Effects of Wet-Pressing and Cross-Linking on the Tensile Properties of Carbon Nanotube Fibers

To increase the strength of carbon nanotube (CNT) fibers (CNTFs), the mean size of voids between bundles of CNTs was reduced by wet-pressing, and the CNTs were cross-linked. Separate and simultaneous physical (roller pressing) and chemical methods (cross-linking) were tested to confirm each method’s effects on the CNTF strength. By reducing the fraction of pores, roller pressing decreased the cross-sectional area from 160 μm2 to 66 μm2 and increased the average load-at-break from 2.83 ± 0.25 cN to 4.41 ± 0.16 cN. Simultaneous injection of crosslinker and roller pressing augmented the cross-linking effect by increasing the infiltration of the crosslinker solution into the CNTF, so the specific strength increased from 0.40 ± 0.05 N/tex to 0.67 ± 0.04 N/tex. To increase the strength by cross-linking, it was necessary that the size of the pores inside the CNTF were reduced, and the infiltration of the solution was increased. These results suggest that combined physical and chemical treatment is effective to increase the strength of CNTFs.


Introduction
Carbon nanotube fibers (CNTFs) can be used as 1-dimensional reinforcements of composite materials and could be the next generation of carbon fibers (CFs), which are already used as commercialized reinforcements [1]. CNTFs have several advantages over CF. First, CNTFs have extraordinarily strong (~100 GPa, [2]) building blocks (CNTs). Second, CNTFs are cheaper to synthesize because the synthesis temperature and time of direct spun CNTF (~1000 • C, s, [3]) are lower and shorter than those of CF (>1500 • C, h, [4]); and CNTFs do not require an oxidative stabilization step. Lastly, CNTF is more flexible than CF judging by the high knot efficiency of 100% than that of CF [5,6]. Therefore, the development of strong CNTF would replace the use of CF to CNTF for composite materials.
The covalent bonds between CNTs (cross-linked CNTs) strengthened CNTFs, in addition to the vdW force. In practice, the maximization of the strength of CNTs requires a combination of both the physical removal of pores and chemical cross-linking. However, the removal of voids impedes subsequent infiltration of crosslinker chemicals into the CNT bundle, and the chemical reaction impedes the subsequent physical removal of voids because fibers become brittle after cross-linking.
Here, we propose a strategy to increase the strength of CNTFs by removing pores and cross-linking CNTs simultaneously. CNTFs were roller pressed while the crosslinker solution was injected (wet-pressing) and the Friedel-Craft acylation reaction was used to cross-link the CNTs (Figure 1a-c). Roller pressure and capillary pressure effectively reduced the porosity and increased the infiltration of the crosslinker solution into CNTFs. This approach enabled the identification of the effects of each treatment (i.e., roller pressing, capillary pressure, cross-linking) and increased the strength of CNTFs.

Materials
CNTFs were synthesized by the chemical vapor reaction [29] and were composed of singlewalled carbon nanotubes (SWCNTs). The average diameter of SWCNT was 1.2 nm (n = 30) ( Figure  S1a,b). Thermogravimetric analysis (TGA) results ( Figure S1c) showed that CNTFs contained an amorphous carbon impurity and absorbed water of 20.6 wt% and a catalytic impurity of 21.2 wt%. The IG/ID ratio was 7.9 ± 0.3 (n = 4) judging by the Raman spectra ( Figure S1d). The tubes were compressed using a custom-made roller machine (Figures 1a and S2) composed of nip rollers (diameter = 9.2 cm) and a take-up roller (diameter = 4 cm). The straight distance between the center of the nip rollers and the surface of the take-up roller was 0.3 m. Rollers were made of Teflon, and the nip rollers were wrapped in Teflon tape (width: 10 cm) to reduce damage to the CNTFs while pressing.

