1. Introduction

Titanium dioxide (TiO₂) is of much interest to many fields due to its unique properties, especially its degradation of H₂O to H₂ and O₂, as well as its UV-absorption ability, that are widely used in photocatalysis and cosmetics [1,2]. As a wide bandgap semiconductor, TiO₂ is also useful to transparent optoelectronic devices owing to its excellent optical properties [3–5], including moderately high refractive index, large bandgap of 3.1∼3.2 eV, low thermal expansion coefficient, negative thermo-optic coefficient [6,7], large Kerr nonlinearity [8,9], etc. In particular, because of its moderately high refractive index and light absorption below 390 nm wavelength, TiO₂ has been added to the family of materials suitable for optical microcavities, such as whispering-gallery-mode resonators (WGMRs) and photonic crystals for control and manipulation of visible light [10]. Therefore, TiO₂ thin films can be applied in visible photonics, optical sensing, photovoltaics, etc. Among its three (anatase, rutile and brookite) phases, TiO₂ in rutile phase has the largest refractive index, (>2 for the amorphous phase, >2.5 for the anatase phase, and >2.7 for the rutile phase), as well as greater thermal stability, so that optical microcavities based on rutile TiO₂ (r-TiO₂) thin film can achieve higher-Q value in visible wavelengths.

Most TiO₂ thin films are fabricated using epitaxial growth methods, such as sputtering, electron beam evaporation, sol-gel, etc, but the resulting TiO₂ thin films are mostly polycrystalline or nanoparticle composites, rather than single-crystalline [11–13]. Although r-TiO₂ thin films homostructure can be formed by metal-organic chemical vapor deposition (MOCVD) [14,15], r-TiO₂ thin film heterostructure still has not been realized. Considering the importance of r-TiO₂ thin film heterostructure with high refractive index contrast in making high-performance photonic devices that depend on optical microcavities with high-Q, fabrication of single-crystalline r-TiO₂ thin film heterostructure is of urgent interest.

Ion implantation as a promising technology has many useful applications [16–21]. In particular, thin film fabrication using ion-implantation assisted with wafer bonding, well-known as “smart-cut” [22,23], should be valuable for integrated optoelectronics. The principle of this method is: light ions such as H⁺ and He⁺ with fluences of 10¹⁶-10¹⁷ ions/cm² implanted into a crystal generate a large concentration of dislocations and defects at the end of ion range, which trap H⁺/He⁺ into the bubbles/cracks [24–26]. Subsequent wafer bonding and annealing treatments can promote the He bubbles to aggregate into large cavities. As a result, the implantation area above the damage layer can be exfoliated from the bulk as a thin film [27–30]. This layer splitting method can be used to fabricate single-crystalline r-TiO₂ thin films. Ion-implantation characteristized by controllability and reproducibility has unique advantage in controlling the thickness of the exfoliated r-TiO₂ films.

Wafer bonding is the key step in the fabrication of uniform sub-micrometer r-TiO₂ thin film. A controlled split-off or exfoliation is possible only after proper bonding of the implanted sample onto a supporting substrate. Traditional direct bonding method that has been successfully applied to fabrication of LiNbO₃ on insulator (LNOI) structure, has rigid requirements on the flatness and smoothness of both the donor and substrate wafer surfaces [31,32]. However, the surface roughness of ion-implanted r-TiO₂ sample cannot satisfy these requirements. Since the thermal treatment at elevated temperature (>400°C) is required for the exfoliation of the implanted r-TiO₂ samples, the glass transition temperature (about 350°C) of the BCB resin as well as the difference in the thermal expansion coefficient between the bonded materials prohibits the indirect bonding method using BCB resin. Thus, both direct bonding and BCB bonding cannot be used for fabricating r-TiO₂ thin film heterostructure. Cu/Sn solid-liquid interdiffusion (SLID) wafer-level bonding is an attractive technique for encapsulation and interconnection of micro-electro-mechanical systems (MEMS) due to its low cost, high temperature stability, high bond strength, and hermeticity [33,34]. The Sn layer melts and the intermetallic compounds (IMCs) solidify isothermally at bonding temperatures ranging from 250 to 300°C. The overall goal of the Cu-Sn bonding process is to achieve a Cu-Cu₃Sn-Cu final bonding layer that is thermodynamically stable (melting temperature higher than 676°C) [35]. Therefore, Cu-Sn bonding technology is suitable for r-TiO₂ thin film exfoliation.

