Generation of deep levels near the 4H-SiC surface by thermal oxidation

Deep levels near the surface of 4H-SiC after dry oxidation were investigated. A large and broad peak appeared in the low-temperature range of deep level transient spectroscopy (DLTS) spectra after oxidation of SiC at 1300 °C, indicating multiple deep levels energetically located near the conduction band edge are generated inside SiC by thermal oxidation. Analyses of the DLTS spectra acquired with changing the bias voltage revealed that the majority of deep levels is located very near the SiC surface, within about 6 nm deep region from the surface. The area density of the observed deep levels is higher than 3 × 1012 cm−2.

Deep levels near the surface of 4H-SiC after dry oxidation were investigated.A large and broad peak appeared in the low-temperature range of deep level transient spectroscopy (DLTS) spectra after oxidation of SiC at 1300 °C, indicating multiple deep levels energetically located near the conduction band edge are generated inside SiC by thermal oxidation.Analyses of the DLTS spectra acquired with changing the bias voltage revealed that the majority of deep levels is located very near the SiC surface, within about 6 nm deep region from the surface.The area density of the observed deep levels is higher than 3 × 10 12 cm −2 .© 2024 The Author(s).Published on behalf of The Japan Society of Applied Physics by IOP Publishing Ltd S ilicon carbide (SiC) power transistors have emerged as next-generation high-voltage and low-loss power switching devices, [1][2][3][4] and recent progress in both the material and device technologies has resulted in volume production of 4H-SiC power MOSFETs. Hwever, a high density of defects present near the oxide/SiC interface severely limit the performance and reliability of SiC power MOSFETs.[5][6][7][8][9][10][11] The atomistic structure of the interface defects is still unidentified despite intensive experimental and theoretical studies in the last decades.[12][13][14][15][16][17][18][19] Although near-interface traps located inside SiO 2 and carbon clusters at the interface have been suggested as the major origins of the interface defects, it is not easy to show a direct proof of these defect models.
The authors' group reported that carbon vacancy defects inside the SiC bulk region can be almost eliminated by thermal oxidation. 20,21)The carbon vacancy elimination by thermal oxidation can be interpreted by a model that excess carbon atoms are emitted into the SiC bulk region by thermal oxidation and the emitted carbon atoms diffuse toward the deep region, leading to filling carbon vacancies. 21)The nearly carbon-vacancy-free region is extended to a more than 100 μm depth from the surface by prolonging the oxidation period. 22)The authors' group also discovered generation of both electron and hole traps near the SiC surface by thermal oxidation. 23,24)The densities of these generated traps are in the 10 12 -10 14 cm −3 range and exhibit a rapid decay along the depth direction in a 3-20 μm range, where the depth profiles become deeper with elevating the oxidation temperature or increasing the period.These depth profiles of trap densities imply that a higher density of traps may exist very near the SiC surface after oxidation.However, deep levels in a sub-surface (<0.1 μm depth) region after thermal oxidation have not been investigated.One of the major reasons for the lack of deep level information in the sub-surface region originates from the use of lightly-doped SiC for deep level transient spectroscopy (DLTS) measurements in previous studies, where the depletion layer is extended to a deep region.
In this study, the authors tried to monitor deep levels near the conduction band edge (E c ) in a sub-surface region after thermal oxidation by using heavily-doped n-type 4H-SiC epitaxial layers.It was revealed that deep levels are actually generated in a sub-surface (<6 nm) region by thermal oxidation.The area density of the generated traps is in the same order as that of electrons induced by the gate voltage in n-channel MOSFETs.
The starting materials were 100 μm thick n-type (nitrogen-doped) 4H-SiC homoepitaxial layers grown on 4°off-axis (0001) wafers.The donor density of the epilayers was 5 × 10 18 cm −3 so that the depletion layer is extended to only a few tens nm from the surface.Figure 1 shows (a) the sample preparation flow and (b) the depletion layer width versus the bias voltage for a Ni/SiC Schottky structure.In the first step, dry oxidation was conducted at 1300 °C for 15 min or 10 h, which yielded an oxide thickness of about 30 nm or 410 nm, respectively.It is noted that post-oxidation nitridation was not performed to investigate purely oxidation-induced defects.After removing the oxide by etching with concentrated HF, Ni Schottky contacts with a diameter of 500 μm were deposited on the surface.The backside ohmic contact was formed by Al deposition.Postmetallization annealing was not performed to avoid potential chemical reactions at the Ni/SiC interface.The depletion layer width shown in Fig. 1(b) was extracted from the capacitance-voltage measurement on the Schottky structure, indicating that the depletion layer width is as thin as 18 nm at zero bias.For reference, a sample without oxidation was also prepared.A native oxide of this reference sample was removed by etching with concentrated HF before Ni deposition.
