Al thin films were deposited on a glass (SCHOTT Nippon K.K., Glass code: D263T) substrate by a radio frequency (RF) magnetron sputtering (SHINKO SEIKI CO., LTD, Type: SRV4320), where the deposition time was extended to 40 min, obtaining 100 nm of Al films. The Al-coated glass substrate was also used as a counter electrode of PEC. ZnO or NbO_x was deposited on the Al/glass substrate by RF magnetron reactive sputtering (ULVAC JAPAN, Ltd., Model: YH-500A or DIAVAC LIMITED, Type: DS-412Z, respectively). Figure 1 shows the structure of ZnO or NbO_x/Al/glass photovoltaic electrode. Table 1 shows the deposition conditions of Al, ZnO and NbO_x, respectively.

Figure 2 shows the schematic of PEC used in photoelectrochemical measurements. KNO₃ solution was commonly used as an electrolyte solution in both the electrochemical and photoelectrochemical measurements.

KNO₃ was dissolved in a distilled water (Wako Pure Chemical Industries, Ltd., CAS No.: 7732-18-5, 200 ml), obtaining 0.1 mol/L electrolyte solution. During the measurements, the electrolyte solution was kept at a constant temperature of 25˚C, and the measurements were done in a dark place.

Firstly, the electrode potential was measured. The ZnO or NbO_x/Al/glass electrode and a carbon counter electrode (NaRiKa Corporation, CAT. NO. B10-2050-09) were immersed in the electrolyte solution. The electrodes were connected to a potentiostat (Bio Logic, SP-50), and an Ag/AgCl electrode in saturated KCl solution (HOKUTO DENKO Co., HX-R6) was also connected as a reference electrode. The potential of the electrodes with respect to the reference electrode was determined by measuring open circuit voltage for 100 s. Subsequently, electrochemical stability, that is, corrosion resistance of the electrodes was evaluated, where a constant load discharge (CLD) mode with an electrical resistance of 100 kΩ installed in the potentiostat was used. At this time, Al/glass soaked in a separate beaker was used as a counter electrode, and the beakers were connected with a KNO₃ salt bridge. Elution of Nb, Zn and Al in the electrodes was investigated after immersing the electrodes for 1 hour. Inductively coupled plasma measurement (SEIKO and VARIAN Inst., Vista-PRO CCD Simultaneous ICP-OES) was used for the detection of the elements eluted in the electrolyte solution.

In the evaluation of photoelectrochemical properties, photocurrent generation was firstly measured by using the PEC shown in Figure 2, where a solar simulator (SAN-EI ELECTRIC, XES-40S1, Irradiance: 1000 W/m²) was used as a light source. The CLD mode with 100 kΩ resistance installed in the potentiostat apparatus was also used in the photocurrent measurements. During the measurements, voltage was not controlled by the potentiostat, and only the current passing through the resistance in the potentiostat apparatus was measured. The ZnO or NbO_x side of the photovoltaic electrodes was exposed to a simulated sunlight generated by the solar simulator. The power supply of the potentiostat was turned on. After waiting for 100 s without irradiation, the light emitted from the solar simulator was irradiated to the photovoltaic electrode for 300 s. The irradiation was repeated for 9 times at an interval of 100 s, and the total measurement time was 3600 s. After the measurements, elution of the electrode constituents was also evaluated by ICP.

Finally, electricity generation property was investigated by measuring current and voltage generated by the light irradiation. In the measurement, the cell setup given in Figure 2 was also used, and a variable resistor (YOKOGAWA ELECTROC WORKS. LTD., Decade Resistance Box Type 2793, Max: 111.1110 MΩ) was, however, connected to the electrodes instead of the constant load. The current and voltage were measured by using multimeters (Agilent Technologies, Inc., Agilent 34,450 A 5 1/2 Digit Multimeter). Photocurrent and photovoltage were measured by changing resistance. Dark current and voltage measurement without light irradiation was also carried out, and net electricity generation was evaluated by subtracting dark current and voltage from photocurrent and photovoltage.

In the previous report [12] , ZnO nano-islands were not distinguished clearly by AFM (Bruker AXS K.K., Dimension Icon). In the present study, a tapping mode AFM apparatus was also used, where a phase mode was applied to distinguish the NbO_x or ZnO deposits from the Al thin film. In the phase mode, the difference in hardness of two substances, that is, metal and oxide, appears in the difference in phase angle, which can be used for the distinction.

