E ﬀ ects of In Situ Co or Ni Doping on the Photoelectrochemical Performance of Hematite Nanorod Arrays

Featured Application: The material reported in this article can be used as a photoanode of a photoelectrochemical cell that splits water into hydrogen and oxygen. In this way, solar energy can be converted into hydrogen energy. Abstract: Co-doped and Ni-doped hematite ( α -Fe 2 O 3 ) nanorod arrays were prepared on ﬂuorine-doped tin oxide (FTO) conductive glass via aqueous chemical growth, in which the doping and the formation of nanorods occurred simultaneously (i.e., in situ doping). These samples were characterized by X-ray di ﬀ raction (XRD), scanning electron microscopy (SEM), ultraviolet (UV)–visible spectrophotometry, linear sweep voltammetry and Mott–Schottky (M–S) measurement. Results showed that the introduction of 5% Co or Ni into α -Fe 2 O 3 (the molar ratio of dopant to Fe is 1:20) did not change its crystal phase, morphology, energy gap and ﬂat band potential. Both the undoped and the doped α -Fe 2 O 3 showed a direct band gap of 2.24 eV, an indirect band gap of 1.85 eV, and a ﬂat band potential of − 0.22 V vs. saturated calomel electrode (SCE). At an applied potential of 0.2 V vs. SCE, the Co-doped and the Ni-doped α -Fe 2 O 3 exhibited a photocurrent of 1.28 mA / cm 2 and 0.79 mA / cm 2 , respectively, which were 2.1 times and 1.3 times that of the undoped α -Fe 2 O 3 . After the Co or Ni doping, the charge carrier concentration increased from 1.65 × 10 25 m − 3 to 3.74 × 10 25 m − 3 and 2.50 × 10 25 m − 3 , respectively. Therefore, the increase in the photocurrent of the doped α -Fe 2 O 3 was likely attributed to their enhanced conductivity.


Introduction
Photoelectrochemical (PEC) water splitting is a promising technology for the direct conversion from solar energy to hydrogen energy. Since Fujishima and Honda proposed electrochemical photolysis of water at a TiO 2 electrode in 1972 [1], extensive studies have been carried out in this field. Photoelectrodes, which are the key component of a PEC system, should meet the following requirements: (1) appropriate energy levels for conduction band and valence band; (2) effective separation of photogenerated electron-hole pairs; (3) resistance to chemical-and photo-corrosion. Various semiconductor materials such as TiO 2 [2,3], α-Fe 2 O 3 [4][5][6], ZnO [7,8], WO 3 [9][10][11], SrTiO 3 [12,13], BiVO 4 [14,15], etc. have been examined for their potential as photoelectrodes. Among them, α-Fe 2 O 3 shows great potential due to its favorable band gap (2.1-2.2 eV), abundant natural reserves, non-toxicity and high chemical stability. Theoretically, its maximum photocurrent density is 12 mA/cm 2 and the maximum energy efficiency is 15% at 1.23 V vs. RHE (reference hydrogen electrode) under one solar irradiation (AM1. 5, In this study, in situ Co or Ni doped α-Fe 2 O 3 nanorod arrays will be prepared on a conductive substrate through aqueous chemical growth followed by annealing under air. The impacts of the Co or Ni doping on the crystalline phase, the morphology, the light absorption and the PEC activity of α-Fe 2 O 3 nanorod arrays will be investigated.

Materials Preparation
Undoped α-Fe 2 O 3 nanorod arrays were grown on fluorine-doped tin oxide (FTO) conductive glass by an aqueous chemical growth method [20]. In a typical process, a sheet of FTO glass (15 mm × 25 mm) was placed in a glass vial of 25 mL, leaning against the inwall of the vial as displayed in Figure 1. An aqueous solution containing 0.1 M FeCl 3 and 1 M NaNO 3 was prepared and its pH was adjusted to 1.25 by HCl. Then a proper amount of the reactant solution was injected into the vial, leaving part of the FTO substrate uncovered. The vial was sealed and heated at 95 • C for 10 h. After reaction, the FTO substrate was rinsed by deionized water and precipitates on its non-conductive side was removed. Last, the FTO substrate was annealed in a muffle furnace at 500 • C for 1 h. Co-doped and Ni-doped α-Fe 2 O 3 nanorod arrays were prepared in a similar process except that small amounts of CoCl 2 or NiCl 2 were added into the aqueous solution. The molar ratio of dopants to photoelectrode materials reported in the literature was normally in the range of 1% to 10% [2,30]. Thus, the molar ratio of the dopants to Fe here was set at 1:20 (i.e., 5%).
