Nanotube Dispersion and Polymer Conformational Confinement in a

Jul 10, 2014 - Yiying Zhang , Navid Tajaddod , Kenan Song , Marilyn L. Minus ... Kenan Song , Yiying Zhang , Jiangsha Meng , Marilyn L. Minus .... A m...
0 downloads 0 Views 6MB Size
Article pubs.acs.org/JPCB

Nanotube Dispersion and Polymer Conformational Confinement in a Nanocomposite Fiber: A Joint Computational Experimental Study Jiangsha Meng,† Yiying Zhang,† Steven W. Cranford,‡ and Marilyn L. Minus*,† †

Department of Mechanical and Industrial Engineering and ‡Department of Civil and Environmental Engineering, Northeastern University, Boston, Massachusetts 02115, United States ABSTRACT: A combination of computational and experimental methods was implemented to understand and confirm that conformational changes of a polymer [specifically polyacrylonitrile (PAN)] vary with the dispersion quality and confinement between single-wall carbon nanotubes (SWNT) in the composite fibers. A shear-flow gel-spinning approach was utilized to produce PAN-based composite fibers with high concentration (i.e., loading of 10 wt %) of SWNT. Dispersion qualities of SWNT ranging from low to high were identified in the fibers, and their effects on the structural morphologies and mechanical properties of the composites were examined. These results show that, as the SWNT dispersion quality in terms of distribution in the fiber and exfoliation increases, PAN conformations were confined to the extended-chain form. Full atomistic computational results show that the surface interaction energy between isolated PAN and SWNT was not preferred, leading to the self-agglomeration of PAN. However, confinement of the polymer chains between SWNT bundles or individual tubes (i.e., molecular crowding) resulted in large increases in the PAN−SWNT interaction energy. In other words, the crowding of polymer chains by the SWNT at high concentrations can promote extended-chain conformational development during fiber spinning. This was also evidenced experimentally by the observance of significantly improved PAN orientation and crystallization in the composite. Ultimately this work provides fundamental insight toward the specific structural changes capable at the polymer/nanotube interface which are important toward improvement of the effective contribution of the SWNT to the mechanical performance of the composite.



INTRODUCTION The inconsistency of mechanical properties often exhibited by polymer/single-wall carbon nanotubes (SWNT) composite fibers has been typically attributed to the difficulties in obtaining highly purified and uniform SWNT materials,1−4 controlling SWNT dispersion quality,5−7 and improving fillers/ matrix interactions during solution preparation,8−10 and it remains a limiting challenge for processing these materials as well as for their subsequent applications. Inconsistencies in the sample quality of nanotube fillers are also typically dependent on different synthesis procedures/parameters. For example, small variations in synthesis routes can lead to large structural variations of SWNT samples (i.e., tube length, chirality distribution, aspect ratio, and morphological consistency), which influence the electrical, thermal, and mechanical performance of both the as-produced tubes (i.e., fillers) and the polymeric composite fibers incorporating them.2,4,11−13 As mentioned, another factor that also contributes significantly to the property variation in polymer-based composites is the irregularity in the dispersion quality of the SWNT within the material (i.e., formation of aggregations and agglomerations, as well as a lack of bundle exfoliation14). It is well-known for SWNT materials that the tubes naturally self-assemble in bundle form due to van der Waals forces.15 One approach for reducing such bundling in a SWNT−solvent © 2014 American Chemical Society

dispersion is the introduction of a strong external shear stress (e.g., generated by sonication processes) between tube surfaces. This mechanism breaks down the aggregates and forms a temporary exfoliated state of the tubes in dispersion. This process allows the surface adsorption of dispersants or polymers to become possible, which may subsequently stabilize the dispersed state.16 However, the bubble implosion process (i.e., cavitation) occurring during sonication can be seen as random and qualitatively depends on conditions such as ultrasound intensity and frequency, pulsing interval and duration, SWNT concentration, external pressure, and temperature.17 For this reason, the homogeneity of the resultant SWNT bundle exfoliation levels can be very difficult to control, especially for the quality of carbon nanotubes needed for composite processing. In particular, for cases of dispersing high concentrations of SWNT (e.g., greater than 5 wt %), tube exfoliation processes may be further limited by the local spatial confinements from other surrounding SWNT bundles and tubes. This lack of control regarding the overall dispersion process for nanotube materials directly affects the ultimate quality of the SWNT distribution in a polymer matrix. For this Received: May 13, 2014 Revised: July 10, 2014 Published: July 10, 2014 9476

dx.doi.org/10.1021/jp504726w | J. Phys. Chem. B 2014, 118, 9476−9485

The Journal of Physical Chemistry B

Article

SWNT dispersion, homogeneity, and distribution throughout the composite fibers. Wide-angle X-ray diffraction analysis (WAXD) was used to study the global crystallization behavior and orientation of the polymer chains as influenced by SWNT dispersion quality. Full atomistic molecular dynamics (MD) computational simulations were used to study free surface and confined PAN−SWNT interactions resulting from the dispersion conditions. Five PAN−SWNT distribution scenarios were designed and modeled to understand the difference in PAN−SWNT interaction energies as a function of the dispersion quality of SWNT. These coupled computational and experimental studies provide direct insight toward molecular conformational changes that affect the PAN− SWNT interphase development of the composite.