Materials
CNTFs were synthesized by the chemical vapor reaction [29] and were composed of single-walled carbon nanotubes (SWCNTs). The average diameter of SWCNT was 1.2 nm (n = 30) ( Figure S1a,b). Thermogravimetric analysis (TGA) results ( Figure S1c) showed that CNTFs contained an amorphous carbon impurity and absorbed water of 20.6 wt% and a catalytic impurity of 21.2 wt%. The I G /I D ratio was 7.9 ± 0.3 (n = 4) judging by the Raman spectra ( Figure S1d). The tubes were compressed using a custom-made roller machine (Figure 1a and Figure S2) composed of nip rollers (diameter = 9.2 cm) and a take-up roller (diameter = 4 cm). The straight distance between the center of the nip rollers and the surface of the take-up roller was 0.3 m. Rollers were made of Teflon, and the nip rollers were wrapped in Teflon tape (width: 10 cm) to reduce damage to the CNTFs while pressing.

Methods
Pristine CNTFs were compared with CNTFs that had been treated in one of four ways (Table 1).
(1) Simple cross-linking (CL): CNTFs were merely immersed in crosslinker solution (azelaic acid dichloride (AAD), 98%, Aldrich) for 10 min and were taken out, then heated in an Ar atmosphere for 1 h at 100 • C (CL100), 130 • C (CL130), or 160 • C (CL160). After the reaction, samples for Fourier transform infrared (FT-IR) and x-ray photoelectron spectroscopy (XPS) were washed several times with acetone during filtration and dried in an oven at 60 • C. (2) Pressed fibers: CNTFs were passed between nip rollers under an applied load of 510 kPa. The velocities of the nip rollers were~2 m/min and the take-up roller always rotated~3 mm/min faster than the nip rollers. (3) Acetone Pressed: CNTFs were passed through the nip rollers while acetone (99.5%, Sigma-Aldrich) was dropped at 0.6 mL/min on the top of them. The acetone-containing sample was then dried at room temperature for 1 h. (4) AAD pressed and CL: CNTFs were passed through the nip rollers while AAD crosslinker was dropped on them at 0.6 mL/min, then the fibers were heated in an Ar atmosphere for 1 h at 160 • C to induce the covalent cross-linking between CNTs.

Characterization
The morphology of CNTFs was determined using field emission scanning electron microscopy (FE-SEM, Philips, XL30 FEG, Amsterdam, The Netherlands). The cross-section was observed using a focused ion beam (FIB, Helios-Pegasus, FEI company, Hillsboro, OR, USA), SEM, and transmission electron microscopy (TEM, JEM-2200FS, JEOL, Tokyo, Japan) [18]. The length or the cross-sectional area A (µm 2 ) in the SEM images were measured using the ImageJ program (LOCI, University of Wisconsin, Madison, WI, USA). The contact angle of the liquid and the CNT film were measured using SmartDrop (SmartDrop Lab, Seongnam, Gyeonggi, Korea). CNT film (width: 9 cm, height: 4 cm) for the measurement of the contact angle was obtained by repeatedly collecting the CNTFs on a bobbin made of Teflon, then pressing them using the nip rollers while acetone was injected. The results of the chemical reactions were confirmed using FT-IR (Varian, 670/620, Varian, Palo Alto, CA, USA), and XPS (ESCALAB 250, VG Scientific, East Grinstead, Sussex, UK). TGA (STA7300, Hitachi, Tokyo, Japan) was operated at air (heating rate: 5 • C/min). Raman spectra (Horiba Jobin-Yvon LabRam Aramis spectrometer, Edison, NJ, USA) were recorded at the 514 nm line of the Ar-ion laser as the excitation source. The load-at-break L B (cN) of CNTFs was measured using a FAVIMAT single-fiber tester (FAVIMAT-AIROBOT2, Textechno, Mönchengladbach, NRW, Germany) at an elongation rate of 2 mm·min −1 (the number of measurements per sample n = 6-8) and a gauge length of 20 mm. The linear density (tex) of CNTFs were measured by the vibroscopic method using the FAVIMAT, and the specific strength SS (N/tex) was calculated [29]. Tensile strength TS (GPa) was calculated as L B /A.

