Localized phase transition of TiO₂ thin films induced by sub-bandgap laser irradiation

Titanium dioxide (TiO₂) is an extensively studied metal-oxide semiconductor owing to its applications in chemistry, electronics, and medicine.^1–7 TiO₂ is an ideal candidate for applications in catalysis, hydrogen production, water treatment, and air purification.^8–12 Due to its high dielectric constant, large refractive index, resistivity, and ease of mass production, TiO₂ is widely used in solar cells, optically active coatings, memory devices, and microelectronic capacitors.^13–15 

TiO₂ is a wide bandgap semiconductor that exhibits different chemical and physical properties depending on its crystal structure.¹⁶ It is polymorphic with three distinct crystalline phases: rutile, anatase, and brookite. The crystal structures are built from networks of distorted TiO₆ octahedral units.¹⁷ Rutile has a tetragonal P4₂/mnm symmetry, whereas anatase is tetragonal I4₁/amd. Anatase shows enhanced catalytic properties^18–23 as the electrons in anatase behave as free electrons rather than polarons.^24–26 However, rutile is more thermodynamically stable.²⁷ 

The oxygen vacancy (V_O) has a low formation energy and in anatase TiO₂, V_O acts as a shallow donor.^28,29 V_O defects are known to be responsible for the n-type conductivity of anatase and are correlated with optical absorption in the visible range.³⁰ Calculations indicate that, although the thermal donor level (0/+) is near the conduction-band minimum, the optical transition threshold is ∼1 eV, which may account for the observed absorption.³¹ In the present study, a 532 nm green sub-bandgap laser was used to irradiate TiO₂ thin films under vacuum and ambient conditions. The laser photons are absorbed by defects, assumed to be V_O, resulting in electronic excitation of the treated region. This electronic excitation leads to a phase transition. Phase transformations of TiO₂ are affected by crystal strain and surface defects such as oxygen vacancies and Ti interstitials.^31–33 In the present work, laser irradiation under different ambient conditions resulted in different phase transitions. The ability to define micrometer-sized anatase and rutile phase regions could lead to enhanced photocatalytic properties^34,35 and improved solar conversion efficiency.³⁶ 

Anatase TiO₂ thin films of 300 nm thickness were deposited onto fused silica substrates by RF magnetron sputtering following an argon (Ar⁺) etch with 300 V etching bias. A gas mixture of 80% Ar and 20% O₂ and a pressure of 10 mTorr were used to carry out the sputter deposition at room temperature for 6.5 h. The TiO₂ target bias voltage was set to 450 V with a 0.8 nm/min deposition rate.

Laser irradiation was performed with a continuous wave (CW) optically pumped semiconductor laser (Coherent Verdi G series) with a 532 nm wavelength and a beam diameter of 2.25 mm. The laser beam was focused to a diffraction-limited diameter of ∼100 μm using a 200 mm focal length biconvex lens. The sample was mounted inside a chamber fitted with a quartz glass window and scanned in a raster pattern using a manual 2D (XY) stage to define regions of transitioned phases. Laser irradiation in vacuum was performed under dynamic vacuum conditions with a residual pressure of ∼2.2 × 10⁻⁵ mbar at room temperature. Laser irradiation in the air was then performed to define phase-transformed stripes on regions previously irradiated in vacuum. 1.8 W laser power in vacuum was necessary to trigger the anatase-to-rutile phase transition, while 3.5 W in air was required to cause rutile-to-anatase transition.

To investigate the phase transformation of the laser treated TiO₂ thin film, a Raman microscope (Renishaw inVia) was used to collect Raman spectra in a backscattering geometry. A He–Ne laser was used as the excitation source with 632 nm excitation wavelength at a magnification of 100×. The Raman spectrometer was adjusted carefully to ensure regions of the samples undergo similar testing conditions with a spatial resolution of about 1 μm. For signal processing, a Savitzky–Golay filter was used to smooth the spectra, followed by subtraction of the splined background.