reason, this lack of homogeneity present in the SWNT dispersion will lead to the increase of defective elements (i.e., aggregates and variation in SWNT bundle size and length leading to edge/end defects) and the decrease of polymer− SWNT interactions, which finally lower the mechanical performance of the composite end product. It is also well-recognized that an interfacial interaction between a polymer matrix and SWNT is an important factor determining the effective utilization of the exceptional mechanical properties of nanotubes within composite fibers.8,18−20 If a homogeneous polymer/SWNT interaction can be achieved throughout the fiber system, then SWNT bundle reaggregation may be prevented by polymer wrapping and the crystallization behavior of the polymer chains, as well as the formation of an ordered polymer interphase region between the SWNT and the bulk matrix.8,18,19 The growth of the ordered interphase region is considered to be the major factor for maximizing the mechanical contribution from the SWNT, since it provides a graded-region between the SWNT and bulkpolymer matrix to improve stress transfer in the composite.14,20 In addition, the mechanical contribution from the ordered interphase region has been shown to be significantly higher than the bulk-polymer (i.e., semicrystalline) region and further enhances composite fiber performance.4,20 As stated, to achieve higher mechanical improvement in the nanocomposite requires (a) an even distribution of SWNT, (b) exfoliation of the SWNT bundles (i.e., no aggregations and entanglements), and (c) the formation of a distinct crystalline polymer interphase structure in the fiber. In particular, for the high SWNT concentration case, it is most probable that some of the SWNT aggregates will not fully exfoliate and interact with the polymer chains due to insufficient polymer/SWNT interactions at various locations. For this reason, highconcentration polymer/SWNT dispersion may separate into SWNT-rich (i.e., aggregate-SWNT interaction dominates) and polymer-rich domains (i.e., interchain interaction is preferred), which ultimately act as defective regions in the fibers and result in early failure at these region boundaries (similar to grain boundaries of polycrystalline materials21). Although composite fibers with similar components, SWNT loadings, and fiber-spinning processes have been produced by various research groups, the mechanical performances of the final products deviate significantly.22−26 For instance, the PVA/ SWNT fibers (>60 wt % of SWNT) produced by polymer flow coagulation method have a modulus of 7−15 GPa and a strength of 0.15 GPa;24 and the fibers with similar SWNT loading and spinning approach achieved a modulus of 40−80 GPa and a strength of 0.3−1.8 GPa.22,23 In another case, the PAN/SWNT fibers (1 wt % of SWNT) produced by the wetspinning method was reported to have a modulus of 10.3 GPa and a strength of 0.56 GPa,26 and the fibers with the same SWNT loading and similar spinning approach achieved a modulus of 28.7 GPa and a strength of 1.1 GPa.25 For this reason, in order to minimize the property deviation, a more fundamental understanding regarding the structure−property relationship as a function of changes in processing parameters (i.e., control thereof) is necessary. In this work, PAN/SWNT composite fibers with 10 wt % SWNT were produced by shear-flow gel-spinning. Fiber sample batches obtained by the same spinning procedure exhibited different mechanical performance due to controlled variations in SWNT dispersion quality. Thermal charring was used to remove the polymer domains in order to visualize the level of



METHODS Materials. PAN-co-MAA (molecular weight 513 000 g· mol−1) was obtained from Japan Exlan Co. SWNT (purity >90 wt %, ash 900 °C in nitrogen, while the degradation temperature of the as-received PAN powder was found to be ∼200 °C in air and ∼250 °C in nitrogen. For this reason, all fibers were heated in a nitrogen environment from 25 to 900 °C at a rate of 5 °C/ min to burn out PAN regions. It is recognized that during the thermal treatment some fiber shrinkage and distortion will occur. However, due to the higher thermal stability of the SWNT under the charring conditions, the PAN/SWNT composites retained its fiber-like form, and in general, the original SWNT morphology was preserved. The charred composite fibers were observed by SEM (Figure 2). After thermal charring, F-1 fibers (Figure 2a) exhibited large voids due to the burnout of PAN-rich regions. This may be

indicative of a poor interaction between the PAN and SWNT during the dispersion process, resulting in separation of the dope into PAN-rich and SWNT-rich domains. This structure was subsequently translated to the spun F-1 fibers. For F-2 fibers, such large voids were not observed as compared to F-1. However, medium sized voids as well as large SWNT aggregations/entanglements (Figure 2b) were observed. The presence of these SWNT aggregates also suggests a poor interaction between PAN and SWNT during dispersion and subsequent fiber formation. As compared to both F-1 and F-2 batches, F-3 fibers (Figure 2c) exhibited a much more homogeneous distribution of the SWNT throughout the composite. There was also no evidence of large void structures arising from polymer-rich regions or SWNT aggregates throughout the fiber. The F-3 morphological features are more consistent with an improved PAN−SWNT interaction as compared to those of F-1 and F-2 fibers. WAXD was used to further study the morphology of the PAN/SWNT fiber batches in terms of the polymer crystal size, crystallinity, and orientation. Parts a, b, and c of Figure 3 show the WAXD patterns of the F-1, F-2, and F-3 fibers, respectively. To distinguish PAN peaks from the carbon reflections, a reference WAXD pattern pertaining to the as-received SWNT powder is also shown in Figure 3d. The PAN crystal sizes (L) for the predominant crystallization plane (110) were calculated on the basis of Scherrer’s approach (eq 1) for all of the composite fibers B(2θ ) =

