Formation Mechanism of BN Flakes on MWCNTs

Abstract

A novel heterostructured hexagonal‐boron nitride (h‐BN) flake‐coating on multi‐wall carbon nanotubes (MWCNT/BN) is reported and synthesized by chemical vapor deposition (CVD). Comprehensive characterization using X‐ray photoelectron spectroscopy (XPS) and scanning transmission electron microscopy (STEM), combined with electron energy loss spectroscopy (EELS), revealed the atomic structure and growth mechanism, which is further validated by molecular dynamics simulations. The resulting MWCNT/BN structure comprises three distinct layers: an inner carbon nanotube (CNT) core, coaxial BN nanotubes (BNNTs) surrounding the CNT core, and outer BN flakes extending from the BNNTs. We propose that BN layers first form coaxial BNNTs on the CNT surface; as deposition proceeds, BN accumulation generate in‐plane and out‐of‐plane compressive stresses in the h‐BN layers. When these stresses exceed a critical threshold, local buckling or cracking occurs, BN flakes emerge and grow further. This work elucidates, for the first time, the formation mechanism of BN nanoflakes on MWCNTs and confirms that the structure is a van der Waals heterostructure. The approach also offers a new route for synthesizing coaxial MWCNT@BN with only a few h‐BN layers. Notably, the BN flake coatings provide efficient phonon transport pathways and a large surface area, making this heterostructure highly promising for applications in thermal dissipation.

This work presents a novel heterostructured h‐BN flake‐coating on multi‐wall carbon nanotubes synthesized via CVD. It details structural characterization supported by molecular dynamics simulations. The mechanism of BN flake formation, stress‐induced buckling, and van der Waals heterostructure design is discussed, highlighting applications in thermal dissipation and advanced material engineering.

1. Introduction

Flake or sheet‐coated carbon nanotube, including single, double and multi‐wall carbon nanotubes (SWCNT, DWCNT, MWCNT respectively), structures are widely explored because they integrate multiple functionalities within a single structure. These structures leverage the advantages of carbon nanotubes (CNT), such as exceptional thermal and electrical conductivity, high mechanical strength, and low density, while also incorporating the benefits of nanosheets (e.g., MnO₂, graphene, and hexagonal boron nitride (‐h‐BN)). This combination results in a significantly increased surface area, enhancing thermal dissipation and electrochemical activity. Such attributes are crucial for applications in detectors, superconductors, supercapacitors, heat management materials, and reinforcement in a polymer matrix.^{[ 1} , ^{2 ]} For example, deposition of MnO₂ sheets and graphene onto CNT surfaces is used in flexible superconductor and supercapacitor applications because of the increased surface area, ion accessibility, and numerous edge planes of MnO₂ sheets and graphene.^{[ 3} , ⁴ , ^{5 ]} Similarly, graphene sheet‐coated CNTs are used in high‐performance electrodes, such as graphene‐wrapped MWCNT@MoS₂ electrodes with superior capacitance and cycle‐life stability.^{[ 6 ]} Coating MWCNTs with graphene has also been shown to improve optoelectronic, gas sensing properties, and thermal conductivity.^{[ 7} , ^{8 ]} Additionally, integrating BN nanosheets with CNTs has enhanced super‐hydrophobic properties, facilitating successful oil/water separation.^{[ 9 ]} Leaf‑shaped CNT@BNNS covalent heterojunctions and dendritic MXene@CNT hybrids show how tailored interfaces deliver recyclable films with high strength and ultralow gas permeability and nanocomposites with simultaneous gains in strength, toughness, barrier properties, and recyclability.^{[ 10} , ^{11 ]} Different routes to producing such structures include exfoliating and blending individual materials to combining them together,^{[ 12} , ^{13 ]} direct synthesis by chemical vapor deposition (CVD)^{[ 14} , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ^{22 ]} and creating bonds among different components by hydrothermal methods.^{[ 1} , ^{2 ]} However, the growth mechanism and physical interactions between the flakes and CNTs remain largely unknown due to the lack of direct, high‐resolution observations of the dynamic growth process and precise modelling methods.

Several hypothetical growth mechanisms for graphene‐coated CNTs (G‐CNT) have been proposed. Muangrat et al. synthesized graphene sheets grown on double‐wall CNT (DWCNT) bundles for use as an ethanol vapor detector because the high edge density of graphene provides high charge density and reactivity.^{[ 23 ]} Muangrat et al. further explained the growth mechanism, suggesting that carbon atoms initially deposit on the surface of CNTs, forming a carbon layer, upon which the graphene sheets subsequently grow. Parker et al. employed a one‐step method to grow graphenated‐CNT (G‐CNT) using microwave plasma‐enhanced CVD.^{[ 24 ]} Their study claims that the outer walls of CNTs grow faster than the inner walls, resulting in excessive stress that causes localized buckling. This leads to the protrusion of a few layers of carbon, upon which further carbon growth occurs at the foliated graphene tips. Another investigation by Rout et al. examined the growth mechanism of graphene‐coated CNT arrays.^{[ 25 ]} They demonstrated that exposing a CNT array to H₂ flow, plasma power, and high temperatures (900 °C) creates defects on the outer layers and tips of MWCNTs, which act as nucleation centres for the growth of graphene layers. However, the proposed mechanisms mentioned above are largely hypotheses and lack direct solid evidence. Further research is needed to provide more evidence for these theories.