Appl. Sci. 2020, 10, x FOR PEER REVIEW 3 of 13 or Ni doping on the crystalline phase, the morphology, the light absorption and the PEC activity of α-Fe2O3 nanorod arrays will be investigated.

Materials Preparation
Undoped α-Fe2O3 nanorod arrays were grown on fluorine-doped tin oxide (FTO) conductive glass by an aqueous chemical growth method [20]. In a typical process, a sheet of FTO glass (15 mm × 25 mm) was placed in a glass vial of 25 mL, leaning against the inwall of the vial as displayed in Figure 1. An aqueous solution containing 0.1 M FeCl3 and 1 M NaNO3 was prepared and its pH was adjusted to 1.25 by HCl. Then a proper amount of the reactant solution was injected into the vial, leaving part of the FTO substrate uncovered. The vial was sealed and heated at 95 °C for 10 h. After reaction, the FTO substrate was rinsed by deionized water and precipitates on its non-conductive side was removed. Last, the FTO substrate was annealed in a muffle furnace at 500 °C for 1 h. Co-doped and Ni-doped α-Fe2O3 nanorod arrays were prepared in a similar process except that small amounts of CoCl2 or NiCl2 were added into the aqueous solution. The molar ratio of dopants to photoelectrode materials reported in the literature was normally in the range of 1% to 10% [2,30]. Thus, the molar ratio of the dopants to Fe here was set at 1:20 (i.e., 5%).

Materials Characterization
Crystalline structures of the as-prepared samples were detected by X-ray diffraction (XRD, Rigaku D-MAX/2550, Tokyo, Japan) with Cu Kα radiation. The microscopic morphology of the samples was characterized by scanning electron microscopy (SEM, JEOL JSM−6360LV, Tokyo, Japan). The optical absorption property of the samples was analyzed by an ultraviolet (UV)-visible spectrophotometer (Shimadzu UV-2401PC, Kyoto, Japan). The instrument was calibrated with blank FTO conductive glass so that the UV-vis absorption spectra of the as-prepared α-Fe2O3 samples alone can be achieved.

Photoelectrochemical Measurements
The preparation of α-Fe2O3 photoanodes is illustrated in Figure 1. After the α-Fe2O3 was formed, the blank of the FTO substrate was stuck to a Cu sheet by silver conductive adhesive. Then the margin and the back of the FTO substrate was coated by insulation paste to achieve an active area of around 1 cm × 1 cm.
A three-electrode cell system was assembled for the PEC measurement as Figure 2 shows. The as-prepared α-Fe2O3 photoanode acted as working electrode, a Pt sheet (1 cm × 1 cm) as a counter electrode, and saturated calomel electrode (SCE) as a reference electrode. 1 M KOH solution was

Materials Characterization
Crystalline structures of the as-prepared samples were detected by X-ray diffraction (XRD, Rigaku D-MAX/2550, Tokyo, Japan) with Cu Kα radiation. The microscopic morphology of the samples was characterized by scanning electron microscopy (SEM, JEOL JSM−6360LV, Tokyo, Japan). The optical absorption property of the samples was analyzed by an ultraviolet (UV)-visible spectrophotometer (Shimadzu UV-2401PC, Kyoto, Japan). The instrument was calibrated with blank FTO conductive glass so that the UV-vis absorption spectra of the as-prepared α-Fe 2 O 3 samples alone can be achieved.