Fourier Transform DLTS measurements 25) were conducted on the fabricated Ni/SiC Schottky structures with changing the filling pulse voltage (V P ) and the reverse-bias voltage (V R ) to obtain the depth profiles of observed trap densities.Figure 2 depicts the bias voltage sequence, the energy band diagrams under the filling pulse and after the filling pulse.In these band diagrams, w P and w R ∞ are the depletion layer widths at the filling pulse voltage and the reverse-bias voltage after sufficiently long time, respectively.And λ is the lambda length where electron emission from traps is negligibly small during the transient time after the filling pulse. 26)In DLTS measurements, only traps located in the depth range from w P − λ to w R ∞ − λ are detected. 26)This phenomenon must be carefully considered to extract the depth profile of trap density.Here λ is defined by the following equation. 26) where E f , E t , and N d are the Fermi level, the trap level, and the donor density, respectively.And ε s , ε 0 , and q are the relative dielectric constant (10.3 for 4H-SiC), the dielectric constant in vacuum, and the elementary charge, respectively.Using Eq. ( 1), λ can be calculated to be 4.8, 8.3, and 10.7 nm for E f − E t = 0.1, 0.3, and 0.5 eV, respectively.With a N d of 5 × 10 18 cm −3 used in this study, E f is located at (0.02-0.05) eV below E c in the temperature range from 100 to 300 K, indicating that E f is very close to E c in the present study.Figure 3(a) shows the DLTS spectra taken from the sample oxidized at 1300 °C for 10 h.Here the filling pulse voltage was varied from 0.8 to 1.5 V with keeping the reverse-bias voltage -1.5 V. Large DLTS signals showing a very broad peak are observed in the spectrum acquired with V P = 1.5 V.The observed broad DLTS peak must be caused by severe overlapping of DLTS peaks originating from several traps at different energy levels.A similar DLTS spectrum was observed for the sample oxidized at 1300 °C for 15 min, though the DLTS signal intensity was about 60%-80% smaller (not shown).It is noted that these DLTS signals come from electron traps present inside the SiC bulk region, because the oxide was removed by HF.The DLTS signals especially in the low-temperature range show a rapid drop with decreasing V P , and a clear DLTS peak around 280-360 K is identified when V P = 0.8 V.In contrast, the DLTS signals acquired from the reference sample (without oxidation) were below the detection limit (trap density <5 × 10 14 cm −3 ). Figure 3(b) shows the w P − λ and w R ∞ − λ as a function of the filling pulse voltage, where a trap level energetically located at E c − 0.10 eV was considered.With this trap level, λ is calculated to be 3.7 nm, where a temperature of 200 K was assumed.The depth w P − λ becomes far from the Schottky interface (SiC surface) with decreasing the filling pulse voltage, whereas the depth w R ∞ − λ is constant (about 22 nm).Since the detection region x is defined as w P − λ < x < w R ∞ − λ as described with Fig. 2, traps located very near the surface (e.g.x < 2 nm) are not detected when the filling pulse voltage is lower than 1.2 V.These results indicate that a high density of traps observed in the low-temperature range of the DLTS spectra, which correspond to energetically shallow traps, are generated very near the surface by thermal oxidation of SiC.Furthermore, judging from the DLTS spectrum acquired with     Figure 4(a) depicts the DLTS spectra taken from the sample oxidized at 1300 °C for 10 h under different measurement conditions from those in Fig. 3.In this measurement, the reverse-bias voltage was varied from −0.5 to −3.5 V while keeping the filling pulse voltage 0.8 V.A major DLTS peak (labeled as OI-3) was observed at 320 K with minor peaks at about 180 K (OI-1) and 220 K (OI-2).From the Arrhenius plot of the emission time constant (not shown), the energy level of the electron traps labeled as OI-3 center was determined to be E c − 0.35 eV.Though the analyses of the minor peaks were difficult, the energy levels were roughly estimated as E c − 0.19 eV for OI-1 and E c − 0.24 eV for OI-2.All these peaks became small with increasing the absolute value of the reverse bias voltage.Figure 4(b) shows the w R ∞ -λ and w P − λ as a function of the reverse bias voltage, where a trap level energetically located at 0.35 eV below the conduction band edge (OI-3 center) was considered.With this trap level, λ is calculated to be 8.1 nm at 300 K.Here the filling pulse voltage was carefully selected so that w P − λ is always apart from the surface (Ni/SiC interface).Thus, no electron emission from the surface states is included in the DLTS signals, indicating that the OI-3 center is a bulk defect inside SiC.Since the observed DLTS peak intensity corresponds to the trap density averaged within the detection region (w P − λ < x < w R ∞ − λ) in DLTS measurements, the smaller DLTS signal intensity with increasing the absolute value of the reverse bias voltage means that the majority of these traps is located very close to the surface (x < 10 nm).
