Graphitizing Non-graphitizable Carbons by Stress-induced Routes

Graphitic carbons’ unique attributes have attracted worldwide interest towards their development and application. Carbon pyrolysis is a widespread method for synthesizing carbon materials. However, our understanding of the factors that cause differences in graphitization of various pyrolyzed carbon precursors is inadequate. We demonstrate how electro-mechanical aspects of the synthesis process influence molecular alignment in a polymer precursor to enhance its graphitization. Electrohydrodynamic forces are applied via electrospinning to unwind and orient the molecular chains of a non-graphitizing carbon precursor, polyacrylonitrile. Subsequently, exerting mechanical stresses further enhances the molecular alignment of the polymer chains during the formative crosslinking phase. The stabilized polymer precursor is then pyrolyzed at 1000 °C and characterized to evaluate its graphitization. The final carbon exhibits a uniformly graphitized structure, abundant in edge planes, which translates into its electrochemical kinetics. The results highlight the significance of physical synthesis conditions in defining the structure and properties of pyrolytic carbons.

Herein, we demonstrate how the synergistic effect of CNT-induced electrohydrodynamic forces and mechanical compressive treatment can overcome the lack of fusion-induced rearrangement in organic precursors, thus significantly enhance their graphitization. In particular, we present a framework by which a traditionally known "non-graphitizing" polymer, Polyacrylonitrile (PAN), may be graphitized at low pyrolysis temperature, thereby challenging the current notion of graphitizing and non-graphitizing polymers. The choice of electrospinning process allows us to use electrohydrodynamic forces, enhanced by the addition of carbon nanotubes, to unwind and orient the polymer molecular chains. The polymer was thermally stabilized under calculated mechanical stress to preserve and enhance the alignment of the precursor's chains. A myriad of characterization methods, including transmission electron microscopy, Raman spectroscopy, ohmic resistance, and electrochemical measurements, were used to measure the degree of graphitization in pyrolyzed PAN specimens. The presented results demonstrate how the physical conditions and treatments applied during the synthesis affect molecular alignment in a polymer precursor to markedly enhance its graphitizability. By revealing unexplored stress-based mechanisms for carbon graphitization, this work suggests that graphitizability is not solely a chemical property of organic precursors.

Results
Unwinding PAN molecular chains. Electrohydrodynamic forces offer a powerful tool to manipulate the configuration of polymer molecules embedded in electrically charged liquids. A method of choice for applying electrohydrodynamic forces is electrospinning, where an electric filed is used to draw out polymer nanofibers from a reservoir into fabric mats 15 .
The underlying mechanism driving the electrospinning process is the interplay between molecular drag forces (F D ) in the electrospun polymer and the external electrostatic forces (F ES ), as illustrated in Fig. 1A. The electrostatic forces deform the meniscus of the polymer fluid protruding from the capillary, which forms a region called the Taylor cone 16,17 (Region 2 in Fig. 1). When the Taylor cone is formed, the cross-sectional area of the polymer jet reduces from that of the capillary tube to the significantly smaller cross section of the traveling jet. This change of cross sections brings about transition regions with drastically different flow velocity: a slow region within the polymer meniscus and a fast region within the traveling jet. The high velocity gradient zones developed at these interfaces result in sizable shear stress fields, which unwind the polymer chains in the jet flow 17 . Small-angle neutron scattering (SANS) and fluid mechanics studies have corroborated this confinement effect, and found a correlation between the gyration radius of the polymer, radius of the electrospun nanofiber, and the alignment of the polymer chains [18][19][20][21] .
The addition of MWCNTs into the polymer flow introduces dielectrophoresis phenomenon into the process. Here, the dielectrophoretic force (F DEP ) stems from the difference in the electrodynamic properties of the MWCNTs, polymer (PAN) and solvent medium (DMF) in the presence of a strong non-uniform electric field (Table in Fig. 1). The direction of the F DEP is determined by the dimensionless Clausius-Mossotti factor (k) which depends on the complex permittivity of the particle ε ε = − By substituting the permittivity values in Fig. 1, the k factor is determined to be positive for MWCNTs and negative for PAN molecular chains. Consequently, the F DEP will drag the CNTs against the jet flow, while it pushes the polymer chains along the jet's flow direction. The interaction of these opposing forces generates additional shear force zones at the surfaces of the MWCNT that further align the polymer chains.

