Direct conversion of h-BN into c-BN and formation of epitaxial c-BN/diamond heterostructures

Synthesis and processing of cubic boron nitride (c-BN) and diamond related composites with immense scientific and technological importance have presented formidable challenges because of nonequilibrium thermodynamics. Boron nitride exists in four polymorphs, namely, hexagonal (h-BN), rhombohedral (r-BN), wurtzitic (w-BN), and cubic (c-BN) zinc-blende structures. Out of these, h-BN (point group = D_6h, space group = P6₃/mmc) and c-BN (space group = Fd3m) have generated tremendous scientific and technological interests due to their analogous structures and properties with graphite and diamond, respectively. The c-BN (lattice constant = 0.361 nm) and diamond (0.356 nm) have the number density of atoms 1.69 × 10²⁹ m⁻³ and 1.77 × 10²⁹ m⁻³, respectively. This fact coupled with strong interatomic potentials leads to the highest hardness (∼7500 kg mm⁻² for c-BN and ∼10 000 kg mm⁻² for diamond) and stiffness (Young's modulus ∼ 700 GPa for c-BN and ∼1000 GPa for diamond). The thermal conductivity of c-BN (12 W cm⁻¹ K⁻¹) is close to that of diamond (20 W cm⁻¹ K⁻¹), which has the highest value of all the known materials, including copper (4 W cm⁻¹ K⁻¹). The coefficient of friction of c-BN and diamond is the lowest (less than 0.1) of all the known materials. However, c-BN coatings have certain advantages over diamond due to higher oxidation resistance than diamond because of protective boron oxide layers. Due to less reactivity even at high temperatures with ferrous alloys, c-BN coatings are ideally suited for machining these alloys. The electrical properties of c-BN are quite similar to those of diamond, where c-BN has Johnson (for high-power devices) and Keyes (for integrated circuits) figures of merit 8200 and 32 compared with 1 for silicon. However, unlike diamond films, which can be doped reliably with p-type dopants only, c-BN can be doped with both n- and p-type dopants and it can have boron oxide as the insulating layer needed for solid state devices. In view of the above properties, c-BN and diamond represent Holy Grail for solid state devices and systems, ranging from high-power devices to cutting tools and biomedical applications.^1–3 

According to well-established carbon phase diagram, the diamond is in the metastable state over the entire temperature range at ambient pressures and it turns into vapor above 5000 K. The graphite–diamond–liquid triple point occurs at 5000 K and 12 GPa (120 000 atmospheric pressures).⁴ Similarly, hBN–cBN–liquid triple point occurs at 3500 K and 9.5 GPa.⁵ Thus direct conversion of carbon into diamond and h-BN into c-BN requires these high temperatures and pressures, corresponding to their triple points. These high-temperatures and pressures can be reduced slightly in the presence of catalysts, which may contaminate the bulk diamond and c-BN. In the presence an epitaxial template, these pressures and temperatures for metastable phase formation of diamond and c-BN can be reduced, as shown for c-BN synthesis.⁶ In addition, CVD of diamond thin films still requires temperatures exceeding 1273 K in the presence of hydrogen.⁷ Most successful methods for c-BN synthesis require energetic introduction of ions with phase purity limited to 85%.^1,2

Recently, we reported a major breakthrough in the synthesis and processing of diamond by direct conversion of carbon into diamond at ambient temperatures and atmospheric pressure in air.^8,9 We showed that amorphous carbon can be melted under a super undercooled state, which can be quenched into a new state of Q-carbon, from which nanodiamonds, microdiamonds, nanoneedles, microneedles, and large-area single-crystal diamond thin films are formed by controlling the template for diamond nucleation and growth.

