Investigation of dominance in near-surface region on electrical properties of AlGaN/GaN heterostructures using TLM, XPS, and PEC etching techniques

To investigate how the electrical properties of AlGaN/GaN HEMTs are dominated by the near-surface region, transmission line method and X-ray photoelectron spectroscopy (XPS) measurements were conducted on three samples. There was one sample with poor ohmic properties. The XPS results indicate that the surface-Fermi-level, E FS, position of the poor-ohmic sample was deeper than the others. When a 5 nm thick surface layer was removed by contactless photo-electrochemical etching, E FS shifted to the same position as the others and the contact resistance improved. These results indicate that the control of the near-surface region of AlGaN can be a dominant factor changing the E FS position, which greatly affects the ohmic properties.


H
igh-electron-mobility transistors (HEMTs) based on gallium nitride (GaN) have made remarkable progress in high-power and high-frequency performance due to the fact that GaN has attractive features such as wide band gap, high breakdown field, 1) and high saturation electron velocity. 2)Furthermore, GaN can also be heterojunctioned with other Ⅲ-nitride alloys, such as AlGaN, InAlN, 3) and AlInGaN, 4) for forming heterointerfaces to make two-dimensional electron-gas (2DEG) densities higher than 1 × 10 13 cm −2 , due to the spontaneous and piezoelectric polarization field as well as large conduction band offset. 5)A record-high cutoff frequency of 455 GHz 6) and output power densities of more than 3 W mm −1 above 90 GHz [7][8][9] have been reported for GaN-based HEMTs. Lw specific onresistance (<4.5 mΩcm 2 ) can also be achieved at a breakdown voltage up to 2 kV 10) due to the material advantage of nitride-semiconductors.However, many challenges related to surface/interface states remain.In compound semiconductors, Fermi-level pinning is a longstanding issue that affects device performance and reliability, and GaN-based materials are no exception. R2) We investigated the correlation between the condition of E FS pinning and electrical properties for AlGaN/GaN HEMTs using X-ray photoelectron spectroscopy (XPS) and the transmission line method (TLM).We also conducted photo-electrochemical (PEC) etching on samples with abnormal pinning position and poor electrical characteristics and investigated the effect by comparing before and after the PEC etching process.
A cross-section of an AlGaN/GaN HEMT structure is schematically shown in Fig. 1.We used three types of HEMT structures with epitaxial layers grown by metalorganic vapor phase epitaxy on semi-insulating SiC or GaN substrates.Each epitaxial structure consisted of a 300 nm Fe-doped GaN layer as a buffer layer, 900 nm undoped-GaN channel layer and AlGaN barrier layer, the Al composition and thickness of which are summarized in Table I.The 2DEG density and mobility of each sample are also summarized in Table I.For the analysis of the electrical properties of each sample (A, B and C), we conducted TLM measurements.First, mesa structures with line widths of 250 μm were formed by inductively-coupled-plasma reactive ion etching using Cl 2 /BCl 3 plasma to electrically isolate the TLM structures.Next, TLM electrode patterns with spacing of 5, 10, 20, 30, 40, and 50 μm were formed using photolithography.After removing any surface oxide layer with HF treatment, Ti/Al/Ti/Au (=20/50/20/50 nm) stacked films were deposited on the AlGaN surface by a vacuumevaporation process such as electron beam evaporation for Ti and Au and resistance heating evaporation for Al.After the lift-off formation of TLM electrodes, a 20 nm thick SiN film was deposited to prevent damage to the AlGaN surface during the annealing process. 13,14)The samples were then annealed at 830 °C for 1 min in N 2 ambient using a rapid thermal annealing furnace to obtain ohmic conduction at each electrode/AlGaN interface.After the ohmic metallization process, the SiN film was removed using a buffered HF solution.Current-voltage (I-V ) measurements were conducted at room temperature using a Keysight B1500A semiconductor-device-parameter analyzer.For the surface analysis on each sample, XPS measurements were conducted at a takeoff angle of 45°using a monochromated Al-Kα X-ray source (1468.6 eV).The binding energy was calibrated by setting Au 4 f 7/2 at 84.0 eV prior to each measurement. 