A. Sample preparation

High purity silica slides (20 mm×20 mm×1 mm) were implanted by 20 kV Ti ions to 1.5×10¹⁷ ions/cm² on an implanter of metal vapor vacuum arc (MEVVA) ion source. The samples were kept rotating in a horizontal plane during ion implantation, with the sample holder being cooled by circulating water. The implanted samples were annealed under O₂ atmosphere in a conventional tube furnace. Group 1 of the samples were annealed for 2 h at 500 ℃, 600 ℃, 700 ℃, 800 ℃, 900 ℃ and 1000 ℃. Group 2 were annealed at 800 ℃ for 2, 4 and 6 h.

B. Sample characterization

Optical absorption spectra of the samples were measured on a UV-vis-NIR dual-beam spectrophotometer (Varian Cary5000) in wavelength range of 200–800 nm. Surface morphologies of the implanted samples before and after thermal annealing were examined by using a scanning electron microscopy (SEM, FEI Sirion). Raman scattering spectra were measured using a micro-Raman microscope (Jobin-Yvon LabRAM HR) equipped with a cooled CCD detector under a backscattering geometry to identify the crystalline phase of TiO₂. The excitation source was the blue line (488 nm) of an Ar⁺ laser powered at 5 mW, with its scan density varying from 100 to 1200 cm^-1. Microstructural characterization of the as-implanted and annealed samples was performed on a JEOL 2010 (HT) transmission electron microscope (TEM) operated at 200 kV. X-ray photoelectron spectroscopy (XPS) analysis was done on a Kratos XSAM800 XPS system with Mg K_α (1253.6 eV) as the radiation source under a vacuum of 6×10^-7 Pa.

A. Optical absorption spectra of Ti ions implanted samples versus annealing temperature

Figure 1 shows optical absorption spectra of the as-implanted sample and the samples annealed for 2 h at 500–1000 ℃. In the absorption spectra of the as-implanted sample, the background absorption increases drastically in the UV and visible regions. This is due to absorption of implantation-induced point defects and the Ti nanoparticles formed in silica substrate. Annealed at 500 ℃, the background absorption reduces drastically due to the defect annihilation by thermal annealing. Also, an abrupt absorption edge appears, due to presence of anatase TiO₂ [19, 20], which was formed upon post-implantation annealing at 500 ℃. For all the annealed samples, the spectra present a trend that the absorption edges shift to longer wavelength with the increase of annealing temperature, implying an increase in mean sizes of the TiO₂ nanoparticles. It is interesting to note that the sample annealed at 1000 ℃ differs obviously from other annealed samples in spectrum with a greater shift of absorption edge to longer wavelength. This remarkable shift of absorption edge suggests that phase transformation of TiO₂ from anatase to rutile may occur at 1000 ℃. Such shifts of the maximum absorption and absorption edge induced by the phase transformation are similar to the results in Refs. [20, 21].

Fig. 1.Full-screen View

(Color online) Optical absorption spectra of samples implanted to 1.5×10¹⁷ Ti ions/cm² and annealed at 500–1000 ℃ for 2 h.

B. SEM images of Ti ions implanted samples versus annealing temperature

Surface morphologies of the samples were investigated by SEM. Figure 2 shows the SEM images of the Ti ions implanted samples annealed for 2 h at 500, 600, 700, 800, 900 and 1000 ℃.

Fig. 2.Full-screen View

SEM images of the samples implanted to 1.5×10¹⁷ Ti ions/cm² and annealed for 2 h at 500–1000 ℃.

On the whole, the surface morphology changes obviously for the samples annealed at different temperatures. At 500 ℃, some small-sized nanoparticles began to appear, indicating the TiO₂ nanoparticles formed at 500 ℃. At 600 ℃, a smooth and homogenous film was formed, with many near-spherical nanoparticles distributed compactly on the substrate surface. The sizes of TiO₂ nanopartices increase with the annealing temperature, with the mean diameter of particles being 25.3, 37.6 and 45.8 nm at 600 ℃, 700 ℃ and 800 ℃, respectively. At 900 ℃, however, the TiO₂ nanoparticles in the surface aggregated and formed cross-linked particles of about 61.4 nm in diameter, as shown in Fig. 2(e). The surface morphology underwent a big change at 1000 ℃, and the TiO₂ nanoparticles in the substrate surface became embedded and undistinguished, as shown in Fig. 2(f).

Interestingly, evolution of surface morphologies of the samples as a function of annealing temperature in Fig. 2 is well related with the results from optical absorption spectra (Fig. 1). The increase in both the particle size and crystal quality as a function of annealing temperature leads to the absorption edge shift towards longer wavelength. The abrupt shift of absorption edge observed at 1000 ℃ is just a result of structure change characterized by SEM, and the change may be due to the phase transformation from anatase to rutile.

