Interfacial phenomena in hematite photoanodes fabricated by directly associating iron oxide suspensions with FTO substrates using a dipping-annealing method

Abstract

The fabrication of solid semiconductor nanoparticles on conductive substrates while retaining their high photocatalytic activity as they were in the dispersed form remains a challenge. In this study, we adopted the idea of used a dipping–annealing (DA) method to associate solid iron oxide nanoparticles directly onto an FTO substrate. We focused on the interfacial phenomenon of the as-fabricated hematite photoanodes to evaluate factors that may affect their photocatalytic performance. A significant sintering effect occurs during calcination; this process converts the iron (hydro)oxide precursor into a hematite structure and removes binder molecules. This sintering effect is more severe for the nanocubes than the nanospheres. However, the sintering effect would induce a size effect, further compromising the photocatalytic performance of the prepared photoanodes. Based on EIS analyses, the deteriorated photocatalytic performance arises from the deactivation that occurs at the exposed facet, increasing the open circuit potential and the resistance of hematite bulk, as well as decreasing the capacitance at the hematite–electrolyte interface.

---

Introduction

Various iron oxide nanoparticles with interesting morphologies have been successfully synthesised.^1–3 Their (photo)catalytic performances were experimentally demonstrated, showing a strong dependency on their morphology, particle size, and architecture.^3–10 However, the development of an appropriate method for preparing photoanodes using these iron oxide nanoparticles that could maintain their performance as they are in the dispersion form remains a challenge. This is because so many factors that could potentially deteriorate the photocatalytic performance would be introduced into the photoanode during the fabrication process. For example, one straightforward fabrication method involves dispersing solids into solvents, followed by annealing. However, the as-obtained film is often so brittle that disintegration occurs, even when gently handled, due to the weak adhesion between the solid particles and the substrate. To enhance the adhesion, some binding molecules, such as hydroxylpropyl cellulose (HPC), are added into the mixture.^11,12 These binding molecules should be capable of complexing with both the iron oxide nanoparticles and the substrate via their hydroxyl groups. Afterwards, these binders are removed during the subsequent calcination process, and the adhesion of the as-obtained photoanode can be greatly improved. However, the adhesion remains inferior to materials that are synthesised via a directly hydrothermal pyrolysis reaction. Meanwhile, although better adhesion can be achieved, morphological control remains difficult when samples are prepared using a hydrothermal pyrolysis reaction.

Consequently, we extended the concept of deposition–annealing (DA) that was originally proposed by Li and coworkers.¹³ The DA method describes a synthesis process during which an FeCl₃ precursor was dissolved in ethanol and subsequently drop-deposited onto an FTO substrate. The FTO substrate was dried under flowing air and was heated to 350 °C, forming an iron oxide precursor layer. The thickness of the film depended on the number of DA cycles and the concentration of the precursor. In this study, we dispersed solid iron oxide nanoparticles including goethite, hematite, and magnetite into a mixture containing HPC, polyethylene glycol (PEG), and H₂O. A hematite photoanode was constructed by dipping an FTO substrate into the above-mentioned mixture before being heated in a box furnace at 200 °C for 15 minutes. This dipping–annealing cycle was repeated from 3 to 13 times; at the end of the DA cycles, the prepared samples were subjected to a two-step calcination treatment (500 °C for 2 hours and then 700 °C for 20 minutes) to convert the iron oxide precursor layer into a hematite structure. The interfacial phenomena were studied and discussed regarding the surface texture, the orientation of the deposited hematite nanoparticles, the open circuit potential, the resistance and capacitance at the grains (hematite), the hematite–electrolyte interface, and the photocurrent density of the as-prepared hematite photoanodes. Our results will facilitate the assembly of functional solid nanoparticles onto a variety of conductive substrates.