In this paper, we report a direct conversion of h-BN into c-BN by a similar nanosecond pulsed laser melting at ambient temperatures and atmospheric pressure in air in accordance with curve 3 of the phase diagram (Figure 1). This extension for BN phase diagram is very similar to the one proposed for carbon.^8,9 Similar to Q-carbon, we have created Q-BN, which is formed as a result of quenching from super undercooled state from which phase-pure c-BN is grown in the form of single-crystal nanodots, microcrystals, nanoneedles, microneedles, and large-area films. The nanoneedles and microneedles are formed as a result of interfacial instability during crystallization from the liquid phase. We have also grown diamond on the top of these structures to create epitaxial c-BN/diamond and diamond/c-BN heterostructures by pulsed laser deposition of carbon. The c-BN/diamond heterostructures are needed for next-generation high-power devices as well as for high-speed machining and cutting tool applications. We discuss the mechanism of diamond growth by laser evaporation of carbon without the presence of hydrogen. We discuss the similarities and differences between h-BN to c-BN and carbon to diamond conversions and accompanying phase transformations.

Hexagonal BN (h-BN) was deposited on c-sapphire, glass, and polymer substrates at room temperature using ArF laser (pulse duration = 20 ns, wavelength = 193 nm, and energy density = 3.0 J cm⁻²) in laser MBE chamber under a vacuum of 3 × 10⁻⁸ Torr. The as-deposited films were nanocrystalline with grain size ∼25 nm. To form Q-BN and c-BN, these films were irradiated using pulsed ArF laser having energy density of 0.3–1.0 J cm⁻². Diamond is grown by pulsed laser deposition (PLD) under oxygen partial pressure of 0.2 Torr at 500 °C. During PLD a diamond-like carbon (DLC) with a mixture of sp³ and sp² bonded carbon is deposited and diamond can grow on sp³ bonded substrate (initially c-BN and later diamond), while oxygen etches off and removes sp² bonded carbon preferentially. State-of-the-art characterization techniques like micro-Raman spectroscopy, high resolution scanning electron microscope (SEM), electron back scattered diffraction (EBSD), high resolution TEM (HRTEM), aberration corrected STEM, and electron energy loss spectroscopy (EELS) were used to characterize these samples. Alfa300 R superior confocal Raman spectroscope with a lateral resolution less than 200 nm was employed to characterize the Raman active vibrational modes in h-BN, Q-BN, and c-BN films. Crystalline Si was used for calibrating the Raman spectra, which has its characteristic Raman peak at 520.6 cm⁻¹. High resolution SEM with sub nanometer resolution was carried out using FEI Verios 460 L SEM to characterize the as-deposited and laser irradiated films. FEI Quanta 3D FEG with dual beam technology employing both electron and ion beam guns was used for preparing cross-sectional TEM samples. EBSD HKLNordlys detector with less than 10 nm lateral resolution was used to map out Kikuchi diffraction pattern in FEI Quanta 3D FEG instrument, thereby elucidating crystal structures and phases of c-BN and h-BN present in the as-deposited and laser annealed thin films. Aberration corrected STEM-FEI Titan 80–300 and JEOL-2010 STEM/TEM (point-to-point TEM resolution 0.18 nm; STEM-Z resolution 0.08 nm with information limit of 0.06 nm) were used in conjunction with electron energy loss spectroscopy (EELS) having a lateral resolution of 0.15 eV to collect high resolution TEM images, micro-diffraction, and EELS spectra for determination of interfacial characteristics and crystal structure in laser annealed c-BN.

Fig. 1 shows the BN phase diagram (P vs. T) in the range of 0–10 GPa pressure and 0–4000 K temperature, containing regions of phase stability for c-BN, h-BN, liquid, and vapor.⁵ According to well-accepted phase diagram of Corrigan and Bundy (curve 1)⁵ c-BN line intersects pressure axis at 1.4 GPa at zero K without meeting the temperature axis, making the h-BN as the stable phase in the entire temperature range 0–3000 K above which BN turns into vapor.⁵ Solozhenko et al. have tried to refine the phase diagram based on experimental data on BN melting and extrapolation of specific heats of various BN polymorphs into high-temperature regions. This modification (curve 2) shifts L-cBN-hBN triple point from 3500 K/9.5 GPa (Corrigan-Bundy P-T diagram) to 3700 K/7.0 GPa, and show cBN to be the stable phase in the temperature range 0–1600 K and h-BN beyond that at atmospheric pressure.^4,10,11 Thermodynamic calculations based upon BN melting and scaling of specific heats C⁰_p(T) (Table I) to high temperatures by Solozhenko et al.^10,11 have suggested curve 2, where triple point is shifted to lower pressures but higher temperatures and made c-BN to be the stable phase 0–1600 K and h-BN beyond that. In view of the multiple fitting parameters in C⁰_p(T) extrapolation to higher temperatures for calculating enthalpies and entropies, the accuracy of free energy values to better than 5% is questionable. It should be pointed out that while the free energies of transformation of h-BN to c-BN and graphite to diamond are large, the differences between the two are small, less than 4% (Table I).

