Electrical and Structural Analysis of β‐Ga2O3/GaN Wafer‐Bonded Heterojunctions with a ZnO Interlayer

Wafer bonding of β‐Ga2O3 and N‐polar GaN single crystal substrates is demonstrated by adding ZnO as a “glue” interlayer. The wafers are fully bonded such that Newton rings are not observed. Temperature‐dependent current‐voltage (I–V) measurements are conducted on the as‐bonded Ga2O3/ZnO/N‐polar GaN test structure and after annealing at 600 °C and 1100 °C. The impact of post‐annealing temperature on the electrical and structural characteristics of the bonded samples is investigated. A consistently ohmic‐like characteristic is obtained by annealing the bonded wafers at 1100 °C in N2, which is in part due to crystallization of ZnO and diffusion of Ga into ZnO which makes it n‐type doped. The wafer bonding of β‐Ga2O3 and GaN achieved in this work is promising to combine the material merits of both GaN and Ga2O3 targeting breakthrough high‐frequency and high‐power device performances.

Achieving -Ga 2 O 3 /GaN heterojunctions by epitaxial growth of high-quality -Ga 2 O 3 on GaN substrates or vice versa is challenging. This is because these two materials have different crystal structures and mismatch in lattice constants. Considering the requirements of large-scale transfer of high-quality epi-structure and low temperature device processing, another alternative approach is the wafer bonding of -Ga 2 O 3 and GaN substrates. Previously, we reported the direct bonding of -Ga 2 O 3 and Npolar GaN single crystal substrates. [25] We showed that the -Ga 2 O 3 /GaN surfaces were atomically bonded without any readily identifiable loss in crystalline quality at the interface. [25] However, the bonded area was only ≈40% and the yield was lower than 50% probably because both GaN and Ga 2 O 3 have high mechanical hardness and any nano-or micro-roughness on the surface would lead to non-uniform bonding. Our experiments showed that the bonding yield is also affected by cleavage planes in the Ga 2 O 3 substrate with (001) orientation. This issue is especially apparent when bonding small pieces, as GaN's square shape with sharp corners creates non-uniform pressure distribution with high pressure points, leading to cracks in Ga 2 O 3 substrate. However, we believe that this problem will be reduced when full wafers are used for bonding and (001) orientations could also be used. Additionally, the bonded interface was very sensitive to thermal treatment even though the ramping rate was as low as 2°C min −1 . This hinders the device fabrication as thermal cycling of the devices such as lithography with the pre-and post-baking of photoresist is routinely involved. Lastly, electrical characterization of the bonded interface was not conducted. In the merged materials we imagine, the bonded interface is within the active region of the device and a barrier for electron flow would be undesirable and needs to be reduced and ideally eliminated.
In this work, we first conducted temperature-dependent current-voltage (I-V-T) measurements on direct-bonded GaN-Ga 2 O 3 samples. Our characterizations revealed a larger barrier to electron flow at the bonded interface in these structures. We therefore implemented ZnO interlayer as a "glue" layer to further improve bonding yield and strength and study its impact on the electron barrier. We investigated the I-V-T characteristics of Ga 2 O 3 /ZnO/GaN test structures varying temperature from room temperature to 650 K. Additionally, the impact of high temperature annealing at 600°C and 1100°C in N 2 on the I-V characteristics was also studied. Scanning transmission electron microscopy (STEM) was utilized to analyze the bonded -Ga 2 O 3 /N-polar GaN interface.