Experimental

All chemicals were analytical grade reagents, and deionised water was used throughout every experiment. To synthesise the Fe₃O₄ nanoparticles, 6.756 g of FeCl₃·6H₂O, 3.426 g of FeSO₄·7H₂O (mole ratio of Fe(II)/Fe(III) = 0.5), and 6.0 g of (NH₂)₂CO were dissolved in 100 mL of deoxygenated water with constant magnetic stirring; the mixture was refluxed for 12 hours.¹⁴ The black precipitate was isolated using a magnetic field, washed with deionised water, and stored at 50 °C. To synthesise the hematite nanoparticles, 2 L of aqueous 0.02 M FeCl₃ solution was sealed in a serum bottle and aged at 100 °C for 2 days.¹⁵ The goethite particles were synthesised by mixing 1.39 g of FeSO₄·7H₂O and 1.36 g of CH₃COONa·4H₂O in 50 mL of deionised water; the mixture was aged for 2 days at 60 °C.¹⁶ The yellowish slurry was collected and washed several times with deionised water before being dried at 50 °C. To prepare the Fe-containing precursor solution for the DA process, 0.1 g of individually synthesised goethite, hematite, and magnetite was dispersed in a mixture containing 0.5 g HPC, 0.5 g PEG (mw 2000), 9 mL deionised water and 1 mL pure acetic acid. The precursor solution was magnetically stirred for at least 1 day until it became viscous. For a typical DA process, a pre-cleaned FTO substrate (2 cm × 1 cm) was immersed into the precursor solution with an area of 1 cm × 1 cm for 10 seconds. After dipping, the FTO substrate was leaned against the neck of the vial to allow the excess precursor solution to drip off of the surface via gravity for at least 3 minutes. This step is a very important because the residual precursor solution would generate an unevenly dispersed surface (a corona-like iron oxide deposition). At the end of the DA process, these photoanodes were subjected to a two-step calcination treatment (500 °C for 2 hours and then 700 °C for 20 minutes). The nomenclature follows the example of “g13,” indicating that this photoanode was produced using a goethite precursor solution with 13 cycles of DA. There were seven photoanodes made during each DA process; the one exhibiting a photocurrent closest to the average value [relative standard deviation (RSD): ca. 25%] was chosen as a representative example for further morphological, orientation, and electrochemical impedance spectroscopy (EIS) analyses.

The sample morphologies were imaged using field-emission scanning electron microscopy (FE-SEM; JSM6700F, JEOL, Japan). The X-ray diffraction (XRD) patterns were measured on a D2 Phaser diffractometer (Bruker, Germany) using Cu Kα radiation (λ = 1.5418 Å); the lattice refinement was performed using the built-in Topas 4.2 software (Bruker, Germany). All (photo)electrochemical measurements were conducted using a three-electrode configuration: a hematite photoanode served as the working electrode, a saturated calomel electrode (SCE) was the reference electrode, and platinum foil served as the counter electrode in 1 M NaOH on a CH Instrument 604E electrochemical workstation. The photocurrent density measurements were performed using linear sweep voltammetry at 20 mV s⁻¹. Sunlight was simulated using a 300 W xenon lamp (Spectra Physics) and an AM 1.5G filter (Oriel); the light intensity was calibrated to 100 mW cm⁻². The EIS was performed with a 10 mV bias potential, and the frequency was scanned between 100 kHz and 0.1 Hz at potentials between −0.5 and +0.5 V (versus SCE) in the dark. The obtained data were fitted to an equivalent circuit model using the Zview software.

Results and discussion

Morphological properties

Our preliminary results indicated that the adhesion between the hematite particles and the FTO substrate decreased as the concentration of the precursor suspension increased. Specifically, the hematite film could be destroyed, even when handled gently, after preparation from a suspension with a concentration above 0.1 g iron oxide/10 mL. As suggested by Li and coworkers, the photocatalytic performance would be improved when the hematite photoanode was prepared using a more diluted precursor solution.¹³ Because additional DA cycles would be required to reach the optimal photocatalytic performance while using a more dilute precursor suspension,¹³ a 0.1 g iron oxide/10 mL mixture was selected for this study.

