CNF- Carbon Nano Fibre

Carbon Nanofibers and Their Composites: A Review of Synthesizing, Properties and Applications

by Arun P R

Abstract: Carbon nanofiber (CNF), as one of the most important members of carbon fibers, has been investigated in both fundamental scientific research and practical applications. CNF composites are able to be applied as promising materials in many fields, such as electrical devices, electrode materials for batteries and supercapacitors and as sensors. In these applications, the electrical conductivity is always the first priority need to be considered. In fact, the electrical property of CNF composites largely counts on the dispersion and percolation status of CNFs in matrix materials. In this review, the electrical transport phenomenon of CNF composites is systematically summarized based on percolation theory. The effects of the aspect ratio, percolation backbone structure and fractal characteristics of CNFs and the non-universality of the percolation critical exponents on the electrical properties are systematically reviewed. Apart from the electrical property, the thermal conductivity and mechanical properties of CNF composites are briefly reviewed, as well. In addition, the preparation methods of CNFs, including catalytic chemical vapor deposition growth and electrospinning, and the preparation methods of CNF composites, including the melt mixing and solution process, are briefly introduced. Finally, their applications as sensors and electrode materials are described in this review article.

1.       Introduction
Since the first carbon fiber (CF), which was prepared by carbonizing cotton and bamboo, was used as the filament of a light bulb in 1879 by Thomas Edison, it has been developed tremendously in both fundamental scientific research and practical applications [1–5]. As one of the most important members of CFs, carbon nanofibers (CNFs) have been applied as promising materials in many fields, such as energy conversion and storage, reinforcement of composites and self-sensing devices [5–10]. There are some differences between conventional carbon fibers (CCFs) and CNF. The first one, also the most obvious one, is their size. The conventional CF has diameters of several micrometers, while CNFs have diameters of 50–200 nm. Figure 1 gives a schematic illustration of the difference between CNF and conventional CF. Except the diameter; the structures of the CNFs are evidently different from traditional carbon fibers. The typical CCFs were prepared from high-strength polyacrylonitrile (PAN) or meso-phase pitch (MP), varying the preparing conditions, including the oxidation atmosphere, the raw materials chosen and the heat treatment temperatures. The different preparing conditions will result in different properties of the prepared conventional CF. However, unlike the CCF, the CNF can be mainly prepared by two approaches: catalytically vapor deposition growth and electrospinning.

2.       Synthesizing of CNFs

This section will discuss the preparation methods of the CNF. Currently, the CNF can be prepared mainly by two methods. One is catalytic thermal chemical vapor deposition growth, and the other one is electrospinning followed by heat treatment.
Two types of CNF can be prepared by catalytic thermal chemical vapor deposition, namely, the cup-stacked CNF and the platelet CNF. The cup-stacked CNF, also called conical CNF, was first found by Ge and Sattler in 1994 [11]. Figure 2a–c gives the schematic demonstration of the formation of the cup-stacked CNF, and Figure 2d gives the schematic illustration of the platelet CNF structures.
For the preparation of CNF by the catalytic vapor deposition growth approach, several types of metal or alloys, which are able to dissolve carbon to form metal carbide, have been used as the catalyst, including iron, cobalt and nickel; chromium, and vanadium. Additionally, the molybdenum, methane, carbon monoxide, synthesis gas (H2/CO), ethyne or ethene are used to provide the carbon sources in the temperature range from 700 to 1200 K [12]. Generally, the structures of the CNF are governed by the shapes of the catalytic nanosized metal particles. The growth mechanism has been proven as the deposition of the hydrocarbons dissolved in the metal particle and precipitated on the metal surface as graphitic carbon [13]. Figures 3 and 4 demonstrate the schematic illustration of the typical growth mechanism of the cup-stacked and platelet CNFs [14,15]. Figure 5 demonstrates the high resolution transmission electron microscope (HRTEM) image of the cup-stacked CNF and the platelet CNF [14,15].
Electrospinning is another widely used method for the preparation of the CNFs. Inagaki [16] recently reviewed CNF prepared by electrospinning process systematically. In this review, the CNFs prepared via electrospinning and carbonization was summarized according to their structure and properties. Most recently, Zhang [17] summarized the preparation and applications of CNFs prepared by electrospinning.
To fabricate the CNFs by the electrospinning method, the polymer nanofibers are required to be prepared as the precursors of the CNFs. The properties of the final CNFs are decided by the types of polymer solution and the processing parameters. PAN and pitches are the most frequently used polymers. In addition, poly(vinyl alcohol) (PVA), polyimides (PIs), polybenzimidazole (PBI), poly(vinylidene fluoride) (PVDF), phenolic resin and lignin were also used [16]. Once the polymer nanofibers have been successfully prepared, a heat treatment will be applied to carbonize the polymer nanofibers to form CNFs. The morphology, purity, crystallinity, diameters and porosity are governed by the parameters of the heat treatment process, such as atmosphere and temperature. Figure 6 shows the schematic demonstration of the electrospinning device for CNF preparation.
After the polymer nanofibers have been successfully fabricated, the carbonization process will be followed by heating the polymer nanofiber up to 1000 °C in a specific environment. Generally, volume and weight change will occur during the carbonization process, which results in the decrease of the diameter of the CNFs. In most cases, the CNFs prepared by the electrospinning method are prone to form web or mat structures. This structure is a good form to be used as electrode materials for batteries or supercapacitors. Due to the overall performances of the batteries largely counting on the transport performances of the ions in the electrolyte, therefore, controlling the pore structure is the most import factor to enhance the performances of batteries.

