Structural and chemical analysis of c-BN/diamond heterostructures

Cubic boron nitride (c-BN) is an ultra-wide bandgap (UWBG) semiconductor with the highest bandgap among the binary nitrides (∼6.4 eV), as well as high thermal conductivity (13 W cm⁻¹ K⁻¹), low dielectric constant (7.1), and high breakdown electric field (8 MV cm⁻¹).¹ Moreover, both n- and p-type doping have been achieved, thus making it highly attractive for high-power and high-temperature electronic applications, especially in harsh environments.¹ Diamond, which has a high thermal conductivity (22 W/cm  K), is a promising substrate candidate for epitaxial growth of single-crystal c-BN due to the small lattice mismatch (1.36%), and its linear thermal expansion coefficient (3.1 × 10⁻⁶/K) is only slightly smaller than that of c-BN (4.7 × 10⁻⁶/K).² Thus, c-BN/diamond heterostructures could potentially provide new and improved options for high-power electronic devices.^3–5 However, the synthesis of wafer-scale c-BN single crystals or epitaxial heterostructures remains in an early stage of development.

Cubic BN has the zincblende crystal structure with a lattice constant $a=3.615Å$⁠, and it belongs to the $F 4 ¯3m$ space group. Boron nitride is polymorphic and crystallizes in sp²-bonded hexagonal (h-BN), rhombohedral (r-BN), and turbostratic (t-BN) phases, as well as in sp³-bonded cubic (c-BN) and wurtzite (w-BN) phases.⁶ This polymorphism poses a major challenge for the growth of the pure cubic-phase material. Moreover, there is considerable ongoing discussion about which BN phase is thermodynamically the most stable.^7–9 

Since the initial discovery of the c-BN phase,¹⁰ its stability has been closely tied to the method of growth. Aggressive growth techniques that induce compressive stress in the growing film, such as high-pressure high-temperature (HPHT),^11–13 ion-assisted pulsed laser deposition (PLD),¹⁴ and physical vapor deposition (PVD),^15–17 have proceeded on the basis that the initial nucleation of amorphous BN and the more stable t- or h-BN phases was inevitable. Thus, highly energetic ion bombardment was considered necessary to achieve the desired transition to the cubic phase.^16,18,19 These results aligned well with the first-ever reported phase diagram of BN,²⁰ which suggested that t-BN or h-BN were the two most stable BN phases under standard temperature and pressure (STP) conditions (273.15 K and 1 atm). Later reports indicated nucleation of the cubic phase under more relaxed growth condition,^21,22 and these results were supported by a modified phase diagram based on theoretical thermodynamic calculations.²³ More recently, advances in growth techniques have demonstrated significant progress in achieving high-quality c-BN films with the presence of minimal sp²-bonded BN phases. For example, fluorine-based chemical vapor deposition (CVD) has been utilized to grow c-BN films with a columnar structure and large crystal sizes (>1 μm) using silicon substrates.²⁴ This method involved the formation of approximately 100 nm of amorphous or turbostratic BN at the interface, followed by the nucleation of a predominantly cubic phase, indicating a substantial reduction in the proportion of non-cubic phases compared to earlier growth methods. Moreover, molecular beam epitaxy (MBE) has achieved epitaxial growth of c-BN films with notable results. One recent study reported films containing over 99% c-BN, with only trace amounts of sp²-bonded BN detected at the interface and at the film surface,²⁵ while another reported the successful growth of epitaxial c-BN films devoid of any sp²-bonded BN at the interface.²⁶ However, simulations based on ab initio framework—fixed-node diffusion Monte Carlo (FNDMC) have suggested that h-BN is more stable than c-BN under STP conditions.²⁷ 

Despite reports of single crystal c-BN films being achieved, there remains a notable absence of any detailed structural analysis. Prior workers have used Fourier transform infrared spectroscopy (FTIR), x-ray diffraction (XRD), and transmission electron microscopy (TEM) for phase identification and texture analysis of nanostructured c-BN films.^14,28–32 These studies have mostly focused on comparing film quality under different growth conditions, with little attention being paid to morphological features present in the films. This present paper describes a detailed TEM investigation of c-BN/diamond heterostructures grown by electron cyclotron resonance-plasma enhanced chemical vapor deposition (ECR PECVD) and focuses on the structural evolution of the BN films and evaluating the defect types and densities as a function of growth conditions. Chemical analysis using electron-energy-loss spectroscopy (EELS) has also been used to differentiate between regions of sp² and sp³ bonding within the BN films.