Fig. 1a shows the appearance of the hematite photoanodes prepared using the DA process with the hematite suspension precursor before (top series) and after (bottom series) calcination. The FTO substrate becomes more reddish after successive DA cycles because additional hematite nanoparticles have been deposited onto the FTO substrate. After calcination, all of the samples exhibit a comparable reddish appearance. Specifically, all of the samples possess a typical hematite absorption spectrum, as confirmed by the UV-vis absorption data (not shown). Therefore, the calcination treatment converts the iron oxide precursor into the hematite phase. Importantly, the as-obtained hematite films may not inherit the morphology of their precursor when the particle size is below 100 nm. The m-series samples show a texture combining both rod- and sheet-like solids with sizes of approximately 200 nm (Fig. 1d), while the magnetite precursors are nanospheres 50 nm in diameter (Fig. 1c). The h-series photoanodes exhibit tetragonal solids approximately 500 nm in size (Fig. 1h), whereas the hematite precursors are nanocubes approximately 100 nm in size. In contrast, the surface of the g-series samples is composed of acicular hematite lying on top of the FTO substrate (Fig. 1f). Therefore, the morphology of the goethite precursors (40 nm in width and 500 nm in length, Fig. 1e) has been retained after calcination. Smaller nanoparticles tend to fuse with each other, increasing their size during calcination to reduce their high surface energy. Size increases frequently occur when converting metals and metal oxides into their corresponding chalcogenides, phosphides, and other derivatives. Therefore, conversion reactions are often performed in the presence of organic stabilizers to control the shape, size, and size dispersion of the nanoparticles.¹⁷ Furthermore, the surface activity of nanocubes appears to be much higher than that of the nanospheres because the former was sintered to form much larger particles than the latter during calcination (bottom series). Observed higher surface activity of nanocube than nanosphere agrees with the literature; at which a gaseous ethanol sensor made with a different hematite morphology showed a sensitivity in the following sequence under a given ethanol concentration: oblique parallelepipeds > quasi-spherical structure.⁶ A similar oxidation performance is also reported when quasicubic hematite nanoparticles could catalytically convert almost 100% of the CO at a much lower temperature than those of the nanophases with flower-like, hollow, and other irregular external morphologies.⁸

Fig. 1 (a) A photograph of h-series samples subjected to different DA cycles before (upper) and after (bottom) calcination; (b) the thickness of the obtained photoanodes after different DA cycles; the SEM images of magnetite (c), goethite (e), and hematite (g) precursors (scale bar: 500 nm); and the representative top views of m7 (d), g7 (f), and h7 (h) (scale bar: 500 nm). The inserted panels are the cross-sectional views (scale bar: 1000 nm).

The thickness of hematite film increases linearly with the number of DA cycles (Fig. 1b). The thickness of the hematite film is determined by the cross-sectional view during SEM analysis (10 different places are measured, and the RSD is approximately 30%). Interestingly, for g-series samples, the thickness seems to be independent of the DA cycles, similar to the m- and h-series samples. This behaviour may reflect the deposition of the precursor solids. For smaller iron oxide particles, such as magnetite nanospheres and hematite nanocubes, the aspect ratios are rather small; these particles stack onto the previous layers, filling the voids created by the previous layer. However, the aspect ratio of goethite is so large that the thickness of the g-series samples strongly relied upon the orientation of the previously deposited layer. Specifically, when the previous layer was constructed such that the long side was tilted toward the substrate, additional goethite particles would deposit in a similar orientation, increasing the thickness by only the “width” of the goethite particles. This behaviour explains the sharp increase in the thickness of the g-series samples during the early DA cycles followed by a milder increase afterward. In summary, the sintering effect is critical for photoanode fabrication. Consequently, combining nanoparticles smaller than 100 nm while maintaining their original shape after calcination is difficult. Additionally, nanoparticles with a large aspect ratio, such as wires or rods, tend to deposit such that their largest facet is parallel to the substrate. Therefore, fabricating a vertically oriented wire and rod array via DA remains a challenge.