Formation of Stabilized PAN fibers.
A critical stabilization phase is usually carried out prior to pyrolysis by heat treatment of the polyacrylonitrile under ambient air pressure (Fig. 2) [24][25][26][27] . In the course of stabilization, PAN molecular chains gain structural stability by cross-linking with the neighboring chains, thus forming conjugated ladder-like structures. The resulting structural stability enables PAN to withstand the extreme heat treatment of the subsequent pyrolysis step 24,28 .
During the stabilization of PAN, the polymer backbone chains tend to lose nitrogen groups in the cyclization process, resulting in sp 3 hybridized carbon bonds (Fig. 2a) 24,28 . This leads to the formation of penta-and hepta-carbon rings, which due to their curved structures will yield amorphous carbon or fullerenes during the pyrolysis. Thermodynamic stability of fullerene structures inhibits the polymer from graphitizing into carbon planes even at extremely high temperatures. In other words, the foundation of the polymer microstructures is cast during the preceding synthesis and stabilization phases. Once the curved sp 3 carbon structures form in the stabilization, the PAN chains are set up to form fullerenes. Accordingly, high-temperature pyrolysis cannot force the underlying carbon foundation to form planar sp 2 carbon planes, rendering the polymer "non-graphitizing".
Considering the impact of the stabilization phase on the orientation of the PAN molecular chains, one expects that the physical parameters of stabilization would influence the graphitization of the polymer precursors substantially. The gains of polymer chain alignment from electrospinning might well be undone by the thermal stresses induced during stabilization, an insight that seems to have been overlooked. Accordingly, during the stabilization process one should aim at providing conditions that preserve or further align the PAN precursor chains, which are already oriented by the shear fields in the electrospinning process. To implement this alignment process, one can apply mechanical forces to maintain/enhance the orientation of the PAN precursor chains during stabilization (Fig. 2b).
This approach simply boils down to the question of "how to mechanically suppress the formation of curved surfaces in the polymer microstructure during the stabilization process". It is thus anticipated that applying surface compression and/or mechanical stretching to a stabilizing polymer mat will translate into the alignment of precursor chains at a molecular-level.
Following the outlined hypothesis, we performed mechanical treatment of the polymer fibers by subjecting the PAN mats to approximately 200 kPa compressive stress while they are being stabilized at 275 °C in an oven. By performing the stabilization process under such mechanical treatment, the carbon chains are maintained partially aligned and kept from curving during this formative step. This in turn reduces the extent of fullerene formation and sets up the stabilized polymer to form more graphitic carbon planes. We also investigated the influence of applying tensile mechanical stress to the stabilizing polymers. Interestingly, changing the axial stress from compressive to tensile affects the microstructure of final carbon fibers. We discuss this effect in our upcoming report.
Subsequently, we pyrolyzed pure PAN, CNT infused PAN (PAN-CNT) and mechanically-treated PAN-CNT (M-PAN-CNT) samples at 1000 °C. The resulting carbons' microstructure and properties were then characterized to investigate the effects of polymer chain alignment on the degree of graphitization.

Raman Spectroscopy and Conductivity Characterization.
Over 100 Raman spectra were collected and averaged across each type of carbon fibers to characterize the microstructure of the resulting carbons. Lorentzian curves were then fitted to the D and G peaks of the representative Raman spectra for each type of synthesis ( Fig. 3a-c). The heights, shapes, positions, and areas of Raman peaks can provide valuable information about structural characteristics of the final carbons 29,30 . Typically, the G peak, centered at 1560-1600 cm −1 , is activated by vibration of carbon atoms in sp 2 -hybridized graphene planes. Thus, the height and sharpness of the G peak reveals the presence of crystalline graphitic carbons. The D peak at 1300-1400 cm −1 is the defect-induced band that indicates the degree of disorder in graphitic microstructures. Accordingly, the ratio of D band to G band peak heights, (I D )/(I G ), has been consistently used to probe the graphitic quality of carbon materials microstructure [29][30][31][32][33] .
A quick examination of Lorentzian fitted Raman spectra indicates that with the application of CNTs during electrospinning, I D decrease with respect to I G and both peaks exhibit approximately the same intensities. The implmentation of mechanical treatment reduces the relative intensity of disorder associated band by more than 30%. Accordingly, (I D )/(I G ) ratio decreases from 1.26 in pure PAN fibers to 0.69 for M-PAN-CNT fibers, revealing a remarkable increase in the graphitic structure of the resulting carbons (Fig. 3a-c).
The evolution in G and D peak areas, from PAN control samples to M-PAN-CNT fibers, supports the same upward trend of graphitization in the carbon fibers (Fig. 3d). Typically, broader Raman peaks with larger areas are correlated with a higher level of disorder and defects in the graphitic materials. The aquired Raman peak areas verify that mechanical treatment has markedly reduced the areas of D and G peaks, and thus resulted in lower level of disorder in the sp 2 network. Interstingly, the transistion from PAN-CNT to M-PAN-CNT exhibits larger is not due to the contribution of nanotubes, as mere addition of CNTs does not change the area of the G peak.
The electrical conductivities of the PAN-based carbons, measured by four-point probe method, confirms the increasing trend of graphitization, from pyrolyzed pure PAN, to PAN-CNT, to M-PAN-CNT, with conductivity values of ~200 S/m, 550 S/m, and 5,000 S/m, respectively (Fig. 3e). The correlation between the microstructures and bulk properties reveals the uniformity of graphitization across the PAN mats, and demonstrates that the enhanced graphitic microstructures translate into improved bulk characteristics in the carbon fabrics.

