Visible-light-accelerated oxygen vacancy migration in strontium titanate

Strontium titanate is a model transition metal oxide that exhibits versatile properties of special interest for both fundamental and applied researches. There is evidence that most of the attractive properties of SrTiO3 are closely associated with oxygen vacancies. Tuning the kinetics of oxygen vacancies is then highly desired. Here we reported on a dramatic tuning of the electro-migration of oxygen vacancies by visible light illumination. It is found that, through depressing activation energy for vacancy diffusion, light illumination remarkably accelerates oxygen vacancies even at room temperature. This effect provides a feasible approach towards the modulation of the anionic processes. The principle proved here can be extended to other perovskite oxides, finding a wide application in oxide electronics.

Scientific RepoRts | 5:14576 | DOi: 10.1038/srep14576 light. Without illumination, a bias voltage up to ± 200 V produces negligible effects on the structure of STO. This can be ascribed to the low mobility of oxygen vacancies (~8 × 10 −12 cm 2 /Vs at room temperature 20 ).As recently reported, the structural deformation only occurs for STO accompanying the electro-migration of V O s 6 . It is easy to calculate the time required for the V O s to drift out of the interfacial layer, and it is ~6200 s under the gate bias of − 200 V if layer thickness is ~2 μ m, well beyond the time window of the ϑ -2ϑ scanning (~180 s). Aided by light, however, a gate bias of − 200 V is high enough to cause sizable structural deformation. As shown in Fig. 1b, an obvious shoulder appears beside the main reflection when illuminated by a light of P = 100 mW (λ = 532 nm), indicating a lattice expansion along [001] axis. Since the unchanged phase can be clearly seen, the deformed phase could be much thinner than the penetration depth of x-ray in STO (~6 μ m), appearing in an interfacial layer. According to reciprocal space mapping of the 103 reflection, fascinatingly, this new phase has an elongated c-axis lattice constant but a unchanged a-axis one (Fig. 1c). In contrast, positive biases produce no effect on lattice even aided by illumination, which means that the field-induced lattice expansion only appears underneath anode (Fig. 1d).
For a thorough exploration of this new gating effect, we measured the XRD spectra for various combinations of V G and P. Fixing P to 100 mW while varying V G , according to Fig. 1d, lattice expansion emerges when V G exceeds − 40 V. From V G = − 40 V to − 200 V, the lattice constant changes from 3.905 Å to 3.922 Å (marked by red arrows), increasing by ~0.4%. While fixing V G to − 200 V but tuning P, sizable gating effect appears when P = 5 mW (Fig. 1e), i.e., the structural distortion caused by the gating effects can be accelerated remarkably even in low light power. Quantifying the lattice expansion in Fig. 1b by a shaded area, we obtain a quantitative description of the gating effects in the V G − P plane (Fig. 1f), Photo-excitation-modulated lattice expansion. To reveal the effect of light illumination, the evolution of the 002 reflection with time is recorded when STO is exposed to different lights in the presence of electrical bias. Figure 2a shows the XRD spectra for two typical cases of P = 0 and 30 mW (λ = 532 nm, V G = − 300 V). Figure 2b is a summary of the data for different wavelengths (P = 30 mW). As shown in Fig. 2b, the acceleration of short wavelength light to gating effect is prompt: Lattice expansion occurs right upon the application of V G and completes within about 10 minutes. Increasing λ leads to an inclining of the c(t) curve, indicating a slowdown of the gating effect. To get a quantitative description of the wavelength effect, in Fig. 2c we present, in semi-logarithmic scale, the dependence of dc/dt on λ , derived from the c(t) curve in the initial stage of structural deformation. A simple calculation shows that dc/dt is ~0.0047 Å/min for λ = 532 nm and ~0.0001 Å/min for λ = 980 nm, depressing nearly 50-fold.
The distinctive dc/dt− λ relation implies that the acceleration of the gating effect may be ascribed to photo-excitation rather than thermal activation. This is understandable since the photon energy used in the present experiments is well beyond the range of phonon energy. In fact, a direct effect of photo-excitation could be observed on photoconductivity. Since STO is insulating, its photoconductivity cannot be directly measured with visible light. We choose a LaAlO 3 /SrTiO 3 interface with a 3-unit-cell-thick LaAlO 3 top layer as a sample. As well established 25 , the interface will be insulating when the LaAlO 3 layer is thinner than 4 unit cells for lack of mobile sheet charge carriers. However, we found that this insulating interface can be driven into conductive state by illuminating. As exemplified by Fig. 2d, from λ = 532 nm to 980 nm (P = 30 mw), the sheet conductance changes from G ≈ 2 × 10 −5 Ω −1 to ~6 × 10 −8 Ω −1 . It is obvious that light illumination generates extra charge carriers by exciting the in-gap states of STO (photon energy is lower than the band gap of STO), leading to the growth of sheet conductance. When the quantum efficiency for photo-excitation decreases with the increase of λ , G accordingly decreases. Surprisingly, the G-λ relation mimics the dc/dt-λ dependence very well, even in detailed features (Fig. 2c). This result strongly suggests that it is the photo-excitation of the in-gap states of STO that accelerates the gating effect. We noted that the light illumination induced conductance enhancement has been observed previously 26,27 . These works revealed the photo-excitation-generated effect. According to our data, we believed that photo-excitation of the in-gap state of STO, which gives rise to photoconductivity as shown in Fig. 2d, causes a transition of V O • to V O •• + e, accelerating vacancy diffusion, as will be discussed later.