The interactions between the nanosheets and CNT components are essential in determining the overall thermal and electrical properties of hierarchical structures. Interface resistance between these materials can significantly impact thermal and electrical conductivities. For example, Li et al. reported that the formation of the covalent bonding between CNTs and BN nanosheets (BNNSs) reduced the thermal contact resistance by one order of magnitude while the volume resistivity increased by two order of magnitude.^{[ 8 ]} In the case of CNT@h‐BN, van der Waals interactions enable efficient thermal conduction while preserving electrical insulation.^{[ 26 ]} Understanding the growth mechanism and interactions within flake‐coated CNT structures is crucial for effectively integrating these components in future applications of hierarchical structure.

Here, we introduce a MWCNT/h‐BN composed architecture, where CNT cores are enveloped by boron nitride nanotubes and h‐BN flakes. One previous study reported this structure as free‐standing BNNSs deposited on CNTs.^{[ 9 ]} However, our research has studied this structure in depth and confirmed the integrated structure. Our comprehensive experimental investigations help elucidate the structural attributes, the growth mechanism, and the interfacial interactions of this hybrid material. Specifically, we confirm that flakes initially grow on the CNT surface as coaxial multi‐wall nanotubes, followed by the deformation of the outermost layer into larger flakes. The simulations suggest the deformation of BNNTs happens at the surface defects and the extent of such deformations is proportional to the radius of the inner tube. Our research demonstrates that the flake‐coated CNT structure synthesized via CVD resolves issues related to the loose connections between CNTs and flakes synthesized by blending and hydrothermal approaches. Additionally, the individual properties of the CNT and h‐BN components are preserved. Unlike ensemble methods which create chemical bonds to hold CNT and BN together, our approach relies on the van der Waals forces, allowing CNT and BN to remain independent. By understanding the growth mechanism and chemical nature of this material, we find that it offers substantial surface area, elevated thermal stability, and significant potential for applications such as heat dissipation, mechanical reinforcement, and electronics. In addition, our method can be a potential new route to synthesize coaxial MWCNT@BN structures with a few BNNT walls. This understanding aids in tailoring experimental parameters to achieve diverse morphologies, catering to various applications.

2. Results and Discussion

We used MWCNTs as the starting material and successfully synthesized BN flake‐coated MWCNTs using CVD at a temperature of 1300 °C (Figure S1, Supporting Information), as shown in Figure 1A,B. For comparison, products obtained at 1200 °C and 1400 °C are shown in Figure S2 (Supporting Information). At 1200 °C, the average CNT/BN diameter increases from ≈40 nm (pure CNTs) to ≈200 nm. Raising the temperature to 1300 °C produces thicker BN flakes, with diameters of ≈300 nm. At 1400 °C, the BN flakes become noticeably thinner and shorter, reducing the average diameter to ∼200 nm, indicating that excessively high temperatures may suppress flake growth. Numerous h‐BN flakes surround the MWCNTs, creating a rough surface structure. Both Raman and Fourier‐transform infrared Spectroscopy (FT‐IR) spectra (Figure S2, Supporting Information) present the formation of BN flakes. Figure 1C presents the Raman spectra of pure MWCNT and MWCNT/BN. In Raman spectroscopy, the first‐order graphene identifying peaks appear around wavenumber ν = 1349 cm⁻¹ (D peak) and 1580 cm⁻¹ (G peak), as seen in the spectrum.^{[ 27 ]} The intensity ratio of the D‐band and G‐band (I_D/I_G) is used to quantify the quality of the synthesized MWCNTs. After deconvoluting the peaks, I_D/I_G was calculated to be 0.33, indicating that CNTs have defects, disorder, and potential functional groups. However, since the G peak remains dominant, the overall quality of the CNTs is still considered good. The frequency of the G peak of the CNT remains unchanged at 1579 cm⁻¹ before and after h‐BN coating. Theoretically, the dominant h‐BN peak is at 1366 cm⁻¹, which corresponds to its in‐plane vibrational mode.^{[ 28} , ²⁹ , ^{30 ]} However, in Figure 1C, the h‐BN peak appears at 1360 cm⁻¹, lower than expected. This shift occurs due to the overlap of the CNT's D peak at 1349 cm⁻¹ with the BN peak at 1365 cm⁻¹, resulting in a lower combined peak.