3.       Preparation of CNF Composites

The overall performances of the CNF/polymer composites are largely governed by the dispersion of the CNF in the polymer matrix. Therefore, the dispersion technique plays a key role in the synthesizing of CNF composites. aterials 2014, 7 3924 The dispersion of CNF in polymer matrix can be realized mainly by two approaches: the melt mixing process and the sonication process in low viscosity solutions. The most widely used method is melt mixing, due to its low cost, simplicity and availability. Generally, extrusion or roll mill [18,19], Haake torque rheometer [20] and mini-max molder [21,22] all belong to the melt mixing method. In this method, to obtain a good dispersion condition for CNF in polymer matrix, a high shear mixing condition is usually required. Although the high shear mixing will lead to a relatively good dispersion of the CNF, the aspect ratio, which is another key parameter governing the overall performances of the CNF/polymer composites, will be decreased during the mixing process. It was found that the decrease of the aspect ratio will result in the degradation of some properties [23,24]. Therefore, the investigation of the relatively low shear mixing approach without sacrificing the dispersion is still a challenge for the preparation of CNF/polymer composites by the melt mixing approach. The chemical surface treatment of CNFs is a promising method to help their dispersion in the polymer matrix. In this process, the compatibility between the grafting functional groups and the polymer matrix is the key factor that decides the CNF dispersion and the overall performances of the composites. In most cases, the treatment process is oxidizing the CNF surface by soaking in sulfuric/nitric acid at various temperatures followed by acylation. After this process, the functional group will be grafted onto the surface of the CNF by the reaction between the oxidized CNF and the functional groups. Li [25] and co-workers prepared and characterized the surface-treated CNF by using diamines or triamines as linker molecules. The amine group acts as a bridge connecting the CNF and the –NH2 to form the CNF–C(O)–NH– structure. Kelarakis and co-workers [26] prepared the CNF/ethylene/propylene (EP) random copolymer composite with a similar process. The as-received CNFs were surface oxidized by sulfuric/nitric acid and then reduced by sodium borohydride in absolute ethanol to form the CNF–OH structure. After that, the hydroxylated CNF was dispersed in dichlorobenzene and a polypropylene-graft-maleic anhydride polymer to form the CNF–O– structure. Apart from the melt mixing processing for the CNF/thermoplastic polymers, the dispersion of the CNF in thermosetting polymers to prepare the CNF/thermosetting polymers (mostly epoxy resins) composites mainly counts on the solution approach with the help of sonication. In this process, the CNFs will be dispersed in the liquid epoxy form by sonication before being mixed with the hardener. Usually, acetone or other solutions are needed to help the effect of sonication. In addition, to avoid the increasing temperature during the sonication process, external cooling devices are necessary in most cases. Pervin and co-workers [27] fabricated nanocomposites composed of SC-15 epoxy and CNF. The mixing process was carried out through a high-intensity ultra-sonication of the CNF and SC-15 epoxy. Once the sonication was completed, the hardener was added in the mixture, followed by high-speed mechanical stirring, and then cured at room temperature. The preparation of the CNF/epoxy nanocomposite by Choi et al. [28] showed that ozone surface treatment of the CNF is helpful for its dispersion in the epoxy matrix. In this study, CNF with and without ozone surface treatment was to investigating the dispersion conditions. The CNFs were dispersed in acetone by sonication and stirring process at room temperature. The epoxy resin was added into the CNF-acetone solution without aterials 2014, 7 3925 stopping sonication and stirring. After this process, the acetone was removed by heating the mixture at 100 °C for 24 h, followed by the addition of the hardener, and then cured at room temperature.

