Novel hydrophilic SiO₂ wafer bonding using combined surface-activated bonding technique

Low-temperature SiO₂ wafer direct bonding with a high-strength and void-free interface is important for manufacturing silicon-on-insulator (SOI), micro-electromechanical systems (MEMS), and three-dimensional (3D) integration.^1,2) Hydrophilic bonding approaches with wet chemical or plasma activation have been widely studied for SiO₂ wafer bonding.^3–15) Typically, hydrophilic wafer bonding is performed in humid air after prebonding surface treatment, followed by postbonding annealing. However, trapped species at the bonding interface may generate bubbles and voids.^16,17) Also, the hydrogen produced through the oxidation of the silicon bulk by excess water molecules results in the formation of voids.^18,19) Futhermore, bonding energy is limited by the stress corrosion effect induced by remaining water at the bonding interface.²⁰⁾ Wafer bonding in vacuum is necessary to minimize the amounts of trapped species and excess water, which easily desorbs from the wafer surface in vacuum, at the bonding interface.

Yu and coworkers studied Si–SiO₂ wafer bonding in moderate vacuum using a wet prebonding cleaning.^21,22) The result showed a very low bonding energy and an increased unbonded area with increasing vacuum level, due to the lack of coverage of molecular water on the bonding surfaces. Suni et al. reported that the bonding energy of Si–SiO₂ pairs with oxygen plasma activation decreases when the bonding is performed in vacuum or at 150 °C in air.²³⁾ However, with nitrogen or argon plasma activation, no dependence of bonding energy on the bonding atmosphere was observed. They suggested that the difference is attributed to the various numbers of surface Si–OH groups. The surface activated by oxygen plasma is saturated with oxygen and cannot bind to –OH before a wet etching step. Nevertheless, the surface activated by nitrogen or argon plasma is enriched in silicon and is readily hydrated stably. The same behavior was observed by Dragoi and Lindner, who also reported that the Si–Si pairs bonded in vacuum show almost the same bonding energy using either oxygen or nitrogen plasma activation.²⁴⁾ These previous works indicate the important role of the number of Si–OH groups in hydrophilic bonding. However, SiO₂–SiO₂ bonding with a high bonding energy is more challenging than Si–SiO₂ and Si–Si bonding, because of the lower capability of a thick SiO₂ layer to accommodate –OH than of a thin silicon native oxide layer. To date, research on low-temperature hydrophilic SiO₂–SiO₂ bonding in vacuum with a high bonding energy has remained insufficient.

The hydrophilic bonding of Cu/SiO₂ wafers by surface activation with Ar fast-atom beam bombardment and water vapor exposure was previously studied in air by Shigetou and Suga.^25,26) In this work, we investigate hydrophilic SiO₂–SiO₂ wafer bonding in vacuum. We propose a new combined surface-activated bonding (SAB) technique involving a combination of ion beam bombardment, in situ silicon deposition, and water vapor exposure processes for surface activation prior to bonding in vacuum. The new bonding technique is expected to increase bonding energy by increasing the numbers of Si–OH on SiO₂ surfaces. An increased surface energy of more than 1 J/m² is achieved by postbonding annealing at 200 °C. Wafer surface analysis, surface energy measurement, and interface analysis were carried out to identify the mechanism of the proposed combined SAB technique.

To develop the present combined SAB technique, we constructed a new bonding apparatus that consists of a wafer transport subsystem, a wafer surface treatment subsystem and an ultrahigh vacuum bonding subsystem, as shown in Fig. 1. In the ultrahigh vacuum bonding chamber, ion beam bombardment and silicon deposition can be performed. The ion beam incident onto the wafer surface is neutralized by an electron beam to prevent charging damage. In the surface treatment chamber, water vapor exposure at a controlled vapor flow rate can be carried out. Wafers with diameters of 50, 75, 100, and 150 mm can be aligned and bonded at temperatures ≤200 °C under a bonding force of up to 20 kN. The bonding apparatus was placed in unclean room atmosphere.