Figure 1.

SEM images of (A) MWCNT and (B) MWCNT/BN. C) Raman spectra of MWCNTs and MWCNT/BN. Schematic drawings (D) VA‐MWCNT. E) coaxial MWCNT@BN. F) BN flakes coated veMWCNT. G) TEM image of MWCNT/BN. BN was painted with red and MWCNT core was painted with green. H) STEM image of MWCNT/BN and (I) zoomed‐in image of the interface of nanotube and the flakes using threshold filtering. J) schematic drawing of (I).

The two‐step schematic growth process of h‐BN on the CNT is illustrated in Figure 1D–F. (D) shows the starting material, a pure MWCNTs. At the beginning of the BN coating process, a few‐walle BNNTs grow outside of CNTs, forming a coaxial CNT@BN structure (Figure 1E). Thereafter, BN flakes develop outside of the CNT@BN, thus forming the final CNT/BN heterostructure (Figure 1F). This two‐step mechanism will be discussed further in details with experimental and modelling data. To obtain detailed structural information, a transmission electron microscopy (TEM) image was obtained in Figure 1G. The h‐BN flakes are depicted as pink, while the inner MWCNT is light green. In this image, numerous h‐BN sheets grown outside of the MWCNT were identified. These h‐BN sheets were found to increase the overall diameter of the composite tube from ca. 40 to 300 nm. To understand the flake formation, high‐resolution scanning transmission electron microscopy (HR‐STEM) was carried out which shows how flakes grow outside of CNTs (Figure 1H). The sample we used for HR‐STEM was coated with h‐BN for 30 min because only a small quantity of h‐BN was formed to avoid overlaid h‐BN sheets as shown in (G). The interface of BN flakes and the coaxial tube was identified in a red square, which is further processed as a close‐up image shown in (I). We observed that BN walls had peeled off from the coaxial structure from the top of the image and some walls were bent. This image reveals that the flakes are not covalently bonded to the tube.

To further investigate the growth of BN structure on CNTs, a parametric study was conducted with varying synthesis durations of 15, 30, 45, and 90 min. The samples were then heated at a temperature of 900 °C to remove the CNT core, leaving only the h‐BN coating. The thermal gravimetry analysis (TGA) analysis of the burning process quantifies the amount of h‐BN coating synthesized at different durations, as shown in Figure S3 (Supporting Information). A revised version of Figure 2 , incorporating noted CNT and BN regions, is provided in Figure S4 (Supporting Information).

Figure 2.

Schematic drawings and ex situ TEM images of (A)‐(D) CNT/BN with 15, 30, 45, and 90 min synthesis duration, respectively, and (E)–(H) h‐BN coatings after burning CNT core of CNT/BN with different synthesis duration at 900 °C in air.

After a 15‐min synthesis, the CNT/BN composite retained a coaxial tube structure, as shown in the TEM image in Figure 2A. However, after oxidation, the h‐BN did not retain the tubular structure due to its thinness. The oxidation of the CNT core resulted in slight damage to the BN, causing a structural collapse and the agglomeration of small BN pieces, as shown in (E).

At a 30‐min synthesis duration, BN retained a tubular morphology after the inner CNT core was oxidized away. Both the inner and outer walls of the CNT/BN structure at this stage were continuous and clearly visible. Following the core's oxidation, the BN coating preserved the nanotube structure, as depicted in Figure 2F.

After a 45‐min synthesis, the outer wall of CNT/BN became rough, and BN flakes started to grow on the surface (Figure 2C). The inner diameter increased post‐oxidation, but BN maintained its tube‐like structures with flakes grown on the outermost layer (Figure 2G). This is similar to the structures observed in Figure 2D and (H), where the inner tube becomes more hollow after burning the CNT core, but the outer flake's structure remains. After a 90‐min synthesis duration, the CNT/BN structure exhibited with both denser and larger flakes.

To summarize, coaxial BNNTs initially grow outside of CNTs, followed by the growth of BN flakes on the outer surface of BNNT. As synthesis duration increases, the density and the size of the flakes increase. Therefore, this method can be used as a novel route to synthesize few‐layered BNNTs and BN flake‐coated BNNTs after the oxidation of the inner CNTs.

Figure 3A shows the structure of the 90‐min synthesized CNT/BN and the general distribution of carbon, boron, and nitrogen within it. The raw and fitted electron energy loss spectroscopy (EELS) scanning transmission electron microscopy spectra are shown in Figure S5 (Supporting Information), with the B‐K edge at 190 eV, the C‐K edge at 285 eV, and the N‐K edge at 400 eV. In Figure 3A(i), the STEM image of CNT/BN shows the dense flakes surrounding the CNTs. Figure 3A (ii)–(iv) present the EELS mappings corresponding to i) with the carbon mapping ii) highlighting the inner core of the CNT. The mappings in iii) and iv) display very similar boron and nitrogen distributions, outlining the outer shell and the flakes with empty cores, confirming the coatings are composed of boron and nitrogen. An overall view of a single CNT/BN heterostructure is presented in Figure S6 (Supporting Information).

