Improvement of the tensile strength of carbon nanotube fibers to 12.5 GPa by fluidics-induced alignment and densification

Abstract

Achieving the superior tensile strength of individual carbon nanotubes (CNTs) in macroscopic fibers depends on how CNTs are assembled. This study reports a fluidics approach: spraying high-velocity ethanol to scour and shrink pre-made CNT fibers. Experimental measurements show that the continuous scouring and shrinking induce significant improvements of the CNT alignment and packing density, leading to large increases of the van der Waals forces between CNTs and thus the tensile strength of the fibers. Under optimal conditions, the new CNT fibers achieve a specific tensile strength of 7.5×10⁶N⋅m⋅kg⁻¹ (or 7.5 N⋅tex⁻¹) and absolute strength of 12.5 GPa, approaching the lower limit of individual CNTs. They also exhibit a high Young’s modulus of 370 GPa, electrical conductivity of ~3×10⁶ S⋅m⁻¹, and thermal conductivity of ~968 W⋅m⁻¹⋅K⁻¹. This fluidics approach offers a promising route to develop high-performance CNT materials comparable to CNT individuals.

Scouring and shrinking using fast flow of ethanol induced large improvements of carbon nanotube alignment and packing density and thus the tensile strength of the fiber to 12.5 GPa.

Introduction

Carbon nanotube (CNT) is considered a one-dimensional molecule with a tubular structure from rolling a hexagonal mesh of sp²-hybridized carbon atoms¹. Pulling apart this structure involves the switching of strong covalent C–C bonds and migration of a series of non-hexagonal rings^2–4. For these reasons, individual CNTs have been shown to possess very high mechanical and physical properties, including a tensile strength (σ_b) of 13–80 GPa accompanied by a high Young’s modulus (or stiffness) (E) and tensile ductility^5–9. It is such intriguing properties that have been providing an impetus to assemble the nano-sized tubes into macroscopic fibers and films for practical uses^10,11. For this purpose, promising techniques have been developed for assembling CNTs, such as the mainstream spinning from CNT solutions^12–15 and vertically aligned CNT arrays^16–19 and direct spinning from CNT aerogels^20,21 or hollow cylinders^22,23. However, the CNT assemblies developed to date exhibit only a fraction of the properties of individual CNTs. Great efforts contributed throughout the years prove that transferring the microscopic property to the macroscopic scale is not as easy as expected.

The mechanics of the CNT assemblies are primarily determined by the stress transfer between CNTs rather than solely the C–C bond and defect structure within CNTs. Thus, the technical issue for the assemblies to achieve a tensile strength comparable to that of individual CNTs is perfectly aligning long CNTs in their longitudinal direction and closely packing them in their transverse direction so that the mutual interaction between individual CNTs via the van der Waals force could become large enough to break them. Longer CNTs are beneficial because their mutual interaction is larger in comparison with their short counterparts. In this sense, although CNTs are well aligned in the fibers spun from solutions and arrays, the tensile strength measured for these fibers are low. It is necessary to investigate if the low strength can be improved with the use of long CNTs. Due to its potential for large-scale production, the direct spinning technique has been widely studied, with CNT alignment and close packing as the main technical hurdles to overcome to improve properties.

CNTs in the fibers and films from direct spinning are usually long and curled and show a low degree of alignment along the axial direction. Among the various methods proposed so far, such as molecular crosslinking, polymer infiltration, and die drawing^10,11, wet stretching was found to be effective in improving the CNT alignment and thus the tensile strength without introducing heterogeneous structures behaving as effective brittle sites or phonon and electron scattering centers and thus reducing the tensile ductility or electrical and thermal conductivities^24–26. Nevertheless, this method has some apparent limitations. One is that only a limited amount of stretching can be applied (up to 15%, engineering strain) without deteriorating the quality of the fiber. Such small stretching cannot straighten the curled CNTs and bring the divergent ones to the axial direction. The other limitation is that stress transfer efficiency is low in the poorly aligned CNTs, especially those that have been completely wet, with little inter-tube bonding. All these could lead to a large portion of CNTs in the fiber being unaffected by the stretching.

To achieve better CNT alignment and close packing, we propose a fluidics approach. That is, a fluid is sprayed onto the fiber at a high velocity. This is to introduce continuous scouring and shearing to all constituent CNTs in the direction of fluidic flow, thus allowing the manipulation of the CNT orientation and packing density. The effects of the flow parameters on the evolution of these microstructures are explored to realize a high tensile strength approaching the lower limit for individual CNTs. The scientific correlation between fluidics-enabled microstructure control and property improvement has been established, paving the way for new science in developing high-performance CNT fibers and films.

Results

Starting materials

A schematic of the present spinning process is shown in Fig. 1a. The reaction solution containing carbon precursor, Fe precursor, and growth promoter, was injected into the reactor for pyrolysis. CNTs formed in the high temperature zone and were integrated into a hollow cylinder in the low temperature region. The CNT hollow cylinder was blown out from the reactor by N₂ carrier gas (Fig. 1b) and then condensed on an Al foil substrate wrapped on the winding drum with a diameter of 1 m (Supplementary Fig. 1a). The CNT film on the substrate had a length of 3.14 m and width of 0.5 m (Fig. 1c) and was cut to a long ribbon-like fiber with a width of 2 mm (Fig. 1d and Supplementary Fig. 1b, c). See Supplementary Videos 1–4 for these experimental processes.

Fig. 1. Morphology and structure of the starting material.

a Schematic diagram of the synthesis of carbon nanotube (CNT) films by floating-catalyst spray pyrolysis. b Hollow CNT assembly being blown out from the reactor. c CNT film with a length of 3.14 m and a width of 0.5 m. d Roll of CNT ribbon of 2 mm wide from cutting the film. e Bright field (BF)-transmission electron microscopy (TEM) image of CNTs. f High resolution BF-TEM image of a double-walled CNT.

