The major deep levels that are observed in as-grown n-type and p-type 4H-SiC epitaxial layers are mostly intrinsic defects, and their energy positions in the bandgap are as shown in Figure 4.33. Among them, the [152] and EH6/7 [153] centers are the dominant thermally stable defects that are most commonly observed with the highest concentration in all as-grown epitaxial layers. These levels are also dominant in ion-implanted, plasma-etched, or particle-irradiated SiC, as described in Chapter 6. Although a few hole traps, which may originate from the intrinsic defects, are observed, such as HK4 (or P1) [154], they can be annealed out at 1450–1550 °C. Thus, the and EH6/7 centers are the most common and important deep levels in SiC. Note that the signals from the EH6 and EH7 centers usually overlap severely in normal DLTS spectra, and these two defect levels are thus often treated as the single EH6/7 center, though the EH7 component is dominant in the “EH6/7 peak”. Laplace-DLTS measurements succeeded in resolving this peak, demonstrating that, in terms of generation upon particle irradiation and thermal stability, there are slight differences between the EH6 and EH7 centers [155]. Here, the features of the and EH6/7 centers are summarized.
The center is the most important defect, because of its abundance and because of its property of being a major carrier-lifetime killer [171, 172]. To obtain a long carrier lifetime, which is beneficial for reduction of the on-resistance in bipolar devices, the density of the center must be decreased to the order of or lower. Thus far, two successful techniques have been proposed to eliminate the center.
In the first technique, excess carbon atoms are introduced from the outside and the diffusion of these carbon atoms into the bulk region is promoted by high-temperature annealing [173, 174]. This can actually be achieved by carbon ion-implantation and subsequent annealing in Ar at 1650–1700 °C. Figure 5.49 shows the DLTS spectra obtained from n-type 4H-SiC epitaxial layers before and after carbon ion-implantation followed by high-temperature Ar annealing. Both the and EH6/7 centers are eliminated by this process down to the detection limit (approximately in this case). Because the created carbon interstitials have large diffusion coefficients above 1400–1500 °C, the excess carbon interstitials diffuse and fill the carbon vacancies, leading to the elimination of the and EH6/7 centers from the surface region to the bulk. After annealing, the defective implanted region near the surface is removed by plasma etching.
In the second technique, thermal oxidation of SiC under appropriate conditions results in the elimination of the and EH6/7 centers. During thermal oxidation, the carbon atoms are mostly removed by the creation of carbon monoxide (CO). However, some portion of the carbon atoms is emitted to the bulk region and diffuses into a deep region of SiC. Because silicon atoms are emitted during thermal oxidation of Si [175], it is expected that carbon (and also silicon) atoms will be emitted during the thermal oxidation of SiC. As in the case of the carbon ion implantation, the diffusion coefficient of the carbon interstitials is so large that the depth to which the and EH6/7 centers are eliminated can exceed , by either the combination of oxidation and subsequent high-temperature Ar annealing or by high-temperature oxidation (1300–1400 °C) [176–178]. Note that the thermal oxide is removed before Ar annealing. It was also reported that the migration energy of the carbon interstitials is much smaller than that of the silicon interstitials [179], and the self-diffusion coefficient of carbon is much larger than that of silicon [180]. Figure 5.50 shows the depth profile of the density for the 4H-SiC epitaxial layers after various thermal treatments (e.g., oxidation, Ar annealing) [178]. By extending the oxidation time or raising the oxidation temperature, the “” region extending from the surface becomes thicker ( or even thicker).
In both the deep-level reduction techniques (carbon ion implantation and thermal oxidation), the depth of the “” region depends not only on the process conditions but also on the initial density. When the initial density is relatively low , it is easy to eliminate these deep levels in regions. However, a thick “” region can hardly be obtained if the initial density is high . The depth profiles of the density after these defect-reduction processes can be predicted through numerical simulation of the carbon diffusion process [178]. It was noted that carbon diffusion into the SiC bulk region results in the generation of a few new deep levels, such as the ON1 , ON2 , and HK0 centers [181, 182], which are commonly observed near the surface in both carbon-implanted SiC and oxidized SiC. The HK0 center is annealed out by thermal treatment at temperatures above 1350–1400 °C, while the ON1/ON2 centers (the same defect, but with different charge states) survive even after annealing at 1800 °C. A carbon-interstitial-related defect such as the carbon di-interstitial was suggested as the origin of the HK0 center [182].
