Introduction

Room-temperature bonding technology has emerged as a challenging process for three-dimensional integration and wafer-level packaging of optoelectronics, MEMS, RF, and CMOS devices. As a technique for direct bonding, hydrophilic bonding via plasma activation is becoming mainstream in the semiconductor industry1,2,3. This method requires annealing at a temperature of a few hundred degrees to achieve a sufficiently strong bond for practical applications. To overcome the severe thermal stress and degradation of bond alignment accuracy, the development of room-temperature bonding technology is essential.

As one solution, surface-activated bonding (SAB)4 is a promising room-temperature direct bonding technology. In this method, wafer surfaces are first irradiated using an argon (Ar) fast atom beam (FAB) to remove organic contaminants and the native oxide layer, and then, the activated wafer surfaces are brought into contact at room temperature. The applicability of this method to various materials has been studied for a wide range of device applications. To date, the SAB method has demonstrated room-temperature bonding of metals (Cu–Cu5,6 and Au–Au7,8) as well as semiconductors (Si–Si9, Ge–Ge10, Si–GaAs11, Si–SiC12, and GaAs–SiC13). However, it is difficult to apply this method to the direct bonding of SiO2 and SiN, which are suitable as insulator layer materials in semiconductor device applications14,15. To overcome this limitation, modified SAB16,17,18,19 and atomic diffusion bonding20,21 have been reported based on the quasi-direct bonding concept using a metal intermediate layer. In addition, a room-temperature bonding method using an Si intermediate layer has recently been reported22,23,24,25. In some electronics applications such as 3D integration using vertical electrical interconnections, these intermediate layers cause serious current leakage. Therefore, the SAB of insulating materials is highly desirable.

Al2O3 is a promising alternative insulator material because of its excellent electrical resistivity and thermal conductivity, which are comparable to those of SiO2. To date, the SAB of single-crystal Al2O326, SAB of Al2O3 deposited by mist CVD27, and hydrophilic bonding of Al2O328, which requires annealing at 100–300 °C after temporary bonding, have recently been examined. Herein, we focus on Al2O3 thin films grown by atomic layer deposition (ALD), which allows for precise growth, atomic-scale thickness control, and good uniformity on large-scale wafers. Previous reports on Al2O3 thin films deposited by ALD, have investigated hydrophilic bonding29,30 and modified SAB31, in which an adhesive of Si is sputter-deposited on ALD-deposited Al2O3 films. SAB32 of AlO films deposited by ion beam sputtering has also been examined. It is interesting to apply ALD Al2O3 thin films to the SAB method.

There have only been a few studies on the application of the SAB method to the direct bonding of ALD Al2O3 thin films without a heating process, and a large bond strength has not yet been achieved. Herein, the applicability of the SAB method to the direct bonding of ALD Al2O3 thin films is investigated in comparison with those of single-crystal sapphire. Furthermore, we demonstrate the room-temperature wafer direct bonding of deposited Al2O3 thin films on a Si thermal oxide wafer using the ALD method. Compared to previous reports, the novelty of this study is the utilization of a surface-activated bonding method using Ar-FAB for Al2O3 thin films deposited by ALD.

Results and discussion

SAB of ALD Al2O3 thin films on 4-inch Si thermal oxide wafers was performed at room temperature. Ar FAB irradiation was used for surface activation. Surface smoothness is one of critical factors for room-temperature direct bonding, and a rms of approximately 0.5 nm or less is typically required33. First, the wafer surfaces were studied before and after Ar FAB irradiation using atomic force microscopy (AFM). Figure 1 shows an AFM image of the ALD Al2O3 surface over a measured area of 1 µm × 1 µm. Consequently, the RMS surface roughness of the ALD Al2O3 thin film was 0.32 nm and 0.25 nm before and after the Ar FAB irradiation, respectively. This indicates that Ar FAB irradiation for surface activation did not roughen the surface significantly, and the surface of the ALD Al2O3 film retained sufficient smoothness.