Physical Effects of Roller Pressing and Liquid Injection
We observed the cross-section of CNTFs and investigated the physical effects of roller pressing and the injected liquid (Figure 2a-h). Pristine CNTF produced by the direct spinning method had irregular cross-sectional shapes and had macroscopic voids, including one >10 µm (Figure 2a,b, white arrows); they occur when CNTFs are synthesized using direct spinning. CNT agglomeration is produced only at a certain radius of the reactor and hence forms CNT sock having a doughnut-shape in cross-section [30]. The voids reduce the load transfer and L B of the fibers; moreover, A increases as the total volume of the pores increases, so TS decreases. Therefore, direct spun CNTFs require post-treatment to reduce porosity and increase strength. We observed the cross-section of CNTFs and investigated the physical effects of roller pressing and the injected liquid (Figure 2a-h). Pristine CNTF produced by the direct spinning method had irregular cross-sectional shapes and had macroscopic voids, including one >10 μm (Figure 2a,b, white arrows); they occur when CNTFs are synthesized using direct spinning. CNT agglomeration is produced only at a certain radius of the reactor and hence forms CNT sock having a doughnut-shape in cross-section [30]. The voids reduce the load transfer and LB of the fibers; moreover, A increases as the total volume of the pores increases, so TS decreases. Therefore, direct spun CNTFs require posttreatment to reduce porosity and increase strength. Physical compression reduced db from hundreds of nanometers to tens of nanometers. The db of CNTFs (schematic: Figure 3a; TEM images: Figure 3b) were measured (n = 1000) ( Figure 3d). In pristine CNTFs, the distribution of db was wide (range, 20 nm to 160 nm). In the CL160 samples, the most frequent range of db was 20-40 nm, but the high end of db (of CL160) still reached 140 nm; i.e., the porosity was lower than in the pristine samples, but the pore distribution remained wide. On the other hand, pressed fibers showed a narrow db distribution, mostly <40 nm. Moreover, both the median (179 to 65 nm) and mean values (206 to 71 nm) decreased to about one third after pressing. When the solution was dropped on the samples during pressing, the overall distribution of db increased slightly compared with the case where the solution was not used, probably due to the surface tension of the liquid caused aforementioned rearrangement, but most db were <40 nm.  (Figure 3d). In pristine CNTFs, the distribution of d b was wide (range, 20 nm to 160 nm). In the CL160 samples, the most frequent range of d b was 20-40 nm, but the high end of d b (of CL160) still reached 140 nm; i.e., the porosity was lower than in the pristine samples, but the pore distribution remained wide. On the other hand, pressed fibers showed a narrow d b distribution, mostly <40 nm. Moreover, both the median (179 to 65 nm) and mean values (206 to 71 nm) decreased to about one third after pressing. When the solution was dropped on the samples during pressing, the overall distribution of d b increased slightly compared with the case where the solution was not used, probably due to the surface tension of the liquid caused aforementioned rearrangement, but most d b were <40 nm. The second purpose of roller pressing was to increase the infiltration of the liquid (acetone or AAD). In this experiment, the liquid (AAD) had two functions: as a capillary-pressure inducer and as a crosslinker. CNTFs can be strengthened by capillary densification [12,14] or cross-linking methods [15][16][17][18][19][20][21][22][23]. However, the liquid cannot smoothly infiltrate the CNTF [18] (Figure 1b top), because CNTs are generally insoluble and difficult to disperse [31]. The process of roller pressing while injecting the liquid was designed to increase the infiltration of the solution. Pore reduction and the infiltration of the liquid were facilitated by the external force of the nip rollers (Figure 1b bottom); these changes would increase the effects of AAD, i.e., the CNT bundle aggregation by capillary pressure and cross-linking.
If the volume and surface area of the pores in CNTFs are constant, the driving force of capillary pressure is γLV·cosθ, where γLV (mN·m −1 ) is the surface tension of the liquid on the liquid/vapor surface, θ (°) is the contact angle of the liquid and CNTFs [18]. Here, we measured the contact angle of the CNT films, which provided easier measuring environments than CNTFs. The measured θ of acetone (9°) and AAD (27.3°) were much lower than water (106.7°) (Table 2, Figure 4a-e). γLV·cosθ of AAD (32.8 mN·m −1 ) was slightly higher than acetone (22.6 mN·m −1 ), but both were much higher than that of water (−20.7 mN·m −1 ). Therefore, acetone and AAD are better than water to aggregate CNT bundles by capillary pressure. The second purpose of roller pressing was to increase the infiltration of the liquid (acetone or AAD). In this experiment, the liquid (AAD) had two functions: as a capillary-pressure inducer and as a crosslinker. CNTFs can be strengthened by capillary densification [12,14] or cross-linking methods [15][16][17][18][19][20][21][22][23]. However, the liquid cannot smoothly infiltrate the CNTF [18] (Figure 1b top), because CNTs are generally insoluble and difficult to disperse [31]. The process of roller pressing while injecting the liquid was designed to increase the infiltration of the solution. Pore reduction and the infiltration of the liquid were facilitated by the external force of the nip rollers (Figure 1b bottom); these changes would increase the effects of AAD, i.e., the CNT bundle aggregation by capillary pressure and cross-linking.
If the volume and surface area of the pores in CNTFs are constant, the driving force of capillary pressure is γ LV ·cosθ, where γ LV (mN·m −1 ) is the surface tension of the liquid on the liquid/vapor surface, θ ( • ) is the contact angle of the liquid and CNTFs [18]. Here, we measured the contact angle of the CNT films, which provided easier measuring environments than CNTFs. The measured θ of acetone (9 • ) and AAD (27.3 • ) were much lower than water (106.7 • ) (Table 2, Figure 4a-e). γ LV ·cosθ of AAD (32.8 mN·m −1 ) was slightly higher than acetone (22.6 mN·m −1 ), but both were much higher than that of water (−20.7 mN·m −1 ). Therefore, acetone and AAD are better than water to aggregate CNT bundles by capillary pressure.     Diameters D (nm) of bundles (Figures 2a-h and 3a,c) were measured (n = 200) and their distributions (Figure 3e) showed the degree of bundle aggregation by capillary pressure. Pristine CNTFs usually had 20 ≤ D ≤ 80 nm (average = 61 nm); after roller pressing, most bundles had 40 ≤ D ≤ 80 nm (average = 64 nm). This result shows that simple roller pressing did not increase the bundle aggregation significantly, although the process decreased the void sizes to tens of nanometers.
In acetone-pressed and AAD pressed and CL specimens, most of the bundles had D > 60 nm (averagẽ 94 nm) (Figure 3e). This increase compared to roller-pressed CNTFs may be a result of capillary pressure and roller pressing. Capillary pressure is a microscopic phenomenon that occurs between CNTs or between CNT bundles and is considered to be effective for bundle aggregation [14,18], whereas roller pressing is a macroscopic phenomenon that effectively reduces the distances between bundles.
Simple immersion of the CNTF in liquid could not effectively induce capillary pressure. In CL160 specimens, which were simply dipped in AAD and cross-linked, D was almost identical to that of pristine CNTFs, i.e., bundles in CL160 did not aggregate, as bundles in acetone pressed or AAD pressed and CL did. This failure may be due to the insufficient infiltration of AAD molecules into the fibers. This result confirms that the concurrent roller pressing and dropping of liquid improved its infiltration effectively.
To summarize, external pressure by roller pressing effectively reduced the voids by several micrometers in size; and internal pressure by the capillary effect caused the aggregation of bundles. Therefore, simultaneous roller pressing and crosslinker liquid infiltration could (i) reduce the void fraction, (ii) aggregate bundles, and (iii) increase the infiltration of the crosslinker.