Zoom In Zoom Out Reset image size

100-mm p-type silicon wafers with a 100-nm-thick thermally grown SiO₂ layer were used for the wafer bonding experiments. The process flow of the combined SAB is shown in Fig. 2. The wafers were irradiated by Ar ion beam bombardment and simultaneously subjected to in situ silicon deposition. Line-type Ar ion beam bombardment at a power of 1.0 kV × 100 mA and a scanning speed of 7.5 mm/s was carried out to irradiate the wafer surface under ultrahigh vacuum at a pressure of less than 10⁻⁴ Pa. In situ silicon deposition was carried out using a silicon sputter target in the chamber. The ultrahigh vacuum prevents the rapid oxidation of wafer surfaces during the surface irradiation and silicon deposition. A 300 mm SiO₂ wafer was used to cover the metal stage to prevent sputtered metal contamination. Then, the wafers were transported into a vacuum chamber with a pressure of 10⁻² Pa. The low pressure is used to reduce the adsorption of oxygen and other contaminants on the wafers. Water vapor and nitrogen carrier gas were introduced into the vacuum chamber at flow rates of 100 and 20 cm³/min, respectively. The wafers were exposed to water vapor for 5 min while the chamber pressure was increased to approximately 10⁻¹ Pa. The flow rates and exposure duration were chosen considering the maintenance of the low chamber pressure. Then, the wafers were transported into a bonding chamber and bonded at room temperature in vacuum at a pressure of 10⁻² Pa. A bonding force of 5 kN was applied for 1 min for the wafers to bond tightly. The bonded wafers were postbonding-annealed in air at 200 °C for 7 h. The relatively long annealing was carried out to ensure the achievement of the saturated bonding energy at the given annealing temperature. For comparison, wafer bonding experiments without in situ silicon deposition and water vapor exposure were also performed using the same experimental parameters.

Zoom In Zoom Out Reset image size

Wafer surfaces were analyzed by X-ray photoelectron spectroscopy (XPS) using a non-monochromatic Mg Kα source to investigate the effects of treatments on the SiO₂ surface. To emphasize the difference in peak position, the spectra were calibrated using a binding energy of C 1s, which came from the organic contaminants on the surface. Bonding energy was measured by the crack opening method in air and at room temperature.²⁷⁾ In this method, the surface energy (i.e., half of the bonding energy) with the debonded length induced by insertion of a blade at the bonding interface was calculated using the following equation:

$$\begin{equation} \gamma=\frac{3Et^{3}y^{2}}{32L^{4}}, \end{equation} \tag{ 1 }$$

where E = 166 GPa is the Young's modulus for silicon wafer, t the thickness of each wafer, y the blade thickness, and L the debonded length measured by infrared transmission imaging.

The bonding interface was inspected by infrared transmission imaging and transmission electron microscopy (TEM) at an acceleration voltage of 200 kV. Elemental analysis of the bonding interface was performed using an energy dispersive spectroscopy (EDS) attached to a TEM system.

3.1. Results

The XPS analysis results of the SiO₂ surfaces are shown in Fig. 3. In the XPS profiles, the peaks detected at binding energies of 103.80 and 103.35 eV correspond to the SiO₂ and Si–OH groups, respectively. Only the SiO₂ peak with a full width at half maximum (FWHM) of 1.77 eV was observed on the surface irradiated by ion beam bombardment. On the surface activated by the combined SAB without the in situ silicon deposition, the peaks corresponding to SiO₂ with a FWHM of 1.83 eV and to Si–OH with a FWHM of 1.43 eV were observed. On the surface activated by the combined SAB, the peaks corresponding to SiO₂ with a FWHM of 1.81eV and Si–OH with a FWHM of 1.48 eV were observed. The XPS results show that the intensity of the Si–OH groups on the surface activated by the combined SAB is significantly greater than that activated by the process without in situ silicon deposition. No peak corresponding to bulk silicon was detected in all cases.

Zoom In Zoom Out Reset image size

Table I shows the surface energies of the SiO₂–SiO₂ interface measured by the crack opening method. The surface energies of SiO₂–SiO₂ pairs with the combined SAB were 1.09–1.54 J/m², which are high enough to sustain postbonding processes such as mechanical grinding and wet-chemical etching.²⁸⁾ The surface energies of SiO₂–SiO₂ pairs bonded using the combined SAB without in situ silicon deposition were 0.45–0.79 J/m². The wafer bonding failed with the bonding process without water vapor exposure.

Table I. Surface energies of SiO₂–SiO₂ pairs bonded by the combined SAB.

	Surface energy γ (J/m2)
	Test (1)	Test (2)	Test (3)	Test (4)
Combined SAB	1.09	1.54	1.09	1.09
Combined SAB without in situ silicon deposition	0.79	0.59	0.59	0.45
Combined SAB without water vapor exposure	Failure

Figure 4 shows the infrared transmission images of a SiO₂–SiO₂ pair bonded with the combined SAB before and after postbonding annealing. Bonding in vacuum prevents the generation of large bubbles owing to gas trapping, although some particle defects were formed at the bonding interface. The particles were adsorbed on the surface during the wafer transfer between the water vapor exposure chamber and the bonding chamber. No annealing voids were visible at the bonding interface.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

TEM was applied to detect sub-micrometer interfacial voids. Figure 5 shows a TEM image of the SiO₂–SiO₂ bonding interface of a sample prepared by combined SAB after 200 °C annealing in air. A void-free intermediate layer with a thickness of about 15 nm was found at the bonding interface. Figure 6 shows the EDS profiles of the SiO₂ layer and bonding interface. The O Kα and Si Kα peaks correspond to silicon oxide. The C Kα peaks must be from the adhesive used to fix the TEM specimen. The Al Kα peaks must be from the sputtered contaminants generated during argon ion slicing for TEM specimen preparation.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

