Influence of thermal growth parameters on the SiO 2 / 4 H-SiC interfacial region

In order to elucidate the origin of SiC electrical degradation from thermal oxidation, 4H-SiC substrates were thermally oxidized under different conditions of time and pressure. Results from nuclear reaction analyses were correlated to those from electrical measurements. Although the increase in the flatband voltage shift and in the film thickness were related to the oxidation parameters, the results exclude the thickness of the SiO2/4H-SiC interfacial region and the amount of residual oxygen compounds present on the SiC surface as the main cause of the electrical degradation from the SiC oxidation.

Silicon carbide (SiC) is a promising semiconductor to replace Si in micro and nanoelectronics device applications that require high-power, high-frequency, and/or high-temperature. 1,2 esides, a SiO 2 film can be thermally grown in a similar way to that on Si, allowing the technology used to produce MOS (metal-oxide-semiconductor) devices to be adapted to the case of SiC. 3 Nevertheless, the oxidation of SiC leads to a higher interface states density (D it ) in the SiO 2 /SiC interface as compared to the Si case. 1 Although successful routes to reduce D it were achieved, like thermal treatments involving NO, N 2 O, and H 2 , [4][5][6] the nature of the defects responsible for the electrical degradation in the SiO 2 /SiC interfacial region is not yet completely understood.
Concerning SiC oxidation, excess of carbon near the formed interfacial region was observed by different techniques, [7][8][9][10] although medium energy ion scattering (MEIS) analysis indicates a stoichiometric SiO 2 film formed from the SiC oxidation. 11Besides, a non abrupt interface between SiO 2 films and SiC was revealed by nuclear reaction profiling (NRP) [12][13][14] differently from the case of silicon oxidation. 12,15 9][20] Such compounds exhibit different properties when using Si or C face terminated substrates. 20These compounds were not observed in the silicon case, 19 being probably related to the presence of silicon bonded to oxygen and to carbon in different stoichiometries, named silicon oxycarbides (SiC x O y ). 21,22 any trials to remove these compounds in wet environments were unsuccessful evidencing their high chemical resistance. 19However, the use of a flux of O 2 bubbled in hot H 2 O 2 proved to be efficient in removing partially these SiC x O y , 23 reducing D it in the SiO 2 /4H-SiC and decreasing the interfacial thickness after further reoxidation steps. 14In order to elucidate how these residual compounds and how the SiO 2 /4H-SiC interfacial region thickness determined by NRP influence the electrical properties of the SiC MOS structures, more investigations must be performed.In this work, we propose to investigate the relation between these characteristics obtained by nuclear reaction analyses and the modification in the electrical properties induced by the thermal growth parameters oxidation time and oxygen pressure.Thus, we expect to achieve a better understanding on the SiC thermal oxidation and on the origin of the electrical defects present in the SiO 2 /SiC interfacial region.
To achieve these goals, different subatmospheric oxygen pressures (of 18 O 2 ) and oxidation times were used to thermally grow thin SiO 2 films on 4H-SiC substrates.Samples were probed by nuclear reaction analysis (NRA) to determine the total amount of oxygen incorporated before and after the removal of the Si 18 O 2 film, and by NRP to determine its depth distribution.Current-voltage (I-V) and capacitance-voltage (C-V) measurements were performed in Al/SiO 2 /4H-SiC MOS structures and correlated with the other results.
Commercial SiC wafers of the 4H polytype, polished in both (0001) and ( 0001) faces (terminated in Si and C, respectively), were employed as substrates.Samples characterized by electrical measurements were 4H-SiC (n-type) commercial epitaxial wafers, 8 • off-axis on the Si face, doped with nitrogen (1.1 × 10 16 cm −3 ), 4.6 μm thick.Wafers were purchased from CREE Inc. Research.All substrates were cleaned in a mixture of H 2 SO 4 and H 2 O 2 followed by the standard RCA (Radio Corporation of America) process. 24Then samples were etched for 60 s in a 1 vol.%aqueous solution of hydrofluoric acid (40 wt.% HF, purchased from Merck) and rinsed in deionized water.Immediately after blow drying with N 2 , 4H-SiC samples were loaded in a static pressure, quartz tube, resistively heated furnace that was pumped down to 10 −7 mbar.SiO 2 films were thermally grown at 1100 • C in different oxygen pressures (50, 100, and 200 mbar) and oxidation times (0.5, 1, 2, 3, and 4 h) of dry O 2 (<1 ppm H 2 O) enriched to 97% in the 18 O isotope, whose natural abundance is 0.2%, named 18 O 2 .Oxygen pressures higher than 200 mbar were not employed in this work due to the use of a N 2 (L) trap to help the base pressure reduction (mainly H 2 O molecules condensation) while keeping O 2 molecules in the gas phase, which would condensate at higher pressures.The use of 18 O is crucial, as the used nuclear reaction analyses allow to distinguish it from oxygen eventually incorporated from other sources (for instance, from exposure to the ambient).The total amount of 18 O in resulting samples was determined by NRA using the 18 O(p,α) 15 N nuclear reaction at 730 keV, 25 referenced to a standard Si 18 O 2 film on Si. 26 The depth distribution of 18 O in the samples was determined by NRP using the narrow resonance at 151 keV in the cross section curve of the 18 O(p,α) 15 N nuclear reaction. 18O concentration profiles were determined from experimental excitation curves (alpha particle yield versus incident proton energy) 27 using the FLATUS code.With the experimental condition used in this work, a sub-nanometric resolution can be obtained near the surface.Al thermal evaporation to obtain MOS structures used a mechanical mask, forming circular capacitors with a diameter of 200 μm.An InGa eutectic was used as back contact.Samples were electrically characterized using a computer-controlled HP4155A Semiconductor Parameter Analyzer for the I-V curves.The C-V curves were taken from inversion to accumulation at 100 kHz with a 0.25 V/s rate using a HP4284A Precision LCR Meter.
