III. RESULTS AND DISCUSSION

Anodization of iron in an EG electrolyte containing F⁻ produces an anodic layer consisted of highly ordered vertically aligned elongated nanopores, as shown in the FESEM image of the film cross-section in Fig. 1(a). The morphology of the film surface is shown in Fig. 1(b). Elongated pores form channels within the layer, as shown by the transmission electron microscopy (TEM) image in Fig. 1(c). The length of the pores corresponds to the thickness of the anodic layer, and is measured to be ∼1.8 µm. The TEM image in Fig. 1(d) shows that the regular cylindrical pores have diameters of ∼30 nm.

FIG. 1.

FESEM images of iron oxide film prepared by anodizing iron at 50 V for 15 min in EG containing 0.35 wt% NH₄F: (a) film cross-section, (b) film surface and TEM images showing (c) the pore channels and (d) the ordered pores.

Current density versus time measurements were used to understand the formation of the nanoporous structure, and the resulting curve is shown in Fig. 2. Proposed stages during anodization are shown schematically in Fig. 3. The initial decay in current density [Fig. 2, region (I)] can be explained by the oxidation of the bare iron surface [Fig. 3, step (I)], with the electrolyte forming a thin iron oxide layer. An anodic layer is formed in buffer solution from the rearrangement of oxygen and iron ions, which may be Fe₃O₄ [Eq. (1)] as proposed by previous authors like Nagayama et al.^{Reference Nagayama and Cohen16} As oxidation proceeds, the outer part of Fe₃O₄ is converted to γ-Fe₂O₃ [Eq. (2)]. It is also possible to achieve oxidation through [Eq. (3)], where γ-Fe₂O₃ formed as the main component of the anodic layer. Similar reactions are thought to occur in the current H₂O-EG electrolyte, forming a barrier layer on the iron surface as shown in Fig. 3, step (II). Further oxidation of γ-Fe₂O₃ may form a cation vacancy structure.^{Reference Nagayama and Cohen16} Eventually γ-Fe₂O₃ will transform to the more stable phase of iron oxide α-Fe₂O₃.

(1)$${\rm{3Fe}}\,{\rm{ + }}\,{\rm{4 }}{{\rm{H}}_{\rm{2}}}{\rm{O}} \to {\rm{F}}{{\rm{e}}_{\rm{3}}}{{\rm{O}}_{\rm{4}}}\,{\rm{ + }}\,{\rm{8}}{{\rm{H}}^{\rm{ + }}}\,{\rm{ + }}\,{\rm{8}}{{\rm{e}}^ - }$$

(2)$$2{\rm{F}}{{\rm{e}}_3}{{\rm{O}}_{\rm{4}}}\,{\rm{ + }}\,{{\rm{H}}_{\rm{2}}}{\rm{O}}\, \to \,3\gamma {\rm{ - F}}{{\rm{e}}_{\rm{2}}}{{\rm{O}}_{\rm{3}}}\,{\rm{ + }}\,{\rm{2}}{{\rm{H}}^{\rm{ + }}}\, + \,{\rm{6}}{{\rm{e}}^ - }$$

(3)$$2{\rm{Fe}}\,{\rm{ + }}\,3{{\rm{H}}_{\rm{2}}}{\rm{O}}\, \to \,\gamma {\rm{ - F}}{{\rm{e}}_{\rm{2}}}{{\rm{O}}_{\rm{3}}}\,{\rm{ + }}\,{\rm{6}}{{\rm{H}}^{\rm{ + }}}\, + \,{\rm{6}}{{\rm{e}}^ - }$$

FIG. 2.

Current density–time curve during the anodization of iron at 50 V for 15 min in electrolyte containing 0.35 wt% NH₄F.

FIG. 3.

Stages in the formation of the nanoporous oxide film on iron by anodization process in fluoride-EG electrolyte.

Growth of the anodic layer proceeds by the inward migration of anions to the iron|iron oxide interface, or outward migration of cations, via anion or cation vacancies, respectively [Fig. 3, step (II)]. The migration of ions is determined by the field within the anodic layer, and largely depends on the applied voltage and migration paths (i.e., the generation and concentration of defects such as vacancies). The barrier layer is susceptible to dielectric breakdown and chemical etching or pitting due to F⁻.^{Reference Szklarska-Smialowska17,Reference Regonini, Bowen, Jaroenworaluck and Stevens18} Pits form randomly throughout the surface of the anodic layer, as shown in Fig. 3, step (II). The current slightly increases upon breakdown of the barrier layer, indicated by region (II) in Fig. 2.

