Growth of bridging carbon nanofibers in cracks formed by heat-treating iron oxide thin sheets in acetylene gas Open Access

Many researchers have studied the growth mechanism of carbon nanotubes(CNTs). The growth of CNTs by catalytic chemical vapor deposition (CCVD) is explained by the nanoparticle catalyst model.^1–9CNT growth on a metal nanoparticle has been observed directly with motion video through environmental transmission electron microscopy(ETEM).^10,11 The metal nanoparticles become soft and fluctuate when CNTs grow on them, and the CNTs swing and rotate as they grow. Researchers have attempted to inhibit this fluctuation to obtain long, straight CNTs.^9,12–14 However, it is difficult to inhibit the fluctuation completely because the growing edge of the CNT is untethered and thus fluctuates easily on a nanoparticle catalyst.

Recently, many researchers have studied carbon nanosheets (CNSs) composed of graphene layers.^15–18 CNS films have been synthesized on template substrates such as copper films using chemical vapor deposition (CVD).^17,18 However, the quality of CNSs is influenced by the surface structure on the substrate. Therefore, it is desirable to grow CNSs without using a template.

Many groups have reported the reduction of iron oxide by a carbon source gas during steel production.^19–23 However, little research has been reported on excess carburizing of the reduced iron in the carbon source gas because the samples break into small pieces. For example, when iron oxide was reduced and carburized under heat treatment for 2 hours at 750 °C in acetylene gas, the sample broke into small pieces with carbonnanostructures.²⁴ 

In this work, we successfully prevent samples from breaking into small pieces during carburization by using pure iron with a high purity of 5N as the starting material and optimizing heat treatment conditions. Heat treatment swelled and deformed samples into large curved shapes, and we observed many cracks in them. Some of the crackscontained carbon nanofibers(CNFs). The CNFs were connected to the walls of the cracks, bridging the gaps between them. In this paper, we study the formation and structure of these bridging-growth CNFs (BG-CNFs) produced without using metal nanoparticles or template substrates. This is the first example of CNFs formed in the cracks of heat-treated iron samples.

High-purity electrolytic iron foil (purity of 5N, Mairon-UHP, Toho Zinc Co., Ltd., Tokyo, Japan) with an area of 10 mm² and thickness of 50–80 μm was used as the starting material. An iron sample was placed at the center bottom of a quartz glass tube with an inner diameter of 26 mm in an electric tubular furnace. Heat treatment was performed in two steps: oxidation and carburization (including reduction). First, the iron foil was oxidized by heat treatment for 1 min at 850 °C in air. Oxidation doubled the thickness of the iron foil from 50 μm to about 100 μm. We analyzed the oxidizediron samples by X-ray diffraction(XRD) and identified them as iron oxide composed of hematite (Fe₂O₃) and magnetite (Fe₃O₄), as shown in Fig. 1(a).

Next, we heated the ironoxide samples to 850 °C at a rate of 10 °C/min and maintained this temperature for 10 min under nitrogen (N₂) gas with a flow rate of 700 mL/min. The gas flow was changed to a mixture of 5% acetylene (C₂H₂) in N₂ with a flow rate of 740 mL/min. Samples A1 and A2 were heat treated for 2 and 10 min, respectively, at 850 °C under acetylene gas flow (Fig. 1(b) and 1(c)). N₂ gas flow was then used to decrease the temperature in the furnace. Samples broke into small pieces when the treatment time for both oxidation and carburization was increased to 1 hour. Samples were characterized by scanning electron microscopy (SEM, S3400N, Hitachi, Tokyo, Japan), transmission electron microscopy(TEM, JEM-4000EX, JEOL,Tokyo, Japan), XRD and Raman spectroscopy (NRS-3100, JASCO, Tokyo, Japan). Samples for TEM were prepared by dividing the sample at a crack with BG-CNFs by bonding it to a TEM grid and then removing part of the heat-treated sample. Raman spectra were acquired using a laser excitation of 532 nm with an incident power of 0.4 mW and spot size of 1 μm.

