1. Introduction

Hydrogen has high-energy capacity and exhibits environment-friendly properties. If produced from a renewable source such as solar energy, hydrogen can be used as a main energy carrier and play an important role in the green economy. Hydrogen generation through photo-assisted electrolysis of water, or photoelectrolysis, using a photoelectrochemical (PEC) cell has been an attractive research topic since Fujishima and Honda produced H₂ and O₂ using titanium dioxide (TiO₂) photoanodes under UV light irradiation in 1972 [1]. However, at least two problems hinder the feasibility of water splitting through the PEC process. One issue is that photoelectrodes easily corrode in acidic or alkaline solutions [2, 3]. Another problem is that the solar-to-hydrogen conversion efficiency of photoelectrodes is insufficient for practical application because the usable solar spectrum is limited [4]. A number of studies have been conducted on hydrogen generation through the PEC process with GaN photoelectrodes, as the band-edge potentials of these photoelectrodes are suitable for overall water splitting. Additionally, these photoelectrodes are potentially resistant to corrosion in aqueous solutions [5–7]. However, GaN only absorbs UV light. Hence, only about 5% of the solar spectrum can be absorbed. Accordingly, PEC cells with photoelectrodes that contain In_xGa_1-xN-based materials have attracted much attention [8–11] because the band-edge potentials of In_xGa_1-xN-based materials can satisfy water splitting conditions [5]. Moreover, the band gap of In_xGa_1-xN can vary from 0.7 eV to 3.4 eV by modifying the content of In to fit most of the terrestrial solar spectrum [6, 9, 12]. However, the development of high-quality In_xGa_1−xN absorption layers with a low band gap (< 2.0 eV) and large thickness remains difficult because of the large lattice mismatch between In_xGa_1−xN and GaN [13]. Highly strained In_xGa_1−xN layers with a high In mole fraction were found to contain V-shaped surface pits or screw treading dislocations, both of which lead to lower surface mobility [14]. For PEC devices, the dense surface pits cause a significant recombination of photogenerated carriers with charged defects, thereby lowering photocurrent and gas generation rate [15]. In contrast to the aforesaid approach, this study incorporated manganese (Mn) atoms into the GaN epitaxial layers to form the intermediate band (IB) within the band gap of GaN to extend the absorption wavelength to the visible region [16, 17]. We demonstrated that the photocurrent density and hydrogen generation rates could be enhanced using PEC cells with Mn-doped GaN photoelectrodes. This concept is similar to the IB solar cells (IBSCs), which aim to boost further the efficiency of the single-gap solar cells using materials with IB [18]. The detailed experimental process and results are presented in the succeeding sections.

2. Experiments

The samples used in this study were all grown on c-face (0001) double-polished sapphire (Al₂O₃) substrates through metal organic vapor phase epitaxy. First, low-temperature GaN nucleation layer with a thickness of 30 nm was deposited on the sapphire substrates, followed by 2-µm-thick unintentionally doped GaN (u-GaN) buffer layer and a 1-µm-thick Si-doped GaN (n-GaN) epitaxial layer. The n-GaN epitaxial layer had a carrier concentration of approximately 1.0 × 10¹⁹ cm³. The wafers that consisted of the aforesaid GaN layers were referred to as the GaN templates. Under visible light irradiation, the experimental wafers that consisted of Mn-doped GaN layers grown on the GaN templates for PEC experiments were used to clarify whether the Mn-doped GaN could drive the water-splitting reaction and enhance the photocurrent density [19]. The wafers consisting of Mn-doped GaN layers were labeled as PEC-I. For comparison, the wafers with undoped GaN layers grown on the GaN templates (Mn-free GaN) were also prepared (labeled as PEC-II). Figure 1(a) and 1(b) show the schematic layer structures of PEC-I and PEC-II, respectively. As shown in both figures, the bilayer metal of Cr (50 nm)/Au (200 nm) was deposited on the exposed n-GaN epitaxial layer to form Ohmic contacts after the Mn-doped GaN was selectively removed through inductively coupled plasma etching [20] because the resistivity(electron concentration) of Mn-doped GaN was significantly larger(lower) than that of n-GaN layer. Therefore, the aforesaid Ohmic contacts allowed most of the photo-generated electrons to transport in n-GaN layer (under the Mn-doped GaN) during the PEC reactions, and eventually the electrons were collected by the Cr/Au metal electrodes.