In the previous study [12] , X-ray diffraction (XRD) measurement (Rigaku Corporation, Ultima IV) indicated that ZnO deposits were crystalline. However, it was guessed that NbO_x deposits were amorphous [27] , and actually sharp peaks were not observed in XRD patterns. It is generally known that band gap of Nb₂O₅ is 3.4 eV, and it is hence expected that the stoichiometry of NbO_x is estimated from the optical absorption edge. Then, optical absorption measurement was performed by using a double-beam spectrophotometer method (Shimadzu Corporation, UV-3100). For the measurement, NbO_x were directly deposited on a quartz glass substrate (Sendai quartz glass Works, Co. Ltd., Synthetic quartz glass, SUPRASIL-P30) under the same conditions given in Table 1. Furthermore, the absorption spectra for the reagents, NbO (Kojundo Chemical Lab. Co., Ltd., Cat. No. NBO01PB), NbO₂ (NBO02PB), and Nb₂O₅ (NBO06PB) were also measured with an integrating sphere (JASCO Corporation, ISN-470) for comparison.

Electrode potentials are shown in Table 2. The electrode potential of Al/glass (−0.70 V vs. Ag/AgCl), which is higher than the standard electrode potential of Al (−1.662 V vs. SHE, standard hydrogen electrode) [28] . The

electrode potentials of ZnO- and NbO_x-deposited electrodes are higher than those of the Al/glass electrodes. In NbO_x/Al/glass electrode, the electrode potential increases with increasing the deposition time, and similar trend is also seen in ZnO/Al/glass electrode. During the potential measurements, the electrode potential of NbO_x-de- posited electrodes was stable and almost constant with small fluctuation. In the case of ZnO-deposited electrodes, however, the electrode potential decreased continuously during the measurements, and the electrode with shorter deposition time had higher rate of decrease in electrode potential. The decrease in the electrode potential during the measurements suggests the two possible changes in the ZnO-deposited electrodes, that is, the reduction of Zn from 2+ to 0 in valence number and the dissolution of ZnO deposits into the electrolyte solution.

After electrochemical corrosion tests, the elution of Al and Zn from ZnO/Al/glass electrodes is confirmed, and such the elution is not observed in NbO_x/Al/glass electrodes. The elution is discussed later.

Figure 3 shows the photocurrent density curves of ZnO- and NbO_x-based PECs. In these PECs, ZnO and NbO_x were deposited for 3, 17 s and 17, 100 s at the apparent deposition rates of 1.00 and 0.17 nm/s, respectively, expecting the comparable thicknesses of ca. 3 and 17 nm. In the previous study [12] , it was confirmed that ZnO deposits were present as nano-islands at these thicknesses. As shown in Figure 3, current generation during the light irradiation is successfully confirmed in the NbO_x-based PECs, where current generation is stable, and very little degradation is observed in current generation even after repeating irradiation. In case of the ZnO deposits, current decreases rapidly and changes into increase just before irradiation. Even after starting irradiation, current increases slowly and continuously, and in the case of 3 s ZnO-deposited PEC, current stops increasing at ca. 1700 s. Even when light was not irradiated, almost the same current curves were obtained, suggesting that most of current obtained in the ZnO-based PECs was caused not by photoelectrochemical but by electrochemical reactions. However, only in 3 s ZnO-deposited PEC, small but clear decrease in current is recognized after stopping 4th irradiation, which indicates photocurrent generation by photo-induced electrochemical reactions. When the light irradiation is continued for a few hours or more, very small and few bubbles are generated at all the electrodes, that is, photovoltaic and counter electrodes in both NbO_x and ZnO-based PECs.

In the previous study of ZnO nano-islands [12] , Al-etching under UV irradiation was observed, and as mentioned, the elution of Zn and Al in ZnO/Al/glass electrode was also confirmed in dark place. Table 3 shows the concentrations of the elements eluted from ZnO/Al/glass electrodes during the electrochemical corrosion tests

and photocurrent measurements. After the photocurrent measurements, Zn and Al were confirmed in the electrolyte solution, and the concentrations were larger as compared with the case without light irradiation. Therefore, it is supposed that the electrochemical reactions resulting in the elution of Zn and Al are accelerated by their radiation. In case of the NbO_x-based PECs, neither Al nor Nb elution was detected in the photovoltaic electrode. However, Al elution from the counter electrodes was confirmed in both PECs.