Air oxidation and acid washing were used for purification. Imaging using secondary electrons (SE) scanning electron microscopy (SEM) and bright field (BF) transmission electron microscopy (TEM) and high resolution (HR) TEM showed individual CNTs with clean surfaces and single- or double-walled structures (Fig. 1e, f). Thermogravimetric analysis (TGA) revealed an iron content of only 3.83%, and Raman spectroscopy, X-ray diffraction (XRD), and X-ray photoelectron spectroscopy (XPS) confirmed good CNT crystallinity and purity (Supplementary Fig. 2). After purification, the linear density of the ribbon was reduced, but the load-carrying capability remained about 1.2 N. Such a ribbon was strong enough to be scoured and stretched and thus used as the starting material for subsequent treatments. Routine experimentation revealed that weaker fibers are not suitable for repeated stretching and scouring without sacrificing their structural integrity.

CNT alignment from scouring

A schematic of the present post treatments is shown in Fig. 2. The raw ribbon-shaped fiber, denoted as R-CNTF, was continuously released at one end and winded at the other. 3 cases were studied: with direct stretching only (D-CNTF), with scouring only (S-CNTF), and with both (DS-CNTF). Figure 3a, b shows the surface morphology of the fiber before and after the direct stretching treatment. A large number of voids existed inside R-CNTF, and the individual CNTs, as well as their bundles, were entangled with each other. Slight alignment originated from the spinning of the original hollow CNT assembly. With direct stretching by a high amount of 15% (engineering strain), some CNTs and bundles were aligned, leading to a small increase of the intensity ratio of G band from polarized Raman spectroscopy (I_G‖/I_G⊥) from 2.38 to 3.2 and Herman’s orientation factor from small-angle X-ray scattering (SAXS) from 0.48 to 0.53 (Fig. 3e, f and Fig. 3i, j). However, many regions were slightly stretched and contained entangled CNTs and pores of varying sizes, indicating that stretching alone had only a limited effect on alignment and densification.

Fig. 2. Schematic diagrams of the carbon nanotube (CNT) materials treated by stretching and scouring and corresponding microstructure evolution.

a CNT ribbon fiber which is released at a velocity V₁ and winded at a velocity V₂ with the stretching strain determined by V₂-V₁. b CNT fiber showing CNT alignment and compaction as it is scoured and stretched. c Segment of a CNT film with one end fixed and the other pulled down. In both cases, ethanol is sprayed at a high speed to introduce continuous scouring.

Fig. 3. Secondary electrons-scanning electron microscopy images, polarized Raman spectra, and small angle X-ray scattering images of different carbon nanotube fibers (CNTFs) having undergone different stretching and scouring treatments.

a, e, i Raw CNTF (R-CNTF), no treatment; b, f, j Directly stretched CNTF (D-CNTF), only stretching; c, g, k Directly scoured CNTF (S-CNTF), only scouring; d, h, l Stretched and scoured CNTF (DS-CNTF), scouring coupled with stretching. Source data are provided as a Source Data file.

In the case of scouring, CNTs became straight, aligned, and densely packed. Tiny pores could be observed only at high magnification (Fig. 3c). The intensity ratio of G band (3.65) and Herman’s orientation factor (0.63) showed a larger increase in comparison with the stretching case (Fig. 3g, k). To further improve the structure, a scouring process was applied along with stretching. Under these circumstances, the densification became apparently higher with much fewer pores visible (Fig. 3d). Meanwhile, the orientation was further improved without entangled CNTs visible. This can also be seen from the much higher intensity ratio of G band of 5.27 and Herman’s orientation factor of 0.72 (or 0.97 based on Gaussian fitting) (Fig. 3h, l), indicating that the scouring coupled with stretching improved the CNT alignment and packing density significantly. Such improvements could cause the increases of the contact area and interaction between CNTs and the effective transfer of load, electron, and phonon between CNTs, resulting in large increases of mechanical, electrical, and thermal properties.

Effects of flow parameters

The flow parameters related to the fluidics of the scouring process are now considered. To induce the scouring effect at both the macroscopic and microscopic scales, the fluid must infiltrate into the CNT fiber. For this reason, ethanol was used rather than water as the CNT fiber is hydrophilic to ethanol but hydrophobic to water. Wetting the entire fiber volume was favorable for inducing scouring and sliding to facilitate CNT reorientation. It was found that both the flowing velocity of the ethanol sprayed on the fiber and the total spraying time affected the scouring effect in a similar way. To unify the effects of these two parameters and quantitatively describe the degree of scouring, we propose a concept termed as “scour distance”. This parameter is defined as the total distance that a segment of the fiber travels during a scouring process, which is measured as the flowing velocity multiplied by the flow time of the ethanol fluid. To maximize the CNT alignment, the coupling effect of stretching was investigated, and the gradient flow effect was examined by scouring an additional ribbon sample with a much thinner thickness.

A series of scouring experiments showed that both the intensity ratio of G band and the orientation factor increased with increasing scour distance, and at the same scour distance, the application of concurrent stretching generated more increment in orientation factor. This can be visualized from the gradual deformation of the isotropic circular SAXS pattern into more anisotropic eclipse patterns, thus suggesting enhancement of the CNT alignment during scouring and that coupled with stretching (Fig. 3, Supplementary Figs. 3 and 4). It appeared that scouring by distances larger than 1.2 km didn’t give rise to further significant enhancement in CNT orientation (Fig. 4a, Supplementary Fig. 5), thus suggesting the culmination of the scouring effect under the present experimental conditions.

Fig. 4. Effects of fluid scouring on microstructures and mechanical properties of the carbon nanotube fiber (CNTF) and additional thinner fiber (THINF).

a Influence of scour distance on CNT orientation. b Variations of maximum tensile load and linear density with scour distance. c Dependence of specific strength upon scour distance. d Comparison of strength between the fibers before and after rolling with the THINF unaffected. The upper limit of an error bar represents the maximum measured value minus the average value, and the lower limit represents the average value minus the minimum measured value. As a result, an error bar may not be symmetric about the reported average value. Black or red circles represent the datum points used for averaging. e Specific stress-strain curves of different samples, showing a little change in specific strength with rolling. f Stress-strain curves of different samples, showing a large increase in strength with the introduction of scouring. Source data are provided as a Source Data file.

Tensile testing showed that, corresponding to the improvement of CNT alignment, the maximum tensile load (the load at breaking) of the fiber increased gradually with increasing scour distance and stretching, and reached a maximum at the scouring distance of 1.2 km. This distance was achieved after several minutes at a high flow velocity. But, if too high a velocity was used, the sample was damaged in a short time. While the maximum tensile load was 1.205 N for the fiber before scouring, it was increased to 2.271 N for the fiber after having been scoured for 1.2 km and to 2.813 N for the fiber having also undergone stretching (Fig. 4b). During all the scouring processes, there was no significant change in linear density. Minor decrease observed from scouring up to 1.2 km may result from the loss of a very small amount of impurity into the flowing ethanol and the small lengthening of the fiber under stretching (Fig. 4b).

