M-center in low-energy electron irradiated 4H-SiC

Electrically active defects in n-type 4H-SiC have been studied in detail for decades. Piece by piece, the puzzle behind the most common and dominant defect traps, such as V_C(=/0), V_C(0/++), and V_Si, has been solved. Part of the puzzle that has kept researchers busy for several years is the study of silicon vacancy and carbon interstitial defects (V_Si and C_i), introduced by radiation. It was well known that electron irradiation,^1–3 proton irradiation,^4–6 neutron irradiation,^7,8 and ion implantation^9,10 introduce two deep level defects in the n-type 4H-SiC material. These levels are located at 0.40 and 0.70 eV below the conduction band and have been labeled either as S_1/2 or EH_1/3. Recently, Bathen et al.⁶ have provided conclusive evidence that S_1/2 deep level defects are related to V_Si, while Alfieri and Mihaila³ have shown that EH_1/3 deep level defects are related to C_i. The EH_1/3 deep level defects are introduced only in the case of the low-energy electron irradiation (<200 keV), since under such conditions, silicon atoms cannot be displaced.^1,2

The V_Si has attracted much attention in recent years due to its physical properties and its potential application for quantum sensing.^6,11–13 The S₁ and S₂ are identified as V_Si (−3/=) and V_Si (=/−) charge transitions, respectively. Bathen et al.⁶ have shown that S₁ (in proton irradiated 4H-SiC samples) has two emission lines originating from V_Si sitting at −k and −h lattice sites. These findings were later confirmed by Capan et al.⁸ when studying 4H-SiC material irradiated with fast neutrons.

Despite their technological importance, C_i defects are not yet fully understood. Coutinho et al.¹⁴ have recently shown that a bi-stable defect in 4H-SiC, known as M-center, is carbon interstitial. Accordingly, the defect is responsible for two pairs of the first and second acceptor transitions^4,5,10,14,15

where configurations A and B were assigned to a carbon interstitial at the hexagonal and cubic sub-lattice sites, $Ci(h)$ and $Ci(k)$⁠, respectively. The two configurations can be interchanged by annealing and applying a reverse bias voltage. Configuration A is obtained when the measurement is performed after annealing the sample just above room temperature under reverse bias voltage. The defect jumps to configuration B after the sample is annealed at ∼450 K without bias.

The evident similarities between M_1/3 and EH_1/3 traps, including their location within the bandgap and their formation conditions, are so striking that we must hypothesize that they may ultimately arise from the same defect. Therefore, the main goal of this work is twofold. By using low energy electrons and low fluence, we intend to introduce only the EH_1/3 deep level defects and verify if there is evidence for bi-stability and the formation of M_1/3. Moreover, since we displace only the carbon atoms, resulting in a very clear EH₁ signal, Laplace-deep-level transient spectroscopy (DLTS) was used to investigate possible superposition of EH₁ and M₁ signals.

In this work, n-type Schottky barrier diodes (SBDs) were fabricated on nitrogen-doped (∼4.7 × 10¹⁴ cm⁻³) 4H-SiC epitaxial layers, with a thickness of approximately 25 μm. The epi-layer was grown on an 8° off-cut silicon face of a 350 μm thick 4H-SiC (0001) wafer without a buffer layer by chemical vapor deposition. The Schottky barriers were formed by thermal evaporation of nickel through a metal mask with a patterned squared aperture of 1 mm edge length, while the Ohmic contacts were formed on the backside of the silicon carbide substrate by nickel sintering at 950 °C in an Ar atmosphere.

Low-energy electron irradiations were performed at Nissin Electric Group (NEG), Kyoto, Japan. The electron energy was 150 keV, and the total fluence was 1 × 10¹⁵ cm⁻². The irradiations have been performed through the Schottky contact (Ni, thickness ≤ 100 nm) at room temperature (RT).