For details of other intrinsic deep levels in SiC, such as [152], UT1 [183], HS2 [165], and HK4 [154], please see the individual references provided or review papers [152, 184]. The exact correlation between the observed deep levels and a specific defect structure (e.g., silicon vacancy, antisite) has not yet been established. Figure 5.51 shows the energy positions of the various intrinsic defects that are expected in 4H-SiC, which were obtained through theoretical calculations [179, 185, 186]. The energy levels of the silicon vacancy , the carbon antisite–carbon vacancy pair , the carbon di-interstitial , and other defects have been reported. The diffusion of these intrinsic defects has also been investigated through calculations [179]. However, the experimental assignment of actual deep levels to these defects has not yet been achieved.
The purity of the SiC epitaxial layers grown by chemical vapor deposition (CVD) is rather high, although the substrates do contain various impurities, including heavy metals. Other than the intentional dopants, the major impurities that are present in SiC epitaxial layers are boron and titanium, as described in Section 4.3.2. The typical densities of these impurities are for boron and for titanium. Several other impurities are also occasionally observed. Figure 5.52 shows the energy positions of the major metallic impurity-related deep levels that are observed in 4H-SiC.
Boron (B) atoms substituting the silicon lattice sites create boron acceptors [187]. However, some boron atoms form pairs with carbon vacancies [188], leading to the formation of a “deep boron” level; this is the D center in 4H-SiC [187]. The energy position of the D center is deeper in 6H-SiC . The D center is also acceptor-like [189], and contributes to the DAP luminescence at relatively long wavelengths. The generation of the D center in SiC epitaxial layers is enhanced under Si-rich growth conditions and is reduced under C-rich conditions [20]. The D center is thermally stable, but can be reduced by thermal oxidation or by carbon ion implantation, which is reasonable because interaction of the D center with a carbon interstitial will naturally annihilate the defect.
Oxygen (O) is a common impurity in almost all semiconductor materials, including Si and GaAs. Because nitrogen is a dominant residual impurity in high-purity SiC, oxygen contamination is naturally expected. However, it was found that oxygen is hardly incorporated in SiC during epitaxial growth, even when an oxygen-containing gas is intentionally introduced into the CVD reactor [190, 191]. Etching of or by hydrogen that is supplied as a carrier gas, or decomposition of at the high growth temperatures may suppress oxygen incorporation. No deep levels that could be attributed to pure oxygen-related defects have been identified. The oxygen density in SiC seems to be very low ( or lower).
Titanium (Ti) is another common impurity in SiC, and it creates very shallow electron traps near the conduction band edge in 4H-SiC [29]. Figure 5.53 shows schematically the energy positions of the Ti levels in the different SiC polytypes. The Ti levels with respect to EV are almost identical when plotted in the bandgaps of the different polytypes, indicating that the Langer–Heinrich rule [192] is valid for Ti in SiC. Ti is electrically active in 4H-SiC, but its acceptor levels are resonant in the conduction band for 6H-, 15R-, and 3C-SiC. As described in Section 5.1.1, Ti exhibits a unique greenish luminescence in 4H-, 6H-, and 15R-SiC, but not in 3C-SiC.
Vanadium (V) is an amphoteric impurity, and it forms an acceptor level in the upper half of the bandgap and a donor level in the lower half of the bandgap. In 4H-SiC, the acceptor level and the donor level are located at and , respectively [31]. Thus, vanadium creates a very deep level, especially in p-type SiC, and is thermally stable. This is why vanadium doping can be used to form semi-insulating SiC, as described in Section 3.4.4. In SiC epitaxial layers grown by CVD, however, the vanadium density is usually less than .
The deep levels of chromium (Cr), molybdenum (Mo), tungsten (W), and iron (Fe) in 4H-SiC have been identified [7, 193–195]. In infrared PL spectra from 4H-SiC, the sharp PL peaks at 1170 and 1171 nm (1.0586 and 1.0595 eV) are well known and have been assigned to a UD center (UD1). In recent years, systematic investigations of this defect center by PL, EPR, DLTS, and theoretical calculations have revealed that the origin of UD1 is actually tungsten [36]. In high-purity SiC epitaxial layers, however, the densities of Cr, Mo, W, and Fe are lower than . It was demonstrated that some transition metals such as Ni and Cr exhibit significant diffusion in SiC at temperatures higher than 1500 °C [196].