Photoelectrochemical Measurements
The preparation of α-Fe 2 O 3 photoanodes is illustrated in Figure 1. After the α-Fe 2 O 3 was formed, the blank of the FTO substrate was stuck to a Cu sheet by silver conductive adhesive. Then the margin and the back of the FTO substrate was coated by insulation paste to achieve an active area of around 1 cm × 1 cm. A three-electrode cell system was assembled for the PEC measurement as Figure 2 shows. The as-prepared α-Fe 2 O 3 photoanode acted as working electrode, a Pt sheet (1 cm × 1 cm) as a counter electrode, and saturated calomel electrode (SCE) as a reference electrode. 1 M KOH solution was used as electrolyte. The PEC cell was continuously purged with Ar gas when it was in operation. Illumination of AM1.5 and 100 mW/cm 2 was provided by a solar simulator (America Newport Oriel69911). Linear sweep voltammetry (LSV) test was performed on an electrochemical workstation (Shanghai Chenhua CHI660B, China). The scanning rate was 5 mV/s and the scanning range was from −0.3 to 0.6 V vs. SCE. The Mott-Schottky (M-S) test was also carried out on an electrochemical workstation (Princeton Applied Research Company PARSTAT2273, TN, USA) at a frequency of 1000 Hz.
Appl. Sci. 2020, 10, x FOR PEER REVIEW 4 of 13 used as electrolyte. The PEC cell was continuously purged with Ar gas when it was in operation. Illumination of AM1.5 and 100 mW/cm 2 was provided by a solar simulator (America Newport Oriel69911). Linear sweep voltammetry (LSV) test was performed on an electrochemical workstation (Shanghai Chenhua CHI660B, China). The scanning rate was 5 mV/s and the scanning range was from −0.3 to 0.6 V vs. SCE. The Mott-Schottky (M-S) test was also carried out on an electrochemical workstation (Princeton Applied Research Company PARSTAT2273, TN, USA) at a frequency of 1000 Hz.  Figure 3a shows XRD profiles of the samples obtained from aqueous chemical growth without annealing. The broad diffraction peak around 20° is assigned to amorphous SiO2 (JCPDS 51-1593), which is the matrix of FTO glass. The diffraction peaks marked with bold dots belong to SnO2 (JCPDS 46-1088), a conductive film on the glass matrix. In addition to these background peaks, characteristic diffraction peaks of crystalline β-FeOOH (JCPDS 34-1266) also exist. Figure 3b displays XRD profiles of the samples after annealing at 500 °C for 1 h. By comparing Figure 3a,b, it was found that annealing led to the transformation from β-FeOOH to α-Fe2O3 (JCPDS 33-064). Such a phase transformation was also indicated by the color change of the sample from yellow to red (see Figure  4). In order to find out whether the α-Fe2O3 are oriented, the intensity ratio of each characteristic peak relative to the (1 0 4) peak, which shows the largest intensity, was calculated, and compared with the standard data in JCPDS 33-064. The two sets of data are almost consistent with each other, indicating the α-Fe2O3 are not oriented. Figure 3 also shows that the XRD profiles of the Co-doped and the Ni-doped samples are almost the same as that of the undoped one. Neither dopant-related phases nor a shift of the α-Fe2O3 diffraction peaks was observed in the XRD profiles of the doped samples. This suggests the doping of Co or Ni has few influences on the crystal structure of α-Fe2O3. This is probably because Co or Ni is successfully doped into the crystal lattice of α-Fe2O3 due to their similar atomic structures and the amount of the dopants was small. Similar results have been reported by others [36,39]. Hou et al. [36] believed that Co could be easily incorporated into the crystal lattice of α-Fe2O3 and hence considered Co to be a suitable dopant for α-Fe2O3. Li et al. [39] synthesized α-Fe2O3 nanowires on a FTO substrate and diffused Sn from FTO to α-Fe2O3 by high-temperature sintering. Their SEM and XRD results revealed that neither the structure nor the crystal phase of the α-Fe2O3 changed after Sn was incorporated.  