Optical spectroscopy study of damage evolution in 6H-SiC by H+2 implantation*

Yong Wang(王勇), Qing Liao(廖庆), Ming Liu(刘茗), Peng-Fei Zheng(郑鹏飞),Xinyu Gao(高新宇), Zheng Jia(贾政), Shuai Xu(徐帅), and Bing-Sheng Li(李炳生),4,†

1China Institute for Radiation Protection,Taiyuan 030006,China

2State Key Laboratory for Environment-Friendly Energy Materials,

Southwest University of Science and Technology,Mianyang 621010,China

3Southwestern Institute of Physics,Chengdu 610041,China

4Engineering Research Center of Biomass Materials,Ministry of Education,

Southwest University of Science and Technology,Mianyang 621010,China

Keywords: SiC,H+2 implantation,Raman scattering spectroscopy,photoluminescence spectrum

1. Introduction

Silicon carbide(SiC),as an important semiconductor material with interesting physical, optical, and electronic properties, is widely used in high-voltage, high-frequency, hightemperature switching devices, as well as in numerous optoelectronic applications.[1]Besides semiconductor industry,SiC can also be used in nuclear power plants because of its low cross-section for neutron capture,excellent chemical,mechanical,and structural stability,as well as low diffusion rate of fission products.[2]In future nuclear reactor,SiC is considered an ideal fuel cladding material for high-temperature gascooled fission reactors.[3]Although there are over 200 polytypes of SiC, only face cubic SiC (3C-SiC), considered for nuclear applications and hexagonal 4H-SiC and 6H-SiC having wide-band gap, high-electron mobility, and low-leakage current used in the semiconductor industry, have been extensively investigated so far. To perform selective-area doping in hexagonal SiC, ion implantation has been widely used in the past.

It is well known that RBS/C and TEM methods have some drawbacks with respect to particle analysis. Generally,RBS/C analysis uses 2.0-MeV He ions and fabrication of TEM samples is needed due to required transparency to 200-keV or 300-keV electrons. After RBS/C analysis,He ion implantation induces surface color being changed from green to black, due to the increase in the absorption coefficient (not shown).[11]Optical characterization is a good way to avoid sample modification/destruction during measurement. Raman scattering spectrometry technology is widely performed to investigate ions implanted SiC,and it can precisely characterize the evolution of lattice damage with an implantation dose,similar to the result obtained via Rutherford backscattering in channeling geometry.[12–16]Furthermore,Raman scattering spectrometry is sensitive to a low defect concentration via the decrease in Raman scattering intensity, broadening of phonon Raman bands, and frequency shift.[12]In this study, confocal Raman scattering spectroscopy and photoluminescence spectrum are employed to investigate the H+2-implanted 6H-SiC followed by thermal annealing. The defect recovery with annealing temperature is characterized by different methods. These two kinds of optical characterizations of lattice damage are compared.

2. Experimental process

Cz-grown N-doped (1018cm−3–1019cm−3), 〈0001〉-oriented single crystalline 6H-SiC wafers with 2 inch(1 inch=2.54 cm)in diameter and 0.35 mm in thickness(supplied by HF-Kejing Inc.), were cut into 1 cm × 1 cm pieces and implanted at room temperature and elevated temperatures (300°C and 500°C) with 134-keV H+2molecular ions at several fluences ranging from 0.5 × 1016H+2·cm−2to 5× 1016H+2·cm−2, respectively. The nominal H concentration depth distribution was obtained from SRIM-2010 using the quick cascade calculation.[17]The mean-projected range Rpwas estimated to 393 nm with a straggling ΔRpof approximately 46 nm(using the density of 3.21 g·cm−3and the threshold displacement energies of 20 eV and 35 eV for C and Si sublattices, respectively). In order to investigate the defect recovery in the sample, isochronal annealing at three different temperatures(700°C,900°C,and 1100°C)for 15 min in vacuum(～10−3Pa)was carried out.Surface blisters and exfoliation were observed by scanning electron microscopy when the sample was implanted to a fluence of 1.5×1016H2·cm−2followed by 1100-°C annealing. An arc-shaped crack located at a depth of approximately 410 nm was formed inside the sample, where is good consistent with the projected range of hydrogen deposition, as shown in Fig. 1. The Raman scattering spectroscopy was performed using a Horiba Join Yvon HR800 spectrometer at room temperature in the backscattering configuration. The excitation wavelength was 532 nm from the Ar+laser with a spectral resolution of approximately 0.5 cm−1,and the measurement range was from 100 cm−1to 1000 cm−1. Spectra were recorded with an acquisition time of 30 s. Micro-Raman spectroscopy was employed and the laser spot was approximately 1 µm using a microscopy lens system.The scattered light was detected by a Jobin Yvon T64000 spectrometer with an LN2cooled charge-coupled device. The penetration depth of the exciting wavelength is expressed as d =1/2α, where α is the absorption coefficient of the crystalline SiC.[18]The absorption coefficient of single-crystalline 6H-SiC is 115 cm−1at 532 nm based on the UV-visible optical transmittance spectrometer measurement.[12]Therefore,the penetration depth of the exciting wavelength in the singlecrystalline 6H-SiC reaches 43µm,which is far larger than the H+2ion penetration depth. Samples were excited by 325-nm light from a suitably filtered He–Cd laser on the Shimadzu RF-5301 photoluminescence spectroscopy. The measurement was carried out at room temperature(RT),and the exciting wavelength was 340 nm. The measurement range was from 350 nm to 1000 nm.

