Impact of BN Buffer Layer Thickness on Interfacial Structure and Band Alignment of a-BN/4H-SiC Heterojunctions

Abstract

This study provides a comprehensive investigation into the growth behavior of boron nitride (BN) buffer layers on Silicon carbide (SiC) substrates and their influence on interfacial band alignment. BN layers were deposited on semi-insulating SiC by RF magnetron sputtering with deposition times of 2.5, 5, and 7.5 min (these deposition times are specific experimental parameters to adjust the thickness of the amorphous BN layer, not intrinsic material properties of BN). Atomic force microscopy revealed that the surface roughness of the BN layers initially decreased and then increased with thickness, indicating an evolution from nucleation to continuous film formation, followed by surface coarsening. Transmission electron microscopy confirmed the BN thicknesses of approximately 3.25, 4.91, and 7.57 nm, showing that the layers gradually became uniform and compact, thereby improving the structural integrity of the BN/SiC interface. Band alignment was analyzed using the Kraut method, yielding a valence band offset of ~0.36 eV and a conduction band offset of ~2.34 eV for the BN/SiC heterojunction. This alignment indicates that the BN buffer layer introduces a pronounced electron barrier, effectively suppressing leakage, while the relatively small VBO facilitates hole transport across the interface. These findings demonstrate that the BN buffer layer enhances interfacial bonding, reduces defect states, and enables band structure engineering, offering a promising strategy for improving the performance of wide-bandgap semiconductor devices.

1. Introduction

Semi-insulating 4H-Silicon carbide (4H-SiC), a representative third-generation wide-bandgap semiconductor material, has attracted considerable attention for power and optoelectronic device applications owing to its wide bandgap of ~3.2 eV, high critical electric field of ~3 MV·cm⁻¹ [1,2,3], and outstanding thermal conductivity and thermal stability. Nevertheless, the performance of 4H-SiC-based devices is often constrained by interfacial states and structural defects, which reduce carrier mobility, increase channel resistance, and compromise device reliability [4,5,6]. Consequently, interface engineering aimed at improving the quality of the 4H-SiC/buffer or passivation layer interface has emerged as a key strategy to enhance device performance [7,8,9,10].

Amorphous boron nitride (a-BN), a layered material with a wide bandgap of ~5.9 eV [11,12], exhibits excellent thermal stability and chemical inertness, rendering it a promising candidate for wide-bandgap semiconductor heterostructures [13,14]. Previous studies have demonstrated that BN films, when employed as buffer layers, can significantly improve interface quality, suppress defect states, and modulate band alignment [12,15,16]. To clarify “surface defect states”, we refer here to point defects and local bonding disorders at the interface (e.g., vacancies, dangling bonds, F-centers, and pit/step features on SiC surfaces), which can introduce midgap states and promote carrier trapping or band bending [5,17], thereby optimizing the Schottky barrier height and electrical properties of devices. In particular, BN interlayers are expected to serve as effective interface modulation layers in wide-bandgap semiconductors such as SiC, enhancing both the power-handling capability and long-term stability of devices [18,19].

Although several studies have reported the growth of BN on SiC surfaces and the basic properties of BN/SiC heterostructures [15,20], systematic investigations of the effect of BN thickness on the interfacial structure and band alignment remain limited. Herein, BN films with different deposition times (2.5, 5, and 7.5 min) were deposited on SiC substrates by magnetron sputtering. Their surface morphology, interfacial structure, and chemical states were comprehensively characterized using Atomic force microscopy (AFM), Transmission electron microscopy (TEM), and X-ray photoelectron spectroscopy (XPS). Furthermore, the band alignment of BN/SiC heterojunctions was determined using the Kraut method, elucidating the influence of BN thickness on the valence band offset (ΔE_V) and conduction band offset (ΔE_C). These findings provide both experimental evidence and theoretical insights into the role of BN buffer layers in the interface engineering of SiC-based systems.