C. Formation and phase of TiO₂ nanocrystals: Raman scattering spectroscopy and XPS measurements

The Raman spectroscopy measurements were made to identify and analyze the formation and phase of TiO₂ nanocrystals after thermal annealing. Figure 3 shows Raman spectra of the bare silica, the as-implanted sample and the Ti ions implanted samples annealed for 2 h at 500–1000 ℃. Some Raman peaks of SiO₂ can be observed in the bare silica sample. However, Raman spectrum of the as-implanted silica became almost featureless due to extensive structural damage induced by ion implantation. The structure of silica is partially recovered by subsequent thermal annealing, and the Raman peaks of SiO₂ in the annealed samples can be observed again. A Raman peak centered at about 140 cm^-1 appears at 600 ℃. This can be assigned as E_g mode of TiO₂ in anatase phase [22, 23]. The Raman peak height increases with temperature from 600 ℃ to 900 ℃, indicating the increased crystallinity and amount of the anatase phase TiO₂. However, for the sample annealed at 1000 ℃, the spectrum shape changed, with three Raman peaks centered at 605, 440 and 230 cm^-1. These can be ascribed to the two-phonon scattering of A_1g and E_g modes, and the second-order effect of the rutile phase TiO₂, respectively [24]. Thus, the anatase-to-rutile phase transformation taking place at the high temperature is confirmed.

Fig. 3.Full-screen View

(Color online) Raman spectra of bare silica, the 1.5×10¹⁷ Ti ions/cm² as-implanted sample, and annealed for 2 h at 500–1000 ℃.

The formation and evolution of the anatase phase in the annealed samples is consistent with the evolution of optical absorption spectra in Fig. 1 and SEM in Fig. 2. The increase in size and crystal quality of the nanoparticles as a function of annealing temperature leads to the red-shift of the absorption edge and the intensification of the E_g Raman mode. However, the absorption edge of TiO₂ in the sample annealed at 500 ℃ was observed, while the crystalline phase in the same sample was not detected by Raman spectroscopy. It is well known that optical absorption spectrum usually gives average structural information, whereas Raman scattering spectrum as a local probe is sensitive to crystallinity and microstructures of materials. That the Raman peaks could not observed implies that annealing at lower temperatures produced not enough TiO₂ nanocrystals, which can be verified by the SEM image in Fig. 2(a). Also, the nanocrystals are not of good crystallinity, though the absorption edges of TiO₂ were confirmed by optical absorption measurement.

By carefully analyzing the change of E_g modes centered at 140 cm^-1 of the anatase phase TiO₂ as a function of annealing temperature, we found that the intensity, frequency and linewidth (FWHM) of the E_g Raman mode are strongly dependent on the annealing treatment. The linewidth of Raman peaks indicates that the TiO₂ nanofilms are of high crystallinity. The intensity of E_g Raman mode increases with annealing temperature, accompanied by a downshift in frequency and a decrease in linewidth. The sample annealed at 900 ℃ exhibits a large downshift in frequency and an obvious decrease in the linewidth. A similar behavior of frequency shift and linewidth decrease of E_g Raman mode was observed in TiO₂ nanoparticles fabricated by other methods [24-28].

Mechanisms proposed to interpret this behavior include the phonon confinement effect [24, 29], internal stress/surface tension effects [28, 30], and nonstoichiometry due to oxygen deficiency [25, 26]. The pressure effect mechanism usually leads to large frequency shifts, and Raman measurements as a function of the external pressure shows that the Raman modes undergo a downshift in frequency with increasing pressure [31]. For instance, the pressure effect induced by the surrounding particles or the interface stress has been successfully used to elucidate the frequency shift of the Raman peaks observed in TiO₂ and PbTiO₃ nanoparticles [32]. The size-induced pressure effect may act in a manner similar to an external pressure on the TiO₂ nanoparticles. The smaller the nanoparticles are, the larger the contribution of pressure is, hence the frequency downshift of the E_g annatase mode [28].

In our experiment, the size of TiO₂ nanoparticles increases with annealing temperature, and the pressure releases gradually with heat treatment, which can lead to the frequency up-shift (towards to higher frequency) of the E_g mode according to the size-induced pressure effect. However, instead of up-shift, a frequency downshift with increasing annealing treatment was observed, implying that the pressure effect can be ruled out. In addition, according to the phonon confinement effects, the increase in crystallite size can cause downshift in frequency and decrease in linewidth of the E_g Raman mode, thus the phonon confinement effects should be taken into account to explain the change of E_g Raman mode.