The BN films were grown on heavily B-doped single-crystal diamond substrates (Technical Institute of Superhard Novel Carbon Materials, TISNCM) using ECR PECVD. The substrates were loaded into the CVD vacuum system, and hydrogen plasma was used for cleaning the substrate surfaces at high temperatures and low pressures prior to initiation of growth (refer to Table I for details). Two different cleaning procedures, labeled here as C1 and C2, were followed. Deposition parameters for the samples are summarized in Table II. Deposition of BN was carried out using gas precursors H₂, BF₃, N₂, and inert gases He and Ar, with a negative bias voltage of −60 V, and microwave power of 1.42 kW. Gas flow rates of N₂, and He and Ar were kept constant at 12.5, 35, and 2.5 SCCM, respectively, during film growth. Substrate temperatures were varied between 735 and 820 °C and the chamber pressure was ∼1.1 × 10⁻⁴ Torr. More extensive details about the substrate cleaning procedure and the growth techniques are given elsewhere.³³ 

Samples suitable for cross-sectional TEM observation were prepared by focused-ion beam (FIB) milling using a Thermo Fisher Helios 5UX dual-beam instrument, with initial thinning done at 30 keV and further thinning at 5 and 2 keV to minimize surface damage induced by ion milling. A Philips-FEI CM-200 FEG TEM operated at 200 kV, and an image-corrected FEI Titan 80–300 operated at 300 kV, were used for imaging, and a JEOL ARM200F operated at 200 kV was used for analysis by electron-energy-loss spectroscopy (EELS).

Cross-sectional TEM images of samples A1 and A2, which compare BN film nucleation at the BN/diamond interface, followed by growth in the intermediate region, and near the top surface, are shown in Fig. 1. The image of sample A1 grown with H₂= 1.5, BF₃ = 2 SCCM in the initial growth stage, and H₂= 2, BF₃ = 2 SCCM in the later stage, while maintaining H/F ratios of 0.75 and 1, respectively, demonstrates the presence of overlapped c-BN twin domains at the interface, and well-oriented c-BN grains in the intermediate and near-top surface regions of the film. The presence of the cubic BN phase is also confirmed by the {111} and {002} diffraction spots visible in the fast Fourier transforms (FFTs) of the respective images. In contrast, sample A2, grown with H₂= 4 SCCM, BF₃ = 4 SCCM throughout the experiment, while maintaining a H/F ratio = 1, shows nucleation of t-BN at the heterointerface. The disordered t-BN phase then continued to grow all the way to the top surface of the BN film.

Figure 1 also compares images taken from near-interface, intermediate, and near-surface regions of samples A1 and A3, which were grown at temperatures of 735 and 820 °C, respectively, together with the corresponding fast Fourier transforms (FFTs) of sample A1. The BN layers near the substrate surfaces of both samples show a complex appearance. Moreover, the corresponding FFTs show a pattern of spots that could at first sight be interpreted as corresponding to hexagonal-close-packing (HCP). However, further examination and analysis reveals that this appearance is yet another example of a well-known imaging artefact originating from the overlap of twins and stacking faults (SFs) in projection along the beam direction in nanograined materials, as recently explained in detail by Cayron.³⁴ A more detailed explanation of the origin of this complex contrast in the case of c-BN during HRTEM imaging will be given elsewhere.³⁵ Intermediate and near-surface regions in the film showed no apparent sign of any t- or h-BN phases and the predominance of the cubic phase in these regions was again confirmed by the {111} and {002} spots visible in the FFTs. Areas further away from the BN/diamond interface in sample A3 had FFTs (not shown here) with sharper {111} and {002} spots and no additional spots or streaking effects, indicating improved film quality in the intermediate and near-surface regions.

Figure 2 shows cross-sectional TEM images of sample A1 and sample A3. Both samples exhibit extensive regions of BN with darker contrast just above the heterointerface, corresponding to highly defective layers that contain twins and stacking faults. These regions occupy roughly comparable proportions of the two films. Thus, it appears that the increased growth temperature for sample A3 had minimal effect on reducing defect densities in regions near the BN/diamond interface. Furthermore, the film growth rate was unaffected by the change in temperature.

Samples A3 and A4 were prepared using almost identical conditions, as found here to be the most suitable for growth of the less defective cubic phase, but the substrates were subjected to different hydrogen plasma treatment. The substrate temperatures during surface cleaning were almost the same for sample C1 and sample C2 (760 and 750 °C). However, the duration of the H₂-plasma treatments was 90 and 10 min, respectively. The initial c-BN layers (⪅30 nm) were compared to assess any differences in morphology. Regions below and above the BN/diamond heterointerface in samples A3 and A4 can be compared in Figs. 3(a) and 3(b), respectively. The regions above the heterointerface in sample A3 displays a higher proportion of film with relatively darker contrast compared to sample A4 which are associated with high density of twins and stacking faults.