The phase pure c-BN can be synthesized only near the triple point at high temperatures and pressures. Recently, phase pure c-BN has been synthesized at slightly lower pressures and temperatures (1700 K/5.5 GPa) by using a lattice matching diamond substrate for c-BN growth from melt.⁶ The c-BN processing at high pressures and temperature involves expensive steps with limited yield. In addition, synthesis and scale-up processing of phase-pure c-BN by both physical vapor deposition (PVD) and CVD methods are still very challenging, and only up to ∼85% c-BN content has been achieved.^1,2 The PVD methods close to equilibrium lead to the formation of h-BN, and the formation of c-BN requires highly nonequilibrium processing under energetic ion bombardment, localized stress, and high concentration of defects.^12–14 These observations tend to favor Corrigan and Bundy phase diagram (curve 1 in Fig. 1) showing h-BN as the stable phase and c-BN metastable in the entire temperature range at ambient pressures.⁵ The PVD methods for c-BN synthesis involve significant levels of ion bombardment during growth in a narrow window of parameters. The yields are very limited due to sputtering losses from the ion bombardment, and CVD methods are not as well established as those for diamond synthesis.⁷ In the case of diamond, thermal and CVD techniques are fairly well established, although these processes occur at high temperatures in the presence of hydrogen, and are very energy intensive with limited yield.

We show a direct conversion of h-BN into c-BN through rapid melting under super undercooled state and quenching. Fig. 2(a) shows the formation of super undercooled BN (referred as Q-BN) from which nano and micro c-BN crystallites are formed. The as-deposited h-BN is in the form of nanocrystalline h-BN, which is melted at atmospheric pressure in air at the estimated temperature of 2800 K. The temperature of 2800 K was estimated by numerical simulation using SLIM computer code developed by Singh and Narayan.¹⁵ The Q-BN is formed near the sapphire interface, similar to Q-carbon, which can break into filamentary structure through interfacial instability.¹⁶ The Q-BN can be converted into nanoneedles and microneedles, as shown in Fig. 2(b). The inset shows characteristic EBSD pattern of c-BN, which establishes the nanoneedles and microneedles to be single-crystal. The length of microneedles of over a micron can occur only under liquid-phase growth with estimated velocity exceeding 5 m/s with estimated growth time of 200 ns. The EBSD pattern, also known as backscatter Kikuchi diffraction (BKD), by using field-emission scanning electron microscope, is a powerful technique to determine the crystal structure of nano- and microcrystalline regions (with resolution of 10 nm) and their relative orientations with respect to the substrate. In the EBSD a stationary electron beam strikes a tilted crystalline sample and the diffracted electrons form a pattern on a fluorescent screen. This pattern is characteristic of the crystal structure and orientation of the sample region from which it was generated. It provides the absolute crystal orientation with sub-micron resolution. Fig. 2(c) shows the formation of large-area 〈111〉 cBN platelet, which has grown epitaxially on (0001) sapphire. Here, c-BN has grown directly as a result of quenching from super undercooled state, where (0001) sapphire has provided a template for (111) c-BN epitaxial growth. The epitaxial growth of (111) c-BN on (0001) sapphire occurs by domain matching epitaxy, where integral multiples of lattice planes match across the film–substrate interface. Fig. 2(c) shows multiple domains that are related by twinning as expected from epitaxial growth of cubic BN on (0001) sapphire having hexagonal symmetry. The formation of twins in c-BN can be beneficial for improving fracture toughness and surface catalytic properties. The details of epitaxial growth will be discussed with HRTEM results. The growth of a very large-area (single domain) epitaxial (111) c-BN single-crystal thin film is shown in Fig. 2(d).

