Correlations between reverse bias leakage current, cathodoluminescence intensity and carbon vacancy observed in 4H-SiC junction barrier Schottky diode

Reverse bias currents of ten commercial junction barrier Schottky diodes were measured, and the dies were studied by scanning electron microscope (SEM) and cathodoluminescence (CL) after the de-capsulation of the diodes. Defect emissions (DEs) of 2.62 eV were observed in all the CL spectra. By comparing the SEM images, the integral CL intensity spatial mappings and the reverse bias leakage currents, correlations between the leakage current, the integral CL intensity and the Al-implantation process were established. The data of reverse bias leakage current against the reverse bias voltage taken at room temperature followed the Poole Frenkel emission from the Z 1/Z 2 carbon vacancy states to the conduction band. The DE at 2.62 eV is associated with the electronic transition from Z 1/Z 2 to the valence band. The current observation also opens up the feasibility of screening off SiC diodes with large leakage current during production by inspecting the CL intensity before the device fabrication is complete.

Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence.Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.

Introduction
Compared to silicon, 4H-SiC has a larger bandgap (3.25 eV), higher breakdown electric field (2.5 MCcm −1 ), higher saturated carrier velocity (2 × 10 7 cm −1 ) and better thermal conductivity (4.9 Wcm −1 K −1 ).These supreme physical properties make SiC a better choice for fabricating high-power devices.Atomic-scale intrinsic defects in semiconductors have deep ionization states in the band gap and a significant influence on the electrical properties of the materials and the corresponding devices.Carbon vacancy is the dominant intrinsic defect in 4H-SiC, having ionization states at 0.6-0.7 eV and 1.5-1.6 eV below the conduction band (denoted by Z 1/2 and EH 6/7 , respectively) [1][2][3][4][5][6].Carbon vacancy has a typical concentration of 10 12 cm −3 in as-grown 4H-SiC epi, and goes up to 10 15 cm −3 after annealing at 1950 • C [3]. V C is reported to have the effect of decreasing the carrier lifetime in SiC devices [7][8][9], influencing the forward current-voltage characteristics of 4H-SiC p-i-n diodes [10] and increasing the reverse bias leakage current [11].Al-ion implantation followed by hightemperature annealing is crucial to introduce p-type doping in 4H-SiC bipolar device fabrication [5].Carbon vacancy was created in the ion-implantation and subsequent high temperature annealing [12].It is thus important to have a probe capable of characterizing carbon vacancy in the line of the SiC device fabrication.
There are many methods for detecting electronic defect states in the band gap in semiconductor materials and devices, such as deep level transient spectroscopy (DLTS) [1,3,4,6,9,13] and luminescence spectroscopy [13][14][15].DLTS can reveal the deep trap activation energy and concentration by monitoring the capacitance transient during the reverse bias pulse period after the forward bias trap filling pulse is applied.Luminescence spectroscopy can be used to detect deep trap states by observing the defect emission (DE) resulting from the electronic transition between states, for example, the conduction band to defect state transition or the defect state to valence band transition.Carrier excitation is needed for luminescence spectroscopy, whereas a laser with photon energy larger than the band gap is used in photoluminescence and accelerating electrons are used in cathodoluminescence (CL).Fabbri et al [13] compared the spectra of CL and DLTS of Alion implanted 4H-SiC diodes, reporting a correlation between a broad DE centered at 2.5 eV and the DLTS signal Z 1 /Z 2 (with activation energy of 0.66 eV).Scanning electron microscopes (SEM) are probes which can revealdefects on a surface via morphological imaging.Defect review SEM is a useful tool capable of being installed in the device production line to reveal the surficial defect image on the wafer [16].It is not easy to detect atomic scale defects such as vacancy by SEM imaging, however, for in-situ detection of deep traps during the manufacturing process, DLTS may not be easily encorporated into the production line, as the method needs to be performed on the rectifying diode.