XRD analysis

After the two-step calcination, all precursors have been completely converted into α-Fe₂O₃ (JCPDS card number 89-2810), as confirmed by the XRD patterns (Fig. 2). For samples subjected to less than 5 DA cycles, the diffractions attributed to the hematite phase are very weak because an insufficient amount of hematite is presented on the FTO substrate. This argument is supported by the observation that the diffraction intensity of the hematite increases after successive DA cycles (DA > 5). Moreover, the hematite film does not have a preferred orientation, as indicated by the appearance of multiple diffractions including the (104), (110), (012), (113), (116), and (214) planes. The orientation of the hematite toward the FTO substrate directly influences the photocatalytic performance of the hematite photoanode. An array of hematite wires or rods that is vertically aligned toward the FTO substrate is a favourable orientation for the photo-generated carriers that diffuse from the hematite particles into the substrate.^18,19 To determine how these iron oxide solids deposit and orient during the DA process, the diffraction intensity of (104)/(110) is arranged and plotted against the DA cycles (Fig. 2d). For the m-series samples, the (104)/(110) ratio is unaffected by the DA cycles due to the spherical geometry of the magnetite precursor (Fig. 1c) because there is no preferred orientation for the spheres. Specifically, the magnetite nanospheres should spontaneously fill the external voids developed by the previous DA process, leaving the orientation unaffected by the following DA process. The (104)/(110) ratio of the g-series remains relatively constant; the same ration for the h-series declines rapidly during the first few DA cycles but seems to be unaffected afterward. The decrease in the (104)/(110) ratio might be the result of fewer exposed {104} facets or additional {110} facets appearing. Similar to the {021} facet, the {104} facet is a high index plane that appears on the truncated cubic and rectangular particles,²¹ such as the acicular goethite precursor. In accordance with our SEM observations, the newly introduced goethite particles would deposit on the previous layer, maintaining the orientation. Conversely, the hematite precursor nanocubes appear randomly orientated during the first few DA cycles, exposing a certain amount of the {104} facet. As the DA process continues, the hematite nanocubes will fill the voids created by the previous DA process, exposing fewer {104} facets. Therefore, the (104)/(110) intensity ratio remains relatively constant after 5 DA cycles.

Fig. 2 The XRD patterns of the hematite photoanodes obtained after different DA cycles with (a) magnetite, (b) goethite, and (c) hematite precursor suspensions; (d) the variation in the diffraction intensity of (104)/(110) against different DA cycles.

Photoelectrochemical behaviour

Fig. 3 displays the photoelectrochemical performance of the hematite photoanodes. Three very distinct and interesting features can be found. First, the photocatalytic activity of the m-series photoanodes is very sensitive to the simulated sunlight illumination, as indicated by the spikes appearing at the light on/off transitions. Second, the photocatalytic performance of the m-series samples is approximately six times higher than those of the g- and h-series. Lastly, none of these samples possess photocatalytic performance as high as those prepared via direct hydrothermal hydrolysis (typical photocurrent density performance of photoanodes prepared from direct hydrothermal hydrolysis falls near mA cm⁻²). Fig. 3d shows that the photocurrent of the m-series samples increases gradually as the DA cycles increase until a maximal photocurrent density of ∼13 μA cm⁻² is reached after 7 DA cycles. Additional DA processes decrease the photocurrent of the m-series samples; after 13 DA cycles, the photocurrent density drops by ∼25% to 9 μA cm⁻². In contrast, the maximal photocurrent density (∼2 μA cm⁻²) appears after 9 and 11 cycles of DA for the h- and g-series samples, respectively.

Fig. 3 The photocatalytic performance of hematite photoanodes obtained after different DA cycles with (a) magnetite, (b) goethite, and (c) hematite precursor suspensions; (d) the variations in the photocatalytic performance of hematite photoanodes against different DA cycles.

The photocurrent is strongly correlated with the thickness and density of hematite films.¹³ For the m-series photoanodes, the initial increase in the photocurrent can be attributed to the increasing amounts of photoactive material, increasing the light absorption and the total number of photoexcited carriers. Increasing film thickness facilitates charge carrier recombination because the carrier diffusion length of hematite is very short. The same argument holds for the h- and g-series samples, even though their photocurrents are much lower compared to the m-series samples. The lower photocurrent of the h- and g-series samples relative to the m-series might arise from the size effect when considering the charge separation and transport at the semiconductor–electrolyte interface.¹³ For small particles, the photoexcited holes collection efficiency is greater in comparison with that of larger particles as the average distance between the holes and the semiconductor–electrolyte interface is shorter. Additionally, the separation of the photoexcited electrons and holes can be improved using the electric field developed by the space-charge layer and the interface dipole at the semiconductor–electrolyte interface. According to the literature, our results confirm that a nanostructured hematite film always shows a much higher photocatalytic performance than does a microstructured one.⁹