The Raman spectroscopy provides a powerful complementary technique to study the bonding characteristics of c-BN and diamond¹⁷ and compare these results with electron energy loss spectroscopy (EELS). Figure 3 shows Raman spectrum using 633 nm excitation wavelength from h-BN before laser annealing which has sharp E_2g peak at 1370 cm⁻¹, characteristic of high-quality h-BN. From high-resolution X-ray diffraction, the average grain size was determined to be 25 nm, in agreement with high-resolution SEM data, after a single laser pulse of ArF laser (in the outer regions of energy density of 0.6 J cm⁻²). The Raman spectra from two different regions contain all characteristic TO and LO peaks of c-BN, where there is very small peak of h-BN (region 1) and is completely gone in region 2, showing a complete conversion of h-BN into phase -pure c-BN. Table II show a detailed comparison of our experimental results of first-order Raman peaks with previous experimental results from second-order peaks¹⁸ and theoretical ab-initio calculations of Raman active modes in c-BN.¹⁹ Such a fine structure in the Raman spectra, besides well known TO(Γ) at 1060 cm⁻¹ and LO(Γ) at 1310 cm⁻¹, is indicative of very high-quality structure of c-BN. Pressure dependence measurements in the range of 0–10 GPa of TO(Γ) and LO(Γ) Raman frequencies have shown a linear relation given by ω = (1054.7 ± 0.6) + (3.39 ± 0.08) p for TO(Γ) and ω = (1305 ± 1) + (3.45 ± 0.07) p, where p is in GPa.²⁰ 

The HRTEM and EELS characterization of c-BN and Q-BN were carried out using the cross-section TEM samples prepared by focused ion beam (FIB) thinning techniques. Figure 4(a) is a cross-section HRTEM image from c-BN thin films, where the 〈110〉 cross-section has two sets of {111} planes and 〈110〉 columns of c-BN are clearly imaged. The characteristic 〈110〉 diamond selected-area-diffraction pattern is included in the inset, showing first order (111) and (002) spots. The HRTEM shows c-BN to be free from defects, such as dislocations, twins, and stacking faults. From the inset selected-area diffraction patterns of c-BN (in Fig. 4(a)) and that of sapphire (in Fig. 4(b)), we derive the following epitaxial relations: 〈111〉 c-BN//〈0001〉 sapphire (out of the plane direction) and 〈110〉 c-BN//〈−2110〉 sapphire and 〈112〉 c-BN//〈$101¯0$〉 sapphire in the plane of the film. This shows epitaxial growth of c-BN by domain matching epitaxy,^21,22 where 15 (d₁₁₀) planes match with 16 (⁠$d2¯110$⁠) planes of sapphire. Figure 4(b) shows HRTEM image from the Q-BN, where it has mostly amorphous structure with a few nanocrystallites embedded into it, as shown by arrows. To investigate the details of bonding characteristics EELS studies were carried out using aberration corrected STEM-FEI Titan 80–300 with an energy resolution of 0.15 eV. Core loss electron energy loss spectroscopy (EELS) is a powerful technique to determine the structure and bonding characteristics in solids, where the electrons are excited from core levels to the unoccupied states and thereby providing a direct way of measuring electronic transitions and structure. An intense sub-angstrom electron probe EELS in conjunction with HRTEM is used to evaluate various band edge structures of B and N in c-BN and Q-BN, formed after laser annealing process. Figures 4(c) and 4(d) represent EELS spectrum of c-BN and Q-BN, respectively. The insets represent B K-edge of c-BN and Q-BN. The c-BN formed after laser annealing has its characteristic B K-edge at 197.5 eV and 216 eV, which is in accordance to the previously reported results.^23,24 There is a complete absence of the 216 eV peak from w-BN due to its different density of states as compared with c-BN.¹⁹ There are also characteristic N K-edges at 408 eV and 415 eV, which show that the c-BN formed after laser annealing technique is pristine. Hexagonal-BN and w-BN have their characteristic B K-edge at 192 eV, which is completely absent in the EELS spectra of laser annealed c-BN. BN is a partially covalent material owing to the different electronegativity values of B and N atoms. In h-BN, there occur σ states on its basal plane with weak interlayer π states thereby having characteristic σ and π edges in EELS spectrum. There are both π edge (at 192 eV) and σ edge (at 197 eV) in h-BN and w-BN, which are affected by the strong attractive interaction between the core hole and excited electron.¹⁹ The EELS spectrum from the uniform layer of Q-BN near the sapphire substrate contains B K edge at 198.50 eV and 218.75 eV, which are broader and slightly shifted c-BN edges (at 198.25 eV and 217 eV). The intensity of higher energy band edge N peaks is low owing to the reduced thickness of the Q-BN layer.