Experimental
The samples are commercial JBS diodes with the detailed fabrication processes given as for the Batch A samples in [17].The 4H-SiC epi has an electron concentration of 10 16 cm −3 .Multi-step Al-implantation was performed in a square ringshaped region to create a ∼1 µm depth box-shaped doping of p ∼ 10 20 Al cm −3 .Thermal annealing at 1950 • C in Ar was then carried out to activate the Al doping.Ti deposition is then performed on the front and back sides for the metal contacts and then capsulated.
The reverse bias leakage current of the samples was measured and monitored at room temperature with a bias of −1700 V.Ten representative samples with leakage current lying in a wide range from ∼1 µA to 100 µA were chosen.The ten samples were de-capsulated by wet chemical etching.The samples were etched by being immersed in H 2 SO 4 /H 2 O 2 for 10 min and HF for 10 min.After removing the surface electrode, the samples were rinsed with deionized water.The size of each die is ∼0.3 mm × 0.3 mm.SEM characterization was carried out using the JEOL field emission SEM JSM-7001-F and the CL study was performed using the attached monochromator GATAN MonoCL3.For each die, SEM and CL studies were conducted at nine different sampling positions of the die after de-capsulation.While carrying out the CL measurements, the power was kept constant at 30 µW and electron acceleration voltage of 10 kV so that the comparison of CL intensity between samples was feasible.

Result
Table 1 tabulates the reverse bias leakage currents of the ten diodes while biased at −1700 V.The leakage current I leakage varied from 0.86 µA to 139 µA.Sample 10 is the sample with the largest I leakage (=139 µA).The SEM images taken at the nine different locations are given in figure S1(a) in the supplementary information figure S1, whereas a typical one is shown in figure 1(a).The nine locations of the CL measurements sampled over the whole area of sample 10 (see figure S1(b)).CL measurements were also conducted in the regions of the SEM images.The spatial mappings of the CL intensity summing over the photon energy of 1.75-4.13eV taken at the corresponding nine different positions are shown in figure S1(b) in the SI.The CL intensity spatial mapping associated with the SEM image in figure 1(a) is given in figure 1(b).The red dotted square labels one of the regions with Al-ion implantation.The SEM images and the corresponding CL intensity spatial mappings of sample 6 (which has a lower I leakage = 12.6 µA) sampled at the nine different positions are shown in figure S2 of the SI.Figures 1(c) and (d) show one of the typical SEM images and the corresponding CL intensity spatial mapping for sample 6.The SEM images and the corresponding CL intensity spatial mappings taken at the nine different positions of   Figure 2 shows the CL spectra (i.e.luminescence intensity versus photon energy) for all the samples in the photon energy range of 1.75-4.13eV, with the intensities obtained by summing over the whole SEM image region.A peak is observed  looks scattered.It can be understood in terms of the sample preparation process for the CL measurement involving acid etching to remove the encapsulation, which would not be able to be precisely controlled upon each of the individual samples.
To further understand the reverse bias current, we studied the reverse bias current against the applied reverse bias (−75 V to −1200 V) at different temperatures (room temperature to 150 • C) on a new commercial JBS diode fabricated by the same process (sample 11). Figure 4 shows the corresponding Ln(J/E) against E 1/2 .E is the electric field of the corresponding electric bias.The Poole-Frenkel (PF) emission current involves carrier emission from a defect to the conduction band, and is given by [18,19]: where K is a constant depending on the carrier concentration and Φ T is the ionization energy of the trap.The linear relation between Ln (J/E) against E 1/2 implies that the PF emission current is the dominant process for the reverse bias leakage current.The ionization energies of the traps are fitted using equation ( 1) on the data and the results are tabulated in table 2. The ionization energy obtained at the RT is 0.65 eV, which coincides well with the E a of the carbon vacancy (Z 1 /Z 2 ) observed in the DLTS measurement [3,4,6,9,13].This implies that the reverse biased leakage current is dominated by the PF emission from the carbon vacancy Z 1 /Z 2 states to the conduction band.It is noticed that the ionization energy obtained from the fitting increases with temperature to value of 0.92 eV at 150 • C. The leakage current may originate from traps other than the carbon vacancy states due to PF emission.One possible candidate is the trap having ionization energy Φ T = 1.03 eV, which is associated with a carbon cluster [20].Higher temperatures enhance the thermal excitation, and thus PF emission from deeper states like carbon clusters emerges, becoming more effective and feasible.The fitted ionization energy is indeed the corresponding effective ionization energy.The observed increase of the fitted effective ionization energy with temperature indicates emission from deeper states at higher temperatures.

