Solution-mediated nanometric growth of α-Fe2O3 with electrocatalytic activity for water oxidation

This paper describes a simple, low-temperature, and environmentally friendly aqueous route for the layer-by-layer nanometric growth of crystalline α-Fe2O3. The formation mechanism involves alternative sequences of the electrostatic adsorption of Fe2+ ions on the surface and the subsequent onsite oxidation to Fe3+. A combination analysis of X-ray diffraction, scanning electron microscopy, UV-Vis spectroscopy, and X-ray photoelectron spectroscopy revealed that α-Fe2O3 is directly formed without post-growth annealing via designed chemical reactions with a growth rate of ca. 1.7 nm per deposition cycle. The obtained α-Fe2O3 layer exhibits electrocatalytic activity for water oxidation and, at the same time, insignificant photo-electrocatalytic response, indicating its defective nature. The electrocatalytic activity was tailored by annealing up to 500 °C in air, where thermal diffusion of Sn4+ into the α-Fe2O3 lattice from the substrate probably provides an increased electrical conductivity. The subsequent surface-modification with Ni(OH)2 lowers the overpotential (250 mV at 0.5 mA cm−2) in a 1 M KOH solution. These findings open direct growth pathways to functional metal oxide nanolayers via liquid phase atomic layer deposition.


Introduction
Ceramics coating is a core technology that affords various functions to a substrate material. Metal oxides are particularly attractive for applications in anticorrosion, 1 catalysis, 2 sensing, 3 energy storage 4 and conversion, 5 optics, 6 and electronics. 7 In an industry context, oxide materials are deposited by vacuum phase deposition techniques such as chemical vapor deposition, 8 pulsed laser deposition, 9 and sputtering. 10 Solution deposition techniques such as sol-gel, 11 electrodeposition, 12 and chemical bath-deposition (CBD) methods 13 are potential alternatives to the above-mentioned vacuum processes as they utilize inexpensive and less toxic solution precursors as well as ambient pressure for oxide deposition, hence making the fabrication more environmentally friendly and more costeffective.
Moreover, a solution process enables the direct deposition of crystalline oxides on the substrate without post-growth annealing. Low-temperature direct deposition is suitable for lm formation on low heat-resistant substrates, expanding the potential scope of functional oxide materials as a component in exible plastic devices and electrochemical devices with indium-tin-oxide substrates. Recent efforts have allowed the fabrication of various oxide (e.g., ZnO, 14 TiO 2 , 15 WO 3 , 16 and SnO 2 (ref. 17)) via aqueous solution routes without the need for a post-annealing treatment. However, so far, the direct solution routes have been adopted for tiny members in the broad family of metal oxides. The major difficulty lies in the growth mechanism; according to classical aqueous chemistry, the metallic ions (M n+ ) in an aqueous solution react with OH À to precipitate hydroxide nucleus on the substrate via a heterogenous nucleation process (1): 18 If the stability of hydroxide is sufficiently low for dehydration, the hydroxide spontaneously transforms into an oxide form during deposition (2): 18 However, hydroxides are oen stable so that a postannealing treatment is inevitable for oxide formation. For example, annealing temperatures over 500 C are necessary to yield Al 2 O 3 (ref. 19) and ZrO 2 (ref. 20) from the hydroxides.
This difficulty also applies to a-Fe 2 O 3 , the target material in the present study. Resultant products from classical precipitation reactions between ferric or ferrous precursors with an aqueous base are a-FeOOH, b-FeOOH, g-FeOOH, s-FeOOH, and Fe 5 OH 8 $4H 2 O, 21-23 while one can nd no direct solution route to a-Fe 2 O 3 . Nevertheless, Fe 3 O 4 (magnetite) can be formed via a co-precipitation process of ferric and ferrous precursors, even at room temperature. 21 In other words, Fe-(OH) 2 -Fe bonds can preferentially convert Fe-O-Fe bonding via dehydration in an aqueous solution, opening the possibility for direct solution deposition of crystalline a-Fe 2 O 3 under suitable reaction conditions. In fact, using the above idea, we recently developed a liquid phase atomic layer deposition (LP-ALD) of a-Fe 2 O 3 via an onsite oxidation and dehydration pathway by using a spin spray technique. 24 In this method, a source solution containing Fe 2+ and an oxidizing solution containing an oxidizer, NaNO 2 , was simultaneously sprayed onto the substrates mounted on a rotating table heated to 95 C. We propose that the deposition mechanism involved alternative sequences of the absorption of Fe 2+ ions onto the surface and the subsequent formation of Fe 3+ -oxygen bonds through reactions with the source and oxidizing solutions, respectively (Fig. 1a). We refer to this deposition process as LP-ALD. Such a non-classical LP-ALD strategy may open direct growth pathways to functional metal oxide nanolayers via aqueous solution chemistry. However, the spin spray technique is available only in the specialized laboratories.
In the present study, we explore a solution-mediated alternate reaction technique, SMART, to further verify, simplify, and generalize the non-classical LP-ALD reaction pathway (Fig. 1a). Briey, lm deposition in SMART proceeds simply by alternate immersion of the substrate in FeCl 2 and NaNO 2 precursor solutions. We demonstrate a primitive beaker process that allows the direct growth of crystalline a-Fe 2 O 3 lms with a growth rate of ca. 1.7 nm per cycle. Resultant a-Fe 2 O 3 thin lms exhibit unexpected electrocatalytic activity for oxygen evolution reactions (OER). The origin of catalytic activity comes from the defective nature of SMART-derived a-Fe 2 O 3 where OH À species are present in the oxide lattice. Sn 4+ -diffusion into the a-Fe 2 O 3 lattice by annealing and surface modication with Ni(OH) 2 further enhance the OER activity, which is superior to state of the art a-Fe 2 O 3 -based catalysts.

