Increasing concerns regarding the escalating demand of energy consumption throughout the world has triggered the needs of developing energy-efficient high-power and high-temperature metal-oxide-semiconductor (MOS)-based devices. It has been projected that gallium nitride (GaN) has the potential of conforming to the needs of these MOS-based devices due to its promising properties, which include wide bandgap (3.4 eV), large critical electric field (3 MV/cm), high electron mobility, as well as good thermal conductivity and stability 1 2 3 4 5 6 . The fabrication of a functional GaN-based MOS device requires a high-quality gate oxide that is capable of resisting a high transverse electric field 7 8 . Native oxide (Ga₂O₃) of GaN 9 10 11 12 13 and a relatively low-dielectric-constant (k) SiN

x

Prior to the deposition of 60-nm thick Y₂O₃ films on the commercially purchased Si-doped (n-type) GaN epitaxial layers with thickness of 7 μm and doping concentration of 1 to 9 × 10¹⁸ cm⁻³ grown on sapphire substrates, the wafer, which was diced into smaller pieces, were subjected to RCA cleaning. Subsequently, these samples were loaded into a vacuum chamber of RF magnetron sputtering system (Edwards A500, Edwards, Sanborn, NY, USA). A comprehensive description on the deposition process of Y₂O₃ films has been reported elsewhere 29 30 . Then, PDA was performed in a horizontal tube furnace at 400°C in different ambients (O₂, Ar, N₂, and FG (95%N₂ + 5% H₂)) for 30 min. The heating and cooling rate of approximately 10°C/min was used for the PDA process. After the PDA process, X-ray photoelectron spectroscopy (XPS) measurements were conducted on the samples at the Research Center for Surface and Materials Science, Auckland University, New Zealand, using Kratos Axis Ultra DLD (Shimadzu, Kyoto, Japan) equipped with a monochromatic Al-K_α X-ray source (hv = 1486.69 eV). The spectra of the survey scan were obtained at a low pass energy of 160 eV with an energy resolution of 0.1 eV, and the photoelectron take-off angle was fixed at 0° with respect to the surface normal. Chemical depth profiling was performed by etching the samples using an Ar ion gun operated at 5 keV in order to identify the boundary of Y₂O₃ and interfacial layer between the oxide and GaN. To further determine the bandgap of Y₂O₃ and IL, a detailed scan of O 1s was first performed at the same pass energy of 20 eV with an energy resolution of 1.0 eV. The energy loss spectrum of O 1s would provide the bandgap of Y₂O₃ and IL by taking into consideration the onset of a single particle excitation and band-to-band transition. Kraut’s method was utilized in the extraction of the valence band offset of Y₂O₃ and IL 34 35 . In order to fabricate MOS test structure, the Y₂O₃ film was selectively etched using HF/H₂O (1:1) solution. Next, a blanket of aluminum was evaporated on the Y₂O₃ film using a thermal evaporator (AUTO 306, Edwards). Lastly, an array of Al gate electrode (area = 2.5 × 10⁻³ cm²) was defined using photolithography process. Figure 1 shows the fabricated Al/Y₂O₃/GaN-based MOS test structure. The current–voltage characteristics of the samples were measured using a computer-controlled semiconductor parameter analyzer (Agilent 4156C, Agilent Technologies, Santa Clara, CA, USA).

Bandgap (E _g) values for Y₂O₃ and IL are extracted from the onset of the respective energy loss spectrum of O 1s core level peaks. The determination of E _g values for Y₂O₃ and IL is done using a linear extrapolation method, wherein the segment of maximum negative slope is extrapolated to the background level 36 . Figure 2a shows typical O 1s energy loss spectra of Y₂O₃ and IL for the sample annealed in O₂ ambient. The extracted E _g values are in the range of 4.07 to 4.97 eV and 1.17 to 3.93 eV with a tolerance of 0.05 eV for Y₂O₃ and IL, respectively, for samples annealed in different post-deposition annealing ambients (Figure 3a).

Typical valence band photoelectron spectra of Y₂O₃ and IL for the sample annealed in O₂ ambient are presented in Figure 2b. By means of linear extrapolation method, the valence band edges (E _v) of Y₂O₃ and IL could be determined by extrapolating the maximum negative slope to the minimum horizontal baseline 36 . The acquired valence band offset (ΔE _v) values of Y₂O₃ and IL with respect to GaN substrate are in the range of −0.04 to −1.43 eV and −0.21 to −3.23 eV with a tolerance of 0.05 eV, respectively, for all of the investigated samples. The ΔE _v values of Y₂O₃/GaN and IL/GaN are shown in Figure 3b as a function of PDA ambient.

