Gallium nitride (GaN) is expected as a photoanode to generate oxygen and hydrogen through a water splitting reaction. This is because the band-gap energy of GaN is 3.4 eV (λ ≤ 365 nm) and the top of the valence band is lower than the oxidation potential of water and the bottom of the conduction band is higher than the reduction potential of protons. However, some holes generated in the light-absorbing layer of a GaN-based photoanode are used not in the water oxidation reaction but rather in the etching reaction due to self-oxidation of the photoanode. As a result, the solar-to-hydrogen conversion efficiency η STH decreases over time. To improve η STH and prevent degradation of the GaN-based photoanode surface, it has been reported that supporting NiO on a GaN-based photoanode surface by spin coating using a metal organic decomposition solution is effective. The NiO/GaN-based photoanode has been evaluated under intermittent light irradiation for over 100 h [1]. To prevent the etching reaction from progressing, we have proposed to use NiO thin film as a protective layer for the photoanode on a sapphire substrate [2-4]. When we used this NiO thin film/GaN-based photoanode, the photocurrent density was 76% of the initial value after 100 h of continuous light irradiation [3]. Since some etch pits were observed after this time, we speculate that the decrease in photocurrent is related to the occurrence of etch pits. A transmission electron microscope (TEM) observation indicated that the etch pits are caused by dislocations that grew from the GaN-based layer to the surface. In this study, we investigated the photoelectrochemical stability to light irradiation for several hundred hours when using a GaN-based layer that was more crystalline than those in the previous studies as a photoanode.Al0.1Ga0.9N (AlGaN)/n-GaN heterostructures were grown by metal organic chemical vapor deposition on n-GaN substrate. Ni thin film (thickness: 1 nm) was fabricated on the AlGaN surface by vacuum evaporation. The sample was then heat-treated at 563 K for 60 min in air to form NiO thin film (thickness: 2.5 nm). The NiO thin film/AlGaN/n-GaN and Pt wire were immersed in 1 mol/L NaOH as the photoanode and cathode, respectively. The photocurrent was measured with a potentio-galvanostat (Solartron Analytical, 1287A) under UV light irradiation at 2.2 mW/cm2 (λ ≤ 365 nm). A Xe lamp (Asahi Spectra, MAX-303 with SHX450 filter) was used as the light source. The irradiated sample area was 1 cm × 1 cm.The photocurrent density 1 min after irradiation was 0.195 mA/cm2. The photocurrent density decreased with irradiation time and was 0.126 mA/cm2 at 300 h. However, in this experiment, the light intensity decreased by 30% after 300 h because the Xe lamp deteriorated with operating time. In order to correct the light intensity effect on the photocurrent density, the light intensity was adjusted to the initial value and the photocurrent was measured after continuous operation. The resulting photocurrent density at 300 h was 0.174 mA/cm2, which was 89% of the initial value. The surface and sectional structure of the photoanode before and after 300 h of irradiation were observed with a scanning electron microscope in order to investigate the structural change caused by the etching reaction. Most of the surface of the photoanode maintained its initial structure, though some hexagonal etch pits were observed. The etch pits occupied 8% of the surface area. The elemental analysis for pit depth direction was performed with TEM and an energy dispersive X-ray spectrometer in order to conform the layer structure at area with no pits and bottom of pit. The no-pit areas consisted of NiO, AlGaN, and n-GaN layers, while the AlGaN layer was not observed at the bottom of the pit. Therefore, the NiO and AlGaN layers occupied 92% of the photanode surface area. From these results, we consider that the loss of the effective reaction area due to the increase in etch pit density is the reason for the decrease in the photocurrent.[1] K. Ohkawa et al., Jpn. J. Appl. Phys., 52, 08JH04 (2013).[2] Y. Ono et al., Book of , ICARP, p4-17 (2017).[3] Y. Ono et al., The 85th ECSJ Spring Meeting, 1Q21 (2018) in Japanese.[4] K. Kumakura et al., The 79th JSAP Autumn Meeting, 19p-PA4-22 (2018) in Japanese. Figure 1