Figure 1 shows deposition rate of undoped layer in H₂-ambient (black-circle) or O₂-ambient (open-circle) and Nb-F doped layer (red-circle) in H₂-ambient at various deposition temperatures in Arrhenius plot, in which the gas flow rate of H₂ or O₂ into the reactor was 4.2 sccm and NbF₅ was evaporated from crucible with the rate about 0.05 mg/min for the deposition of doped layer. Deposition rates of the undoped layers were increased with the temperature and then saturated to the rate about 8 nm/min at the temperature above 360˚C, in which the activation energy of 228 kJ/mol in H₂ was clearly larger than that of 138 kJ/mol in O₂. Here, it is considered that TTIP-dissociation in H₂-ambient is not owing to chemical reaction on the deposition surface because the higher activation energy is similar to activation barrier of 238 kJ/mol to form Ti(OC₃H₇)₃(OH) through monomolecular dissociation of TTIP calculated by quantum chemical calculations using B3LYP. The surface chemical states have not been identified but it is expected that the surface is terminated by hydroxyls which can be formed during TTIP-dissociation and/or by hydrogenation of oxygen at the surface [12] . On the other, when NbF₅ was simultaneous supplied during the deposition in H₂, the deposition rate was increased with the activation energy of 120 kJ/mol and then saturated to 18 nm/min above 380˚C. The activation energy and the saturation temperature were smaller and higher than that for the undoped layer in H₂-ambient respectively, which was resulted in higher saturated deposition rate. The results clearly indicate TTIP-dissociation in H₂ was promoted by NbF₅. Here, it is difficult to recognize reduction of the activation energy by NbF₅ supply in H₂ is due to fluorination of TTIP because the fluorination reduces the deposition rate as shown previously for the deposition in O₂-ambient [10] . Further, it is mentioned sticking probability of TTIP on the deposition surface was increased by NbF₅ contribution because the saturated deposition rate at high temperature was increased by NbF₅ supply. As the results, it should be concluded that the TTIP-dissociation was promoted by reactive site on the surface contributed by NbF₅.

Deposition feature of the doped layer was also influenced by NbF₅ and H₂ gas supply rate. Figure 2(a) shows dependence of TiO₂ deposition rate at 380˚C in H₂ with the flow rate of 0.75 (open-circle), 4.2 (solid-circle) and 8.3 sccm (solid-triangle) on NbF₅ evaporation rate from crucible, where TTIP supply rate was same for the depositions. The deposition rate was linearly increased and then gently decreased with NbF₅ evaporation rate de-

pending on H₂ flow rate. It is easily expected that F play an important role for TTIP-dissociation but not the fluorination. Further, it is difficult to recognize that the significant increase of deposition rate was due to the reaction between TTIP and F in vapor phase because the density of NbF₅ was far lower than the TTIP, in which the NbF₅/TTIP supply ratio can be roughly estimated as a few % from the TiO₂ deposition rate and the NbF₅ evaporation rate with taking the reactor structure into account. Therefore, it should be discussed the TTIP-dissociation contributed by F on the deposition surface as shown in the next section. It is however considered the excessive NbF₅ fluorinates TTIP, which is resulted in decrease of the deposition rate. Since structural disordering in the layer is enhanced by the fluorination of alcohol ligands of TTIP as previously suggested [10] , the excess NbF₅- supply on the deposition surface is insufficient for the deposition. The deposition rate linearly increased with NbF₅ evaporation rate was decreased with H₂ flow rate as shown in Figure 2(b), in which incremental rate of deposition rate due to NbF₅ evaporation rate is shown for H₂ flow rate. The result suggested density of the reactive species for TTIP-dissociation on the deposition surface is decreased with H₂ density on the surface, which can be recognized that surface F is non-activated as HF by H.

