Abstract
Multiwalled carbon nanotubes (MWCNTs) were synthesized by CVD on industrially manufactured highly crystalline vapor-grown carbon fibers (VGCFs). Two catalyst metals (Ni and Fe) and carbon precursor gases ( and CO) were studied. The catalysts were deposited on the fibers by sputtering and experiments carried out in two different reactors. Samples were characterized by electron microscopy (SEM and TEM). Iron was completely inactive as catalyst with both and CO for reasons discussed in the paper. The combination of Ni and was very active for secondary CNT synthesis, without any pretreatment of the fibers. The optimal temperature for CNT synthesis was , with total gas flow of 650 of , , and Ar in 1.0:6.7:30 ratio.
1. Introduction
Carbon nanotubes (CNTs) have in recent years stimulated intensive scientific research due to their exceptional physical and chemical properties [1, 2]. A few different techniques, such as carbon arc-discharge [3, 4], laser ablation [5, 6], and substrate supported and aerosol chemical vapor deposition (CVD) [7–13], have been successful in producing lab-scale quantities of material.
CNT components have been used in a large range of applications: in light-emitting diodes [14], transistors [15], filters [16], field emitters [17], and photovoltaic devices [18]. Most of the proposed applications utilize the unique high strength, small size, large aspect ratio and surface area, and the exceptional chirality-determined electrical properties of nanotubes. In particular, the advantages gained by incorporating CNTs in fuel cells [19–22] and supercapacitors [23, 24] rely on the tubes’ characteristic high surface area, combined with their good electrical conductivity.
Vapor-grown carbon fibers (VGCFs) are a type of carbon fiber produced by the CVD method, similar to nanotube synthesis [25]. A specific manufacturing process [26] commercialized by Showa Denko yields industrial-scale amounts of high quality material. Many promising applications for VGCFs have been implemented in, for example, fuel cells [27], supercapacitors [28], and polymer composites [29].
Materials combining structures of different-size scales can be called hierarchical. The special case considered in this study is a combination of structures on two different-size scales: long, large diameter primary (or “first generation”) nanofibers, and shorter, smaller diameter secondary (or “second generation”) carbon nanotubes, attached to the primary fibers. Using the nomenclature of Teo and Sun [30], this can be characterized as the hierarchical “1D on 1D” structure CNT VGCF. The secondary tubes could, theoretically, be either multiwalled or single-walled.
In addition to work published about CNT growth on graphite [31], carbon cloth or paper [32–34], and MWCNTs [35, 36], carbon nanotubes have previously been synthesized on various carbon fibers [33, 37–42].
Most of the above methods either require complicated procedures for preparing suitable catalyst particles or extensive pretreatment of the fibers, or produce low yields and poor-quality tubes. We provide a simple method for synthesizing CNTs on untreated industrially manufactured highly crystalline Showa Denko VGCFs, and further study the role of catalyst particles and precursor gases. Two catalyst metals—nickel and iron—were studied, with two carbon precursor gases, namely carbon monoxide (CO) and acetylene , used in two different reactors. Iron was found to be totally inactive for secondary tube growth, while nickel proved very active at suitable conditions.
2. Results and Discussion
2.1. Acetylene Experiments
The effect of various experimental parameters was extensively studied in the horizontal reactor used (Section 3.2), and a successful combination was found. Practically, no secondary nanotubes were produced at C, confirming the findings of Duesberg et al. [43]. Of the three higher temperatures studied (700, 750 and, C), the best results were clearly obtained at C. As for reduction time, the shortest tested period of 2.5 minutes seemed to be too short to form good catalyst particles, as activation (defined as a visual estimation of the density and length of produced secondary tubes in SEM) was poor. A reduction time of 5 minutes—which has been successfully utilized for iron [44] in a different setup—worked best with nickel. Based on 24 catalyst particles measured from TEM images of the 5-minute samples, the average diameter of nickel particles after synthesis on the Showa Denko fibers was nm, while on the grid surface their average diameter was only nm. After 7.5 minutes of reduction even though activation was in some places good, the produced secondary tubes were very short. Moreover, many particle sites were inactive and covered with amorphous carbon. The effect of gas composition was found to be less important within the range we studied, and the optimum flow rates were determined to be 30, 200, and for acetylene, hydrogen, and argon, respectively.