4.       Properties of CNF Composites

 4.1. Electrical Conductivity

One of the most important properties of CNF composites is their electrical conductivity. When the CNF composites are applied as electrical devices, sensors, electromagnetic shielding or electrodes for batteries or supercapacitors, the electrical conductivity is always the first priority need to be considered. In CNF composites, when the electrical conductivity measured as a function of the filling content of CNF, a typical “S” shape curve appears, due to the critical percolation phenomenon [29]. It follows a power law behavior expressed as[30]: ; (1) where σreal is the real electrical conductivity of the CNF composites; σc is the electrical conductivity of the CNF; f is the filling content of the CNF; fc is the percolation threshold, which is defined as the critical value of the CNF filling content forming continuous network; and t and s are the electrical conductivity critical exponents above and below the percolation threshold, respectively. In this equation, σreal was experimentally measured, and the values of σc and fc are constants. According to the numerical and experimental results, it was claimed that, in the bond and site percolation lattices, the values of the conductivity critical exponents were universal, such that t ≈ 1.3–1.4, s ≈ 0.5 (in two dimensions) and t ≈ 1.6–2.0, s ≈ 0.6 (in three dimensions), based on the renormalization group theory [29,30], and in practical applications, they were usually considered belonging to the same universality system, as well. Some experimental and numerical results, however, have indicated that the practical application problems and the simulated lattice percolation problems do not belong to same universality systems. Although the percolation phenomenon has been studied for decades, the non-universality of the critical conductivity exponents observed experimentally has remained difficult to explain. The Kogut and Straley (KS) model is a milestone for quantitatively analysis of the critical conductivity exponents regarding the non-universality in a percolation system [31]. It first claimed that the universality of the conductivity exponents would be broken if the low-conductance bonds in percolation networks were characterized by an anomalous conductivity distribution. This model was derived from the mean field theory by assigning each neighboring pair in a regular lattice. It was claimed that, in a lattice percolation system, if a bond with finite conductivity, g, with probability μ and zero conductivity with probability 1 − μ, the bond conductivity distribution function can be written as: (2) where δ(g) is the Dirac delta function and h(g) is the distribution function of the finite bond conductivity. If h(g) has a power law divergence for small g of the form: (3) aterials 2014, 7 3926 where , then the universality of the conductivity critical exponents will be lost with sufficiently large values of exponent α. Based on the KS model, a well-known model, namely the “tunneling model”, was introduced by Balberg [32], corresponding to granular materials and carbon/polymer composites. Figure 7 shows the schematic demonstration of the tunneling model, which is the main conductivity mechanism of the carbon/polymer composites. As demonstrated in this figure, the black spheres represent the carbon materials, and the grey circles around the black spheres represent the polymers. In this model, the electrical conductivity of the carbon/polymer composites is governed by the width of the tunnel, which means the thickness of the polymer layer on the surface of the carbon materials.
In CNF composites, the tunneling effect is the main mechanism of electrical conduction; therefore, the electrical conductivity of the CNF/polymer composites are affected by the thickness of the polymer layer on the CNF surfaces, which is decided by the surface treatment methods and the polymer types. Although the tunneling model has been widely used in describing the non-universality of the percolation system; the conductive fillers were assumed as spherical particles; thus, it is insufficient to describe the percolation system with non-spherical fillers, such as CNF-filled polymer composites. In recent years, a comprehensive understanding of the mechanism of conduction in CNF/polymer composites was developed rapidly thanks to the development of random media physics [33–37]. The electrical conductivity was systematically investigated via theoretical models and the microstructure dependence of the carbon polymer composites. Lux [38], Kirkpatrik [39], Clerc [40] and Nan [29] reviewed the physics of percolation theory and the physics of inhomogeneous materials. Another important model to describe the percolation phenomenon in a conductor-insulator composite is the general effective media (GEM) model, which was described by McLachlan [41] in detail. In the GEM model, the composite was considered as a symmetric medium, in which the conductor with an ellipsoidal or spherical shape was embedded in the insulator matrix. Under this condition, the relationship of the conductivities between the composite, the insulator matrix and the conductive fillers can be written as: (4) where p is the conductor’s filling content; pc is the percolation threshold; and σh, σl, and σm are the electrical conductivities of the conductor, insulator and the composite, respectively
