Insight into the energy level structure and luminescence process of color centers at SiO₂/SiC interfaces

A single-photon emitter (SPE) that can emit a photon with high probability at desired times is important in various applications, including quantum computing, communication, and sensing. Optically active point defects (i.e., color centers) in semiconductors are an ideal candidate for SPEs. In this regard, the most studied defect is the negatively charged nitrogen-vacancy (NV) center in diamond. Quantum interference between indistinguishable photons emitted by remote NV centers has been demonstrated.¹ The NV center is also ideal for nanoscale sensing and imaging applications that use electron spins.² In addition to diamond, silicon carbide (SiC) has shown potential as a host for color centers due to its well-established crystal growth, conductivity control, and process technologies.³ Various color centers such as negatively charged silicon vacancy (V_Si),^4–7 neutral divacancy (V_CV_Si),^8,9 and neutral nitrogen-vacancy complex (N_CV_Si)^10–12 have been reported thus far in SiC. Apart from these bulk color centers, highly bright color centers were found at the interface between silicon dioxide (SiO₂) and SiC.¹³ Their radiative lifetime is estimated to be as short as about 2 ns,¹⁴ resulting in a brightness higher than the diamond NV center at room temperature.¹⁵ Although these interface color centers can be easily formed by the thermal oxidation of SiC, each color center should be spatially separated to suppress the overlap in their luminescence.^16,17 The second-order correlation value, g⁽²⁾(0), of photons emitted from interface color centers increases due to the luminescence overlap, which makes individual color centers no longer function as SPEs. Thus, having control over the density of interface color centers is critical for them to function as SPEs. Post-oxidation annealing in nitric oxide (NO),¹⁸ argon (Ar),¹⁹ and carbon dioxide (CO₂)¹⁶ has been reported to be effective in controlling the density of interface color centers. Density control was also achieved by changing the oxidation temperatures.^17,20 An advantage of interface color centers over bulk color centers is the compatibility with evolving large-scale integration technology. Interface color centers can be readily implemented in metal–oxide–semiconductor (MOS) devices, enabling electrical control over their optical properties.^15,21 Because the interface color centers are naturally localized at the interface, there is no need to control their position in three dimensions as with bulk color centers, which seems advantageous for device integration. Despite the favorable properties of interface color centers, their photoluminescence (PL) spectrum strongly varies from one defect to another,^{15–17,19,20} hindering their use in practical applications. To control the PL spectrum and increase photon emission rates, integrating interface color centers into optical cavities such as nanopillars and photonic crystals, as demonstrated for bulk color centers, is a promising solution.^22,23 A method to create intended color centers is required for photonic integration. However, the origin of the color centers has yet to be clarified. It is known that a high density of electron traps is located at the SiO₂/SiC interface, especially near the conduction band edge (E_C) of SiC.^24–27 Carbon-related defects formed by the thermal oxidation of SiC have been discussed as their possible origin both experimentally^28–30 and theoretically.^31–33 Thus, they are likely a candidate of the color centers found at the SiO₂/SiC interface. A recent theoretical study calculated the electronic and optical properties of carbon cluster emitters in SiC as possible candidates of the color centers.³⁴ However, the lack of experimental information on the color centers makes it difficult to identify their origin. While studies have found that the color centers originate from deep electron traps,^16,20 the energy levels responsible for the luminescence are completely unknown. Identifying the energy level structure is crucial in understanding the origin of the interface color centers and controlling their optical properties.

In the present study, we investigated the defects at the SiO₂/SiC interface both electrically and optically, aiming to clarify the energy level structure and luminescence process of the color centers found at the interface. A set of samples with various defect concentrations were fabricated to draw a comprehensive picture of the electronic and optical properties of the color centers. We then varied the oxidation temperature and oxygen partial pressure [P(O₂)] over a wide range and prepared the SiO₂/SiC samples. Here, samples that did not exhibit active oxidation³⁵ and had a smooth surface and interface were carefully selected for further investigation. The correlation between the color centers in confocal PL mapping images and the electrical defects evaluated from the capacitance–voltage (C–V) characteristics of SiC metal–oxide–semiconductor (MOS) capacitors was investigated to reveal the properties of color centers.