In addition to the size effect, the orientation of the hematite particles and exposed facets would influence the photocatalytic performance. The {104} and {110} facets possess high Fe(III) density (10.3²¹ and 10.1⁴ atoms per nm², respectively) due to their ridge-and-valley topography. Unlike the octahedrally coordinated Fe(III) with six oxygen anions presented at the {001} facet, the surface activity of the less coordinated Fe(III) facets will be more active and hence more catalytic.²¹ Additionally, the unique geometry of these low coordinated facets (i.e., the {104} facet exhibits a low angle of ∼78° between the dangling Fe–O and surface μ3-O–Fe bonds)²⁰ renders more catalytic sites accessible, enhancing the overall photocatalytic performance. For example, for the visible light-induced RhB dye photodegradation using different morphologies of hematite, the observed reactivity expressed in terms of exposed facets is as follows: {110} > {012} ≫ {001}.⁴ Therefore, the efficiency of the RhB photodegradation could be improved by exposing additional low coordinated facets.⁹ This rationale is exactly the motive of this study; we would like to understand the interfacial phenomenon fully. This relevant information would allow us to fabricate hematite photoanodes that expose additional low coordinate Fe(III) facets in the future. Unfortunately, the sintering effect appears to be a severe problem; attempts to retain the nanoparticle texture, particularly for nanocubes, after calcination remain on-going. Once the sintering effect is minimised, we would be able to compare the photocurrent performance between different facets of hematite nanoparticles quantitatively.

Electrochemical impedance spectroscopy

Fig. 3 indicates that the size effect might be the primary issue affecting the photocurrent performance of the as-obtained hematite photoanodes. In this section, we examine the interfacial electrochemical behaviours of these hematite photoanodes using electrochemical impedance spectroscopy. We initially observed the varying open circuit potential (OCP) of these photoanodes against time, as shown in Fig. 4. The open circuit potential reflects the degree of surface potential developed via protonation–deprotonation reactions.²² A negative Eocp value is accompanied by a deprotonated (negatively charged) surface that occurs where the pH > the point of zero charge potential (for hematite, ∼pH 9). In an alkaline environment, such as the one at which the OCP was measured (1.0 M NaOH), the hematite particles should be fully deprotonated (≡Fe_xO⁻, where x could be 1, 2, or 3 because our samples are terminated with some low coordinated facets, as indicated by XRD). In Fig. 4, the OCP values are negative, but the equilibrated OCP values are different between the samples. From Fig. 4d, the OCP values of both the g- and m-series samples gradually decrease along with the increasing DA cycles, whereas for h-series samples, the OCP increases as the DA process increases. The OCP decreased as the DA cycles increased because more hematite surfaces are deprotonated. Accordingly, for the h-series samples, the increasing OCP indicates that increasingly fewer hematite surfaces are exposed to the electrolyte even though the surface activity varies between the surfaces. Nevertheless, having additional available surfaces able to make contact with the electrolyte does not guarantee that a higher photocurrent can be achieved (i.e., OCP of m11 < m7; g13 < m13; h3 < m3). Therefore, the size effect related to the facets exposed to the electrolyte is the primary factor affecting the photocurrent performance.

Fig. 4 The open circuit potential of the hematite photoanodes obtained after different DA cycles with (a) magnetite, (b) goethite, and (c) hematite precursor suspensions; (d) the variation in the open circuit potential of the hematite photoanodes after different DA cycles.