High-Resolution Transmission Electron Microscopy. High-Resolution Transmission Electron
Microscopy (HRTEM) allows for detailed study of carbon microstructure in the pyrolyzed PAN specimen. Visual examination of lattice fringes in TEM micrographs illustrates the evolution of carbon microstructures developed by the outlined processes. In Fig. 4a we show the HRTEM micrograph of the carbon fibers produced by pyrolysis of pure PAN at 1000 °C. Consistent with the agreed-upon notion that PAN is "non-graphitizing", this micrograph reveals that the structure of the pyrolyzed PAN is highly disordered and randomly curled. The Fast Fourier Transform (FFT) of the micrograph yields symmetric rings, further evidencing the amorphous nature of the carbon fiber.
The micrographs in Fig. 4b and c reveal the effects of electrohydrodynamic molecular alignment and mechanical treatment on the structure of the resulting pyrolyzed PAN. As can be seen in Fig. 4c, the structure of M-PAN-CNT specimen exhibits aligned carbon lattice fringes, which are stacked rather tightly together. Here, the carbon lamellae are generally well-oriented with slight undulations, demonstrating imperfect graphene planes. The fast Fourier transform of this image supports our observation by revealing directionally aligned carbon planes with a fringe separation-distance of 3.6 Å.
By comparison, the nanostructure of PAN-CNT appears to be in transition from an amorphous to a more graphitic material (Fig. 4b). Here, the carbon fringes are somewhat visible but more fragmented and randomly oriented. Moreover, there is a noticeable increase in the tortuosity of the carbon planes, compared to M-PAN-CNT. In some domains, the fiber nanostructure resembles that of pure pyrolyzed PAN of Fig. 4a. This is particularly meaningful considering that randomly oriented curves in atomic planes impede the development of tightly-packed parallel carbon planes; instead, they lead to the formation of fullerenic structures 4,5,34 . Such "wrinkles" are mechanically "flattened" during the stabilization processes of M-PAN-CNT (Fig. 4c). A higher level of tortuosity also increases the fringe separation, as manifested by the 3.8 Å interlayer spacing acquired from the FFT of the TEM image (Fig. 4b inset). Figure 4d, TEM micrograph of pure MWCNT, is included for comparison with the pyrolyzed PAN samples. Expectedly, the CNT exhibits well-aligned, tightly stacked graphitic lattice planes with a minimal to zero tortuosity and fringe separation of 3.4 Å.
Here, n is the number of electrons transferred, F is the Faraday's constant, v is the scan rate, R is the universal gas constant, and D O and D R are oxidation and reduction diffusion coefficients for [Fe(CN) 6 ] 3−/4− , with values of 7.26 × 10 −6 cm 2 /s and 6.67 × 10 −6 cm 2 /s, respectively 41 . Ψ is a dimensionless function of ∆E peak (distance between the cathodic and anodic current peaks) found for the conditions of a single step, single electron transfer process at the temperature (T) of 300 K. It is important to note that the observed linear dependency between the current density of voltammograms' peaks and the square root of the scan rate for all carbon electrodes (insets of Fig. 6) is a characteristic of standard bulk diffusion, for which Nicholson formula is applicable.
The mechanical treatment of PAN electrodes has significantly boosted their kinetics, increasing the k°a pp eleven-fold from 0.0029 cm/s in PAN to 0.0312 cm/s in M-PAN-CNT (Table 1). The enhanced kinetics is also evident from the lack of peak shift with increasing scan rate, indicative of quasi-reversible electrochemical reaction 38 . The improvement in electrochemical performance stems from the abundance of graphitic edge planes in the carbon fiber microstructure 35,36,39 . Indeed, Compton et al. were able to model the electrochemical performance of carbon electrodes solely based on the quantity of graphitic edge domains on their surface 35,42,43 . Consistent with these studies, the k o app of M-PAN-CNT electrodes demonstrates an order-of-magnitude enhancement over pure CNT-based electrodes 44,45 , and approaches that of graphene and reduced graphene oxide electrodes 46,47 .
All carbon fiber electrodes were tested in a 2 M KCl solution immediately after the electrochemical characterization in K 3 [Fe(CN) 6 ] batch (Fig. 6d). Compared to pure PAN and PAN-CNT electrodes, the peaks in M-PAN-CNT's voltammograms are more pronounced, pointing to stronger adsorption of [Fe(CN) 6 ] 3−/4− onto the surface of the mechanically treated fibers. This observation is consistent with previous studies, which reported that the adsorption of [Fe(CN) 6 ] 3−/4− onto carbon occur predominantly on functionalized graphitic edge sites 39,42,48 . It is noteworthy that the observed enhanced kinetics and adsorption capacity in M-PAN-CNT are achieved without any additional activation process. The inherent aptitude of M-PAN-CNT for functionalization is particularly attractive because it addresses a main challenge in utilizing graphitic carbons for various electrochemical devices 49-53 .

