A bell-jar type reactor with the base pressure under 1 × 10⁻⁵ torr by a combination of diffusion pump and a rotary pump was used for LPCVD of titanium-oxide. Titanium tetra-iso-propoxide (TTIP, Ti(O-i-C₃H₇)₄: 99.7% purity) was used as source gas after purification in vacuum. Details of the apparatus configuration and the purification sequence of TTIP were already shown elsewhere [9] . High purity hydrogen gas (99.99999%) was simultaneously introduced into the reactor through an individual gas inlet during the deposition. In the case of the deposition in O₂, high purity oxygen gas (99.9999%) instead of H₂ was introduced into the reactor. The gas sup- ply ratio of TTIP/H₂ or TTIP/O₂ were controlled by monitoring the reactor pressure using Shultz gage when TTIP and H₂ or O₂ were introduced into the reactor, in which sensitivity of the ion gage for H₂ and O₂ was taken into account to determine the gas supply ratio to TTIP. Substrate of 1 mm-thick quartz plate with optically flat surface was mounted on a substrate holder after chemical cleaning. Temperature of the substrate holder on resistive heater was monitored by K-type thermo-couple and controlled by PID-system.

Thickness of the layer was checked by a surface profiler (Veeco, DEKTAK150). Resistivity was evaluated by Van Der Pauw (VDP) method using symmetric four ohmic contacts of Indium-dots, in which the temperature was varied from 10 to 300 K in vacuum and dark by a cryogenic system (Janis Research, CCS-150). Crystallographic behavior was examined by θ - 2θ X-ray diffraction (RIGAKU: RAD-C) using CuK_α. UV-Vis optical transmission spectra were obtained by UV-Vis spectrometer (OCEAN OPTICS: USB-2000) using a light source of Halogen lamp.

Figure 1(a) shows Arrhenius plot of TiO₂ deposition rate in O₂ (open-circle) and H₂ (solid-circle) ambient for the deposition temperature, in which supply rate of the gas to TTIP was kept at 1. The deposition rate of the layer deposited in O₂ (TiO₂/O₂) and in H₂ (TiO₂/H₂) was increased with the temperature and then saturated above 360˚C. At such low temperatures, the layers were poly-crystallized into anatase-phase as shown by XRD spectra in Figure 1(b), in which the spectra of TiO₂/O₂ and TiO₂/H₂ deposited at 360˚C is shown by black-line and red-line respectively. In the both spectra, diffraction peaks originated from (101) and (200) anatase-TiO₂ were observed but not any peaks due to another phase such as rutile and brookite. The diffraction peak intensities of the TiO₂/H₂ layer were weaker than that of the TiO₂/O₂ layer but the peak angle shift was not observed in the spectrum. Activation energy of 138 kJ/mol for the deposition of TiO₂/O₂ obtained the result in Figure 1(a) showed that TTIP-dissociation feature was scarcely influenced by O₂ because the energy was similar to CVD by TTIP single precursor [10] . In this case, it can be recognized TTIP is thermally dissociated then anatase-TiO₂ is formed with propene and isopropanol at such low temperatures as follow reaction [11] .

,

where it is expected Ti=O bonds are formed on the deposition surface [12] . Further, it is considered the surface Ti=O bonds act as adsorption-site of the precursor and enhance the dissociation of TTIP as depicted in Figure 2(a). In contrast, the activation energy for TiO₂/H₂ was obtained as 230 kJ/mol. Previously, Shmakov et al. demonstrated quantum chemical calculations using B3LYP for monomolecular dissociation of TTIP and suggested