2. Experimental Procedure

2.1. Sample Preparation

The semi-insulating SiC substrates were used in this study. Prior to deposition, their surfaces were ultrasonically cleaned in organic solvents and rinsed with deionized water to remove contaminants. BN films were deposited by magnetron sputtering using a high-purity BN target. In the sputtering process, the RF power was set to 150 W, the chamber pressure was stabilized at 1 Pa, and the substrate temperature was maintained at ~200 °C. A mixed Ar/N₂ atmosphere with a flow ratio of 30:10 was employed, and the target was pre-sputtered for 20 min to eliminate surface impurities and stabilize the plasma environment. Film thickness was tuned by varying the deposition time to 2.5, 5, and 7.5 min under identical conditions, ensuring that thickness was the only variable. A sample with 0 min deposition (bare SiC substrate) served as the control to evaluate the influence of BN insertion on the interfacial structure and band alignment. After deposition, all samples were immediately transferred and stored in a clean environment to minimize contamination before characterization.

2.2. Characterization Methods

The samples were characterized in terms of morphology, structure, and electronic properties. The surface morphology of the SiC substrates and BN films with different deposition times was examined using anAFM (Dimension ICON, Bruker, Billerica, MA, USA), providing two-dimensional (2D) and three-dimensional (3D) images along with height distribution profiles; all AFM images were acquired over 5 × 5 μm² scan areas. For each sample, three independent regions were randomly selected and measured. The reported surface roughness values are given as the mean ± sample standard deviation (mean ± SD) of these three measurements. Surface roughness and related parameters were extracted to quantitatively analyze the influence of BN thickness on morphological evolution. Cross-sectional observations were obtained using TEM (Talos F200X, Thermo Fisher Scientific, Waltham, MA, USA) to evaluate film continuity, interface features, and thickness. Furthermore, TEM specimens were prepared usingdual-beam focused ion beam (FIB) system (Helios 5 CX, Thermo Fisher Scientific, Waltham, MA, USA) to ensure clear visualization of the interface. Chemical composition and bonding states were investigated by XPS (AXIS Supra+, Shimadzu, Kyoto, Japan) with a monochromatic Al Kα source (1486 eV). Survey scans were acquired with a step size of 1 eV, while high-resolution scans were recorded with a step size of 0.05 eV. Surface contamination resulted in a C 1 peak at 284.8 eV, which was used for charge correction, while 5 keV Ar⁺ ion sputtering provided depth profiling, a setting optimized for efficient removal of surface contamination and for depth profiling in our instrument. XPS data were employed to analyze the composition and bonding of BN films, as well as the band alignment of the BN/SiC heterojunction. The peak value of the valence band was determined through linear extrapolation of the valence band spectrum, and band offsets were calculated using the Kraut method.

3. Results and Discussion

3.1. Analysis of Surface Morphology and Structural Characteristics

Figure 1a presents the AFM-based characterization of the SiC substrate before BN deposition, including 2D surface morphology, 3D topography, and height profiles [21]. The SiC surface appears smooth overall, exhibiting only a few uniformly distributed step-like features and shallow grooves, which originate from the intrinsic steps of the SiC crystal and residual polishing effects [22,23]. The 3D topography further reveals minimal surface undulation, with an absence of large particles or island-like features, confirming the flatness and uniformity of the substrate.

Height profiles reveal that surface variations are primarily confined to the nanometer scale and change gradually, with a surface roughness of 0.089 ± 0.018 nm, representing an exceptionally smooth surface. This highly smooth surface provides favorable conditions for subsequent BN nucleation and epitaxial growth, facilitating the formation of uniform and continuous thin films.

Compared with the bare SiC substrate, the BN film deposited for 2.5 min exhibits significant changes in surface morphology (Figure 1b). The 2D AFM image shows numerous uniformly distributed granular structures [24,25,26], indicating that BN nucleates in an island-like form on the SiC surface during the initial deposition stage. The 3D topography further reveals height variations among these granular features, resulting in an overall increase in surface undulation. Height profile analysis indicates that the surface roughness has increased notably compared with the bare substrate, reaching 0.267 ± 0.03 nm, with more pronounced peak-to-valley variations. These observations suggest that, at the early stage, the BN film only but exists as dispersed islands. The morphology is consistent with a typical “nucleation–island growth” mode, in which BN preferentially nucleates at regions of lower local energy or surface defects and gradually expands to form larger, interconnected domains. This early-stage island structure establishes the foundation for the subsequent formation of a continuous BN film.