15)The shift in the binding energy due to charging was corrected by setting the C-C bond component peak of C 1s core-level spectra to 284.8 eV. 16)igure 2 shows the total resistances as a function of the gap spacing of TLM patterns for Samples A and C. The inset shows typical I-V characteristics obtained on TLM patterns at the gap spacing of 10, 30, and 50 μm.As can be seen, each curve has a linear characteristic, and the slope decreases as the gap spacing increases.This indicates that ohmic contact was indeed obtained in both samples.The solid lines show the data for least-squares fitting.The sheet resistance was approximately 520 Ω sq −1 for both devices, which is close to that predicted from the 2DEG density and mobility.However, the contact resistance (R C ) was different, i.e. 2.7 and 7.8 Ωmm for Samples A and C, respectively.The R C of Sample B (not shown) was 2.1 Ωmm, which was almost same as that obtained for Sample A. Since the sheet resistance is reasonable, the anomalously large R C observed only in Sample C can be attributed to the top surface of the AlGaN layer, not to the 2DEG channel region.
To clarify what differences exist in AlGaN surface, XPS analysis was conducted on each sample.Figures 3(a)-3(c) show Ga 3d, Al 2p, and N 1s core-level spectra, respectively, and the data obtained from Samples A, B, and C are compared.The intensity was normalized to the maximum peak intensity among samples.In Samples A and B, the spectra were consistent for all core levels, but for Sample C, the peak of the core-level spectra was observed to shift toward the lower binding energy by 0.3 eV.Using Sample A as an example, let us derive the location of E FS from the Ga 3d and valence band spectra obtained from XPS measurements.The E FS position relative to the valence band maximum (VBM), E V , can be derived as 17,18) ( ) ( ) () where E Ga3d is the Ga 3d core-level energy, and B Ga3d and B VBM are the binding energies of Ga 3d and VBM, respectively.Since (B Ga3d -B VBM ) is known as a materialspecific constant, (E FS -E V ) can be determined from measuring B Ga3d .In the example of Sample A shown in Fig. 4, B Ga3d was 19.96 eV and (B Ga3d -B VBM ), which is constant for all samples, was 17.06 eV.Thus, the B VBM of Sample A with respect to the Fermi level was estimated to be 2.9 eV, which was also obtained from extrapolation of the lower edge of the binding energy of the VBM spectra.When the band gap of AlGaN was assumed to be 3.8 eV, 5,19) E FS was found to be at E C −0.9 eV.As mentioned above, the location of E FS can be determined only from B Ga3d , so the E FS of Samples B and C were determined to be E C −0.9 eV and E C −1.2 eV, respectively, from the XPS measurements of B Ga3d in Fig. 3.We found that E FS of Sample C was pinned at 0.3 eV deeper energy in the band gap compared with those of Samples A and B, resulting in a larger potential barrier formed at the interface between the ohmic metal and AlGaN layer.These XPS analysis results are very consistent with the TLM results showing that R C was anomalously large only in Sample C.
To remove the Fermi-level-pinning factor present near the topmost surface, we conducted PEC etching on Sample C. The PEC etching has the advantage of low-damage processing compared with dry-etching.][26] A 20 nm thick Ti film was formed on the edge of the sample surface by electron beam evaporation as a cathode pad for CL-PEC etching; the role of the cathode pad is described in previous studies. 26,27)CL-PEC etching was conducted by immersing the sample in 0.025 mol L −1 K 2 S 2 O 8 solution (pH = 2 ∼ 3) under irradiation of UV light.Figure 5(a) shows the capacitance-voltage (C-V ) characteristics of the Schottky diodes formed on Sample C with and without (w/o) CL-PEC etching.A Ni circular gate with a diameter of 200 μm was prepared on the AlGaN surface.The positive thresholdvoltage shift observed in the sample with CL-PEC etching was due to the decrease in 2DEG density because the AlGaN layer thickness became thinner.Figure 5(b) shows the carrier density distribution calculated from the C-V curve. 5)From the comparison of peak positions, the CL-PEC etching depth of the AlGaN layer was estimated to be 5 nm.The 2DEG density calculated by integrating the carrier profile for each sample was 6.0 × 10 12 cm −2 for the sample without CL-PEC and 5.0 × 10 12 cm −2 with CL-PEC etching.
