4H-type silicon carbide (4H-SiC) is almost unanimously regarded as the best material for the next generation of power electronic devices. In fact, in some specific application fields (energy conversion, power supply, consumer electronics, etc.), 4H-SiC metal–oxide–semiconductor field-effect transistors (MOSFETs) can be excellent candidate to replace silicon insulated gate bipolar transistors (IGBTs) in power modules, enabling a lower power consumption at high switching frequencies and junction temperatures above 200 °C.^[1–3] However, the potential of MOSFETs on 4H-SiC has not been fully tapped because of the low channel mobility resulting from an unacceptable high density of interface traps (D_it) near the conduction band edge of 4H-SiC.^[4,5] Electrons trapped by these interface states will lead the free carriers density to decrease in the inversion layer, which reduces the channel conductivity. Additionally, the electron mobility in the inversion layer is lowered due to coulomb scattering by the trapped charges at the interface states. In this sense, a significant improvement of the SiO₂/SiC interface quality has been obtained by introducing annealing steps of the gate oxide in NO or N₂O, which enables channel mobility to reach a value in a range of 20 cm²⋅V⁻¹⋅s⁻¹–50 cm²⋅V⁻¹⋅s⁻¹.^[6–8] The further increase of the channel mobility (up to 89 cm²⋅V⁻¹⋅s⁻¹) in 4H-SiC lateral MOSFETs has been demonstrated by employing post-deposition-annealing (PDA) of the gate oxide in POCl₃.^[9,10] Besides, combined NO and forming gas annealing (FGA) treatment and high temperature (over 800 °C) annealing in diluted H₂ were reported to be efficient for SiC/SiO₂ interface passivation.^[11,12] During the annealing treatment, NO, N₂O, H₂ or POCl₃ molecules decompose. Atoms of N (H/P) diffuse through the silicon–dioxide layer into the 4H-SiC/SiO₂ interface, and react with carbon-related defects there. As a result, the energy position of the interface states shifts into the conduction band of SiC,^[13] which reduces the efficiency of carrier capture at traps. Although some improvements have been achieved in channel mobility, it is still far from the ideal one. It is thought that electrons injected into the interface will be captured by the traps, reducing the effective D_it, or will change the stoichiometry of the interface, improving the interface quality.^[14] Therefore, we present the electrical properties and x-ray photoelectron spectroscopy (XPS) results of the nitrided SiC MOS capacitor after being irradiated by electrons, indicating that electron irradiation can lead the NO-nitrided passivation to be improved.

The starting substrate was an n-type 4H-SiC (0001) wafer covered by a 12-μm epilayer (N_D ∼ 1 × 10¹⁶ cm⁻³). Before dry oxidation, the SiC wafers were cleaned by the standard Radio Corporation of America (RCA) cleaning procedure. Thermal oxides were grown by dry oxidation at 1300 °C for 30 min in pure oxygen atmosphere in HT-RTP4000 cold wall furnace. Afterwards, samples were annealed in NO ambient for 2 h at 1350 °C. The samples were finally pulled out from the furnace in 100% N₂ ambient in 30 min. The final thickness of SiO₂ was measured by UVISEL spectroscopic ellipsometry to be 60 nm. Circular gate electrode and backside ohmic contact were formed by aluminum evaporation and the diameters of the electrode were 100 μm and 300 μm respectively. The electron irradiation of the devices was carried out at Shanghai SN Irradiation Technology Co., Ltd., China at 10-MeV energy with using four different absorbed doses: 0 kGy, 2 kGy, 20 kGy, and 100 kGy. During the irradiation, no bias was applied to the MOS capacitors. XPS analyses were performed using a PHI Quantera scanning x-ray microprobe with a monochromatic Al x-ray source (hν = 1486.7 eV) with an energy resolution of 0.1 eV. Ar sputtering was used to slowly remove the SiO₂ layer. It has been shown that preferential sputtering did not occur on SiC surfaces under Ar bombardment. After each Ar sputtering step, the XPS spectra were recorded by performing multiple scans in each binding energy range of interest with a pass energy of 55 eV. Specified analyzed area was about 100 μm in diameter. The Si 2p, O 1s, C 1s, and N 1s core-level spectra were recorded. Interface state density (D_it) was estimated by the high (1 MHz)–low (quasi-static) method at room temperature through using a computer-controlled Agilent B1500A LCR meter under dark condition. The dielectric breakdown properties were also investigated.

The measured high-frequency C–V curves of SiC MOS capacitors irradiated by electrons at doses of 0 kGy, 2 kGy, 20 kGy, and 100 kGy are shown in Fig. 1(a). The 0-kGy sample corresponds to the non-irradiated reference sample. There is no large flatband voltage (V_FB) shift in the samples subjected to electron irradiation at doses of 2 kGy and 20 kGy compared with the non-irradiated reference sample; however, a positive shift of ∼ 0.6 V is found in the 100-kGy irradiated sample, which indicates a considerable increase on the negative charges generated in the oxide during irradiation. The contribution of oxide-trap charges and interface-trap charges to the V_FB shift are estimated to be ΔV_ot and ΔV_it, respectively. As shown in Fig. 1(a), V_FB and V_ot are determined from the flatband capacitance (C_FB = 3.7 pF) and midgap capacitance (C_mg = 1.5 pF@E_C – E = 1 eV) by the standard formula.^[15,16]V_it indicates the stretch-out between V_FB and V_ot along the gate voltage axis, commonly referred to | V_FB – V_ot|. Figure 1(b) shows the calculated values of ΔV_ot and ΔV_it for the irradiated samples in contrast to the non-irradiated reference. The values of ΔV_ot are quite close to ΔV_FB, and all the samples show quite small ΔV_it values of ∼ 0 V, which suggests that the V_FB shift mainly results from the increase in oxide trap charge density. It could also be concluded that no interface state is generated by electron irradiation in the studied dose range. However, the oxide layer is changed to negatively charged after 100-kGy irradiation. Mobile and fixed oxide charges are positive, whereas the polarities of interface traps may be positive or negative due to holes or electrons trapped depending on their relative positions to the Fermi level. It has been proposed that the interface traps in the lower half of bandgap be negatively charged and the interface traps between midgap and the Fermi level be neutral for the n-type SiC MOS capacitor. Only the traps above the Fermi level are positively charged.^[17] Therefore, the increase in negative charge density is probably attributed to deep interface trap buildup and the reduction of shallow interface traps. Storasta et al.^[14] have reported that low energy electron irradiation can induce the hole trap (HS2 (E_v – 0.39 eV)) in SiC layer, which may be one of the reasons for the negative charges.