Furthermore, Figure 1c depicts the surface morphology of the BN film after 5 min of deposition. The 2D AFM image shows that the gaps between surface particles have decreased [27], and the film exhibits higher coverage, indicating the partial growth and coalescence of the islands. The 3D topography clearly demonstrates a transition from the initially isolated islands to a more continuous film, accomplished by reduced surface undulation compared to the 2.5 min sample. Furthermore, height profile analysis reveals a reduction in surface roughness to 0.225 ± 0.01 nm, with peak-to-valley variations becoming more moderate, indicating a denser surface. Based on nucleation and growth mechanisms, extended deposition drives the BN film from the “island nucleation stage” to the “island coalescence” stage, promoting continuous coverage by the lateral growth and merging of particles. At this stage, enhanced surface flatness facilitates interface formation and enables subsequent band alignment studies.

The surface morphology of the BN film further evolved after the deposition time was extended to 7.5 min (Figure 1d). The 2D AFM image shows that the film has achieved near-complete coverage, with no exposed substrate areas and overall uniform morphology [28]. However, compared with the 5 min sample, localized regions exhibit larger protrusions, suggesting additional particle accumulation or secondary nucleation following continuous layer formation. The 3D topography reveals increased surface undulation relative to the 5 min sample, demonstrating that excessive deposition induced local roughness. This observation is further confirmed by height profile analysis, showing more pronounced peak-to-valley variations and an increase in surface roughness to 0.285 ± 0.008 nm. Based on growth mechanisms, the BN film at 7.5 min enters an “overgrowth” stage. Although the layer is continuous, atomic diffusion and local secondary nucleation lead to a rebound in roughness. This evolution indicates that the optimal surface flatness is achieved at intermediate deposition time (~5 min), whereas excessive deposition results in increased roughness.

Furthermore, Figure 2 presents cross-sectional TEM images of the SiC substrate and BN films deposited on the substrate for different durations [22,29,30]. The bare SiC substrate (Figure 2a) displays a dense crystalline structure with well-defined layers, confirming good crystallinity and surface flatness prior to deposition. In contrast, Figure 2b–d show the cross-sectional morphologies of BN films sputtered on SiC for 2.5, 5, and 7.5 min, respectively.

From the TEM analysis, the thicknesses of BN layers at different deposition times are 3.25 nm, 4.91 nm, and 7.57 nm, respectively. These thickness values were extracted from representative cross-sectional TEM images and may have an uncertainty of approximately 0.2-0.3 nm. At 2.5 min (Figure 2b), the BN layer exhibits local discontinuities and non-uniform thickness [31,32], forming island-like thickened regions, consistent with the higher surface roughness observed using AFM. Prolonging the deposition time to 5 min (Figure 2c) produces a thicker and more continuous BN layer with a relatively flat interface, indicating the merging of nucleated islands into a dense coverage. Furthermore, complete coverage of the BN layer is observed at 7.5 min (Figure 2d), although localized thickness variations and protrusions indicate secondary nucleation or overgrowth following continuous film formation. These results align with the increase in the roughness measured by AFM. Based on these results, the estimated deposition rates are ~1.25, 0.99, and 1.02 nm·min⁻¹ for 2.5, 5, and 7.5 min, respectively, with an average rate of ~1.09 nm·min⁻¹. This indicates that under identical sputtering conditions, the BN film grows at a generally stable rate of ~1 nm·min⁻¹. Overall, the interface remains dense and well-bonded, demonstrating the stable deposition characteristics of BN on SiC under the applied magnetron sputtering conditions.