Samples heat-treated for more than 7 min under acetylene gas flow were significantly swollen and deformed into curved shapes. The surfaces of heat-treated curved samples were covered with cracks, as shown in Fig. 2(a). Figure 2(b) shows sample A2, which has many CNFs that bridge the walls of cracks. BG-CNFs grew almost parallel to each other and perpendicular to the crack walls. The BG-CNFs in the rectangular area in Fig. 2(b) are shown in higher resolution in Fig. 2(c). Both ends of the BG-CNFs were connected to the walls of a crack. The BG-CNFs had various shapes and their widths ranged from μm to nm. Such BG-CNFs appeared reproducibly when iron samples were heat-treated for more than 7 min in acetylene gas flow.

We observed the structure of the BG-CNFs by TEM, and confirmed that the BG-CNFs were composed of carbon through energy dispersive X-ray spectroscopy (EDX) and electron diffraction (ED) analysis. TEM images revealed that the BG-CNFs had two shapes, tubular (hereafter abbreviated as ‘BG-CNT’, which stands for bridging-growth carbon nanotube) and sheets(Figs. 3 and 4, respectively). The BG-CNT shown in Fig. 3(a) has a length of over 15 μm. This image was obtained by combining two images of the same BG-CNT. The root of the BG-CNT was connected to the wall of a crack, and the top end was free and open. The ED pattern obtained for the top end (Fig. 3(c)) indicates that the BG-CNT is composed of graphene layers. The open edge had a wavy,fluctuating cross section. We observed iron nanofilaments inside the middle of the BG-CNT (Fig. 3(a)). The BG-CNT was seamless over the iron nanofilaments, indicating that the iron nanofilaments formed inside the BG-CNT as it grew.²⁵Carbonnanoparticles adhered to the outside of the BG-CNT. It seems that the carbonnanoparticles were generated by decomposition of acetylene gas on the surface of the iron substrate.

TEM also revealed that CNSs with a length of over 6 μm formed. The CNS shown in Fig. 4was over 1 μm wide, had flat and scrolled structures and was connected perpendicular to the substrate. The top end was broken with a horizontal straight line, as shown in Fig. 4(b). ED analysis on the flat area of the top end of the BG-CNS indicated it was composed of graphene layers, as shown in Fig. 4(c).

We analyzed the structure of the BG-CNFs by Raman spectroscopy.^26–29 In the Raman spectra of carbon nanomaterials, the G- and D-bands that are related to graphite structure and defects, respectively, can be used to evaluate crystal phase and defect concentration. Figure 5(a) shows an SEM image at a crack with BG-CNFs in sample A2, which was heat-treated for 10 min in acetylene gas flow. The BG-CNFs had carbonnanoparticles on their surface, as shown in Fig. 5(b). Figure 5(c) shows a Raman spectrum of a BG-CNF in the crack. The D/G ratio was about 0.2. The amorphous, defect-rich structure is attributed to carbonnanoparticles generated by decomposition of acetylene gas. The BG-CNF is composed of multiple layers of graphene,^28,29 which is consistent with the ED results presented above.

Figure 6 shows SEM images of sample A2. We observed the BG-CNFs at a crack with a sector form at the lower left side in Fig. 6(a). BG-CNFs formed along the trajectory of the opening motion of each crack, as shown in Fig. 6(b). It seems that BG-CNFs were drawn out from the crack walls at different places and with different speeds according to the motion of the crack. A number of BG-CNFs were broken around their middle because of the tension caused by movement of the crack walls. This shows that the BG-CNFs are connected to the crack walls more strongly than within their body.

The effect of heat treatment time on CNF formation was investigated. After heat treatment for 2 min in acetylene gas flow, no BG-CNFs were observed in the cracks of the substrate. The structure of sample A1 was analyzed by XRD and identified as a mixture of wüstite(Fe_0.925O), iron(α-Fe), cementite (Fe₃C) and graphite (C) (Fig. 1(b)). When heat treatment lasted for 10 min, the wüstite phase disappeared from the XRD spectrum, as shown in Fig. 1(c). It is known that porous wüstite is generally obtained following reduction of ironoxide to steel.^19–21,24

Figure 7 shows cross-sectional scanning ion microscopy (SIM)images of sample A1 embedded in resin. Reduction and carburization progressed from the top to the bottom of the sample because it was placed at the bottom of the quartz glass tube of the electric furnace. Reduced iron and carbon were observed near the top of the sample and iron oxide(mainly Fe₃O₄) was observed near the bottom. The reduced iron phase was almost continuous, but contained numerous holes and cracks.Carbon precipitated inside the small holes and cracksnear the top of the sample, as shown in Fig. 7(b).