the activation barrier to form Ti(OC₃H₇)₃(OH) and propene via O-H-CH₂ interaction in isopropyl group of TTIP is 56.9 kcal/mol (238 kJ/mol) [13] . The calculated value is in good agreement with the activation energy for the deposition TiO₂/H₂, however, very high as compared to that for TiO₂/O₂ and the previous report using TTIP single precursor. It can be recognized the large different of the activation energy is owing to hydrogen at the deposition surface as shown in Figure 2(b). When H₂ gas is simultaneously supplied with TTIP, methyl-group of TTIP cannot be allowed in the reaction to Ti-O or Ti on the deposition surface, because the surface consists of Ti=O is terminated by hydrogen. Therefore, the adsorption of the precursor is achieved after monomolecular dissociation to form Ti(OC₃H₇)₃(OH), then the adsorbed species is dissociated to TiO_x. It is noted Ti-H is probably remained in the deposited layer and the hydrogen is partially desorbed by thermal dissociation. In the scheme of the hydrogen-desorption, it is expected Ti⁴⁺ is reduced to Ti³⁺ and resulted in generation of donor- defect. Figure 3 shows dependence of resistivity of TiO₂/H₂ on the deposition temperature, in which the layers were deposited with the thickness about 200 nm and the resistivity was evaluated by VDP method at room- temperature. Although all of TiO₂/O₂ layer was with high resistivity above 100 Ω cm beyond the evaluation limit of our system, resistivity of the TiO₂/H₂ layer deposited in H₂ was decreased with the deposition temperature and then gradually increased above 360˚C. The activation energy of 244 kJ/mol obtained below 350˚C suggested that donor concentration was increased with the temperature. Recently, it was reported that hydrogen acts as donor in TiO₂ by experimental results for anatase-TiO₂ powder (Degussa P25) treated by atomic-hydrogen [14] and molecular orbital theoretical results [15] , in which hydrogen was found to be bound to lattice oxygen in TiO₂. In this work, it is however considered that the door species in the layer was not the hydrogen bound to O, because the absorption due to Ti-OH (Chemisorbed OH) which was identified in IR-spectrum of amorphous Ti- O_x layer deposited by plasma-assisted deposition at room-temperature [16] could not be observed in the TiO₂/H₂ layers. On the other, Nakamura et al. studied reduction treatment of anatase-TiO₂ powder by radio-frequency hydrogen-gas plasma at 400˚C to enhance the visible light activity for NO removal [17] . In the report, ECR spectra indicated the anatase-TiO₂ is reduced by hydrogen excited in the plasma and oxygen-vacancies are for- med with the double defect levels of 0.75 and 1.18 eV below the conduction band minimum. Although non-ex- cited hydrogen gas was supplied below 380˚C during the deposition in this work, it can be considered hydrogen was adsorbed to Ti as shown in Figure 3 and maybe form oxygen-vacancies in the layer. However, the defect level is too deep to decrease the resistivity at room-temperature. Therefore, the defect to reduce the resistivity is not oxygen-vacancy but Ti³⁺ with the defect level shallower than the complex centers predicted by theoretical calculations [15] . Indeed, the activation energy of 244 kJ/mol obtained by the result in Figure 3 was similar to the dissociation energy of 208 kJ/mol for Ti-H bond [18] .

Figures 4 shows optical absorption spectra in UV-Vis region. Dot-lines in Figure 4(a) show transmission spectra of TiO₂/H₂ (red-line) and TiO₂/O₂ (black-line) layers grown at 360˚C with the thickness about 200 nm. Gra- dual increase and relatively low value of optical transmittance in Vis-region suggested deep-level defects were included in the forbidden energy gap. The spectra were successfully deconvoluted into two-spectra with the peak at 2.2 eV and 3.0 eV, which were described by Gaussian-function, and near band edge absorption obtained after removal the two-spectra from the raw transmission spectrum. In Figure 4(a), the deconvoluted spectra and

the fitting curves are shown by solid-line (red-line: TiO₂/H₂, black-line: TiO₂/O₂) and blue-lines, respectively. Previously, Sekiya et al. suggested the origin of absorption at 3.0 eV and in between 2.2 and 2.3 eV, which were coincident to peak energies of the deconvoluted spectra in Figure 4(a), is attributed to oxygen-vacancies with two-trapped electrons and electron transition of Ti³⁺ from ²T₂ to ²E state contributed by crystal field interaction with surrounding oxygen, respectively [19] . They also showed intensity of the absorption in between 2.2 and 2.3 eV is decreased by post-annealing above 300˚C with decreasing electrical conductivity. In this work, the absorption peak at 2.2 eV was significantly increased in the intensity for TiO₂/H₂ whereas the absorption peak at 3.0 eV was scarcely different as compared with TiO₂/O₂. From these results, it is considered that increase of Ti³⁺ density in TiO₂/H₂ was independent from the increase of oxygen-vacancy and resulted in the reduction of resistivity. It is noted significant broadening of the absorption around 2.2 eV was also observed for the TiO₂/H₂ in comparison with TiO₂/O₂ as shown in Figure 4(a), in which the full-with at half-maximum (FWHM) was 1.4 eV and 0.86 eV for the TiO₂/H₂ and TiO₂/O₂ respectively. The absorption broadening could be recognized crystal field around Ti³⁺ was more fluctuated in the layer as compared with the TiO₂/O₂, that is, structural disordering of TiO₂/H₂ was higher than that of TiO₂/O₂. Indeed, XRD spectra showed crystallinity of TiO₂/H₂ was inferior to that of TiO₂/O₂ as shown in Figure 1(b). Further, it has been reported structural disordering causes band gap narrowing in crystal. Figure 4(b) shows Tauc plot in (αhν)^1/2 vs. photon energy (hν) to determine the optical band gap of TiO₂/H₂ (red-line) and TiO₂/O₂ (black-line), where α is absorption coefficient. The band gap of 3.16 eV given by the extrapolated linear line (black dot-line) for the TiO₂/O₂ was scarcely smaller than that of 3.2 eV for anatase-TiO₂. In contrast, the band gap was significantly reduced to 3.0 eV for the TiO₂/H₂. The estimated band gap of 3.0 eV was similar to that for rutil-TiO₂ but the lowering of band gap could be considered to be originated from defect states in the band gap, because any XRD diffraction peak corresponding to rutile phase could not be observed in the layer. On the other, it has been recognized localized defect states originated from structural disordering are extended form an optical absorption tail to lower energy side below the band gap and resulted in origin of the optical band gap narrowing, in which the tail is referred as Urbach tail represented by α = α_oexp(hν/E_u) using Urbach energy (E_u) [20] . Figure 4(c) shows absorption coefficient of the TiO₂/H₂ (red- line) and TiO₂/O₂ (black-line) in log-scale against photon energy, where two-absorption spectra centered at 2.2 eV and 3.0 eV were numerically removed. Urbach energy obtained by the linear relationship in log(α) vs. photon energy below band gap was 153 meV and 114 meV for the TiO₂/H₂ and TiO₂/O₂, respectively. Previously, Choudhury et al. [21] reported the Urbach energy is dependent on anneal condition for sol-gel TiO₂, and the energy of 140 meV for the anatase-TiO₂ annealed in vacuum and about 100 meV for anatase/rutile mixed TiO₂. They concluded such increase of the Urbach energy is come from oxygen-vacancies or structural disordering at interface with different crystal phases. In this work, Urbach energy for the TiO₂/H₂ was higher than that for the TiO₂/O₂, however, the density of oxygen-vacancy was similar for the both layers as shown by the absorption intensity around 3.0 eV in Figure 4(a). In contrast, from the result that the absorption around 2.2 eV was increased in the intensity for the TiO₂/H₂, it can be suggested not only the oxygen-vacancy but also the Ti³⁺-re- lated defects and/or the complexes contributed the band gap narrowing.