Illumination-accelerated oxygen vacancies.
A further question is how photo-excitation accelerates the gating effect. As suggested by Hanzig et al. 6 , the structural deformed phase forms while oxygen vacancies drift along electric field. In this scenario, photo-excitation may affect the gating effect through accelerating vacancy diffusion. To reveal the effect of light illumination on oxygen vacancies, transient current, i(t), through an electrically biased STO is studied in the presence (absence) of light illumination. Figure 3a is a schematic experimental setup. Figure 3b,c display the transient current as a function of time, recorded by fixing V G to − 80 V at a temperature between 328 K and 408 K. With the increase of t, the transient current undergoes first a low to high and then a high to low transition, leaving a current peak at τ p . This feature is observed in all i(t) curves acquired at different temperatures. Comparing with the data without light, we found that light illumination causes an obvious left shift of current peak. This phenomenon is particularly evident when temperature is not very high. For example, at a temperature of 328 K, τ p is ~2100 s in light but ~24970 s without light, i.e., the time required to reach the current peak has been reduced by more than one order of magnitude by light illumination. This result implies an acceleration of vacancy diffusion. As well established, the transient current is carried by V O s, and the transient current gets a maximum when the oxygen vacancies underneath anode arrive at cathode 6,28 .
As predicted by space-charge-limited (SCL) theory, the peak position of i(t) has a close relation to the mobility, μ , of oxygen vacancies following the equation 28,29 .
where d is the thickness of the sample. If the transient current is exclusively due to ionic charges, we will have the following equation where E A is the activation energy for oxygen vacancy diffusion, k B is the Boltzmann's constant, T is the temperature, q is the charge of oxygen vacancy, and γ is a pre-exponential factor. Here the Einstein relation between mobility and diffusivity, μ = qD/k B T, has been adopted for the derivation of Eq. (2). Based on Eq. (1), μ as a function of temperature can be deduced from the data in Fig. 3b,c. A direct calculation shows that the mobility is, for example, for T = 328 K, ~3.8 × 10 −10 cm 2 /Vs without light and ~1.8 × 10 −9 cm 2 /Vs with light. Extrapolating the data in Fig. 3b,c to room temperature (296 K), we obtain the vacancy mobility of ~1.8 × 10 −11 cm 2 /Vs without light, which is comparable to the value deduced from ref. 20 (~8 × 10 −12 cm 2 /Vs).
Remarkably, the μ-T relation is well described by Eq. (2). This, in addition to confirming the ionic origin of the transient current, reveals the activation nature of the charge transport process. From the slope of the semi-logarithmic μT-T plot in Fig. 3d, the activation energy can be deduced. It is ~0.84 eV without light and ~0.56 eV in a light of, for example, 100 mW and 532 nm. Light illumination reduces E A by ~34%. These results indicate that, indeed, light illuminating mobilizes oxygen vacancies. Here we emphasize that the activation energy may be continuously depressed by illumination, as suggested by the progressive left shift of the current peak as P grows (Fig. 3e).
In fact, the diffusion of oxygen vacancies in STO has been intensively studied, and the activation energy obtained by different groups concentrated on two distinct categories around 0.6 eV and 1 eV (Fig. 3f) 19,20 . It is instructive to note that the activation energies deduced here (0.56 eV and 0.84 eV) also fall into these two groups (Fig. 3f). As will be discussed later, this may be an indication for illumination-induced changes in the V O -complexes that are believed to widely exist in STO, and this could be the underlying reason for the mobilization of V O s.