The total amount of 18 O and the corresponding Si 18 O 2 film thickness of the SiC samples before etching are presented in top panel of Figure 1.These results are presented as a function of the product of pressure and time (p × t).The motivation for such plot was that for the Si and for the SiC oxidations, despite being valid for a different thickness range (films thicker then ∼25 nm 28,29 ), a given SiO 2 film thickness can be reached by maintaining the product of oxygen pressure and time constant.For the SiC case, the p × t dependence was investigated for oxygen pressures higher than those of the present work, presenting deviation from this behavior for pressures higher than 1 atm in the case of Si face samples.In the present samples, whose thicknesses are in the 3-8 nm interval for the Si face and 7-24 nm for the C face, a linear behavior of the amount of incorporated 18 O can be observed in both faces, although it was expected a more rapid oxide growth rate up to ∼10 nm for both Si and SiC oxidations. 30,31 t is possible to convert the 18 O amount into oxide film thickness by assuming a given density to the oxide film.In the present case, it was assumed 2.21 g/cm 3 , typical of silicon dioxide films thermally grown on Si.However, this conversion might not be accurate due to modifications in the oxide density in the initial oxidation steps. 32Nevertheless, what is being highlighted from these results is that the linear dependence of the 18 O incorporated amount in the silicon oxide films on (p × t) is still valid for the thermal oxidation conditions tested for both Si and C faces of SiC.It means that for the initial stages of oxidation, this relation between oxygen pressure and oxidation time can be used to determine the SiO 2 film thickness on SiC.The residual amount of 18 O after etching in aqueous HF as a function of oxidation time is presented in bottom panel of Figure 1 for samples synthesized under different time oxidation conditions.As already observed, 20 the Si face presents higher residual oxygen amounts than the C face.However, in present results no relation was observed with oxidation time, indicating that the amount of residual compounds is not affected by the oxidation parameter tested.
Figures 2(a) and 2(b) show the excitation curves of the 18 O(p,α) 15 N nuclear reaction and the 18 O concentration profiles obtained for 4H-SiC samples submitted to different 18 O 2 pressures and oxidation times at 1100 • C. The horizontal lines at 97% observed in Figure 2 profiles refer to a stoichiometric Si 18 O 2 film, while the decrease in the 18 O concentration towards zero refers to its concentration in the SiO 2 /4H-SiC interfacial region.Interfacial region thicknesses around 3 nm (considering the thickness interval from when the 18 O concentration starts to decrease until when this concentration is negligible) can be observed for all C-face oxidized samples, in good agreement with our previous results. 13,14 or Si-face oxidized samples, only the sample oxidized under 50 mbar presented a thicker interfacial region (around 3.8 nm).The decrease in the thickness of the SiO 2 /SiC transition layer as the oxide film thickness is increased during the initial stages of oxidation can be attributed to a smoothing effect of the interface, as suggested by Szilágyi et al. 32   modifications in the SiO 2 /SiC interfacial region thickness attributable to the oxygen pressure and oxidation time were observed.
To investigate the SiO 2 /SiC interfacial region thickness in conditions of a longer oxidation time, a sample was oxidized for 10 h in 100 mbar of 18 O 2 at 1100 • C. To avoid a degradation of depth resolution around the interfacial region due to the thicker film, the upper part of the film was partially removed with a controlled etching 19 for 290 s.In this way, the final film thickness was almost the same of the sample oxidized for 1 h, simplifying the comparison.The excitation curve of the 18 O(p,α) 15 N nuclear reaction and the 18 O concentration profile on the C-face are presented and compared to a sample oxidized for 1 h in 100 mbar of 18   no major modification was observed in the SiO 2 /SiC interfacial region thickness, confirming the absence of influence of the oxidation time in this property.
I-V and C-V curves for samples oxidized at 1100 • C for 1 h in 100 mbar, 1 h in 200 mbar, and 4 h in 100 mbar are presented in Figure 3, and their results are summarized in Table I.The I-V curves presented breakdown fields around 8.0 MV/cm, almost independent of the oxidation parameters, indicating minor modifications in the nature of the SiO 2 from the oxidation conditions tested.C-V curves indicate that the increase of the flatband voltage (V fb ) follows the p × t behavior, and the increase of both oxidation parameters induced a higher effective charge concentration, although not in the same proportion, with oxygen pressure being more important than oxidation time in inducing the highest effective charge concentration.Nevertheless, this is considered an important finding, since the presence of effective negative fixed charge plays a major role in the effective mobility of SiC based MOSFETs. 33These results reveal that a higher oxygen pressure also induces a larger electrical degradation, similar to the well-known effect of longer oxidation times. 34,35 hus, the electrical degradation seems to be controlled by the p × t parameter, i.e., an oxidation parameter that accelerates the SiO 2 film growth should lead to a larger electrical degradation.Therefore, alternative ways to obtain SiO 2 films on SiC such as thermal growing of a very thin and stoichiometric SiO 2 film  in a minimal oxidation condition followed by the SiO 2 film deposition, 36 or the oxidation of a Si/SiC heterojunction produced by a layer-transfer process, 37 or even the direct deposition of the SiO 2 film on the SiC substrate, reducing the influence of the substrate in the thermal oxide formation 38 should be investigated in order to minimize the formation of electrical active defects in the SiO 2 /SiC structure.
Results presented in this work exclude the thickness of the SiO 2 /4H-SiC interfacial region and the amount of residual compounds present on the SiC surface as the main cause of the electrical degradation originated from SiC thermal oxidation.Concerning the presence of effective negative fixed charge observed in our electrical measurements, Ebihara and co-workers 39 recently attributed the presence of negative fixed charge to be from CO 3 -like moiety formed from the interaction of the SiO 2 film with residual carbon atoms.Therefore, a possible explanation for the origin of the electrical degradation caused by SiC thermal oxidation is the interaction of the oxidation by-products with the SiO 2 bulk. 10By-products formed in this work would contain 18 O and can be incorporated in the solid phase in all depth regions of the film.When submitting the film to HF etching, the only compounds that remain in the sample are those insoluble in the previous near-interface region.Thus, the fact that no modification in the amount of residual compounds was observed after the removal of the oxide film in HF corroborates this hypothesis.
In summary, this work presented the influence in the electrical and structural properties of low oxygen pressures and different times of thermal oxidation of 4H-SiC on both Si and C faces.Although the product oxygen pressure with oxidation time controlled the total amount of oxygen incorporated in both faces, it does not affect the amount of residual compounds after etching the sample in HF or the SiO 2 /4H-SiC interfacial region thickness.On the other hand, the V fb was influenced by those parameters inducing higher negative effective charge concentrations, indicating that increasing oxygen pressure during thermal oxidation can induce an electrical degradation in a similar way to oxidation time.The possibility that the origin of this electrical degradation is in the interaction of SiC oxidation by-products with the SiO 2 bulk should not be disregarded.