Experimental Section
Commercially available 680 μm-thick UID (201) -Ga 2 O 3 substrates with a carrier concentration ≈2.3 × 10 17 cm −3 and N-polar on-axis 413 μm-thick (0001) GaN substrates with a carrier concentration ≈1 × 10 18 cm −3 were used for this work. Figure 1 shows the schematic process flow of the bonding of N-polar GaN and Ga 2 O 3 substrates with a ZnO interlayer. Both 5 × 5 mm 2 GaN and 10 × 10 mm 2 Ga 2 O 3 substrates were cleaned in acetone, isopropyl alcohol, and de-ionized water with ultra-sonication, followed by soaking in buffered hydrofluoric acid (BHF) for 30 s. The samples were then loaded immediately into the Atomic Layer Deposition (ALD) chamber to minimize any surface contamination. Ten nanometer thick ZnO was deposited on both N-polar GaN and (−201) Ga 2 O 3 samples via thermal-ALD at 200°C using Diethyl Zinc (DEZ) and water as the precursors. Next, the 5 × 5 mm 2 GaN sample was flipped on top of the 10 × 10 mm 2 Ga 2 O 3 sample and transferred to an EVG 510 bonding chamber. The bonding was conducted at 550°C under a pressure of 4 MPa in vacuum for 3 h. To investigate the impact of post-annealing temperature on the electrical quality of bonding interfaces, the N-polar GaN/ZnO/Ga 2 O 3 test structure was annealed in N 2 for 30 min at 600°C and 1100°C with ramping up and down rates of 2°C min −1 . After post-bonding annealing, Ti/Au (20/200 nm) was deposited on both the front and back sides of the sample. The cross sectional schematic of the test structure is shown in Figure 2.
The electrical properties of the bonded interfaces were investigated by measuring the I-V-T characteristics in vacuum using a Keysight B1500A Semiconductor Parameter Analyzer in conjunction with an MMR Technologies Variable Temperature Microprobe System. The temperature ramping rate was 2°C min −1 . The Ga 2 O 3 side was grounded, and the voltage bias was applied on the GaN side. High-resolution STEM-bright field (BF) imaging was performed on the N-polar GaN/ZnO/ -Ga 2 O 3 bonded interfaces. The specimens for STEM characterization were prepared by dicing followed by micro-sampling method using a Thermo-Fisher G4 650 Xe plasma dual-beam focused ion beam (FIB) instrument. A JEOL-JEM-3100R05 transmission electron microscope (TEM) equipped with double aberration correctors was used for STEM imaging. Figure 3 shows the optical image of Ga 2 O 3 (-201) bonded with on-axis N-polar GaN substrate with ZnO as a "glue" layer. In our previous work on the direct bonding of Ga 2 O 3 and N-polar GaN, only ≈40% contact area was bonded and Newton's rings were observed where the surfaces were not in contact with each other. [25] Here, the surfaces were fully bonded. No Newton's rings were observed on the sample indicating a flat and clean interface. The 100% bonding area in this work is attributed to the ZnO "glue" layer, by which the bonding uniformity can be improved by the conformal and smooth ZnO surfaces at the bonding interface of ZnO/Ga 2 O 3 and ZnO/GaN. The hardness of ALD-ZnO thin film was reported to be 5.6 GPa, [26] which is significantly lower than the hardness of GaN (10.2 GPA) and -Ga 2 O 3 (8.9 GPa). [27,28] The soft 20-nm thick ZnO interlayer can probably compensate the nano-roughness on the surface, resulting in a flatter surface during bonding compared with the direct bonding. Figure 4 shows the I-V-T measurements performed on the as-bonded GaN/Ga 2 O 3 and GaN/ZnO/Ga 2 O 3 structures for   comparison at temperatures ranging from 300 to 650 K with a step increment of 25 K. For these measurements, Ga 2 O 3 was grounded, and the bias was applied to GaN. The I-V curves measured on the GaN/Ga 2 O 3 direct bonding samples resemble that of a Schottky contact (Figure 4a,b), indicating that there is a barrier between GaN and Ga 2 O 3 . A turn-on voltage of ≈2.2 V was measured at room temperature which was reduced to 1.5 V at 650 K. The current density also increased by increasing the temperature as expected for thermionic emission of electrons over a barrier. A conduction band barrier forms at the GaN and Ga 2 O 3 interface, which prevents electrons moving from GaN to Ga 2 O 3 . An electron barrier of ≈0.40 eV was extracted from the Richardson model. The theoretical band offset (based on the electron affinity rule, which is extremely approximate) was calculated to be 0.6 eV from the difference of the electron affinity of GaN (4.1 eV) and Ga 2 O 3 (3.5 eV), [29,30] respectively. However, we recognize that the band offset is almost always different from the theoretical calculation in practice due to the interface dipoles and so we use the calculated number merely as a guide. [31] Similarly, the as-bonded sample using ZnO as the glue layer demonstrated a barrier limited transport behavior (Figure 4c,d), suggesting a barrier between GaN and Ga 2 O 3 , with more current flowing from Ga 2 O 3 to GaN (forward bias) compared with that flowing from GaN to Ga 2 O 3 (reverse bias). The test structure was measured again at room temperature after measurements were performed up to 650 K to determine whether hightemperature measurements affected the device performance. As shown in Figure 5a, the test structure showed deviation of I-V curves before and after temperature-dependent measurements (up to 650 K) as compared in Figure 5b and showed close to an ohmic behavior after the device was cooled down to room temperature and was measured again. However, the I-V characteristic  almost recovered to its initial Schottky-like behavior after 48 h. This could be explained by the occupancy of slow traps in amorphous ALD-ZnO relaxing over time.