In order to investigate the timeline of CNT growth, we carried out experiments for different synthesis times. The samples were prepared in the same manner as before (Section 3.1), but acetylene and hydrogen flows were stopped at the allotted time instead of the typical 10 minutes, and the reactor temperature was lowered. SEM observations showed that after 2.5 minutes of synthesis, growth has just barely begun, with very little activation on the Showa Denko fibers. After 5 minutes, activation was already significant, but the lengths of the synthesized secondary tubes were short. Activation after 15 minutes was significantly higher than activation after 10 minutes, suggesting that the growth was still proceeding at some rate after 10 minutes. The difference between 15- and 20-minute samples was not clear by visual observation. This behavior is in agreement with the findings of Krestinin et al. for methane pyrolysis [38].
In the optimal synthesis conditions, the VGCFs were well covered with secondary multiwalled tubes (Figure 1). Some secondary tubes had catalyst particles embedded at their ends, as found by TEM observation (Figure 2(a)). Most of them had between 10–20 walls and reasonably good wall structure (Figure 2(b)). A statistical survey of 20 MWCNTs observed in TEM found the average outer diameter to be nm, while the inner was nm.
(a)
(b)
(a)
(b)
Switching to iron catalyst and using the same gas mixture that had proven successful with nickel, synthesis was attempted at temperatures of 700, 750, and C. The catalyst particles on the substrate activated CNT growth effectively, with activation increasing with temperature. However, activation of the catalyst particles on the VGCFs was nonexistent. The good activation on the substrate itself proves that reactor conditions were perfectly suitable for growing tubes with acetylene from the iron catalysts. Furthermore, at the synthesis temperature of C, carbon nanotube forests were grown from iron catalyst on the substrate. At many places on the substrate, forests of several hundred micrometers in height were growing directly adjacent to the catalyst-containing fibers on which the catalysts were totally inactive (Figure 3).
(a)
(b)
2.2. Carbon Monoxide Experiments
Experiments with carbon monoxide using the vertical reactor (Section 3.3) with iron catalyst were unsuccessful in producing hierarchical structures. The iron catalyst particles were again completely inactive on the fibers. We carried out statistical measurements from TEM images of the samples prepared with 5-minute reduction and 30-minute growth time at C. We measured 72 inactive iron particles: 40 on the Showa Denko fibers and 32 on the grid surface. The average diameter of iron particles on the fibers was nm, while on the grid surface it was nm.
2.3. Discussion
Synthesis of hierarchical nanostructures using iron catalyst was consistently unsuccessful with both carbon monoxide and acetylene. Variation of the experimental conditions together with the prolongation of synthesis time up to 2 hours did not result in any significant improvement. However, as noted above, reactor conditions were obviously appropriate for the CNT growth at least in the case of acetylene and iron, since CNT forests were successfully synthesized on the silicon wafer in the close vicinity of inactive fibers (Figure 3).
Let us discuss possible reasons for the catalytic inactivity of iron for the formation of secondary tubes on the surface of untreated Showa Denko VGCFs, and compare them with successful experiments with nickel and literature data. It is worth noting that in spite of the similarities between Fe and Ni, some of their properties are significantly different. Of main interest here, the solubility of carbon in bulk iron is much higher than that in nickel [45, 46]. Size effects alter the properties of Fe [47, 48] and Ni [49] nanoparticles significantly from those in the bulk. However, even for small nanoparticles, crucial differences between the two metals are preserved, with Amara et al. [49] explicitly underlining the importance of limited carbon solubility in Ni. Also, with iron, the higher solubility can result in the rapid formation of iron carbide, which is—conversely to nickel carbide—quite stable at the experimental conditions [50]. The formation of iron carbide has been proposed to poison the catalyst [51, 52], inhibiting CNT growth [9, 53]. This view has been called into question, however, by recent in situ TEM observations of nanotube growth from C nanoparticles [54].