Figure 3a shows XRD profiles of the samples obtained from aqueous chemical growth without annealing. The broad diffraction peak around 20 • is assigned to amorphous SiO 2 (JCPDS 51-1593), which is the matrix of FTO glass. The diffraction peaks marked with bold dots belong to SnO 2 (JCPDS 46-1088), a conductive film on the glass matrix. In addition to these background peaks, characteristic diffraction peaks of crystalline β-FeOOH (JCPDS 34-1266) also exist. Figure 3b displays XRD profiles of the samples after annealing at 500 • C for 1 h. By comparing Figure 3a,b, it was found that annealing led to the transformation from β-FeOOH to α-Fe 2 O 3 (JCPDS 33-064). Such a phase transformation was also indicated by the color change of the sample from yellow to red (see Figure 4). In order to find out whether the α-Fe 2 O 3 are oriented, the intensity ratio of each characteristic peak relative to the (1 0 4) peak, which shows the largest intensity, was calculated, and compared with the standard data in JCPDS 33-064. The two sets of data are almost consistent with each other, indicating the α-Fe 2 O 3 are not oriented. Figure 3 also shows that the XRD profiles of the Co-doped and the Ni-doped samples are almost the same as that of the undoped one. Neither dopant-related phases nor a shift of the α-Fe 2 O 3 diffraction peaks was observed in the XRD profiles of the doped samples. This suggests the doping of Co or Ni has few influences on the crystal structure of α-Fe 2 O 3 . This is probably because Co or Ni is successfully doped into the crystal lattice of α-Fe 2 O 3 due to their similar atomic structures and the amount of the dopants was small. Similar results have been reported by others [36,39]. Hou et al. [36] believed that Co could be easily incorporated into the crystal lattice of α-Fe 2 O 3 and hence considered Co to be a suitable dopant for α-Fe 2 O 3 . Li et al. [39] synthesized α-Fe 2 O 3 nanowires on a FTO substrate and diffused Sn from FTO to α-Fe 2 O 3 by high-temperature sintering. Their SEM and XRD results revealed that neither the structure nor the crystal phase of the α-Fe 2 O 3 changed after Sn was incorporated. Appl. Sci. 2020, 10, x FOR PEER REVIEW 5 of 13   Figure 5 displays the microscopic morphology of the as-prepared β-FeOOH and α-Fe2O3. The β-FeOOH is presented as a nanorod array that vertically grows on the substrate (Figure 5a). The α-Fe2O3 samples also show a similar morphology except for a slight volume contraction (Figure 5b), suggesting a topological transformation from β-FeOOH to α-Fe2O3. When Co or Ni dopant was introduced, the morphology of the α-Fe2O3 remained almost the same as exhibited in Figure 5c,d. The three α-Fe2O3 samples were all comprised of bunches of nanorods. Each bunch of nanorods had a diameter of about 50 nm and each nanorod had a diameter of about 5-10 nm. The small diameter is beneficial to the transport of photoinduced holes to the semiconductor-electrolyte interface, although the α-Fe2O3 nanorods do not have the merit of a high conductivity since they are not oriented in the (0 0 1) direction on the substrate [6,17,39]. after before    after before   Figure 6a shows the UV-visible absorption spectrum of the α-Fe 2 O 3 nanorod array. The absorption band is mainly located between 300 nm and 550 nm. Electron transition on the energy levels of α-Fe 2 O 3 can be divided into direct transition and indirect transition, corresponding to O 2− →Fe 3+ and Fe 3+ 3d→3d, respectively [40]. The energy gaps of direct transition and indirect transition could be estimated by Equations (4) and (5), which were derived from Equations (2) and (3). The plots of (Ahv) 2 or (Ahv) 0.5 against hv (i.e., Tauc plots) of the undoped α-Fe 2 O 3 are displayed in Figure 6b.