Figure 1
figure 1

AFM image of the deposited Al2O3 thin film using ALD method.

Figure 2 shows the infrared (IR) transmission image of the SiO2–SiO2 interface bonded with ALD Al2O3 thin films, showing no large voids in the entire bonded wafer. The bond strength was evaluated using the crack-opening method34 and the half-cutting dicing test. In the crack-opening method, a razor blade is inserted into the bond interface, and the surface energy, which indicates the bond strength, can be calculated using the propagated crack length. Using the measured crack length, the estimated surface energy of ALD Al2O3–ALD Al2O3 was approximately 1.5 J/m2. The dicing test of the bonded wafer was then performed using a dicing saw. The bonded wafer was cut into 10-mm-square chips, and then one chip was half-cut into 0.5-mm-square pieces, leaving the bottom of the chip. The durability against the applied stress during the dicing process indicates the bond strength. As shown in Fig. 3, perfect dicing was successful with no debonding or chipping observed. These results indicate that a strong bond between ALD Al2O3 thin films can be achieved, which may be sufficient for device applications.

Figure 2
figure 2

IR image of 4-inch-scale bonded thermal oxidized Si–Si wafer with ALD Al2O3 thin films at room temperature.

Figure 3
figure 3

Image of the diced 0.5 mm × 0.5 mm chips, showing perfect dicing without debonding.

The nanostructure of the cross-sectional bonding interface was observed using transmission electron microscopy (TEM). Figure 4a,b show the TEM images obtained at low and high magnifications, respectively. The dotted line in Fig. 4b shows the initial bond interface of the ALD Al2O3 thin film.

Figure 4
figure 4

(a) Low- and (b) high-magnification cross-sectional TEM images of the bonding interface.

As shown in Fig. 4b, a void-free bonding interface could be achieved at the atomic level, indicating the presence of a strong bond. In addition, we can see no damaged layer by Ar FAB irradiation, whereas this damaged layer clearly exists in the SAB of sapphire–sapphire.

These results indicate that the bonded ALD Al2O3 thin films functioned as adhesives and facilitated strong bond formation between two SiO2 films. Figure 5 shows the results of elemental analysis obtained using energy-dispersive X-ray spectroscopy (EDX). Figure 5a shows a scanning TEM (STEM) image; the numbers in the image indicate EDX measurement points, and Fig. 5b shows the element concentrations at each measurement point. The elemental analysis of the bonded ALD Al2O3 in the depth direction (Fig. 5b) shows that there is little change in composition from the bonding interface in the depth direction (measurement points 4–6).

Figure 5
figure 5

Elemental analysis using cross-sectional STEM-EDX.

The bond strengths of the different Al2O3 were investigated. Figure 6 compares the bond strengths of sapphire–sapphire (SA–SA), sapphire–ALD Al2O3 (SA–ALD), and ALD Al2O3–ALD Al2O3 (ALD–ALD). Using the crack-opening method, the estimated surface energies of the bonded SA–SA, SA–ALD, and ALD–ALD were approximately 2.0, 1.9, and 1.5 J/m2, respectively. It is well-known that the typical Al2O3 thin film deposited by ALD is amorphous. As shown in Fig. 4, no lattice structure was visible in the bonded ALD Al2O3 thin film, indicating that it was amorphous. As shown in Fig. 7, the measured X-ray diffraction (XRD) results also indicated that the Al2O3 thin film deposited by ALD is amorphous.

Figure 6
figure 6

Comparison of the bond strengths of different Al2O3 samples.

Figure 7
figure 7

XRD pattern of ALD Al2O3 thin film.

