1. Introduction

Graphene consists of a single layer of carbon, and graphene oxide (GO) is an oxidation derivative of graphene. Graphene is intrinsically a zero-gap semi-metal, and the oxygen functional groups of GO open its gap [1, 2]. The bandgap formation leads to some useful applications [3–6]. We would like to explore the potential of GO for solar water splitting by combining it with other existing structures for water splitting. Solar water splitting can produce H₂ and O₂ without using the external bias or sacrificial reagents. The generated H₂ can be used in fuel cells for the ultimate clean and sustainable energy [7–9]. GO has been combined with cadmium sulfide (CdS) as the photocatalysts for water splitting [10]. In this study, we want to incorporate GO with a more promising photocatalytic material, copper oxide (Cu_xO), since Cu is abundant and nontoxic [11]. In the past, only reduced GO (r-GO) was combined with Cu_xO for water splitting, and r-GO acted as the metal [12, 13]. We wondered if GO, with an opened gap, could contribute the additional absorption (band gap of GO is much different from Cu_xO as described later) when combined with Cu_xO. However, the reflectance spectra analyzed in this study showed that the reflectance of visible light, which comprised the majority of white light, didn’t decrease with GO deposition (which meant that GO didn’t contribute to additional absorption). The H₂ evolution rate still obviously increased, and it should result from another mechanism. We suspect that it comes from the passivation effect contributed by GO, and it’s proved by the photoluminescence (PL) study.

2. Experiments

The schematic flow diagram to fabricate samples in this study is shown in Fig. 1. The detailed procedure to obtain the base W/Au/Cu_xO structure before GO deposition is described in [14]. First, a tungsten (W) plate was cleaned successively in acetone, ethanol, and de-ionized water. An Au layer was deposited onto the W substrate by evaporation, and then a Cu layer was sputtered by plasma bombardment onto the Au layer. The thicknesses of the Au film and Cu film were both 20 nm. The sample was annealed at 190°C for 10 hours in air to introduce inter-diffusion between Au and Cu and to oxidize Cu into Cu_xO. The inter-diffusion made the Au film become nanoporous, which induced the surface plasmon assisted absorption and declined the surface reflectance [14]. Therefore, the promising base W/Au/Cu_xO structure with plasmonic nanoporous Au was adopted.

 

Fig. 1 The schematic flow diagram to fabricate the Cu_xO/GO sample. Control Cu_xO sample without the procedure of GO deposition was also fabricated for comparison.

Download Full Size | PDF

After the demonstration of the base W/Au/Cu_xO structure, the GO was prepared in order to deposit GO on the base substrate. The first step to prepare GO was to obtain the graphite oxide powder by the modified Hummers method [15]. 0.03g of graphite oxide powder and 20 mL DI water were stirred at 500 rpm for 30 minutes; later, the mixture was treated by ultrasonic shock and centrifugation. Only the upper solution after centrifugation was collected. Ultrasonication and centrifugation were repeated, and then the tiny flakes of GO were obtained in the suspension. The GO suspension was finally dropped on the base W/Au/Cu_xO structure and dried naturally.

Cu_xO includes copper (I) oxide (cuprous oxide, Cu₂O) and copper (II) oxide (cupric oxide, CuO). Cu₂O, which is one of the major oxides of copper, has a cubic crystal structure and a direct bandgap of ~2 eV [16]. CuO, in contrast, has a monoclinic structure and an indirect bandgap of 1.4 eV [17]. We would like to investigate the contribution of GO, which has a larger band gap (2.3~2.7eV [4]), to assist water splitting by Cu_xO.

3. Photocatalytic performance

Scanning electron microscope (SEM) was used to observe the surface morphology of the samples. Figure 2(a) shows the morphology of sample W/Au/Cu_xO which is uniformly distributed with Cu_xO. In Fig. 2(b), the SEM image of W/Au/Cu_xO/GO reveals randomly aggregated GO flakes on Cu_xO. Elemental analysis shows increases in both C and O amounts in the sample Cu_xO/GO (the W/Au/Cu_xO/GO structure) in Fig. 2(d) as compared to sample Cu_xO (the W/Au/Cu_xO structure) in Fig. 2(c).

 

Fig. 2 (a) SEM image of the Cu_xO structure (b) SEM image of the Cu_xO/GO structure (c) and (d) are the elemental analysis of Cu_xO and Cu_xO/GO, respectively.