 

Fig. 1 Schematic layer structures of the working electrodes (a) with Mn-doped GaN layer (PEC-I) and (b) without Mn-doped GaN layer (PEC-II).

Download Full Size | PDF

A spectrophotometer (Hitachi U-4100) was used to measure the transmittance of epitaxial wafers. A potentiostat (Autolab-PGSTAT128N) was used to supply the external bias and to measure the current density for the evaluation of the electrical properties of the PEC cells. A platinum (Pt) wire was used as the counterelectrode. A 300 W Xe lamp was used as the light source, and 1 M NaCl was used as the electrolyte. In addition, we used a potentiostat and a monochromator (Jobin Yvon-TRIAX 320) to measure the spectral responses of the photocurrent.

3. Results and discussions

Figure 2 shows the typical photocurrent density as a function of external bias voltage (V_CE). V_CE was the applied voltage between the GaN photoelectrode (i.e., working electrode) and the Pt counterelectrode. As shown in Fig. 2, the photocurrent densities of PEC-I were markedly higher than that of PEC-II by at least one order. For a water-splitting system using PEC cells, the higher magnitude of the photocurrent density indicates faster PEC reaction rate and larger amount of generated H₂. In this study, the photocurrent densities increased with the increasing external applied voltage because the driving force of the photogenerated carriers increased with the applied voltages [21]. In addition, an increased photocurrent density was observed when the Mn atoms were added into the GaN layers. This result can be attributed to the Mn-related intermediate levels within the band gap of GaN (i.e., PEC-I) that allowed more photons to be absorbed, thereby generating more electron-hole pairs under illumination. Unlike in PEC cells made of Mn-free GaN photoelectrodes, the up-conversion of photons may occur in Mn-doped GaN photoelectrodes, thereby contributing excess electron-hole pairs in the PEC cells. The spectral responses of PEC-I and PEC-II were measured by combining a 300 W Xe arc lamp with a calibrated monochromator as light source to clarify whether the Mn-doped samples could absorb photons with energies below the band gap of GaN (Eg _GaN).

 

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

Download Full Size | PDF

Figure 3 shows the spectral responses of PEC-I and PEC-II; 1 M NaCl solution at zero bias was used as electrolyte. As shown in Fig. 3(a), PEC-II showed a typical characteristic with a cutoff at around 365 nm. This result was similar to those of the typical GaN-based p-i-n PDs [22] that had high UV (365 nm)-visible (450 nm) rejection ratio. In addition to the absorption in the UV region, the Mn-doped GaN samples (i.e., PEC-I) exhibited a significant response between 380 and 600 nm. To clarify further whether this visible response resulted from the up-conversionof photons, a long-pass filter with a cutoff at 400 nm was placed in front of the light source to ensure that the incident photon energies were less than the Eg _GaN. As shown in Fig. 3(b), no detectable responsivity close to the background of the detection system was observed in the PEC-II. However, PEC-I exhibited a significant spectral response in the visible light range (400 nm < λ < 600 nm), indicating that photons with energy below Eg _GaN (λ > 400 nm) were absorbed by the Mn-doped GaN photoelectrodes to generate the photocurrent. This below-band gap spectral response verified the aforesaid contention that incorporating Mn in GaN induces considerably deep levels or IB within the band gap to cause the transition between the band edge and Mn-related states.

 

Fig. 3 Typical spectral responses of PEC-I and PEC-II (a) without a 400 nm long-pass filter and (b) with a 400 nm long-pass filter.

