Prominent luminescence of silicon-vacancy defects created in bulk silicon carbide p–n junction diodes

We investigate fluorescent defect centers in 4H silicon carbide p–n junction diodes fabricated via aluminum-ion implantation into an n-type bulk substrate without the use of an epitaxial growth process. At room temperature, electron-irradiated p–n junction diodes exhibit electroluminescence originating from silicon-vacancy defects. For a diode exposed to an electron dose of 1×1018cm-2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1 \times 10^{18}\,{{\mathrm{cm}}}^{-2}$$\end{document} at 800keV\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$800\,{{\mathrm{keV}}}$$\end{document}, the electroluminescence intensity of these defects is most prominent within a wavelength range of 400–1100nm\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1100\,{{\mathrm{nm}}}$$\end{document}. The commonly observed D1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\mathrm{D}}}_1$$\end{document} emission was sufficiently suppressed in the electroluminescence spectra of all the fabricated diodes, while it was detected in the photoluminescence measurements. The photoluminescence spectra also displayed emission lines from silicon-vacancy defects.


Results and discussion
shows the sample structure fabricated in this study. A single p-n junction was formed via Al-ion implantation into an n-type 4H-SiC substrate followed by annealing at 1650 • C . A simulated Al concentration 18 in a doping layer is shown in Fig. 1b. Electron irradiation at the energy of 800 keV was performed to introduce intrinsic defects. We prepared four groups of samples with different irradiation doses (sample A: 1 × 10 17 cm −2 , B: 5 × 10 17 cm −2 , C: 1 × 10 18 cm −2 , and R: non-irradiated reference). The current-voltage characteristics of all the samples are shown in Fig. 1c. A higher electron-irradiation dose increases the electrical resistance. We attribute such an increase to the generation of the non-radiative recombination centers of Z 1/2 and EH 6/7 19 . The high nitrogen concentration in the substrate prevented the samples from reached a semi-insulating state. Figure 2 shows the room-temperature EL spectra at a driving current of 20 mA . As can be seen in Fig. 2, sample R shows a broad emission band around 720 nm . In addition to this band, another emission band with a peak wavelength at 910 nm appeared for samples A, B, and C. The emission at 910 nm is a characteristic of V Si defects 21 . The V Si -related emission dominates the spectra of samples A, B, and C. In particular, for sample C, it exceeds the other emission peaks by a factor of 11. Conversely, the V Si emission is negligibly small for sample R. Because the post-implantation annealing temperature of 1650 • C was sufficiently higher than the anneal-out temperature of the V Si defects, i.e., 750 • C 22 , the V Si concentration introduced by the Al-ion implantation was significantly reduced during the annealing process. Therefore, the V Si defects in samples A, B, and C were generated purely via the electron irradiation. Figure 3 shows the integrated intensity of the V Si emission as a function of the injection current I. Saturated behavior for large I is observed for samples A, B, and C. The fit of the equation 23 P ∝ (1 + I 0 /I) −1 , where P is the EL integrated intensity and I 0 is the saturation current, yields I 0 = 13.7 mA for sample C. This value is larger than the value of I 0 = 5.3 mA for the nitrogen-vacancy-related EL of diamond 24 but comparable to the value of I 0 = 10 mA for the V Si -related EL of 6H-SiC 13 .
We now turn to the 720 nm band observed in the EL spectra in Fig. 2. The origin of this band seems neither carbon antisite-vacancy pair ( C Si V C ) defects 25,26 nor D 1 defects, but annealing-related defects 20 . The C Si V C defect is reported to completely disappear with high-temperature annealing around 1100 • C 26 which is much lower than the employed annealing temperature. In addition, their zero-phonon lines, AB lines 25,26 , were not detected in the photoluminescence (PL) measurements at 10 K (see Supplementary Fig. S1a online). Regarding the D 1 defects, the annealing out of them should not occur in our experiments because these defects are thermally stable up to 1700 • C 27 . However, although the D 1 emission intensity should increase with electron irradiation 28 , Supplementary Fig. S2 shows that their EL intensity is not affected by electron irradiation. Thus, there is no good reason to attribute the 720 nm band to C Si V C or D 1 defects. The annealing-related defects may be the origin of this band, however, their varied emission wavelength for each defect 20 makes further identification difficult. www.nature.com/scientificreports/ In the following, we discuss plausible mechanisms underlying the absence of D 1 -originated EL. Unlike the EL measurements, the D 1 emission was observed in the PL measurements for samples R and C even at room temperature, and was not observed in a bare SiC substrate (see Supplementary Fig. S1b online). Therefore, D 1 defects were introduced at least to the penetration depth of excitation light from the top-surface via Al-ion implantation and electron irradiation. Based on the above results, the deteriorated crystal quality in the vicinity of the p-n junction by the high-fluence implantation of Al ions may be the mechanism of the smeared D 1 band in all of the EL spectra. It has been established that variations in the stacking faults of SiC are sensitive to the wavelength of the D 1 -emission line 14 ; therefore, this interpretation is reasonable. According to this mechanism, our EL results suggest that D 1 defects are more sensitive to the crystal quality than V Si because V Si -related EL can still be observed in the electron-irradiated, i.e., additional defect-introduced, samples. Also, defect diffusion is another possible mechanism. The depth profile of Al-implanted 4H-SiC, as detected by cathode luminescence 29 , revealed that high-temperature annealing induces the diffusion of D 1 defects from the surface-implanted area, with a diffusion length of up to several micrometers, while the implanted Al atoms remain in the vicinity of the surface. The low concentration of D 1 defects in the p-n junction layer then would result in the absence of D 1 -related EL. This mechanism, however, leaves an open question as to why the D 1 -emission band does not appear after electron irradiation (see Supplementary Fig. S2 online); electron irradiation should generate the D 1 defects 28 . Lastly, note, thermal activation of the defect-bound electrons in the conduction band 13 due to current injection is not the mechanism of the absence of D 1 EL, because the comparison of the EL spectra between the smallest ( I = 1 mA ) and largest ( 30 mA ) currents revealed that there was no significant difference in the shape of EL spectra and the D 1 EL was not detected for the both currents (see Supplementary Fig. 3 online).
The PL spectra of samples R and C with a 785-nm excitation laser are shown in Fig. 4. Except for the Raman scattering peaks of 4H-SiC at 836 nm and 850 nm 30 , as shown in the upper panel of Fig. 4a, no significant emission www.nature.com/scientificreports/ peaks were observed for sample R at 10 K . As can be seen in the lower panel of Fig. 4a, the PL spectrum at 300 K does not show a large difference. These results indicate the absence of V Si defects in sample R. Conversely, as shown in the upper panel of Fig. 4b, the 862-nm emission line, labeled V1, appears in the PL spectrum of sample C at 10 K . This emission line originates from the V 1 zero-phonon line of the V Si defects in 4H-SiC 31 . The V 2 emission line at 917 nm was too weak to be assigned compared to the large phonon sideband (PSB), which corresponds to the large Huang-Rhys (HR) factor (the small Debye-Waller factor 32 ). Note, V ′ 1 line, which is associated with a second excited state of the V Si defect 33 , was not detected possibly due to a low measurement temperature 31 . The 300-K PL spectrum of sample C in the lower panel of Fig. 4b shows a broad PSB originating from the V Si defects with its peak wavelength around 940 nm , which is slightly longer than the typical wavelength for the V Si PL. However, the PSB peak around 950 nm was observed with the experimental setup similar to ours 34 , and is reasonable for the large HR factor. The discrepancy of the peak wavelength of the V Si -related PSB between the EL and PL spectra is perhaps due to the different local crystalline structures under detection. Therefore, both the EL and PL measurements substantiate the feasibility of efficient V Si luminescence emitted from a bulk SiC substrate with no epitaxial layers.

