanoparticles. These differences become manifest as a particle size-dependent reactivity, producing different nanotube products. Stated differently, the parameters of reactant gas composition and temperature determine the reactivity differences between Fe and Ni nanoparticles of comparable size. This catalytic preference toward SWNTs or MWNTs may reflect the susceptibility of the metal nanoparticle toward restructuring induced by adsorption of CO.38 Thus Fe, known to undergo

Vander Wal and Ticich such restructuring upon CO adsorption, begins to catalyze SWNTs at an earlier particle growth stage (smaller particle). In constrast, Ni, not undergoing such restructuring upon CO adsorption, remains inactive toward catalyzing nanotube growth from CO. Rather, Ni nanoparticles eventually become encapsulated. In comparison, C2H2 does not induce such restructuring (of either Fe or Ni).39 Consequently small Fe nanoparticles do not readily initiate SWNTs with C2H2 as reactant, reflecting both catalyst particle size and associated reactivity. Instead, Fe nanoparticles become encapsulated. Small Ni nanoparticles are not particularly active toward either CO or C2H2. This lack of preferential reactivity is consistent with neither reactant inducing restructuring of the nascent Ni nanoparticles. Larger Ni nanoparticles appear to possess a catalytically active form/structure (perhaps similar to that of the bulk material). As a result, they are catalytically active toward C2H2 (and weakly toward CO). Given their size, MWNTs/nanofibers result. Additionally, the higher nanotube yield at 1000 °C may reflect the kinetics of the individual reaction steps in nanotube growth. Dissociative adsorption of CO via the Boudard reaction is wellknown to be facilitated by elevated temperatures.40 Surface or interstitial diffusion of carbon to the growing nanotube/nanofiber wall(s) would also increase with increasing temperature. These processes could also be aided by restructuring of the catalyst particle. Thermally induced restructuring, aided by the elevated temperature, may convert the amorphous particle to a more crystalline structure.41 Moreover, adsorption of CO may also render the particle more susceptible to thermally induced restructuring.38 This combination of restructuring processes could increase the rate of dissociative decomposition and carbon atom diffusion (through or around the particle perimeter). With a higher carbon atom supply rate, a higher SWNT yield at 1000 °C compared to 700 °C would result within the furnace for Fe. In contrast, the high pyrolysis rate of C2H2 simply leads to PAHs42 which could readily deactivate catalyst particles, thereby accounting for the lack of increasing nanofiber production by Ni at 1000 °C. Finally, in addition to adsorbate and thermally induced restructuring, the catalyst particle size-related structure may impose steric restrictions upon molecular adsorbates. For example, CO is known to adsorb end-on while unsaturated hydrocarbons such as acetylene adsorb sideways upon transition metal surfaces, based on bulk single-crystal studies.39 Thus smaller particles (of Fe and Ni) may be inherently less active toward C2H2 dissociative adsorption than larger particles, simply based on steric factors. Conversely, large particles may possess suitable active site geometry and thus be reactive towards both CO and C2H2. Other factors, as discussed would then determine the relative reactivity of the catalyst nanoparticles toward these two gases. In summary, the difference in reactivity as a function of particle size for Fe and Ni toward CO and C2H2 reflects a combination of steric factors, adsorbate- and thermally induced restructuring and potentially nascent nanoparticle structure. These structural differences then impart a size-dependent reactivity which translates into different nanotube catalysis mechanisms (and products). That different products are observed (SWNTs or nanofibers) using CO or C2H2 with a given catalyst metal, e.g., Fe, then becomes a consequence of the nanotube dimensions (and type) reflecting the catalyst particle size. In light of the above considerations, the synthesis of a doublewalled nanotube shown in Figure 4 reflects a fortuitous combination of particle size and lattice plane spacing that

Synthesis of Carbon Nanotubes and Nanofibers

J. Phys. Chem. B, Vol. 105, No. 42, 2001 10253

Figure 5. A HRTEM image of an Fe-catalyzed SWNT produced within the pyrolysis flame with a CO/H2/Ar mixture. The flow rates were 0.5, 0.5 and 0.5 slm, respectively.