Figure 5(a) shows dependence of resistivity evaluated by VDP method for TiO₂/H₂ layer. The electrodes using Indium-dots were ohmic at overall temperatures as shown in Figure 5(b), in which the current-voltage characteristics at 10 K (open-circle) and 300 K (solid-circle) are plotted. The saturated feature at low temperature indicated the layer was degenerated by the high density of conduction electron originated from shallow donors. In contrast, the resistivity was exponentially decreased with increasing temperature above 100K in Arrhenius relationship with the activation energy of 96.1 meV. Since it is difficult to attribute the activation energy to donor-level energy below conduction-band minimum in the degenerated semiconductor, influence in the grain boundary region should be taken into account in the electric conduction of the layer. In the grain boundary region, it is easy to consider deep level defects are formed by significantly disordered structures. The defects in the boundary region trap the carriers or compensate the shallow donors. In the case that the carriers are traps in the deep defects, depletion region is introduced toward the kernel. Previously, Blom et al. presented a conduction model for poly-crystalline layer consists of kernels and grain-boundaries and applied the analysis for temperature dependent conduction of ZnO layer [22] . In the model, resistivity of the layer parallel to the substrate is limited by that of grain boundaries surrounding kernel, because Schottky barrier is formed by traps in the grain- boundary because the resistivity of kernel can be expected to be much lower than that of the boundary region. Since the density of thermally emitted carrier across the barrier is increased with temperature, resistivity of the layer is decreased with temperature and then saturated to the resistivity due to the kernels at high temperature. However, the model is insufficient to the result in this work because the resistivity was saturated at low temperature. On the other, in the case that the carrier density is significantly reduced by compensation of shallow donors in the boundary region, resistivity in the region is much higher than in the kernel and dependent on the temperature. The resistivity in grain-boundary region is expected to be decreased with temperature. In this case, although resistivity of the layer limited by the resistance in grain-boundary is saturated at low temperature and reduced with temperature, the activation energy experimentally obtained by the result of Figure 5(a) is too large to explain the result of temperature dependent resistivity. Here, it should be noted that defects at deep level are also formed at surface. When electrons are captured by the traps at the surface, depletion region is extended in the boundary region with low carrier density from the TiO₂ surface. The depletion length can be estimated by density of the trapped electron in the surface state (N_sse) and effective donor density in the boundary region (N_bd) in the relationship of N_bd/N_sse. The depletion region acts to cut off carrier transport across the boundary region and the depleted length is decreased with temperature due to activation of electrons captured the traps, which is resulted in decrease of the resistivity with temperature. Details based on the model including surface treatment of the layer are in progress, but the simulated result can explain well the temperature dependent resistivity as shown by solid-line in Figure 5(a). The result indicated N_bd/N_sse, peak energy of the surface trap below conduction band minimum, FWHM of surface trap distribution for electron energy and resistivity without the cut-off of electron transport in grain boundary is 1.8 × 10⁻⁵, 195 meV, 136 meV and 4.2 × 10⁻³ Ω cm, respectively. The obtained parameters suggested resistivity of the kernel is much lower than the effective resistivity limited by the grain boundary.