It has been widely accepted that, if the effective conductivities are given by the average conductance values, from the perspective of the geological structures, the conductivities can be divided into two types: the series and the parallel. In CNF composites, one extreme case is that all CNFs in the composite connected in parallel form, where the equivalent effective conductivity can be written as: (5) In this assumption, if all CNFs are geometrically identical, the σeff will be proportional to the second sum. Another extreme case is that all CNFs are connected in series form, where the equivalent effective conductivity should be written as: (6) Of course, in a real CNF composite, the resistance of the whole system is not able to be reflected only by parallel or series form; therefore, the real connection form or conduction channel has to be analyzed. In addition, there is a geometrical restriction in this case, that the CNFs must all be congruent, which means all the CNFs have to be of the same size and shape; otherwise, the geometrical factors have to be included in the analysis [31]. In a CNF composite, the continuous network of the CNF can be categorized into two types: the backbone and dangling ends [42,43],which show different properties. The percolation backbone is to demonstrate the real path that carries the current transport. In previous studies, it was found that the effective path or minimum length of the conduction was governed by the backbone based on widely simulated results in different lattice percolation systems [43]. Although these results are able to reflect the backbone characteristics to some extent [44], they were still hardly able to analyze the real conductor-insulator composite, due to the backbone structure being very hard to be directly observed experimentally. In a CNF/polymer composite, near the percolation threshold, not all CNFs belong to the continuous percolation backbone, because some of them still belong to the isolated clusters or form dangling ends. Therefore, the contribution to the composite properties of the percolation backbone density, which is defined as the proportion of the continuous percolation network in the whole percolation infinite clusters, is not the same as the isolated cluster. We presented a new model to describe a possible non-universal behavior in a conductor-insulator composite system [45]. In this model, the backbone and dangling end masses, MB and MD, were presented as the key parameters to describe the backbone structure. The backbone or dangling end density is defined as the portion of the total backbone or dangling ends that belong to the percolation infinite clusters, respectively. In a percolation system, the conductance between two randomly selected nodes (g) can be expressed as: (7) where g0 is a constant; MD and MB are dangling ends and backbone masses belonging to two randomly selected nodes; and κ is called the “structural factor” that represents the geometry and topology aterials 2014, 7 3928 structure of the conductor. According to this assumption, the value of κ was defined as the function of the aspect ratio of the fillers and expressed as: (8) where f is the filling content of the conductive fillers; i is the index of the filler types; a/b represents the aspect ratio of the filler; and λ is 1, 0 and −1 correspond to a > b, a = b and a < b, respectively. In this equation, the tunneling conductance parts should be considered as a part of the backbone. Our recent study demonstrates that the conductive distribution function of the percolation network, H (MD/MB), and the dangling ends and the backbone masses could be expressed with the exponential form [45]: (9) where η is the average value of , written as: (10) Therefore, the critical exponent, t, can be obtained as: (11) where tun is the “universal value” of a percolation system based on the effective medium theory [31]. In this equation, if there is no backbone in the system, the t value goes to infinity, and the electrical conductivity goes to zero. In this extreme case, the system is an insulator. If there are no dangling ends in the system, the t value goes to zero, which means that the electrical conductivity goes to σc. In this extreme case, the system is a pure conductor. According to this model, the effective electrical conductivity of the system increases with the decreasing of the conductivity critical exponent, which is a key factor to reflect the backbone density of the fillers. Based on this analysis, the backbone variation trend with increasing filling content of the CNF was analyzed [46]. Figure 8 gave the schematic illustration of the backbone structure variation mechanisms with the increasing content of the fillers with the long aspect ratio above the percolation threshold. This figure shows the original infinite cluster of the filler with long aspect ratio in a composite. As can be seen in the figure, the loop, ABCDE, belongs to the backbone, MB, and AAʹ, AAʹʹ, BBʹ, CCʹ, CCʹʹ, DDʹ, EEʹ and EEʹʹ belong to the dangling ends, MD, in the infinite cluster. As such, the ratio of the backbone and the dangling ends is: (12) Figure 8b–f categorized the backbone structure variations with increasing filling content based on Figure 8a. As shown in Figure 8b, the added part, xy, was located on the percolation loop and connected the previous dangling ends, BBʹ and EEʹ.