Chemical Cross-Linking
The vdW force (2-10 kJ/mol) acts between CNTs, but is relatively weak compared to covalent bonds (150-500 kJ/mol) [19]. Therefore, a method to introduce covalent bonds between the CNTs was attempted using the 1-step Friedel-Craft reaction (Figure 1c) with AAD. Among the chemicals having acyl-chloride functional groups at both ends of the chain, AAD was chosen for the following criteria: whether it is liquid at room temperature (for easier handling than vapor and better infiltration than solid), the length of the chain, and the price of the chemicals. The reaction was confirmed by FT-IR spectroscopy and XPS.
The cross-linking reaction was successfully performed at temperatures ≥130 • C. In the pristine sample and CL100 samples, the FT-IR and XP spectra showed no peaks. In the CL130 and CL160 samples, a ketone peak (C=O) occurred at 1710 cm −1 and the C-Cl peak was barely observed (Figure 5a); this result means that the C-Cl bond broke and a ketone formed on the CNT surface. The peaks with very low intensities at 721 cm −1 (C-Cl) would be caused by the one-end reacted crosslinker. Raman spectra results showed that the I G /I D ratio decreased by about 10% from 7.9 ± 0.3 (pristine) to 7.1 ± 0.5 (CL160) (n = 4) (Figure 5b). The decreased intensity of the G-peak at 1588 cm −1 represents the increased defects after the chemical reaction. The increased defects on the surface of CNTs provides indirect evidence that the chemical reaction has occurred. The XPS C1s spectrum results also confirmed the success of the reaction at 160 • C (Figure 5c,d). In the pristine sample, the ratio of the sp 2 carbons, which represent C = C bonds in CNTs, was dominant (Figure 5c). After the chemical reaction, the ratio of sp 2 carbon decreased and that of sp 3 carbon increased (Figure 5d). Additionally, a ketone peak at 289.1 eV appeared; which was almost absent in the pristine sample. Lastly, TGA curves also showed the indirect evidence of the successful chemical reaction judging by the decreased thermal stability of CNTs ( Figure S3).
Raman spectra results showed that the IG/ID ratio decreased by about 10% from 7.9 ± 0.3 (pristine) to 7.1 ± 0.5 (CL160) (n = 4) (Figure 5b). The decreased intensity of the G-peak at 1588 cm −1 represents the increased defects after the chemical reaction. The increased defects on the surface of CNTs provides indirect evidence that the chemical reaction has occurred. The XPS C1s spectrum results also confirmed the success of the reaction at 160 °C (Figure 5c,d). In the pristine sample, the ratio of the sp 2 carbons, which represent C = C bonds in CNTs, was dominant (Figure 5c). After the chemical reaction, the ratio of sp 2 carbon decreased and that of sp 3 carbon increased (Figure 5d). Additionally, a ketone peak at 289.1 eV appeared; which was almost absent in the pristine sample. Lastly, TGA curves also showed the indirect evidence of the successful chemical reaction judging by the decreased thermal stability of CNTs ( Figure S3).