Experimental section
Preparation of the source and oxidizing solutions  was purged into the solutions at a rate of 1.0 L min À1 while stirring in order to prevent the oxidation of the reactants in the solution.

Deposition process in SMART
A soda-lime glass substrate (25 mm Â 25 mm Â 1 mm) and a transparent conductive oxide (TCO) coated glass substrate (FTO/ITO, Type-0052, 10 U sq À1 , Geomatec Co., Ltd.) were cleaned in an ultrasonic bath with distilled water, ethanol, and acetone for 10 min each. The glass substrate was further cleaned by treatment in a bath of methanol (Wako Pure Chemical Industries)/HCl (Wako Pure Chemical Industries) (1/1 in volume) for 1 day to obtain a hydrophilic surface, while the TCO substrate was immersed in the 0.1 M HCl (Wako Pure Chemical Industries) for 1 day to obtain a hydrophilic surface.
The surface-modied substrate was rst immersed in the source solution heated to 75 C for 1 min, followed by rinsing with ethanol. Then the substrate was immersed in the oxidizing solution heated at 75 C for 1 min and rinsed with ethanol again. A series of these operations was repeated from 1 to 90 times to control the lm thickness.
Ni(OH) 2 surface-modication a-Fe 2 O 3 layer deposited on TCO was immersed in 0.1 M Ni(NO 3 ) 2 aqueous solution (Wako Pure Chemical Industries, Ltd., Japan) for 1 min without heating followed by rinsing with water. Then, the substrate was immersed in 1 M KOH solution without heating for 1 min, and then rinsed with water.

Fabrication of Ni(OH) 2 layer by a successive ionic layer adsorption and reaction (SILAR)
The surface-modied substrate was rst immersed in a 0.1 M Ni(NO 3 ) 2 aqueous solution heated to 75 C for 1 min, followed by rinsing with water. Then the substrate was immersed in a 1 M KOH aqueous solution heated at 75 C for 1 min and rinsed with water. The process was repeated 30 times to obtain Ni(OH) 2 layer.