Quasi-static C–V (QSCV) measurement is carried out for D_it estimation. The results are shown in Fig. 2(a). There is no significant V_FB shift until an irradiation dose rises up to 100 kGy as happened to C–V curves measured at high frequency illustrated in Fig. 1(a). A frequency dispersion is obtained at the onset of depletion in QSCV measurement, which indicates a reduction in deep D_it. This behavior was not observed in the high frequency C–V results. Thereafter, the D_it value is calculated using high–low method with the expression below

$$ \begin{eqnarray}{D}_{{\rm{it}}}=\displaystyle \frac{1}{q}\left[{\left(\displaystyle \frac{1}{{C}_{{\rm{QS}}}}-\displaystyle \frac{1}{{C}_{{\rm{ox}}}}\right)}^{-1}-{\left(\displaystyle \frac{1}{{C}_{{\rm{HF}}}}-\displaystyle \frac{1}{{C}_{{\rm{ox}}}}\right)}^{-1}\right],\end{eqnarray}$$ (1)

$$ \begin{eqnarray}{E}_{{\rm{C}}}-E=e(0.19\,{\rm{V}}-{\psi }_{{\rm{s}}})\end{eqnarray}$$ (2)

Furthermore, the effect of electron irradiation on near interface trap is evaluated statistically through the hysteresis of high frequency C–V data from the following expression:

$$ \begin{eqnarray}{N}_{{\rm{NIT}}}=\displaystyle \frac{|\Delta {V}_{{\rm{Hys}}}|{C}_{{\rm{ox}}}}{qS},\end{eqnarray}$$ (3)

Figures 3(a) and 3(c) show the Si 2p core-level spectra of the SiO₂/SiC interfaces at sputtering time of 612 s for the NO annealed sample and the electron irradiated samples with NO annealing, respectively. The Si 2p spectrum of the NO annealed sample can be fitted with three Gaussian peaks which are marked as Si-1 [100.1 eV, full width at half maxima (FWHM) = 2 eV], Si-2 (101.8 eV, FWHM = 2 eV), and Si-4 (103.2 eV, FWHM = 2 eV), which relate to SiC from the substrate, Si≡N bond and SiO₂, respectively.^[19] After 100-kGy electron irradiation, the energy position of the peak of Si-4 shifted to the 102.8 eV (Si-3, FWHM = 2 eV), indicating that silicon oxide states exist in the interface region.^[19] It has been reported that O–Si–N defects exist in the crystalline lattice in the NO-annealing treated SiC/SiO₂ interface.^[20] Additionally, it has been shown that the O–Si–N defects are efficient electron trap centers.^[21] Therefore, when a neutral O–Si–N molecule near an interface defect is electrically activated, a negative SiO-charge is created at the SiC/SiO₂ interface. This N-depassivation reaction can be described with the following diffusion-limited electrochemical reaction:^[21]

$$ \begin{eqnarray}{\rm{O}}-{\rm{Si}}-{\rm{N}}+{\rm{e}}-\leftarrow \to {\rm{O}}-{\rm{Si}}-{+{\rm{N}}}^{^\circ }.\end{eqnarray}$$ ()

The electrical breakdown properties of the dielectrics are also studied for all MOS structures after being irradiated by electrons. Figure 4(a) shows the typical current–electric field (I–E) characteristics of samples investigated for each set. The current begins to increase exponentially at electric field above 6 MV/cm, which indicates that the current conduction follows the Fowler–Nordheim (FN) mechanism given by the following equation, which does not consider the Frenkel-pool emission:

$$ \begin{eqnarray}J=\displaystyle \frac{{q}^{3}(m/{m}^{\ast })}{8\pi h{\varPhi }_{{\rm{B}}}}{E}^{2}\exp \left(\displaystyle \frac{-8\pi {(2{m}^{\ast })}^{1/2}{\Phi }_{B}^{3/2}}{3hqE}\right),\end{eqnarray}$$ (4)

In summary, we find that the electron irradiation with a dose of 100 kGy further improves the electrical properties of NO Passivated SiC/SiO₂ interface. The D_it value decreases by about one order of magnitude, specifically, from 3 × 10¹² eV⁻¹ to 4 × 10¹¹ cm⁻²⋅eV⁻¹ at an energy level E_C – E = 0.2 eV despite N_NIT slightly increasing from 4.5 × 10¹⁰ cm⁻² to 5.3 × 10¹⁰ cm⁻². In the case of current-voltage characteristic, electron irradiation helps increase the barrier height and statistical uniformity, however, makes no contribution to E_BD degradation (∼ 9 MV/cm).