High specific strength

With the tensile load and linear density known, the specific strength was determined, showing the same trend of variation as the tensile load with scouring and stretching. That is, while the fiber before scouring had a specific strength of 2.61 × 10⁶N⋅m⋅kg⁻¹ (or 2.61 N⋅tex⁻¹), the fiber after scouring for 1.2 km had a specific strength of 5.33 × 10⁶N⋅m⋅kg⁻¹, and that one also undergoing stretching 7.03 × 10⁶N⋅m⋅kg⁻¹ (Fig. 4c). Since the scouring induced little change in linear density, the increase in specific strength was a result of the increase of tensile load. This means that as CNTs were aligned by scouring and stretching, inter-tube bonding and load transfer between CNTs was promoted, leading to a strengthening effect for the CNT fiber.

In order to examine the effect of the gradient of flow rate from the surface to the center of the ribbon center, a similar film was prepared but with a thickness of only 152 nm, differing from the above ribbon fiber with a starting thickness of 921 nm. This film was also cut to narrow ribbons and scoured with stretching for a scour distance larger than 1.2 km. All the results are presented in Supplementary Fig. 6. After the stretching-coupled scouring, CNTs were observed to be highly aligned and densely packed, and the intensity ratio of G band increased from 2.76 to 5.12. These microstructural changes led to large increases of maximum tensile load from 1.50 N to 3.75 N and specific strength from 3.5×10⁶N⋅m⋅kg⁻¹ to 7.5×10⁶N⋅m⋅kg⁻¹. Although there was a large difference in thickness (921 nm vs. 152 nm), a close specific strength was achieved by the thick fiber (CNTF) and the thin fiber (THINF) (7.03 × 10⁶N⋅m⋅kg⁻¹ vs. 7.5 × 10⁶N⋅m⋅kg⁻¹). This observation suggests that both the surface and the interior of the fiber or film were equally scoured, and the fluidics-driven ordering of CNTs took place over their whole volumes, regardless of the difference in thickness. This advantage of the little effect of flow gradient may originate from the thorough infiltration of ethanol molecules into the fiber, producing a scouring shear stress on all CNTs.

High strength and conductivity

With CNTs having been aligned and better packed, the fiber was further densified by mechanical rolling. With a proper control of the rolling process, the tensile load at break and specific strength changed little but the elongation at break was reduced (Fig. 4d, e). Since the linear density was not changed, the specific strength remained unaltered. However, due to the large reduction of cross-sectional area, the absolute strength was increased significantly. For example, for DS-CNTF, the cross-sectional area was reduced from 7.34 µm×85.13 µm to 0.44 μm×541.9 µm, the calculated strength was increased from 4.92 GPa to 11.7 GPa (Fig. 4e). But for the much thinner sample (DS-THINF), the strength remained unchanged at ~12.5 GPa. This may result from the fact that the ribbon sample was as thin as 0.15 μm, so that it had already been densified sufficiently by scouring and stretching, and/or the thickness of the sample was beyond the rolling limit of the rolling machine used.

Figure 4f summarizes the typical stress-strain curves of rolled R-CNTF, S-CNTF, and DS-CNTF and unrolled DS-THINF with more tensile curves and cross-sectional images presented in Supplementary Figs. 7 and 8, respectively. Both the DS-CNTF and DS-THINF exhibited continuous strain hardening and then abrupt failure. For these two high-strength samples, the Young’s modulus was 370 GPa and 260 GPa, and tensile elongation at break was 4.5% and 7.0%, respectively. In addition to these high mechanical properties, these scoured and stretched materials were also measured to possess a high electrical conductivity of 2.4 ~ 2.8 × 10⁶ S⋅m⁻¹ (specific conductivity of 1500 ~ 1750 S⋅m²⋅kg⁻¹) along the fiber axis. Since the sample of DS-THINF could be made to have a width of 3 cm and length of 10 cm, its in-plane thermal conductivity at the orientation of CNT alignment was measured by a laser flash method to be 867.96 ~ 968.03 W⋅m⁻¹⋅K⁻¹ (specific conductivity of 541 ~ 605 mW⋅K⁻¹⋅m²⋅kg⁻¹).

Comparison with previous studies

A radar plot was created to compare the performances of R-CNTF, S-CNTF, and DS-CNTF (Fig. 5a). The introduction of scouring is effective in improving the mechanical and physical properties. Figure 5b–f and Supplementary Tables 1–3 show comparisons of these properties from this study with previous results on materials of graphenes^27–30, CNTs^{17,24,25,31–50}, carbon fibers^51,52, and metals²⁰ published in the literature. The performance of the present fiber in terms of tensile strength and specific tensile strength is the highest among those ever reported to date. The physical properties of electrical and thermal conductivities of the present fiber and film also lie at high levels for the carbon materials produced by the leading methods. These results indicate that the present study provides a simple method for preparing high-performance CNT materials.

Fig. 5. Performances of the fibers prepared with or without scouring and their comparisons with other materials.

a Radar diagram showing the performance difference among the present different fibers. b Comparison in specific strength versus specific modulus. c Comparison in strength versus Young’s modulus. d Comparison in strength versus electrical conductivity. e Comparison in strength versus elongation at break. f Comparison in strength versus thermal conductivity. The result of thermal conductivity higher than 1000 W m⁻¹ k⁻¹ for graphene fibers is not included. : Graphene film; : Graphene fiber; : Floating catalyst chemical vapor deposition (FCCVD)-grown carbon nanotube (CNT) film; : Array-based CNT film; : CNT film from solution; : FCCVD-grown CNT fiber; : Array-based CNT fiber; : CNT fiber from solution; : Carbon fiber; : Metal; : This work (stretched and scoured CNT fiber). All data used for comparisons are included in Supplementary Tables 1–3.