Download Full Size | PDF

To elucidate further whether the visible response originated from the Mn-related states to the conduction or valence bands, the transmission spectra obtained from the Mn-doped GaN layers (PEC-I) were grown on a double-polished sapphire substrate. For comparison, n-GaN layers without Mn doping (PEC-II) grown on double-polished sapphire substrate were also prepared. As shown in Fig. 4 , PEC-II exhibited a typical spectrum with a sharp transition edge at around 365 nm (i.e., the band gap of GaN). However, PEC-I clearly showed a strikingly different transmission spectrum. The inset in Fig. 4 shows a schematic band diagram of GaN with Mn energy levels and possible absorption routes (labeled as absorptions A, B, and C). The absorption edge around 365 nm (absorption A) that corresponded to 3.4 eV of energy was caused by the transition from the valance band to the conduction band. The broad band (absorption B) with a threshold of 650 nm corresponded to an energy of around 1.9 eV and a tail extending to the band gap of GaN [23]. The presence of absorption B could be attributed to the electrons undergoing excitation from the Mn-related IB in the band gap, which originated from the Mn-induced structural defect and Mn impurity levels, to the conduction band. Meanwhile, absorption C centered at around 830 nm corresponding to an energy of 1.5 eV [23] that could be attributed to the transitions between the valance band and the Mn-related IB. Both A and B absorptions observed in the transmittance spectroscopy of the Mn-doped GaN films were consistent with the spectral response of PEC-I. However, as the incident wavelength as longer than 700 nm, the responsivity of the PEC-I maintained a flat spectral response. In other words, absorption C in the Mn-doped GaN (Fig. 4) was not reflected on the spectral response of PEC-I. This result could be attributed to fact that the deep Mn-related states filled with electrons because the Mn-doped GaN had a n-type property. In other words, the Fermi level above the IB left a few available states to mediate the electronic transition from the valance band to the Mn-related IB. Although absorption B left behind holes in the Mn-related IB, the heavier holes allowed the injection of excess electrons from the external circuit to maintain electrical neutrality, thereby filling up the Mn impurity band with electrons. A similar mechanism was revealed in a GaN-based heterojunction phototransistor [24]. By contrast, the transmission spectrum of Mn-doped GaN (PEC-I) was measured under bias-free conditions. Therefore, the aforesaid mechanism of excess carrier injection was absent, and absorptions B and C were simultaneously observed in the transmission spectrum of PEC-I. In addition, the photogenerated holes in the Mn-related IB with a short lifetime barely contributed to the photocurrent because the carriers recombined with the defect-related trap states before reaching the electrodes. Therefore, in the spectral response of PEC-I, the cutoff wavelength corresponding to absorption C in the transmission spectrum was absent. Based on the measured responsivity and transmittance, the internal quantum efficiency was estimated to be around 61% under 450 nm light illumination for the PEC-I. In this study, H₂ was generated at the Pt electrode when PEC-I was applied at 1.2 V under visible light illumination. By contrast, no detectable H₂ was observed in PEC-II even when the applied voltage was higher than 4 V. The Mn-doped layer in PEC-I was substantially corroded during hydrogen generation because the surface of Mn-doped GaN was textured after the PEC reactions. Owing to the n-type property, photo corrosion is an inevitable problem for the use of Mn-doped layer to serve as photoanodes for water splitting. In other words, it is necessary to make a compromise between photocurrent density (i.e., hydrogen generation rate) and the stability (i.e., the lifetime) when the hydrogen generates on the photoanode. Concerns about the stability of semiconductor working electrodes, the Mn-doped p-GaN served as photocathode may be a solution to prevent the working electrodes from photo corrosion.

 

Fig. 4 Transmittance spectra of GaN epitaxial layers grown on double-polished sapphire substrate. The inset shows the schematic energy diagram for GaN with Mn-related intermediate band.

Download Full Size | PDF

4. Conclusions

We demonstrated the visible light response of PEC cells made of Mn-doped GaN photoelectrode. The transmittance spectra indicated that Mn-doped GaN could absorb UV and visible light because of the Mn-related IB that formed in the bad gap of the GaN crystal. Based on the spectral responses of Mn-doped GaN photoelectrodes, the Mn-related absorption could generate electron-hole pairs and create photocurrent under visible light illumination. The photocurrent densities of Mn-doped GaN photoelectrode PEC cells were remarkably higher than that of the undoped GaN under visible light illumination. The Mn-related IB in GaN facilitated hydrogen generation through the photoelectrolysis of water under visible light illumination.