Download Full Size | PDF

Raman spectroscopy with a 632.8 nm He-Ne laser was used to characterize the GO flakes (on a p-type Si substrate), sample Cu_xO/GO, and sample Cu_xO as shown in Fig. 3(a). The curve of the GO flakes revealed the common D (~1350 cm⁻¹), G (~1580 cm⁻¹ for the undoped case, and large wavenumber for the doped case), and 2D (~2650 cm⁻¹) bands, and each band was related to different properties of the GO layers [3, 18]. The vibration of sp3 defects is referred to as D band, and the Raman result demonstrates the presence of a high density of defects in GO due to oxidation. For the Cu_xO curve, two groups of peaks could be observed. One group of peaks located at around 300, 340, 600cm⁻¹ as shown in Fig. 3 (b), which was the enlarged view of Fig. 3(a), and these peaks were close to the reported peaks in the previous literature [19,20]: The peaks at 300 and 340 cm⁻¹ came from CuO, and the peak centred at 600 cm⁻¹ was ascribed to Cu₂O. Both CuO and Cu₂O existed in the sample. The other group was a broad packet centred at ~2500 cm⁻¹. We hypothesized that this peak was due to the fluorescence emission from Cu_xO, and this could be proved with the PL measurement, as discussed later. When the Cu_xO sample was deposited by GO as shown in the Cu_xO/GO curve, the Raman intensity increased. This meant that there was interaction between Cu_xO and GO. The original D and G peaks of GO could still be observed in the Cu_xO/GO curve.

 

Fig. 3 (a) The Raman spectra of GO, Cu_xO and Cu_xO/GO. (b) the enlarged view of (a) from 200 cm⁻¹ to 800 cm⁻¹.

Download Full Size | PDF

Both samples, Cu_xO/GO and Cu_xO, were put directly into water without adding any sacrificial reagent. The photocatalytic reaction was performed in a cylindrical quartz cell installed with a water cooling device and a 300 watt white light lamp (Philips). Water displacement caused by generation of H₂ and O₂ was recorded. Quantitative amounts of generated H₂ and O₂ were analyzed by gas chromatography (GC), and GC showed that the ratio of H₂ evolution to O₂ evolution was 0.7 in the first 15 min for the Cu_xO photocatalyst in water. The H₂ evolution was then estimated by water displacement multiplied by the ratio of evolution of H₂ to evolution of H₂ and O₂ obtained by GC. Figure 4 shows H₂ evolution rates of samples with and without GO. The initial H₂ evolution rates of the sample Cu_xO/GO and Cu_xO were 140 and 100 μmol/hr, respectively, for a device area of 0.25cm². The enhancement is as large as 40%.

 

Fig. 4 The H₂ evolution rates of samples with and without GO. The inset depicts that the negative fixed charge in GO can repel photo-generated electrons from the surface, and reduce the probability of recombination of electrons-hole pairs at the surface.

Download Full Size | PDF

In the past, several groups report the H₂ generation via Cu_xO [21, 22]. Corrosion would occur during the process of water splitting, and the self-oxidation (Cu₂O transferred to CuO) results in the degradation of the ability as a photocatalyst [23,24]. In our experiments, the H₂ evolution rate by Cu_xO decays more than 75% for half hour. Nevertheless, it has been demonstrated that the photostability can be much improved by depositing NiO_x onto Cu₂O [25]. We believe that the stability of our system can be also improved by using a similar approach. Our study finds that the H₂ evolution rate of Cu_xO can be enhanced by drop casting of GO. In the future, such improvement may be applied to Cu_xO in methanol/water solutions [26], Cu_xO composite structures (such as Si NW/Cu₂O [27] and Cu₂O/La₂CuO₄ [28]), or even other p-type semiconductor photocatalysts.

4. Photoluminescence study

Since GO is a semiconductor, it might be assumed that this enhancement is simply due to the additional absorption contributed by GO. However, the results of reflectance spectra of samples Cu_xO and Cu_xO/GO shown in Fig. 5 conflict with this assumption. For wavelengths below 534 nm, the reflectance of Cu_xO and Cu_xO/GO was similar. For wavelengths between 534 to 695 nm, Cu_xO/GO showed even larger reflectance as compared to Cu_xO. This indicates that the GO doesn’t contribute additional absorption for the most part of white light. We then inferred that the H₂ evolution enhancement is contributed by the passivation effect due to the negative fixed charge in GO [29]. The negative charge in GO will repel the minority electrons of Cu_xO (p-type as mentioned in [14]) far from the surface as shown in the inset in Fig. 4. More photo-generated electrons can be collected by the metal instead of recombining at the Cu_xO/water surface. In order to prove the mechanism, PL study was performed.

 

Fig. 5 Reflectance spectra of the Cu_xO and Cu_xO/GO. GO doesn’t contribute additional absorption for the most part of white light.