Figure 3.

A) (i) STEM images of CNT/BN. EELS mapping over the area shown in Figure (A) for (ii) carbon, (iii) boron, and (iv) nitrogen. B)zoomed‐in STEM image of the edge of CNT/BN with EELS line scan shown by the blue line. C) EELS line scan profile showing the distribution of C, B, and N along the blue line. D) d‐spacing calculation of BN area shown in (C).

To examine the CNT and h‐BN interface, a scanning transmission electron microscopy (STEM) image of a sample coated with h‐BN for 30 min was taken (Figure 3B) to minimize additional BN flake growth that might blur STEM images due to overlapping flakes. A close‐up STEM image in Figure 3B reveals the parallel outer walls of the heterostructure, indicating coaxial BNNT growth on the CNTs. A zoomed‐in image of the red squared area in (B) was colorized for clearer visualization. To determine the elemental composition of this region, an EELS line scan was conducted along the blue line in Figure 3B, showing the distribution of carbon, nitrogen, and boron along this line in Figure 3C. The amount of carbon decreases while those of boron and nitrogen increase from left to right along the line scan, indicating the transition from a pure carbon nanotube to a coaxial boron nitride nanotube. Figure 3D illustrates the intensity change of the walls along the same blue line based on the zoomed‐in image in (B). Since (C) and (D) share the same x‐axis scale bar, the BN‐rich region in (D) was identified to range from 5.402 to 7.484 nm, a difference of 2.082 nm. Given that there are six walls within this 2.082 nm counted in (D) and further confirmed in (B), the d‐spacing of BNNT was calculated by dividing 2.082 nm by 6, resulting in 0.347 nm, corresponding to the (002) plane of h‐BN.^{[ 31 ]} We acknowledge the possibility that, at the CNT–BNNT interface, a portion of the CNT wall may have been converted into BN, as suggested by the EELS line scan in Figure 3C, which does not exhibit a sharp compositional boundary between C and BN. However, the measured interlayer spacing at the interface remains consistent with the theoretical value for van der Waals–bonded layers. This indicates that, even if a few residual carbon atoms are present within the innermost BNNT wall, the primary interaction governing the interface is still van der Waals in nature. To validate this possibility, future studies should employ in situ high‐resolution microscopy combined with elemental analysis.

X‐ray photoelectron spectroscopy (XPS) analysis was conducted to assess surface information on the CNT/BN composite. XPS survey spectra of CNT and CNT/BN are shown in Figure 4A (i) with N1s, C1s, and B1s at energies of 398.3, 285.1, and s190.9 eV, respectively. The intensities of the B1s and N1s peaks increase significantly after h‐BN coating. In fact, the atomic percentage of N and B is 95.6% in the CNT/BN with the atomic percentage of C dropped from 98.5% to 2.3% after h‐BN synthesis, which confirms the high yield of the BN coating. The existence of the O1s peak in the survey might result from a small number of impurities including the absorption of H₂O and O₂ on the surface of the sample.^{[ 32 ]} The corresponding high resolution XPS spectra were deconvoluted. In the C1s spectrum shown in Figure 4 (ii), there are four deconvoluted peaks, O─C═O, C═O, C─B, and C─C/C─H. Because of the surface‐sensitive nature of XPS, the small amount of carbon detected is attributed to the outer BN flake surface rather than the CNT–BN interface. This carbon may originate from CO₂ adsorption during air exposure and minor carbon incorporation into the BN lattice, as further supported by the corresponding B 1s and N 1s spectra. Additionally, a small number of carbon atoms are bonded with boron, which explains why the boron atomic percentage is slightly more than nitrogen by 4.5%. It has been observed that the atomic percentage of B is slightly more than N in the C@BN coaxial structure.^{[ 33 ]} This can potentially be explained by the fact that with a small amount of carbon in the system, carbon atoms prefer to form C‐B bonds because the defect energy of the boron site is higher than that of nitrogen.^{[ 34 ]}

Figure 4.

A) (i) XPS survey spectra of CNT and CNT/BN from 800 to 0 eV showing O1s, N1s, C1s, and B1s. CNT/BN's high resolution XPS spectra and peak fitting of (ii) C1s and 4 deconvoluted peaks of C─B, C═O, C─C/C─H, and O─C═O, (iii) B1s and 4 deconvoluted peaks of B─N, B‐N₂O, B‐NO₂, and B‐N₂C, (iv) N1s and 3 deconvoluted peaks of N─H. N─C, and N─B. B) XRD spectrum of (i) CNT and (ii) CNT/BN over 2θ from 20° to 60°.