In this paper, low energy He⁺ ions with high ion fluence were implanted into rutile TiO₂ single crystal and aggregated at the end of the ion range (approximate 680 nm), in combination with Cu-Sn bonding and annealing, a heterostructure of single-crystalline r-TiO₂ thin film onto SiO₂ substrate can be formed. Scanning electron microscopy (SEM), energy dispersive spectrometer (EDS), X-ray diffraction (XRD) and atomic force microscopy (AFM) were used to examine the properties of exfoliated r-TiO₂ thin film. Rutherford backscattering spectroscopy/channeling (RBS/channeling) and transmission electron microscopy (TEM) were applied to analyze the exfoliation mechanism of the ion-implanted r-TiO₂.

2. Materials and methods

2.1 Fabrication of the r-TiO₂ thin film heterostructure

200 keV He⁺ ions with ion fluence of 8×10¹⁶ ions/cm² were implanted into rutile TiO₂ single crystals along the [001] axis at liquid nitrogen conditions to inhibit dynamic thermal effect during ion-implantation process. Before the implantation, the sample with size of 5mm×5mm×0.5 mm was optically polished and cleaned. During the implantation, the ion beams were electrically scanned to ensure a uniform implantation over the sample, and the sample was tilted by 7^o off the beam direction in order to minimize the channeling effect. The implantation process was performed at an LC-4 Ion Implanter at the Institute of Semiconductors of the Chinese Academy of Sciences.

After the implantation, the implanted sample was cleaned using the standard RCA recipe. Atomic force microscopy (AFM) and optical microscope were used to check the intactness of the sample surface, showing that the surface roughness before and after implantation is almost the same, lower than 0.5 nm. After the cleaning process film coating was implemented. A 2μm SiO₂ thin film was deposited on the implanted sample surface using plasma-enhanced chemical vapor deposition (PECVD) to form a heterostructure with high refractive index contrast. Then the surfaces of both the deposited SiO₂ and a 10mm×10mm×0.5 mm virgin r-TiO₂ single crystal acting as a receptor were successively covered by 100 nm chromium (Cr) and 5μm Cu with magnetron sputtering, and 1.5μm Sn layers with thermal evaporation. Considering the thermal expansion coefficients matching, it is better to use the same donor and receptor sample for Cu-Sn bonding. The thickness of Cr, Cu and Sn were chosen according to the simulation model established in Ref. [35] that predict the intermetallic compounds development during the bonding process, and Cr layer was used as an adhesive to attach the r-TiO₂-SiO₂ heterostructure with Cu-Sn bonding layer strongly [33]. The coating processes were conducted at low temperature without imposing influence on the damage region in the He⁺-implanted r-TiO₂ samples. After the coatings, the coated surface of both the donor and receptor r-TiO₂ samples were brought into contact and pressed together at 150°C, and Cu-Sn bonding strongly was completed after annealing at an elevated temperature of 270°C for 20 minutes. R-TiO₂ thin film was exfoliated from the bulk material after annealing at 400°C for 10 hours. The receptor r-TiO₂ bulk can act as a substrate to facilitate the r-TiO₂ thin film exfoliation as well as preserve the intactness of the TiO₂ thin film. This thin film fabrication process is illustrated in Fig. 1.

 

Fig. 1. Schematic presentation of r-TiO₂ thin film heterostructure fabrication using ion-implantation together with Cu-Sn bonding: (a) 200 keV He⁺-implantation into r-TiO₂ donor wafer at ion fluence of 8×10¹⁶ ions/cm². (b) Film coating on both the r-TiO₂ donor and receptor wafer. (c) Cu-Sn bonding of the coated donor and receptor wafers together. (d) r-TiO₂ thin film exfoliation from bulk after annealing.

Download Full Size | PDF

2.2 Characterization of the r-TiO₂ thin film

The lattice damage induced by the ion-implantation was analyzed using Rutherford backscattering spectroscopy/channeling (RBS/channeling) measurement, which was performed by 2 MeV He²⁺ ions at a scattering angle of 165^o in the 1.7 MV tandem accelerator at Peking University. The sample was mounted on a three-axis high precision goniometer in a vacuum chamber, which can be carefully angle oriented to minimize backscatter for the channeling measurements. The backscattering counts (yield) were recorded as a function of the channel (energy) of the different target atoms. Stopping and Range of Ion Matter (SRIM 2008) was used to simulate the distribution of implanted He⁺-ions in r-TiO₂. The inner lattice structural properties in cross-section, especially in the damage layer, were observed with transmission electron microscopy (TEM) using the Tecnai G2 F20 S-Twin at 200 kV with a field emission gun. The TEM samples were prepared using Focused Ion Beam (FIB) technique. The surface morphology of the exfoliated TiO₂ thin film was investigated using optical microscopy (OM) and atomic force microscopy (AFM) methods. AFM was performed under ambient conditions on a Bruker Multi-mode VIII microscope, operating in tapping mode at a cantilever frequency of 250 ± 10 kHz. Scanning electron microscopy (SEM) and Energy Dispersive X-ray Spectroscopy (EDS) were used to examine the properties of the fabricated r-TiO₂ thin film heterostructure. The crystal structure of the r-TiO₂ thin film was investigated by XRD (Empyrean) using CuK-α radiation working in the θ-2θ mode.