Figure 6. A HRTEM image of a nongraphitic Fe-catalyzed MWNT produced within the pyrolysis flame using a C2H2/H2/Ar mixture. The flow rates were 0.25, 0.25, and 0.5 slm, respectively.

matches the diameter of the helical benzene ring structure that forms the nanotube walls. Formation of a SWNT of diameter matching either the inner or outer diameter of the double-walled nanotube in Figure 4 ought not to incur excessive bond strain, i.e., unfavorable energetics, yet does not occur.43 It is possible that the two concentric nanotubes provide a stabilizing influence for each other.44 Of significance here is that the catalyst metal nanoparticles need not be produced in-situ within the reactive growth environment but can be synthesized separately. Hence a wide variety of catalyst preparation techniques may be used for nanotube growth. As no particle size selection process was imposed here, our results reveal the range of reactivities brought about by catalyst particles of different sizes. Size selection processes as used for aerosol processing then offers possibilities for governing the size distribution and morphology of the nanotubes produced. C. FlamesFe. Analogous to the tube furnace results, SWNTs were produced by Fe particles with CO/H2 gas mixtures in the pyrolysis flame. We observed a strong dependence of the relative yield with total gas flow within the pyrolysis flame and with the relative concentrations of CO and H2. Figure 5 shows a HRTEM image of an Fe-catalyzed SWNT, as sampled directly from the flame. As observed with CO/H2 mixtures in the tube furnace, there was a lack of amorphous carbon coverage on the SWNT walls, indicative of an absence of pyrolysis products within the reacting flow. In contrast to the SWNTs produced with CO/H2/He mixtures, C2H2/H2/He mixtures resulted in amorphous carbon nanotubes (nanofibers) that possess nongraphitic walls, but in far lower abundance. Figure 6 shows that the tube walls contain short, discontinuous, randomly oriented graphene segments or short stacks of graphitic lamella. The degree of disorder varies but in no instance did these nanofibers exhibit an extended graphitic order. RelatiWe Yield. The relative reactivity of Fe toward CO and C2H2 gas mixtures mirrors the results obtained using the furnace. Similarly, this difference in reactivity likely reflects a combination of steric factors, adsorbate- and thermally induced restructuring and potentially nascent nanoparticle structure. The general absence of amorphous carbon upon the nanotubes and nanoparticles within CO/H2 mixtures illustrates yet an additional factor. Production of gas-phase C2 species and derived pyrolysis products (PAHs) that can then deposit inert graphitic segments

is highly unfavorable relative to other reactions involving atomic carbon deposition directly from CO.45 Thus within the CO/H2 mixture, there is an absence of C2 species and PAHs as witnessed by the general lack of amorphous carbon upon either the nanotubes or nanoparticles. Catalyst particles commensurate in size with SWNTs likely possess only a few sites active toward dissociative adsorption. Blockage of these sites by PAHs and their partial dehydrogenation products can completely deactivate a nanometer-sized particle toward SWNT growth.46,47 An absence of PAHs would then permit catalyzed growth of SWNTs. In contrast, a larger particle may possess many active sites for dissociative adsorption and carbon dissolution. Blockage of one or two of these sites may inhibit or shut down carbon atom layer production that forms a portion of the nanofiber walls while still allowing the rest of the structure to form. Thus MWNTs could still grow in C2H2/H2/He mixtures, although in likely far lower quantities. Thus the different products produced by Fe within CO/H2 versus C2H2/H2 mixtures may also partially reflect the deactivation of the nascent Fe nanoparticles by PAHs arising through pyrolysis of C2H2. The lack of deactivation by CO/H2/He mixtures could also account for the relative abundance of SWNTs compared to MWNTs with CO mixtures. In the absence of deactivating species, small Fe particles are active toward SWNT synthesis upon achieving sufficient size (around 1 nm). Their high catalytic activity (enhanced by previously discussed factors) precludes further particle growth to form larger particles that could catalyze MWNTs. With most Fe particles producing SWNTs, the relative yield of MWNTs is low. In contrast, C2H2/H2 mixtures, forming PAHs,42 readily deactivate nascent Fe nanoparticles. With the catalytic activity of small nanoparticles being highly sensitive to deactivation, SWNT synthesis is virtually shut down. Large particles, possessing many more active sites, are less sensitive to adsorption of a PAH molecule. Blockage of some sites active toward dissociative decomposition or precipitation of dissolved carbon would not prevent nanofiber growth. With undulating discontinuous carbon lamella forming the nanotube walls, there is not a structural or size constraint between the nanotube and catalysis nanoparticle as for SWNTs. Rather, changes in the carbon atom supply would be reflected in the nanofiber morphology. The relative nanofiber yield would then reflect the ratio of active sites to those deactivated by continued deposition and deactivation by PAHs. If, however, the PAH concentration