The determination of both E _g of Y₂O₃ and IL as well as ΔE _v of Y₂O₃/GaN and IL/GaN enables the calculation of the conduction band offset (ΔE _c) of Y₂O₃/GaN, IL/GaN, and Y₂O₃/IL using the following equation: ΔE _c(oxide or IL) = E _g(oxide or IL) − ΔE _v(oxide/GaN or IL/GaN) − E _g(GaN), where E _g(GaN) is 3.40 eV for GaN 37 . The obtained values of ΔE _c(Y₂O₃/GaN), ΔE _c(IL/GaN), and ΔE _c(Y₂O₃/IL) for all of the investigated samples are presented in Figure 4. In general, a reduction in E _g(Y₂O₃), E _g(IL), ΔE _c(Y₂O₃/GaN), and ΔE _c(IL/GaN) is observed when different PDA ambients are performed, as indicated by O₂ > Ar > FG > N₂. The IL has been proven using XPS to be comprised of a mixture of Ga-O, Ga-O-N, Y-O, and Y-N bonding (HJQ and KYC, unpublished work). The detection of Ga-O and Ga-O-N bonding in the region of IL indicates that the oxygen dissociated from Y₂O₃ during PDA in different ambients would diffuse inward to react with the decomposed GaN substrate. During PDA in O₂ ambient, an additional source of oxygen from the gas ambient has contributed to the formation of Ga-O and Ga-O-N bonding in the region of IL. Sample subjected to PDA in O₂ ambient attains the largest E _g(Y₂O₃) and E _g(IL) as well as the highest values of ΔE _c(Y₂O₃/GaN) and ΔE _c(IL/GaN). This is related to the supply of O₂ from the gas ambient during PDA, which has contributed to the reduction of oxygen-related defects in the Y₂O₃ film and the improvement in the compositional homogeneity of the oxide film. The absence of O₂ supply during PDA in Ar (inert) and reducing ambient, such as FG and N₂, may be the reason contributing to the attainment of lower E _g(Y₂O₃), E _g(IL), ΔE _c(Y₂O₃/GaN), and ΔE _c(IL/GaN) values than the sample annealed in O₂. The presence of N₂ in both FG and N₂ ambient has caused the formation of O₂-deficient Y₂O₃ film, wherein N atoms dissociated from N₂ gas may couple with the oxygen-related defects in the Y₂O₃ film 30 38 . In addition, the presence of N₂ in both FG and N₂ ambient is also capable of performing nitridation process to diminish the tendency of O₂ dissociated from the Y₂O₃ film during PDA to diffuse inward and react with the GaN substrate 30 . Thus, the interfacial layer formed in between the Y₂O₃/GaN structure for these samples could be O₂ deficient. Despite the fact that FG and N₂ ambient are capable of providing nitridation and coupling process, the percentage of N₂ in FG ambient (95% N₂) is lower than that in pure N₂. Hence, PDA in N₂ ambient will enhance the nitridation process and coupling of N atoms with the oxygen-related defects in Y₂O₃, which leads to the formation of more O₂-deficient Y₂O₃ film and IL when compared with the sample annealed in FG ambient. This could be the reason leading to the attainment of the lowest E _g(Y₂O₃), E _g(IL), ΔE _c(Y₂O₃/GaN), and ΔE _c(IL/GaN) values for the sample annealed in N₂ ambient.

A schematic illustration of the energy band alignment of the Y₂O₃/IL/GaN structure that had been subjected to different PDA ambients is presented in Figure 5. Three different energy band alignment structures were obtained due to the effect of PDA ambient. It is noticed that the conduction band edge of IL is higher than that of Y₂O₃ for the sample annealed in O₂ ambient, but it is lower in samples annealed in Ar, FG, and N₂ ambient. This band alignment shift would influence the leakage current density-electrical field (J-E) characteristics of the samples (Figure 6). The dielectric breakdown field (E _B) is defined as the electric field that causes a leakage current density of 10⁻⁶ A/cm², which is not related to a permanent oxide breakdown but representing a safe value for device operation 39 . Of all the investigated samples, the sample annealed in O₂ ambient demonstrates the lowest J and the highest E _B (approximately 6.6 MV/cm) at J of 10⁻⁶ A/cm². This might be attributed to the attainment of the largest E _g(Y₂O₃) and E _g(IL) as well as the highest values of ΔE _c(Y₂O₃/GaN) and ΔE _c(IL/GaN), while for other samples, a deterioration in J and E _B is perceived. The reduction is ranked as Ar > FG > N₂.