5.       Applications

 5.1. Sensors

Generally, the self-sensing function of the CNF composites is realized by testing the variation of electrical properties that has resulted from the change of the external conditions, including stress/strain and the gas environment. The electrical conductivity of the CNF composites is able to be reversibly changed by several orders of magnitude with the reversible change of the external conditions. Zhu [78] prepared CNF/elastomer (VM2) composites with a percolation threshold of 1 wt% as strain sensors for reflecting large mechanical deformation. The electrical conductivity of the composite is able to reversibly change 102 –103 orders of magnitude upon stretching to 120% strain and recovery to 40% strain. CNF/poly(acrylate) was prepared by Li [79] as gas sensors. In this study, the vapor can be detected via a five orders of magnitude change of the electrical conductivity. Other than vapor sensors, the CNF/(polypyrrole) PPy coaxial nanocable toxic gas sensor was fabricated to perceive irritant gas, such as aterials 2014, 7 3936 NH3 and HCl, via the one-step vapor deposition polymerization method [80]. The structure of the sensor is composed of an ultrathin and uniform PPy layer on the surface of CNF, shown in Figure 11. By the change of the oxidation level of the PPy layer resulting in the reaction between the PPy and the NH3 or HCl, the electrical conductivity of the composite decreased because of the decrement in the charge carrier density.
Apart from the strain/stress or gas sensing, the temperature, humidity, magnetic field and light are all important factors that are required to be sensed in many applications. The CNF and their composites for sensing these factors still need to be investigated deeply in future studies.