Correlation between the carrier lifetimes and the density of the center was suggested at an early stage of the studies of SiC [167, 197] and, in recent years, the center has been unambiguously identified as a major carrier lifetime killer, at least, in n-type 4H-SiC [171, 172]. Figure 5.54 shows the inverse of the carrier lifetime versus the measured density of the center for n-type 4H-SiC epitaxial layers [172, 198]. When the density is higher than , the inverse of the carrier lifetime is proportional to the density, indicating that the lifetime is governed by Shockley–Read–Hall (SRH) recombination via the center. However, the correlation between the lifetime and the density is unclear when the density is in the range. As described in Section 5.1.4, multiple recombination processes exist, including SRH recombination and several other recombination processes. Thus, the carrier lifetime can be expressed by the following equation:
where is the SRH lifetime governed by the recombination centers, and is the carrier lifetime that is governed by the other recombination processes, such as surface recombination, recombination in the substrate, Auger recombination, and recombination at extended defects. Here, the inverse of is proportional to the density of the recombination centers (, where is a constant, and is the density of the center), while can be assumed to be independent of the density. By using the model expressed by Equation 5.19, the experimental data can be fitted. The fitted result is shown in Figure 5.54 as a solid line for , and as two broken lines for and . Here the line is determined by recombination at the surface and in the substrate, due to the limited thickness of epitaxial layers ( in this plot), and this component decreases by using thicker epitaxial layers. The detail was explained in Section 5.1.4.4.
Based on the results described above, a simple relationship between the high-injection bulk lifetime and the density can be established as follows:
Note that this equation is valid when Auger recombination and the recombination at extended defects are less important, and it is independent of both the epilayer thickness and the SRV (surface recombination velocity). Instead, the factor of is a function of the excitation intensity (i.e., the injection level) and the temperature.
In semiconductor textbooks, it is stated that a midgap level is an effective recombination center and can thus be a lifetime killer [157]. In 4H-SiC, however, the midgap level of the EH6/7 center does not work as a carrier lifetime killer [199]. This can be deduced by accounting for the properties of the and EH6/7 (EH7) centers. As described in Section 5.3.1, the origin of both the and EH7 centers is a carbon vacancy with different charge states:
In n-type 4H-SiC, the Fermi level is well above the center, and the defect is charged to . When excess electron–hole pairs are generated, a hole is first captured at the defect, because of the large hole capture cross-section (the attractive Coulomb force between and the hole). Then, the defect changes its charge state as follows: . If the hole density is extremely high, then may proceed. Under these circumstances, the capture cross-sections of the electrons for and are very large, because of its property. Thus, the defect immediately tends to revert to the original state; or , and subsequently . These capture and recombination processes are repeated upon electron–hole excitation. Therefore, only the center is involved in the recombination process. The EH7 center does not appear during this process. This is why the EH7 center is not important in carrier recombination in n-type 4H-SiC. The EH7 center may, however, be important in p-type 4H-SiC, where the Fermi level is close to the valence band and the defect is charged to .
Figure 5.55 shows the decay curves at room temperature that were obtained from a n-type 4H-SiC epitaxial layer [100]. The decay curves for the as-grown material and that after reduction through thermal oxidation at 1400 °C for 48 h are shown. For the as-grown epitaxial layer, the measured lifetime is , while the lifetime has improved remarkably to after the defect reduction process. The depth of the region of this particular sample is estimated to be about , indicating that the defect is eliminated throughout the entire thickness of the epitaxial layer. After lifetime measurement of the defect-eliminated sample, the surface was passivated with a 20-nm-thick deposited oxide and annealed in nitric oxide (NO) at 1250 °C for 30 min. After this surface passivation, the lifetime increased to , and the lifetime increased further to at a measurement temperature of 200 °C. Because a relatively low interface state density is obtained over the whole energy range of the bandgap by this passivation process [200], surface recombination may be suppressed, leading to a longer carrier lifetime. Similar carrier lifetimes can be achieved by carbon ion implantation and subsequent Ar annealing [201].