To study the spatial distribution of phases within the laser-irradiated TiO₂ thin film, Raman measurements were performed using a Klar Mini Pro microscope. The microscope was equipped with a fiber-coupled 532 nm CW laser and an Ocean Insight QE Pro spectrometer. A total of 15 876 individual spectra were scanned for each 100 × 100 μm² area of the sample with a spatial resolution of 0.8 μm. The sample thickness of 300 nm was less than the depth resolution of both microscopes. The laser power at the sample was 9 mW. Despite the relatively high laser power density, the microscope laser did not cause phase transitions. The acquisition time was set to 5 s per point for the region laser irradiated in vacuum and 8 s per point for the region treated in air. The detected spectral region was 100–4000 cm⁻¹. Each spectrum was fit with eight Lorentzian peaks (144, 197, 236, 395, 436, 515, 608, and 636 cm⁻¹), known to be the typical Raman peaks of TiO₂ anatase and rutile phases (Fig. 1).

Anatase and rutile both have a tetragonal symmetry with six active Raman modes in anatase: A_1g + 2B_1g + 3E_g. The A_1g, B_1g, and E_g Raman modes are related to asymmetric bending, symmetric bending, and symmetric stretching vibrations of O–Ti–O bonds, respectively.³⁷ The three E_g modes have frequencies 144, 197, and 640 cm⁻¹. The strongest mode at 144 cm⁻¹ that corresponds to the symmetric lattice angular vibration is a characteristic peak of anatase TiO₂. Rutile exhibits four active Raman modes: A_1g + B_1g + B_2g + E_g with frequencies 143, 235, 445, and 608 cm⁻¹.³⁴ Modes E_g (445 cm⁻¹) and A_1g (608 cm⁻¹) are characteristic peaks of rutile TiO₂.³⁵ 

As shown in Fig. 2(a), laser irradiation in vacuum caused a visible change in the TiO₂ thin film. The irradiated region had a cellular structure [Fig. 2(b)] with feature sizes ∼10 μm. When this region was then irradiated in air [Fig. 2(c)], the material darkened and formed spherical regions ∼3.5 μm in diameter [Fig. 2(d)].

Figure 3 shows Raman spectra of TiO₂ that was untreated, laser irradiated in vacuum, and laser irradiated in air. Five Raman-active vibrational modes at frequencies 143, 197, 395, 516, and 636 cm⁻¹ confirm the anatase phase for the as-deposited, untreated region of the sample. Laser irradiation under vacuum causes a phase transition from anatase to rutile. Three Raman-active modes at 241, 436, and 607 cm⁻¹ confirm the phase transformation. The spectra also consist of a faint peak at 147 cm⁻¹. The broad peak around 241 cm⁻¹ is a specific feature of the rutile spectrum originating from strong second-order Raman scattering.³⁶ 

To transform the phase back to anatase from rutile, laser irradiation in the air was performed on the rutile region. The observed rutile-to-anatase transformation is remarkable because rutile is thermodynamically stable. Five active Raman modes at frequencies 147, 195, 400, 518, and 641 cm⁻¹ confirm the rutile-to-anatase phase transition. Peaks around 147, 400, 518, and 641 cm⁻¹ are all blue-shifted except the peak at 195 cm⁻¹. The slight shifts in vibrational modes are consistent with disorder due to residual strain and V_O concentration.^34,35,38

Micro-Raman mapping was performed under ambient conditions on a 100 × 100 μm² area. First, we examined the region that was irradiated in vacuum. Figure 4(a) shows a Raman map of the E_g mode of anatase TiO₂ (144 cm⁻¹) in red and the A_g mode of rutile TiO₂ (608 cm⁻¹) in green. The spatial Raman intensity distribution shows that anatase indeed transformed to rutile but there are a few residual anatase regions. The strongest rutile signal comes from the cell edges. Some regions in the cell interiors showed a weak, broad spectrum consistent with amorphous TiO₂.