Kλ L cos(θ)

(1)

where λ is the X-ray wavelength of 0.1541 nm; K is the Scherrer constant, taken to be 0.9; θ is the Bragg angle in radians; and B is the full-width at half-maximum (fwhm) at the 2θ angle for the (110) plane in radians. The average crystal sizes for the 9479

dx.doi.org/10.1021/jp504726w | J. Phys. Chem. B 2014, 118, 9476−9485

The Journal of Physical Chemistry B

Article

Figure 3. Wide-angle X-ray diffraction (WAXD) patterns for F-1 (a), F-2 (b), and F-3 (c) fibers and (d) SWNT material. (e) The comparison of PAN crystallinities and crystal sizes in F-1, F-2, and F-3 fibers as a function of the SWNT dispersion qualities, respectively. (f) Azimuthal scans and fitted curves for the (110) and (g) (020) diffraction planes from WAXD patterns.

(110) plane for the F-1, F-2, and F-3 fibers were calculated to be 11.2, 11.5, and 11.8 nm, respectively (Figure 3e). By fitting both crystalline and amorphous peaks from the WAXD patterns, the crystallinity values for F-1, F-2, and F-3 were determined to be 51.7%, 50.2%, and 53.8%, respectively (Figure 3e). As the SWNT dispersion quality improves from F-1 to F-3 batches, the polymer crystal size and crystallinity changes only slightly, indicating that the SWNT does not have a major influence on the PAN crystal growth process. The influence of the SWNT dispersion quality on the overall polymer orientation was also studied by calculating the Herman’s orientation factor (f) from the PAN (110) and (020) WAXD diffraction planes using Wilchinsky’s equation35 (eq 2)

⟨cos2 θ ⟩c − axis = 1 −

(1 − 2 sin 2 ρ2 )⟨cos2 φ1⟩ − (1 − 2 sin 2 ρ1)⟨cos2 φ2⟩ sin 2 ρ1 − sin 2 ρ2 (2)

where ⟨cos2 φ⟩ is determined using eq 3: 2

⟨cos φ⟩(hkl) =

∫0

π /x

I(φ) cos2 φ sin φ dφ

∫0

π /x

I(φ) sin φ dφ

(3)

The Herman’s orientation factor (f) for the polymer chains is determined with ⟨cos2 θ⟩ obtained from Wilchinski’s approach using eq 4: 9480

dx.doi.org/10.1021/jp504726w | J. Phys. Chem. B 2014, 118, 9476−9485

The Journal of Physical Chemistry B

Article

3⟨cos2 θ ⟩ − 1 (4) 2 The subscripts 1 and 2 correspond to the (110) and (020) planes, respectively, and ρ1 and ρ2 are the angles between the plane normal of these two planes and the b-axis. I(φ) is the integrated azimuthal intensity for each crystalline peak with respect to the azimuthal angle (φ). The azimuthal scans for both (110) and (020) planes are plotted in parts f and g of Figure 3, respectively. On the basis of the azimuthal curves, the fwhm decreases as the orientation factor increases and is most pronounced for both the (110) and (020) planes for the F-3 fibers. The parameters and calculated orientation factors for all the composite fibers are listed in Table 2. As compared to the

alignment. These interactions are further explored using full atomistic computational analysis, but first we assess the composite system theoretically. Theoretical Analysis. In order to analyze the stiffness contribution from the SWNT in the composites, the theoretical rule-of-mixture treatment was used (eq 5) to calculate the effective modulus contribution of the SWNT filler in all composite fibers:

f=

Ec = (ESWNT − eff − E PAN)Vf + E PAN − eff

Vf is the volume fraction of SWNT in the composite fibers, and EPAN−eff and ESWNT−eff are the effective Young’s modulus of the PAN and the SWNT, respectively. As discussed, the average PAN chain molecular orientation was calculated by WAXD analysis. This orientation data was used to determine the effective Young’s modulus values for PAN (EPAN−eff) in each of the fibers (eq 6) to be used in conjunction with eq 5:36

Table 2. Parameters Used for Calculating the Effective Young’s Modulus and Orientation Factor (f) of PAN, as Well as the Effective Young’S Modulus of Contribution of SWNT in All Three Composite Fibers Which Show a Variation in Dispersion of the Filler parameters

F-1

F-2

Parameters for Calculating f PAN and EPAN−eff E1 (GPa) 22.1 22.1 (control fiber modulus) 50 10 10 E2 (GPa) ν12 0.5 0.5 G12 (GPa) 1 1 ⟨cos2 θ110⟩ 0.06 0.06 ⟨cos4 θ110⟩ 0.01 0.01 ρ1 (deg) 59.1 59.1 ρ2 (deg) 0 0