Figure 4 shows a comprehensive EELS spectrum analysis from the diamond-BN interface region of sample A3. Figure 4(a) is a spectrum image that combines EELS data for energies over the range from 180 to 460 eV. It is overlaid with the sp²-B (blue) and sp³-B (green) signal. The white bars in the image, labeled A, B, and C, are positioned on c-BN, interface, and diamond regions, respectively, and shadow the pixels that were used for generating the spectra shown in Fig. 4(b). The spectra confirm the presence of boron and nitrogen atoms to the left of the interface (growth direction), and carbon to the right of the interface, and also highlight the different bonding states of these elements. The prominent π* and σ* peaks of boron at around 191 and 197 eV, carbon at around 286 and 294 eV, and nitrogen at around 401 and 405 eV, which are characteristic of sp²- and sp³-bonding, respectively, are then used to provide sp²–sp³ maps of the region. The signal intensities of these maps are summed perpendicular to the growth direction and then normalized to produce the plots shown in Fig. 4(c). The sp² and sp³ curves of boron and nitrogen both reveal transitions in the type of bonding on moving from nucleation layer to the epilayer. Similarly, the carbon bonding changes from the sp³-bonded state of diamond cubic material to an sp²-bonded state close to the BN/diamond heterointerface. Notably, bluish regions observed near the interface in Fig. 4(a) correspond to the higher proportion of lower intensity sp² EELS signal, which reveals some local variability in the width of the transition region.

In order to estimate the effective width of the nucleation layer, the boron curves were selected for curve fitting since lower energy ranges in an EELS spectrum yield higher signal counts with lower background noise. Figure 4(d) shows the fitted gaussian and sigmoidal curves based on normalized signal intensities of the sp²- and sp³-bonded boron across the interface. The corresponding adjusted R-squared values are 0.991 and 0.999, respectively. Using the fitted sigmoidal curve based on the sp³-bonded boron signal intensities, the lateral separation from 10% to 90% of the highest intensity level was estimated to be 2.5 ± 0.3 nm. Since the effect of probe size and beam spreading have not been taken into account, this figure represents an upper estimate for the width of the nucleation layer where there are transitions in bonding characteristics for each of boron, nitrogen, and carbon.

In this study, it was not expected that different substrate orientations would contribute to a change in phase of the BN films because of the very small lattice mismatch with the substrate. Thus, the structure of the BN films grown on (111) surface (sample A1) and (001) surface (sample A2) can be compared based solely on the effect of different gas precursor concentrations and ratio.

An increase in the H₂ concentration increases the concentration of NH_x species, which is essential for BN film growth, while it decreases the concentration of free F radicals, which are crucial for the activation of H-terminated N surface atoms and for etching away of any non-cubic BN constituent.^42,43 The experimental results here confirm that a nucleation step with optimized H/F ratio and absolute concentrations of gas species during the growth step are both crucial for achieving the desired plasma chemistry that is needed to grow BN films with a dominant cubic phase.

The effect of etching due to different H₂/BF₃ ratios in the gas mixture was also apparent from the different film thicknesses of samples A1 and A2. Sample A1 had a greater concentration of F radicals in the gas mixture with H₂/BF₃ = 0.75 compared to sample A2 (H₂/BF₃ = 1), during the initial growth stage, and it had lower film thickness overall (150 nm vs 250 nm).

It has been established that a hydrogen-limited environment is more conducive for growth of the BN cubic phase.^9,41 Thus, the substrate temperature was varied here to determine any other morphological changes while keeping the same gas precursor concentrations. The primary defects observed in the resulting c-BN films were twins and stacking faults, albeit in different proportions.

Figure 2 presents a comparative analysis of the c-BN films in samples A1 and A3. Notably, the darker contrast observed in regions just above the BN/diamond interface is indicative of a high density of twin-related c-BN domains and stacking faults. Despite the higher growth temperature for sample A3, a comparative assessment of the size of these darker, defect-rich regions showed no significant difference. This observation suggests that the types and densities of defects occurring near the BN/diamond heterointerface are minimally affected by changes in growth temperature. In contrast, as observed in multiple TEM images of the two samples, regions farther from the heterointerface in A3 exhibited larger domain sizes compared to A1. This enlargement is most likely due to the enhanced thermal energy available to B and N adatoms at higher temperatures, which increases their in-plane mobility on the growth surface, leading to more ordered and hence less defective cubic phase growth. This difference was further shown by the sharper diffraction spots in the corresponding FFTs of these images, taken from identically sized regions while moving away from the substrate. To confirm this qualitative distinction, additional observations across extensive cross sections of both films are required.

It was observed that growth at higher temperatures resulted in larger c-BN domains and a reduction in defect density. However, the initial layers of c-BN were predominantly characterized by high densities of twin domains and stacking faults. EELS experiments were thus conducted to provide an understanding of the nucleation and early stages of c-BN layer growth, and its dependence on substrate surface quality.