Fig. 5 shows Raman spectra from c-BN/diamond epitaxial thin film heterostructure. These lattice matching c-BN/diamond epitaxial heterostructures were created by growing diamond films on our phase-pure cBN films. The diamond films were deposited by pulsed laser evaporation of amorphous carbon target using KrF (248 nm laser wavelength) laser with energy density 3–4 J cm⁻², pulse duration 25 ns, and 10 Hz with oxygen partial pressure of 0.2 Torr at 500 °C. The Raman spectra results are shown in Fig. 5, where there are phase-pure c-BN peaks 1076 cm⁻¹ and strong diamond peak at 1336 cm⁻¹. The position of LO (Γ) at 1311 cm⁻¹ is masked by the much stronger diamond peak at 1336 cm⁻¹. The Raman shift in diamond¹⁷ (Δω) is related to Δω (in cm⁻¹) = 2.2 ± 0.10 cm⁻¹ GPa⁻¹ along the [111] direction, Δω (in cm⁻¹) = 0.73 ± 0.20 cm⁻¹ GPa⁻¹ along the [100] direction, and Δω (in cm⁻¹) = 3.2 ± 0.23 cm⁻¹ GPa⁻¹ for the hydrostatic component in c-BN.¹⁷ The biaxial stress in thin films can be described as a combination of two-thirds hydrostatic and one-third uniaxial stress. The biaxial stress can be estimated using σ = 2μ ((1 + ʋ)/(1 − ʋ))Δε, where μ is shear modulus, ʋ is Poisson's ratio, and Δε is the in-plain strain. Thin film strain has three components: planar (lattice) misfit, thermal, and defects. The planar misfit strain between diamond and c-BN is estimated to be 1.4%, and thermal misfit 0.02%. The thermal is ΔαΔT, where Δα is the difference in thermal coefficient of expansion and ΔT is the change in temperature. Assuming defect strain to be very small, the biaxial stress in the diamond layer is estimated to be about 2.2 GPa, which can explain the diamond peak shift from 1332 cm⁻¹ to 1336 cm⁻¹.

Fig. 6 shows EELS spectra from c-BN and diamond interface (a), diamond (b), amorphous DLC (c), and Q-BN (d). The inset figure identifies the various regions from which EELS spectra were acquired. Fig. 6(a) contains characteristic c-BN EELS spectrum with B K-edge at 197.5 eV and 216 eV, and the characteristic diamond spectrum with a sharp edge at 288 eV and a peak at 292 eV, corresponding to sp³ (σ*) bonding. Away from the c-BN/diamond interface, the EELS spectrum contains only the signature EELS spectrum for diamond with a sharp edge at 288 eV and a peak at 292 eV, as shown in Fig. 6(b). Fig. 6(c) shows EELS spectrum from the diamond surface, which is grown by pulsed laser deposition under oxygen partial pressure of 0.2 Torr at 500 °C. This EELS spectrum shows diamond-like carbon with a mixture of sp³ and sp² bonded carbon from which diamond can grow on sp³ bonded substrate (initially c-BN and later diamond), while oxygen etches off and removes sp² bonded carbon preferentially. The EELS spectrum near the sapphire interface (Fig. 6(d)) shows the formation of Q-BN with B K-edge peaks at 195 eV and 204 eV. The c-BN embedded into Q-BN provides a template for the growth of diamond by pulsed laser deposition. This epitaxial growth of diamond pulsed laser deposition carbon without the presence of hydrogen or any catalysts is quite interesting. In this case, energetic carbon atoms (with average energy of 16 eV) from PLD form sp³ bonded carbon species which are incorporated into the substrate templates, such as c-BN, diamond, and sapphire.