Discussion
In this work, we discussed how electro-mechanical parameters of synthesis influence molecular alignment in a polymer precursor to enhance its graphitization. We focused on carbonization of polyacrylonitrile as the carbon precursor, since it is generally considered a "non-graphitizing polymer", and is a main source for commercial carbon fibers. The application of electrospinning allowed us to utilize electrohydrodynamic forces, boosted by addition of carbon nanotubes, to align polymer molecular chains within the spun fibers. The polymer chain alignment was then preserved by mechanical compression of the polymer fabrics during the formative stabilization step. Subsequently, a myriad of material analysis techniques, including Raman spectroscopy, transmission electron microscopy, electrical conductivity, and electrochemical measurements allowed us to evaluate the graphitization level of the pyrolyzed PAN specimen. Pyrolyzing the curved polymer chains lead to formation of amorphous carbon or fullerenic structures, whereas mechanically aligned carbon precursors are poised to form more graphitic carbon planes. This result underlines the correlation between the physical conditions exerted on a  stabilizing polymer and its graphitization, thereby, emphasizing that the foundation of the final carbon structure is cast during the stabilization of the precursor polymer.
Graphitization of organic precursors is a complex phenomenon and it is unlikely to find a comprehensive solution to develop graphitic carbons from different polymer precursors. This work addresses the physical aspect of polymer graphitization and explores the influence of the synthesis parameters on the microstructure of the pyrolytic carbons. By investigating the effects of applied electro-mechanical stresses during the carbon synthesis process, the presented results reveal further insight into the underlying mechanism of graphitization and provides an attractive pathway for tailoring the microstructure of carbon materials.