Effects on Ultraviolet (UV)-Visible Absorption
The intersection of the linear extension of the Tauc plots with x axis represents the energy gap of direct transition or indirect transition [41]. They are 2.24 eV and 1.85 eV, respectively, in the present work. The energy gap of the direct transition estimated by the Tauc plot is close to its theoretical value (2.1-2.2 eV). Vayssieres et al. [42] observed that the energy gap of α-Fe 2 O 3 nanorods increased by 0.3-0.6 eV compared with α-Fe 2 O 3 bulk due to the quantum confinement effect of nanorods. However, the quantum confinement effect was not observed in this study.   Figure 6a shows the UV-visible absorption spectrum of the α-Fe2O3 nanorod array. The absorption band is mainly located between 300 nm and 550 nm. Electron transition on the energy levels of α-Fe2O3 can be divided into direct transition and indirect transition, corresponding to O 2− → Fe 3+ and Fe 3+ 3d→3d, respectively [40]. The energy gaps of direct transition and indirect transition could be estimated by Equations (4) and (5), which were derived from Equations (2) and (3). The plots of (Ahv) 2 or (Ahv) 0.5 against hv (i.e., Tauc plots) of the undoped α-Fe2O3 are displayed in Figure  6b. The intersection of the linear extension of the Tauc plots with x axis represents the energy gap of direct transition or indirect transition [41]. They are 2.24 eV and 1.85 eV, respectively, in the present work. The energy gap of the direct transition estimated by the Tauc plot is close to its theoretical The UV-visible absorption spectra of the doped samples are not displayed in Figure 6 as they overlap with the undoped one. As the UV-visible absorption spectra are almost the same this suggests that the Co or Ni doping has little influence on the optical absorption of α-Fe2O3. Inducing a red shift of absorption onset is one of the reasons for an enhanced PEC performance [4,36]. However, in this study, this reason was excluded. Similar to this study, other researchers [5,43] also observed the doping of foreign elements did not change the energy gap of α-Fe2O3.  Figure 7 shows the I-V plots of the as-prepared α-Fe2O3 samples. The equilibrium potential of water oxidation (i.e., E(H2O/O2)) is 0.1572 V vs. SCE in 1 M KOH electrolyte. For the undoped α-Fe2O3 in dark, the initial potential for the occurrence of current is 0.36 V, more positive than the equilibrium potential due to the existence of overpotential. When the undoped α-Fe2O3 sample is under solar irradiation, the initial potential moves to −0.22 V, much more negative than the equilibrium potential, indicating an obvious PEC effect. The doped α-Fe2O3 samples exhibit a similar initial potential to the undoped α-Fe2O3 under illumination. The initial potential of photocurrent is generally considered as flat band potential. The three samples have similar flat band potentials, which suggests that the doping of 5% Co or Ni does not change the energy band structure of the α-Fe2O3 nanorod array [42]. This conclusion is consistent with that derived from the Tauc plots (see Section 3.3). The UV-visible absorption spectra of the doped samples are not displayed in Figure 6 as they overlap with the undoped one. As the UV-visible absorption spectra are almost the same this suggests that the Co or Ni doping has little influence on the optical absorption of α-Fe 2 O 3 . Inducing a red shift of absorption onset is one of the reasons for an enhanced PEC performance [4,36]. However, in this study, this reason was excluded. Similar to this study, other researchers [5,43] also observed the doping of foreign elements did not change the energy gap of α-Fe 2 O 3 . Figure 7 shows the I-V plots of the as-prepared α-Fe 2 O 3 samples. The equilibrium potential of water oxidation (i.e., E(H 2 O/O 2 )) is 0.1572 V vs. SCE in 1 M KOH electrolyte. For the undoped α-Fe 2 O 3 in dark, the initial potential for the occurrence of current is 0.36 V, more positive than the equilibrium potential due to the existence of overpotential. When the undoped α-Fe 2 O 3 sample is under solar irradiation, the initial potential moves to −0.22 V, much more negative than the equilibrium potential, indicating an obvious PEC effect. The doped α-Fe 2 O 3 samples exhibit a similar initial potential to the undoped α-Fe 2 O 3 under illumination. The initial potential of photocurrent is generally considered as flat band potential. The three samples have similar flat band potentials, which suggests that the doping of 5% Co or Ni does not change the energy band structure of the α-Fe 2 O 3 nanorod array [42]. This conclusion is consistent with that derived from the Tauc plots (see Section 3.3).

