Interfacial band parameters of ultrathin ALD-ZrO₂ on Ga-polar GaN through XPS measurements

With high breakdown electric field and high electron saturation velocity, gallium nitride (GaN) has been a promising competent semiconductor material for high-performance radio-frequency electronic devices.^1–4 Researchers have exhibited well-developed unipolar GaN devices since its first demonstration decades ago.⁵ It revealed that GaN has excellent potential for both radio-frequency and power electronic applications. However, given the low hole mobility and difficulties in both making good p-type ohmic contact and achieving high p-type doping concentration, the possibility of implementing practical GaN-based bipolar devices is restricted.⁶ Other than improving the p-type doping technique in GaN, another approach to figure out the above-mentioned problem is to replace p-type GaN materials with other available semiconductor materials to form GaN-based bipolar transistors. A recent study demonstrated the grafting approach to integrate high-quality p-type non-nitride single crystalline semiconductors with n-type GaN to form epitaxy-like p-n junctions, with an ultrathin dielectric (e.g., Al₂O₃) layer deposited by atomic layer deposition (ALD) as the interfacial layer.^7–10 The ultrathin interfacial ALD layer in these heterostructures serves as a quantum tunneling layer and passivates both the surfaces of GaN and attached p-type materials. In this sense, understanding the interfacial band parameters between ultrathin ALD dielectrics and Ga-polar GaN is critical in regard to the design and optimization of grafted GaN-based heterostructures.

Recently, the interfacial band parameters between ultrathin ALD-Al₂O₃, ALD-HfO₂, or plasma-enhanced ALD-AIN and n-type Ga-polar GaN have been studied, showing that ultrathin ALD layers could adapt to the band-bending structure significantly when the thickness of the material is below 2 nm.^11,12 On the other hand, with excellent thermal, mechanical, and electrical properties, zirconium oxide (ZrO₂) is considered a strong candidate in many applications, including ceramics and optical devices.^13,14 Several studies have shown that high-k ZrO₂ introduces a small density of states and small leakage current for GaN high-electron-mobility transistor devices when serving as the gate dielectric,^15–17 indicating the outstanding passivation capability of ZrO₂. This makes ZrO₂ a potential candidate for the interfacial layer in high performance GaN-based grafted devices. ZrO₂ thin films can be prepared by traditional physical vapor deposition and chemical vapor deposition, which are unable to produce precise thicknesses of the deposited film. Instead, ALD can provide excellent uniformity and accurate control of the thickness of the deposited layer at low temperatures.¹⁸ A few studies on the interfacial band parameters of GaN with ALD-ZrO₂ were reported in the past,^19,20 but they lacked the variation in the dielectric thickness, especially in the range below 2 nm, which is of great interest to the application of ZrO₂ as the interfacial layer in grafted devices. Therefore, a detailed study of ultrathin ALD-ZrO₂ with well-selected deposition conditions and thickness variation is necessary for the benefit of further development of grafted heterostructure devices.

In this work, we studied the influence of ultrathin ALD-ZrO₂ on the band-bending of the Ga-polar GaN surface through x-ray photoelectron spectroscopy (XPS). ALD-ZrO₂ showed the largest suppression on the GaN upward band-bending from 0.88 to 0.48 eV with five-cycle ALD, and the effect of suppression decreases as the ZrO₂ thickness increases. In addition, the band-bending of bare GaN has a minor difference after the removal of the ultrathin oxide layer, indicating that the suppression effect is from the ultrathin ALD-ZrO₂ layer, and the surface state can be recovered. Furthermore, the bandgap of ultrathin ALD-ZrO₂ of various thicknesses was derived from the O 1s spectrum from XPS results. The overall bandgap values are consistent with other published results, ranging from 5.1 to 5.8 eV.^21–23 Finally, the valence band offset (VBO) and conduction band offset (CBO) were derived. The possible mechanism behind the change in band-bending is discussed.

The n-type GaN used in this work was diced from the same epi-wafer used in our previous publications, and the detailed growth conditions and characteristics were discussed.^11,12 To clarify the structure of the sample, first, a 2.5-µm thick GaN buffer layer was grown on the two-inch wafer of the c-plane sapphire substrate. 1.5-µm GaN with a Si doping concentration of 1.5 × 10¹⁹ cm⁻³ and 0.25-µm Si-doped GaN with a doping concentration of 1 × 10¹⁸ cm⁻³ were grown on the substrate.