3. Results and discussion

Different implantation parameters, including ion fluences, energy, temperature, flux, can induce different lattice damages, and for good layer splitting there exist critical values of the ion fluence and annealing temperature. A wide range of implantation parameters were applied to the r-TiO₂ samples. Bubbles and craters were found on the surface of the samples implanted with 200 keV He⁺ ions at fluences larger than 7×10¹⁶ ions/cm² after annealing at temperature of 550°C for 1 hour [21]. With increase of the annealing temperature and time, the number of bubbles and craters increases. High annealing temperature can also cause breakage of the Cu-Sn bonding. We found that the critical annealing temperature for layer splitting decreases with increasing ion fluence, and to realize wafer exfoliation, 200 keV He⁺ ions with 8×10¹⁶ ions/cm² fluence at liquid nitrogen temperature should be used for the implant. This phenomenon can be attributed to the high concentration of defects created by He⁺-implantation. It is well known that implanted ions lose their energy through nuclear interactions with the lattice atoms at the end of the ion range, which can generate vacancy-interstitial pairs. Ions with higher fluences can generate more vacancies, and the liquid nitrogen environment can inhibit the dynamic thermal effects during the implantation that can promote the out-diffusion of implanted ions. Therefore, with low-temperature thermal treatment, high density of vacancies is more likely to trap the implanted He⁺-ions into bubbles. However, He⁺-ions with higher ion fluences may also induce damage in the exfoliated r-TiO₂ thin film, which would weaken its single-crystalline properties. Thus, 8×10¹⁶ ions/cm² fluence at liquid nitrogen temperature is appropriate implantation parameters to fabricate r-TiO₂ thin film in good quality. The implantation parameters mentioned in this paper are examined to be suitable for r-TiO₂ thin film exfoliation. Figure 2 shows nothing happened on the ion-implanted r-TiO₂ sample surface before thermal treatments, but the bubbles and craters appearing on the implanted-sample surface after annealing at only 400°C for 1 hour, and the number of craters increases with increase of the annealing time.

 

Fig. 2. Surface morphology of ion-implanted r-TiO₂ sample before (a) and after annealing at 400°C for 1 hour (b) and 5 hours (c).

Download Full Size | PDF

Following the fabrication process shown in Fig. 1, the implanted sample was attached to the substrate to form the heterostructure through Cu-Sn bonding, and r-TiO₂ thin film with thickness of 680 nm was exfoliated from bulk after annealing at 400°C for 10 hours. The cross-section of the fabricated r-TiO₂ thin film heterostructure detected by SEM is shown in Fig. 3. We see that it is composed of the exfoliated r-TiO₂ thin film, the deposited SiO₂ layer, the Cr layer, and the Cu-Sn compound. The thickness of the exfoliated r-TiO₂ thin film shown in Fig. 3 is in consistent with the ion range simulated with SRIM2008, which confirms that the aggregation of implanted He⁺-ions contributes to thin film exfoliation. EDS was used to identify the compositions of the heterostructure, as shown in Fig. 4, which clearly shows the staggered layers in the r-TiO₂ thin film heterostructure. The surface roughness inherently induced by the straggling of the implanted ions was smoothed using Ar⁺-ion etching. Finally, high quality single-crystalline r-TiO₂ thin film heterostructure was fabricated. Rocking curve, θ-2θ measurements with Cu Kα x rays were used to analyze the lattice properties of the exfoliated thin film, as shown in Fig. 5. As the x rays probe depth is limited, excluding the r-TiO₂ substrate, the detected diffraction peak at 2θ=62.75^o corresponding to (002) crystal plane, demonstrates the good crystalline property of our exfoliated r-TiO₂ thin film. The other two small peaks are attributed to SiO₂ and Cu-Sn compound. As the x-rays probe range exclude the r-TiO₂ substrate, so this (002) diffraction peak comes from r-TiO₂ thin film.

 

Fig. 3. Cross-section images of the fabricated r-TiO₂ thin film heterostructure detected by SEM.

Download Full Size | PDF

 

Fig. 4. The component elements in each layer of r-TiO₂ thin film heterostructure detected by EDS.

