Cathodoluminescence Study of Damage Formation and Recovery in Si-ion-implanted β-Ga2O3

Ion implantation and activation annealing are key processes in the creation of an ideal free carrier distribution in semiconductor devices. Ultra-wide-bandgap (UWBG) semiconductors, such as gallium oxide (Ga<sub>2</sub>O<sub>3</sub>) and aluminum nitride (AlN), have many advantages suitable for power-device applications. We implanted silicon (Si) at doses ranging from <inline-formula> <tex-math notation="LaTeX">$1 \times 10^{11}$ </tex-math></inline-formula> to <inline-formula> <tex-math notation="LaTeX">$1 \times 10^{15} \mathrm{~cm}^{-2}$ </tex-math></inline-formula> into <inline-formula> <tex-math notation="LaTeX">$\beta-\mathrm{Ga}_2 \mathrm{O}_3$ </tex-math></inline-formula> (−201) wafers, and annealed them at 800 and 1000° C in N<sub>2</sub> atmosphere. Secondary ion mass spectrometry (SIMS) and cathodoluminescence (CL) were used to evaluate the dopant profile and damage resulting from ion implantation. The CL intensity decreased rapidly with the dose. Even at a dose of <inline-formula> <tex-math notation="LaTeX">$1 \times 10^{11} \mathrm{~cm}^{-2}$ </tex-math></inline-formula>, the intensity dropped by nearly half of its value for the bare wafer. The CL intensity recovered after annealing at all doses; however, the CL intensity did not fully recover even after annealing at 1000°C. Moreover, the CL-depth profile at a dose of <inline-formula> <tex-math notation="LaTeX">$1 \times 10^{15} \mathrm{~cm}^{-2}$ </tex-math></inline-formula> after annealing at 1000°C showed pronounced intensity decay near the Si-diffused region. The CL-intensity decay was strongly correlated with Si diffusion. This phenomenon suggests that high-temperature annealing at high dose not only activates the Si dopant through the interaction of the interstitial Si and Ga vacancies but also causes interstitial Si atoms to diffuse into deeper regions. CL spectroscopy is very sensitive to the implantation damage and can be used for optimization of ion implantation and annealing processes in <inline-formula> <tex-math notation="LaTeX">$\beta_{-\mathrm{Ga}_2 \mathrm{O}_3}$ </tex-math></inline-formula>.


I. INTRODUCTION
Gallium oxide (Ga 2 O 3 ) is one of the ultra-widebandgap (UWBG) semiconductors with many advantages suitable for power-device applications [1]- [7]. The monoclinic β phase is thermally stable and its bandgap is approximately 4.7-4.9 eV. Control of the free carrier distribution is important to realize ideal device characteristics suitable for the power devices. Ion implantation and activation annealing are key processes for the creation of an ideal local-carrier distribution in the device. N-type regions have been successfully fabricated using Si-ion and other group-IV-ion implantations [8]- [15]. However, the damage generated by ion implantation may cause serious degradation and failure of the devices. Activation annealing can be used to activate dopant atoms and eliminate the generated crystal defects; however, such annealing may also diffuse the dopant atoms [10]- [12], [14] and transform the point defects into more stable defects [12], [14], [16], [17]. These changes during annealing may affect the device characteristics.
Luminescence spectroscopy is one of the powerful methods to characterize crystal defects. Photoluminescence (PL), which uses light as an excitation source, is widely used to characterize the defects in semiconductors. However, UWBG semiconductors require short-wavelength excitation sources such as vacuum ultraviolet (VUV) light. VUV excitation is effective for UWBG semiconductors; however, VUV systems require a special spectrometer and light source, and the operation conditions are only met in a synchrotron radiation beamline. Cathodoluminescence (CL), which uses electron beams, is suitable for defect characterization in UWBG semiconductors because conventional scanning electron microscope (SEM) can be used as the excitation source, and the exciting-beam energy exceeds the bandgap energies of UWBG semiconductors [11], [18]- [22]. We reported CL analysis on the Si-ion-implanted β-Ga 2 O 3 [22]; however, only a cross-sectional measurement at a dose of 1 × 10 15 cm −2 , which was relatively high dose implantation, were shown in the report.
In this study, we implanted Si at doses ranging from 1 × 10 11 to 1 × 10 15 cm −2 into β-Ga 2 O 3 wafers, and annealed them at 800 and 1000 • C under N 2 atmosphere. Secondary ion mass spectrometry (SIMS), plan-view CL, and cross-sectional CL were used to evaluate the implanted damage and recovery. Based on the intensity change of the CL, we discuss a possible underlying mechanism for dopant diffusion and point-defect evolution during annealing.