In both PECs, electrons flowed from the counter electrode into the photovoltaic electrode through the external circuit during light irradiation. The direction of electron flow was the same even in the electrochemically-unst- able ZnO-based PECs. According to the direction of electron flow, following reduction and oxidation reactions are expected at the photovoltaic and counter electrodes, respectively.

Photovoltaic electrode:

,

Counter electrode:

,

However, the current flow observed in the present study is opposite to the direction commonly observed in wet-type solar cells with semiconductor electrodes, such as TiO₂ [29] -[31] .

Figure 4 shows the relation between photocurrent density, J and photovoltage, V (J-V curve) at different deposition time of NbO_x. By subtracting dark current and dark voltage from photocurrent and photovoltage obtained at the same electrical resistance, the net photocurrent and photovoltage generated by light irradiation were estimated, obtaining the net J-V curve. In general, gradual and steep decreases in photocurrent density at lower and higher photovoltage regions are expected in typical PECs, and in the PECs shown in Figure 4, however, an

opposite change in J-V curve is commonly observed, that is, steep and gradual decreases in J are seen at lower and higher V regions, respectively.

Then, fill factor, FF is calculated from the equation, FF = P_max/(J_sc × V_oc), where P_max is the maximum power density, J_sc is the short circuit current density, and V_oc is the open circuit voltage. Figure 5 shows J_sc, V_oc, P_max and FF of the NbO_x-based PECs at the different NbO_x deposition time. J_sc and P_max reach respective maxima of ca. 100 nA/cm² and 2.9 nW/cm² at the deposition time of 17 s. FF reaches maximum of ca. 25% at the deposition time of 15 s. V_oc reaches maximum of ca. 270 mV at the deposition time of 150 s. J_sc and P_max indicate second maxima at the deposition times of 100 and 150 s, respectively. The detail of photoelectrochemical conversion mechanism is discussed later.

Figure 6 shows the height images (a)-(e) and the phase images (f)-(j) obtained by AFM for the NbO_x/Al/glass electrodes before the photoelectrochemical measurement. Comparing the height images, surface roughness of NbO_x-undeposited Al/glass substrate (Figure 6(a)) and NbO_x-deposited NbO_x/Al/glass substrates (Figures 6(b)-(d) at deposition time ≤ 40 s) is almost unchanged. As for the phase images, the NbO_x-undeposited Al/glass substrate indicates the phase angle of ca. 45˚ (Figure 6(f)), which may correspond to oxidized surface layer of Al film. In the NbO_x-deposited specimens (Figures 6(g)-(j)), components with the phase angles of −60˚ and 0˚ - 45˚ are confirmed. It is thought that these components are metallic Al and NbO_x, respectively, because the 0˚ - 45˚ component increases with increasing the deposition time. It is supposed that the oxidized surface Al layer (phase angle = 45˚) is removed or reduced to metallic state (−60˚) by NbO_x sputter deposition. In the phase images, nano-islands of NbO_x deposits are also confirmed at the deposition time of 10 s. The further deposition results in the growth and densification of NbO_x islands at the deposition time of 40 s (Figure 6(i)), and the surface of Al/glass substrate is almost completely covered with a film of NbO_x at the deposition time of 100 s (Figure 6(j)).

In Figure 6, two-dimensional phase images (k-n) are also shown, in which points with the phase angles larger than −20˚ are shown in white, which indicate NbO_x. At the deposition time ≤ 20 s, growth of NbO_x in an island state is confirmed. At the longer deposition time = 100 s, NbO_x are in a film state, and the underlying Al/glass substrate seems like lakes. Surface area of NbO_x increases continuously with increasing the deposition time.