A cross-sectional SIM image of sample A2 following focused ion beam (FIB) milling is depicted with a tilt angle of 45° in Fig. 8. The reduced iron was broken into small pieces and the interspaces were filled by carbon. When heat treatment was increased from 2 to 10 min in acetylene gas flow, the reduced iron broke into smaller pieces and the volume of the carbon precipitate increased. This indicates that the breakup of reduced iron into smaller pieces is caused by precipitation of carbon from inside the iron. Generally, carburizing heat treatment of pure iron just produces cementite (Fe₃C) such as pearlitic steel,as shown in the phase diagram of Fe-C.³⁰Therefore, precipitation of carbon from inside the iron causes the excess carburization of iron oxide. We assume that carbon precipitation proceeds as follows. When iron oxide is reduced by acetylene gas, the reduced iron contains many transient vacancies and defects after losing oxygen atoms through topochemical reactions.^24,31 At that time, excess carbon can penetrate into the vacancies and defects in the reduced iron.Carbon then precipitates from inside the iron,causing it to break as time progresses.

It is difficult to explain the growth mechanism of a BG-CNF using the usual growth model for CCVD on a metal nanoparticle. In particular, it is a challenge to understand the growth of BG-CNS structures with widths of over 1 μm because the contact with the catalyst would need to maintain a wide, thin structure after division of the substrate. We propose that the growth of BG-CNFs proceeds as shown in Fig. 9. After heat treatment in acetylene gas reduces the iron oxide(Fe₂O₃ and Fe₃O₄) sheet to porous wüstite(Fe_1-xO) and then to porous reduced iron, BG-CNFs are formed by division at the instant of carbon precipitation from inside the reduced iron that contains a large amount of carbon as a solid solution, as indicated above in SIM images. SEM images demonstrating the growth process of BG-CNFs are shown in Fig. S1 in the Supplementary Materials.³² 

The shape of a BG-CNF depends on the contact structure with the divided walls of the crack and the tension induced by the opening motion. It seems that a BG-CNT forms when the divided edge has a cap structure. The BG-CNTs and iron nanofilaments with a variable cross section (Fig. 3(a)) may be formed by fluctuation of the cap structure during growth caused by changes in tension. A BG-CNS forms when the edge contacting the substrate is a line rather than a point. The sheet shape is maintained by tension. The formation mechanism of BG-CNFs is associated with the precipitation of carbon in the reduced iron and dividing of the substrate by tension. In situ observation by a technique such as ETEM should provide further insight into the growth mechanism of BG-CNFs.

We discovered that BG-CNFs grow in the free space generated by cracks formed in substrates under tension during carburization. If control of the quality, cross-section shape and length of BG-CNFs can be achieved by arranging the structure of the contact point with the substrate during heat treatment, a new technique to produce carbon wires will be realized. The BG-CNFs are arranged between the divided substrates on the macro scale, so they can easily be moved arbitrarily by holding both sides of the substrates. It is expected that the growth of CNSs can be terminated using a template substrate such as copper sheets and transcription to polymer films.¹⁸ 

BG-CNFs have the advantages that both ends are fixed and growth can be controlled by tension. This technique may lead to long CNFs, considering that CNTs with a length of 20 cm have been fabricated by ‘Kite-mechanism’(Refs. 9 and 14)through tension induced by gas flow. In addition, We have proposed the independent control of catalyst functions including absorption of a carbon source and excretion of CNTs at the catalyst structure as the ‘Carbon transmission method’ in our previous paper.²⁵ Further development of BG-CNF technology may provide a new way to arrange long CNFs for various wire applications.

We fabricated BG-CNFs by oxidizing high-purity iron foil and then heat-treating it in acetylene gas flow. Cracks formed in the samples with BG-CNFs connected to the walls of the cracks. BG-CNFs were drawn out from the walls of the cracks as they opened. This is a new way to grow and arrange CNFs using bulk iron sheets, and does not use metal nanoparticles or template substrates. Optimizing this process may provide a new way to produce organized carbon nanostructures for various applications.

Dr. Hitoshi Kawanowa and Dr. Ken-ichi Yoshida at Ion Technology Center Co., Ltd. are thanked for their technical support with TEM and Raman observation and analysis.