Previous CNT fibers achieved strengths up to 8 GPa with the use of horizontal stretching in a static liquid for CNT alignment (e.g., refs. ^41,44). But the present fiber achieved a much higher strength of 12.5 GPa due to the use of vertical stretching in a flowing liquid. The flowing liquid induced microscopic shearing for CNT alignment and densification, and the external stretching only played an assisting role in these processes. In comparison with carbon fibers, the present CNT fiber has additional advantages in tensile ductility, toughness, and electrical and thermal conductivities. These advantages originate from the present CNT fiber having a graphitic tubular structure favorable for migration of mobile defects, electrons, and phonons, whereas carbon fibers have an amorphous carbon structure, which is intrinsically brittle and electron- and phonon-scattering. In the case of graphene fibers, significant progresses have been made by microstructure control in recent years^53–56. These include, but are not limited to, molecular reorientation of graphene sheets enabled by plasticization processing, their bidirectional ordering by multiple shear flows, and their dense packing by domain-folding. Such microstructure control has led to enhanced mechanical and physical properties, such as the mechanical strength to 3.4−5.19 GPa, electrical conductivity to 0.3−1.2 × 10⁶ S⋅m⁻¹, and thermal conductivity to 1480−1660 W⋅m⁻¹⋅K⁻¹. Further work is needed to improve the mechanical strength comparable to other carbon-based fibers.

Discussion

In order to transfer the microscopic property of CNTs to the macroscopic scale, it is necessary to have them fully aligned and closely packed so that the load transfer between CNTs can be enhanced to suppress inter-tube sliding and the inter-tube bonding can be maximized to induce the breaking of CNTs. The original fiber and film constituted CNTs, curled and poorly oriented. The application of fluidic flow enabled CNTs to be aligned and their assemblies to have high mechanical and physical properties.

Origin of CNT alignment

Aligning CNTs includes straightening CNTs and orientating them to the axial direction of the fiber. Previous mechanical stretching, combing, extrusion, and rolling are all inadequate in aligning some CNTs without destructing others. The present fluid flow provided a shear stress on wetted CNTs and straighten and direct them downstream. The increase in orientation factor with scour distance can be attributed to the continuous action of a shear stress from the flowing fluid that straightened and rotated the one-dimensional CNTs to be parallel to the flow direction. This shear stress is proportional to the flow rate and thus appears larger in the surface region than in the interior of the fiber. But with the proper control of scouring rate and time, this flow gradient didn’t cause an apparent microstructural gradient as can be seen from the similar scouring results from the thick ribbon fiber (DS-CNTF) and its much thinner counterpart (DS-THINF). Application of small stretching during scouring supplied an additional shear stress at the axial direction and promoted the CNT alignment further. The straightening and rotation of the CNTs were eventually limited by the free space available inside the fiber, leading to saturation of CNT ordering.

Enhancement of load transfer

As CNTs were aligned and porosity was mitigated, more and more CNTs were brought into parallel alignment and into contact with each other. These microstructural changes facilitated load transfer across CNTs and suppressed inter-tube sliding. The result was that all CNTs started to support externally applied load together and underwent breaking at the same time (Supplementary Fig. 9). In situ Raman spectroscopy (Supplementary Fig. 10) showed that the G’ band for DS-CNTF shifted to low frequencies during the whole straining process (up to 4.6%) at a high rate of 7.73 cm⁻¹/1%, but that for R-CNTF shifted to low frequencies only within the elastic straining range (up to 1%) at a low rate of 2.23 cm⁻¹/1%. These observations suggest that the load transfer in the aligned fiber took place across all CNTs simultaneously and continued to breaking strains, but that in the unaligned fiber occurred only across some CNTs after elastic straining due to the early breaking of others.

The enhancement of load transfer was further examined from fracture surfaces (Fig. 6). The R-CNTF sample illustrated a saw-toothed and uneven morphology, featuring characteristic slip and large laddering of CNT bundles, thus indicating widespread inter-tube sliding and pulling-out during the breaking stage. In contrast, the DS-CNTF sample demonstrated a neat and even fracture morphology with limited laddering, suggesting that the whole structure was broken with much less inter-tube sliding. We propose that these differences in fracture surface are also clear evidence of the effects of scouring. When CNTs were aligned and packed tightly, they transferred and supported the external load as a whole and maintained the homogeneity and integrity of the fiber until failure took place.

Fig. 6. Fracture surfaces of the fibers with carbon nanotubes (CNTs) unaligned and well aligned.

a, c–e Schematic and secondary electrons scanning electron microscopy (SE-SEM) images of unaligned raw CNT fiber (R-CNTF), showing wide inter-tube sliding and pulling-out of CNTs at a fracture surface; b, f–h Schematic and SE-SEM images of aligned stretched and scoured CNT fiber (DS-CNTF), showing little inter-tube sliding and even breaking of CNTs at a fracture surface. I, j High-resolution bright field transmission electron microscopy (BF-TEM) images of double-walled CNT bundles, which had collapsed and been broken during tensile testing. The red lines in (a, b) represent individual CNTs.

Strong inter-tube interaction

The shear strength between neighboring CNTs is the embodiment of CNT alignment and load transfer. Using the thin film samples with a width of 2 mm, two strips of these samples were overlapped over different lengths and pulled in tension to break to examine the shear strength between the strips (Fig. 7a, b). For the aligned strips (scoured and stretched) (DS-THINF), breaking occurred within the overlapping area, and the load at break increased as the overlapping length L_o increased. At a length of 40 mm and beyond, however, the breaking took place outside the overlapping area, and the breaking load reached the level of the strips used (Fig. 7c, e). For the unaligned strips from the original film (raw THINF), similar breaking phenomena were observed except that the strip itself was broken at a much smaller overlapping length of 20 mm The shear strength between CNTs was determined from the breaking load and shear area of the 2 mm wide strip (L_o×2 mm²). While the shear strength was 0.035 MPa for the aligned strips, a much lower shear strength of 0.005 MPa was observed for the unaligned ones (Fig. 7g, h). This observation evidences the importance of alignment for inter-tube strengthening.