Characterization
The crystalline phases of the deposited lms were identied by X-ray diffraction (XRD, MultiFlex, Cu Ka, 40 kV and 40 mA, Rigaku). The surface morphologies and textures of the lms were observed using a scanning electron microscope (SEM, SU-8020, Hitachi High-Technologies). The elemental distribution was observed an energy dispersive X-ray spectroscopy (EDS, JEOL, JSM-7600). The light absorbance of samples in the ultraviolet-visible (UV-Vis) region was evaluated by the visible absorption spectroscopy (UV-Vis, UV-1280, Shimadzu). X-ray photoelectron spectroscopy (XPS, JPS 9010 TR, JEOL) was conducted to investigate the chemical state of the samples. All measured XPS spectra were calibrated corresponding to the value of the C 1s peak at 284.4 eV using Mg Ka X-ray source with 1253.6 eV. Raman spectroscopy measurements were made LabRam Armis, Horiba Jobin Yvon instrument equipped with 532 nm laser and a microscope to focus the laser light on the lm surface.

Electrochemical and photoelectrochemical measurements
The OER measurements were performed in 1 M KOH aqueous solution using a three-electrode conguration, with a Pt wire counter electrode and an Ag/AgCl, KCl reference electrode. All potentials have been referenced to the reversible hydrogen electrode (RHE) by the expression: The linear sweep voltammograms (LSV) was performed for 20 cycles with a scan rate of 5 mV s À1 . Photoelectrochemical measurements were performed at the same condition, while visible light (wavelength above 400 nm) was irradiated for the measurements. The 200 W Xeon lamp (Asahi Spectra. Co) was used for the measurements.

Characterization of SMART-derived thin lms
The reaction pathway of SMART is designed as follows. In the rst step, the substrate is immersed in a FeCl 2 solution with pH 4 to form a Fe 2+ adlayer onto the substrate surface. When an oxide surface is negatively charged at the pH, a double layer is formed on the surface where Fe 2+ forms an inner layer (Stern layer) and the Cl À from the FeCl 2 precursor forms a chargebalancing outer layer. In the following step, the substrate is rinsed with ethanol so that only the immobile double-layer remains on the substrate surface. Subsequently, the substrate is immersed in the NaNO 2 solution and heated to 75 C. The NO 2 À in the solution is diffused onto the surface to oxidize the adsorbed Fe 2+ to the Fe 3+ state. At this stage, hydrolysis and dehydration simultaneously occur onsite, resulting in the formation of the rst O-Fe-O bonds. In a nal step, the substrate is rinsed again to remove the ions from the diffusion layer. In principle, the repletion of these cycles leads to a layerby-layer deposition of the a-Fe 2 O 3 layer. Fig. 1b shows the X-ray diffraction pattern from the sample prepared by the SMART on the glass substrate aer 90 cycles of deposition. The pattern displays broad peaks from the glass substrate and sharp peaks corresponding to 104 and 110 reections of the a-Fe 2 O 3 phase (JCPDS 33-0664) without an impurity phase such as Fe(OH) 2 , Fe(OH) 3 , FeOOH, Fe 3 O 4 , and g-Fe 2 O 3 . According to Scherrer's equation using a 104 peak, the crystallite size is calculated to be 47.4 nm. 25 Raman spectroscopy was employed to further investigate the phase purity of the resultant lms. As shown in Fig. 1c, all the detected peaks are assigned as two A 1g modes (231 and 473 cm À1 ) and four E g modes (255, 297, 411, and 576 cm À1 ), 26,27 in support of the lm being composed of a-Fe 2 O 3 domains. Note that Raman spectra up to 1400 cm À1 shown in Fig. S1, ESI † also exclude the formation of iron-based impurity phases. 28 Fig. 1d displays the cross-sectional and surface SEM images of the SMART-derived a-Fe 2 O 3 lm. The cross-sectional image reveals the formation of a dense layer with a relatively uniform thickness of 150 nm on average. Thus, the growth rate can be estimated at approximately 1.7 nm per deposition cycle, if the lm thickness linearly increases in each deposition cycle. The surface image presents a continuous lm consisting of dense grains with an average size of ca. 50 nm. The grain size roughly matches the calculated crystallite size (47.4 nm) from the XRD pattern, indicating that each grain consists of a single crystalline domain. Indeed, the SEM observation with a lower magnication conrms that the lm is free from cracks (Fig. S2, ESI †). Owing to the high uniformity, the lm was transparent with a yellowish red color, as shown in the inserted photograph in Fig. 1a. The lm exhibited good adhesion to the substrate aer scotch tape testing, indicating the presence of chemical bonds at the interface between the substrate and the a-Fe 2 O 3 layer.

