1. Introduction

At present, hydrogen generation from water photoelectrolysis has been receiving increasing attention due to environmental issues and the energy crisis. Direct photoelectrolysis of water from solar power is a promising method of hydrogen generation because solar energy is a renewable energy source [1, 2]. Water, the product of hydrogen burning, will not cause pollution and can be returned to the photoelectrolysis system via an appropriate feedback mechanism. In 1972, Fujishima and Honda successfully achieved hydrogen generation with titanium dioxide (TiO₂) photoanode under ultraviolet light irradiation; they were the first to demonstrate the transformation and storage of solar energy in the form of hydrogen through a photoelectrochemical (PEC) water splitting reaction [3]. However, at least two problems about PEC water splitting need to be addressed: (1) photoelectrodes were easily corroded in acidic or alkaline solution [4, 5]; and (2) photoelectrodes have an insufficient solar-to-hydrogen conversion efficiency for practical application because of they are limited by usable solar spectrum [1]. Recently, In_xGa_1-xN-based materials for PEC water splitting [6–9] have been investigated because their band-edge potentials can satisfy the conditions for water splitting [10] and they are potentially resistant to aqueous solutions [11]. In_xGa_1-xN materials with a band gap ranging from 0.7 to 3.4 eV can satisfy most of the terrestrial solar spectrum [7, 11]. Although n-type In_xGa_1-xN materials have been used as photoelectrodes for PEC water splitting, they corroded during hydrogen generation [6–8, 12]. Fortunately, p-type semiconductors that serve as photocathodes for water splitting are not susceptible to photocorrosion, since the cathodic current may, to some extent, protect the surface of the semiconductor from oxidation [13]. For this reason, p-type semiconductors can be expected to be more stable and have more potential for water splitting than n-type semiconductors. Fujii et al. reported that the photocurrent density of p-type GaN is less than that of n-type GaN by approximately one order. From the view of thermal dynamics, both n- and p-type GaN should have equivalent efficiencies because their band-edge potentials are the same. However, the fast recombination process of the photogenerated carrier in p-GaN leads to small photocurrent density [14]. They considered that the suppression of carrier recombination may increase the photocurrent of the p-GaN photoelectrode. According to our previous studies, the resistivity of photoelectrodes is a key factor that directly affects the magnitude of the photocurrent [15]. Thus, a decrease in resistivity of the photoelectrodes leads to an increase of the photocurrent. We know that the resistivity of p-type GaN is greater than that of n-type GaN due to the low activation efficiency of Mg dopants, so that there is lower collection efficiency of the photo-generated carriers and smaller photocurrent. During the water splitting process, the photo-generated holes and electrons transport to the external circuit through the p-GaN and n-GaN photoelectrodes, respectively. Since photo-generated holes have shorter diffusion length than electrons, they have difficulty reaching the external circuit through p-GaN photoelectrodes. Gas generation rate with n-GaN photoelectrodes have been shown to be markedly improved by the use of immersed Cr/Au Ohmic contacts [16]. Indium tin oxide (ITO) is a well-known transparent and conductive optical film [17, 18]. In this study, we aimed to overcome the inherent problem of p-GaN and enhance the hydrogen generation rate by utilizing p-GaN as photocathodes with immersed finger-type ITO ohmic contacts.

2. Experiments

In this study, Mg-doped GaN (p-GaN) epitaxial layers were grown on a c-plane sapphire substrate. Before the growth of p-GaN epitaxial layers, a 30-nm thick low-temperature GaN nucleation layer and a 2 μm-thick undoped GaN were in subsequently grown on the sapphire substrate. The p-GaN epitaxial layers had a carrier concentration of ~5.8 × 10¹⁷/cm³ and a thickness of 0.2 μm. To determine whether immersed finger-type ITO ohmic contacts enhance photocurrent density and hence gas generation rate, we designed two different p-type GaN photoelectrodes for PEC water-splitting experiments. PEC1 consisted of the p-GaN photoelectrode with finger-type ITO ohmic contacts covered with a SiO₂ layer to prevent the ohmic contact from contacting the electrolyte and generating current leakage. Figure 1(a) and 1(b) show the schematic diagram and the cross-section of a local area of PEC1, respectively. The schematic structure of the ITO ohmic contact and SiO₂ protection layer on the p-GaN layer are also shown. PEC2 was composed of p-GaN photoelectrodes with ITO ohmic contacts outside of the working area, so that the ITO ohmic contacts were not immersed in the electrolyte during PEC experiments (Fig. 1(c)). For PEC1, the space between the ITO stripe contacts was 200 μm and the width of each stripe contact was 20 μm. Furthermore, two p-GaN photoelectrodes (e.g. PEC3 and PEC4) with layer thickness greater than that of PEC1 were also prepared. The p-GaN epitaxial layers of PEC3 and PEC4 had thicknesses of 1.5 and 2.5 μm, respectively. The photoelectrodes of PEC3 and PEC4 had the same schematic structure as PEC1. A potentiostat (Autolab-PGSTAT128N) was used to supply the external bias and to measure the current density in order to evaluate the electrical properties of the PEC cells. A platinum (Pt) wire was used as counterelectrode. A 300 W Xe lamp was used as light source and 1 mol/L NaCl was used as electrolyte.