Fig. 7. Analyses of shear strength between carbon nanotubes (CNTs).

a, b Schematic diagrams of the breakings along and outside the overlapping regions when the two CNT ribbons are overlapped over a length L_o < a critical length L_c and a length L_o ≥ L_c, respectively; c, d Measured load-strain curves of the stretched and scoured-thin film (DS-THINF) sample (with aligned CNTs) and the raw THINF one (without aligned CNTs). In both cases, the samples were tested with different lengths of overlapping (L_o = 5─40 mm) or no overlapping (pristine). e, f Observed breakings at different locations (red circle) for the samples of DS-THINF and raw THINF. In the DS-THINF case, the breaking took place within the overlapping region (red rectangle) when L_o = 10 mm, but outside it when L_o = 40 mm. In the raw THINF case, the breaking also occurred within the overlapping region when Lo = 10mm, but outside it when Lo was only 20mm. For both cases, the breaking of the pristine sample without overlapping (L_o = 0) is also shown. g, h Measured relationships between break load vs. overlapping area for samples of DS-THINF and raw THINF. Datum points are from panels (c and d), and the slope of the line best fitting these datum points represents the shear strength between CNT layers. Source data are provided as a Source Data file.

Previous experimental measurements and theoretical calculations have led to a large disparity in the magnitude of inter-tube shear strength. The general finding is that such strength depends upon a number of factors, including tube configuration (single-walled or multi-walled, bundle or yarn), graphitic defects in the tube structure, microstructural defects at the tube ends or walls (e.g., pores, functional groups), tube alignment, and even tensile velocity. While the shear strength measured in the present study is at the same level as those measured for basal-plane shear in single crystal graphite⁵⁷ and inter-tube sliding within multiwalled CNTs^58,59, it is significantly smaller than those measured for CNT bundles with perfect parallel alignment and minimal inter-tube porosity^60–63. The main reason is that the present measurement was made on macroscopic CNT strips with CNTs that are by no means perfectly aligned and densified, and the measured shear strength consisted of the contributions of inter-tube boundary, end, and impurity effects. It is believed that as the CNT alignment is further improved (with an orientation factor larger than the present 0.72) and other crucial structures further optimized (e.g., tube length, packing density, and uniformity), the CNT fiber could achieve a tensile strength higher than the present observation (12.5 GPa) and further narrows its disparity from perfect CNT bundles with an engineering tensile strength of 43 GPa⁶³.

CNT length effect

The planar shear strength measured for the CNT strips was high, a result of good alignment and packing of CNTs. To attain a high tensile strength for the macroscopic fiber, CNT length must be large enough to minimize free inter-tube sliding and suppress inter-tube fracturing. Similar strengths were measured at gauge lengths of 2, 10, 20, and 30mm, and only slightly lower strengths were observed at 40mm and 50mm gauge lengths (Supplementary Fig. 11). These results indicate a long length for the present CNTs. Efforts to track the CNT length in SEM led to no clear results. Instead, the ends of CNTs, wherever they were found, appeared to be bonded with the ends or side walls of other CNTs through the graphitic walls on the Fe catalysts. This suggests the joining of different CNTs to have a long length, a unique process taking place when the CNT cylinder is formed and is continuously blown out from the reactor. An increase in CNT length not only enhances the inter-tube contact area and thus the effective load transfer between nanotubes but also reduces the number of tube-ends per unit length of the fiber, both of which result in a higher tensile strength.

Mechanistic analysis of strengthening

An ultimate situation is that when the CNTs are so long that the inter-tube shear force becomes large enough to break individual CNTs rather than just induce inter-tube sliding. Consider a ribbon with a width of w and a thickness of t, which consists of n layers of parallel and closely packed CNTs with n = t/d, where d is the diameter of CNTs (Fig. 8). A tensile stress is applied on the ends of the ribbon, resulting in failure when the fibrous elements have slid out of contact or even broken at a high stress. Since the shear strength between adjacent layers τ_f is known from the above overlapping tensile testing, determination of the failure strength of the fiber-like ribbon only requires an estimate of the total area of shear failure. If we define a reference plane perpendicular to the axis, which is at the center of the fracture, then each CNT will fail on the side of the fracture plane where it is shorter. Hence, the mean sliding length of each CNT will be L/4, where L is the length of the CNT. The average sliding area per interlayer shearing will then be w×L/4.

Fig. 8. Schematics of a mechanistic model for carbon nanotube (CNT) sliding and breaking.

a Ribbon with n layers of closely packed CNTs. W: width, t: thickness, d: CNT diameter. White dots in a blue matrix represent omitted CNTs. b Uniform arrangement of CNTs with a length L. The average sliding length is L/4 at fracturing.

From further considering that the total force is n times of that of each shearing interlayer (τ_f) and the total cross-sectional area of the ribbon is w×t, the fracture strength of the ribbon σ is thus

$$\sigma =\left(L/d\right)\left({\tau }_{\mathrm{f}}/4\right)$$ 1

In order to break individual CNTs with a strength of σ_c, σ should be at least =0.7 σ_c, where 0.7 is a factor introduced with the consideration of the effective cross-sectional area occupied by closely packed CNTs. That is

$$L/d=2.8{\sigma }_{\mathrm{c}}/{\tau }_{\mathrm{f}}$$ 2

Using τ_f = 0.035 MPa (Fig. 7g) and d = 5 nm (Supplementary Fig. 2j), the length of CNTs must be larger than 12 mm and 20 mm for breaking individual CNTs with strengths of 30 GPa and 50 GPa, respectively. Both the tensile testing of the aligned sample at different gauge lengths and the overlapped ones with different overlapping lengths suggest a large CNT length over 20 mm. Thus, the additional origin of the high strength observed in this study is the large CNT length and the consequent large inter-tube shear force. The inter-tube shear force was so large that individual CNTs could be broken. This can be manifested from the fracture surface, showing the breaking of flattened CNT bundles (Fig. 6i, j). It should be noted that flattened CNT bundles were observed only at some locations on fracture surfaces but never in the sample without tensile deformation. The primary reason for the measured high strength is the overall well-aligned and densified CNT structure, which had a high load-bearing capability. The load-bearing capability at some locations might be so high that the CNT bundles could be flattened and even broken up.