We used 4H-SiC(0001) substrates with n-type epilayers (donor density: 1 × 10¹⁶ cm⁻³). The samples were cleaned with acetone, piranha, SC1, SC2, and hydrofluoric acid and rinsed with ultrapure water. After that, SiO₂ films were formed by thermal oxidation at temperatures ranging from 1400 to 1600 °C in an Ar-diluted oxygen ambient [P(O₂): 0.05%–100%] as shown in Fig. 1(a). The oxidation time was controlled for each oxidation condition so that the SiO₂ thickness fits within 9–23 nm. As shown in the optical microscope images in Figs. 1(b) and 1(c), the samples oxidized at 1600 °C under P(O₂) of 0.05%–0.1% exhibited rough surfaces due to the active oxidation. These two samples were then excluded from further investigation. For electrical measurements, circular aluminum (Al) gate electrodes (diameter: 100 μm) and Al Ohmic contacts were formed by vacuum evaporation to fabricate the SiC MOS capacitors. The oxide on the back side of the samples was removed by scratching the surface with a diamond pen before forming the Ohmic contacts. Then, C–V measurements were performed to evaluate the interface properties. The probe frequency was set to 1 MHz in the measurements. Optical measurements were performed on SiO₂/SiC samples without electrodes. We used a home-built confocal microscope setup as described in our previous study.¹⁶ The setup is equipped with a green excitation laser (wavelength: 532 nm), an objective lens (NA = 0.9), and silicon-avalanche-photodiode-based single-photon-counting modules. The objective lens is placed on top of a piezo stage, which enables three-dimensional scanning of PL signals. The setup also has a functionality to measure the correlation between the photons with Hanbury-Brown and Twiss (HBT) measurements. Using this setup, PL mapping images were acquired by measuring the luminescence of wavelengths above 600 nm. The excitation laser power was set to 0.5 mW in the measurements, and the instruments for optical measurements were controlled by the MAHOS software.³⁶ All measurements were performed at room temperature.

First, electrically active defects at the SiO₂/SiC interface were evaluated from C–V measurements. Figure 2 shows the typical C–V characteristics of samples oxidized at 1500 °C under different P(O₂) conditions. The MOS capacitors formed by oxidation under 100% P(O₂) exhibited a positive shift in the flatband voltage (V_FB) due to a plateau in the characteristics. This indicates that a high density of defects is present at the interface. As the P(O₂) lowered, the plateau was reduced, and the V_FB approached its ideal value. Hence, the reduction in the interface defects was achieved by lowering the P(O₂). Thus, we successfully prepared a set of samples with different defect concentrations by controlling the oxidation conditions.

Next, optically active defects at the SiO₂/SiC interface were evaluated by confocal PL mapping measurements. Figures 3(a)–3(d) show the confocal PL mapping images of samples oxidized at 1500 °C under different P(O₂) conditions. The sample oxidized under 100% P(O₂) exhibited numerous luminescent spots in the image. As the P(O₂) decreased, the number of spots was reduced. To quantify the luminescence, we evaluated the histograms of PL intensity in the mapping images as shown in Fig. 4. The sample formed by oxidation under 100% P(O₂) showed a broad histogram with a maximum frequency at about 16 kcps. As the P(O₂) was lowered, the count rates of the spots were reduced, and the distribution became narrower. When P(O₂) reached 0.05%, the peak count rate (2.0 kcps) was similar to that of the as-cleaned SiC surface without SiO₂ (1.8 kcps). To summarize, the color centers were reduced by lowering P(O₂), similarly to the electrically active defects (Fig. 2). For the typical spots, we confirmed by HBT measurements that they have g⁽²⁾(0) values below 0.5, confirming that the color centers observed in this study function as SPEs when the overlap in their luminescence is suppressed. Although we did not measure the emission wavelength of color centers in this study, we found in our previous studies that varying the oxidation or annealing conditions does not resolve the variation in the emission wavelength.^16,20 The room-temperature PL spectra vary approximately in the same energy range (1.6–2.2 eV), regardless of oxidation or annealing conditions. The generation of interface color centers without emission wavelength variation remains an issue to be resolved.