3.2. Study of Band Alignment in BN/SiC Heterojunctions

In wide-bandgap semiconductor heterostructures, the band alignment directly governs carrier transport across the interface, ultimately affecting device performance [5,33,34]. The ΔE_V and ΔE_C govern the interfacial charge distribution and barrier height, also determining key device parameters such as the Schottky barrier height [35,36], breakdown voltage, and photodetection range. For SiC-based devices, performance is often constrained by carrier scattering and recombination losses, arising from interface states and band misalignment. Therefore, the introduction of an appropriate buffer layer and its precise tuning offers a promising approach to optimize the band alignment at the BN/SiC interface, thereby enhancing both electrical and optoelectronic properties.

Additionally, the synthesized materials were analyzed using XPS. Figure 3 presents the XPS results of BN films deposited on SiC substrates [37,38], including the survey spectrum, valence band spectrum, and high-resolution core-level spectra. In the survey spectrum (Figure 3a), distinct B 1s and N 1s peaks are observed, indicating the film is composed mainly of boron nitride. Slight surface oxidation or environmental contamination leads to weak O 1s and C 1s signals.

The valence band spectrum (Figure 3b) shows a valence band maximum (VBM) of 2.68 ± 0.05 eV with respect to the Fermi level, determined by linear extrapolation, the uncertainty in VBM determination, including fitting residuals and instrument resolution, is estimated at ±0.05 eV, consistent with the wide-bandgap nature of BN and providing essential parameters for subsequent band alignment analysis. The high-resolution N 1s spectrum (Figure 3c) can be deconvoluted into three components. A main peak at 398.57 eV corresponding to N-B bonds, accounting for ~81.6% of the signal and indicating that the BN film is predominantly composed of B-N bonds. Furthermore, a shoulder at 400.20 eV (~15.2%) attributed to N-C bonds, possibly originating from the substrate or residual contamination. Another minor component at 401.84 eV (~3.2%) is assigned to N-O bonds, arising likely due to slight surface oxidation. Similarly, the high-resolution B 1s spectrum (Figure 3d) exhibits clear peak splitting. The main peak at 190.42 eV corresponds to B-N bonds (~81.9%), confirming the dominant B-N composition of the film. The two minor components are also observed at 188.90 eV (B-C, ~13.8%) and 191.55 eV (B-O, ~4.3%), indicating the presence of trace impurity bonds or surface oxidation.

XPS analysis of the BN films indicates that the predominant bonding is B-N, with minor contributions from B-C, N-C, and oxidation-related B-O and N-O bonds. The small amounts of carbon- and oxygen-related bonds likely arise from interactions with the substrate or the environment during film growth; however, their overall content is low and does not affect the intrinsic properties of the boron nitride film. Previous studies report reference values for SiC substrates, supplying essential energy-level information, including the valence band maximum, which is critical for investigating the band alignment in BN/SiC heterojunctions, as summarized in

Table 1. Bandgap, valence band maximum (VBM), core-level binding energies, and the energy difference between the core levels and the corresponding VBM energy in BN and SiC samples.

Sample	Eg	EV	ECL	ECL − EV
BN	5.9 eV	2.68 eV	398.57 eV (N 1s)	395.89 eV
4H-SiC	3.2 eV	1.93 eV	152.04 eV (Si 2s)	150.11 eV

.

Additionally, the synthesized BN/SiC interface was examined by XPS (Figure 4). The high-resolution core-level spectra and peak deconvolution reveal the chemical bonding and interfacial interaction mechanisms [39].