Fig. 1. Cross-sectional scanning electron microscopy, (a) secondary electron image, (b) backscattering electron image of 6H-SiC implanted with 134-keV H+2 ions at a fluence of 1.5×1016 H+2·cm−2 at 300 °C followed by 1100 °C annealing for 15-min observed by secondary electron (a) and backscattering electron(b),superposed the distribution of damage and hydrogen deposition of the H+2-implanted 6H-SiC simulated by SRIM-2010(c).Blisters and exfoliation are observed in panel(a)and a crack exhibiting black contrast is clearly observed in panel(c).

3. Results and discussion

Figure 2 presents the change of photoluminescence spectra after H+2-ion implantation followed by the thermal treatment. An emission peak located at 590 nm was clearly observed. In the present experiment,n-type 6H-SiC due to nitrogen doping was chosen. Meanwhile, using boron as acceptor doping compensates nitrogen doping. Kawai et al.[19]argued that the emission peak located at 590 nm is attributed to donor–acceptor pair emission in a SiC crystal. This emission in the range of visible light is convenient for various optical devices,with the intensity of the emission peak being an important parameter in the fabrication process. It is generally regarded that the donor–acceptor pair emission itself is very strong in highquality SiC.[19]As a result,the emission intensity strongly depends on the quality of the H+2ions-implanted SiC. Figure 2 shows that the location of emission peak did not change significantly after the H+2ion implantation,except for 134-keV H+2-implanted SiC at 500°C after 1100-°C annealing,where large blisters and craters are formed on the surface.[6]This result demonstrates that the donor–acceptor pair emission plays the main role in photoluminescence emission of SiC before and after H+2ion implantation. Moreover, the emission intensity decreased significantly after H+2implantation. The detailed change of the emission intensity versus fluence and temperature is presented in Fig.3.

The Raman spectra of 6H-SiC samples before and after the 134-keV H+2-ion implantation to fluences ranging from 0.5×1016H+2·cm−2to 5×1016H+2·cm−2at RT or 2.5×1016H+2·cm−2at RT,300°C,and 500°C,were investigated.In the range of 100 cm−1–1000 cm−1, the group-theoretical analysis indicates that the Raman-active modes of wurtzite structure(C6vsymmetry for hexagonal polytypes)are A1,E1,and E2modes.[20]In addition, A1and E1phonon modes are split into longitudinal(LO)and transverse(TO)optical modes,respectively.In backscattering configuration,Al(LO),E1(LO),and E2phonons are observable in Raman spectra.

Figure 3 presents the Raman spectra of 6H-SiC samples, observed Raman peaks located at 144 cm−1/150 cm−1,266 cm−1, 505 cm−1/513 cm−1, 768 cm−1/790 cm−1,798 cm−1, 968 cm−1are attributed to E2(TA), E2(TA),A1(LA), E2(TO), E1(TO), A1(LO) phonons, respectively.[21]It can be seen that Raman peaks shift to a lower wavenumber after H+2-ion implantation, as shown in Fig. 3(h). The red shift can be attributed to the hydrostatic strain caused by implantation-induced defects.[22]It is interesting to note that one broad band located at 578 cm−1is observed in the H+2ions implanted 6H-SiC after 1100-°C annealing(like 8#SiC),except for 5#SiC.This phonon peak is attributed to the vibration mode of CSiVCcomplexes.[23]It is consistent with recent research indicating that many carbon antisite–vacancy pairs formed after 1100-°C annealing measured by transmission electron microscopy and positron annihilation spectrum.[24]It can be seen that the background intensity increased after the H+2ion implantation. This is related to Rayleigh scattering from the implantation-induced damage.However,background intensity increased in the H+2ions implanted 6H-SiC with the thermal treatment, except for 5#SiC. It is associated with the vibrations of silicon and disorder silicon carbide.[21]In addition,the intensities of Raman peaks decreased after H+2ion implantation. The intensities of Raman scattering spectroscopy and photoluminescence spectrum were analyzed and the result is present in Fig.4.