Previously, it was reported that HF interacts to hydroxyl groups and oxygen on TiO₂ surface by hydrogen- bonding in weak acidic HF solution and hydroxyl groups on TiO₂ surface is easily exchanged to F in strong acidic HF solution [13] . Although HF is not preliminary supplied on the deposition surface in the LPCVD pro- cess, it is expected that Ti-F is formed with NbF₄OH by the reaction between NbF₅ and hydroxyls on the surface, then HF is produced after TTIP is adsorbed at the reactive site of surface F in Ti-F. According to the reaction, a possible scheme for the deposition with NbF₅-supply can be depicted in Figure 3. When NbF₅ is supplied on surface hydroxyl groups in Scheme 1, the OH is exchanged to F with formation of NbF₄OH (Scheme 2). The next coming TTIP adsorbed to the surface F is dissociated with forming HF and propene, and then the partially dissociated TTIP is coordinated to surface Ti (Scheme 2 & Scheme 4) as O-Ti-(OR)₃ which then form oxo-bridging between the neighbors via the thermal dissociation. Here, the produced HF interacts to the neighbor hydroxyl groups (≡Ti-OH) and oxygen (Ti-O) (Scheme 4), then ≡Ti-F is formed with product of H₂O (Scheme 5), in which the reactions from Scheme 2 to Scheme 5 are sequentially processed on the deposition surface. As previously discussed, TTIP-dissociation is owing to the monomolecular thermal process due to the low sticking probability on the hydrogen-terminated surface (Scheme 1), which is resulted in the significant increased activation energy for the deposition rate in H₂-ambient [12] . In contrast, it is expected the surface F with high electron negativity acts as catalyst for partial dissociation of TTIP (from Scheme 3 to Scheme 4). As the result, activation energy for the deposition rate is significantly decreased by NbF₅-supply as shown in Figure 1 and the deposition rate is increased with the density of F (NbF₅ evaporation rate) as shown in Figure 2(a), whereas the of F-density is far lower than that of TTIP. It is noted that hydrogen supplied as H₂ gas acts to form HF on the surface before TTIP is adsorbed to the surface F in Scheme 2 or Scheme 5, which is resulted in decrease

of the deposition rate with increasing the H₂ flow rate as shown in Figure 2(b). On the other, excessively supplied NbF₅ probably causes fluorination of alcohol ligands in Scheme 4 or Scheme 5, which is insufficient to control property of the layer because significant structural disordering is introduced into the layer.

Figure 4 shows resistivity of undoped (black-circle) and doped (red-circle) layers with the thickness about 200 nm deposited at various temperatures in H₂ with the flow rate of 4.2 sccm, where NbF₅ evaporation rate was fixed at 0.05 mg/min for the doping. Resistivity of the undoped layer was decreased with increasing the deposition temperature according to the activation energy of 244 kJ/mol but increased beyond 360˚C. In case of the doped layer, the resistivity higher than the undoped layer at low temperature was also decreased with the deposition temperature below 380˚C with larger activation energy (319 kJ/mol) than the undoped layer. As the result, resistivity of the doped layer deposited at temperature beyond 370˚C was more reduced than the undoped layer. We previously speculated that Ti³⁺ is formed in the layer with oxygen-vacancy by desorption of hydrogen from H-Ti≡ formed in H₂-ambient, in which the resistivity is decreased with the temperature according to the dissociation energy of H-Ti≡ [12] . It is however difficult the same scheme for the layer deposited by NbF₅ supply in H₂ because the activation energy was larger than the dissociation energy. When NbF₅ is simultaneously supplied to the deposition surface, F and Nb substituted to O-site and Ti-site are expected to act as relatively shallow donors in TiO₂ respectively, where the F generates Ti³⁺ without O-vacancies and the Ti³⁺ forms Ti-3d¹ band [14] . In contrast, interstitial F and anti-site Nb reduce the carrier density because such defects act as acceptors. In case of the doping in O₂-ambient, F is substituted to the O-site but the crystallinity was significant degraded by fluorination of TTIP due to oxidation of NbF₅ [10] . In contrast, it is considered the fluorination can be prevented in H₂-ambient because of removal the oxidant for NbF₅ but Nb reduced from NbF₅ by hydrogen maybe substitute the O-site. Indeed, the activation energy of 319 kJ/mol is close to the bond dissociation energy of Nb-Ti (302 kJ/mol) [15] . On the other, the resistivity was increased at high temperature beyond 360˚C for the undoped layer and 380˚C for the doped layer. It is interesting the deposition rate was limited by TTIP-supply rate at such high temperatures, which indicates the most efficient doping can be achieved at the highest temperature in surface reaction limited region for the deposition rate.

Figure 5(a) shows resistivity of the layer as a function of NbF₅ evaporation rate in H₂ with the flow rate of 0.75 (open-circle), 4.2 (solid-circle) and 8.3 sccm (solid-triangle), where the layers were deposited with the thickness about 200 nm at 380˚C. The results can be recognized not only NbF₅ evaporation rate but also H₂ flow rate affects to the doping feature. For the NbF₅ evaporation rate, the resistivity was decreased with increasing the evaporation rate but the feature was different in the high evaporation rate. Since the deposition feature was disturbed in the high NbF₅ evaporation rate as shown in Figure 2(a), it is not easy to discuss the dependence of resistivity on the doping in the high NbF₅ evaporation rate. In contrast, the resistivity in the low NbF₅ evaporation rate is uniquely decreased with the evaporation rate, where the dependence should be discussed for NbF₅ density on the deposition surface. The NbF₅ supply rate on the deposition surface () is proportional to [NbF₅]/DR, because the density is increased with NbF₅ evaporation rate ([NbF₅]) and decreased with deposition rate of the layer (DR). Figure 5(b) shows resistivity of the doped layer for [NbF₅]/DR in log-log scale. Resistivity of the doped layer (ρ_doped) was decreased with [NbF₅]/DR, which was resulted in the relationship of ρ_doped ~ ()ⁿ. If