Discussion
The results presented above provide evidence that the reverse bias current at room temperature of the device is dominated by the PF emission from the carbon vacancy Z 1 /Z 2 states to the conduction band.It is also worth discussing the origin of the 2.62 eV emission peak in the CL spectrum.Comparing the SEM images with the corresponding CL intensity spatial mappings in figures 1, S1 and S2, the CL is associated with the Al implantation.A similar observation in the correlation between CL (∼2.5 eV) and the Al implantation was observed by Fabbri et al [13].Furthermore, Fabbri et al [13] observed the correlation between the CL intensity and the DLTS signal of Z 1 /Z 2 .The band gap of 4H-SiC is 3.26 eV at room temperature.Carbon vacancies have trap states Z 1 /Z 2 at 0.6-0.7 eV below the conduction band E C .The energy difference between the Z 1 /Z 2 and the valance band is thus ∼2.5-2.6 eV, and thus the 2.62 eV DE observed in the current study could be attributed to the electronic transition from Z 1 /Z 2 to valance band transition.Chen et al [21] reported introduction of DEs (2.56 eV and 2.81 eV) in the CL spectra of 4H-SiC after the Cu implantation.The 2.56 eV DE was associated with Z 1 /Z 2 .
The correlation between the reverse bias leakage current and the integral CL emission intensity observed in the current study can now be explained.Both the reverse bias leakage current and the integral CL emission intensity are associated with the carbon vacancy through the PF emission and electronic transition from the conduction band to the carbon vacancy state.Thus, the higher concentration of V C leads to larger reverse bias leakage current and integral CL emission intensity.
Atomic scale defects in materials have a significant influence on the device performance.For example, V C is commonly found in 4H-SiC epi materials.It can be thermally generated during the high temperature fabrication process, such as the annealing activation of the Al p-type dopant after the Alion implantation [12].The presence of V C would degrade the device performance, such as decreasing the minority carrier lifetime [7][8][9] and increasing the reverse bias leakage current of the JBS diode [11].However, characterizing atomic scale defects during in-line device manufacturing is not trivial.DLTS is a commonly used method to characterize deep traps in semiconductors including 4H-SiC [4,22].This method involves the filling of deep traps while the diodes are in forward or zero, followed by the measurement of the capacitance transient in the depletion region while the diode is put under reverse bias.Characterizing defects by DLTS requires the rectifying diode and thus is not easy during the on-line device fabrication process before the device is completed.SEMs are now adopted by all advanced wafer fabrication lines for imaging and reviewing sub-micron defects [23].With advanced classification schemes, yield limiting defects (killers) are separated from non-yield limiting defects (non-killers).However, SEMs can only detect sub-micron size defects but are not able to detect electrical active atomic-scale defects which would significantly influence the device performance.There is currently no available methodology used in fabrication sites during the on-line fabrication to detect atomic scale defects which significantly influence the performance of the resultant devices.The present study reveals the correlation between the 4H-SiC diode leakage current and the integral CL intensity, while the reverse bias leakage current and the integral CL intensity of the 2.62 eV peak are associated with the carbon vacancy via the PF emission and electronic transition involving the carbon vacancy.Carbon vacancies are introduced in many energetic processes and thermal processes during SiC device fabrication, such as Al-ion implantation and high temperature annealing.The current results open up the feasibility of employing CL as a probe in the manufacturing line to detect carbon vacancies and screen off the devices in production that would yield large reverse bias leakage currents.Through the spatial mapping of the integral CL intensity, the spatial distribution of the carbon vacancy can also be obtained.Moreover, CL can be installed in the SEM used to carry out defect inspection during the fabrication process.

Conclusion
In summary, correlations between the reverse bias current, CL intensity and carbon vacancy were observed, whereas the CL originates from the electronic transition from the conduction band to the carbon vacancy Z 1 /Z 2 states and the leakage current is dominated by the PF emission from the carbon vacancy Z 1 /Z 2 states to the conduction band.This also opens the feasibility of screening off the devices with large reversed bias current during the production line using on-line CL inspection before the device fabrication processes complete.

Figure 1 .
Figure 1.Shows the SEM images and the corresponding CL intensity spatial mappings of sample 10 ((a) and (b)); sample 6 ((c) and (d)); and sample 2 ((e) and (f)).The red dotted lines indicate one of the Al-ion implantation regions.

Figure 2 .
Figure 2. Shows the CL spectra of the different samples.The CL signal intensity of a particular photon energy shown is obtained by summing the CL intensity over the whole area of the corresponding SEM images taken at nine different locations of the sample.

Figure 3 .
Figure 3. Shows the integral CL intensities against the reverse bias leakage currents for each of the samples.The reverse bias applied is −1200 V.The integral CL intensity corresponds to the area of the 2.62 eV CL peak shown in the CL spectrum (figure 2).

Figure 4 .
Figure 4. Shows the against E 1/2 of Sample 6 taken at different temperatures from room temperature to 150 • C. The straight lines refer to the fitted lines with equation (1).

Table 1 .
The reverse bias leakage current of ten diodes at −1700 V.

Table 2 .
PF emission fitting results in different temperature.