 

Fig. 1 Schematic diagram of the photoelectrochemical cells (a) PEC1, (c) PEC2. Figure 1(b) shows the schematic structure of ITO/SiO₂ staked layers on the p-GaN.

Download Full Size | PDF

3. Results and discussions

Figure 2 shows photocurrent density as a function of external bias voltage (V_CE). The V_CE refers to the applied voltage between the p-GaN working electrode and the Pt counterelectrode. With an illumination intensity of 1.38 W/cm², the photocurrent densities with the applied voltages ranged from 0.4 to −2.0 V. Photocurrents of PEC1 were markedly higher than that of PEC2 when the negative bias voltages were higher than −0.4 V. This result can be attributed to the beneficial effect of immersed finger-type ITO ohmic contacts of PEC1 to the collection of photo-generated holes. The photocurrent density of PEC1 was 96-fold of that of PEC2 when the illumination intensity and V_CE were 1.38 W/cm² and −0.6 V, respectively. The enhancement of photocurrent density increased rapidly with the increase of applied bias. For instance, when V_CE was increased to −1 V, the photocurrent density of PEC1 increased significantly to 464-fold of that of PEC2. This finding can be attributed to the uniform distribution of the applied fields in PEC1 over the working area due to the immersed finger-type ITO ohmic contacts, thereby leading to shorter transit time of the holes compared with PEC2 [16]. This observation also implies that immersed finger-type ITO ohmic contacts are indeed pathways for photo-generated holes, since the holes in PEC1 can reach the external circuit more readily under the same bias. Although immersed finger-type ITO ohmic contacts could effectively enhance the photocurrent of PEC cells with p-GaN as photocathode, the cells still required considerable external bias to generate enough gas. When light intensity was 1.38 W/cm², the critical biases for hydrogen generation were −1.2 and −3 V for PEC1 and PEC2, respectively (Table 1 ). These findings are consistent with the bias-dependent photocurrents shown in Fig. 2. In principle, p-type photocathodes should not be susceptible to photocorrosion [13], so that illumination intensity on the photocathodes can be increased to improve photocurrent density and hydrogen generation rate. Clearly, the critical bias for hydrogen generation was reduced from −1.2 to −0.6 V when illumination intensity was increased from 1.38 to 1.83 W/cm².

 

Fig. 2 Typical photocurrent density-bias curves of the experimental PEC cells.

Download Full Size | PDF

Table 1. Critical Bias for Hydrogen Generation of the Experimental PEC Cells

View Table

As mentioned above, photocurrent densities could be increased obviously under more dense illumination. In fact, p-GaN photoelectrodes associated with neutral electrolyte (e.g., NaCl solution) did not suffer from the corrosion during the photoelectrolysis process because of the reduction reaction rather than the oxidation reaction. Therefore, worrying about the consumption of the p-GaN layer due to corrosion may be not be necessary when gas generation rate is enhanced by increasing light illumination intensity. Furthermore, the immersed ITO finger ohmic contacts can overcome the inherent weakness of high resistivity, low mobility, and short diffusion length of carriers in p-GaN. Our findings suggest that p-In_xGa_1-xN/ITO photoelectrodes associated with neutral electrolytes is a promising scheme for hydrogen generation from water splitting. The design of the immersed ITO finger ohmic contacts can reduce the critical applied bias for hydrogen generation from −3 to −1.2 V. However, the photocurrent observed from PEC1 or PEC2 was not high enough for the massive photoelectrolysis of water. In general, we expect that photocurrent can be increased by moderately increasing the thickness of p-GaN layer to enable the absorption of more photons.