Materials and Methods
Carbon Synthesis. Electrospinning. For electrospinning PAN fibers, polyacrylonitrile (Sigma Aldrich) with molecular weight M w = 150,000 g/mol was dissolved in DMF, N, N-dimethylformamide (Sigma Aldrich), at 40 °C for 24 hours to a final concentration of 8 wt.%. The solution for electrospinning CNT-infused PAN (PAN-CNT) nanofibers was prepared by dispersing MWCNT, multi-walled carbon nanotubes (Sigma Aldrich) with a diameter of 110-170 nm and a length of 5-9 µm, in DMF under ultra-sonication for 1 hour then stirred for 24 hours at 40 °C. Subsequently, PAN was dissolved into the DMF-CNT solution by stirring at 40 °C for 24 hours, to achieve a final concentration of 8 wt.% PAN and 1 wt.% MWCNT. Electrospinning for PAN and PAN-CNT solution was performed by loading the PAN or PAN-CNT solutions, respectively, into a 1 mL syringe mounted on a syringe pump, which provided a flow rate of 0.7 mL/hour. An operating voltage of 15 kV was applied between a 21-gauge needle (820 µm outer diameter) and an 8 × 8 cm 2 aluminum plate, which was held at 15 cm away from the dispensing needle. The electrospinning process was run for 1 hour to form a ~100 µm thick PAN and PAN-CNT nanofiber mats, respectively.
Stabilization. PAN and PAN-CNT mats were peeled off the aluminum plate and then thermally stabilized on top of an alumina plate at 275 °C in an oven for 5 hours. For mechanically-treated PAN-CNT (M-PAN-CNT) samples, a PAN-CNT mat was subjected to a 200 kPa compressive stress between two alumina plates, while being stabilized in an oven at 275 °C for 5 hours.
Pyrolysis. All stabilized PAN nanofibers were pyrolyzed in a tube furnace under continuous flow of Nitrogen at an approximate flow rate of 4,600 sccm. First, the stabilized samples were heated to 300 °C at a 4.5 °C/min ramp rate and maintained at that temperature for one hour. Next, the temperature was increased to 1000 °C at a 2.5 °C/ min ramp rate and held at that temperature for an hour before cooling down to ambient temperature at 10 °C/ min rate.

Characterization. The pyrolyzed polymer samples characterized by Raman spectroscopy with a Renishaw
InVia Raman Microscope, equipped with 532-nm excitation laser set, to assess the graphitization of the pyrolyzed PAN samples. Over 100 Raman spectra were acquired and averaged across each type of carbon fibers to evaluate the graphitic quality of the resulting carbons. The Raman spectra were then fitted using Lorentzian curves and analyzed. Optical confocal microscopy (Keyance, VK-X) and scanning electron microscopy (FEI, Quanta 3D FEG Dual Beam) were employed to observe the morphology of the carbon nanofibers. 4-point probe station (Signatone, S-302-4, with a SP4 probe head) was used to measure the conductivity of the pyrolyzed PAN samples.
High-resolution transmission electron microscopy (HRTEM; FEI, Titan 80-300 kV S/TEM) operated at 300 kV was used to visualize and evaluate the microstructure of the pyrolyzed PAN samples. TEM samples were prepared by drop-casting pyrolytic carbon sample, suspended in toluene, onto TEM grids (PELCO CM20 purchased from Ted Pella, Inc.). Fast Fourier Transform (FFT) was applied on the collocated TEM micrographs to further measure the crystallinity and the fringe separation distance of the resulting carbons.
Cyclic voltammetry (CV) was used to characterize the electrochemical behavior of the three distinct types of carbon fiber mats(CFM): pure PAN, PAN-CNT and M-PAN-CNT. CVs of the CFM were first run in a 1 mM potassium ferricyanide (Fischer Scientific) in 2 M potassium chloride (Sigma Aldrich) solution using a traditional 3-electrode configuration with a Ag/AgCl (BASi) reference electrode and a 1 cm 2 polished glassy carbon (Structure Probe, Inc) counter electrode. Potassium ferricyanide was chosen as the redox couple to characterize the CFM as it is particularly sensitive to the presence of graphitic edge planes, while remaining relatively insensitive to most other surface modifications 36,38 . To further distinguish the effects of graphitic edge planes, a subsequent CV is run in a 2 M KCl aqueous solution. The presence of redox peaks in this second CV affirms the presence of ferricyanide ion adsorption along the edge planes 35,36 . To characterize the electrochemical performance of the pyrolyzed PAN electrodes, the heterogeneous electron transfer rate, k o app , was calculated using the Nicholson method 40 ,  (3), n is the number of electrons transferred, F is the Faraday's constant, v is the scan rate, R is the universal gas constant, and D O and D R are oxidation and reduction diffusion coefficients for Ferricyanide redox couple, with values of 7.26 × 10 −6 cm 2 /s and 6.67 × 10 −6 cm 2 /s, respectively 41 . Ψ is a dimensionless function of ∆E peak (distance between the cathodic and anodic current peaks) found for the conditions of a single step, single electron transfer process at the temperature (T) of 300 K. k o app is determined by evaluating the ∆E peak at a scan rate of 100 mV/s of each pyrolyzed PAN nanofiber mats.