Download Full Size | PDF

 

Fig. 5. XRD results of the exfoliated r-TiO₂ thin film.

Download Full Size | PDF

Since the ion range of the implanted He⁺-ions is within 1μm, RBS/Channeling technique can be used to investigate the lattice damage in the ion-implanted r-TiO₂ sample. For demonstrating the lattice damage, the He⁺-ions backscattered from the target Ti and O atoms were collected. The backscattering yield of the ion-implanted r-TiO₂ sample with ion fluence of 8×10¹⁶ ions/cm², as well as the aligned and random spectra from a virgin r-TiO₂ sample detected by RBS/Channeling are shown in Fig. 6(a). The same minimum yields of around 2% at the surface region of both virgin and as-implanted r-TiO₂ samples indicate the good quality of our exfoliated r-TiO₂ thin film. In contrast with the aligned spectra, there is a clear bulge in the backscattering spectra of the as-implanted r-TiO₂ sample. It can be attributed to the disordered lattice induced by nuclear energy deposition during the ion-implantation. The damage profile, i.e., the relative number of displaced lattice atoms, was extracted from the RBS spectra by using a multiple-scattering dechanneling model based on Feldman’s procedure and applied for all target elements in the crystal [36,37]. The peak damage ratio of nearly 50% was detected at the depth of around 680 nm below the sample surface, which is consistent with the SRIM result shown in Fig. 6(b). This indicates that the density of defects (in the damage layer at the damage ratio of 50%) is high enough to trap the implanted He⁺-ions into bubbles or cracks, which facilitates layer splitting.

 

Fig. 6. RBS/Channeling spectra (a) and damage profile (b) of 200 keV He⁺-implanted r-TiO₂ at ion fluence of 8×10¹⁶ ions/cm². In (a), the aligned and random spectra from the virgin r-TiO₂ are shown for comparison. The distribution of the implanted He⁺-ions as simulated by SRIM is shown in (b) by the solid red curve.

Download Full Size | PDF

The number of displacements per atom (dpa) vs. the depth (z) can be calculated using the expression of dpa(z)=N_displ(z)×N₁/N₀, where N₁ is the ion fluence, N_displ is the number of displacements per incident ion and unit length (taken from SRIM 2008), and N₀ is the atomic density of r-TiO₂. The calculated dpa peak is 1.65 for the ion-implanted r-TiO₂ sample with ion fluence of 8×10¹⁶. This dpa number is for layer splitting in ion-implanted r-TiO₂.

To better understand the nature of the inner-structure, TEM measurements were done for the He⁺-implanted r-TiO₂ sample, and the results are presented in Fig. 7. All these TEM images are in cross-section geometry, and the sample surface is marked in the figure. The damage layer at depth of 680 nm below sample surface can be clearly seen in Fig. 7(a). We see that it is consistent with the thickness of exfoliated r-TiO₂ thin film, as well as the RBS result. Figures 7(b) - (e) show that the damage layer has a high concentration of He nanobubbles as well as dislocations induced by nuclear displacement, and the thickness of damage layer is about 80 nm, consistent with the RBS result. Figure 7(b) is a high-resolution TEM micrograph of the damage layer. The wavy-strings oriented along energetically preferred planes are resulted from the aggregation of He⁺-ions. The enlarged image in Fig. 7(c) show that it consists of He bubbles. One can see in Fig. 7(d) that the lattice structures on both sides of the string are dislocated and somewhat tilted, suggesting that these bubble strings lay on twin boundaries. He bubbles of around 4 nm size are also detected in the damage region, as shown in Fig. 7(e). Figure 7(e) demonstrates that, consistent with RBS results, in spite of the severely disordered lattice structure in the damage layer, it is not fully amorphous and unbroken lattice structure can still be clearly seen. Figure 7(f) shows that the lattice structure near implanted sample surface corresponding to the exfoliated thin film region is in good single-crystalline, indicating that the exfoliated r-TiO₂ thin film is almost perfect and can preserve good optical and physical properties.

 

Fig. 7. TEM images of the internal structure of the ion-implanted r-TiO₂ sample.

Download Full Size | PDF

4. Conclusion

Heterostructure composed of r-TiO₂ thin film on SiO₂ substrate with high refractive index contrast is fabricated by means of 200 keV He⁺-implantation at ion fluence of 8×10¹⁶ ions/cm² together with Cu-Sn bonding. The exfoliated r-TiO₂ thin film is investigated and shown to be in good quality, and the layer-splitting process induced by ion-implantation is investigated. The fabricated r-TiO₂ thin film heterostructure here should be useful in many applications, such as fabrication of optical microcavities with high Q in visible wavelengths.