During pitting, the presence of water in the electrolyte aids the dissolution of the oxide, forming an aquo complex. F⁻ migrates inward to maintain charge neutrality within pits, which increases the local F⁻ concentration.^{Reference Schmuki19} This leads to dissolution of the barrier layer, exposing the bare metal surface; however, the bare metal surface continuously oxidizes because of the applied anodic potential and in the presence of H₂O, oxide forms. This process is accompanied by the release of H⁺, and therefore an increase in the local H⁺ concentration [Eqs. (1)–(3)]. The high concentration of F⁻ and H⁺ within pits leads to rapid dissolution of the growing oxide thus the formation and growth of pores [Eq. (4)]. This is illustrated in Fig. 3, step (II). Complexation of Fe³⁺ with the F⁻ species in the solution can also be expected to produce the initial pores.

(4)$${\rm{F}}{{\rm{e}}_2}{{\rm{O}}_{\rm{3}}}\,{\rm{ + }}\,12{{\rm{F}}^ - }\,{\rm{ + }}\,6{{\rm{H}}^{\rm{ + }}}\, \to \,{\rm{\ 2}}{\left[ {{\rm{Fe}}{{\rm{F}}_{\rm{6}}}} \right]^{{\rm{3}} - }}\, + \,3{{\rm{H}}_{\rm{2}}}{\rm{O}}$$

The porosification of iron oxide can therefore be thought of as autocatalytic, whereby initial dissolution initiates further dissolution at the electrolyte|oxide interface within pits. This process deepens and enlarges the pits, forming the stable pores shown in Fig. 3, step (III). The active dissolution front is at the electrolyte|oxide interface at the bottom of the pores, so pores grow inward and become elongated. This process is independent of those occurring at the oxide surface, in which dissolution and/or deposition can occur. The formation of the nanopores is time-dependent, so longer anodization duration deepens the pores (Fig. 3). However, dissolution can occur at the surface because the electrolyte used is rather corrosive [Eq. (4)]. Mild dissolution at the entrances to pores is however required to reveal the pores, and surface etching is sufficient to achieve this in the fluoride-EG electrolyte used.

The FESEM images in Fig. 1 show that the pore size is uniform. Pores can be classified as weak or strong pores.^{Reference Yasuda, Macak, Berger, Ghicov and Schmuki20} Significant H⁺ is produced in strong pores, but a lesser amount is produced in weak pores, so their growth rate slows. Weak pores eventually stop growing, because the neighboring metal has already been oxidized by strong pores, which results in the uniform pore size.

Anodic iron oxide is amorphous^{Reference Prakasam, Varghese, Paulose, Mor and Grimes10,Reference Zhang, Hossain and Takahashi12} and may contain a small amount of fluoride phases like FeF₂.^{Reference Albu, Ghicov and Schmuki6,Reference Rozana, Azhar, Anwar, Kawamura, Khairunisak, Matsuda and Lockman13} For instance, Konno et al. suggested that anodic film is amorphous, hydrated, and highly contaminated with fluoride as well as carbon species.^{Reference Konno, Tsuji, Skeldon, Thompson and Habazaki9} Fluoride or carbon (from EG used) species can be removed by annealing the anodic oxide at high temperature. Moreover, annealing induces crystallization of the oxide and may transform γ-Fe₂O₃ and hydrated phases to the photoactive α-Fe₂O₃ phase. Crystalline α-Fe₂O₃ is preferred as a photoanode in a PEC cell, so annealing is necessary to induce phase formation and crystallization. The annealing conditions must be however carefully controlled as to avoid the oxidation of the underlying iron, thickening of the barrier layer and growth of the surface oxide which can eventually collapse the nanoporous structure. Annealing in the current study was carried out in nitrogen, to avoid growth of new oxide layer. Annealing was carried out at 300, 450, 500, 600, and 900 °C for 2 h.

Morphologies of anodic iron oxide annealed at 300–900 °C in nitrogen are shown in Fig. 4. It is obvious that the morphology is significantly affected by the annealing temperature whereby the nanoporous structure retains when the sample was annealed at ≤ 450 °C. Sample annealed at 500 °C has much reduced pore diameters, indicating the growth of the pore wall shrinking the pore diameters. The pores are completely destroyed leaving large grains when sample was annealed at 600 °C. These grains grow into even large crystals upon annealing at 900 °C. Grain growth is inevitable at such high temperatures as diffusion of iron ions from the parent metal leads to the formation of new, possibly iron-rich grains. Annealing temperature must therefore be kept at <500 °C.

FIG. 4.

FESEM images of iron oxide films annealed for 2 h at different temperatures in nitrogen: (a) as-anodized, (b) 300 °C, (c) 450 °C, (d) 500 °C, (e) 600 °C, and (f) 900 °C.

