Twisted-layer boron nitride ceramic with high deformability and strength – Nature

Sample preparation

The oBN nanoprecursors were prepared by the chemical vapour deposition method^16,20,29; the raw materials used in the current work were trimethyl borate and ammonia. The oBN particle size ranged from 50 to 500 nm, with an average size of around 180 nm (Supplementary Fig. 1). A DR.SINTER SPS system and HIGH MULTI 10000 hot-pressing sintering device were used to sinter the precursors, respectively. For SPS sintering, a pressure of 50 MPa was applied first, followed by rapid heating to the target temperature at a rate 100 °C per min. Temperature was monitored with an on-line infrared thermometer during sintering. After 5–10 min at the target temperature, the power was cut off, and the pressure was released. The as-sintered specimens were left in the SPS until they had cooled to room temperature; then, they were taken out and polished. TS-BN ceramics can be synthesized between 1,600 and 1,800 °C by SPS. The ceramic synthesized at 1,500 °C was a composite consisting of TS-BN and residual untransformed oBN (Supplementary Fig. 5). For hot-pressing sintering, the same sintering pressure of 50 MPa was applied first, followed by gradual heating to the target temperature at a rate of 10 °C per min. The holding time was set to 5 min, and then the heating was stopped, and the pressure was released. Densities of the as-sintered specimens were measured according to the Archimedes principle.

XRD and Raman spectroscopy

XRD was used to characterize both the oBN nanoprecursors and the as-sintered BN ceramics, using a Rigaku diffractometer (SmartLab, Rigaku) with Cu Kα radiation (λ = 0.15418 nm). The applied voltage and current were 40 kV and 40 mA, respectively, with a step size of 0.02° at a scanning rate of 1° per min. Raman spectra were also collected at room temperature using a Horiba Jobin Yvon LabRAM system with a laser wavelength of 473 nm. The size of the laser spot was approximately 1 μm.

TEM sample preparation

oBN nanoparticles were dispersed in ethanol solution by ultrasonic treatment, drop-casted on to a carbon-coated copper grid and then dried before TEM observation. Sintered BN ceramics were first crushed and ground in an agate mortar; then, small nanoplates from ceramics were used to prepare TEM samples in the same way as above for oBN. In addition, thin foils were cut from as-sintered bulk samples for TEM observation using a focused ion beam (Helios 5 CX DualBeam, ThermoFisher). The foils were further milled to less than 100 nm and polished by Ar-ion milling (NanoMill; Model 1040, Fischione) to remove surface damage.

Microstructure characterization

We used scanning electron microscopy (Verios, ThermoFisher) to characterize the oBN nanoparticles and fracture morphology of BN ceramics. More detailed microstructure was characterized with a scanning transmission electron microscope (Talos F200X, ThermoFisher) operated at an accelerating voltage of 200 kV and a spherical-aberration-corrected scanning transmission electron microscope (Themis Z, ThermoFisher) operated at an accelerating voltage of 300 kV. HAADF images were collected by combining 20 frames from acquired series with drift correction (DCFI in Velox software, Thermo Fisher). The probe convergence angle was set to 25 mrad, and the collecting angle for HAADF was set to 65–200 mrad to eliminate the coherent scattering effect.

First-principles calculations

We constructed twist-layer BN structures using the Materials Visualizer module of the Materials Studio software³⁰. Calculations were performed on the basis of DFT as implemented in the CASTEP code³¹. Ultrasoft pseudopotentials were used^32,33. We used the local density approximation exchange-correlation functional of Ceperley and Alder parameterized by Perdew and Zunger^34,35 to perform structural optimization and calculations of total energies and elastic properties. A k-point sampling³⁶ of 0.04 × 2π Å⁻¹ and a plane-wave cutoff of 570 eV were applied. The Broyden–Fletcher–Goldfarb–Shanno³⁷ minimization scheme was used for geometry optimization. Structural relaxation was stopped when the total energy changes, maximum ionic displacement, stress and ionic Hellmann–Feynman force were less than 5.0 × 10⁻⁶ eV per atom, 5.0 × 10⁻⁴ Å, 0.02 GPa and 0.01 eV Å⁻¹, respectively. The elastic moduli of the investigated structures were calculated in the linear elastic strain range. Selected calculation parameters were tested to ensure that energy convergence was less than 1 meV per atom. To validate our computational scheme, benchmark calculations were conducted for the hBN structure. The calculated lattice parameters of a = 2.49 Å and c = 6.48 Å were in good agreement with experimental values of a = 2.50 Å and c = 6.66 Å (ref. ³⁸). The calculated bulk modulus (27.9 GPa) of hBN was in agreement with the experimental value (25.6 GPa)³⁹.