However, experimental results were not consistent with this expectation; thicker p-GaN resulted in lower photocurrent density and higher dark current density. Figure 3 shows the photocurrents of PEC with different layer thicknesses of p-GaN. Obviously, PEC1 with a p-GaN layer thickness of 0.2 μm exhibited relatively higher photocurrent compared with PEC3 and PEC4, which had layer thickness of 1.5 and 2.5 μm, respectively. In this study, PEC working electrodes with p-GaN layer were all grown on unintentionally doped-GaN(u-GaN) buffer layers that exhibited n-type conductivity with an electron concentration of 5 × 10¹⁶ cm⁻³. The absorption coefficient of GaN is ~5 × 10⁴ cm⁻¹ when incident light wavelength is around 365 nm [19]. For PEC1, in addition to carrier generation by light absorption in the p-layer region, the incident photos could also generate carriers in the u-GaN layer because of the thin p-type layer. Although the incident depths of photons in the PEC samples were the same, photons could not reach the u-GaN layer of PEC3 and PEC4 when incident light wavelength was less than or equal to 365 nm because the p-GaN layer thicknesses exceeded 1 μm. On the other hand, photo-generated electrons tend to be transported vertically through the p-GaN layer into the electrolyte, thereby leading to hydrogen generation at the electrolyte/p-GaN interface. However, the quality of p-GaN is less than that of the u-GaN layer, so that the diffusion length of photo-generated carriers in the p-GaN layer is shorter than that in the u-GaN layer. In PEC3 and PEC4, photo-generated carriers were almost completely produced in the p-GaN layer and were transported therein. The inset in Fig. 3 shows the dark I-V characteristics of the experimental PEC samples. For a given bias, the dark current increased with greater layer thickness of the p-GaN layer. Due to the relative lower material quality of the p-GaN layer, photo-generated carriers might recombine with defect-related states before they reach the electrolyte and/or ITO electrode contacts. For PEC1, most photo-generated carriers were generated in the u-GaN underlying layer and were transported therein due to the thin p-GaN layer. Therefore, PEC1 exhibited relatively higher photocurrent and lower dark current than PEC3 and PEC4 (Fig. 3).

 

Fig. 3 Photocurrent and dark current densities as a function of applied bias V_CE of PEC1, PEC3, and PEC4.

Download Full Size | PDF

In theory, the reduction reaction at the p-GaN working electrode should protect it from corrosion during the PEC reaction. To evaluate the morphologies of the p-GaN working electrodes, scanning electron microscopy (SEM). Figure 4(a) and 4(b) correspond to the p-GaN surfaces of PEC1 before and after hydrogen generation, respectively. No detectable change between p-GaN surfaces before and after hydrogen generation was observed. To further inspect p-GaN surface morphology, atomic force microscopy (AFM) images were also taken from the surface of p-GaN. The insets of Fig. 4(a) and 4(b) show the top-view AFM images. The root-mean-square (RMS) roughnesses of p-GaN surfaces for PEC1 before and after hydrogen generation were 0.45 and 0.48 nm, respectively. Only a tiny change between p-GaN surfaces before and after hydrogen generation was observed. In contrast, n-GaN photoelectrodes have a short lifetime due to corrosion during hydrogen generation [6–8, 12]. Findings of this study suggest that the ITO contacts on p-GaN epitaxial layers scheme can serve as potential reliable photocathodes for hydrogen generation from the photoelectrolysis of water.

 

Fig. 4 SEM images of the surface of PEC1 (a) before and (b) after photoelectrochemical measurements. The insets of Fig. 4(a) and 4(b) show the top-view of the AFM images of PEC1 before and after photoelectrochemical measurements, respectively.

Download Full Size | PDF

4. Conclusions

In this study, we utilized immersed ITO finger ohmic contacts to overcome the inherent weaknesses of high resistivity and poor carrier mobility of p-GaN. Immersed ITO finger ohmic contacts were beneficial to the collection of photo-generated carriers, leading to higher photocurrent density and lower critical bias for hydrogen generation. In addition, we found that the thickness of the p-GaN layer is not the key factor for improving photocurrent during water splitting reaction. Photocathodes with the ITO/thin p-GaN/u-GaN scheme did not suffer from significant corrosion and the photocurrent density increased remarkably with higher illumination intensity. Our findings indicate that the use of p-GaN working electrodes with finger-type ITO contacts is a promising approach for hydrogen generation using PEC cells.