We have investigated the details of Q-BN formation upon quenching from super undercooled state of h-BN, and epitaxial growth of diamond, as shown in Fig. 7. The EELS spectrum from the uniform layer of Q-BN (Fig. 7(a)) near the sapphire contains B K-edge at 198.50 eV and 218.75 eV, which are slightly shifted from c-BN edges (198.25 eV and 217 eV) in Fig. 7(b). This Q-BN layer contains c-BN regions, which appear somewhat darker in the high angle dark field (HADF) STEM-Z contrast compared with the Q-BN contrast, indicating lower specific number density of atoms or shorter bond lengths in Q-BN compared with c-BN. The darkest band over the Q-BN layer B K-edges at 193.0 eV and 201 eV, which correspond to h-BN (Fig. 7(c)). This is consistent with the fact that c-BN atomic density is higher by 54.7% than that of h-BN. The epitaxial diamond has a sharp edge at 286.5 eV, with characteristic σ* peaks at 291.75 eV, 297.25 eV, and 304.75 eV. It is interesting to note that the EELS spectra of c-BN and Q-BN are remarkably similar with only σ* peaks, suggesting only sp³ bonding in both of these structures. In the case of Q-carbon, on the other hand, EELS spectrum contained a small π* peak, corresponding to 10%–15% sp² bonded carbon. The presence of sp² bonded carbon and sp²/sp³ interfaces were found to be critical for the creation of unpaired spins and observation of ferromagnetism in Q-carbon. Thus, observation of diamagnetism and the absence of ferromagnetism in Q-BN are consistent with absence of sp² bonded (π* peak) in Q-BN. Fig. 8 shows high-resolution STEM-Z (HAADF, high-angle annular dark-field) image, and EELS images corresponding to carbon, boron, and oxygen edges from the area shown in Fig. 7. It is interesting to note darker regions in Q-BN, which correspond to c-BN with lower number density of atoms. The presence of c-BN nanocrystallites in Q-BN matrix is consistent with HRTEM results, shown in Fig. 4(b). Thus Q-BN, similar to Q-carbon, is denser than c-BN and may have shorter bond length with higher hardness.

We have shown that nanocrystalline h-BN can be converted directly into phase pure c-BN at ambient temperature and atmospheric pressure in air. By using nanosecond laser pulses, we create super undercooled molten state of BN which is quenched into a new state of BN, which we have named Q-BN, analogous to the formation of Q-carbon. High-quality phase-pure c-BN is formed from Q-BN and by controlling the nucleation and growth of c-BN, we can form nanodots, microcrystallites, nanoneedles, microneedles, and large-area thin films. Large-area epitaxial c-BN thin films are formed in the presence of planar (like sapphire) or lattice matching (like copper) substrates, which provide a template during growth from super undercooled liquid. The atomic structure of h-BN, c-BN, and Q-BN and bonding characteristics have been studied by high-resolution SEM, HRTEM, STEM, EELS, and Raman spectroscopy. We have also created c-BN/diamond epitaxial composites with profound implications for high speed machining and next-generation microelectronic, photonic, and power devices. The preliminary studies have shown Q-BN to exhibit superior field emission, hardness, and optical and electronic properties, analogous to the properties of Q-carbon. Since diamond and c-BN are grown from super undercooled liquid, they can be doped with both n- and p-type dopants. The details of these properties of Q-BN and c-BN will be reported shortly.

We are grateful to Fan Family Foundation Distinguished Chair Endowment for Professor J. Narayan, and this research was partly funded by the National Science Foundation (Grant No.: DMR-1304607). We are also very pleased to acknowledge the technical help and useful discussions with John Prater, Jim LeBeau, Roger Narayan, and Jerry Cuomo.