To understand the electronic properties of the hematite photoanodes, we recorded the individual electrochemical impedance spectra in the dark after the OCP measurement (at this point, the OCP is equilibrated, and we thus applied a bias equivalent to its corresponding OCP value). To interpret the obtained EIS results quantitatively, the data were fitted with an equivalent circuit proposed by Boily and coworkers (model 1, as shown in the inset in Fig. 5a).²² Because no 45° slope for the impendence in the high frequency region was observed, the Warburg element was removed. We also removed the ohmic resistance parallel to the Warburg element because extracting any meaningful information from two resistances in series is difficult when only one semicircle appeared (model 2).¹⁹ We integrated the diffusion layer capacitance (Cdl) and a CPE in the series of a Rad (referring to the resistance induced by the adsorbed ions) into the CPE_interface to emphasise the capacitance phenomenon for the hematite–electrolyte interface. Accordingly, the remaining elements in the solid-phase include R_hem (referring to the ohmic resistance + the charge transfer resistance inside hematite) and the CPE_hem (the capacitance behaviour from the space charging layer) in our reduced equivalent circuit (model 3). The solution-side contained elements that include the R_interface that is ascribed to the resistance caused by the adsorbed ions and the CPE_interface, accounting for the capacitance behaviour from the diffusion layer. We selected a CPE element rather than the capacitance to account for the imperfect capacitance induced by the existence of an electrical double layer at each interface. The admittance (reciprocal of impedance) of CPE is defined as follows:

Y_CPE = Q₀(ωi)ⁿ (1)

where a value of n = 1 represents an ideal capacitor and a value of 0 describes a pure resistor.²³

Fig. 5 (a) an illustration of the hematite–electrolyte interface and its corresponding equivalent circuit; the obtained electrochemical impedance spectra of the hematite photoanodes obtained by subjecting different DA cycles with (a) magnetite, (b) goethite, and (c) hematite precursor suspensions.

Fig. 5 indicates that the EIS shapes for the m-, g-, and h-series samples are quite different. For the m-series samples, the EIS shapes appear to be a simple depressed semicircle (Fig. 5b). In the g-series samples, the EIS shapes appear to be small semicircles at the high frequency region, and high capacitance behaviour is apparent in the low frequency region (Fig. 5c). For the h-series, the EIS shapes are more complicated and are obviously affected by the number of DA cycles (Fig. 5d).

Fig. 6 shows the variation in the R_hem, R_interface, CPE_hem, and CPE_interface of the hematite photoanodes prepared using various precursor suspensions and different DA cycles. Notably, fitting the low frequency region was quite difficult, and a great uncertainty was introduced. Using g3 and g5 as examples, the low frequency tail might arise from the large resistance at the hematite–electrolyte interface, electrode polarisation, or carrier diffusion. In the latter case, this tail would show a 45° straight tail; a Warburg element could be adopted to account for this semi-infinite linear diffusion. This assignment requires the precise acquisition of more reliable data because these low-frequency data would be rather erratic if no appropriate shielding was applied. From this point of view, we preferred to assign this low-frequency region as a resistor with a CPE in parallel. If this tail stemmed from carrier diffusion, an n value of 0.5 (eqn (1)) could be observed.

Fig. 6 The variation in (a) R_hem, (b) R_interface, (c) CPE_hem, and (d) CPE_interface for the hematite photoanodes obtained from various precursor suspensions against different DA cycles.

The resistance of R_hem in the m-series samples decreases by approximately one order of magnitude as the DA cycles increase. In contrast, the R_hem of the h-series samples appear to gradually decrease, whereas the R_hem of the g-series appears to be less affected by increasing the DA cycles (Fig. 6a). From Fig. 1, the decreasing R_hem could be attributed to the sintering effect. During calcination, small particles sinter to form large ones such that the resistance contribution from the grain boundary decreases.²⁴ The decreased grain boundary resistance also accounts for the lower R_hem of the h-series samples than the m-series because the former possesses larger hematite particles after calcination. However, the relatively unaffected R_hem might reflect the fact that the calcination does not significantly change the particle size when converting goethite into hematite. Importantly, the m-series samples possess a higher resistance than the rest of the samples in most cases, explaining the lower photocurrent of the samples prepared via the DA process relative to those produced using hydrothermal hydrolysis. However, these results also suggest that (samples prepared by the DA process) the resistance of hematite particles prepared using our system is not the most critical factor regarding the photocurrent performance. Regarding the R_interface, the R_interface of both the m- and g-series samples gradually decreases as the DA cycles increase; however, the R_interface of the h-series samples exhibits a converse trend toward the DA cycles (Fig. 6b). The variation in the R_interface shows similar features to that of the OCP (Fig. 4d). This similarity reveals that, in addition to the photocatalytic performance, the surface activity is another factor that is closely correlated with the hematite–electrolyte interface resistance. Therefore, exposing the active surface is very important for reducing the hematite–electrolyte interfacial resistance.