5.2. Batteries

Currently, the CNF composites as electrode materials for batteries and supercapacitors have been widely studied worldwide. The main requirement for high performance batteries and supercapacitors is high porous electrode materials, which are able to contain enough electrolytes and satisfy the fast and long-term ion transport. Ji [81] prepared a porous CNF by the carbonization of electrospun PAN/SiO2 composite nanofibers followed by resolving the SiO2 nanoparticles with hydrofluoric acid (HF). This porous CNF has magnified surface areas and defects. The specific porous structure guarantees that this CNF is able to be used as an anode material for lithium ion batteries directly, without adding any polymer binder or non-active carbon black. Figure 12 shows the schematic illustration of the porous CNFs. A facile way to synthesize N-dope porous CNF webs as anode materials for lithium ion batteries via using polypyrrole as a precursor was reported by Qie [82]. The high-level N-doping and the nanostructure of the CNF webs guaranteed a reversible capacity of 943 mAh/g with a current density of 2 A/g, even after 600 cycles. Figure 13 gives the SEM microstructure of the (1) PPy nanofiber webs and (2) CNF webs. A 3D CNF/graphene nanosheets hybrid material was prepared via the chemical vapor decomposition approach by Fan [83]. This is a special structure with the 1D CNF grown on the 2D graphene nanosheets that contains sufficient cavities, open tips and exposed edges of the graphene sheet. Figure 14 gives the schematic illustration of the structure with the 1D CNF grown on the 2D aterials 2014, 7 3937 graphene nanosheets. Due to this 1D–2D hybrid structure, the lithium ions could be stored in these spaces more efficiently, which guaranteed the high reversible capacity (667 mAh/g), high-rate performance and cycling stability. In addition, the direction of the CNF axis is vertical to the graphene nanosheets, which is able to control the diffusive orientation of the lithium ions.

6.       Conclusions and Future Perspectives

 Generally, two preparation approaches, namely catalytic chemical vapor deposition growth and electrospinning, are the mainly effective pathways to fabricate CNFs. In the catalytic chemical vapor deposition growth method, some metals and alloys, including Fe, Co, Ni, Cr and V, which can dissolve carbon to form metal carbides, were able to be chosen as the catalysts, and the molybdenum, methane, carbon monoxide, synthesis gas (H2/CO), ethyne or ethane are able to be used as carbon sources. Generally, the structures of the CNF are decided by the shapes of the catalytic nano-sized metal particles. In the electrospinning process, the polymer types and the carbonization process play the most important roles in the quality of the prepared CNFs. The overall properties of the CNF composites are largely governed by the dispersion condition of the CNFs in the matrix materials. To prepare the CNF composites with a good CNF dispersion, the melt mixing and solution process are the most widely used approaches. The melt mixing method, which was realized by high shear mixing, can effectively disperse the CNF in polymer matrix, while it is not able to guarantee the original aspect ratio and shapes of the CNFs. The solution process method, which is widely applied to disperse the CNF in a thermoset polymer matrix, is realized by sonication of CNFs in various solutions, followed by a curing process. In both of these methods, the chemical surface treatments of the CNFs are effective ways to realize their good dispersion in the matrix materials. The electrical property of the CNF composites largely counts on the dispersion and percolation condition of the CNFs in the matrix. The percolation theory and fractal method are the decisive tools to evaluate the percolation threshold, the percolation backbone structure and the percolation critical exponents, which are the key factors to enhance the electrical properties of the CNF composites. Near the percolation threshold, most of the CNFs are not able to form a continuous network, and the aterials 2014, 7 3940 dangling end parts are in the majority. The main electrical transport mechanism is the tunneling effect. Therefore, the surface treatment methods, the dispersion approaches and the polymer types are extremely important for the enhancement of the electrical properties. How to quantitatively determine the backbone structure of the CNFs in the matrix materials is still a challenging topic. The fractal analysis is an effective way of quantitatively characterizing the structure of the CNFs, and related models are waiting to be developed in future works to combine the overall performances and the microstructures of the CNF composites. In addition, the relationship between the thermal properties, the mechanical properties and the microstructures of the CNFs also need to be investigated more deeply, due to few studies demonstrating the effects of the CNF structures on the overall performance of the CNF composites. As for applications, the CNFs and their composites are able to be used in many fields, including sensors, electrode materials and electromagnetic shielding. The sensitivity of the CNFs and their composites mainly count on their electrical performances. Therefore, how to accurately and quantitatively reflect the real situation by the electrical performances is the most important issue. Other than the vapor and strain/stress sensing, the humidity and temperature sensing capability of the CNFs and their composites are also waiting for development in future studies. As electrode materials, special structural designs and realization to guarantee high specific areas without satisfying mechanical performances is the key factor to enhancing the performances of the current materials.