Discussion
After annealing at 600°C in N 2 for 30 min, the I-V curves showed more ohmic-like behaviors in both heating up (from 300 to 650 K) and cooling down (650 to 300 K) measurements (Figure 6a,b). The maximum current density of 0.02 A cm −2 at V bias = 6 V was measured at room temperature. The current density biased at negative voltage was increased to be a similar level to that biased at positive voltage. Then, the test structure was again measured at room temperature at the end of cooling down measurements performed from 650 to 300 K. Figure 6c shows less current deviation before and after high temperature measurement compared with Figure 4c. This motivated us to further anneal the devices at a higher temperature. Figure 7a,b presents the I-V-T characteristics of test structure after annealing at 1100°C for 30 min with ramping up and down rates of 2°C min −1 . The device demonstrated the desired ohmiclike characteristics. This time, only a small shift of I-V curves under both positive and negative voltage biases could be observed with subsequent elevated temperatures ramped up to 650 K and then ramped back down to 300 K. The maximum current density of 1.24 A cm −2 at V bias = 6 V and the resistance of 65 Ω were achieved at room temperature. This resistance could be dominated by the contact resistance on UID (−201) -Ga 2 O 3 substrate with a carrier concentration ≈2.3 × 10 17 cm −3 . Heavily doped Si implantation should be performed to reduce this resistance in the future work. As depicted in Figure 7c, the device almost preserved its I-V characteristics after high-temperature measurement, exhibiting the annihilation of ultra-slow traps at the interface by the 30 min anneal at 1100°C in N 2 .
To understand what caused the barrier lowering, highresolution STEM-BF imaging and elemental mapping were performed on the GaN/ZnO/Ga 2 O 3 test structure after annealing at 1100°C. As shown in Figure 8, the ZnO interlayer was fully crystallized as if it has been grown epitaxially. In addition, the annealed ZnO is clearly observed to be separated into two sections. The part attached to GaN had a thickness of 12.2 nm whereas the part attached to Ga 2 O 3 had a thickness of 34.1 nm. This is surprising because ZnO films with equal thicknesses of 10 nm were deposited on GaN and Ga 2 O 3 by ALD. The reason could be an increase in the volume due to intermixing of ZnO with the underlying Ga 2 O 3 substrate and formation of a different phase. Further investigations are necessary to fully understand the phase and atomic composition of this interlayer which is beyond the scope of this work and will be reported separately. Figure 9 shows element maps collected from the bonded interface. Figure 9a exhibits that the ZnO layer has intermixed with the Ga 2 O 3 . Ga is a shallow n-type dopant for ZnO and so this could lead to the ZnO becoming n-type, [32][33][34] which combined with an intermixing of the interface can be contributing reasons for the barrier between GaN and Ga 2 O 3 to be reduced as witnessed from the I-V characteristics.  . It shows that the ZnO grown on Ga 2 O 3 has been fully crystallized. There is a "gap" between the two ZnO layers.

Conclusion
In this work, the heterogeneous integration of N-polar GaN and Ga 2 O 3 substrates was demonstrated via the ZnO soft "glue" layer, which helped to achieve a fully bonded interface and improve tolerance to thermal treatment. No Newton's rings were observed on the sample which indicates a flat and clean bonded interface. Extensive I-V measurements were conducted on the as-bonded Ga 2 O 3 /ZnO/N-polar GaN test structure and after annealing at 600°C and 1100°C. A consistently ohmic-like characteristic was achieved by annealing the bonded wafers at high temperature in N 2, which is probably due to crystallization of ZnO and diffusion of Ga into ZnO which makes it n-type doped. The electrical and structural analysis of -Ga 2 O 3 /GaN wafer-bonded heterojunctions with ZnO interlayer provides promises for the development of novel high-power high-frequency electronic devices.