Fig.2. Photoluminescence spectra from 134-keV H+2 ions-implanted 6H-SiC to different fluences after thermal annealing for 15 min. Note that 5#SiC,6#SiC,8#SiC,and 10#SiC stand for the implantation fluence of 5×1015 H+2·cm−2,1.5×1016 H+2·cm−2,5×1016 H+2·cm−2,and 2.5×1016 H+2·cm−2 at RT,respectively. 12#SiC and 13#SiC stand for the implantation fluence of 2.5×1016 H+2·cm−2 at 300 °C and 500 °C,respectively.

Fig.3. Raman scattering spectra from 134-keV H+2 ions-implanted 6H-SiC to different fluences after thermal annealing for 15 min in vacuum. Note that 11#SiC stands for the implantation fluence of 3.75×1015 H+2·cm−2 at RT.The left shift of the A1(LO)peak after H+2 ion implantation in the 5#SiC sample before and after annealing is observed in panel(h). (i)The expanded version of Raman spectra in panel(d)shows a broad band at 578 cm−1indicated by a black arrow.

H+2-ion implantation can induce dense lattice defects,which destroy lattice structure and decrease the intensity of observed peaks. In general,researchers use the integrated intensity,but not peak intensity,to calculate observed Raman peaks after ion implantation.[11–13]In this study,integrated intensity ΔA of the photoluminescence emission from 400 nm to 850 nm is normalized to the value ΔAas−grownof the as-grown sample,and the result is shown in Fig.4(a).Similarly,integrated intensity ΔA of the Al(LO)phonon from 940 cm−1to 1000 cm−1is normalized to the value ΔAas−grownof the as-grown sample and the result is shown as Fig.4(b).Compared with the relative intensity between photoluminescence emission and Raman scattering, a similar change trend as a function of the annealing temperature is observed. After 700-°C annealing, the profile slope is positive, except for 12#SiC. Meanwhile, the profile slope is small, which is indicated that the defect recovery is not significant after 700-°C annealing. As for 12#SiC,the relative intensity decreases after 700-°C annealing. It is a reverse annealing phenomenon,which is usually observed by gas ion implantation,such as H+,He+. The growing gas bubbles during high-temperature annealing squeeze around C and Si sublattice atoms, which result in punching out self-interstitials,known as loop punching.[25–27]Because the threshold temperature for Si vacancy migration in SiC is above 900°C and C vacancy needs higher temperature,[28]it is reasonable to regard that bubbles cannot be formed after 700-°C annealing.The possible reason is the migration and accumulation of interstitials into interstitial-type clusters after 700-°C annealing.The intensity of the photoluminescence emission depends on the donor–acceptor pair number,while the intensity of the Raman scatting depends on the optical absorption coefficient.The decrease in relative intensity indicates the reduction in the donor–acceptor pair number and the increase in the optical absorption coefficient via these interstitial-type clusters.[12,16,21]With increasing annealing temperature from 700°C to 900°C,all profiles have positive slopes,indicating the defect recovery after 900-°C annealing. With increasing annealing temperature from 900°C to 1100°C,the profile slope is positive,except for 13#SiC.It can be attributed to the growth of vacancytype defects, like micro-cracks after 1100-°C annealing. The microstructure in the H+2ions-implanted 6H-SiC after 1100-°C annealing is reported in our recent article.[6]The difference between the two methods is the range of the relative intensity.In the whole, the relative intensity of the photoluminescence emission is smaller than that of Raman scattering, which indicates that lattice defects have more influence on the number of donor–acceptor pair than the optical absorption coefficient.As for an important semiconductor and optical material,fabrication of p–n junction via ion-doping into SiC,implantationinduced lattice defects should be taken into consideration.

