Synthesis of a single-crystal Fe2O3 nanowire array based on stress-induced atomic diffusion used for solar water splitting

In this study, we successfully fabricated a single-crystal Fe2O3 nanowire array based on stress-induced atomic diffusion and used this array as the photoelectrode for solar water splitting. With the surface polishing treatment on the sample surface, the density of the Fe2O3 nanowire array reached up to 28.75 wire µm−2 when heated for 90 min at 600°C. The photocurrent density of the optimized sample was 0.9 mA cm−2 at 1.23 V versus a reversible hydrogen electrode in a three-electrode system under AM 1.5 G illumination. The incident photon-to-electron conversion efficiency was 6.8% at 400 nm.


Introduction
A clean and sustainable energy resource is indispensable for human beings because of the massive fossil fuel consumption that comes along with industrial development. Hydrogen is a suitable candidate. However, traditional production methods use fossil fuels to obtain it, which would not eventually solve the energy crisis. Solar water splitting is a clean and sustainable production method to produce hydrogen, because it uses only sunlight and water, which are the two most abundant natural resources. The anode in a solar water splitting system can be a semiconductor material with a suitable bandgap, while the cathode can be a metallic material. Both the anode and cathode are placed in water or an electrolyte solution. When the photoanode is illuminated by sunlight, the photons with an energy larger than the bandgap of the semiconductor will excite the electrons from the valence band to the conduction band. The excited electrons will then move to the cathode and react with the electrolyte solution to produce hydrogen [1].
2018 The Authors. Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited.

Experimental approach 2.1. Nanowire fabrication
A commercial iron plate with 99.95% purity was used as the substrate. The iron plate was 0.1 mm thick and measured 10 × 10 mm 2 . Two types of samples were prepared to investigate the effect of surface roughness on the nanowire growth. One was polished by a rasp, while the others were unpolished. The two types of samples were then put in alcohol and cleaned by an ultrasonic cleaner (AS ONE, ASU-2D). Subsequently, all samples were heated by a ceramic heater (SAKAGUCHI, SCR-SHQ-A) in air. The heating temperatures were changed to determine the optimized condition for the nanowire growth (table 1). microscope (TEM) observation was performed to obtain a high-resolution picture of the single nanowire and identify the crystalline structure of the Fe 2 O 3 nanowires. An X-ray diffraction (XRD) analysis was also performed to study the structure and phase of the nanowire arrays.

Photocurrent measurements
The photocurrent density, incident photon-to-current conversion efficiency (IPCE), and stability measurements were conducted using a three-electrode system. A platinum wire was used as the counter electrode in a 1 mol l −1 NaOH solution electrolyte. A scanning potentiostat (ALS DY2325) was used to measure the photocurrent density. Light was provided by a solar simulator (SERIC XC-100EFSS) with the same spectrum as sunlight. The optical power density was 100 mW cm −2 at the test position equivalent to AM 1.5 G illumination. The IPCE measurements were performed using the same solar simulator with a monochromatic filter from 400 nm to 650 nm.

Results and discussion
3.1. Experimental results Figure 1 shows the SEM image of the Fe 2 O 3 nanowire arrays fabricated at 500°C, 600°C and 700°C with and without the surface polishing treatment. The SEM images showed that the sizes of the nanowire arrays fabricated under different temperatures were very different. Two types of nanowire shapes (i.e. wire-and leaf-shaped nanowires) were found in the sample heated at 500°C (figure 1a,d, respectively). The nanowire length for samples heated at 600°C was much longer than that heated at 500°C. Moreover, the diameter became smaller. The nanowires heated at 600°C on the unpolished sample had a long leaf shape, whereas those on the polished sample had a wire shape with an average diameter of 29 nm (figure 1b,e, respectively). Figure 1c,f shows the nanowires heated at 700°C on the unpolished and polished samples, respectively. The nanowires obtained from both the polished and unpolished samples had a long triangular shape. The average density, length and diameter of the nanowire arrays herein were evaluated based on the SEM observation results (figures 2-4, respectively). Figure 2 depicts that the largest density of 28.75 wire µm −2 was found in the polished sample heated at 600°C for 90 min. This value was five times higher than the density of the unpolished sample (i.e. 5 wire µm −2 ) heated at the same condition. In the sample heated at 500°C, the density of the polished samples was higher than that of the unpolished samples. The nanowire arrays fabricated under all the experimental conditions were uniform, except for the unpolished sample heated at 700°C.
Meanwhile, one-dimensional nanostructures were reported to reduce the path length of the hole transport to improve the charge carrier recombination through a high aspect ratio and large surface area [25]. We compared the length and the diameter of the nanowires fabricated under different conditions (figures 3 and 4, respectively) to find the highest aspect ratio of the nanowire arrays. Consequently, the average length of the nanowires on both the polished and unpolished samples became longer with the increase in the heating temperature. The longest nanowires were obtained at 700°C from the unpolished sample of 3.37 µm. Furthermore, the nanowire length for the polished and unpolished samples was almost the same under the same temperature, indicating that the heating temperature was the key factor of the nanowire length. As for the diameter of the nanowire used for the photoanode in water splitting, a small diameter could reduce the diffusion distance of the photogenerated minority        carriers from the centre to the semiconductor-electrolyte interface [26]. The smallest average diameter of 29 nm was obtained for the nanowires fabricated at 600°C for 90 min from the polished sample. The nanowire arrays fabricated at 500°C showed a wide range of diameters from 26 to 180 nm because the diameter of the wire-shaped nanowire was much smaller than that of the leaf-shaped nanowire.  from the result because the thickness of the formed Fe 2 O 3 layer was smaller than the penetration depth in the XRD observation. Figure 7 shows the investigated photocurrent density of the Fe 2 O 3 nanowire arrays fabricated at different experimental conditions. Under each temperature, the polished samples showed a higher photocurrent density than that of the unpolished samples, which could be explained by the higher density of the Fe 2 O 3 nanowire arrays in the polished sample providing more photogenerated carriers, which consequently enhanced the photocurrent density. The samples heated at 600°C with the surface polishing treatment exhibited the highest photocurrent density of 0.9 mA cm −2 at 1.23 V versus RHE, which was thrice that of the unpolished sample fabricated at the same temperature. This value was higher than that reported in other literatures [19][20][21]. Figure 8 presents the IPCE of the nanowire array fabricated at 600°C from the polished sample, which was 6.8% at 400 nm at 1.23 V versus RHE. This value was higher than that of the polycrystalline Fe 2 O 3 nanowire array that we previously reported (i.e. IPCE of 5.54% at 400 nm) [23]. Moreover, the  IPCE value reported herein was relatively high compared with the other undoped Fe 2 O 3 photoanodes fabricated with metallic substrates, such as undoped Fe 2 O 3 nanoflakes on an Fe foil with an IPCE of 3.43% at 400 nm [27], a haematite nanotube array on an Fe foil with an IPCE of 3.2% at 400 nm [28], and an Fe 2 O 3 nanorod array on a Ti plate with an IPCE of 4.8% at 400 nm [29]. It is considered that the high-density Fe 2 O 3 nanowire array could provide a higher absorption of the light and the single crystal structure is helpful to reduce the recombination of photogenerated electrons and holes happening in the photoanode. However, a nanostructured Fe 2 O 3 photoanode fabricated on the fluorine-doped tin oxide (FTO) substrate also showed a higher IPCE value. Liao et al. [30] reported an IPCE value of 23% at 400 nm. Normally, an FTO substrate is helpful in increasing the photoelectrochemical performance because it could facilitate the electron transport in the photoanode. Therefore, further increasing the IPCE value of the Fe 2 O 3 nanowire array photoanode is possible using an FTO substrate and surface function with dopant doping. Figure 9 illustrates the photocurrent stability. The Fe 2 O 3 nanowire array photoanode fabricated at 600°C from the polished sample was illuminated with 10 s on/off for 300 s by a chopped light. The result denoted a photocurrent density of 0.75 mA cm −2 1.23 V versus RHE during the illumination. Moreover, the photocurrent of the prepared sample quickly changed with the light switched off, thereby showing good photoresponse properties.