Current-voltage (I-V) Plots
In addition to the initial potential, photocurrent at a certain potential is another index of the PEC activity. At 0.2 V vs. SCE, the photocurrent density of the Co-doped and the Ni-doped α-Fe 2 O 3 are 1.28 mA/cm 2 and 0.79 mA/cm 2 , respectively, which are 2.1 times and 1.3 times that of the undoped α-Fe 2 O 3 . This result suggests the incorporation of Co and Ni into α-Fe 2 O 3 nanorod array could enhance its PEC activity. The photocurrent densities of doped and undoped α-Fe 2 O 3 reported in the literature are summarized in Table 1. In addition to the initial potential, photocurrent at a certain potential is another index of the PEC activity. At 0.2 V vs. SCE, the photocurrent density of the Co-doped and the Ni-doped α-Fe2O3 are 1.28 mA/cm 2 and 0.79 mA/cm 2 , respectively, which are 2.1 times and 1.3 times that of the undoped α-Fe2O3. This result suggests the incorporation of Co and Ni into α-Fe2O3 nanorod array could enhance its PEC activity. The photocurrent densities of doped and undoped α-Fe2O3 reported in the literature are summarized in Table 1. In order to explain why the PEC activity of α-Fe2O3 was enhanced by Co or Ni doping, M-S plots of both doped and undoped α-Fe2O3 samples were measured and shown in Figure 8. All the samples exhibited a positive slope, indicating they are all n-type semiconductors with electrons as majority charge carriers. The slope of the linear part of the M-S plots could be used to estimate the concentration of charge carriers (Nd) according to Equation (6).

Mott-Schottky Plots
In order to explain why the PEC activity of α-Fe 2 O 3 was enhanced by Co or Ni doping, M-S plots of both doped and undoped α-Fe 2 O 3 samples were measured and shown in Figure 8. All the samples exhibited a positive slope, indicating they are all n-type semiconductors with electrons as majority charge carriers. The slope of the linear part of the M-S plots could be used to estimate the concentration of charge carriers (N d ) according to Equation (6).
In this equation, C is the space charge capacitance per m 2 active area, e 0 is the elementary charge (1.6 × 10 −19 C), ε is the dielectric constant of α-Fe 2 O 3 (14.2), ε 0 is the permittivity of vacuum (8.85 × 10 −12 F/m), E is the applied potential, E fb is the potential of flat bands, k is Boltzmann's constant (1.3806505 × 10 −23 J/K), and T is the absolute temperature (298 K).
Appl. Sci. 2020, 10, x FOR PEER REVIEW 9 of 13 In this equation, C is the space charge capacitance per m 2 active area, e0 is the elementary charge (1.6 × 10 −19 C), ε is the dielectric constant of α-Fe2O3 (14.2), ε0 is the permittivity of vacuum (8.85 × 10 −12 F/m), E is the applied potential, Efb is the potential of flat bands, k is Boltzmann's constant (1.3806505 × 10 −23 J/K), and T is the absolute temperature (298 K). The concentration of charge carriers derived from the slope and the Efb derived from the intercept of the M-S plot are listed in Table 2. The Efb of both doped and undoped α-Fe2O3 samples is around −0.22 V, indicating the doping has little influence on the Efb. This conclusion is consistent with that obtained from I-V plots (see Figure 7). The charge carrier concentration of the undoped α-Fe2O3 is 1.65 × 10 25 m −3 , which is in the same order of magnitude as those reported in the literature [5,6,39,44]. After Co or Ni doping, the charge carrier concentration increased to 3.74 × 10 25 m −3 and 2.50 × 10 25 m −3 , respectively, which are 2.3 times and 1.5 times that of the undoped sample. The increase in the charge carrier concentration means the improvement of the electrical conductivity, which may contribute to the enhanced PEC activity of the doped samples. Zhao et al. [45] reported that the heterogeneous interface of nickel and iron oxide exhibited a high electrocatalytic activity for hydrogen evolution reaction in alkaline media. In the present work, whether the Co or Ni doping promotes the electrocatalytic performance of α-Fe2O3 nanorods has not been investigated. It is necessary to figure this question out in future because an improved electrocatalytic activity may be one of the potential reasons for the enhanced photocurrents observed for the doped α-Fe2O3 samples. This question could be solved by