The EELS analysis shown in Fig. 4 established that the initial layers of the film predominantly contained sp²-bonded BN, and transitioned to almost entirely sp³-bonded BN later in the growth process. Similarly, there is a shift of carbon bonding from exclusively sp³ in diamond to a mixture of sp² and sp³ in the nucleation layer. These observations suggest the possibility that the diamond surface undergoes some disordering either as a result of the cleaning process or during the early stages of BN growth. Other potential causes for the transition could be the influence of very low levels of residual hydrogen and fluorine contaminants or irradiation with H⁺ ions.

A mechanism has been proposed previously explaining how growth temperature can affect the development of multi-domain heteroepitaxial c-BN (111) structures grown on diamond (111) substrates.²⁶ This same mechanism can be adapted here to explain the formation and relative density of twin-related c-BN domains depending on the substrate cleaning method. The c-BN (111) plane is polar, while the bond energies for C–B and C–N are 444 and 305 kJ/mol, respectively,^44,45 suggesting that C–B bonds are more likely to form at the c-BN/diamond interface.

Figure 5(a) shows an HRTEM image of c-BN multi-domain structures near the c-BN/diamond interface in sample A4. Epitaxial c-BN domains are enclosed with yellow dashed lines and a rotated c-BN domain is demarcated with a teal dashed line. The red line segment(s) highlights the structural relationship and the presence of a lattice defect in the region. For N-polar c-BN (111), as sketched in Fig. 5(b), the stacking sequence would depend on the location of incoming nitrogen atoms. For sp³-bonded BN, the zincblende structure is more stable than the metastable wurtzite structure.^46,47 Considering probabilistic events, nitrogen atoms can attach at specific sites (a′) for the wurtzite structure during growth, resulting in stacking faults (SFs) that would, in turn, lead to the formation of rotational twins. This process can be considered as equivalent to inserting a single layer of wurtzite structure with the 2H stacking sequence into the 3C stacking sequence, as sketched in Fig. 5(c). The SF at the heterointerface inverts the 3C stacking sequence from “…cbacba” to “abcabc….” For a specific growth temperature, surface amorphization due to H₂-plasma cleaning is postulated to increase the potential barrier for the migration of N atoms to c sites (zincblende structure) from the a′ sites (wurtzite structure). This phenomenon could account for the observed higher proportion of twin-related c-BN domains (darker contrast regions above the heterointerface), especially in sample A3, which was subjected to prolonged cleaning, potentially leading to greater surface roughness.

The morphology of BN films grown on diamond substrates using ECR PECVD has been investigated using cross-sectional TEM, focusing on phase formation, defect characteristics, and structural evolution during the initial stages of growth. Within the studied range of growth temperatures (735–820 °C) and using a specific hydrogen plasma substrate cleaning method, it was established that the concentrations of gas precursors played a crucial role in determining whether the primary phase of the BN film was cubic or turbostratic. H₂-deficient growth (nucleation step) leads to essentially phase-pure c-BN. Twins and stacking faults were predominantly found in the bulk of c-BN films grown with a nucleation step. Higher growth temperatures resulted overall in a reduction in defect densities and larger c-BN domains, in regions of the film away from the interface. Surface cleaning proved to be critical for the nucleation of heteroepitaxial c-BN domains. Aggressive cleaning tended to roughen the substrate surface and contributed to a higher proportion of rotational twins. A plausible mechanism correlating surface roughness with rotation twin formation was presented. In order to further improve film quality and achieve single-domain epitaxial c-BN formation, it is recommended that the BN films should be grown at higher temperatures. The surface preparation should also be improved in order to minimize the potential barriers for N adatoms to migrate to stable sites more suitable for the growth of the zincblende structure.

This work was supported as part of ULTRA, an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, Basic Energy Sciences under Award No. DE-SC0021230. The authors acknowledge the use of facilities within the John M. Cowley Center for High Resolution Electron Microscopy at Arizona State University. We also thank Professor Martha R. McCartney for assistance with some microscopy.

The authors have no conflicts to disclose.

Saurabh Vishwakarma: Conceptualization (lead); Data curation (lead); Formal analysis (lead); Investigation (lead); Writing – original draft (lead). Avani Patel: Investigation (equal); Writing – review & editing (supporting). Manuel R. Gutierrez: Investigation (supporting); Writing – review & editing (supporting). Robert J. Nemanich: Investigation (equal); Supervision (equal); Writing – review & editing (equal). David J. Smith: Conceptualization (equal); Investigation (equal); Supervision (equal); Writing – original draft (equal); Writing – review & editing (equal).

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