10254 J. Phys. Chem. B, Vol. 105, No. 42, 2001

Vander Wal and Ticich

Figure 7. A HRTEM image of nanotubes and graphitically encapsulated Ni nanoparticles captured from the pyrolysis flame using a CO/ H2/Ar mixture. The flows were 0.5, 0.5, and 0.5 slm, respectively.

is too large, deactivation will be complete. Once a PAH molecule adsorbs, its partial decomposition can provide additional sites for reactive hydrocarbon deposition leading to further carbonaceous layer growth.42,45 The production of MWNTs/nanofibers of varying graphitic quality within the same gas environment may again reflect a combination of factors related to the particle structure and gasphase environment. The one constant factor is that the nanotube morphology will reflect the structure of the catalyst nanoparticle. Thus restructuring processes leave their legacy. If partially decomposed PAHs block some active sites, a partial termination of carbon atom layer plane production will be observed as a discontinuation of lattice planes in the MWNT/nanofiber wall, as illustrated in Figure 6. D. FlamesNi. The Ni results in the flame environment parallel some of the Fe results, with the production of some nanostructures dependent on the reactant gas identity. Ni in the CO/H2 mixture produced a few SWNTs, though with far less yield than Fe. The SWNT synthesis with Ni was less dependent upon the CO/H2 ratio. We did occasionally observe a variety of other structures. These included nanotubes, but mainly encapsulated nanoparticles (Figure 7) and a few MWNTs. In the CO-fueled flame, there was a general absence of amorphous coverage upon either the nanotubes or encapsulated particles. Also similar to observations with Fe, the encapsulation of the particles was highly graphitic. C2H2/H2 mixtures resulted in an absence of SWNTs although MWNTs were found, as was the case for Fe, but in much greater abundance. Unlike the Fe results, the graphitic quality of the MWNTs did not appear to depend strongly upon the reactive gas composition for either CO- or C2H2-fueled flames. As Figure 8 illustrates, a range of MWNT structures were observed with C2H2-fueled flames. The structure seen in Figure 8a is best described as a hollow, nongraphitic nanofiber (HNGF) in which relatively unordered graphene segments comprise the tube wall. The degree of order, although greater than that of amorphous carbon, is low judging from the length of the individual graphene segments. Figure 8b shows a structure with a clearly discernible central channel. The “tube” wall consists of several layers of turbostratically oriented, partially graphitic carbon platelets which undulate frequently along the tube axis. The yield and purity of these nanofibers were highly dependent upon the relative C2H2-to-H2 ratio and C2H2 concentration. Higher concentrations of C2H2 relative to H2 resulted in

Figure 8. HRTEM images of MWNTs/nanofibers of varying graphitic quality produced within the pyrolysis flame using a C2H2/H2/Ar mixture with Ni catalyst using flows of 0.5, 0.5, and 0.5 slm, respectively.

nanofibers encased in amorphous carbon and an increasing prevalence of nongraphitic, amorphous structures. Consistent with this is that increasing concentrations of C2H2 led to more nanoparticles encapsulated in amorphous carbon. In general, these results for Ni in a flame parallel those observed using the high-temperature furnace. The markedly lower nanotube/nanofiber yield with Ni may reflect a combination of inherently lower catalytic activity (particle structure affecting both activity toward dissociative adsorption and activity towards nanofiber growth) and absence of adsorbate plus thermally induced restructuring. The variety of morphologies and relative lack of graphitic quality observed for nanofibers catalyzed by Ni with C2H2/H2 stands in contrast to the Fe results, where SWNTs and occasionally highly graphitic MWNTs were observed. These observations suggest that the Ni nanoparticles do have a different structure than those of Fe, manifesting itself in the nanotube/nanofiber morphology, in agreement with previous findings48,49 using these metals as catalysts. Apparently this structure also possesses a lower relative reactivity toward both SWNT and nanotube or nanofiber catalysis. E. FlamesFe and Thiophene. Sulfur is a well-known promoter of nanotube growth as demonstrated frequently by thiophene addition to reacting mixtures for nanotube synthesis involving high-temperature tube furnaces.50-52 Well-accepted

Synthesis of Carbon Nanotubes and Nanofibers

J. Phys. Chem. B, Vol. 105, No. 42, 2001 10255

Figure 9. A HRTEM image of a cluster of Ni-catalyzed, hollow, nongraphitic carbon nanotubes produced within the pyrolysis flame using a mixture of CO/C2H2/H2/Ar.