Download Full Size | PDF

Figure 6 compares PL spectra for both Cu_xO and Cu_xO/GO. Both peaks centre at the wavelength of ~750 nm. Since both samples show PL peaks at the same position, it indicates that the peaks come from the common semiconductor material, Cu_xO. CuO owns an indirect bandgap of 1.4 eV. It would correspond to a PL band-edge emission with a cutoff wavelength of 886 nm. It matches the cutoff shown in Fig. 6. Therefore, it is CuO rather than Cu₂O contributing to this PL peak. The intensity of PL of Cu_xO/GO can be enhanced as much as 2.85 fold as compared with Cu_xO. The absorbance of our GO suspension is shown in Fig. 7. The main absorption region is UV rather than visible light. Therefore, we can infer that the excitation wavelength of 632.8 nm in the PL measurement was mainly absorbed by Cu_xO instead of GO. Therefore, the PL intensity increase of Cu_xO/GO in Fig. 6 is not mainly due to increased absorption contributed by GO, but due to the lifetime of excited electron-hole pairs being increased. The probability of non-radiative recombination at the surface is reduced by the negative fixed charge in GO, and more electron-hole pairs can then contribute to radiative recombination.

 

Fig. 6 PL spectra of samples Cu_xO and Cu_xO/GO. The intensity of PL of Cu_xO with GO can be enhanced as much as 2.85 fold as compared with Cu_xO without GO.

Download Full Size | PDF

 

Fig. 7 The absorbance of GO suspension. The main absorption region is UV rather than visible light.

Download Full Size | PDF

The excitation wavelength of both Raman and PL measurement locates at 632. 8 nm, and the positive wavenumber shift in the Raman spectrum in Stock mode corresponds to a red shift in the PL peak. The PL peak of Cu_xO is ~750 nm (red shift as compared to 632. 8 nm) which matches the broad packet at ~2500 cm⁻¹ in the Raman spectrum. The broad packet in the Raman spectrum comes from the fluorescence emission of Cu_xO.

It is interesting why the PL-intensity enhancement ratio (2.85 fold) between Cu_xO/GO and Cu_xO is much larger than the hydrogen-evolution-rate enhancement ratio (1.4 fold) between Cu_xO/GO and Cu_xO. The hydrogen evolution is mainly contributed by Cu₂O instead of CuO in Cu_xO [24]. Cu₂O is a direct bandgap material. On the other hand, the PL peak at 750 nm is contributed by CuO in Cu_xO. CuO is an indirect bandgap material. Indirect bandgap material is more sensitive to the defects than the direct bandgap material because defects (or surface states) can provide the needed momentum in the non-radiative recombination process for the indirect bandgap material [30]. Therefore, the presence of GO repels the minority from the surface, and the non-radiative recombination at the surface states of CuO (indirect bandgap material) is greatly reduced, which increases the PL intensity substantially. For hydrogen evolution by Cu₂O (direct bandgap material), the benefit of reduced non-radiative recombination is shared by radiative recombination and carrier separation for water splitting. Hence, the improvement ratio on water splitting by GO is not as large as PL results.

In [10], GO acted as the cocatalyst with CdS, and GO is the supporting matrix and electron acceptor in the GO-CdS nanocomposites. Their reflectance absorption spectra showed that their GO contributed the light absorption in the range of 500-800 nm. They also found that their GO owned a small bandgap of 1.65 eV as compared with the commonly reported values (2.4-4.3 eV). They attributed their small bandgap to the low oxidation degree of GO in their process. Our absorption and reflectance spectra all demonstrate that our GO owns a common and larger bandgap. Our GO doesn’t obviously contribute to more absorption of the main part of white light (or solar spectrum) in our case. The PL results further confirm the proposed passivation mechanism. One may wonder if our GO acts as the electron acceptor in our Cu_xO/GO system just like the case in [10]. However, our PL intensity contributed by Cu_xO doesn’t decrease when GO is loaded (If photo-generated electrons transport to GO, radiative recombination in Cu_xO should decrease). Therefore, the high-oxidation-level GO in our case is not similar to the low-oxidation-level GO as the electron acceptor in [10]. The GO in [10] behaves like the r-GO in [12,13].

5. Conclusion

GO is an effective material to passivate Cu_xO in various applications, such as photocatalysts and light emission. If Cu_xO is utilized in H₂ production, GO can obviously improve the evolution rate. If Cu_xO is targeted to be developed as a light source such as light emitting diodes, GO can increase the illuminating intensity. Such outstanding enhancement from a low cost process makes GO a promising assisting material for Cu_xO.