In order to determine whether the E _B of the investigated samples is either dominated by the breakdown of IL, Y₂O₃, or a combination of both Y₂O₃ and IL, Fowler-Nordheim (FN) tunneling model is employed to the extract barrier height (Φ_B) of Y₂O₃ on GaN. FN tunneling mechanism is defined as tunneling of the injected charged carrier into the conduction band of the Y₂O₃ gate oxide via passing through a triangular energy barrier 7 8 30 . This mechanism can be expressed as J _FN = AE ²exp(−B/E), where A = q ³ m _o/8(hmΦ_B, B = 4(2 m)^1/2 Φ_B ^3/2/(3qh/2), q is the electronic charge, m is the effective electron mass in the Y₂O₃ (m = 0.1m _o, where m _o is the free electron mass), and h is Planck’s constant 8 40 . In order to fit the obtained experimental data with the FN tunneling model, linear curve fitting method has been normally utilized 8 20 41 . Nevertheless, data transformation is needed in this method owing to the limited models that can be presented in linear forms. Hence, nonlinear curve fitting method is employed using Datafit version 9.0.59 to fit the acquired J-E results in this work with the FN tunneling model. It is believed that the extracted results using nonlinear curve fitting method is more accurate due to the utilization of actual data and the minimization of data transformation steps required in the linear curve fitting 42 43 . Figure 7 shows the J-E results for the samples annealed in O₂, Ar, and FG ambient, which fitted well with FN tunneling model. The extracted Φ_B values of these samples are presented in the Figure 4. The highest Φ_B value attained by the sample annealed in O₂ ambient (3.72 eV) was higher than that of metal-organic decomposed CeO₂ (1.13 eV) spin-coated on n-type GaN substrate 20 . No Φ_B value has been extracted for the sample annealed in N₂ ambient due to the low E _B and high J of this sample, wherein the gate oxide breaks down prior to the FN tunneling mechanism.

Table 1 compares the computed ΔE _c values from the XPS characterization with the Φ_B value extracted from the FN tunneling model. From this table, it is distinguished that the E _B of the sample annealed in O₂ ambient is dominated by the breakdown of IL as the obtained value of Φ_B from the FN tunneling model is comparable with the value of ΔE _c(IL/GaN) computed from the XPS measurement. For samples annealed in Ar and FG ambient, the acquisition of Φ_B value that is comparable to the ΔE _c(Y₂O₃/GaN) indicates that the E _B of these samples is actually dominated by the breakdown of bulk Y₂O₃. Since the leakage current of the sample annealed in N₂ ambient is not governed by FN tunneling mechanism, a conclusion in determining whether the E _B of this sample is dominated by the breakdown of IL, Y₂O₃, or a combination of both cannot be deduced. Based on the obtained values of ΔE _c(Y₂O₃/GaN), ΔE _c(IL/GaN), and ΔE _c(Y₂O₃/IL), the E _B of this sample is unlikely to be dominated by IL due to the acquisition of a negative ΔE _c(IL/GaN) value for this sample. Thus, the E _B of this sample is most plausible to be dominated by either Y₂O₃ or a combination of Y₂O₃ and IL. However, the attainment of ΔE _c(Y₂O₃/IL) value which is larger than that of ΔE _c(Y₂O₃/GaN) value obtained for the samples annealed in Ar and FG ambient eliminates the latter possibility. The reason behind it is if the E _B of the sample annealed in N₂ ambient is dominated by the combination of Y₂O₃ and IL, this sample should be able to sustain a higher E _B and a lower J than the samples annealed in Ar and FG ambient. Therefore, the E _B of the sample annealed in N₂ ambient is most likely dominated by the breakdown of bulk Y₂O₃.

^aNot influenced by FN tunneling. Therefore, barrier height is not extracted from the FN tunneling model.

In conclusion, three different energy band alignment models of Y₂O₃/interfacial layer/GaN structure subjected to post-deposition annealing at 400°C in different ambients (O₂, Ar, forming gas (95% N₂ + 5% H₂), and N₂) have been established using X-ray photoelectron spectroscopy. It was proven that the dielectric breakdown field (E _B) of the sample annealed in O₂ ambient was dominated by the breakdown of IL, while the E _B of the samples annealed in Ar, FG, and N₂ ambient was dominated by the breakdown of bulk Y₂O₃. The sample annealed in O₂ ambient demonstrated the best leakage current density-breakdown field due to the attainment of the largest bandgap, the largest conduction band offset, and the highest barrier height value.

The authors declare that they have no competing interests.

HJQ carried out all of the experimental work, data analysis of the obtained experimental results, and drafting of the manuscript. KYC had played a vital role in assisting HJQ in the experimental work and data analysis as well as in revising and approving the submission of the final manuscript for publication. Both authors read and approved the final manuscript.

HJQ received his MSc degree in 2010 from Universiti Sains Malaysia, Penang, Malaysia, where he is currently working on a PhD degree in Materials Engineering in the School of Materials and Mineral Resources Engineering. KYC received his PhD degree from the School of Microelectronic Engineering, Griffith University, Brisbane, Australia, in 2004. He is currently an associate professor with Universiti Sains Malaysia, Penang, Malaysia.