The high‐resolution B1s spectrum in Figure 4 (iii) shows four bonds: B‐N, B‐N₂O, B‐NO₂, and B‐N₂C. The deconvoluted spectrum of N1s is in Figure 4 (iv) which reveals three fitted peaks of N─B, N─H, and N─C. After taking C─ into account, the stoichiometry of B/N is 1.08, which agrees approximately with the theoretical stoichiometric component of h‐BN. The C─C/C─H peak corresponds to the sp² C─C peak in the hexagonal atomic lattice of MWCNTs. Spectra in (iii) and (iv) confirm the majority presence of the B─N peak in the CNT/BN because the peak intensity reflects the amount of the corresponding element. Therefore, XPS spectra confirm the formation and high purity of BN in CNT/BN.

Figure 4B (i),(ii) show the XRD spectra of pure CNT and CNT/BN, respectively. The XRD spectrum of CNT exhibits peaks at the Bragg angles of (002), (100), and (004).^{[ 35 ]} In comparison, the CNT/BN spectrum shows three peaks at (002), (100), and (101). Due to the similar diffraction patterns of CNT and h‐BN, interpreting CNT and h‐BN heterostructures has been challenging. However, comparing the spectra of CNT and CNT/BN helps to understand the linity changes during the synthesis. The (002) peak of h‐BN coated CNT structures is typically more asymmetric than that of pure CNT.^{[ 36} , ^{37 ]} Because the (002) shifts of CNT and h‐BN are slightly different from each other, combining these two peaks results in an asymmetric structure. as confirmed in the spectrum of CNT/BN in Figure 4B.^{[ 37 ]} Additionally, the appearance of higher resolved peaks at (110), and (101) as well as high‐resolution TEM image (Figure 3A) suggests high crystallinity of h‐BN.^{[ 38 ]}

$$d_{002}=\frac{n\lambda }{2\sin \theta },n=1$$ (1)

Based on Bragg's Law (Equation 1) where the X‐ray wavelength λ = 0.154 nm for a Cu K‐alpha source, and θ as the Bragg angle. the interlayer d‐spacing of CNT and CNT/BN is approximately 0.341 nm, which is close to the d‐spacing of 0.347 nm calculated from the STEM image in Figure 3B–D. The d‐spacing is slightly higher than the theoretical value of the h‐BN d‐spacing which ranges between 0.33 and 0.34 nm,^{[ 39 ]} likely due to a turbostratic structure and wall curvature.^{[ 40 ]}

Molecular dynamics (MD) simulations were used to examine the growth of boron nitride nanotubes on the surface of the carbon nanotubes. In this study, MD simulations were conducted using LAMMPS.^{[ 41 ]} Further simulations were performed on the analogous CNT/BNNT layers as described below. The initial C─C and B─N bond lengths were 1.421 and 1.470 Å, respectively. The initial separation between adjacent nanotubes (in the radial direction) or between nanosheet layers was 3.33 Å. The Tersoff potential was employed to describe the atomistic interactions in each layer or nanotube,^{[ 42 ]} while the interlayer van der Waals interactions between carbon and BN were modelled using a hybrid potential that combined the Tersoff potential and the h‐BN interlayer potential.^{[ 43 ]}

The experimental results discussed in Figure 2 indicate that coaxial multi‐wall BNNTs are initially synthesized on the outermost layer of multi‐wall CNTs. The bottom‐left of Figure 5A shows a typical starting configuration comprising a “core” multi‐wall CNT (here containing three‐layer CNTs, periodically repeated along the z axis direction) coated with a single‐layer BNNT. The BNNT has a finite length of 100 Å, a chiral vector of (90,0), and an initial radius R ₁ of 35.25 Å. A range of simulations were performed on CNT/BNNT composites with different radii, varying the chiral vector of the BNNT from (20,0) to (200,0). In this study, we propose that the nucleation of BN flakes result from the growth of BNNTs on the outer sidewall, which induces residual stress in the BNNT walls, leading to bending and buckling. This deformation facilitates the nucleation of BN flakes. A similar mechanism has been reported in analogous systems, such as graphene‐coated CNTs, where stress‐driven buckling of carbon layers leads to graphene sheet formation.^{[ 24 ]} In this case, BNNTs were subsequently compressed from both ends along the cylinder's major axis at a rate of 0.01 nm/ps by displacing by 0.01 Å every 1 ps (100 MD steps). The applied axial compression serves as a physical proxy for residual stress arising from lattice mismatch and non‐uniform BN deposition during CVD growth. In all cases, the system was maintained at a temperature of 300K using thermostats stimulated.

Figure 5.