In the N 1s spectrum, the dominant peak at 398.03 eV corresponds to N-B bonds (88.3%), indicating the predominant bonding between nitrogen atoms and boron at the interface, thereby preserving the primary BN structure. A shoulder at 399.34 eV is attributed to N-Si bonds (~12.7%), suggesting a certain degree of chemical interaction between BN and the SiC substrate through bonding between nitrogen and silicon. In the B 1s spectrum, the main peak at 190.80 eV corresponds to B-N bonds (83.6%), confirming the predominant presence of BN bonding at the interface. A minor component at 189.50 eV (~16.4%) is attributed to B-C bonds, indicating the possibility of bonding between boron atoms and carbon atoms on the SiC surface. In the Si 2s spectrum, the principal peak at 151.89 eV is assigned to Si-C bonds (~92.5%), representing the primary SiC lattice structure, while a minor peak at 153.85 eV (~7.5%) corresponds to Si-O bonds, likely arising from slight oxidation at the interface or surface. In the C 1s spectrum, the primary peak at 283.13 eV corresponds to C-Si bonds (~87.8%), consistent with the SiC lattice, while additional peaks at 284.80 eV (C-C, ~9.5%) and 285.68 eV (C-N, ~2.7%) are also observed. The C-C peak originates from minor surface contamination, whereas the presence of C-N bonds indicates partial chemical interactions between carbon atoms and nitrogen in the BN film, providing further insight into interfacial reactions. The interfacial chemical species we observe (B-C, N-Si and minor Si-O) are consistent with localized chemical interactions that likely coexist with defect states and influence the electronic structure. Further work, including targeted defect spectroscopy and electrical characterization (e.g., EPR, DLTS, STM/STS), is needed to quantify their roles for device performance [5,17]. And the weak Si-O signal detected by XPS suggests the potential presence of an extremely thin oxide layer at the interface. Although its impact on band shifts is limited, this interfacial layer may introduce minor deviations, necessitating more systematic verification in subsequent studies using methods such as TEM-EELS or XRR. Additionally, in future experiments we will include low-energy sputter tests, angle-resolved XPS (ARXPS) and complementary techniques such as AES or Raman spectroscopy to further validate the chemical assignments.

Overall, the XPS peak deconvolution of the BN/SiC interface reveals distinct chemical interactions. The BN film retains a stable structure dominated by B-N and Si-C bonds. Conversely, the detection of N-Si, B-C, and C-N bonds indicates localized chemical reactions occurring at the interface. These bonding features suggest the formation of strong interfacial coupling between BN and SiC, enhancing adhesion, modulating dipoles and band offsets at the interface, and thereby providing essential information for subsequent band alignment calculations.

As shown in Figure 5, the measured binding energies of N 1s and Si 2s are 398.03 eV and 151.89 eV, respectively, resulting in a core-level energy difference (ΔE_CL) of 246.14 eV.

According to the Kraut method, the ΔE_V of the BN/SiC heterojunction is determined by analyzing the difference in energy between VBM and the corresponding core levels (E_CL), as described below [40,41,42,43]:

$$\Delta {\mathrm{E}}_{\mathrm{V}}=\left({{\mathrm{E}}_{\mathrm{N}1\mathrm{s}}^{\mathrm{B}\mathrm{N}}-\mathrm{E}}_{\mathrm{V}}^{\mathrm{B}\mathrm{N}}\right)-\left({{\mathrm{E}}_{\mathrm{S}\mathrm{i}2\mathrm{s}}^{\mathrm{S}\mathrm{i}\mathrm{C}}-\mathrm{E}}_{\mathrm{V}}^{\mathrm{S}\mathrm{i}\mathrm{C}}\right)-\Delta {\mathrm{E}}_{\mathrm{C}\mathrm{L}}$$

(1)

$$\Delta {\mathrm{E}}_{\mathrm{C}\mathrm{L}}=\left({{\mathrm{E}}_{\mathrm{N}1\mathrm{s}}^{\mathrm{B}\mathrm{N}}-\mathrm{E}}_{\mathrm{S}\mathrm{i}2\mathrm{s}}^{\mathrm{S}\mathrm{i}\mathrm{C}}\right)$$

(2)

In Equation (1), the first and second terms represent the energy differences between the core levels and the VBM of the BN film and the SiC substrate, respectively. Furthermore, in Equation (2), ΔE_CL corresponds to the difference in energy between the core levels of BN and SiC at the heterojunction interface. Using the measured values of ΔE_g and ΔE_V, the CBO ($\Delta {\mathrm{E}}_{\mathrm{C}})$ at the BN/SiC interface is calculated according to Equation (3):

$$\Delta {\mathrm{E}}_{\mathrm{C}}={\mathrm{E}}_{\mathrm{g}}^{\mathrm{S}\mathrm{i}\mathrm{C}}{-\mathrm{E}}_{\mathrm{g}}^{\mathrm{B}\mathrm{N}}-\Delta {\mathrm{E}}_{\mathrm{V}}$$