1 E PAN − eff

=

Ec (GPa) Vf ESWNT−eff (GPa)

0.75 0.77 7.3 7.3 Parameters for Calculating ESWNT−eff 12.8 ± 0.9 18.4 ± 2.2 0.128 0.128 50.4

94.2

⎛ 1 2v 1 2⎞ +⎜ − 12 − ⎟⟨cos2 θ ⟩ E1 ⎝ G12 E1 E2 ⎠ ⎛1 2v ⎞ 1 1 +⎜ + − + 12 ⎟⟨cos 4 θ ⟩ E2 G12 E1 ⎠ ⎝ E1

F-3 22.1

(6)

⟨cos θ⟩ and ⟨cos θ⟩ values are determined using eqs 7 and 8. 2

10 0.5 1 0.04 0.01 59.1 0

4

1/2π

2

⟨cos θ ⟩ =

∫0

I(θ ) cos2 θ sin θ dθ

1/2π

∫0

1/2π

4

⟨cos θ ⟩ = f PAN EPAN−eff (GPa)

(5)

∫0

(7)

I(θ ) cos 4 θ sin θ dθ

1/2π

∫0

0.79 8.2

I(θ ) sin θ dθ

I(θ ) sin θ dθ

(8)

E1, E2, G12, and ν12 are the longitudinal modulus, transverse modulus, in-plane shear modulus, and the Poisson’s ratio of PAN, respectively. All parameters as well as the calculated EPAN−eff values are listed in Table 2. By using the experimentally measured values for the composite fiber modulus (i.e., Ec), the effective contribution of SWNT pertaining to each batch (i.e., F-1, F-2, and F-3) is determined (Table 2). The effective SWNT moduli values are within the range of the experimentally determined properties for SWNT ropes.37 This analysis also shows that the effective modulus contribution of the SWNT improves as the dispersion qualities is observed to increase. This increase in SWNT dispersion and distribution quality throughout the fiber also affects the PAN chain alignment. The increase of the effective SWNT contribution is indicative of better stress transfer from the PAN matrix to the tubes. The highest effective contribution of SWNT (∼197 GPa) was found in F-3 fibers, which exhibits the most uniform SWNT dispersion and highest PAN orientation, crystal size, and percent crystallinity. This theoretical treatment also supports the notion that PAN− SWNT interactions improve with dispersion quality. Computational Analysis. Both the experimental and theoretical sections support the evidence that, as the SWNT dispersion improves, the composite mechanical properties increase. To complement experimental observations and theoretical predictions, full atomistic computational analysis is implemented to fundamentally explore how very slight changes in SWNT dispersion lead to significant differences in PAN− SWNT interactions, as well as in PAN conformational changes,

32.4 ± 3.8 0.128 197.4

crystal size and crystallinity, which only exhibit slight changes as the SWNT dispersion quality increased, the improvement in orientation of the polymer chains is much more pronounced. This increase can be attributed to the improvement in homogeneity of SWNT bundle-size and the reduction of the SWNT agglomerations, as well as the increase in PAN−SWNT interactions (i.e., no evidence of PAN-rich regions) within the fibers. All three features may play a role in improving PAN chains alignment and promoting PAN−SWNT interfacial interactions during processing. On the basis of this structural analysis, it can be concluded that the improvement in the mechanical performance of the fiber batches from F-1 to F-3 is related to the improvements of PAN orientation, SWNT orientation, and SWNT dispersion quality. The increase in modulus is related to an increase of orientation for filler and polymer, while the increase in strength is due to a decrease in defective features such as voids and SWNT aggregations. Although the crystallinity and crystal size exhibits only slight improvement, the global distribution of these regions is more uniform as the filler distribution becomes more homogeneous. It is noted that as the SWNT dispersion quality improves, the PAN−SWNT interaction becomes more pronounced and the SWNT may also act as a template for PAN 9481

dx.doi.org/10.1021/jp504726w | J. Phys. Chem. B 2014, 118, 9476−9485

The Journal of Physical Chemistry B

Article

Figure 4. Molecular dynamics simulation snapshots of PAN/SWNT free interactions under 300 K with the following initial configurations: (a1) case 1, three PAN chains located on one side of a SWNT; (b1) case 2, three PAN chains located on one side of a SWNT bundle; (c1) case 3, three PAN chains located inside and outside of the dispersed SWNT bundles with different bundle sizes; (d1) case 4, three PAN chains located within the space of the mixture of SWNT bundles and individual SWNT; and (e1) case 5, three PAN chains located within the space of the dispersed individual SWNT. The final configurations of each case after 2 ns are shown in parts a2, b2, c2, d2, and e2, respectively.