In order to estimate the depth profile of the observed deep levels, an analytical approach to reproduce the experimental results 27) was taken.In DLTS measurements, the trap density averaged within the detection region  N T (w P , w R ) is given by the following equation. 27) Here ΔC and C R are the DLTS peak intensity of the trap and the steady-state capacitance at the reverse-bias voltage V R , respectively.N T (x) is the actual trap density as a function of the depth x.When an appropriate function form for N T (x) with a few fitting parameters is assumed, the trap density averaged within the detection region  N T (w P , w R ) can be calculated.By fitting the calculated  N T (w P , w R ) to the relationship between the averaged trap density [(2ΔCN d )/C R ] and the depletion layer width (w P or w R ) experimentally obtained by DLTS measurements, the fitting parameters can be determined, leading to extraction of the depth profile of the actual trap density N T (x).
In the present case, of course, it is unknown which function form is appropriate to reproduce N T (x).As a trial, the authors assumed a half Gaussian function for N T (x) of the OI-3 center in this study.This assumption is based on the fact that the depth profiles of several deep levels generated by thermal oxidation of SiC can be mostly expressed by this function, 23,24) although the detected deep levels are located in a much deeper region (several μm from the surface) in the previous reports.This half Gaussian function could reflect the depth profile for some kind of defects which diffuse from the surface to the bulk region; one potential example is diffusion of carbon atoms during oxidation of SiC. 20,21)Therefore, N T (x) of the OI-3 center was assumed by where the trap density at the surface (x = 0) N 0 and the standard deviation σ are fitting parameters.respectively.From the fitting, the parameters were determined as N 0 = 1.2 × 10 19 cm −3 and σ = 2.0 nm.The extracted N T (x) of the OI-3 center is shown as the inset of Fig. 5.It is noted that other function forms for N T (x) such as a single exponential function (∼exp(-ax): a is a constant) and a step function did not yield a better fitting result than that shown in Fig. 5.The extracted N T (x) demonstrates an extremely high trap density over 1 × 10 19 cm −3 at the surface.This defect center (OI-3) is localized very close to the surface (x < 6 nm), which corresponds to a depth, where an inversion layer is formed in MOSFETs.Furthermore, its area density is estimated to be as high as 3 × 10 12 cm −2 , being in the same order as the electron density in an inversion layer for the on-state of a MOSFET.Therefore, the observed deep levels must severely work as electron traps, leading to a very low effective mobility in SiC MOSFETs.][30] The microscopic structures of the OI-1, OI-2, and OI-3 centers are unknown, but a recent first-principles calculation study suggested that a dicarbon anti-site defect, (C 2 ) Si , in 4H-SiC creates a defect level at E c − 0.4 eV, 17) which does not contradict with the obtained energy level of the OI-3 center (E c − 0.35 eV).
It is noted that the OI-1, OI-2, and OI-3 centers contribute to only a small fraction of the whole DLTS spectra acquired with a large filling pulse voltage shown in Fig. 3(a).Figure 3(a) indicates that there exist an even higher density of different electron traps, which are energetically located at shallow levels near E c because of the continuous DLTS signals at lower temperatures from 100 to 250 K.And these shallow levels must be more localized near the surface than the OI-1, OI-2, and OI-3 centers, as discussed with Fig. 3(b).It was, however, difficult to distinguish individual DLTS peaks at present, due to the featurelessly broad spectrum in the low-temperature range.
In summary, the authors discovered that several defect levels are generated inside the SiC bulk region very close to the surface by dry oxidation.One of the detected deep levels, the OI-3 center, is energetically located at E c − 0.35 eV and is localized very close to the SiC surface within about 6 nm.There exist other shallower levels and the total area density of these electron traps exceeds 3 × 10 12 cm −2 .