Growth rate and mechanism of SMART process
We verify the growth mechanism by monitoring the change of UV-Vis spectra with an increase in the number of deposition cycles. Fig. 2a shows a change in the UV-Vis absorption spectra aer 1, 3, 5, 10, 15, 20, 25, and 30 cycles on the glass substrate. In general, the absorption gradually intensied with the number of deposition cycles, while the spectral feature, e.g., absorption onset, was not largely changed aer at least 3 cycles, indicating that the a-Fe 2 O 3 layer was deposited throughout the cycles. Fig. 2b displays plots of absorption intensity at 400 nm versus the number of deposition cycles. The plots were tted with lines at a slope of ca. 0.01 per cycle. Note that the absorption coefficient for 90 cycles of deposited a-Fe 2 O 3 was 1.12 at 400 nm (Fig. S3, ESI †). From this value, the slope was calculated to 0.012 per cycle, which accords well to the slope value obtained from up to 30 cycles. Thus, the UV-Vis absorption data supports that the a-Fe 2 O 3 lm is deposited in a layerby-layer manner. The calculated crystallite size, 47.4 nm from the XRD pattern (Fig. 1b), was 28 times larger than the growth rate (ca. 1.7 nm). Thus, the Fe 2+ species in the Stern layer was mainly consumed for crystal growth rather than the heterogeneous nucleation process. We conrm that the slope value that tted with the UV-Vis data was unchanged when the TCO substrate was used as the substrate (Fig. 2b). This result is reproducible. The thickness of a a-Fe 2 O 3 layer on an TCO was observed to approximately 50 nm aer 30 deposition cycles (Fig. S4, ESI †). This further conrms that the growth rate was about 1.7 nm per cycle.
We employed X-ray photoemission spectroscopy (XPS) to detect the changes in surface states of the rst cycle. Fig. 2c displays Fe 2p XPS spectra of the deposited layer aer the reaction with FeCl 2 followed by the rinse step, and one subsequently reacted with the NaNO 2 solution followed by a second rinse step. Note that the Sn 3p 3/2 background signal was extracted from the as-obtained data to better understand the Fe 2p spectra see Fig. S5, ESI. † The Fe 2p spectra involve multi-components including Fe 2p 1/2 , Fe 2p 3/2 , and their satellite peaks, while it can be quantitively described that the peaks shied toward the higher energy side aer reaction with NaNO 2 . Indeed, a peak at around 719 eV, attributable to a Fe 2p 1/2 satellite of Fe 3+ , was pronounced aer the reaction. 29,30 These changes demonstrate that the oxidation of Fe 2+ occurred by reactions with the NaNO 2 solution. Besides, the intensity of the Cl 2p peak (Fig. 2d) decreased aer the second step. This supports the notion that the Cl À ions involved in the outer layer of the doublelayer were replaced by O 2À or OH À species binding with Fe 3+ aer the oxidation step. Thus, all the analytical results support that the a-Fe 2 O 3 layer could be deposited, according to the designed SMART concept (Fig. 1a). The direct formation of a-Fe 2 O 3 demonstrates that controlling surface redox reactions in the growth process plays a critical role in crystallization. The idea would be applied to bring intriguing aqueous routes to crystalline metal oxides based on multivalent metallic components such as Cu 1+/2+ , Co 2+/3+ , and Mn 2+/3+/4+ .