Figure 4(b) shows the Raman spectra of regions A–D. Regions A and B are on the boundary of a cell while region C is inside. Regions A–C confirm the localized phase transition from anatase to rutile due to laser irradiation under vacuum. Region D, which is in the interior of a cell, shows no Raman peaks, characteristic of amorphous TiO₂.

Laser irradiating the rutile region in air resulted in a rutile-to-anatase phase transition. Figure 5(a) shows a Raman map of rutile and anatase modes. The spatial Raman intensity distribution clearly shows that the anatase signal comes from spherical regions. The Raman-active modes disappear as we move from the center toward the edge of the structures. Our results suggest that laser irradiation of the rutile region in air produces anatase spheres surrounded by amorphous material.

Figure 5(b) shows the Raman spectra of regions A–D. Regions A and B are inside a sphere while regions C and D are located outside. Region A is at the center while B is near the edge and both show the Raman spectra of anatase TiO₂. Region C is adjacent to a cluster and displays a weak anatase peak around 147 cm⁻¹ while region D is far away and appears to be amorphous.

Prior work on laser irradiation of TiO₂ nanostructures demonstrated that the atmosphere affects the crystal structure. Irradiation with a high-intensity laser source in the air increases the crystallinity of anatase TiO₂ nanoparticles in the air³⁹ while irradiation under vacuum, even with relatively low laser power density, degrades the crystalline phase to an amorphous phase.⁴⁰ In both cases, the underlying mechanism is athermal and surface oxygen plays a vital role in defining the crystalline structure.^39–41 For the anatase-to-rutile transition under dynamic vacuum, optical excitation generates surface defects, while in air, surface defects are annihilated due to interactions with the surrounding oxygen.^42–45 

In the present study, laser irradiation under vacuum results in an amorphous state and stimulates the desorption of oxygen molecules from the surface. As a result, there is a rapid increase in the oxygen vacancy concentration.⁴⁶ These vacancies act as nucleation sites for the stable rutile phase.^47,48 While the anatase-to-rutile transition is relatively straightforward, the reverse transition appears to go against thermodynamic stability. However, Raman mapping revealed that the anatase crystals are embedded in an amorphous matrix. It is, therefore, conceivable that laser irradiation (1) amorphizes the rutile regions and (2) anatase crystals nucleate within the amorphous material.

A schematic diagram for this process is shown in Fig. 6. Irradiating the rutile region in air amorphizes the area and leads to localized oxygen adsorption. An oxygen molecule is adsorbed at a surface oxygen vacancy site and stabilizes the surface structure by associating two adjacent Ti sites.⁴⁹ This facilitates the nucleation of anatase TiO₂. Our model is consistent with prior work on amorphous TiO₂ thin films, which can transform to anatase after annealing.⁵⁰ The fundamental idea is that the laser energy goes into amorphizing the material. The phase of the crystal regions formed within the amorphous matrix depends on the ambient conditions: high(low) oxygen partial pressure results in anatase(rutile).

In conclusion, anatase thin films transform to a mixture of rutile and amorphous material when irradiated by laser light in vacuum. Subsequent irradiation in air leads to an amorphous/anatase mixture. This latter process is unusual because the rutile phase is thermodynamically stable. The key to the rutile-to-anatase transition is likely the formation of an intermediate amorphous phase. In the presence of oxygen, anatase crystals are nucleated in the amorphous matrix.

Raman mapping revealed that the rutile phase has a cell-like structure while the anatase phase occurs in spherical regions. The reason behind these different geometries is not clear but may be related to the laser power distribution during irradiation. Along these lines, the highly focused Raman laser spot does not induce a phase transition, which suggests that the specific beam profile parameters are important.

This work was supported by the National Science Foundation (NSF) under Grant No. DMR-2109334. Violet Poole is an employee of Klar Scientific and Matthew McCluskey owns equity in the company. Yi Gu is an advisor for Klar Scientific and has equity interest in the company.

The data that support the findings of this study are available from the corresponding author upon reasonable request.