To examine the strong inter-tube strength and load bearing capability at a much larger scale, a 75 cm long rope was made by slightly twisting a film with a width equivalent to about 50 pieces of the thin sample, and hanged with a total weight of 20 kg (Fig. 9 and Supplementary Fig. 12). Since the film was thin and wide (~10 cm), slight twisting was applied so that it could be fastened with the weight at one end and the upper supporting metal bar at the other. Since the twisting was slight, it didn’t provide additional, at least not large, load-bearing capacity. Loading and unloading for a few times or loading for a period of time longer than one week (Supplementary Video 5) showed no breaking. These observations indicate that the high load-bearing capability of the light CNT material can be extended to very large lengths and remain stable for a very long time. If the inter-tube strength were not large, easy sliding would take place, leading to shear breaking, soon after the application of the large load.

Fig. 9. Hanging demonstration.

The carbon nanotube rope with a length of 75 cm was hanged with 6 weights amounting 20 kg in total.

The completeness, validity, and limitations of the model based on Eq. 1 need some discussion. This model estimates CNT fiber strength based on experimentally measured interlayer shear strength, centering on interfacial debonding as the dominant failure mechanism. Mechanistically, it captures the key feature of CNT assemblies: intrinsic CNT tensile strength far exceeds van‑der‑Waals‑governed interlayer shear strength, making interfacial sliding and debonding the rate‑determining failure mode for most low‑to‑medium‑density fibers. Built on classic shear theory and using directly measured shear strength rather than idealized parameters, the model ensures physical consistency and avoids theoretical uncertainty. Experimentally, fractographic observations (bundle pull‑out, delamination) (Fig. 6) confirm interfacial failure, and interlayer shear strength correlates well with macroscopic tensile strength, supporting its predictive validity. Limitations exist for highly densified/cross‑linked fibers (where CNT rupture prevails) and extreme loading conditions, defining its applicable scope. Overall, the model is mechanistically complete and experimentally valid within its intended range, offering a reliable tool for strength estimation and performance optimization of CNT fibers.

Mechanistic analysis of reduction of elongation at break

During the treatment of pre‑stretching and scouring, the axial tensile force as well as scouring-induced shearing induce sliding, straightening, and directional alignment of CNT bundles, internal pores of the fiber are compacted, the CNT network becomes more regular, and the degree of orientation is significantly enhanced. All these changes can be visualized clearly from the set of SEM images of the samples that have been treated by different degrees of stretching and scouring (Fig. 3 and Supplementary Fig. 4). Elongation at break quantifies the total plastic deformability prior to fracture, which is governed collectively by CNT deformation modes, inter‑bundle sliding capacity, and spatial distribution of micro‑defects. The treatment of pre‑stretching and scouring modifies these key structural features and consequently lowers the elongation at break for the following reasons:

Exhaustion of slidable and relaxable deformation space

A substantial fraction of the elongation observed in as‑prepared disordered fibers (R-CNTF) (Fig. 4f, 10.2%) originates from extrinsic plastic deformation mechanisms, including the straightening of curled CNTs, relative sliding between bundles, and void compaction (analogous to untangling and straightening a disordered fibrous mass). The treatment exhausts such contributions in advance: curled CNTs are straightened, bundles are pre‑aligned and pre‑slid, and voids are densified (Fig. 3a–d). In subsequent tensile testing, no additional relaxation or sliding remains available, and deformation is restricted to intrinsic axial elastic stretching of individual CNTs. Fracture initiates once the load exceeds the ultimate strength of nanotubes or inter‑bundle interfaces, leading to a pronounced reduction in macroscopic elongation at break (Fig. 4f, DS-CNTF, DS-THINF, 4.2-6.0%).

Shift from multi‑modal to uniaxial axial deformation

In disordered fibers, CNTs undergo multiple deformation modes under tension, such as bending, torsion, lateral sliding, bundle disentanglement, and local buckling. These pathways facilitate energy dissipation and stress redistribution, yielding high macroscopic deformability. But in highly aligned fibers, CNTs are nearly perfectly oriented along the axial direction, which suppresses bending, torsion, and lateral motion; deformation is limited to pure axial tension. Given that CNTs are quasi‑one‑dimensional materials with inherently high modulus and low fracture strain (individual CNTs exhibit an elongation at break of approximately 5–10%), increasing orientation causes the fiber‑scale mechanical response to converge toward the intrinsic behavior of isolated nanotubes, resulting in decreased elongation at break.

Enhanced interfacial adhesion and intensified stress concentration

Pre‑stretching and scouring compact CNT bundles, enlarges van der Waals contact areas, and strengthens inter‑bundle adhesion, thereby restricting relative sliding. In disordered microstructures, inter‑bundle sliding serves as a stress‑buffering mechanism; in highly aligned, strongly bonded networks, stress propagates rapidly without sliding‑mediated dissipation. Local nanotube fracture induces instantaneous stress concentration, which triggers cascading failure and premature rupture. Concurrent with densification, pre‑stretching and scouring may cause micro‑defects (e.g., nanotube ends, broken segments, and inter‑bundle gaps) to distribute preferentially along the fiber axis (Fig. 3d, elongate voids parallel to aligned CNTs). Under high orientation, these defects act as stress‑concentration sites, initiating early fracture and further reducing elongation at break.

Transition from flexible network to rigid axial structure

As‑prepared disordered CNT fibers exhibit a flexible network‑like architecture, analogous to a spring, enabling large‑scale deformation through network reconfiguration. The treatment of pre‑stretching and scouring transforms this into a rigid, continuous axial structure comparable to a taut steel cable, which raises axial stiffness and diminishes plastic deformability. This trend conforms to a general materials‑mechanics principle: increased structural orientation correlates with elevated modulus and tensile strength, but reduced elongation at break (a behavior analogous to carbon fibers, which achieve ultra‑high orientation, high strength, and extremely low elongation after the pre‑oxidation and carbonization of precursor).

However, it should be noted that the reduction in elongation at break is not always monotonic with pre‑stretching magnitude. Low pre‑stretching ratios (<5 %) optimize orientation and densify voids while maintaining considerable ductility, yielding moderate strength enhancement with limited loss in elongation. Conversely, high pre‑stretching ratios (>10 %) induce excessive nanotube straightening, strong interfacial bonding, and full exhaustion of deformable space, leading to a drastic drop in elongation at break. To achieve a favorable balance between alignment and elongation at break, several strategies can be implemented, including CNT surface functionalization (to introduce flexible interlayers), fabrication of hierarchically oriented architectures, and incorporation of flexible polymeric phases, which collectively preserve inter‑bundle sliding capacity and mitigate stress concentration.