XRD was used to obtain structural information of the annealed samples as shown in Fig. 5. For the ease of observation, XRD of as-anodized and 300 °C samples were plotted together as shown in Fig. 5(a) whereas XRD patterns for all samples are shown in Fig. 5(b). For the as-anodized sample γ-Fe₂O₃ (JCPDS no. 39-1346) peaks are observed. Formation of this phase is discussed as shown in [Eq. (3)]. Peaks from FeF₂ can also be seen. Our findings are consistent with previous reports. For example, Schmuki et al.^{Reference Albu, Ghicov and Schmuki6} reported on nanoporous iron oxide containing FeF₂ (JCPDS no. 03-1324) when iron was anodized in a fluorinated bath, and that the fluoride content decreased upon subsequent annealing. The XRD pattern of our sample annealed at 300 °C contains some very weak peaks which can be indexed to Fe₃O₄ and α-Fe₂O₃ but in general the oxide annealed at this temperature seems to be rather amorphous [α-Fe₂O₃ (JCPDS no. 33-0664) and Fe₃O₄ (JCPDS no. 19-0629)]. The FeF₂ peak then disappeared. Annealing at 450 °C crystallizes the oxide whereby peaks of α-Fe₂O₃ and Fe₃O₄ are clearly observed and indexed according to the JCPDS cards as mentioned above. By using with Rietveld refinement method the sample appears to have 88.6% α-Fe₂O₃ and 11.4% Fe₃O₄. Annealing at 500 °C resulted in the formation of more Fe₃O₄ indicating either phase transformation or formation at this temperature (% of phases calculated by Rietveld method is on Table I). Annealing at 600 °C resulted in further crystallization of α-Fe₂O₃ enhancing the XRD peaks of this oxide phase along with the Fe₃O₄ phase. Even though α-Fe₂O₃ is a fully oxidized form of iron oxide, and is the most stable, at the interface between α-Fe₂O₃ and the metal foil, Fe₃O₄ may have formed and similar to thermal oxidation of iron, the formation of Fe₃O₄ is temperature dependent. Various authors have discussed on this transformation especially during high temperature oxidation of iron as reviewed by Chen et al.^{Reference Chen and Yeun21} Raman spectra were recorded to further investigate the crystallinity of the anodic iron oxide, and are shown in Fig. 6. The Raman peak positions of α-Fe₂O₃ are similar for samples annealed at 300–900 °C and as in agreement with the XRD, sample annealed at 300 °C contains weak peaks of α-Fe₂O₃. However in general this sample seems to be rather amorphous. A broad peak of Fe₃O₄ can be detected at 660 cm⁻¹ for the sample annealed at 450–600 °C. Broad Raman peaks indicate structural disorder which may be due to the incorporation of foreign ions and/or formation of vacancies within the crystals as well as the small crystallite size. The observed Raman peaks are summarized in Table II.^{Reference Bersani, Lottici and Montenero22–Reference Oh, Cook and Townsend26}

FIG. 5.

XRD patterns of nanoporous anodic iron oxide films: (a) as-anodized and 300 °C (b) 450 °C–900 °C in nitrogen.

TABLE I.

Percentage ratio between α-Fe₂O₃ and Fe₃O₄ in the annealed samples.

FIG. 6.

Raman spectra of nanoporous anodic iron oxide films annealed in nitrogen.

TABLE II.

Raman shift of anodic iron oxide annealed at 300–900 °C.

An HRTEM image of the sample annealed at 450 °C for 2 h in nitrogen is shown in Fig. 7(a). Clear lattice fringes with 0.272 nm spacing is observed, corresponding to (104) planes of α-Fe₂O₃. (300), (113), and (012) planes of α-Fe₂O₃ and (220) plane of Fe₃O₄ are also observed as listed in Table III. The selected area electron diffraction (SAED) pattern in Fig. 7(b) shows many diffraction spots, indicating that α-Fe₂O₃ is polycrystalline. Based on all these findings, annealing in nitrogen appears to have induced crystallization of the desired α-Fe₂O₃ phase.

FIG. 7.

(a) HRTEM image of anodic iron oxide annealed for 2 h at 450 °C in nitrogen, and (b) corresponding SAED pattern.

TABLE III.

List of lattices fringes observed in Fig. 7(a).

PEC measurements of the nitrogen-annealed anodic iron oxide film on iron foil were recorded in solution of 1 M NaOH. Figure 8 shows the photocurrent versus potential measured relative to an Ag/AgCl standard electrode, under AM 1.5 simulated solar irradiation in 1 M NaOH. The dark current is shown for comparison, and no significant current is generated in the absence of light. Upon irradiation, the sample annealed at 300 °C (barrier layer thickness ∼190 nm) shows a low photocurrent of 0.07 mA/cm² at 0.5 V versus Ag/AgCl. Recall that the XRD pattern in Fig. 5 reveals that the sample is rather amorphous; suggesting that the crystallinity of the anodic layer affects the generation of photocurrent.