the donor was simply increased in the density with, the resistivity should be decreased in the factor n of −1. In contrast, the factor n less than −1 and beyond −1 suggests the crystallinity including the grain boundary was improved and degraded with the donor density, respectively. In the result of Figure 5(b), the factor for the deposition in H₂ with the flow rate of 0.75, 4.2 and 8.3 sccm was −0.94, −1.7 and −0.45 respectively. Figure 5(c) shows variation of the factor n for the H₂ flow rate, in which the factor n was decreased with the H₂ flow rate but notably increased by excess H₂ flow rate. It should be mentioned here that the factor n was less than −1 for the deposition with H₂ supply rate of 4.2 sccm. The interesting result suggests the doping with sufficient H₂ supply not only increased donor density but also improved the crystallinity. In previous report, we speculated oxygen-vacancies increase resistivity of the doped layer [10] , in which the defects probably form complex centers with deep levels. In contrast, it is expected that F doping into O-site in TiO₂ achieves to reduce the O-vacancies in the density and creates Ti³⁺ with shallow donor level. Of course, the doping feature is owing to the dissociation scheme of the dopant. The result of the factor n dependent on H₂ flow rate indicates sufficient hydrogen supply is required for the efficient doping but excess hydrogen is resulted in invalidation of the doping. The insufficient doping in excess H₂ is probably caused by anti-site Nb which is introduced by species excessively reduced from NbF₅ as speculated in the Section 3.1.1.

Figure 6(a) shows transmission spectra in UV-Vis region of undoped layer (black-line) and Nb-F doped layer (red-line) deposited in H₂, in addition to the spectrum of Nb-F doped layer deposited in O₂ (blue-line) [9] . The layers were deposited with the thickness about 200 nm by the condition optimized to reduce the resistivity and

resulted in the resistivity of 0.2, 0.05 and 0.2 Ω∙cm for the undoped and the Nb-F doped layers in H₂ and the Nb-F doped layer in O₂ respectively. Here, the spectrum of undoped layer in O₂ was similar to the blue line. As previously reported, the undoped layer in H₂ showed broad optical-absorptions around 3.0 eV and 2.2 eV due to deep levels concerned with oxygen-vacancies and Ti³⁺ [10] respectively. In contrast, although significant multi-reflection was included, such absorptions were remarkably prevented in the spectrum of the doped layer deposited in H₂. Especially, the absorption around 3.0 eV was obviously decreased in the doped-layer, which indicated density of oxygen-vacancy in the doped layer was notably reduced in comparison to the undoped layer. It is not difficult to anticipate O-vacancy is easily induced into the oxide in H₂-ambient but is not originated from the reduction of TiO₂ in the CVD-process. The defect is probably formed by TTIP-dissociation via monomolecular thermal dissociation as suggested previously [12] . In the case of simultaneous supply of NbF₅, the dissociation scheme is quite differed as shown in Figure 3, in which TTIP can be chemically dissociated at the reactive site of F on the deposition surface. As the result, the deposition can be performed with preventing O-vacancy by the dissociation supported by the surface F. In addition, F-doping into the O-site may reduce the density of O-vacancy. It is noted that the optical transparence of the Nb-F doped layer deposited in H₂ was better than the doped layer in O₂. Commonly, O₂-gas has been used for CVD of TiO₂ to improve the property aiming at removal oxygen-deficiency and residual carbon-related impurities. However, the results for optical absorption revealed H₂ simultaneously supplied with NbF₅ is more effective than O₂ to reduce optical absorption due to deep-levels especially originated from oxygen-vacancies. It is believed the reduction of the deep-level density was come from the sufficient TTIP-dissociation on the deposition surface and F substituted to the oxygen-site. On the other, structural disordering causes the optical band gap narrowing due to absorption tail below the fundamental absorption edge which is referred as Urbach tail [16] . Figure 6(b) is Tauc-plot in (αhν)^1/2 vs. photon energy, where α, h and ν is absorption coefficient, Plank constant and wavenumber respectively. Optical band gap of 3.0 eV for the undoped layer in H₂ was narrower than 3.2 eV for anatase-TiO₂. In contrast, the band gap energy of 3.27 eV for the doped layer was in good agreement with 3.26 eV for anatase-TiO₂ [17] , which indicated structural disordering in the doped layer was significantly improved as compared with in the undoped layer. Here, the band gap energy of 3.20 eV for the doped layer in O₂ was slightly large as compared with the energy of 3.16 eV for the undoped layer in O₂ [12] , which suggested the structural disordering can be improved also in O₂ by NbF₅-supply. However, the larger energy of 3.26 eV for the doped layer in H₂ indicated NbF₅-supply in H₂ is more effective than that in O₂.