Deformability factor calculation

The ability of layered vdW materials to deform without fracture can be characterized by the deformability factor Ξ = (E_c/E_s)(1/Y) (in units of GPa⁻¹)¹⁸, where E_c and E_s are the cleavage energy and slipping energy, respectively, and Y is the in-plane Young’s modulus along the slip direction. The E_c/E_s ratio quantifies the plasticity of the material that conforms to the criterion proposed by Rice et al.^40,41. Interlayer interactions and the relative glide of the twist-stacked BN structures were simulated on the basis of DFT calculations. The slipping step was kept at 0.3 Å during the simulation. For each step, the energy of the most stable configuration was obtained by geometrically optimizing only the interlayer distance. The (001) plane uniformly slipped along the [210] direction, which is considered to be the lowest-energy sliding direction in hBN⁴². We obtained the energy as a function of the slip distance over the range of periodic distances. The energy difference between the slip distance at maximum energy (E_max) and no slip (E₀) was used to represent the energy barrier to overcome resistance to slip, that is, E_s = E_max − E₀ (ref. ¹⁸). The energy difference between the infinite interlayer distance (E_inf) and E_max was considered to represent the cleavage energy, E_c = E_inf − E_max (ref. ¹⁸). An interlayer distance of 10 Å was used to calculate E_inf, which safely ensured that there was no interlayer interaction.

Molecular dynamics simulation

The phase transition process from oBN to a twisted-layer structure was simulated by molecular dynamics with the large-scale atomic/molecular massively parallel simulator code⁴³. An extended Tersoff potential was chosen to describe the interatomic interaction⁴⁴; this has been widely used to investigate the microstructural evolution of hBN^45,46. In this work, a 10 × 10 × 4 nm³ supercell containing a two-shell BN nano-onion at the centre was constructed. The two-shell BN nano-onion structure was constructed using the method reported in our previous work²¹. The outer (inner) shell corresponded to B₇₅₀N₇₅₀ (B₄₆₀N₄₆₀), and the diameter of the BN onion was 3.61 nm. Periodic boundary conditions and isothermal–isobaric (NPT) ensemble were applied in the simulations. Each supercell was first optimized with the conjugate gradient algorithm and then relaxed for 20 ps at room temperature. Following the relaxation, the supercell was compressed uniaxially along the z direction to a given pressure (6 GPa) within 200 ps and finally heated to the target temperature (1,500 K) within 2 ns. Atomic configurations were visualized and analysed with the help of the Open Visualization Tool package⁴⁷. The local structural environment of the atoms was identified using the polyhedral template matching algorithm⁴⁸.

Mechanical property characterization

Uniaxial compression tests were performed in the MTI mechanical property testing system (MTII/Fullman SEMtester 2000) at room temperature with a strain rate of 1 × 10⁻⁴ s⁻¹. Ceramic specimens were machined into cylinders with diameter of 2.7 mm and length of 4.0 mm (length to diameter ratio: approximately 1.5). Both ends of the cylinders were polished with diamond powder of around 0.5 μm. Parallelism between the two ends was within 0.01 mm. A thin copper foil was placed between the sample and the tester to reduce stress concentration on the contact area. Thin nickel (Ni) film was deposited on the sample surface to form markers. Each compression process was recorded in situ with a digital video recorder. Sample strain was estimated by measuring the change in distances between Ni markers. We conducted at least five tests for uniaxial compressive properties on bulk samples obtained from SPS and hot-pressing sintering.

Tensile tests were carried out at a strain rate of 1 × 10⁻⁴ s⁻¹. The specimens were processed into I-shape with an effective tensile length of 6 mm and a rectangular cross section of 1 mm × 1.5 mm. Flexural strength was measured using the three-point bending method. The specimens were processed into cuboids with a size of 1 mm × 2 mm × 12 mm. The span length was 11 mm, and the loading rate of the indenter was set to 0.1 mm min⁻¹. Flexural strength (σ_f) was calculated as follows:

where F is the maximum load until the specimen fractured, L is the span length, and b and h are the width and height of the specimen, respectively.