Tensile Properties
Pressing increased L B of CNTF from 2.83 ± 0.25 cN to 4.41 ± 0.16 cN (Table 3). This change was a result of the elimination of the micrometer-size pores and the reduction in d b . However, roller compression does not always increase L B. The effect of roller pressing on L B was affected by the number of pressings, and on the material used to perform them. For example, when the fibers were repetitively pressed, L B was instead reduced ( Figure S4). Additionally, the roller pressing of CNT film between two metal plates decreases L B but increases TS by greatly decreasing A [32]. In the present study, the increase of SS was small (from 0.40 ± 0.05 N/tex to 0.47 ± 0.01 N/tex after pressing), but the increase in L b is meaningful (Figure 6a, Table 3). Pressing increased LB of CNTF from 2.83 ± 0.25 cN to 4.41 ± 0.16 cN (Table 3). This change was a result of the elimination of the micrometer-size pores and the reduction in db. However, roller compression does not always increase LB. The effect of roller pressing on LB was affected by the number of pressings, and on the material used to perform them. For example, when the fibers were repetitively pressed, LB was instead reduced ( Figure S4). Additionally, the roller pressing of CNT film between two metal plates decreases LB but increases TS by greatly decreasing A [32]. In the present study, the increase of SS was small (from 0.40 ± 0.05 N/tex to 0.47 ± 0.01 N/tex after pressing), but the increase in Lb is meaningful (Figure 6a, Table 3).    The combination of physical compression and chemical cross-linking increased the strength of the CNTF. In the Pressed sample, the distribution of SS narrowed ( Figure 6a); i.e., the structure and strength of the CNTF became uniform along its length. Uniformity is important in the fiber because it is directly related to its strength [33]. Therefore, roller pressing has the advantage that it imparts uniformity. The Acetone pressed sample had an SS = 0.49 ± 0.03 N/tex, which was similar to that of the pressed sample (Figure 6a), but the AAD pressed and CL sample had SS = 0.67 N/tex, which was 67% higher than the SS of the pristine sample, 43% higher than the SS of the Pressed CNTFs, and 37% higher than the SS of Acetone pressed. The strength of the CL160 sample, which was cross-linked only without pressing, did not increase compared to that of pristine CNTFs (Figure 6a). This failure of strengthening may be due to the lack of infiltration of the AAD in the CL160 sample. Therefore, the reduction of the void fraction and increase in the crosslinker molecules' infiltration are both important factors in the cross-linking method to increase strength.
The stress-strain curve shows the effect of physical and chemical treatments on strain-at-break S B (%) and toughness (Figure 6b). S B increased from 3.4 ± 0.5% to 4.3 ± 0.2% after pressing by rollers, then decreased to 3.5 ± 0.4% in the AAD pressed and CL sample. Toughness showed a similar trend: it increased from 0.74 ± 0.05 N/tex to 1.43 ± 0.09 N/tex after pressing and became 1.36 ± 0.22 N/tex after AAD pressing and cross-linking. The ductility and the energy absorbed by CNTF before fracture, i.e., toughness, increased after roller pressing; this change may be due to the reduced void fraction and increased shear force. Due to van der Waals forces, CNTFs break as bundles slip and result in the tapered broken end [21]. Therefore, increased van der Waals forces would increase a slip range, which increases the strain-at-break. Generally, after the cross-linking reaction, CNTFs become strong but also become brittle [21,22,25,34,35]. However, in this study, physical properties such as S B or the toughness of the AAD pressed and CL sample were similar or improved compared to the pristine CNTF because chemical treatments were conducted after the structural change by roller pressing.
The maximum tangent modulus was 0.26 N/tex in the pristine CNTFs, 0.32 N/tex in the pressed samples, and 0.37 N/tex in the AAD pressed and CL samples. (Figure 7a-c). The slight increase in the maximum tangent modulus after roller pressing may be a result of the increase in shear force as the pore size decreased (Figure 7a,b). In the pristine and pressed samples, the modulus remained low (<0.1 N/tex) at a strain of >1.5%, whereas in the AAD pressed and CL samples, it was >0.1 N/tex until the strain reached 3%, then it tended to decrease linearly; this result means that the resistance of the CNTF to deformation increased after chemical cross-linking, i.e., a covalent bond formed (Figure 7c). The shape of the modulus-strain curve changes depending on the internal structure of CNTFs [36]; i.e., when the individual CNTs in the CNTF do not act independently but act en masse, the modulus-strain curve changes from a sigmoid decay after the peak at low strain (Figure 7a,b) to a linear decay after the peak (Figure 7c). This change supports the hypothesis that roller pressing reduced the void fraction in the CNTF and that a reaction cross-linked the CNTs.
CNTF to deformation increased after chemical cross-linking, i.e., a covalent bond formed (Figure 7c). The shape of the modulus-strain curve changes depending on the internal structure of CNTFs [36]; i.e., when the individual CNTs in the CNTF do not act independently but act en masse, the modulusstrain curve changes from a sigmoid decay after the peak at low strain (Figure 7a,b) to a linear decay after the peak (Figure 7c). This change supports the hypothesis that roller pressing reduced the void fraction in the CNTF and that a reaction cross-linked the CNTs.