After the photoelectrochemical measurement (Figure 3), AFM observation was performed again. Figure 7 shows AFM images and two dimensional images of the NbO_x/Al/glass electrode at the NbO_x deposition time of 17 s as-deposited and after 9 repetitions of photo irradiation in the photoelectrochemical measurement. During the repetitions, the electrode was kept in the electrolyte solution without being exposed to the atmosphere. In the AFM images, nano-island structures are confirmed, and quite little change is observed as compared with the as- prepared specimen. On the other hand, in the Al/glass counter electrode (Figure 7(c)), surface roughness increases slightly as compared with the as-deposited one (Figure 6(a)), and the surface with phaseangle of ca. 0˚ is newly found other than that of 45˚ (Figure 7(f)). It is probably due to the Al elution. Figure 8 shows the AFM images of the ZnO/Al/glass electrode with the ZnO deposition time of 3 s. In the as-prepared electrode, nano- islands of ZnO deposits are confirmed (Figure 8(d)), and in the electrode after the 1st photoelectrochemical measurement, however, larger islands are also observed in the phase image (Figure 8(e)), and the surface rough- ness given in Figure 8(b) indicates a small change. The maximum phase angle increases slightly from 50 to 70˚ in the electrodes before and after the photoelectrochemical measurement, suggesting that the surface is covered with some different substance other than ZnO. After 9 repetitions of the light irradiation, the nano-islands grow higher, and the ditches become deeper, as shown in the height image (Figure 8(c)), and the surface is covered with a substance with large phase angle of 70˚ (Figure 8(f)). The ditches reach a depth of −16 nm, corresponding to 1/6 of the thickness of Al film (100 nm). As mentioned, Al and Zn elution from ZnO/Al/glass electrodes is confirmed, and it is therefore supposed that Al film and ZnO deposits are eluted in the electrolyte solution, and the eluted Al and Zn species are re-precipitated as solids, probably Al(OH)₃ and Zn(OH)₂, which cover for the electrode surface.

Figure 9 shows the optical absorption spectra of NbO_x deposited on a quartz glass substrate, in which those of the reagent powders of NbO, NbO₂, and Nb₂O₅ are also shown for comparison. In the absorption spectra of

NbO_x deposits, a leading edge is commonly confirmed at ca. 2.9 eV, which is close to the absorption edge of Nb₂O₅ reagent powder. For the NbO_x deposits, a steep increase in the slope of absorption curves is also recognized at 3.5 - 3.7 eV, which is close to the band gap of 3.4 eV generally known for Nb₂O₅. In our measurement, however, the optical band gap of Nb₂O₅ reagent powder is ca. 3.1 eV, which is smaller than the energy gap of 3.4 eV given in literatures, and it is hence suggested that the NbO_x deposits have different electronic state or atomistic structure from Nb₂O₅ crystal. Actually, no crystalline diffraction peaks are observed in XRD patterns of the NbO_x deposits, indicating that the NbO_x deposits are in amorphous state. It is supposed that the amorphous state of the NbO_x deposits is due to oxygen deficiency, and it is also expected that part of Nb ions are reduced to 4+ or 2+ from 5+. The reagent powders of crystalline NbO₂ and NbO consisting of lower valency of niobium ions show strong absorption in almost the entire region. The NbO_x deposits indicate quite different absorption spectra from NbO₂ and NbO. Therefore, it is difficult to estimate the valence state of niobium ions in the NbO_x deposits, but it is supposed that most of Nb ions are present as 5+.

As mentioned, the elution of Al and Zn from ZnO/Al/glass electrodes is observed regardless of light irradiation,

and such the elution is not recognized in NbO_x/Al/glass electrodes, which are explainable by the electrochemical corrosion based on potential-pH diagrams [32] . It is known that ZnO is dissolved in acidic and alkaline water but is not soluble in neutral water [33] [34] . It is also known that metallic aluminum is easily corroded by water, and in practice, however, surface Al is oxidized to be covered with insoluble and passive Al₂O₃ skins, which prevents Al corrosion in neutral water [35] [36] . Even in neutral water, Al may be corroded by some corrosion processes, such as galvanic, crevice and pitting corrosions. In galvanic corrosion, a base metal with lower electrode potential will work as anode and corrodes in the combination with noble materials having higher electrode potential. As shown in Table 2, Al/glass has the lowest electrode potential among the electrode constituents. If the galvanic corrosion were predominant, Al elution from NbO_x/Al/glass electrodes would be observed. However, Al elution is confirmed only in the ZnO/Al/glass electrodes, and it is hence supposed that the galvanic corrosion is not responsible for the Al elution from the ZnO/Al/glass electrodes.

Figure 10 shows that the magnified AFM images of the ZnO/Al/glass and NbO_x/Al/glass electrodes at the