In general, the experimentally determined carrier lifetimes are underestimated, because of parasitic recombination paths, as described in Section 5.1.4. In Figure 5.56, the measured lifetimes are plotted as a function of the epitaxial layer thickness. The results for samples taken before oxidation (as-grown), after oxidation, and with the subsequent surface passivation are indicated by closed circles, open circles, and closed triangles, respectively. In the oxidized samples, sufficiently long-term oxidation was performed at 1300–1400 °C to ensure that the center is eliminated throughout the entire thickness of the epitaxial layers. The lifetimes of the oxidized samples increase rapidly with increasing epitaxial layer thickness, while the thickness dependence is very small for the as-grown samples. This result indicates that the measured lifetime is severely affected by the carrier recombination in the substrate (or near the epitaxial layer/substrate interface) when the bulk lifetime becomes very long because of elimination, which is consistent with Figure 5.17. In Figure 5.56, the simulated dependence of the lifetime on the epilayer thickness is also indicated by dashed lines for the various bulk lifetimes of the epitaxial layers. The SRV was assumed to be . As determined from Figure 5.56, the real bulk lifetime will be longer than . The ambipolar diffusion length is given by the following equation [1, 151]:
Here, and are the diffusion coefficients of the electrons and the holes, respectively. Ambipolar diffusion is described in greater detail in Section 7.3.1. By accounting for the diffusion coefficients and assuming a long carrier lifetime of , the diffusion length of the carriers is estimated to be longer than in the region.
In Si, heavy metallic impurities, such as Au, Pt, and Fe, create midgap levels that act as efficient recombination centers [151]. Although the deep levels of several metallic impurities (V, Cr, W) are known in SiC, the densities of these impurity levels determined by DLTS measurements are usually close to the detection limit (about ). Therefore, it can be concluded that the SRH recombination center (i.e., the lifetime killer) is indeed the center in n-type 4H-SiC.
The carrier lifetime in p-type SiC is more complicated. Because the Fermi level is close to the valence band in p-type SiC, the carbon vacancy defect must be positively charged (: EH7 center) in equilibrium. Thus, it is expected that any excess electrons are quickly trapped by the defect, and the charge state will change to neutral. However, accurate capture cross-sections for the electrons and holes are not yet known for this neutral defect. Figure 5.57 shows the decay curves at room temperature that were obtained from a , lightly-doped p-type 4H-SiC epitaxial layer [202, 203]. The decay curves for the as-grown material and for the material after carbon-vacancy reduction via thermal oxidation at 1400 °C for 48 h are shown. The measured lifetime of the as-grown epilayer is , and the lifetime increased to after the defect reduction process. The lifetime improved slightly to following surface passivation with a nitrided oxide [202, 203]. Carbon ion implantation followed by Ar annealing at 1650 °C, instead of thermal oxidation, provides a very similar result. Thus, reduction of the carbon vacancies (i.e., the and EH6/7 centers) is effective for improving the carrier lifetimes in p-type SiC, but the level of improvement is much smaller than that in n-type SiC. Further investigations are required to identify the lifetime killer and enhance the carrier lifetime in p-type SiC.
The effects of extended defects on carrier lifetimes have also been investigated [68, 204, 205]. All the extended defects cause local reduction of the carrier lifetimes near the defects. This is already indicated in the PL mapping data shown in Figure 5.26, where the PL intensities of the free exciton peaks are greatly reduced near the dislocations. The magnitude of the reduction in PL intensity is much larger near TSDs than TEDs. Figure 5.58 shows the PL decay curves that were measured from a n-type 4H-SiC(0001) epitaxial layer at room temperature. The measurements were conducted at a defect-free region, near a TSD core, a TED core, and inside a single SSF expanded from a BPD. The carrier lifetime obtained from the PL decay was from the defect-free region, near the TED, near the TSD, and for the SSF. Thus, the stacking faults (nucleated from BPDs) and the TSDs are the most detrimental defects for the carrier lifetimes in SiC. The effects of these extended defects depend strongly on the crystal quality, and are relatively larger in high-quality SiC with long carrier lifetimes.