On the other hands, although sapphire has a single-crystalline structure, an amorphous-like layer (thickness ≈ 1 nm) exists at the bond interface owing to Ar FAB irradiation27. We can see that this crystal defect layer has little effect on the bonding strength because sufficient bond strength can be achieved compared with that of amorphous Al2O3 thin films. We confirmed that the Al2O3 thin film-sapphire bond was weaker than the sapphire–sapphire bond. These results suggest that the crystallinity of the activated surface affects the bond strength of Al2O3. In addition, the experimental results demonstrated that the ALD method is suitable for the SAB of Al2O3 thin films.

Next, SAB of SiO2–SiO2, SiO2–ALD Al2O3 (SiO2–ALD), SiO2–SA (SiO2–sapphire), and ALD Al2O3–ALD Al2O3 (ALD–ALD) was performed under the same bonding conditions. Figure 8 shows a comparison of the bond strengths between SiO2 and different Al2O3 samples. The low bond strengths of SiO2–SiO2 and SiO2–ALD were confirmed, with the bond strength of SiO2–ALD being slightly larger. Comparatively, the bond strength of SiO2–SA was larger than both. This indicates that simply changing one of the wafers to one with better crystallinity tends to increase bond strength. However, it was necessary to form ALD Al2O3 films on both SiO2 sides and bond them to achieve a sufficient strong SiO2–SiO2 bond interface for device applications. These results also suggest that the single-crystalline phase of Al2O3 is more effective, and crystallinity is important from the viewpoint of bond strength.

Figure 8
figure 8

Comparison of bond strengths between SiO2 and different Al2O3 samples.

Conclusion

Herein, the SAB method was successfully applied to achieve the room-temperature wafer bonding of ALD Al2O3 thin films. The bonded ALD Al2O3 thin films appear to work well as nanoadhesives and form a strong bond between SiO2 surfaces. The bonded wafer was perfectly diced into dimensions of 0.5 mm × 0.5 mm, and the surface energy, indicative of the bond strength, was estimated to be approximately 1.5 J/m2. These results indicate that strong bonds that are sufficient for device applications can be achieved. In addition, the applicability of different Al2O3 microstructures to the SAB method was investigated, and the effectiveness of applying ALD Al2O3 was experimentally verified. This successful SAB of Al2O3 thin film, which is a promising insulator material, opens the possibility of future room-temperature heterogenous integration and wafer-level packaging. In additions, the results of this work will be of significant use in the development of fabrication technique for X materials-on-insulators using room temperature bonding method with ALD Al2O3 thin film, and not only SiO2–SiO2 bond interface.

Methods

In our experiments, we used an ultrathin Al2O3 film (thickness = 6 nm) deposited by atomic layer deposition (ALD) on a 4-inch wafer with a 1-µm thermal oxide film. The ALD deposition process used trimethylaluminum (TMA) as a precursor and ozone (O3) as an oxidant. The deposition temperature was 400 °C. Post-deposition annealing was not performed.

Bonding was performed using a wafer-level bonding apparatus (MWB-08-AX, NIDEC MACHINE TOOL CORPORATION). For the bonding process, two wafer surfaces were simultaneously irradiated with Ar-FAB in a vacuum chamber at 5.1 × 10−6 Pa. The FAB irradiation conditions were as follows: voltage was 1.5 kV, current was 100 mA, Ar flow rate was 13 sccm, and FAB irradiation time was 30 s.. After FAB irradiation, the FAB-irradiated surfaces of the wafers were immediately brought into contact with each other in a vacuum apparatus, and a load of 9.8 × 104 N was applied for 10 s. for bonding.

X-ray diffraction (XRD) method was used to evaluate the crystallinity of ALD Al2O3 thin films. In-plane measurements were performed with an X-ray incidence angle of 0.5 degrees (the critical angle was 0.4 degrees).

The roughness of the deposited surface was evaluated using atomic force microscopy (AFM, DimensionIcon, Bruker). The nanostructure and elemental composition of the bonding interface were analyzed by TEM (Hitachi Hitech H9500) and STEM-EDX (Hitachi Hitech G4000).