(3)

${\mathrm{E}}_{\mathrm{g}}^{\mathrm{S}\mathrm{i}\mathrm{C}} \mathrm{a}\mathrm{n}\mathrm{d} {\mathrm{E}}_{\mathrm{g}}^{\mathrm{B}\mathrm{N}}$ represent the bandgaps of the SiC substrate and the BN film at room temperature, respectively. The calculated ΔE_V is −0.36 ± 0.08 eV, and the ΔE_C is −2.34 ± 0.10 eV [41,43]. The uncertainties were derived from error propagation of VBM fitting, core-level fitting, charging correction, and bandgap variability. To verify the reliability of the present results, a comparative analysis was conducted by correlating with the density functional theory (DFT) simulation study reported by Bo et al. [16]. In their work, the CASTEP software combined with the HSE06 hybrid functional was employed to construct van der Waals (vdW) heterojunction models of BN/SiC with lattice strains of 7.37% and 5.93%, respectively. The calculated band gaps for these two models were 0.851 eV and 1.373 eV, which revealed a clear trend that “an increase in lattice strain leads to a decrease in band gap”. This conclusion is in good agreement with the experimental phenomena observed in this study.

Figure 6 presents a schematic of the BN/SiC heterojunction band alignment, calculated using the Kraut method. The results indicate that the VBM and CBM of BN are shifted by ~0.36 eV and ~2.34 eV, respectively, relative to SiC. This band alignment corresponds to a type-I heterojunction, with a large CBO forming a significant barrier to electron transport, while the VBO is comparatively small. For devices, the substantial CBO (~2.34 eV) effectively suppresses electron leakage at the interface and minimizes interfacial recombination, thereby enhancing breakdown voltage and overall stability. Meanwhile, the modest VBO (~0.36 eV) presents a low barrier to hole transport at the interface. This asymmetric band offset is advantageous for optimizing carrier distribution and regulating interfacial charge transport, making the BN layer a key functional component for enhancing the performance of SiC-based power and optoelectronic devices.

4. Conclusions

This study systematically investigates the growth characteristics of BN buffer layers on SiC substrates and their influence on interfacial band alignment. AFM measurements reveal that the surface roughness of BN films initially decreases and then increases with deposition thickness, reflecting a progression from initial nucleation to continuous coverage, followed by surface coarsening. TEM analysis further shows that, with increasing deposition time, the BN film thickness sequentially grows to around 3.25 nm, 4.91 nm, and 7.57 nm, gradually forming a uniform and dense layer, enhancing the structural integrity of the BN/SiC interface. Moreover, XPS characterization indicates that B-N bonds dominate the BN film, with minor B-C, N-Si, and C-N bonds present at the interface, confirming localized chemical interactions between BN and SiC. Based on the Kraut method, band alignment calculations reveal the $\Delta {\mathrm{E}}_{\mathrm{V}}$ of approximately 0.36 ± 0.08 eV and the $\Delta {\mathrm{E}}_{\mathrm{C}}$ of 2.34 ± 0.10 eV for the BN/SiC heterojunction. These results demonstrate that the BN buffer layer establishes a substantial electron barrier at the interface while maintaining a relatively low valence band barrier. This asymmetric band alignment effectively suppresses electron leakage and minimizes interfacial recombination, providing a potential strategy for tuning carrier transport. In summary, the BN buffer layer plays a promising role in enhancing the structural stability of the SiC interface, reducing interfacial defect density, and enabling precise band engineering. This study provides a practical strategy and experimental foundation for interface optimization in SiC-based wide-bandgap semiconductor devices, delivering valuable guidance for improving the performance of future high-power and optoelectronic devices.