II. EXPERIMENTAL SET-UP
Unintentionally doped β-Ga 2 O 3 (−201) substrates from Novel Crystal Technology, Inc were used [23]. We implanted Si at an energy of 100 keV and doses ranging from 1 × 10 11 to 1 × 10 15 cm −2 with a 7 • tilt with respect to the beam. All samples were implanted at room temperature. Subsequently, the samples were annealed at 800 or 1000 • C in N 2 atmosphere for 30 m. SIMS measurements were performed with a magnetic sector CAMECA IMS-7f using a 14.5-keV Cs + ion beam at a 24 • incident angle. Before SIMS measurements, a thin platinum (Pt) film was deposited on the sample to avoid charging effect. For CL measurements, a JEOL JSM-7100F/TTLS SEM was used as an excitation source. The emitted light was analyzed using a HORIBA fiber-optic CL-detection system with an iHR-320 monochromator that was equipped with a cooled charge-coupled device. We performed all CL measurements at room temperature. We set the beam energy of the excitation electrons to 2 keV for plan-view measurements and 1 keV for cross-sectional measurements with a penetration depth of approximately 45 nm and 14 nm, respectively, for β-Ga 2 O 3 based on the Kanaya-Okayama model [24], [25]. For the cross-sectional CL measurements, we used the cleavage (100) surface instead of mechanically polished faces to avoid processing damage due to cutting. The detailed experimental conditions of the cross-sectional measurements were described in [22]. Figure 1 shows the SIMS depth profiles of Si implanted in β-Ga 2 O 3 substrates at various doses with no annealing. We calculated the depth profile at a dose of 1 × 10 15 cm −2 using SRIM simulation [26]. The maximum concentration of Si atoms was located at 80 nm beneath the surface. The  experimental depth profiles were in good agreement with the SRIM simulation. The bare wafer with no implantation had a Si concentration of approximately 1-3 × 10 17 cm −3 . The detection limit of Si atoms in the SIMS measurements was estimated to be approximately 1 × 10 16 cm −3 according to the raster change technique [27]. The detection limit is an order of magnitude smaller than the concentration of the bare wafer. The high Si concentration near the surface does not originate from the β-Ga 2 O 3 wafers but from the contamination of the Pt film deposited on the samples. Figure 2 shows the SIMS depth profiles after annealing at 800 • C. No significant change in the depth profiles was found. However, diffusion of Si atoms was observed after annealing at 1000 • C, as shown in Fig. 3. The Si atoms reached up to 400 nm for the sample implanted at 1 × 10 15 cm −2 . Diffusion was also observed in the sample implanted at 5 × 10 13 cm −2 . The diffusion after 1000 • C is significant. The diffusion phenomenon seems strongly dependent  on concentration. The reason of the observed diffusion is not fully understood at present; however, Sharma et al. also observed a similar diffusion phenomenon in O 2 or N 2 ambient after annealing at 1100 • C. They explained the phenomenon using a defect-assisted process [14]. Figure 4 shows the CL spectra with no annealing. All samples showed a broad emission near 400 nm with a small emission around 700-800 nm. The small emission comes from the Cr 3+ impurity [28]; it is not from the energy level of the crystal defect formed in β-Ga 2 O 3 . The main emission near 400 nm is typical for β-Ga 2 O 3 . In general, compound semiconductors such as gallium nitride (GaN), gallium arsenide (GaAs), and indium phosphide (InP), show sharp near-band emission; however, the luminescence spectra of β-Ga 2 O 3 generally show broad emissions below the bandgap [17], [29]- [38]. Owing to a strong interaction between a free hole and the lattice, the hole tends to localize and form a polaron, and the band-to-band emission shows a large Stokes shift. The shoulder peak around 360 nm comes from the self-trapped exciton (STE) consisting of a self-trapped hole (STH) and a bound electron [32]- [36]. The main emission at 400 nm is considered to be donor-acceptorpair (DAP) emissions [37], [38]. The donors and acceptors responsible for DAP emissions were not clearly identified; however, Onuma et al. indicated that substitutional Si atoms (Si Ga ) and deep acceptors such as Ga vacancies (V Ga ) and V O -V Ga complexes are possible candidates for DAP emission in Si-doped β-Ga 2 O 3 [38]. The observed emissions in Fig. 4 can be interpreted as a mixture of STE and DAP emissions.