Figure 10. A HRTEM image of a portion of the nongraphitic nanotubes seen in Figure 9 illustrating the amorphous structure of the nanotube walls.

explanations include restructuring of the particle or steric blockage of specific surface sites. Selective blockage of surface sites can steer adsorbing species to locations preferential toward dissociative adsorption or inhibit formation of graphitic segments that could block active decomposition sites. Therefore, we entrained thiophene vapor at 0.8% in the reactant gas to test its effects with our methods. Our results show that thiophene did not improve the relative abundance of the nanotubes nor alter their morphology. In addition, the extent of graphitization for encapsulating carbon remained the same in those structures. That no increase in MWNT production was observed suggests that the putative mechanisms for the enhancing role of thiophene described above are not operative under our conditions. F. FlamesNi with Bicomponent Reactive Gas Mixture. Given the differences between the nanostructures obtained for Ni with CO and C2H2, we tested reactive gas mixtures with both gases present. Figure 9 shows that hollow nongraphitic nanofibers (HNGFs) were produced with CO and C2H2 copresent in the reactive gas mixture. The nanofibers appeared as nested mats reflecting agglomeration within the flame flow. Although similar structures were occasionally obtained with C2H2 mixtures, the co-presence of CO and C2H2 produced structures that were considerably longer and an order of magnitude more abundant than those obtained with C2H2 as the only growth reagent. Graphitic structure was nearly absent, as shown in Figure 10. Their relative yield depended on the ratio of CO to C2H2. Higher C2H2 concentrations led to shorter nanofibers and a lower overall abundance with a greater number of Ni particles becoming encapsulated with amorphous carbon. Only a few SWNTs were observed at high CO concentrations, but with far lower abundance than observed using Fe and the CO/H2 gas mixture as noted previously. We obtained an optimum production of HNGFs using a CO/C2H2 ratio of one and a combined flow rate of 1 slm of these two gases for our flame conditions. Most studies of nanotube synthesis either use a reactive mixture that includes CO at temperatures in excess of 700 °C or else use an unsaturated hydrocarbon at temperatures below 750 °C to minimize the extent of pyrolysis. Indeed, CO is considered inactive toward nanotube growth within furnaces at temperatures below 700 °C.53 Though Ni may not undergo restructuring upon CO absorption, CO could still alter the dissociative decomposition pathways (or probability) of other

species through electronic interaction effects mediated by the metal particle.54 Similar effects are well-known for H2 enhancing dissociative adsorption of CO on metal surfaces.36,39 Alternatively, CO can promote HNGF growth by physically preventing the blockage of surface sites critical to dissociative adsorption.55 Instead of enhancing particle reactivity, adsorbed CO could play a protective role for the nascent Ni nanoparticles. Indeed, C2H2/H2 mixtures generally led to deactivation of the Ni nanoparticles by formation of an amorphous carbon layer resulting from adsorption and partial dehydrogenation of PAHs and/or C2H2 on the particles.56,57 Adsorbed CO could prevent the formation of this layer. Thus the particles survive to achieve catalytic activity towards nanofiber synthesis. In a reciprocal fashion, the absence of any SWNTs (using the binary gas mixtures) in the presence of C2H2 indicates an inhibitory effect of C2H2 towards SWNT growth, as discussed previously. Conclusions While our results do support studies that show that the catalyst particle composition/identity controls nanotube structure and growth rates, we also find that the reactant gas composition plays a deciding role. Based on results presented here, CO promotes single-walled nanotube (SWNT) formation with Fe nanoparticles while C2H2 favors multiwalled nanotubes/nanofibers (MWNTs) with Ni nanoparticles. Separation of the catalyst formation step from the nanotube growth step allows investigation of catalyst particle size dependencies upon nanotube growth unlike methods that employ in-situ generation and concurrent growth. In particular, sizedependent reactivities of Fe and Ni nanoparticles are clearly shown. For Fe, this size-dependent reactivity may derive from a combination of adsorbate- plus thermally induced restructuring and adsorbate-particle steric effects affecting particle structure and hence reactivity. Thus small nanoparticles (∼1 nm) are catalytically active. Fe nanoparticles, attached to SWNTs are then usefully deactivated towards MWNT/nanofiber catalysis. The net effect leads to a relatively high yield of SWNTs compared to MWNTs with Fe in CO/H2/He gas mixtures. In contrast, the activity of Ni nanoparticles is not enhanced by these factors and they remain relatively inactive toward CO/ H2/He gas mixtures. Small (