We also conducted TLM and XPS measurements to investigate the effect of CL-PEC etching.Figure 6(a) shows the Ga 3d spectra obtained from Sample C with and without CL-PEC etching.With CL-PEC sample, we observed that B Ga3d shifted by 0.3 eV toward higher binding energy.This   Since the AlGaN barrier layer is very thin in HEMT structures, it is very difficult to evaluate crystalline quality such as defects and impurity concentration.We conducted secondary ion mass spectrometry and transmission electron microscopy as well as XPS to identify the reason the E FS -pinning position differs only in Sample C but were not able to determine it.However, as shown in Fig. 6, the ohmic properties and E FS position clearly changed before and after CL-PEC etching, indicating that the origins for the deeper E FS -pinning position in Sample C (w/o CL-PEC) located not inside the epitaxial layer but in the near-surface region within 5 nm depth.If the defects or dislocations that distributed inside the epitaxial layer are the main origin of E FS pinning, the 5 nm etching of the surface layer is not expected to change the E FS -pinning position.We believe that the origin of E FS pinning observed in this study is the surface-localized gap states (surface states) induced by the surface disorder layer including displacement of the atomic-bond angle. 28)he surface state density distribution consists of U-shaped state continuum based on the interruption of crystal    091002-3 © 2023 The Author(s).Published on behalf of The Japan Society of Applied Physics by IOP Publishing Ltd periodicity and discrete levels caused by surface defects and/ or specific atomic bonding. 29)Such surface states cause band bending at the surface to maintain charge neutrality, resulting in E FS being pinned in position to maintain the charge balance at the surface.When the state density is high, E FS is known to be strongly pinned near the charge neutrality level (E CNL ). 28,29)On the basis of this perspective, highdensity surface states should form in Sample C, causing Fermi-level pinning at an energy level different from those of the other samples.The electron mobility of Sample C before PEC etching (w/o CL-PEC) was high despite the pinning position deep in the AlGaN surface, indicating that the AlGaN/GaN heterointerface did not electrically degrade.In Sample C, the crystalline strain may have been concentrated at the surface rather than at the heterointerface by some mechanism, causing disorder at the AlGaN surface.It remains unclear how the differences in surface conditions exist for Sample C compared to the other samples.Since the nature of the surface state is very sensitive to the growth termination process, one possibility is that the crystal growth conditions, such as growth recipe and post-growth processing, might be insufficient in sample C. The CL-PEC etching process reduced the surface disorder layer remaining after crystal growth, reducing the gap-state density at the AlGaN surface, which may have changed E FS .
In summary, we conducted TLM measurements and XPS analysis on three different AlGaN/GaN HEMT structure samples to investigate the correlation between the electrical and surface properties.All samples showed very high 2DEG mobility of about 2000 cm 2 V -1 s −1 , indicating the high crystal quality at the AlGaN/GaN heterointerface.However, there was one sample with poor ohmic characteristics (Sample C).XPS measurements showed that the surface Fermi level of Sample C was 0.3 eV deeper than that of the other samples, resulting in a higher potential barrier formed at the AlGaN surface.When the 5 nm thick surface layer was removed from Sample C by CL-PEC etching, the E FS position was shifted to the same position as those of the other samples, and the contact resistance also improved to the same level as others.These results indicate that the difference in the conditions of the topmost AlGaN surface can be a dominant factor that changes the E FS -pinning position and greatly affects the ohmic properties.

Fig. 2 . 2 ©
Fig. 2. Total resistance versus gap spacing plot for TLM measurements of Samples A and C. Solid line indicates data for least-squares fitting.The inset shows I-V characteristics at gap spacing of 10, 30, and 50 μm.

Fig. 4 .
Fig. 4. Ga 3d and VBM spectra measured for Sample A. Binding energy was calibrated by setting the C-C component of C 1s at 284.8 eV.

Fig. 5 .
Fig. 5. (a) C-V characteristics of AlGaN/GaN Schottky diode with and without PEC etching.(b) Distributions of carrier density estimated from C-V profiles.

Fig. 6 . 4 ©
Fig. 6.(a) Ga 3d XPS core-level spectra of AlGaN surface with and without PEC etching.(b) Comparison of R C among all samples.

Table I .
Nominal characteristics of samples.