Next, we examined the correlation between the density of the electrically active defects and the PL intensity of the color centers at the SiO₂/SiC interface. To measure electrically active defects, we evaluated the effective fixed charge density (N_eff) from the shift of V_FB from its ideal value. V_FB was determined as the voltage where the capacitance equals the theoretical flatband capacitance. In the case of n-type SiC MOS structures with a high density of traps, N_eff generally represents the total amount of acceptor-type interface defects. Figure 5(a) shows the oxidation temperature and oxygen partial pressure dependence of the effective fixed charge density (N_eff). N_eff decreased as the oxidation temperature increased, but the change was rather limited; for example, for P(O₂) = 2%, N_eff was reduced from 2.6 × 10¹² to 1.5 × 10¹² cm⁻² as the oxidation temperature increased from 1400 to 1600 °C. In contrast, N_eff was significantly affected by P(O₂); for example, for oxidation at 1500 °C, N_eff was reduced from 6.5 × 10¹² to 4.3 × 10¹¹ cm⁻² as the P(O₂) was lowered from 100% to 0.05%. As a measure of color centers, we evaluated the integrated PL intensity by summing the PL intensities of each pixel in the mapping image. Figure 5(b) shows the integrated PL intensity as functions of oxidation temperature and oxygen partial pressure. A comparison of the optical measurement [Fig. 5(b)] and electrical measurement [Fig. 5(a)] results shows that they seem to be closely correlated. To visualize this clearly, Fig. 6 plots the integrated PL intensity as a function of N_eff. The integrated PL intensity was found to strongly correlate with N_eff with a correlation coefficient of 0.95.

We then discuss the luminescence process of color centers. Since we used n-type samples, the bulk Fermi level of SiC is located near the E_C at room temperature (i.e., E_C − 0.18 eV). In such a case, the acceptor level (Fig. 2) is filled with electron(s), and an optical transition is expected between the acceptor level and the E_C, which should give rise to PL with an energy no higher than 0.92 eV. This is because the energy of PL is generally smaller than the energy difference between the electronic energy levels due to the effect of electron–hole coupling. However, the energy of PL from the color centers at the SiO₂/SiC interface is about 1.6–2.2 eV (560–780 nm)^{15–17,19,20} at room temperature. To resolve this contradiction, we propose that band bending exists in near-interface SiC as shown in Fig. 10. If band bending is present, the Fermi level near the interface becomes lower, and the acceptor level can become empty. As shown by the results of C–V measurements of MOS capacitors (Fig. 2), the V_FB is slightly positively shifted compared with the ideal position even before the acceptor level is filled, regardless of the oxidation conditions. This indicates the presence of negative fixed charges at the SiO₂/SiC interface, which causes the band bending. Therefore, it is reasonable that the band bending also exists for the SiO₂/SiC samples without electrodes used in the PL measurements. Under the assumptions that the surface bending exists and that the acceptor level is empty, we built a model to explain the PL of the color centers. Figures 11(a) and 11(b) show the schematic single particle energy diagrams describing the possible luminescence process of color centers. Strictly speaking, the charge transition level does not necessary match the energy level in the single particle picture, but the acceptor level should be located at about (E_C − 0.65)–(E_C − 0.92) eV. We consider both cases where the detected acceptor level is or is not involved in the optical transition. In the former case [Fig. 11(a)], the transition occurs between the E_V or a defect level close to it and the acceptor level. For the latter [Fig. 11(b)], the transition occurs between a defect level in the lower half of bandgap and the E_C. The presented models reasonably explain the relatively large PL energy (∼1.6–2.2 eV)^{15–17,19,20} of the color centers at the SiO₂/SiC interface. Considering the energy level of color centers (Figs. 9 and 10), switching the luminescence with charge state control should be possible by controlling the Fermi level of SiC near the interface. In particular, either the luminescence is switched off or the change in its wavelength is expected by filling the acceptor level with electron(s) with the application of external bias. If another hidden defect level is present [Fig. 11(b)], the luminescence is also affected when the Fermi level crosses the defect level. Such a property is favorable for electrically controlling/switching the single-photon emissions from the color centers at the SiO₂/SiC interface.