The capacitance of the hematite (CPE_hem) increases as the DA cycles increase for the m- and g-series samples; however, the h-series samples show a converse trend (Fig. 6c). In contrast, the capacitance at the hematite–electrolyte interface (CPE_interface) for the g- and h-series samples is not sensitive toward the DA cycles. However, the m-series samples are highly sensitive to the DA cycles (Fig. 6d). Based on the reduced equivalent circuit model, the capacitance contributed by the space charging layer from the m- and g-series samples is gradually accumulated as the DA cycles increase. The higher capacitance can be regarded as a sign of the numerous mobile charge carriers based on the Mott–Schottky model. However, a higher resistance will compromise the photocurrent performance because it eventually enhances the rate of the photoexcited electron–hole recombination. When comparing Fig. 6a with Fig. 6c, the increased DA cycles are accompanied by the decreasing resistance (R_hem) and the increasing capacitance of hematite (CPE_hem), explaining the higher photocurrent performance of the m-series samples over the others. This concurrent behaviour is an important feature that is frequently observed on highly photocatalytic hematite photoanodes.¹⁹ This behaviour also explains the lower photocurrent of the g- and h-series samples. However, increasing the DA cycles increases the amount of hematite deposited on the photoanode; therefore, a higher capacitance should be observed (additional iron oxide particles should be able to absorb more light) as the DA cycles increase. However, the hematite capacitance (CPE_hem) of the h-series samples displays obviously conflicting features. A highly possible reason might come from the consideration of their hematite–electrolyte interface behaviours. When comparing Fig. 6b with Fig. 6d, the interface resistance (R_interface) dramatically increases, while the capacitance from the diffusion layer (at interface, CPE_interface) gradually decreases. The decreasing capacitance (CPE_interface) at the hematite–electrolyte interface along with the increasing OCP indicated by Fig. 4d can be rationalised as a consequence of the increasingly fewer available active surfaces. Because the (104)/(110) ratio of the m-series remains relatively unaffected by the DA cycles (Fig. 2d), the surface that is exposed to the electrolyte is gradually deactivated such that fewer charge carriers are able to be transported into the bulk hematite. Although additional experiments are necessary, the deactivation at the {104} facet might be responsible for this phenomenon (Fig. 2d).

Conclusion

In this study, we attempted to fabricate hematite photoanodes with a specific surface texture by directly associating different iron oxide particles with distinct morphologies. After calcination, a significant sintering effect occurred such that only the samples fabricated using the goethite suspension inherited their morphology from their precursor. Sintering affected the morphology of the nanocubes more than the nanospheres. Additionally, the size effect seemed to be the primary issue affecting the photocatalytic performance of the as-obtained photoanodes. Studying the interfacial phenomena further with respect to the photoanode preparation process indicated that factors including the size, thickness, orientation, and exposed facets were closely related to the photocurrent performance. The exposed facets that were in contact with the electrolyte seemed to be the second important issue because the surface activity influenced the resistance and the capacitance behaviour at the bulk hematite and at the hematite–electrolyte interface.

Acknowledgements

We thank the National Science Council (NSC102-2113-M-007-003-MY2) and the Bureau of Energy, Ministry of Economy Affair, Taiwan, the Republic of China, for supporting this study. We also thank Elsevier Language Service for English correction.