A) The scale of the wrinkles, defined by their maximum height ΔR = (R₂ − R₁), in the CNT/BNNT composites is proportional to the curvature of the BNNT cross‐section (1/R₁). Here, R₁ is the initial BNNT radius, and R₂ is the maximum wrinkle height before cracking. B) The relationship between average input energy (energy per atom, in eV/atom) and time (ps) for CNT/BNNT composites with different proportions of atomic vacancies in the BNNT, from pristine BNNT to those with vacancies occupying up to 1/2 of the circumferential length. All BNNTs are compressed axially at 0.01 nm ps⁻¹. C) Scenario 1 – Compression of a layered BNNT/CNT composite (modelled as several BN nanosheets lying directly on carbon nanosheets) at the same rate shows three typical wrinkle morphologies: (i) Ripple (t = 20 ps), (ii) Wrinkle (t = 80 ps), and (iii) Standing collapsed wrinkle (t = 160 ps). D) Scenario 2 – (i) Compression of two separate BN nanosheet layers illustrates an alternative flake‐formation pathway: (ii) Sprout stage (t = 50 ps), BN layer begins to peel away from the underlying CNT surface; (iii) Flake grows further outward (t = 500 ps) as newly deposited BN and van der Waals interactions between layers drive continued separation. In both scenarios, the models are simplified representations of local regions of large‐diameter nanotubes, capturing the essential stacking, stress accumulation, and deformation features observed experimentally. In all cases, carbon, nitrogen, and boron atoms are depicted as yellow, red, and blue circles, respectively.

The compression of the boron nitride on the CNT surface leads to the formation of wrinkles in the BN surface. These wrinkles increase in size under compression until they reach a maximum radius R₂ (of approximately R₂ = 40.68 Å for the (90,0) BNNT highlighted in left‐bottom Figure 5A), at which point cracks began to appear at the peak of the largest wrinkle.

To describe the maximum spatial extent of the final wrinkles (i.e., prior to “cracking”) a simple metric, the difference (ΔR) in the radius of the wrinkle, R ₂, and the starting radius, R ₁, is calculated and is shown in Figure 5A. R ₂ is calculated by determining the average distance from the five outermost points of the wrinkles to the centre of mass of the nanotube. The main panel of the Figure 5A shows the maximum spatial extent of the wrinkle, ΔR, plotted against the reciprocal of the initial BNNT radius, R ₁, that is, the curvature of the circular cross‐section of the BNNT. Figure 5A indicates that, as the radius of the BNNT increases, the spatial extent of the formed wrinkles also increases, reaching a limit at ∆R∼6.4 Å.

The SEM results (Figure S8, Supporting Information) indicate that the multi‐wall CNTs have an average diameter of 37.2 ± 11.6 nm, while the average diameter of the coated tubes expands to approximately 255 nm, which means the scale of the BN flakes is significantly larger than the original diameter of the carbon nanotubes. This suggests that, after reaching their maximum size, wrinkles may lead to the formation of cracks, which then serve as nucleation points for flake formation. Rather than expanding into large‐scale structures, the wrinkles give rise to cracks where boron nitride continues to deposit and accumulate. This deposition and accumulation at the crack sites facilitate the formation of larger flakes through compression at these locations.

The broadened peaks in the XRD spectrum and the multiple deconvoluted peaks in the XPS spectrum (Figure 4) indicate that the coaxial CNTs and BNNTs show numerous defects. To investigate the potential role of defects, a series of simulations were performed using the same CNT/BNNT composites, in which different proportions of both B and N atoms were removed from a circular cross‐section of the BNNT. Figure 5B shows an example of the evolution of the total system energy as the compression proceeds for the pristine and three defective single layer BNNT at the outside of three layers co‐axial CNTs. The average input energy per atom is shown as the difference with respect to the initial total energy of the thermalised system divided by the numbers of the total atoms from each system. For the defective BNNTs the fraction of atoms removed from a circular cross‐section is 1/4, 1/3 and 1/2 respectively, with the width of the atomic vacancy ca. 2 Å. The inset to Figure 5B, for example, shows images of the CNTs/BNNT composite for 1/4 of the atoms removed. For the defective BNNTs the evolution of the energy on compression shows several stages. At short time (t<∼60 ps) the energy rises slowly for the defective systems compared with a stronger near‐linear rise for the pristine BNNT. For the defective BNNTs this short time regime corresponds to the defect region effectively closing on compression (point 1). At a relatively well‐defined “critical” time (t∼80 ps) the defective BNNTs show a sharp increase in energy, corresponding to the initial formation of wrinkles on the CNT surface (point 2). As time progresses further wrinkles are formed which coalesce to form a single wrinkle (around the location of the original defects) and finally cracks at the edge of initial atom vacancy (point 3) at time t∼100 ps. The key observation is that the timescale at which the wrinkles form for the pristine BNNT appears significantly longer, highlighting the potential role of the defects in facilitating wrinkle growth and subsequent flake formation. Furthermore, the energy required to generate new cracks on BNNTs with existing defects is much lower than pristine BNNT, as more input energy required with smaller initial atom vacancy.