Also, non-stoichiometry observed in nanophase TiO₂ samples prepared by other methods [24-26] plays an important role in frequency shift of Raman peaks, in which the frequency downshifts with decreasing oxygen deficiency. Although the TiO₂ samples used in this work are prepared by Ti-ion implantation and subsequent annealing in an oxygen atmosphere, oxygen deficiency, as the most common form of nonstoichiometry in TiO₂, can not be ruled out. Moreover, the downshift in frequency of the E_g annatase mode with increasing annealing temperature (the decrease of oxygen deficiency) is well consistent with the results reported by other authors [24-26]. The oxygen defiency decreases with increasing annealing temperature. Therefore, the nonstoichiometry should be taken into account in explaining changes in the E_g Raman mode in our TiO₂ films.

To further confirm the formation of TiO₂ nanofilms and measure the atomic ratio of Ti and O, the XPS spectra of the implanted samples annealed at 800 ℃ was measured (Fig. 4). The peaks centered at 458.7 eV and 464.5 eV binding energies correspond to the 2p 3/2 and 1/2 core levels of Ti⁴⁺, and the peaks centered at 530.0 eV and 532.6 eV correspond to the binding energies of Ti-O and Si-O, respectively, agreeing well with the values reported by other authors [3]. The atomic ratio of Ti and O was calculated as 1:1.82, proving the nonstoichiometry of TiO₂. The color of the annealed samples changes from light-blue to almost white at 900 ℃, indicating a decrease in oxygen vacancies [30] and a change of films thicknesses according to interference phenomena in thin films.

Fig. 4.Full-screen View

(Color online) XPS Ti 2p (a) and O 1s (b) core-level spectra of the 1.5×10¹⁷ Ti ions/cm² implanted sample and annealed at 800 ℃.

D. Mechanism for the formation of TiO₂ nanofilms: TEM Characterization

To figure out growth process and formation mechanism of the TiO₂ nanofilms, the TEM study on microstructure of the samples was carried out. Figure 5 shows the cross sectional TEM images of the as-implanted sample and the sample annealed at 800 ℃ for 2 h. A dark layer with tiny Ti nanocrystals below the surface was observed in the as-implanted sample (Fig. 5(a)). After annealing at 800 ℃ for 2 h, a nanofilm in thickness of 10 nm was formed on the sample surface (Fig. 5(b)). The selected area electron diffraction (SAED) pattern confirms the formation of an anatase phase TiO₂ nanofilm. Meanwhile, the Ti atoms in the silica substrate aggregate into larger nanocrystals according to the SAED pattern. This indicates that during annealing in O₂ atmosphere, the implanted Ti atoms diffused out to the silica surface and were oxidized into TiO₂ nanoparticles. Meanwhile, the embedded Ti nanocrystals form larger ones due to the Ostwald ripening effect.

Fig. 5.Full-screen View

Cross sectional TEM images of the 1.5×10¹⁷ Ti ions/cm² as-implanted sample (a) and sample annealed at 800 ℃ for 2 h(b).

The solid phase growth process of TiO₂ nanofilms by ion implantation and subsequent thermal oxidation is similar to the results reported on Zn-implanted silica and sapphire to form high-quality ZnO nanofilms on the substrate surface [14-16]. The formation of metal oxide semiconductor by metal ion implantation and subsequent thermal oxidation was due to out-diffusion of the implanted Ti ions to the substrate surface, where they were oxidized into TiO₂.

E. Growth of the TiO₂ nanofilms: Influence of the annealing time

The annealing time was increased to 4 h and 6 h, so as to further study the growth process and formation mechanism of the TiO₂ nanofilms. The cross sectional TEM images of the Ti ions implanted samples annealed at 800 ℃ for 4 h and 6 h are shown in Fig. 6. In Fig. 6(b), almost all the Ti atoms diffused out of the substrate to form a thicker TiO₂ nanofilm with a thickness of 50 nm, and the TiO₂ nanocrystals increased greatly in size. It should be mentioned that the TiO₂ nanofilms have a slow growth rate of about 0.15 nm/min to ensure the high crystal quality and high density of the formed nanofilms. Figure 7 shows SEM images of the samples annealed at 800 ℃ for 4 h and 6 h. The film of 6-h annealing has flatter film surface and larger nanocrystal size than the film of 4-h annealing.

Fig. 6.Full-screen View

Cross sectional TEM images of the 1.5×10¹⁷ Ti ions/cm² implanted samples and samples annealed at 800 ℃ for 4 h (a) and 6 h (b).

Fig. 7.Full-screen View

SEM images of the 1.5×10¹⁷ Ti ions/cm² implanted samples annealed at 800 ℃ for 4 h (a) and 6 h (b).

Combining the TEM with SEM results, annealing time was expected to affect the surface morphology, structure and thickness of the formed TiO₂ nanofilms when the samples were annealed at 800 ℃. Thickness of TiO₂ nanofilms increases with annealing time until all of the Ti atoms diffuse out of the substrate to form a thicker TiO₂ nanofilm. As discussed above, the absorption edge and phase transformation of TiO₂ are annealing temperature dependent, and the thickness of the TiO₂ nanofilms is annealing time dependent.