Because the density of the (and the EH6/7) center can be intentionally increased by low-energy electron irradiation [158, 165], lifetime control is easily achieved in n-type SiC. Although several deep levels, which may be ascribed to the interstitial-related defects, are also generated by electron irradiation, any deep levels other than the and EH6/7 centers are eliminated by annealing at 900–1000 °C [159, 165]. Figure 5.59 shows the density of the center generated by electron irradiation (after 950 °C annealing) versus the electron fluence [158, 163]. The electron irradiation was carried out at energies of 116, 160, 200, or 250 keV. In either case, the center density increases almost in proportion to the electron fluence. By increasing the fluence from to along with the irradiation energy, the center density can be controlled over a very wide range from approximately to .
Figure 5.60 shows a map of the carrier lifetimes for n-type 4H-SiC epitaxial layers after electron irradiation and subsequent annealing in Ar at 950 °C for 30 min. The starting material was a 4H-SiC epitaxial layer, grown on an n-type substrate and doped to . Electron irradiation was performed at 160 keV with the fluence range from to on the sample being controlled by using a copper mask during irradiation. Because selective electron irradiation was performed, different densities ranging from to over six different areas were realized in the sample. The carrier lifetime is shorter in the area where the electron fluence is higher (i.e., the density is higher). More importantly, a highly uniform carrier lifetime can be realized in each area. These results indicate that the carrier lifetime can be controlled reasonably well by the electron irradiation process. This technique is useful for obtaining very uniform carrier lifetimes or for the reduction of the switching losses in SiC bipolar devices. Control of the carrier lifetimes by similar low-energy electron irradiation is also possible for p-type 4H-SiC [203], suggesting that a carbon vacancy (e.g., the or EH6/7 centers) can be a lifetime killer when its density is sufficiently high .
Basic characterization techniques have been described in this chapter with emphasis placed on the special care that is required for the characterization of SiC. PL is a powerful technique for characterization of optical properties, and can reflect the incorporation of specific impurities (including dopants), point defects, and any other localized levels. Raman scattering is useful for the identification of the SiC polytypes and for characterization of strains. measurements give the net doping density , while the carrier density can be measured directly only by Hall effect measurements. The carrier lifetimes can be determined by several techniques, but the decay time obtained does not always mean the carrier lifetime. Special attention must also be paid to the effects of surface recombination and recombination in the underlying substrate.
Extended defects have been detected via destructive methods such as molten KOH etching. High-resolution X-ray topography and PL mapping/imaging serve as attractive techniques for nondestructive identification of the locations and types of extended defects. Deep levels, which originate from point defects or impurities, can be characterized by DLTS, and high-temperature (up to 800 K) measurements are required to monitor the midgap levels in SiC. To identify the origin of a deep level, a detailed comparison study of the results of DLTS, EPR measurements, and theoretical calculations is required.
Because micropipe defects have been eliminated, the macroscopic defects that are generated during epitaxial growth, such as triangular defects, carrot defects, in-grown stacking faults, and particles, are the most detrimental defects for any SiC device. A BPD is split into two partial dislocations with a single SSF in between them. This stacking fault is then expanded upon carrier injection and recombination, leading to a considerable reduction of the carrier lifetimes and increased leakage current. Therefore, BPDs cause the degradation of bipolar-type SiC devices (but not unipolar devices). The effects of TSDs on the breakdown voltage and leakage current are very small when no surface pits are formed at the TSD locations. The effects of TEDs are mostly negligibly small. The major carrier lifetime killer has been identified as the center , which is the acceptor level of a carbon vacancy, in n-type 4H-SiC. In p-type 4H-SiC, however, the carrier-lifetime killers have not yet been conclusively identified. TSDs and TEDs cause local reduction of the carrier lifetimes, while BPDs induce expansion of the SSFs upon carrier injection, leading to significant reduction of the carrier lifetimes in the fault region. However, further investigations are required to fully clarify and understand the behavior of the extended and point defects in SiC.
The present chapter did not describe the characterization of SiC by other techniques, such as standard X-ray diffraction, TEM, secondary ion mass spectrometry (SIMS), X-ray photoelectron spectroscopy (XPS), and Auger electron spectroscopy (AES). These techniques are frequently used for material characterization but special care and unique analyses are not always required for SiC. Please see individual textbooks for details of these techniques.