It is hypothesized that when excess boron nitride is deposited on the nanotube surface, the resulting wrinkles reach their physical limits, leading to the formation of new cracks. Once these cracks emerge, the accumulated boron nitride exerts compressive stress on these cracked regions, driving further structural changes. In previous simulations, BN deposition has been observed to occur at the edges of existing BN structures, particularly at crack sites, serving as nucleation points.^{[ 44 ]} To better connect our simulation with experimental reality, we adopted a simplified but physically meaningful approach in which axial compression was applied to BNNTs to approximate the mutual compression generated by BN accumulation during growth. This compression process is considered a key mechanism in BN flake formation. Two possible scenarios are considered, as shown in Figure 5C–D. Scenario 1 represents stress‐induced buckling and cracking of BN layers, while Scenario 2 represents multi‐point BN nucleation on BNNT surfaces followed by compression and imperfect merging. The further growth of boron nitride in these two types of cracks may lead to compression and ultimately result in the formation of flake‐like structures. To explore these possibilities, we conducted simplified simulations for each scenario. As shown by the Figure S9A (Supporting Information), when the diameter of the BNNTs is greater than 0.6 nm, the difference in average per‐atom energy between the nanotube and the boron nitride nanosheet is less than 1%. This indicates that when the diameter of the nanotube is sufficiently large, meaning the curvature is sufficiently small, its surface can be approximated as a planar surface.

To highlight the potential mechanism of flake formation, a further series of simulations were performed. In these simulations, a BN/C composite nanosheet layer was compressed under conditions analogous to those described for the BNNT/CNT composites. Figure 5C illustrates the first scenario, where cracks along the nanotube's axial direction release the ring‐shaped stress generated around the wrinkle tips due to compression. This shows three snapshots from the compression process at t = 20 ps i), 160 ps ii), and 600 ps iii), respectively. These layers represent the limit of a BNNT/CNT composite with an effectively infinite radius. The panels of Figure 5C highlight (from left to right) the growth of a single wrinkle. The standing collapsed wrinkle in Figure 5C (iii) can be easily understood at a qualitative level as resulting from the competition between elastic bending and van der Waals binding. In the figure, we can clearly see that the compression‐induced wrinkle forms a bubble shape, and due to the presence of van der Waals forces, the middle parts converge. The ripple structure minimizes bending distortion. However, when there is a large amount of excess material, collapsing it to form bilayers provides additional van der Waals binding, though this comes at the cost of progressively increased bending.^{[ 24 ]} In Figure 5C (iii), we can see that due to the influence of van der Waals forces, a folded wrinkle structure ultimately forms.

To reveal the potential mechanism of flake formation following the cracking of wrinkles, Figure 5D presents the results of the second scenario. In this case, cracks form along the circumferential direction of the inner tube, so the compressive force generated by the excess BN can be seen as two distinct BN nanosheets pressing against each other. In these simulations, due to the van der Waals forces between the BN nanosheets, the surface tension experienced by the carbon nanosheet varies at the crack formation site, leading to a buckled‐up cracking structure (Figure 5D (i)). In this structure, two separate BN layers were compressed toward each other under the same conditions as before. As the layers meet, a single layer is projected away from the initial BN plane, and at t = 100 ps, it forms a clear flake structure as seen in Figure 5D (ii). Due to the van der Waals forces between the BN nanosheets, they bond and form a new layered structure, which serves as the basis for stable flakes. This structure continues to grow away from the BN plane, and by t = 500 ps, the flake structure is evident, as shown in Figure 5D (iii).

It is reasonable to hypothesize that on the surface of graphene or BNNTs, a composite pattern of the aforementioned two scenarios can occur. Such a composite mechanism—where stress‐induced buckling and cracking act in concert with multi‐point nucleation and imperfect merging—is likely responsible for the complex flake morphologies observed experimentally. The formation of flakes in such a composite pattern is a highly complex structural phenomenon results in complex surface structures, observed in the SEM images shown in Figures 1B and 2. To further support the physical basis of our model, Figures S11 and S1 (Supporting Information) present TEM evidence of ripple and wrinkle formation on BN layers, consistent with the stress‐induced deformation captured in our simulations. In the future, employing real‐time in situ electron microscopy techniques could provide crucial insights into the dynamic transformations governing BN flake formation.^{[ 45 ]}