Based on the information collected in the present study and from the literature, we discuss the possible candidates of color centers found at the SiO₂/SiC interface. According to our results, the color center should have a single acceptor transition level at energies within (E_C − 0.65)–(E_C − 0.92) eV (Fig. 9). Given the fact that the emission and absorption dipoles of the color centers are aligned with the SiC crystal,^14,17 it is likely that they are located in near-interface SiC (not in SiO₂). Table I summarizes the (0/−) charge transition level of the interface color center obtained in this study together with that of the candidate defects in bulk SiC determined by theoretical calculations. For candidate defects, only those that have a single acceptor level in the upper half of bandgap are included in this table. As discussed in the introduction, carbon-related defects are a primary candidate of the color center. Among the carbon-related defects, the dicarbon antisite [(C₂)_Si] in its high-spin ground state has a (0/−) acceptor level at (E_C − 0.65)–(E_C − 0.77) eV,³⁴ agreeing well with the experimentally detected acceptor level [(E_C − 0.65)–(E_C − 0.92) eV] in the present study. The calculated zero-phonon lines of this defect is at 1.54–1.60 eV,³⁴ smaller compared to the experimentally observed PL energy (∼1.6–2.2 eV).^{15–17,19,20} However, since the local structure of defects can change in the vicinity of the interface due to the presence of stress,³⁸ the PL energy of defects may differ from those of the bulk defects. Thus, we consider this defect as a main candidate of color centers found at the SiO₂/SiC interface. A recent deep level transient spectroscopy study has also discussed the formation of (C₂)_Si by the oxidation of SiC.³⁹ While larger carbon clusters are also possible, most of them only create defect levels near E_V, and few have levels in the above half of the bandgap.³⁴ Among native point defects in SiC, carbon vacancy (V_C) at the hexagonal site has a single (0/−) level at E_C − 0.70 eV.⁴⁰ Oxygen-related defects have also been discussed as possible candidates of color centers in bulk SiC^41,42 and at the SiO₂/SiC interface.^43,44 According to a calculation of oxygen-related defects in SiC using a hybrid functional, (O_C)₂V_Si and O_C(C_Si)₂ have acceptor levels at E_C − 0.04 eV and E_C − 0.37 eV, respectively.⁴² Otherwise, there are few defects that have a single acceptor level in the upper half of the bandgap. Thus, we consider the (C₂)_Si defect as the main candidate of color centers at the SiO₂/SiC interface, although further theoretical considerations and experimentation are required to support this model.

This study investigated the correlation between the electrically active defects and the color centers at the SiO₂/SiC interface to clarify the optical and electrical properties of the color centers. SiO₂/SiC structures were formed under various oxidation temperature and oxygen partial pressure conditions to prepare a set of samples with various defect concentrations and draw a comprehensive picture of the characteristics of the color centers. The results indicated that the integrated PL intensity of color centers is closely correlated with the effective fixed charge density at the interface. The effective fixed charges are mostly related to an acceptor level that causes a plateau in the C–V characteristics, which is located at energies within (E_C − 0.65)–(E_C − 0.92) eV. Therefore, the origin of color centers lies in a deep acceptor-type defect. Based on the results obtained, we discussed the luminescence process of color centers at SiO₂/SiC interfaces and their possible origin referring to previous theoretical studies. Our experimental results contribute to clarifying the microscopic origin of highly bright color centers found at the SiO₂/SiC interface. To use the interface color centers as a quantum light source, further studies are needed to suppress the variation in their emission wavelength and to form them at the intended locations. Once these issues are resolved, the implementation of the color centers in metal–oxide–semiconductor devices will pave the way for scalable quantum applications.

This work was partly supported by JST, PRESTO (Grant No. JPMJPR22B5) and JSPS KAKENHI (Grant No. 24H00046).

The authors have no conflicts to disclose.

Kentaro Onishi: Data curation (lead); Formal analysis (lead); Investigation (equal); Validation (equal); Visualization (equal); Writing – original draft (lead); Writing – review & editing (equal). Takato Nakanuma: Formal analysis (supporting); Investigation (equal); Validation (equal); Visualization (equal); Writing – review & editing (equal). Haruko Toyama: Investigation (equal); Validation (equal); Writing – review & editing (equal). Kosuke Tahara: Investigation (equal); Software (lead); Validation (equal); Visualization (equal); Writing – review & editing (equal). Katsuhiro Kutsuki: Investigation (equal); Resources (equal); Validation (equal); Writing – review & editing (equal). Heiji Watanabe: Funding acquisition (equal); Investigation (equal); Project administration (equal); Resources (equal); Supervision (supporting); Validation (equal); Writing – review & editing (equal). Takuma Kobayashi: Conceptualization (lead); Funding acquisition (equal); Investigation (equal); Project administration (equal); Resources (equal); Supervision (lead); Validation (equal); Writing – review & editing (equal).

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