FIG. 8.

Photocurrent density-potential characteristics of nitrogen-annealed samples in 1 M NaOH.

The sample annealed at 450 °C generates a photocurrent of ∼1.04 mA/cm² at 0.5 V versus Ag/AgCl. Our result is higher than a photocurrent of ∼0.26 mA/cm² at 0.6 V versus Ag/AgCl, reported by previous authors (for anodic nanoporous iron oxide annealed at 400 °C for 30 min)^{Reference Prakasam, Varghese, Paulose, Mor and Grimes10} or a photocurrent of ∼0.45 mA/cm² at 0.6 V versus Ag/AgCl for air-annealed sample.^{Reference Rozana, Azhar, Anwar, Kawamura, Khairunisak, Matsuda and Lockman13} This anodic nanoporous iron oxide was predominantly α-Fe₂O₃ with a small amount of Fe₃O₄, and had a barrier layer thickness of ∼800 nm.^{Reference Prakasam, Varghese, Paulose, Mor and Grimes10}

The higher photocurrent for our sample may be attributed to the small thickness of the barrier layer for the nitrogen-annealed sample which is ∼200 nm, as shown in Fig. 9(a). In Fig. 9(b), the cross-section of an air-annealed sample is shown to have barrier layer thickness of >1 µm. We compared the XRD patterns of the air-annealed and the nitrogen-annealed samples as in Fig. 10. From this figure, Fe₃O₄ peaks for air-annealed sample are seen to be more intense than the sample annealed in nitrogen indicating phase transformation (or formation) to Fe₃O₄ during annealing in air. The thicker barrier layer for the air-annealed sample is thought to have consisted of Fe₃O₄ and as Fe₃O₄ is not exactly a good electronic conductor, transport of electrons to the back contact through this layer is more sluggish. Moreover, at high temperature FeO as well as nonstoichiometric α-Fe₂O₃ may also have formed. For ease of comparison, the photocurrent versus voltage plots of the air-annealed sample and nitrogen-annealed samples are shown in Fig. 11. As can be seen, the air-annealed samples have a photocurrent of 0.13 mA/cm² at 0.5 V versus Ag/AgCl. To further investigate the characteristic of nitrogen-annealed sample, UV–Vis absorbance spectra of sample were produced [Fig. 12(a)]. From here the band gap of the sample was measured to be ∼2.14 eV [Fig. 12(b)]. Annealing in nitrogen does not affect the energy gap of the α-Fe₂O₃ very much.

FIG. 9.

FESEM images of nanoporous iron oxide annealed for 2 h at 450 °C in (a) nitrogen and (b) air.

FIG. 10.

XRD of nanoporous iron oxide annealed for 2 h at 450 °C in nitrogen and air.

FIG. 11.

Photocurrent comparison of sample annealed in nitrogen and air at 450 °C.

FIG. 12.

Optical property of nanoporous iron oxide annealed for 2 h at 450 °C in nitrogen: (a) optical absorption spectra and (b) differential reflectance plotted as a function of energy.

From Fig. 8, it is obvious that photocurrent for sample annealed at 500 °C is lower compared to sample annealed at 450 °C. And samples annealed at even higher temperatures have even smaller photocurrent. Table IV summarizes the photocurrent of all samples at 0.5 V versus Ag/AgCl. High temperature annealing may increase the barrier layer thickness hence lower photocurrent values are produced for these samples. Moreover, annealing at high temperature induces grain growth. Recall that for samples annealed at > 600 °C, the nanoporous structure has completely destroyed for all samples, leaving a flat oxide layer. Such samples have lower surface areas and are similar to flat electrode, it provides longer distance for holes to travel hence much of the photogenerated carriers have recombined, thus lowering the photocurrent. At high temperature, similar to the above mentioned argument, annealing will induce the formation of FeO or Fe₃O₄.

TABLE IV.

Photocurrent performance of nitrogen-annealed samples in 1 M NaOH under AM 1.5 (100 mW/cm²).

Annealing in nitrogen of anodic nanoporous iron oxides film is suggested to be used to generate high photocurrent as it avoids forming a new oxide barrier layer but the temperature must be less than 500 °C but above 300 °C. Too high of a temperature as mentioned would induce fast grain growth, diminishing the nanoporous structure and aid to the transformation of semiconducting α-Fe₂O₃ to the nonphotoconductive FeO and Fe₃O₄.