Conclusion
In conclusion, we successfully fabricated p-n junction diodes including V Si defects on a bulk 4H-SiC substrate without epitaxial layers. The PL spectra demonstrated the formation of V Si defects. The EL measurements revealed that the luminescence intensity of V Si was most prominent within the measured wavelength range. Further, the commonly observed D 1 emission was sufficiently suppressed in the EL spectra, while it was detected in the PL measurements. The absence of the D 1 emission in the EL spectra suggests that the deteriorated crystal quality due to ion implantation affects its luminescence. Our approach provides a foundation for novel applications of V Si defects in SiC using simple manufacturing processes.

Methods
We employed an n-type 4H-SiC substrate with a nitrogen concentration on the order of 10 19 cm −3 purchased from SiCrystal AG. Multi-energy Al-ion implantation was performed at a substrate temperature of 500 • C to form a p-type region on the top-surface of the substrate. The fluence and energy values of the Al implantation were 2.0 × 10 14 , 2.0 × 10 14 , 2.5 × 10 14 , 2.5 × 10 14 , and 8.0 × 10 14 cm −2 at 10, 30, 50, 70, and 100 keV , respectively. According to an SRIM-code simulation 18 , the Al atoms should be flatly distributed from the surface to a depth of 130 nm with an Al concentration of 1 × 10 20 cm −3 . Post-implantation annealing was performed at 1650 • C for 30 min in an argon atmosphere. Subsequent to annealing, the substrate was diced into chips. These chips were divided into four groups. The first three groups were irradiated with 800-keV electrons at doses of 1 × 10 17 (sample A), 5 × 10 17 (sample B), and 1 × 10 18 cm −2 (sample C). The remaining group (sample R) was not irradiated by electrons and served as a reference. Finally, Ti/Ni/Au Schottky contact metal was deposited on the top and bottom surfaces of the chips via electron-beam evaporation. The top-surface electrodes were circularly shaped with a diameter of 300 µm.

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