Discussions
The above experiments suggest that light illumination accelerates field-induced lattice expansion via accelerating oxygen vacancy diffusion (Figs 1 and 3), and this acceleration process occurs accompanying the photo-excitation of in-gap states of STO (Fig. 2). A further question is how oxygen vacancy diffusion related to photo-excitation.
In general, it is believed that oxygen vacancies are doubly ionized at high temperatures, existing in the form of isolated V O •• s 30 . According to Cordero 19 , the binding energy of a V O -complex is ~0.2 eV. A direct calculation shows that the probability for a V O -complex to be disassembled at 1000 °C is about 400-fold larger than that at 25 °C. At low temperatures, however, a singly ionized state could be favorable. By analyzing the relation between carrier concentration (n) and δ for SrTiO 3-δ , Moos et al. 31 declared that oxygen vacancies were singly ionized near room temperature, i.e., each V O donates one charge carrier in the meantime trapping an electron. In their experiments, a well linear n− δ relation was observed from δ = 0.03 down to 0.01, without any signatures of inflection.
All of our experiments were conducted near room temperature. In this case, the content of singly ionized vacancies could be considerable. As well documented 15,16,19,30 , singly ionized vacancies prefer to form V O -complexes by sharing their electrons with the two closest titanium atoms, yielding in-gap states (Fig. 4a) 15,19,20 . Since the photon energy used here is much lower than the band gap of STO, photo-illumination will only affect the V O -complexes, exciting the shared electrons to the conduction band of STO. Without the shared electrons, however, the V O -complexes become unstable, disassembling into isolated V O •• s (Fig. 4b). In this manner, photo-excitation accelerates the V O diffusion thus the gating effect. This inference is in consistent with the observed illumination-induced reduction of E A shown in Fig. 3d,f. To summarize, the present work reveals the close relation between photo-excitation and oxygen vacancy diffusion, paving the way towards the tuning of the anionic processes which could be important for oxide electronics.

Methods
Sample fabrication. Samples are (001)-orientated STO substrates (5 × 5 × 0.5 mm 3 ) with an atomic level flat top surface. As electrodes, two 30-nm-thick Ti layers were deposited respectively on two surfaces through magnetron sputtering in an Ar atmosphere of 0.5 Pa. For comparison study, a LaAlO 3 / SrTiO 3 sample with a LaAlO 3 overlayer (3 unit cells in thickness) was prepared using the pulsed laser Scientific RepoRts | 5:14576 | DOi: 10.1038/srep14576 (248 nm) ablation technique. In the deposition process, the temperature was kept at 800 °C and the oxygen pressure at 10 −5 mbar. The fluence of the laser pulses was 0.7 Jcm −2 , and the repetition rate was 1 Hz. The layer thickness of LaAlO 3 was in situ monitored by the RHEED (reflected high energy electron diffraction) technique. After deposition, the sample was annealed in 200 mbar of O 2 at 600 °C for one hour, and then cooled to room temperature in the same oxygen pressure.
Measurements. The structure of SrTiO 3 was measured in the presence of a gate voltage and light illuminating, by a Bruker diffractometer (D8 Discover, Cu K α radiation) with the x-ray being sparallelized and monochromatized by an asymmetric Ge 2202-Bounce monochromator. A transverse electrical field was applied to the bottom electrode of SrTiO 3 , while the top electrode was grounded. Lasers adopted in the present experiments have a wavelength between 532 nm and 980 nm. The spot size of the laser beam is 4 mm 2 , focusing on the regions where the x-ray was reflected. Ultrasonic Al wire bonding (20 μ m in diameter) was employed for electrode connection for the LaAlO 3 /SrTiO 3 sample. The four welding spots were well aligned, and the separation between neighbouring spots was ~0.4 mm. Laser beam focused on the area between two inner leads. The applied current for conductance measurements was 1 μ A. All data were acquired at ambient temperature.