Fracture toughness of the specimens was determined using the single-edge V-notched beam method. The specimens were processed into cuboids with a size of 1 mm × 1.5 mm × 5 mm. A straight-through V-notch with a depth of approximately 0.5 mm was cut in the specimen using a femtosecond laser (Astrella-1K-USP). The radius of the notch tip was less than 10 μm, the span length was 4 mm and the crosshead loading rate was 0.1 mm min⁻¹. Fracture toughness (K_IC) was calculated using the following equations:

where F is the maximum load until the specimen fractured, L is the span length, B is the specimen width, W is the specimen height and a is the notch depth. Tests for tensile strength, flexural strength, Young’s modulus and fracture toughness were repeated at least five times.

In situ synchrotron radiation XRD measurements

In situ triaxial compression tests were performed using a deformation-DIA apparatus coupled with synchrotron X-rays, at the GSECARS 13-BM-D beamline of the Advanced Photon Source at the Argonne National Laboratory, USA. Details of the deformation-DIA module and the sample assembly can be found elsewhere^28,49. The specimens were cylinders of 2.5 mm in diameter and 3.5 mm in length. The specimen was first compressed to a hydrostatic pressure of around 1.5 GPa; then, the differential rams were advanced to shorten the sample at a constant speed (strain rate = 1 × 10⁻⁵ s⁻¹) at room temperature.

The incident monochromatic beam (45 keV) was collimated to tungsten carbide (WC) slits of 200 mm × 200 mm. Detector tilt and rotation relative to the incident beam were calibrated with a CeO₂ standard using the Dioptas program⁵⁰. XRD patterns and radiographs of the sample were collected automatically during deformation. Exposure times for XRD and radiographs were 300 s and 10 s, respectively.

The strain of specimen was defined as ε = (l−l_ε)/l, where l and l_ε are the sample lengths at the initial state and under compression, respectively. Lattice strain of the (002) and (100) reflections was defined as

where d_P(hkl) is the hydrostatic pressure d-spacing, for which the right-hand side of equation (4) is zero. For each plane (hkl), d_(hkl) and φ were measured from the two-dimensional diffraction pattern (with cosφ = cosθcosδ), δ being the true azimuth angle. Q_(hkl) and d_P(hkl) were extracted by fitting d_(hkl) versus φ according to equation (4). Differential stresses σ_(hkl) were calculated for planes from lattice strains Q_(hkl) according to:

where the ‘effective moduli’ G_hkl were calculated with elastic compliances S_ij from inversion of the stiffness tensor (C_ij) for hBN. In the current work, elastic constants of c₃₃ = 27 GPa and c₁₁ = 811 GPa (ref. ³⁹) were used to calculate σ₍₀₀₂₎ and σ₍₁₀₀₎ of layered BN structures, respectively, without considering pressure effects on these moduli.

In situ TEM nanopillar compression

Nanopillars with diameter of around 200 nm and aspect ratio of around 2:1 from the sintered bulk ceramic were fabricated using a Ga ion beam at a voltage of 30 kV in a FEI Helios focused ion beam instrument. Initially, the samples were processed into pillars with a cross-sectional width of approximately 5 μm using relatively large currents from 21 nA to 7 nA. Subsequently, the pillars were milled to cylinders with diameter of 1 μm using low currents from 5 nA to 1 nA. Finally, the pillars were polished to the desired size of approximately 200 nm using small currents from 500 pA to 7.7 pA to minimize the damage layer.

In situ uniaxial compression tests were performed with a Hysitron Picoindenter instrument inside a transmission electron microscope (FEI Titan ETEM G2) operated with an accelerating voltage of 300 kV. The Hysitron PI-95 holder was equipped with a diamond punch joined to a MEMS transducer. In situ compression experiments were carried out in a displacement control mode, which has been proved to be more sensitive to transient phenomena. The displacement rate was kept at 10 nm s⁻¹ during compression, corresponding to a strain rate of 1 × 10⁻² s⁻¹. The whole process was recorded using a digital video recorder for observation of the evolution of the microstructure.