deposition times of 3 and 17 s, respectively. Nano-dips with the depth of ca. 1 nm at brinks of nano-islands (vertical lines in Figure 10) are observed in both electrodes, which are probably produced by sputtering damages. At the dips in Al films, passive Al₂O₃ skins should be removed, and metallic Al might be exposed. Some of the dips should be covered with ZnO and NbO_x deposits, and therefore, pitting and/or crevice corrosions are more likely to occur at the Al/glass surface. The supposed corrosion mechanism for ZnO/Al/glass electrode is illustrated in Figure 11. The pitting and crevice corrosions are initiated by an anode reaction, Al → Al³⁺ + 3e⁻. The electrons migrate toward ZnO deposits, because ZnO has higher electrode potential than Al (Table 2). When oxygen is dissolved in neutral water, a cathode reaction, O₂ + 2H₂O + 4e⁻ → 4OH⁻ is expected at the surface of ZnO deposits, resulting in Al(OH)₃ precipitation [36] . After consuming dissolved oxygen in pits and crevices, a hydrolysis reaction of Al³⁺ ions, Al³⁺ + 3H₂O → Al(OH)₃ + 3H⁺ will be induced, and pH in the pits and crevices subsequently reduces. In acidic water in the pits and crevices promotes the corrosion of Al and dissolution of Al(OH)₃. At the same time, ZnO deposits which cover the crevices should be also dissolved by the acidic water, ZnO + 2H⁺ → Zn²⁺ + H₂O [37] . It is known that Nb₂O₅ is soluble in neutral water [38] , and as shown in Table 2, however, NbO_x/Al/glass electrodes have higher electrode potential than Al/glass substrate. It is hence supposed that NbO_x deposits are less corroded than Al films on glass. If Al films are corroded and pH in crevice water reduces, NbO_x deposits which cover the crevices should not be dissolved, because Nb ions in NbO_x deposits are almost present with valency of 5+, and Nb₂O₅ is insoluble and passive for acidic water [38] . In these ways, the elution of Al and Zn only from ZnO/Al/glass electrodes is explainable based on the pitting and/or crevice corrosions and the subsequent pH change of water in the pits and crevices. The passivation of NbO_x deposits is also described based on the potential-pH diagram.

As shown in Figure 5, the maximum power density, P_max is not proportional to the deposition time of NbO_x deposits, and the maximum P_max is obtained at the deposition time of 17 s. As shown in Figure 9, the optical absorption increases almost linearly with increasing the deposition time of NbO_x, and shift of the optical band gap is not observed, suggesting that the valence state and atomistic structure of NbO_x deposits remain unchanged even in the longer deposition time.

As above mentioned, during light irradiation, electrons flow from the counter electrode into the photovoltaic electrode through the external circuit, and at the same time, very small and few bubbles are formed. Moreover, Al elution from the counter electrodes is observed in the ZnO- and NbO_x-based PECs. Hence, electrochemical reactions given in Equations (1) and (2) are suggested, that is, reduction of nitrate ions and H₂ generation by reduction reactions at the photovoltaic electrodes and Al elution and O₂ generation by oxidation reactions at the counter electrode.

In the case of NbO_x deposits, the electrons consumed in the reduction reactions are probably provided from the NbO_x deposits, in which the electrons are excited into the conduction band by absorbing the irradiated light. It is also supposed that the reduction reactions occur at the interface between NbO_x deposits and electrolyte

solution. However, the maximum power density is not proportional to the NbO_x deposition time, and it is hence suggested that the reduction reactions occur only at the restricted surface of the NbO_x deposits. It is conse- quently considered that most of the electrons excited by optical absorption are consumed to recombine with the holes created at the same time, and the residual holes are presumably filled with the electrons supplied from the external circuit through the Al film on the substrate electrode.

NbO_x is probably an n-type semiconductor as well as Nb₂O₅. When a semiconductor electrode is soaked in an electrolyte solution with different electrostatic potential, a potential bending occurs in the electrode surface, which induces migrations of electrons and holes. In case of the NbO_x-based PEC, it is expected that electrons are concentrated at the surface of NbO_x deposits, and if present, holes migrate toward the underlying Al film. In the case of nano-islands, the thickness is too small to bend the surface potential, resulting in almost no potential gradient, which is similar to a flat-band state. In flat-band state, mobility and migration distance of electrons and holes are quite small, and it is hence expected that only the holes produced very near the Al film recombine with electrons which are supplied through the Al film. It is consequently supposed that the photoelectrochemical reactions proceed in the immediate vicinity of the boundary among NbO_x nano-island, Al film, and electrolyte solution.

Then, the boundary is estimated from the two-dimensional AFM phase images given in Figure 6, in which the boundary between the NbO_x deposits and the underlying Al film also means the boundary with the electrolyte solution. Figure 12 shows the boundary length obtained from an image analysis of Figure 6. For comparison, projected area of the NbO_x deposits was estimated from Figure 6, and the results are also given in Figure 12. The boundary length becomes the longest at the NbO_x deposition time of 17 s, which is consistent with the deposition time obtaining the maximum photocurrent density (J_sc) and power density (P_max) shown in Figure 5. The NbO_x projected area increases continuously with increasing the deposition time, and the slope changes at around the deposition time of 40 s, suggesting the change in growth direction of the islands from horizontal to vertical.