individual PAN chains and (2) SWNT bundles to PAN chains. Experimentally, these two major polymer−SWNT interactions have been recognized.38−42 Similarly, some studies have also been performed from a computational viewpoint in order to obtain molecular level information or to study conditions that are not accessible experimentally.43−48 Here, the computational models are defined by taking both the experimental and theoretical results into consideration in order to assess the best possible representation of the PAN−SWNT molecular interactions observed in the composite fiber. The PAN−SWNT molecular interactions can vary greatly depending on either simple surface or confined interactions between the polymer chains and the nanotubes. To evaluate these molecular effects, the interaction energy between the PAN chains and SWNT is studied in five scenarios. These scenarios include two surface variations, (a) PAN chain interactions with and individual SWNT (small diameter) and (b) PAN chain interactions with a SWNT bundle (19 tubes of total 8.9 nm diameter), as well as three conf inement variations, (c) PAN chains confined by SWNT bundles, (d) PAN chains

which will contribute to the property improvements. While the recognition for the positive influence of a good SWNT dispersion in the composite is not new, the combination of the experimental, theoretical, and computation approaches used here provides insight regarding the codependency of atomistic and nanoscale phenomenological changes that occur to and between the components, which include (1) polymer orientation, (2) SWNT exfoliation, and (3) SWNT distribution, to form the ultimate macroscopic composite morphology. Using full atomistic molecular models allows precise control of the SWNT dispersion, enabling the exploration of minute changes in system conformation, unable to be achieved and/or controlled experimentally. To understand the fundamental issues relating processing approaches to the final fiber structure as well as the ultimate properties, the molecular interactions between the PAN and SWNT are explored at the smallest molecular interfacial scale. To simplify the computational models used here for investigating such interactions, the assumption is made that the PAN−SWNT interactions occur in primarily two main combinations: (1) individual SWNT to 9482

dx.doi.org/10.1021/jp504726w | J. Phys. Chem. B 2014, 118, 9476−9485

The Journal of Physical Chemistry B

Article

confined by a mixture of SWNT bundles and individual tubes, and (e) PAN chains confined by individual (i.e., exfoliated) SWNT. The confinement variations provide a local evaluation of molecular interactions, which are more likely to occur in a composite material, where the polymer is actually surrounded (i.e., confined) by the SWNT. In order to understand the energetic and conformational changes as a result of the variation of PAN−SWNT interactions (i.e., different degrees of interactions due to the dispersion development), a series of full atomistic MD simulations were performed to study both the PAN−SWNT and PAN−PAN interaction energies for each of the aforementioned scenarios (Figure 4). The two scenarios exploring the basic PAN−SWNT surface interactions are shown in Figure 4a,b. Three PAN chains are located near the surface of an individual SWNT (case 1, Figure 4a1) and a SWNT bundle (case 2, Figure 4b1). Both systems were allowed to freely interact at 300 K for 2 ns while the interaction energies were recorded. The final configurations for cases1 and 2 are shown in parts a2 and b2 of Figure 4, respectively. The average PAN−SWNT and PAN−PAN interaction energies per PAN monomer after three simulations for each case were also recorded (Figure 5a−c). On the basis of the simulation results analyzing surface variations, case 1 (Figure 4a1,a2) shows that while there is some degree of PAN−SWNT interaction, the PAN−PAN interaction energy was predominant, e.g., polymer self-aggregation is energetically favorable. For this reason, a local PAN-rich region is formed near the SWNT surface. Case 2 (Figure 4b1,b2) exhibits an improvement of the PAN−SWNT interaction and a decrease in PAN−PAN interaction energy. Due to the increase in graphitic surface available, the polymer chains have more opportunity to interact with the SWNT bundle. It is understood that the predominant PAN−PAN interactions throughout the composite fiber system would lower the stress transfer efficiency from the matrix to the SWNT. However, better matrix/filler interaction will lessen the PAN interchain interactions. These simple surface studies show that, in isolation, the PAN chains prefer polymer−polymer interactions rather than polymer−SWNT interactions. However, increasing the presence of SWNT (and thus the available surface area) in the near vicinity can reduce PAN chains self-interactions by increasing the interfacial polymer−nanotube conformational space. It is also noted that the PAN chains exhibit a coiled conformation on the SWNT surface. These results may provide some insight toward the experimental observation of polymerrich and SWNT-rich regions forming in the composite during processing, as the inherent PAN−SWNT interaction is relatively weak. To study the conf inement variations on the PAN−SWNT interactions, three additional cases with different SWNT dispersion qualities were constructed and analyzed. The differences between these three cases are in relation to the size and distribution as well as dispersion of the SWNT bundles and individual tubes. Three PAN chains are surrounded (i.e., confined) by dispersed SWNT bundles (case 3, Figure 4c), a mixture of dispersed SWNT bundles/tubes (case 4, Figure 4d), and a dispersed individual SWNT (case 5, Figure 4e), respectively. Similar to the surface studies, all systems were allowed to freely interact at 300 K for 2 ns while the interaction energies were recorded. The final configurations for cases 3, 4, and 5 are shown in parts c2, d2, and e2 of Figure 4, respectively. The average PAN−SWNT and PAN−PAN interaction energies per PAN monomer of three simulations for each case are

Figure 5. (a) PAN−SWNT and (b) fitted PAN−PAN interaction energies for all cases were plotted. (c) A summary bar chart for all interaction energy values, where PAN−PAN interaction energies increase (weaken) and PAN−SWNT interaction energies decrease (strengthen) as the SWNT dispersion quality and distribution increases.