DLTS measurements were performed using a Boonton 7200 capacitance meter (Boonton Electronics, New Jersey, USA) and an National Instruments PCI-6521 data acquisition device (NI, Austin, USA). Conventional DLTS measurements were carried out in the temperature range of 100 to 450 K with a temperature ramp rate of 2 K/min. Reverse voltage, pulse voltage, and pulse width were V_R = −4V, V_P = 0 V, and t_P = 10 ms, respectively. For the Laplace-DLTS measurements, the following acquisition settings were used: the number of samples 3 × 10⁴, the sampling rate 10–80 kHz, and the number of averaged scans 100–800. The numerical routine FLOG¹⁶ was used to calculate Laplace-DLTS spectra.

M-center was transformed to configuration B and configuration A by annealing at 450 K (for 20 min) and cooling the SBD without applying a bias voltage (0 V), or by annealing at 340 K for 20 min and cooling down the SBD with an applied bias voltage of −30 V, respectively.

Figure 1 shows the DLTS spectrum for the as-grown n-type 4H-SiC SBD. In the as-grown sample, only the Z_1/2 peak is present. The Z_1/2 has already been reported in numerous studies and assigned to V_c(=/0).^17–19 

Figure 2 shows the DLTS spectrum for the n-type 4H-SiC SBD irradiated with low-energy electrons in configurations A and B. The DLTS measurement of the as-irradiated sample (without pre-measurements bias and annealing settings needed for the configuration A and configuration B) is not shown here. The measurements did not reveal any new information, and they are similar to the spectra shown in Fig. 2.

The low-energy electron irradiation has increased the concentration of Z_1/2. The obtained concentration of ∼10¹² cm⁻³ is in a good agreement with previously reported studies.³ Moreover, irradiation introduced two deep-level defects, whose positions (0.41 and 0.70 eV) are close to those that are usually labeled as EH₁ and EH₃. However, the observed peaks are also metastable and consistent with the properties of the M-center.^4,5,10,14,15 The spectra depend on the applied bias voltage and annealing. The bistable defect, known as M-center, introduces four electrically active deep-level defects. M₁ and M₃ overlap with EH₁ and EH₃ in configuration A, while M₂ overlaps with Z_1/2 in configuration B (Fig. 2). M-states and their bi-stability are more clearly observed if the DLTS signal difference (configuration A − configuration B) is plotted, as shown in Fig. 3. As previously reported, it is not possible to observe the M₄ with DLTS due to the technical limitations, but M₄ has been observed in ion-implanted¹⁰ and neutron-irradiated⁸ 4H-SiC using the isothermal DLTS.

We have estimated the activation energies for M₁, M₂, and M₃ and obtained the following values 0.42, 0.73, and 0.90 eV, which are consistent with previously reported values.^4,5,10,14,15

The DLTS data obtained in this study are in perfect agreement with previous results from DLTS studies on low-energy electron irradiated 4H-SiC samples.¹ However, triggered by the recent advancements in understanding and identifying the M-center,¹⁴ accompanied by the recent progress in understanding the EH_1/3 and S_1/2 defects,^3,6 we have applied the Laplace-DLTS technique to study the EH₁ in more details and to investigate the correlation between the EH₁ and the M₁ defect.

Figure 4 shows Laplace-DLTS spectra in configurations A and B, measured at the temperature of the EH₁ DLTS peak (T = 200 K). The Laplace-DLTS results clearly indicate that EH₁ defect has a single emission line, with no splitting of the emission line due to the different lattice sites (−h and −k) as is the case for S₁ and/or Z_1/2 (EH_6/7).^6,8,17,18,20 Thus, our results clearly show that Laplace DLTS can be used as a tool to distinguish S₁ (not shown here) and EH₁, which give identical signals in the DLTS spectrum. This is the first time that EH₁ has been directly measured with Laplace-DLTS.