Electrocatalytic properties of SMART-derived thin lms
Subsequently, we investigated the electrochemical and photoelectrochemical catalytic performance of SMART-derived a-Fe 2 O 3 for an oxygen evolution reaction (OER) to nd any structure-performance correlations. As is well known, OER is the essence of renewable fuel generation in water electrolysis, and development of stable, cost-effective, and environmentally- Nanoscale Advances Paper friendly OER catalysts is a key challenge. [31][32][33][34][35] This sheds light on a-Fe 2 O 3 as one of the most suitable materials. 36,37 Fig. 3a shows plots of linear sweep voltammograms (LSV) of a a-Fe 2 O 3 deposited TCO substrate aer 30 cycles of SMART along with a bare-TCO reference. In this case, the measurements were performed without light-irradiation. A bare TCO shows a tiny cathodic current (0.18 mA cm À2 at 1.8 V), while a current of more than one order of magnitude higher is gained at the same potential aer a-Fe 2 O 3 deposition. The overpotential is about 390 mV at 0.5 mA cm À2 . We also performed electrochemical measurements under light-irradiation to investigate catalytic activity of photoelectrochemical water oxidation (Fig. 3b). As a result, a tiny increase of cathodic current in the mA cm À2 range was observed, which demonstrates that most photogenerated electron/hole pairs were expensed for the recombination pathways rather than for water oxidation. To our knowledge, there has been no report of a-Fe 2 O 3 -based materials simultaneously exhibiting good electrocatalytic activity (measured without lightirradiation) and photo-electrocatalytic activity (measured with light-irradiation) for water oxidation. This is most probably because they are in a trade-off relationship; namely, defects increase the electron conductivity required for the former catalysis, while they induce recombination pathways for the photogenerated carriers that are avoided for the latter. Considering that SMART was conducted at a low temperature of 75 C, a SMART-derived a-Fe 2 O 3 layer would be more defective than those synthesized by high-temperature methods, which could be the main reason for the remarkable catalytic activity of SMART-derived a-Fe 2 O 3 .
Subsequently, we investigated the effects of thermal annealing in air on the electrochemical properties of SMARTderived a-Fe 2 O 3 to understand the correlation between the catalytic activity and local structures in a-Fe 2 O 3 layers. Fig. 4a shows the LSV curves of the SMART-derived a-Fe 2 O 3 layer, asdeposited and subsequently annealed at 300 C and 500 C. Aer annealing at 300 C, the cathodic current density increased slightly. The catalytic activity was signicantly enhanced with the overpotential of 370 mV at 0.5 mA cm À2 aer annealing at 500 C. We assumed that the activity would decrease aer the annealing in air due to the elimination of defects. However, the trend of the experimental results was the opposite of our expectations, indicating that the a-Fe 2 O 3 layer remained defective aer annealing. In fact, the photo-response remained negligible aer annealing at 500 C (Fig. 4b).
XRD analyses were performed to collect information about the local structure changes of a-Fe 2 O 3 layers aer annealing  This journal is © The Royal Society of Chemistry 2020 Nanoscale Adv., 2020, 2, 3933-3941 | 3937