FIG. 1 .
FIG. 1. (Top)18 O amounts obtained by NRA from samples submitted to different oxidation times (0.5, 1, 2, 3, and 4 h) and oxygen pressures (50, 100, and 200 mbar) at 1100 • C on 4H-SiC on both Si and C faces.Film thicknesses were calculated assuming that the SiO 2 density on SiC is 2.21 g cm −3 .(Bottom) Amount of residual18 O after etching in aqueous HF of samples oxidized at 1100 • C and 100 mbar of 18 O 2 under different oxidation time conditions.Error bars correspond to experimental accuracy of 5%.

FIG. 2 .
FIG. 2. Experimental (symbols) excitation curves of the 18 O(p,α) 15 N nuclear reaction around the resonance at 151 keV and the corresponding simulations (lines) for 4H-SiC samples oxidized for 1 h in 50 (triangle and dashed line), 100 (circle and solid line), and 200 mbar (square and dotted line) of 18 O 2 , and for 3 h at 100 mbar (inverted triangle, dotted/dashed line) at 1100 • C. (a) For the Si-face and (b) for C-face samples.(c) Excitation curves and the corresponding simulations for the same nuclear reaction for a 4H-SiC (C-face) sample oxidized for 10 h in 100 mbar of 18 O 2 at 1100 • C followed by aqueous etching in HF for 290 s (diamond and short dotted line) and of a 4H-SiC (C-face) sample oxidized for 1 h in 100 mbar of 18 O 2 at 1100 • C (circle and solid line).Insets: 18 O profiles obtained from the simulation of excitation curves using the same line types.

FIG. 3 .
FIG. 3. (a) I-V curves and (b) C-V curves (open symbols are experimental data and line is the ideal curve.The arrow indicates where V fb was extracted for the sample oxidized for 4 h.For the other samples, it is not shown since it is alike) of Al/SiO 2 /4H-SiC structures.Oxygen pressure and oxidation time are indicated.Oxidation temperature was 1100 • C for all samples.

TABLE I .
Electrical parameters obtained from C-V and I-V measurements for Al/SiO 2 /4H-SiC Si face, n-type structures.