3. Conclusion

In summary, our comprehensive experimental and simulation studies of the structure of h‐BN flake‐coated CNT have elucidated the h‐BN flakes growth mechanism, structural attributes, and interfacial interactions of this hybrid material. We have demonstrated that coaxial BNNTs form as an intermediate state, with flakes nucleating either through stress‐driven buckling or multi‐point domain merging, both enabled by compressive stress from deposition dynamics. Instead, a coaxial BNNT tube forms first, from which h‐BN flakes buckle off and grow further. Our simulations reveal that deformations in BNNTs originate at surface defects and that the degree of this deformation increases with the radius. Our findings can potentially be extended to other flake‐coated nanotube structures, such as graphene sheet‐coated CNTs and MnO₂ sheet‐coated CNTs. In addition, our method provides a new route to synthesize coaxial CNT@BN structures with a few BNNT walls by controlling the synthesis duration. Due to their exceptional surface area, these structures hold significant promise as materials for heat dissipation and energy storage applications.^{[ 46 ]}

4. Experimental Section

Materials

Ferrocene (sigma‐aldrich, 99%), toluene (sigma‐aldrich, anhydrous, 99.8%), B₂O₃ (sigma‐aldrich, 99.98%), Fe₂O₃ (sigma‐aldrich, nanopowder, <50 nm particle size), and Si wafer.

CNT Synthesis

CNTs were synthesized using aerosol‐assisted chemical vapor deposition (AACVD).^{[ 46} , ⁴⁷ , ⁴⁸ , ⁴⁹ , ^{50 ]} A precursor of 5% ferrocene and toluene mixture as the carbon source was prepared. The aerosol of the mixture solution was introduced into a quartz tube in a horizontal furnace at 800 °C, for 1.5 h. After the experiment, CNTs were collected from the inner wall of the quartz tube.

Post‐Synthesis Treatment of CNTs

CNTs were placed in an alumina boat in a furnace and were burned at 550 °C for 1 h in air, followed by annealing at 1200 °C for 2 h in Ar. Burning in air and annealing in Ar significantly improved the crystallinity of CNT carpets. (Figures S10; Table S1, Supporting Information)

BN Coating

A mixture of B, MgO, and Fe (III) powders in a molar ratio of 1:1:2 was placed into an alumina boat.

The alumina boat was carefully transferred into the middle of a horizontal furnace. The system was purged with Ar at 400 sccm for 40 min, followed by an increase of temperature to 1300 °C at 10 °C min⁻¹. When the temperature reached 1300 °C, ammonia was introduced into the furnace at 240 sccm and Ar flow was lowered to 80 sccm. After 90‐min reaction, the furnace was cooled to room temperature and samples were taken out of the furnace.

The samples were characterized by scanning electron microscopy (SEM), transmission electron microscopy (TEM), Raman spectroscopy, scanning transmission electron microscopy (STEM), electron energy loss spectroscopy (EELS), X‐ray photoelectron spectroscopy (XPS), X‐ray diffraction (XRD), thermal gravimetry analysis (TGA), and Fourier‐transform infrared Spectroscopy (FT‐IR). Sample preparation and operating conditions are described below:

Instruments and Characterization

To prepare for SEM analysis, BNNT samples were coated with an 8 nm layer of platinum using a platinum coater.

SEM images were obtained using a field‐emission SEM (Zeiss Merlin Analytical) with an operating voltage of 5.14 kV at 195 pA beam current. The average BNNT diameters (including coating) were obtained by fitting the BNNT diameter histogram based on 100 unbiased counts from SEM images using ImageJ. The chemical compositions were characterized using Oxford instruments Xmax 150 EDX.

TEM imaging was performed on JEOL JEM‐2100 with a LaB₆ source and an operating voltage of 200 kV. A single tilt (±25 degrees) specimen holder was used, and a Gatan Orius CCD camera is fitted. Gatan Digital Micrograph was used for image processing.

Raman spectra were collected over the spectral range of 1000–2000 cm⁻¹ using a JY Horiba LabRAM ARAMIS Imaging Confocal Raman Microscope equipped with a wavelength of 532 nm. An objective lens of 50x, 532 nm laser wavelength, and a 10% laser intensity of were used to acquire the Raman spectra.

STEM images were collected using a JEOL ARM200F (E01) at 80 kV. A Gatan Quantum Dual EELS spectrometer was used for electron energy loss spectroscopy and spectrum imaging.

XPS analysis was conducted using a Thermo Scientific K‐Alpha XPS, which is equipped with a microfocused monochromated Al X‐ray source for Surface specific analysis.

XRD. XRD diffractograms were obtained using a Bruker D8 ADVANCE Eco diffractometer in reflection mode using Cu Kα radiation (wavelength λ = 0.154 nm).

TGA was employed using a Perkin Elmer Pyris TGA. CNT/BN samples were heated in air to 900 °C at 10 °C min⁻¹ to quantify BN content, and held at 900 °C for 1 h.

FT‐IR spectra were recorded over the range of 500–3500 cm⁻¹ using a Varian Excalibur FTS 3500 FT‐IR Spectrometer.

Conflict of Interest

The authors declare no conflict of interest.

Supporting information

Supporting Information

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.