Besides the lattice defects,the focused condition also affects the Raman scattering intensity. In order to reduce the influence of the focused condition on the Raman peaks,the peak intensity of E2(TO)phonon divides by Al(LO)phonon(Δh=hE2/hA1,and relative intensity is Δh/Δhas−grown)is chosen to calculate the defect recovery after annealing. One can see a sharp difference between ΔA/ΔAas−grownand Δh/Δhas−grown.First, the relative intensity is different. The integrated intensity is near zero for 11#SiC and 8#SiC,while the peak intensity of E2(TO)phonon divides by Al(LO)phonon is near 60%for the same sample. The microstructure shows an amorphous layer formed in the 11#SiC and 8#SiC,resulting in the disappearance of characteristic SiC Raman scattering peaks. Secondly,the change of the profile slope is different.For example,there is no change basically with annealing temperature for the 5#SiC via the calculation of Δh/Δhas−grown,while the relative intensity increased with annealing temperature via the calculation of ΔA/ΔAas−grown. Generally, the as-implanted sample has the largest concentration of survival defects,while Frenkel pairs migrate and recombine each other to reduce the survival defects after 700-°C annealing. As for the 5#SiC,lattice disorder recovery after the thermal treatment is precisely characterized by ΔA/ΔAas−grown,but not Δh/Δhas−grown. In addition,the ΔA/ΔAas−growndecreases after 1100-°C annealing for the 13#SiC. On the contrary, the Δh/Δhas−grownincreases. Our previous result shows an obvious micro-crack formed in the 13#SiC after 1100-°C annealing. There are dense frank loops formed around the micro-crack.[6]These frank loops and some undiscovered defects induce the decrease in the Raman scattering intensity and the increase in the asymmetry of Al(LO)phonon. The asymmetry of Al(LO)phonon cannot be characterized by Δh/Δhas−grown. Therefore,using the peak intensity of E2(TO) phonon divides by Al(LO) phonon is not accurate for characterizing the lattice disorder.

To investigate the asymmetry of Al(LO) phonon, the intensity of the asymmetry is described by Δτ = (Ileft−Iright)/Iright,where I is the intensity of Raman scattering baseline. Based on experimental results, Irightat x=1000 cm−1,and Ileftat x = 936 cm−1(the Al(LO) peak located at 968 cm−1) are observed and the calculation result is shown in Fig. 4(d). It can be seen that there is a linear change with annealing temperature for the 6#SiC and 13#SiC, while Δτ increases initially and then decreases with the annealing temperature for the other samples. It is a surprise that the relative intensity of 6#SiC is the largest compared with other as-implanted samples. Because 6#SiC is implanted with a fluence of 1.5×1016H+2·cm−2,the implantation-induced lattice defects are less than the case of 10#SiC. The asymmetric broadening of the A1(LO) phonon can be accounted for as a“spatial correlation”model that implantation-induced defects can induce q-vector relaxation.[29]The increase in defect density induces an asymmetry of A1(LO) phonon with an obvious tail in the range of 900 cm−1–967 cm−1. Two potential reasons are used to explain why the number of Δτ for the 6#SiC is the largest. The first reason is the growth of the optical absorption coefficient with increasing fluence. Raman scattering intensity decreases with increasing optical absorption coefficient.[12,16,21]Meanwhile, the decrease in Irightis faster than that of Ileft. The Raman scatting results show that the intensity of Raman scattering baseline does not increase with growing lattice disorder. The second reason is that the lattice defects enhance Raman activation, such as a peak located at 935 cm−1–940 cm−1due to the vibration of disordered SiC.[21]Based on the above analysis, using(Ileft−Iright)/Irightcannot precisely characterize the asymmetry of Al(LO)phonon.

Figure 4(e) shows the relative integrated intensity of the sample implanted at RT without annealing. It can be clearly seen that the curve exhibits an exponential decrease with increasing the fluence. It is consistent with previous research that disorder accumulation can be explained by direct impact/defect simulation model or multi-step transformation process.[30,31]Figure 4(f) shows the relative integrated intensity of the sample implanted at elevated temperature without annealing. It exhibits a linear growth of intensity with increasing the implantation temperature. This result indicates the recovery of the same type of lattice defects during implantation at 300°C and 500°C,where only interstitial-type defects can migrate.

4. Conclusion

In this study,the formation and evolution of lattice disorder in the H+2ions-implanted 6H-SiC are investigated by using Raman scattering spectroscopy and photoluminescence spectrum. The influence of lattice disorder on the intensity of observed peaks is discussed in details. The lattice disorder has a more pronounced effect on the intensity of photoluminescence emission, compared to the intensity change measured by Raman scattering. This conclusion demonstrates that the number of donor–acceptor pair decreases significantly when SiC crystal contains lattice defects.It is recommended to pay particular attention to lattice damage in performing selective-area doping technique via ion implantation. It is the integrated intensity of Raman scattering,but not the peak intensity of E2(TO)phonon divided by Al(LO)phonon and(Ileft−Iright)/Iright,that can precisely characterize the evolution of lattice disorder in SiC.

Acknowledgement

The authors appreciate the Laboratory of 320-kV High-Voltage Platform in the Institute of Modern Physics, Chinese Academy of Sciences for the helium implantation. We would like to thank Dr. Vladimir Krsjak for the writing revision.