shown and summarized in parts a−c of Figure 5, respectively. It should be noted that, for all current simulations, no solvent molecules were included in the models. As such, the absolute values of interaction energies may change with the presence and type of solvent. That being said, the trends of the energy changes from case to case should remain qualitatively consistent, since the same solvent effect would be equally imposed throughout all cases. Both case 3 and 4 exhibit PAN−SWNT and PAN−PAN interaction energy levels similar to those of case 2 (PAN− 9483

dx.doi.org/10.1021/jp504726w | J. Phys. Chem. B 2014, 118, 9476−9485

The Journal of Physical Chemistry B

Article

SWNT composites. These results also provide insight toward the formation of PAN ordered interphase structures in the nanocomposite fiber, as well as the importance of a good dispersion of the nanofiller. Such interactions can contribute to the improvement of the fiber mechanical properties, maximizing the stress transfer from the matrix to the fillers.

SWNT bundle surface interactions). As the dispersion level was increased to full exfoliation of the SWNT tubes for case 5 (Figure 4e1,e2), the highest PAN−SWNT bonding energy and the lowest PAN−PAN interaction energy were observed (Figure 5). Simulation snapshots also depict that the PAN chain conformation becomes more extended due to confinement between the SWNT. As the SWNT dispersion becomes more uniform, this conformational confinement effect is more pronounced. Case 5 exhibits fully extended chain PAN morphology in the vicinity of the SWNT surface (Figure 4e). As discussed in the Introduction, this extended-chain morphology is considered to be a major factor in maximizing the stress-transfer from a matrix to SWNT and significantly enhances the composite fiber performance. It has also been experimentally confirmed by observing the crystallization behavior of PAN in a dilute SWNT dispersion.19 The simulation results suggest that, as SWNT exfoliation and dispersion (i.e., distribution of PAN and SWNT) improve, the PAN molecular conformation and its interaction with the SWNT increased from case 1 to case 5. Therefore, by combining the experimental evidence (i.e., increase in mechanical properties, crystal size, orientation, and crystallinity as the SWNT dispersion quality improved) and the SWNT effective modulus analysis (i.e., SWNT effective modulus increased as the dispersion quality improved) deduced by theoretical analysis, the difference in the mechanical performance from F-1 to F-3 fibers may be explained by the various compositions of different PAN−SWNT interaction levels within composite fibers. Understanding the importance of improving polymer−SWNT interactions for composite performance enhancement has been recognized for some time, and this study provides critical and complementary computational and experimental evidence indicating how molecular conformation and confinement evolve and vary with nanotube dispersion and translate into structural features within the material.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Phone: (617) 373-2608. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS The authors thank the Air Force Office of Scientific Research (AFOSR FA 9950-11-1-0153) for the financial support of this research work. Simulations were performed on high-performance multicore Linux workstations from NEU’s Laboratory for Nanotechnology In Civil Engineering (NICE). Simulation visualization has been carried out using the VMD visualization package.49



REFERENCES

(1) Andrews, R.; Weisenberger, M. C. Carbon Nanotube Polymer Composites. Curr. Opin. Solid State Mater. Sci. 2004, 8, 31−37. (2) Hu, J. T.; Odom, T. W.; et al. Chemistry and Physics in One Dimension: Synthesis and Properties of Nanowires and Nanotubes. Acc. Chem. Res. 1999, 32, 435−445. (3) Qian, H.; Greenhalgh, E. S.; et al. Carbon Nanotube-Based Hierarchical Composites: A Review. J. Mater. Chem. 2010, 20, 4751− 4762. (4) Song, K.; Zhang, Y.; et al. Structural Polymer-Based Carbon Nanotube Composite Fibers: Understanding the Processing− Structure−Performance Relationship. Materials 2013, 6, 2543−2577. (5) Zhang, Y.; Minus, M. L. Study of Dispersion Effects on Exfoliation and Length Reduction of Single-Walled Carbon Nanotubes towards Fabricating Polymer-Based Composites. In Proceedings of the ACS 2012 Spring Meeting; San Diego, CA, 2012. (6) Fu, S. Y.; Chen, Z. K.; et al. The Reduction of Carbon Nanotube (CNT) Length during the Manufacture of CNT/Polymer Composites and a Method To Simultaneously Determine the Resulting CNT and Interfacial Strengths. Carbon 2009, 47, 3192−3200. (7) Green, M. J. Analysis and Mesurement of Carbon Nanotube Dispersion: Nanodispersion Versus Macrodispersion. Polym. Int. 2010, 59, 1319−1322. (8) Minus, M. L.; Chae, H. G.; Kumar, S. Interfacial Crystallization in Gel Spun Poly(vinyl alcohol)/Single-Wall Carbon Nanotube Composite Fibers. Macromol. Chem. Phys. 2009, 210, 1799−1808. (9) Barber, A. H.; Cohen, S. R.; et al. Interfacial Fracture Energy Measurements for Multi-Walled Carbon Nanotubes Pulled from a Polymer Matrix. Compos. Sci. Technol. 2004, 64, 2283−2289. (10) Roy, D.; Bhattacharyya, S.; et al. Measurement of Interfacial Shear Strength in Single Wall Carbon Nanotubes Reinforced Composite using Raman Spectroscopy. J. Appl. Phys. 2010, 107. (11) Salvetat, J. P.; Bonard, J. M.; et al. Mechanical Properties of Carbon Nanotubes. Appl. Phys. A: Mater. Sci. Process. 1999, 69, 255− 260. (12) Wong, E. W.; Sheehan, P. E.; et al. Nanobeam Mechanics: Elasticity, Strength, and Toughness of Nanorods and Nanotubes. Science 1997, 277, 1971−1975. (13) Yu, M. F.; Files, B. S.; et al. Tensile Loading of Ropes of Single Wall Carbon Nanotubes and Their Mechanical Properties. Phys. Rev. Lett. 2000, 84, 5552−5555. (14) Coleman, J. N.; Khan, U.; et al. Small but Strong: A Review of the Mechanical Properties of Carbon Nanotube−Polymer Composites. Carbon 2006, 44, 1624−1652.