Fig. 1 .
Fig. 1.(a) Sample preparation flow and schematic illustration of a fabricated sample.(b) Depletion layer width versus bias voltage for a Ni/SiC Schottky structure used in this study.

Fig. 2 .
Fig. 2. Bias voltage sequence in DLTS measurements, energy band diagrams under the filling pulse [A] and after the filling pulse [B].In these band diagrams, w P and w R ∞ are the depletion layer widths at the filling pulse voltage and the reverse-bias voltage after sufficiently long time, respectively, and λ is the lambda length.In DLTS measurements, only traps located in the depth range from w P − λ to w R ∞ − λ are detected.

Fig. 3 . 2 ©
Fig. 3. (a) DLTS spectra taken from the sample oxidized at 1300 °C for 10 h.Here the filling pulse voltage was varied from 0.8 to 1.5 V with keeping the reverse-bias voltage -1.5 V.A DLTS spectrum of the reference sample is also shown.(b) w P − λ and w R ∞ − λ as a function of the filling pulse voltage, where a trap level at E c − 0.10 eV was considered.The observed DLTS peak intensity corresponds to the trap density averaged within the detection region (w P − λ < x < w R ∞ − λ) in DLTS measurement.

Figure 5
shows the averaged density of the OI-3 center  N T versus the depletion layer width under the reverse-bias voltage w R ∞ , where the experimental data and calculated (fitting) result are denoted by red open circles and black line,

Fig. 4 .
Fig. 4. (a) DLTS spectra taken from the sample oxidized at 1300 °C for 10 h.In this measurement, the reverse-bias voltage was varied from -0.5 to -3.5 V while keeping the filling pulse voltage 0.8 V. (b) w R ∞ − λ and w P − λ as a function of the reverse bias voltage, where a trap level of E c − 0.35 eV (OI-3 center) was considered.The observed DLTS peak intensity corresponds to the trap density averaged within the detection region (w P − λ < x < w R ∞ − λ) in DLTS measurement.

Fig. 5 . 3 ©
Fig. 5. Averaged density of the OI-3 center  N T versus depletion layer width under the reverse-bias voltage w R ∞, where the experimental data and calculated (fitting) result are denoted by red open circles and black line, respectively.The inset shows the extracted depth profile of actual trap density N T (x) for the OI-3 center.041004-3