As shown in Figure 12(a), boundary length decreases continuously with increasing the NbO_x deposition time > 17 s. In J_sc and P_max, however, second maxima are observed at the NbO_x deposition times of 100 and 150 s, respectively (Figure 5). As mentioned, NbO_x deposits grow in the vertical direction at the deposition time > 40 s, and it is hence supposed that the thickness of NbO_x islands and films increases continuously. In these specimens, NbO_x deposits should have enough thickness to yield potential bending, from which electrons and holes generated by photo irradiation are able to migrate in the thickness direction and reach the outermost surface and innermost bottom surface of NbO_x deposits, resulting in the photoelectrochemical conversion.

Figure 13 shows the photoelectrochemical conversion mechanism proposed for the NbO_x-based PECs. In the case of nano-islands, it is supposed that the photoelectrochemical reactions in the photovoltaic electrodes take place just around the boundary, and hence most of the deposits far from the boundary should have little contribution to the electric power generation. In the case of film, it is also suggested that the photoelectrochemical conversion reactions occur at the film surface, and the reactions, however, should be restricted at part of the

surface, because the maxima in J_sc and P_max of nano-island state are larger than those of film state (Figure 5) even though the absorption of irradiated light, that is, the amount of electrons excited into conduction band increases continuously with increasing the NbO_x deposition time (Figure 9). To improve the power density, it is necessary to increase the boundary length by suppressing the growth of islands parallel to substrate to prevent the contact between the islands despite increasing the number of islands and promoting the growth of islands perpendicular to substrate to achieve the potential bending to improve the carrier mobility.

In the case of ZnO, various photo corrosion reactions by holes have been proposed.

Han et al. [39] :

Rao et al. [40] :

Fruhwirth et al. [37] :

A corrosion reaction, Al + 3h⁺ → Al³⁺ is also assumed [12] . As above mentioned, the elution of Zn and Al from ZnO/Al/glass electrode is surely observed, and the elution rate increases during the light irradiation, which is probably due to the photo corrosion reactions concerning holes.

When combined with the results from the AFM observation (Figure 8), it is suggested that ZnO nano-islands and underlying Al film are once eluted due to the photo-induced electrochemical corrosions, and the eluted Zn²⁺ and Al³⁺ ions are promptly re-precipitated as hydroxides of Zn(OH)₂ and Al(OH)₃ due to the change in pH of the electrolyte solution, which is due to the OH⁻ forming and H⁺ consuming reactions given in Equation (1). At the deposition time of 3 s, the precipitates seem to cover all the surfaces after 9th light irradiation (Figure 8(c), Figure 8(f)). It is known that Zn(OH)₂ easily dehydrates to change into ZnO [41] , and it is therefore supposed that part of the Zn(OH)₂ precipitates change into ZnO. In case of the deposition time of 3 s, the dehydrated ZnO layer is probably thin but is enough to form potential bending so that electrons and holes can migrate between underlying Al film and electrolyte solution, resulting in photocurrent generation after repeating photo irradiation (Figure 3). As for the ZnO-based PEC at the deposition time of 17 s, the Zn(OH)₂ insulating layer must be thicker than the case of 3 s deposition, and hence photocurrent generation does not happen even though the ZnO layer is formed on the Zn(OH)₂ layer.

According to a potential-pH diagram of Nb system [38] , Nb₂O₅ is passive and insoluble in acidic water but is soluble in neutral and alkaline water. According to Equation (1), pH of electrolyte should be increases with proceeding the photoelectrochemical reactions. Actually, slight increase in pH ~0.2 is confirmed after the photocurrent measurements. Nb⁵⁺ ions are probably dominant in NbO_x deposits, and hence NbO_x deposits should be dissolved in the electrolyte solution. However, Nb elution is not confirmed by ICP measurement. It is therefore supposed that some protection layer is formed on the surface of NbO_x deposits. According to the potential-pH diagram, NbO₂ and NbO with lower valence states are passive and insoluble at the entire pH region. In the present PECs, the NbO_x/Al/glass electrodes are negatively polarized, and electrons are concentrated at the electrode surface. It is consequently expected that some of Nb⁵⁺ ions are reduced to 4+ or lower valence state to form a highly passive layer, from which high photo corrosion resistance of NbO_x/Al/glass electrodes should be achieved.