CONCLUSION Three sets of PAN/SWNT composite fibers with a high concentration of SWNT (10 wt %) were produced using a flow gel-spinning apparatus. Although following the same fiberspinning procedure, significant difference in fibers mechanical performances was observed. SEM images of all fibers after a thermal charring process revealed different SWNT dispersion qualities within composite fibers. A Young’s modulus ratio of 1:1.4:2.5 and a tensile strength ratio of 1:1.9:3.2 were found for the first (F-1), the second (F-2) and the third (F-3) sets of composite fibers, respectively. WAXD analysis suggested slight increases in PAN crystal size and crystallinity, as well as a significant increase in polymer orientation as SWNT dispersion quality improved. The mechanical contributions from SWNT bundles for all composite fibers were calculated using the ruleof-mixture expression by taking orientation factor into account. The calculated values showed an increase in SWNT contribution as the dispersion quality improved. To provide a molecular perspective, full atomistic MD simulations on various SWNT dispersion conditions showed a positive correlation between PAN−SWNT interaction and the SWNT dispersion quality. Confinement of PAN chains between well-dispersed SWNT resulted in conformational changes of the polymer leading to extended-chain structures. We therefore demonstrate that SWNT dispersion is a critical processing condition to enable the tuning of the mechanical properties of polymer/ 9484

dx.doi.org/10.1021/jp504726w | J. Phys. Chem. B 2014, 118, 9476−9485

The Journal of Physical Chemistry B

Article

Rheological Properties of Polyethylene/CNT Composites. Physica E 2008, 40, 2440−2445. (39) Ci, L. J.; Bai, J. C. The Reinforcement Role of Carbon Nanotubes in Epoxy Composites with Different Matrix Stiffness. Compos. Sci. Technol. 2006, 66, 599−603. (40) Barber, A. H.; Cohen, S. R.; et al. Measurement of Carbon Nanotube−Polymer Interfacial Strength. Appl. Phys. Lett. 2003, 82, 4140−4142. (41) Liu, L. Q.; Wagner, H. D. Rubbery and Glassy Epoxy Resins Reinforced with Carbon Nanotubes. Compos. Sci. Technol. 2005, 65, 1861−1868. (42) Song, Y. S.; Youn, J. R. Modeling of Effective Elastic Properties for Polymer Based Carbon Nanotube Composites. Polymer 2006, 47, 1741−1748. (43) Szleifer, I.; Yerushalmi-Rozen, R. Polymers and Carbon NanotubesDimensionality, Interactions and Nanotechnology. Polymer 2005, 46, 7803−7818. (44) Zaminpayma, E.; Mirabbaszadeh, K. Interaction between Singlewalled Carbon Nanotubes and Polymers: A Molecular Dynamics Simulation Study with Reactive Force Field. Comput. Mater. Sci. 2012, 58, 7−11. (45) Li, Y.; Liu, Y.; et al. Pull-out Simulations on Interfacial Properties of Carbon Nanotube-Reinforced Polymer Nanocomposites. Comput. Mater. Sci. 2011, 50, 1854−1860. (46) Minchaca, J. I.; Castillejos, A. H.; et al. Size and Velocity Characteristics of Droplets Generated by Thin Steel Slab Continuous Casting Secondary Cooling Air-Mist Nozzles. Metall. Mater. Trans. 2011, 42B, 500−515. (47) Shokrieh, M. M.; Rafiee, R. Prediction of Mechanical Properties of an Embedded Carbon Nanotube in Polymer Matrix Based on Developing an Equivalent Long Fiber. Mech. Res. Commun. 2010, 37, 235−240. (48) Tallury, S. S.; Pasquinelli, M. A. Molecular Dynamics Simulations of Stiff Backbone Polymers Interacting with Single-Walled Carbon Nanotubes. J. Phys. Chem. B 2010, 114, 9349−9355. (49) Humphrey, W.; Dalke, A.; et al. VMD: Visual Molecular Dynamics. J. Mol. Graphics Modell. 1996, 14, 33−38. (50) Sreekumar, T. V.; Liu, T.; et al. SWNT/PAN Composite Fibers. Adv. Mater. 2004, 16, 1583−1583.