References

1. X. L. Mou, Y. Li, B. S. Zhang, L. D. Yao, X. J. Wei, D. S. Su and W. S. Shen, Eur. J. Inorg. Chem., 2012, 16, 2684–2690.
2. Z. Liu, B. L. Lv, Y. Xu and D. Wu, J. Mater. Chem. A, 2013, 1, 3040–3046.
3. X. L. Mou, X. J. Wei, Y. Li and W. J. Shen, CrystEngComm, 2012, 14, 5107–5120.
4. X. M. Zhou, J. Y. Lan, G. Liu, K. Deng, Y. L. Yang, G. J. Nie, J. G. Yu and L. J. Zhi, Angew. Chem., Int. Ed., 2012, 51, 178–182.
5. H. M. Chen, Y. Q. Zhao, M. Q. Yang, J. H. He, P. K. Chu, J. Zhang and S. H. Wu, Anal. Chim. Acta, 2010, 659, 266–273.
6. Y. Yang, H. X. Ma, J. Zhuang and X. Wang, Inorg. Chem., 2011, 50, 10143–10151.
7. Y. Liu, C. Y. Yu, W. Dai, X. H. Gao, H. S. Qian, Y. Hua and X. Hu, J. Alloys Compd., 2013, 551, 440–443.
8. Y. H. Zheng, Y. Cheng, Y. S. Wang, F. Bao, L. H. Zhou, X. F. Wei, Y. Y. Zhang and Q. Zheng, J. Phys. Chem. B, 2006, 110, 3093–3097.
9. X. M. Zhou, H. C. Yang, C. X. Wang, X. B. Mao, Y. S. Wang, Y. L. Yang and G. Liu, J. Phys. Chem. C, 2010, 114, 17051–17061.
10. Y. Y. Xu, G. Y. Zhang, G. X. Du, Y. Q. Sun and D. Z. Gao, Mater. Lett., 2013, 92, 321–324.
11. Y. Zhao, E. A. Hernandez-Pagan, N. M. Vargas, J. L. Dysart and T. E. Mallouk, J. Phys. Chem. Lett., 2011, 2, 402–406.
12. S.-H. A. Lee, A.-M. S. Jackson, A. Hess, S.-T. Fei, S. M. Pursal, J. Basham, C. A. Grimes, M. W. Horn, H. R. Allcock and T. E. Mallouk, J. Phys. Chem. C, 2010, 114, 15234–15242.
13. G. M. Wang, Y. C. Ling, D. A. Wheeler, K. E. N. George, K. Horsley, C. Heske, J. Z. Zhang and Y. Li, Nano Lett., 2011, 11, 3503–3509.
14. S. Y. Lian, E. B. Wang, Z. H. Kang, Y. P. Bai, L. Gao, M. Jiang, C. W. Hu and L. Xu, Solid State Commun., 2004, 129, 485–490.
15. W. R. Zhao, J. L. Gu, L. X. Zhang, H. G. Chen and J. L. Shi, J. Am. Chem. Soc., 2005, 127, 8916–8917.
16. Q. Y. Hao, S. Liu, X. M. Yin, Z. F. Du, M. Zhang, L. M. Li, Y. G. Wang, T. H. Wang and Q. H. Li, CrystEngComm, 2011, 13, 806–812.
17. Y. Vasquez, A. E. Henkes, J. C. Bauer and R. E. Schaak, J. Solid State Chem., 2008, 181, 1509–1523.
18. D. D. Qin, C. L. Tao, S. I. In, Z. Y. Yang, T. E. Mallouk, N. Z. Bao and C. A. Grimes, Energy Fuels, 2011, 25, 5257–5263.
19. T. H. Wang, M. C. Huang, W. S. Chang, J. C. Lin, C. H. Lee and C. F. Wang, ACS Appl. Mater. Interfaces, 2013, 5, 7937–7949.
20. I. V. Chernyshova, S. Ponnurangam and P. Somasundaran, Phys. Chem. Chem. Phys., 2013, 15, 6953.
21. W. Wu, R. Hao, F. Liu, X. T. Su and Y. L. Hou, J. Mater. Chem. A, 2013, 1, 6888–6894.
22. K. Shimizu, A. Lasia and J.-F. Boily, Langmuir, 2012, 28, 7914–7920.
23. J. R. Macdonald, Impedance Spectroscopy, Wiley-Interscience, New York, 1987.
24. S. M. Haile, G. Staneff and K. H. Ryu, J. Mater. Sci., 2001, 36, 1149–1160.

---