comparing the I-V plots of the doped samples and the undoped sample in dark. Furthermore, the higher photocurrent of the Co-doped α-Fe2O3 compared with the Ni-doped sample may be attributed to a larger carrier density (see Figure 7 and Table 2). However, this observation could not justify that the Co doping outperformed the Ni doping since the dopant concentrations were not known. In fact, the PEC activity of a doped photoelectrode depends not only on the identity of the dopant but also on the dopant concentration, distribution and the morphology and microstructure of the doped photoelectrode [46]. In order to compare the effects of The concentration of charge carriers derived from the slope and the E fb derived from the intercept of the M-S plot are listed in Table 2. The E fb of both doped and undoped α-Fe 2 O 3 samples is around −0.22 V, indicating the doping has little influence on the E fb . This conclusion is consistent with that obtained from I-V plots (see Figure 7). The charge carrier concentration of the undoped α-Fe 2 O 3 is 1.65 × 10 25 m −3 , which is in the same order of magnitude as those reported in the literature [5,6,39,44]. After Co or Ni doping, the charge carrier concentration increased to 3.74 × 10 25 m −3 and 2.50 × 10 25 m −3 , respectively, which are 2.3 times and 1.5 times that of the undoped sample. The increase in the charge carrier concentration means the improvement of the electrical conductivity, which may contribute to the enhanced PEC activity of the doped samples. Zhao et al. [45] reported that the heterogeneous interface of nickel and iron oxide exhibited a high electrocatalytic activity for hydrogen evolution reaction in alkaline media. In the present work, whether the Co or Ni doping promotes the electrocatalytic performance of α-Fe 2 O 3 nanorods has not been investigated. It is necessary to figure this question out in future because an improved electrocatalytic activity may be one of the potential reasons for the enhanced photocurrents observed for the doped α-Fe 2 O 3 samples. This question could be solved by comparing the I-V plots of the doped samples and the undoped sample in dark. Furthermore, the higher photocurrent of the Co-doped α-Fe 2 O 3 compared with the Ni-doped sample may be attributed to a larger carrier density (see Figure 7 and Table 2). However, this observation could not justify that the Co doping outperformed the Ni doping since the dopant concentrations were not known. In fact, the PEC activity of a doped photoelectrode depends not only on the identity of the dopant but also on the dopant concentration, distribution and the morphology and microstructure of the doped photoelectrode [46]. In order to compare the effects of different dopants, Co-doped and Ni-doped α-Fe 2 O 3 nanorod arrays with the same dopant concentration and distribution should be synthesized in future.

Conclusions
In situ doping of 5% Co or Ni into α-Fe 2 O 3 nanorod arrays was performed by a facile aqueous chemical growth method followed by annealing at 500 • C. The doping did not change the crystal structure, morphology, energy gap and flat band potential of the α-Fe 2 O 3 nanorod arrays, but enhanced its PEC activity. At 0.2 V vs. SCE, the Co-doped and the Ni-doped α-Fe 2 O 3 exhibited a photocurrent density of 1.28 mA/cm 2 and 0.79 mA/cm 2 , respectively, which were 2.1 times and 1.3 times that of the undoped α-Fe 2 O 3 . The charge carrier concentration of the Co-doped and the Ni-doped α-Fe 2 O 3 were 2.3 times and 1.5 times that of the undoped α-Fe 2 O 3 . Therefore, the increase in the photocurrent density of the doped α-Fe 2 O 3 was likely attributed to the enhanced conductivity. In the future, the following work will be conducted. (1) Materials characterization such as energy dispersive X-ray spectroscopy (EDS), X-ray photoelectron spectroscopy (XPS) or inductively coupled plasma-mass spectrometry (ICP-MS) will be performed in order to provide direct evidence of the incorporation of Co or Ni. (2) The influence of dopant concentrations on the PEC performance will be investigated and the optimal concentration will be obtained. (3) The effects of Co doping and Ni doping will be compared by synthesizing samples with the same dopant concentration and distribution.