(15) Gee, R. H.; Lacevic, N.; et al. Atomistic Simulations of Spinodal Phase Separation Preceding Polymer Crystallization. Nat. Mater. 2006, 5, 39−43. (16) Huang, Y. Y.; Terentjev, E. M. Dispersion of Carbon Nanotubes: Mixing, Sonication, Stabilization, and Composite Properties. Polymers 2012, 4, 275−295. (17) Huang, Y. Y.; Terentjev, E. M. Dispersion and Rheology of Carbon Nanotubes in Polymers. Int. J. Mater. Forming 2008, 1, 63−74. (18) Chae, H. G.; Minus, M. L.; Kumar, S. Oriented and Exfoliated Single Wall Carbon Nanotubes in Polyacrylonitrile. Polymer 2006, 47, 3705−3710. (19) Zhang, Y.; Song, K.; et al. Tailoring Polyacrylonitrile Interfacial Morphological Structure by Crystallization in the Presence of SingleWall Carbon Nanotubes. ACS Appl. Mater. Interfaces 2013, 5, 807−14. (20) Meng, J.; Zhang, Y.; et al. Forming Crystalline Polymer-Nano Interphase Structures for High-Modulus and High-Tensile/Strength Composite Fibers. Macromol. Mater. Eng. 2014, 299, 144−153. (21) Xu, J.; Ren, Y.; et al. Molecular Dynamics Simulation of Macromolecules using Graphics Processing Unit. Mol. Simul. 2010, 36, 1131−1140. (22) Dalton, A. B.; Collins, S.; et al. Super-Tough Carbon-Nanotube Fibres. Nature 2003, 423, 703−703. (23) Dalton, A. B.; Collins, S.; et al. Continuous Carbon Nanotube Composite Fibers: Properties, Potential Applications, and Problems. J. Mater. Chem. 2004, 14, 1−3. (24) Vigolo, B.; Penicaud, A.; et al. Macroscopic Fibers and Ribbons of Oriented Carbon Nanotubes. Science 2000, 290, 1331−1334. (25) Chae, H. G.; Minus, M. L.; et al. Stabilization and Carbonization of Gel Spun Polyacrylonitrile/Single Wall Carbon Nanotube Composite Fibers. Polymer 2007, 48, 3781−3789. (26) Mikolajczyk, T. Effect of Spinning Conditions on the Mechanical Properties of Polyacrylonitrile Fibers Modified with Carbon Nanotubes. J. Appl. Polym. Sci. 2009, 115, 3628−3635. (27) Plimpton, S. Fast Parallel Algorithms for Short-Range Molecular Dynamics. J. Comput. Phys. 1995, 117, 1−19. (28) Dauber-Osguthorpe, P.; Roberts, V. A.; et al. Structure and Energetics of Ligand Binding to Proteins: Escherichia coli Dihydrofolate Reductase−Trimethoprim, a Drug−Receptor System. Proteins: Struct., Funct., Bioinf. 1988, 4, 31−47. (29) Chenoweth, K.; van Duin, A. C. T.; et al. ReaxFF Reactive Force Field for Molecular Dynamics Simulations of Hydrocarbon Oxidation. J. Phys. Chem. A 2008, 112, 1040−1053. (30) Stuart, S. J.; Tutein, A. B.; et al. A Reactive Potential for Hydrocarbons with Intermolecular Interactions. J. Chem. Phys. 2000, 112, 6472−6486. (31) Brenner, D. W.; Shenderova, O. A.; et al. A Second-Generation Reactive Empirical Bond Order (REBO) Potential Energy Expression for Hydrocarbons. J. Phys.-Condens. Mater. 2002, 14, 783−802. (32) Los, J. H.; Ghiringhelli, L. M.; et al. Improved Long Range Reactive Bond Order Potential for Carbon (LCBOPII). Part I: Construction. Phys. Rev. B: Condens. Matter 2005, 72. (33) Cranford, S. W.; Buehler, M. J. Variation of Weak Polyelectrolyte Persistence Length through an Electrostatic Contour Length. Macromolecules 2012, 45, 8067−8082. (34) Cranford, S. W.; Ortiz, C.; et al. Mchanomutable Properties of a PAA/PAH Polyelectrolyte Complex: Rate Dependence and Ionization Effects on Tunable Adhesion Strength. Soft Matter 2010, 6, 4175− 4188. (35) Wilchinsky, Z. W. Measurement of Orientation in Polypropylene Film. J. Appl. Phys. 1960, 31, 1969−1972. (36) Wei, C. Y.; Srivastava, D.; et al. Thermal Expansion and Diffusion Coefficients of Carbon Nanotube−Polymer Composites. Nano Lett. 2002, 2, 647−650. (37) Salvetat, J. P.; Briggs, G. A. D.; et al. Elastic and Shear Moduli of Single-Walled Carbon Nanotube Ropes. Phys. Rev. Lett. 1999, 82, 944−947. (38) Valentino, O.; Sarno, M.; et al. Influence of the Polymer Structure and Nanotube Concentration on the Conductivity and 9485

dx.doi.org/10.1021/jp504726w | J. Phys. Chem. B 2014, 118, 9476−9485