The Fermi level (E_F) was calculated to be 0.005 eV below the conduction band minimum (E_CBM) based on the doping concentration of the top Si-doped GaN layer.¹¹ The same epitaxial n-GaN sample was used for all the experiments in this study. The cleaning process followed the same procedure: the sample was first cleaned ultrasonically in acetone, IPA, and DI water each for 10 min followed by 5 min of piranha cleaning (96% H₂SO₄:30% H₂O₂:H₂O = 1:1:8), 10 min of diluted hydrochloric acid etching (0.1 normal HCl:H₂O = 1:1), and 5 min of diluted hydrofluoric acid etching (49% HF:H₂O = 1:1) to remove the native oxide and surface oxide that disturbs the XPS results. The epi sample was loaded into the ALD chamber (Savannah thermal ALD system) directly after cleaning to avoid contact, and the same procedure when it was transferred for XPS analysis was followed. Tetrakis(dimethylamino)zirconium (TDMA-Zr) and H₂O were used as precursors since TDMA-Zr can produce a deposited layer with high volatility and good thermal stability.²⁴ During the ALD deposition, the sample holding stage was heated to 250 °C, and five cycles of TDMA-Zr precursor flow were performed to remove the in situ oxide. ZrO₂ was then deposited using TDMA-Zr and H₂O vapor as precursors at temperatures less than 250 °C with different cycles for different recipes. The deposition temperature is decided within the temperature limit of the ALD equipment at 250 °C since the thin ALD ZrO₂ film deposited at this temperature has a more uniform structure than those at lower temperatures.²⁴ The expected ALD-ZrO₂ growth per cycle (GPC) is around 0.1 nm/cycle based on the calibration from the manufacturer of the ALD equipment.

To prevent the possible oxidation from air, the sample was loaded in a N₂ bag after cleaning and transferred directly to XPS analysis. XPS characterizations were performed ah the Thermo Scientific K Alpha XPS with an Al K_α x-ray source at 1486.6 eV. The following settings were applied: a spot size of 400 µm, pass energy of 10 eV, dwell time of 40 s, and step size of 0.02 eV. The Au 4 f_7/2 peak was used to calibrate the XPS spectrometer,²⁵ and the C 1s peak at 284.6 eV was used for charge correction.^26,27 The total Gaussian resolution is around 0.20 eV, which includes the uncertainty from Gaussian phonons and instrumental uncertainty from the spectrometer.^11,12 Clear Zr 3d_5/2 and 3d_3/2 peaks were fitted from the measured XPS spectra of the 5-to-20-cycle ALD-ZrO₂ coated GaN with an average spin–orbit splitting of 2.39 eV (Fig. S1).

The surface roughness of the sample coated with 0, 5, or 20 cycles of ALD-ZrO₂ was measured with a Bruker Dimension Icon AFM system (Fig. S2). No significant change in the surface roughness was observed after the deposition of ultrathin ALD-ZrO₂.

It is noted that we used the same secondary-ion mass spectroscopy profile as our prior work¹¹ to calculate the Fermi energy level in this study as the sample used in the present study and that used in Refs. 11 and 12 are from the same GaN-on-sapphire epi wafer. However, we did notice that the bare sample in this study showed a band-bending of 0.88 eV, which is significantly higher than that of the sample (also bare) used in the prior work.¹¹ To ensure the reliability of our XPS measurements, we have re-measured the sample used in this study after finishing the entire measurement sequence (0.89 eV) and again after more than half a year (0.68 eV). It is concluded that despite the potential measurement variation that can be caused by instrument conditions, there is some non-uniformity across the epi wafer. Since all measurements were performed on only one specific sample in each study, the band-bending results shown in each of the studies, particularly the band-bending trend as a function of the ALD coating conditions, are considered consistent and meaningful.