Mechanistic analysis of the improvement of conductivities

Electrical and thermal conductivities of CNT fibers are determined by the intrinsic conductivities of individual CNTs for electrons and phonons, respectively. But they are mainly governed by the inter‑tube resistance to conduction and the continuity of the conductive network. The improvement of these conductivities of the present CNT fibers is thus closely related to the improvement of CNT alignment and packing density. With enhanced alignment, CNTs are directionally arranged, enabling long‑range ballistic electron or phonon transport along the CNT axis and reducing transport conduction paths and scattering losses. Meanwhile, the contact mode between CNTs changes from point contact to parallel surface contact, which enlarges the contact area and significantly lowers the interfacial contact resistance or thermal barrier. On the other hand, increased packing density reduces the inter‑CNT spacing and void fraction, lowers the electron tunneling barrier or enhances phonon coupling between adjacent tubes, eliminates conductive islands caused by insulating voids, and forms a continuous directional percolation network. The synergistic optimization of alignment and packing enhances the intrinsic ballistic conduction of CNTs and inter‑tube tunneling conduction or phonon coupling, improves long‑range conductive pathways, reduces the total resistance, and thereby remarkably boosts the macroscopic electrical or thermal conductivity of the fibers.

In conclusion, to tackle the longstanding issue of CNT misalignment associated with macroscopic CNT fibers, a fluidics approach was demonstrated. This was done by spraying ethanol onto fibers at high velocity to induce axial shearing of CNT individuals at the microscopic scale and compaction of the fiber at the macroscopic scale. Experimental results showed that both the CNT alignment and packing density could be improved significantly by continuous fluidic flow-induced shearing and compaction. Consequently, CNT fibers achieved high specific and absolute tensile strengths of 7.5 × 10⁶N⋅m⋅kg⁻¹ (or 7.5 N⋅tex⁻¹) and 12.5 GPa, respectively, approaching the lower limit for individual CNTs, together with a high Young’s modulus of 370 GPa, electrical conductivity of ~3 × 10⁶ S⋅m⁻¹, and thermal conductivity of ~968 W⋅m⁻¹⋅K⁻¹. Mechanistic study showed that good alignment, dense packing, long length are key factors for significantly enhancing the contact area and interfacial interaction between CNTs. As a result of the refinement of these CNT structures, the shear strength between CNTs was enhanced to the level for the basal planes in single crystal graphite and inter-tube sliding within multiwalled CNTs. It is such inter-tube strengthening that led to the observed high mechanical and physical properties. Therefore, our study provides new techniques for developing CNT materials with high performances for high-end uses, especially in weight-sensitive applications demanding combined electrical and mechanical functionalities.

Methods

Fabrication of CNT films

CNTs were synthesized by pyrolysis of a solution consisting of a liquid feedstock of carbon source (ethanol) with dissolved ferrocene and thiophene (~99.99% purity). They self-assembled to a hollow cylinder in a tube reactor, and the cylinder was introduced and condensed on a winding drum operated in an open-air environment. Ribbon-like fibers 2 mm wide were cut from the wound film. The obtained fibers and films were heat-treated at 350 °C for 1 h to remove the carbon shell coated on iron particles and the amorphous carbon on CNTs. Then they were soaked in 4 M H₂SO₄ solution for 6 h for dissolution of iron particles. Finally, they were cleaned with ethanol (99.9 purity) and distilled water, and dried in a vacuum oven at 100 °C overnight.

Scouring and stretching processes

The bobbin with CNT fibers was placed on the fiber-supplying system. First, the fiber was introduced onto two guide rails, ensuring that the fiber was suspended vertically between the two guide rails. A peristaltic pump was used to spray ethanol solution on the fiber. Finally, the fiber was winded up at the fiber-collecting system (Fig. 2a). The fiber supplying rate V₁ and collecting rate V₂ were set to result in 0% or 15% stretching on the entire fiber. The scouring and stretching processes involved applying a scouring process to improve the arrangement of CNT with concurrent gradual stretching to 15% strain (Fig. 2b). The stretching rate and strain were controlled by the rotating motor, and the scouring velocity was determined by the peristaltic pump flow and the diameter of the latex tube. The scouring velocity and time were varied, and their combined effects were described in terms of scouring distance, which is the scouring velocity multiplied by scouring time. The total scouring time of a segment of CNT fiber varied from seconds to minutes, depending on the scouring velocity applied. Similarly, a raw film was loaded onto a stretching machine fixture to scour and stretch the CNTs. The subsequent treatment was consistent with the above, and the obtained film was denoted as DS-THINF (Fig. 2c).

Mechanical testing

The tensile testing was conducted using a fiber tensile tester. The tensile tester was equipped with a loading system for measuring forces up to 15 N with a precision of 0.0001 N. The digital output of the load cell was routinely calibrated by hanging different known weights to ensure accurate force measurements, and the displacement of the crosshead was recorded by a displacement transducer. 2 mm wide strips were cut from CNT films, or short segments of ~5 cm long were cut from long fibers and used for tensile testing. Such samples were clamped directly to the tester at both ends after the tester was calibrated. After the upper part of the sample had been clamped, a small weight (0.5 g) was applied on the lower end to ensure that the sample was taut. Then, the lower part of the sample was clamped at the position to have a desired gauge length (GL).

The tensile testing was usually performed at a displacement rate of 20 mm min⁻¹ and a GL of 10 mm, which gave rise to an engineering strain rate of ~3.33×10⁻² s⁻¹. To probe data reliability and microstructural uniformity, 3-5 samples were cut from the different locations of the same CNT material (usually tens of centimeters apart) and tested under the same conditions. In order to study the length dependence of the fracture load, samples of the same group were also tested at other GLs of 2 mm, 20 mm, 30 mm, 40 mm, and 50 mm. In these cases, the displacement rate was adjusted to match the strain rate used in testing at 10 mm GL.

Tensile strength was calculated from the measured load and cross-sectional area of the sample. Since the cross-section of the fiber before rolling was usually irregular, the area of such a cross-section was calculated from its SEM images using ImageJ software. Specific strength was determined by dividing the maximum tensile load by the linear density of the sample.