3.2. Discussion

The present combined SAB technique was successfully demonstrated to be feasible for hydrophilic wafer bonding in vacuum with a high bonding energy. The wafer surfaces are irradiated by ion beam bombardment and simultaneously deposited with silicon by in situ silicon deposition, and then terminated with Si–OH groups by water vapor exposure. The failure of SiO₂ bonding in the combined SAB without water vapor exposure may be due to the absence of Si–OH groups on the wafer surfaces.²⁹⁾ However, the ion beam bombardment is necessary because it removes surface adsorbates and contaminants such as hydrocarbons, which may generate unbonded areas or bubbles. When in situ silicon deposition is employed during the Ar ion beam bombardment, the deposited silicon with dangling bonds are reactive for hydroxylation during the water vapor exposure step, which converts the deposited silicon to Si–OH. As a consequence, the number of Si–OH groups terminated on the surface increases, as indicated in the XPS spectra. Figure 7 shows a schematic diagram of the Si–OH groups on the silicon deposited on SiO₂ surface. Because of the short distances between the deposited silicon sites, the Si–OH groups can interact with each other through hydrogen bonds.³⁰⁾ We can assume that each Si–O–Si bond is transformed from one Si–OH pair after postbonding annealing, and that the surface energies of 0.79 and 1.54 J/m² correspond to numbers of Si–OH of 2.17 and 4.23 per square nanometer, respectively.⁴⁾ These indicate that the deposited silicon is much more reactive for surface hydroxylation than SiO₂ and that the number of Si–OH can be almost doubled. To further increase the number of Si–OH, a longer water vapor exposure can be carried out to achieve a higher degree of surface hydroxylation. A Si–OH number of more than 6.87 per square nanometer is desired to obtain a surface energy close to a silicon fracture energy of approximately 2.5 J/m².⁴⁾

Zoom In Zoom Out Reset image size

The infrared transmission images of the bonded wafer show no unbonded areas or bubbles generated by gas trapping or annealing. The voids generated from particles can be eliminated by performing the bonding in clean-room atmosphere. Also, since most of the silicon deposited on the wafer surfaces has been converted to Si–OH prior to wafer bonding, no voids generated from hydrogen produced by silicon oxidation reaction are formed at the interface.

The bonding intermediate layer consists of two bonded amorphous SiO₂ layers with different contrasts from the SiO₂ bulk in the TEM image. The EDS profiles shown in Fig. 6 suggest that the interface mainly consists of Si and O, which confirms that the bonding interface is silicon oxide, not silicon. No ion-beam-induced metal contaminants such as Fe was detected at the bonding interface, since the metal stage is covered with a SiO₂ wafer. The amorphous layer is induced by Ar ion beam bombardment, with a typical thickness of below 5 nm. The amorphous intermediate layer facilitates the diffusion of generated water from the bonding interface to outside. The total thickness of the intermediate layer is about 15 nm, which could be greater than that prepared by the vapor-assisted bonding method owing to the deposited silicon.²³⁾ The thickness can be modified by controlling the power of the ion beam source and the durations of the ion beam bombardment and silicon deposition process. To achieve a high coverage of Si–OH on the surface, one or more continuous silicon monolayers are prefered.

The combined SAB is promising for SiO₂–SiO₂ direct wafer bonding at an even lower temperature than for bonding with plasma activation, in which case the surface energy is usually limited to below 1.5 J/m² with 300 °C annealing.^10,14) Futhermore, because the surface coverage of the water monolayer is not necessary during the bonding step, this hydrophilic bonding approach can minimize the number of water at the interface. It is also useful for other void-free silicon-based wafer bonding, such as Si–SiO₂ and Si–Si bonding, in which cases the hydrogen generated through the reaction between silicon and excess water is the main origin of the interface voids.

In this work, we utilized a new combined SAB technique for SiO₂ wafer bonding in vacuum at a low temperature. In situ silicon deposition is combined with ion beam bombardment and water vapor exposure for surface activation prior to wafer bonding. The wafer surface is irradiated by ion beam bombardment and simultaneously subjected to in situ silicon deposition, and then terminated with Si–OH groups by water vapor exposure. The deposited silicon is reactive for hydroxylation by water vapor, so the number of Si–OH groups terminated on the bonding surface increases. The activated surfaces are bonded in vacuum at room temperature and postbonding annealed in air. A surface energy above 1 J/m² was realized by 200 °C annealing. An oxide intermediate layer without any voids was confirmed at the bonding interface. This combined SAB technique is useful for the development of novel void-free low-temperature wafer bonding approaches with a high bonding energy.