The $EG,ZrO2$ values were derived from the binding energy differences between the O 1s core level peak and the energy loss onset of the asymmetric O 1s XPS spectra (Fig. 4).³⁰ The inelastic collisions between the energetic electrons and other electrons induce energy loss of the electrons before they are detected by the XPS detector. If there is a significant number of electrons losing energy due to inelastic collisions, there will be an additional energy loss spectrum at higher binding energy next to the corresponding photon-excited electron peak (e.g., O 1s) due to the reduced kinetic energy of these electrons. For wide bandgap semiconductor materials such as ZrO₂, the lower limit of the kinetic energy loss of these electrons is equal to the bandgap energy,³⁰ indicating that the bandgap energy can be extracted by the onset of electron energy loss. To determine the value of the ZrO₂ bandgap and its uncertainty, a linear fitting was made with 95% confidence intervals at the measured O 1s spectrum near the onset of the inelastic losses. The background level was determined by the Shirley background fitting. The binding energy of the onset of electron energy loss was then determined by the intersection between the linear fitting and the background level. Correspondingly, the uncertainty of the ZrO₂ bandgap can be determined by the upper and lower intersections.^29,30 The detailed extraction of the ZrO₂ bandgap value and its uncertainty using the onset of electron energy loss of O 1s spectra are illustrated in Fig. 4, along with the plot of the bandgap values as a function of the number of ALD cycles. As shown in Fig. 4(e), the bandgap values of ALD-ZrO₂ are 5.18 ± 0.26, 5.09 ± 0.21, and 5.19 ± 0.28 eV for ALD cycles of 5, 10, and 15, respectively. For the 20 cycle ALD-ZrO₂ coating, a higher bandgap value of 5.8 ± 0.22 eV was obtained. Such a change in the bandgap caused by the change in the ALD material thickness of below 2 nm has also been reported previously.^12,31 The measured bandgap values are comparable to the reported values of 5.2 and 5.8 eV measured by ultraviolet/visible/near-infrared spectroscopy,³² 5.3 eV from electron energy-loss spectroscopy,³³ 5.1 eV²⁹ and 5.6 eV³⁴ from XPS, and 5.5 eV from ultraviolet photoelectron spectroscopy and inverse photoelectron spectroscopy.³⁵ 

With the value of $EG,ZrO2$ extracted, the CBO [Fig. 5(a)] and VBO [Fig. 5(b)] can be calculated based on Eqs. (3) and (4), respectively. The measured values of qϕ_S,GaN, the surface band-bending energy values of Ga-polar GaN qψ_S, the bandgap values of the ALD-ZrO₂ $EG,ZrO2$⁠, and the extracted VBO and CBO values along with their uncertainties are listed in Table I.

As shown in Table I, the CBO and the suppression effect on upward band-bending decrease simultaneously when the number of ALD cycles is less than 20. From the analysis in previous research,^11,12,36,37 the electron states whose energy ranges between the conduction band minimum of ALD oxide and its counterpart in GaN show positive charge, which means that a larger CBO introduces more positive charge at the interface to suppress the upward surface band-bending of Ga-polar GaN. However, when the number of deposition cycles reaches 20, it shows a larger CBO and larger upward band-bending than the other results, indicating a different mechanism becoming dominant. This can be related to the external negative charge introduced by the border traps in ALD-ZrO₂.³⁸ Therefore, the change in band-bending vs the number of ALD cycles is a mixed effect from the two above-mentioned mechanisms.

We characterized the interfacial band parameters of ultrathin ALD-ZrO₂ on Ga-polar GaN using XPS measurements. The surface band-bending of Ga-polar GaN decreased from 0.88 eV of bare GaN to 0.48 eV with five cycles of ALD deposition of ZrO₂, then gradually increased to 0.60 eV at 15 cycles of deposition, and hit the reverse point somewhere between 15 and 20 cycles with a band-bending of 0.90 eV, surpassing the value of bare GaN. A combinational effect of the charge compensation from the electron states between the two conduction band minima and the border traps in ALD-ZrO₂ is speculated. ALD-ZrO₂ provided good recovery ability to the surface band-bending of bare GaN with negligible changes in band-bending of 0.006 eV after removal of oxide compared to the origin value. The study can be a useful guide to the future development of GaN-based devices with ultrathin ALD-ZrO₂.

See the supplementary material for the Zr 3d XPS spectra and AFM images.

S. Qiu was supported by the China Scholarship Council. J. Gong, J. Zhou, R. Singh, M. Sheikhi, and Z. Ma were supported by the Air Force Office of Scientific Research (AFOSR), under Grant No. FA9550-21-1-0081. B.S.O. and T.K.N. gratefully acknowledge the KAUST funding: BAS/1/1614-01-01 and REP/1/5313-01-01.

The authors have no conflicts to disclose.

S.Q. and J.G. contributed equally to this work.

Shuoyang Qiu: Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – original draft (lead). Jiarui Gong: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – review & editing (lead). Jie Zhou: Methodology (supporting). Tien Khee Ng: Methodology (supporting). Ranveer Singh: Data curation (supporting); Methodology (supporting). Moheb Sheikhi: Methodology (supporting). Boon S. Ooi: Funding acquisition (supporting); Project administration (supporting); Resources (supporting); Supervision (supporting). Zhenqiang Ma: Conceptualization (equal); Funding acquisition (lead); Project administration (lead); Resources (lead); Supervision (lead); Validation (lead); Writing – review & editing (supporting).

The data that support the findings of this study are available from the corresponding authors upon reasonable request.