According to the recent findings of Coutinho et al.,¹⁴ the M-center was assigned to C_i. The four states arising from the M-center are assigned to different charge states located at the different lattice sites. M₁ and M₃ are assigned to $Ci=h$ and $Ci0(h)$⁠, while M₂ and M₄ are assigned to $Ci=k$ and $Ci0 k$⁠. The occupancy of the −h or −k lattice sites is controlled by the reverse bias voltage anneals. For example, if we perform zero bias annealing (configuration B), the −k sites will be dominantly occupied, leading to the appearance of M₂ and M₄ in the DLTS spectra. However, if we perform reverse bias annealing (configuration A), then the occupation of the −h sites prevails, which gives rise to M₁ and M₃ in the DLTS spectra.

According to the Laplace-DLTS measurements at temperatures around 200 K (i.e., temperature at which EH₁ has the peak maximum in the DLTS spectrum), only one emission line is observed in both configurations, A and B. There is no convincing evidence or even suggestions for two overlapping defects with identical emission lines that are unresolved by Laplace-DLTS. Let us assume that M₁ is indeed EH₁. The difference in the intensity of EH₁ peak (DLTS spectra) in configurations A and B is not due to the overlap of an additional defect, such as M₁, but to the different occupancy of $Ci=(h)$ sites. For configuration A, as explained above, this is more favorable than −k sites. Therefore, we can speculate that EH₁ is the same defect as M₁ and assigned to $Ci=(h)$⁠.

Based on the difference signal (configuration A—configuration B), as shown in Fig. 3, the concentrations of M₁ and M₂ are identical within the measurements error margin. M₂ has been recently assigned to $Ci=(k).$⁷ This leads us to conclusion that conversion $Ci=h$ ↔ $Ci=k$ occurs more easily than anticipated. As mentioned above, we control the occupancy of the −h and −k sites by different bias voltage annealing. From isothermal annealing, the conversion from configuration A to configuration B was previously measured to be activated with a barrier of 1.4 eV, while the conversion from configuration B to configuration A is activated with a lower barrier of 0.9 eV⁴. It should be noted that these values were estimated in the study of the MeV proton implanted 4H-SiC material. The MeV implantations result in the introduction of the S_1/2 defects (V_Si). By varying the filling pulse length while maintaining the measurement temperature at the S₁/M₁ peak position, Martin et al.⁴ have clearly shown that contributions from at least two different defects, S₁, and M₁ are present in configuration B. These results were recently confirmed by Capan et al.⁸ as S₁ and M₁ have directly been measured with Laplace-DLTS. Not only do we have contributions from S₁ and M₁, but S₁ is additionally resolved into two components. Although the analysis of the conversion barriers (configuration A ↔ configuration B) has been done using the “difference” DLTS signal (presumably this is the case where the signal is only due to the M-center),⁴ we should not completely underestimate the fact that the conversion barriers were not determined under conditions equivalent to those reported in this study. The conversion barriers $Ci=h$↔ $Ci=k$ could be lower than previously assumed.

All these results imply that EH_1/3 and M-center are indeed carbon interstitials, and they are all arising from the same defects. Thus, we conclude that EH₁ and EH₃ are identical to M1 and M3 and assign them to $Ci=(h)$ and $Ci0 h,$ respectively.

Unfortunately, Laplace-DLTS cannot provide conclusive information about EH₃ as is the case with EH₁, since the EH₃ is too close to Z₁(=/0) and Z₂(=/0) in the Laplace DLTS spectrum, and they overlap. The conversion $Ci0 h ↔ Ci0k$ should follow the same path. Further studies with isothermal DLTS are needed since this is the only way to measure M₄ directly.

In this work, we have used DLTS and Laplace-DLTS to study carbon interstitial-related defects (EH_1/3 and M_1/3) in the low-energy electron irradiated 4H-SiC material. Based on the results obtained in this study, we propose that EH₁ and EH₃ are identical to M₁ and M₃ and assign them to $Ci=(h)$ and $Ci0 (h)$⁠, respectively.

This work was supported by the North Atlantic Treaty Organization Science for Peace and Security Program through Project No. G5674.