Determination of density

The linear density of a sample was routinely measured by weighing the fiber or strip of meters long using an electronic microbalance. Such obtained linear density was expressed as the mass per unit length, g⋅m⁻¹, which was converted to g⋅km⁻¹ (tex) or kg⋅m⁻¹ (common SI unit) for practical uses. In addition, a vibroscopic method, based on a sample of 5 cm long, was also used to determine the linear density of the sample, but usually for checking the uniformity of the linear density along the length of the sample or measuring the linear density of the sample to be tested in tension.

The volume density of a fiber was calculated from its linear density divided by the cross-sectional area measured from SEM. Therefore, such an obtained density represented an average density of the fiber of meters long. The possible variation in the actual density along the fiber length originates from the possible variation in the cross-sectional area. And the variation of the cross-sectional area was ultimately reflected in the variation of the maximum tensile load that could be read from tensile load-strain curves.

The densities for the pristine, scoured, and rolled samples were all determined. The rolling was carried out to further densify the scoured sample using a rolling press. A two-roller configuration was adopted, and the gap between the two rollers was set at 0.09 mm or less, depending on the progression of rolling-induced densification. In order to prevent direct contact between the fragile CNT sample and the hard steel roller, an aluminum foil was placed on the top and bottom of the sample. Experimentation revealed the importance of such Al buffering for non-destructive mechanical rolling. For routine operation, the sample was rolled at a speed of 1.53 m⋅min⁻¹ one time. Using a higher rolling speed could easily damage the sample.

Measurements of transport properties

The square electrical resistance of the fiber or film, R_S, was measured using a four-point probes meter. In doing this, the sample was laid firmly on a glass substrate with a negligible electrical conductivity. R_S was measured at least 10 times at different locations along the axial direction of the sample, and an average value is reported. The electrical conductivity was calculated from 1/(R_S×d) with the consideration of the sample thickness d.

The in-plane thermal diffusivity (α, mm² ⋅s⁻¹) of the film after scouring (using a circular sample with a diameter of 25.4 mm) was measured by a laser flash thermal conductivity meter at 25 °C (LFA467, Netzsch, a laser voltage of 220 V, a pulse width of 50 μs). The measurement was made along the orientation of CNT alignment for 3-5 times, and the average value is used. The thermal conductivity (κ) was then calculated from the following equation

$$\kappa =\alpha \times C_{\mathrm{p}}\times \rho$$ 3

where C_p is the heat capacity (J⋅g⁻¹⋅K⁻¹), and ρ is the density of the sample measured (g⋅cm⁻³). Detailed procedures for the measurement and data reduction are presented in Supplementary Fig. 13.

Small angle X-ray scattering

CNT fibers were subjected to small-angle X-ray scattering measurements (SAXS, Xeuss 2.0, France), and two-dimensional diffractograms were recorded. The Herman’s orientation factor was used to quantify the degree of orientation relative to the central axis using the equation given below

$$f=\frac{3⟨{\cos }^2\phi ⟩-1}{2}$$ 4

where φ is the angle at the azimuthal plot, cos²φ was calculated as follows:

$${\cos }^2\phi =\frac{\underset{0}{\stackrel{\pi /2}{\int }}I\left(\phi \right){\cos }^2\phi \sin \phi \mathrm{d}\phi }{\underset{0}{\stackrel{\pi /2}{\int }}I\left(\phi \right)\sin \phi \mathrm{d}\phi }$$ 5

where I(φ) is the intensity at the azimuthal angle (φ). f was determined based on Lorentzian fitting. Alternatively, f could also be determined in a similar way by using Gaussian fitting for data processing, but often leading to a higher value. All the f values reported in this study refer to those from Lorentzian fitting unless otherwise noted.

In-situ stretching Raman spectroscopy

In situ Raman spectroscopy was used to detect the micro-mechanical information of fibers during deformation. Raman scattering of the sample was excited by a laser with a wavelength of 785 nm, with the polarization direction parallel to the central axis of the fiber. During the testing process, the fiber was fixed on a manual stretching table to detect the Raman signal. Common Raman characteristics of CNTs include the disorder-induced (D) band, the graphite (G) band, and its overtone (G’) band. Due to the sensitivity of the G’ band to external stress, its Raman shift was analyzed to obtain the strain distribution in the fiber.

Other structural characterizations

The microstructure and morphology of the sample were characterized using transmission electron microscopy (TEM, JEOL JEM-2100, 200 kV) and scanning electron microscopy (SEM, S4800). The cross sections of CNT fibers and films were prepared by a focused ion beam (FIB, GAlA3). The graphitic degrees of the fibers and films were investigated by Raman spectroscopy (Raman, Bruker Senterra R200-L), and the alignment degree along the axial direction was evaluated by polarized Raman spectroscopy. For the polarized Raman spectroscopy analysis, the intensity of G peak was measured by obtaining the incident polarized laser beams parallel and perpendicular to the fiber axis, respectively (I_G‖ and I_G⊥). When the axial direction of CNTs was parallel to the polarization direction of the exciting laser, stronger G-band scattering intensity was obtained. The higher the I_G‖/ I_G⊥ value, the higher the orientation degree of the CNT fiber and film. The crystalline structure was also analyzed by X-ray diffraction (XRD, Rigaku D/max 2550VB/PC). The surface elemental composition and existing forms were analyzed by X-ray photoelectron spectroscopy (XPS, ESCALAB 250). The contents of Fe and CNT in the fiber were quantitatively analyzed by thermogravimetric analysis (TGA, PerkinElmer TGA 8000) from room temperature to 1000 °C in an air atmosphere with a heating rate of 10 °C min⁻¹. The heat capacity of CNT films was measured by a differential scanning calorimeter (DSC 204 F1, NETZSCH).

Author contributions

J.N.W. conceived and supervised the research, designed the experiments, and performed some of them. Y.T.C., H.Zhang, and H.Zhan performed the experiments and characterization. J.N.W. and Y.T.C. analyzed the data and prepared the manuscript. Q.X.W., J.C., X.L., J.F.Z., and K.F.L. participated in experimental work and data analysis. All authors reviewed the manuscript.

Peer review

Peer review information

Nature Communications thanks Xide Li and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

Jian Nong Wang was funded by the National Key R&D Program of China (2022YFA1203303) and National Natural Science Foundation of China (52172089).

Data availability

The data generated in this study are provided in this published article and its Supplementary Information file. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-75150-1.