Mechanism
The fabrication mechanism of the metal oxide nanowire arrays with similar methods was also reported [31,32]. In our previous study, we studied the nanowire growth mechanism based on oxidation-assisted  stress-induced atomic diffusion. Tensile stress was generated at the Fe plate surface because of the different molar volumes of Fe 2 O 3 and Fe. Thus, a stress gradient was generated from the centre of the Fe plate to the Fe 2 O 3 -Fe interface as the driving force for the nanowire growth [23].
In the present study, the density of the nanowire array fabricated at 600°C on the polished sample was 28.75 wire µm −2 , which was much higher than that of the nanowire array fabricated from the unpolished sample. Figure 10 shows the measured surface roughness of polished and unpolished samples obtained by a white light interferometer (ZYGO new view 6000). The surface roughness of the polished and unpolished samples is 2.459 µm and 0.452 µm on average, respectively. The surface polishing treatment provided three advantages for the nanowire growth. First, the polishing treatment induced an initial compressive residual stress on the Fe plate surface, which obstructed the volume expansion of the oxide layer formed on the sample surface during the heating process. Therefore, a higher effective vertical stress gradient would occur for the same oxide volume expansion, which induces more and faster diffusion of Fe atoms to the sample surface, thereby increasing the nanowire array density. The residual stress has been measured by XRD and calculated by an XRD analyse software (JADE 6.5), as shown in figure 11. The residual stress is −125.3 Mpa in the polished sample and −28.99 Mpa in the unpolished sample, respectively. Second, the polishing treatment will deform the surface layer of the Fe plate, leading to dislocations and intercrystalline failure at the Fe plate surface. Recrystallization will happen if the polished Fe plate is annealed at a high temperature [33]. During the recrystallization process, the grain size of Fe will decrease and the number of grains will increase, which will finally lead to an increase in the oxide volume expansion and the number of weak spots of the Fe 2 O 3 layer. Third, the surface polishing treatment increases the roughness of the Fe plate surface, which could enlarge the contact area between the sample surface and the air, thereby increasing the volume of the Fe 2 O 3 layer on the Fe plate. Compared to that of the unpolished samples, the volume expansion of the Fe 2 O 3 layer for the polished sample was larger during the heating process, which enhanced the tensile stress at the Fe plate surface and increased the driving force for the Fe atomic diffusion ( figure 12). In summary, the surface polishing treatment could provide a larger driving force and increase the number of weak spots, which could eventually increase the nanowire array density.

Conclusion
In this study, high-density single-crystal Fe 2 O 3 nanowire arrays were fabricated based on the oxidationassisted stress-induced atomic diffusion method from the substrate with a surface polishing treatment. The surface polishing treatment could provide a larger driving force and increase the number of weak spots on the sample surface, leading to an increase in the nanowire density. The nanowire array with the highest density of 28.75 wire µm −2 was obtained from the polished sample heated at 600°C for 90 min. A photocurrent density of 0.9 mA cm −2 at 1.23 V versus RHE and an IPCE of 6.8% at 400 nm were also obtained. The stability test results showed good photocurrent response and stability of the single-crystal Fe 2 O 3 nanowire photoanode, which indicated a good potential for the solar water splitting application.
Data accessibility. The datasets supporting this article have been uploaded as part of the electronic supplementary material.