Another reason for different products can be attributed to the different diffusion of catalyst particles on the VGCFs surfaces. Iron particles coalesced during the reduction step and CNT growth initiation period, producing excessively large particles for CNT growth. As mentioned above, the average diameter of both Fe and Ni catalyst particles was much larger on the Showa Denko fiber surfaces than on the TEM grid surface after synthesis (13.5 and 9.5 nm on the fibers and 6.4 and 5.5 nm on the grids for Fe and Ni, resp.). However, it is further evident that the Fe particles on the fibers were significantly larger than the Ni particles. Theoretical considerations [55] would expect particles of the sizes observed to be solid, especially at C used in the acetylene experiments.
However, as noted in [55], observing the size of catalyst particles either before or after synthesis will not give a realistic picture of their true state during synthesis, as it is known that they are not static during reduction and nucleation. In situ measurements would be needed to observe the evolution of the catalyst particle sizes during reduction and synthesis, and possible implications to the state of the particles and their inactivity. It is noteworthy that shortening the reduction period in either the Fe/ or the Fe/CO experiments did not enable the production secondary tubes.
Importantly, data found in the literature indicates that pretreatment of the supporting carbon structure is vital when using iron [38, 56], as illustrated by the works listed in Table 1. The table displays a collection of relevant parameters from experiments successful in the synthesis of secondary nanotubes from iron catalyst on carbon fibers [33, 37–42]. Several successful special CVD methods are also included.
As an example, Zhu et al. [41] have demonstrated that there exists a limited temperature range where CNTs can grow from iron particles on noncrystalline carbon fibers, using low pressure methane gas as carbon precursor. At low temperatures, only a carbon layer is formed on the fiber surface. At higher temperature, the diffusion rate of iron particles into carbon fibers is enhanced. Small particles diffuse into the fibers before the carbon precursor is introduced, while larger ones nucleate only carbon clusters.
The case above and most of the others listed in Table 1 differ from our experiments either by a radically modified CVD procedure or by the lower crystallinity of the supporting fiber—a property which is altered by pretreatment. A more crystalline support could affect both proper carbon diffusion and the diffusion of catalyst particles on the fibers. Interestingly, Li et al. [31] have found that when growing CNTs from a stainless steel film was deposited on a graphite foil surface, phase segregation happens during the formation of catalyst particles, and only Fe/Ni alloy particles rich in Ni are responsible for the CNTs growth.
3. Experimental
3.1. Sample Preparation
For our supported CVD synthesis of hierarchical carbon nanostructures, industrially manufactured Showa Denko VGCFs were used as the primary level of the hierarchical structure. The tubes were delivered as a powder of very crystalline and pure nanofibers with an average diameter of around 150 nm. The fibers were dispersed in (1,2)-dichloroethane (DCE) in a low-power ultrasonic bath at C, with concentrations ranging from 0.5 to .
For acetylene experiments, silicon wafers were used to support the VGCFs. The sizes of the substrates cut from the wafers were about 0.5 . The pieces were cut from either a (111) /Si wafer with a thermally grown 260 nm layer, or from a (100) Si wafer with a native oxide layer. Droplets of DCE solution of VGCF material were deposited on the substrate surfaces by drop-drying. After synthesis, all samples were observed by field emission scanning electron microscope (Leo Gemini DSM982). Selected samples were also observed by field emission transmission electron microscope (Philips CM200 FEG) after transferring the fabricated material onto TEM grids (SPI, 100 Ni /SiO). This was done by wetting the silicon substrate surfaces with DCE and wiping them with TEM grids. For carbon monoxide experiments, the samples were grown directly on TEM grids, which were inserted into a vertical reactor using a specially designed holder.
Sputtering was used to deposit the catalyst particles needed for the CVD synthesis. For the experiments, the sputtering targets used were nickel and iron, while iron was used for synthesis with CO. A magnetron sputter coater (AGAR Auto Sputter Coater 108A) was used to produce nanoparticles on the carbon fibers deposited on the substrates. A sputtering current of 20 mA maintained for 15 seconds was used for both targets, as this has been found to be optimal for CNT production with the reactors used [57].