Conclusions
The strength of a CNTF can be increased by introducing covalent bonds among CNTs in the CNTF. The prerequisite for this approach, however, is the deep infiltration of crosslinker molecules into the voids in the CNTF. In this paper, we demonstrated that wet-pressing using two rollers is an effective method for this purpose. Deep infiltration was achieved because the average size of the voids is reduced by the physical pressing, resulting in a higher capillary force acting on the narrower gaps. The chance of cross-linking also increases due to the reduction in distance between CNTs and/or CNT bundles. Therefore, wet-pressing and cross-linking, which is the combination of void reduction by the physical method and increase in covalent bonds by the chemical method, effectively increases the strength of CNTFs.

Conclusions
The strength of a CNTF can be increased by introducing covalent bonds among CNTs in the CNTF. The prerequisite for this approach, however, is the deep infiltration of crosslinker molecules into the voids in the CNTF. In this paper, we demonstrated that wet-pressing using two rollers is an effective method for this purpose. Deep infiltration was achieved because the average size of the voids is reduced by the physical pressing, resulting in a higher capillary force acting on the narrower gaps. The chance of cross-linking also increases due to the reduction in distance between CNTs and/or CNT bundles. Therefore, wet-pressing and cross-linking, which is the combination of void reduction by the physical method and increase in covalent bonds by the chemical method, effectively increases the strength of CNTFs.