III. RESULTS AND DISCUSSION
Note also in Fig. 4, that the CL intensity decreased rapidly with the dose. However, the peak wavelength and the spectral shape showed no evident change. This means that the dominant process during Si implantation was the generation of nonradiative recombination centers, and that the total number of such centers increased with the dose. The detailed nature of the nonradiative recombination centers is not well known; however, the ion implantation generates native interstitials (Ga i , O i ), vacancies (V Ga , V O ), interstitial Si (Si i ) atoms, and their complexes through the bombardment of β-Ga 2 O 3 by the Si atoms. These point defects are possible candidates for nonradiative recombination centers. Even at a dose of 1 ×10 11 cm −2 , which is equivalent to the level of Si in the bare wafer with no implantation, the intensity drops by nearly half of its value for the bare wafer. This indicates that the CL spectra are very sensitive to the implantation damage. After annealing at 800 and 1000 • C, the CL intensity recovered, as shown in Fig. 5. The annealing process decreased the nonradiative recombination centers and promoted crystal quality. The spectra after annealing did not show any spectral changes and new emissions, except for the peak intensity. Figure 6 shows a summary of the CL-peak intensity after annealing at 800 and 1000 • C at various doses. The CL intensity recovered after annealing at all doses. The intensities after annealing at 1000 • C were stronger than those after  annealing at 800 • C for all doses. However, the CL intensity did not fully recover even after annealing at 1000 • C. It suggests that the nonradiative recombination centers were still included, and that the Si atoms did not fully activate after high-temperature annealing.
Note that the intensity of the bare wafer increased after annealing. This indicates that thermal annealing can improve the crystal quality of the bare wafer. Two mechanisms are possible for this phenomenon. One is the crystal recovery by the reduction of nonradiative recombination centers such as the inactive Si atoms, and the other is the reduction of surface defects by annealing. We conducted cross-sectional CL measurements to investigate the depth distribution of the defects. Figure 7 shows the CL-intensity profiles of the bare wafer and the implanted sample at a dose of 1 × 10 15 cm −2 as a function of depth. The intensity profiles are nearly the same even after annealing, although the intensity of the bare wafer increased in the deeper range more than 100 nm after annealing at 1000 • C. This suggests that the bulk defects decrease by annealing, and the activation of unintentionally doped Si atoms is considered to be a possible process for the intensity increase of the bare wafer by annealing.
In Fig. 7, the as-implanted sample showed lower intensity than the annealed sample in the range from 0 nm to 1200 nm. This indicates that implanted damage existed at a deeper region than that of the ion-beam range, which was within 200 nm from the SIMS results in Fig. 1. The detailed mechanism of damage formation in deeper region is not unambiguously identified at this stage; however, the same phenomenon was also observed in silicon carbide (SiC) [18], [39] and GaN [20] and mainly explained based on the diffusion and interaction of the mobile point defects. In SiC and GaN, the diffusion coefficients of the dopants are relatively small, and dopants cannot diffuse easily even after high-temperature annealing. In contrast, some types of defects generated by the collision of dopants can move into deeper region and interact with other atoms and defects easily. The diffusion and the interaction of mobile point defects generated by the first collision of dopants are considered to be a possible mechanism for β-Ga 2 O 3 .
The CL intensity at a dose of 1 × 10 15 cm −2 increased after annealing; however, the intensity profile after annealing had an intensity dip around 200-400 nm. This region corresponds to the Si-diffused region. Based on these SIMS and CL results, a possible underlying mechanism for the diffusion and activation of Si atoms is hypothesized as follows.
(1) Ion implantation generates a high concentration of Si i within the ion-beam range. At the same time, many types of point defects such as Ga i , O i , V Ga , V O , and their complexes can be formed by Si-ion bombardment. These point defects are mainly formed within the ion-beam range; however, the generation area is not limited to this range.
(2) The interactions between the point defects are promoted during low temperature annealing at 800 • C. The Si ions can be partially activated mainly by the interaction of Si i and V Ga .
(3) Nearly the same interactions are promoted during high temperature annealing at 1000 • C. Both the interaction of Si i and V Ga , and the diffusion of Si i into deeper regions occur because the binding energy of Si i is weaker than that of Si Ga , and Si i becomes more mobile under 1000 • C annealing because it receives thermal energy. Some portion of Si i substitutes for Ga sites during the diffusion, and stops moving. However, the diffused Si atoms may not be fully activated because the CL-intensity profile at a dose of 1 × 10 15 cm −2 has a dip around 200-400 nm for high-temperature annealing at 1000 • C.

IV. CONCLUSION
We implanted Si at doses from 1 × 10 11 to 1 × 10 15 cm −2 into β-Ga 2 O 3 wafers, and annealed them at 800 and 1000 • C under N 2 atmosphere. The SIMS depth profiles showed no significant change after annealing at 800 • C; however, significant diffusion of Si atoms was observed at a dose of 1 × 10 15 cm −2 after annealing at 1000 • C. The CL intensity decreased with the dose after the implantation and recovered after the annealing at all doses. However, the CL intensity did not fully recover even after annealing at 1000 • C, suggesting that the nonradiative recombination centers were still included and that the Si atoms did not fully activate after annealing. The intensity of the bare wafer increased after annealing. This suggests that thermal annealing can improve the crystal quality of the bare wafer. The cross-sectional CL at a dose of 1 × 10 15 cm −2 after annealing at 1000 • C showed intensity decay near the Si-diffused region, showing that CL-intensity decay was strongly correlated with Si diffusion. The high-temperature annealing at high dose not only activated the Si dopant through the interaction of Si i and V Ga but also caused the Si i atoms to diffuse into deeper regions. CL spectroscopy can be used for the optimization of the ion implantation and annealing processes in β-Ga 2 O 3 . RYUICHI SUGIE received the B.S. and M.A. degrees from Osaka University, Japan, in 1992 and 1994, respectively, where he studied Raman spectroscopy of compound semiconductors, and the Ph.D. degree in engineering from Osaka University in 2011. He spent four years (from 1995 to 1998) as a Researcher with Toray industry Inc. In 1998, he joined Toray Research Center Inc. He has been engaged in the study of semiconductors via Raman and luminescence spectroscopies. His research interest includes the basic physics in semiconductors, optimization of device process parameters using Raman and luminescence, failure analysis of semiconductor devices. He is a member of the Japan Society of Applied Physics and the Japanese Society of Microscopy.