3.2. Acetylene Reactor
The synthesis of carbon nanotubes with was carried out in a horizontal reactor. It consisted of a 40 cm long quartz tube with an inner diameter of 1.2 cm, set inside a ceramic tube with an inner diameter of 3.0 cm. The tube was further placed inside a 34 cm long furnace, which was heated to a high temperature needed for the synthesis. Temperature calibration was done with a digital thermometer (Fluke 52 K/J) attached to a K-type metal wire thermocouple. The temperature inside the reactor at growth conditions was measured to be constant for a distance of several centimeters around the sample position.
For the synthesis, the samples were put into a ceramic boat, one end of the tube opened, and the boat pushed to the high temperature zone of the reactor with a steel rod. The reactor was then sealed. Before the introduction of the carbon source gas, deposited Ni or Fe catalysts were reduced at the selected synthesis temperature in a hydrogen-containing atmosphere, typically for 5 minutes. After reduction, acetylene was introduced and kept at a constant flow for the desired period of time (typically 10 minutes). The flows were controlled using a mass flow controller (Aalborg GFC17) or critical orifices calibrated with a flow calibrator (Gilian Gilibrator-2).
After the synthesis period, acetylene and hydrogen flows were stopped and the reactor temperature was lowered. Argon flow rate was increased to 700 to hasten the cooling. After a cooling period of about 1.5 hours, at which point the sample temperatures were well below C, the ends of the quartz tube were opened and the boat pushed out of the quartz tube with the steel rod.
The main parameters varied in the acetylene experiments were synthesis temperature (650–C), reduction time (2.5–7.5 minutes), acetylene flow rate (10–50 , corresponding to concentrations of 1.6–13.0% depending on the gas mix), and synthesis time (2.5–20 minutes). Furthermore, in some experiments the flow rates of (0, 130, and 200 ) and Ar (0, 270, 420, and 700 ) were varied to study the possible effects of mixing and gas composition.
3.3. Carbon Monoxide Reactor
The synthesis of CNTs with CO was attempted in a vertical laminar flow reactor. It consisted of a ceramic tube with an inner diameter of 2.2 cm inside a 44 cm long furnace. A stainless steel rod of 6 mm in diameter with a TEM grid holder at one end was used for inserting the samples. The reactor design and procedure for the CO experiments were identical to previous work [57].
Briefly, a temperature of C had been found to be optimal for iron-catalyzed growth on wafers, with a reduction time of 5 minutes with and Ar flows (200 and 400 ) passed into the reactor. Ar was then replaced with CO (400 ), while the flow remained constant. A small amount of (8 ) was introduced into the reactor for the whole duration of the experiment, as this had been found to have a critical effect on nanotube growth [58]. In this study, typical synthesis times were 30 minutes, and the samples were characterized by SEM and TEM.
4. Conclusion
Two different carbon precursor gases, carbon monoxide and acetylene, were studied for the synthesis of secondary carbon nanotubes on the surface of highly crystalline Showa Denko vapor-grown carbon fibers. Two different catalyst metals were examined, namely nickel and iron with acetylene, and iron with carbon monoxide. Of these, the combination of nickel and acetylene proved to be the only working solution.
Using the presented simple and straightforward nickel-catalyzed synthesis on untreated VGCF surfaces, the achieved optimal secondary tube yield was high, with good activation and coverage on the carbon fibers. With view of applications, either replacing the Showa Denko fibers with multiwalled carbon nanotubes grown directly to a conducting substrate to act as the primary structure, or replacing the VGCFs used in current applications by our material would be logical next steps. Potential applications for this nanomaterial would include catalyst supports in fuel cells, powerful supercapasitors, and possibly gas sensors. One further interesting application could be the use of hierarchical nanostructures as reinforcement material. The secondary tubes could act as “anchors” to prevent the slipping of MWCNT or fiber composites, thus improving the mechanical properties significantly.
Acknowledgments
The authors thank Mr. Prasantha R. Mudimela for his help during sputtering condition investigations. The authors acknowledge the Academy of Finland for financial support (Grant no. 118445).