Paper
Nanoscale Advances ( Fig. 5a and b). First, no additional peaks emerged aer annealing up to 500 C, supporting that the as-deposited lms were mainly composed of crystalline a-Fe 2 O 3 . Second, the average crystallite size slightly increased to 54.6 nm from 47.4 nm aer annealing at 500 C. Third, the d-value calculated from the 104 peak position corresponded to 2.719Å, 2.709Å, and 2.707Å for the as-deposited lm and ones subsequently annealed at 300 C and 500 C, respectively, while the d-value obtained from a reference bulk crystalline a-Fe 2 O 3 was 2.703Å (Fig. S6, ESI †). Thus, the lattice expansion occurred in asdeposited a-Fe 2 O 3 , and the lattice shrank to the bulk value aer the annealing. XPS was performed to detect the change in chemical states of the a-Fe 2 O 3 aer annealing. Fig. 5c-f show Fe 2p and O 1s spectra of SMART-derived a-Fe 2 O 3 with and without annealing at 300 C and 500 C. In the Fe 2p spectra, a Sn 3p 3/2 peak located at 716 eV appeared aer annealing at 500 C. It was also found that Sn 3d peaks in the wide-scan spectrum were pronounced aer the annealing (Fig. S7, ESI †). Thus, Sn 4+ ions, involved in the conducting substrate, would be thermally diffused in the a-Fe 2 O 3 lattice. In fact, this phenomenon can be found in the literatures. [38][39][40] Sn 4+ exhibits a similar ionic radius and Pauling electronegativity to Fe 3+ ions, which facilitates the substitution of Fe 3+ to Sn 4+ in a-Fe 2 O 3 . Importantly, Sn 4+ -doping has been found to be an effective approach for tailoring the electronic properties of a-Fe 2 O 3 . [41][42][43] In the present case, the catalytic activity was likely enhanced by an increased electron conductivity by Sn 4+ -doping. 43 Two binding energies of O 1s (529.5 eV and 531.0 eV) were assigned to the O 2À and OH À , respectively. 44,45 The ratio of OH À /O 2À decreased aer annealing at 300 C. Thus, we suggest that the observed lattice expansion in the as-deposited a-Fe 2 O 3 layer was caused by the OH species. 46,47 Finally, the intensity of the Cl 2p was quite weak in the as-deposited sample. The peak was almost undetectable aer annealing at 300 C, excluding the signicant contributions of Cl À to change the catalytic activity upon annealing (Fig. S8, ESI †).

Enhanced catalytic activity at Ni(OH) 2 /a-Fe 2 O 3 heterointerface
We revealed that defect-engineering by annealing is effective in enhancing the electrocatalytic activity of a-Fe 2 O 3 . Here, we further extended the LP-ALD concept to tailor the catalytic activity, where the Ni(OH) 2 layer was decorated onto the surface of the a-Fe 2 O 3 lm aer annealing at 500 C. The Ni(OH) 2 layer was deposited SILAR method, 18,48 referred to as the most relevant deposition technique to SMART. In SILAR, metal ions were adsorbed onto the surface followed by rinsing with water. In the next step, metal cations reacted with an alkaline solution to form a metal hydroxide layer via classical precipitation reactions (eqn (1)). In fact, this attempt signicantly improved the OER activity; as shown in Fig. 6a, the overpotential was lowered to 250 mV at 0.5 mA cm À2 aer the Ni(OH) 2 -modication. The overpotential at the same current density was ca. 50 mV lower than those from the best a-Fe 2 O 3 -based catalysts, Ni-or Zn-doped a-Fe 2 O 3 , reported so far. 36 No degradation of catalytic performance was observed aer 100 scans, which indicated the catalytic durability. The Ni(OH) 2 -modied a-Fe 2 O 3 showed better catalytic activity than SILAR-derived Ni(OH) 2 , where the overpotential of a Ni(OH) 2 layer obtained aer 30 deposition cycles was 320 mV at 0.5 mA cm À2 . Fig. 6b and c show the Fe 2p and Ni 2p XPS spectra of Ni(OH) 2 -modied a-Fe 2 O 3 , respectively. There is no remarkable change in the features of the Fe 2p spectra, indicating that the Fe-O-Fe framework containing oxygen vacancies was not altered by Ni(OH) 2 -modication. Based on the relative peak intensity of the Ni 2p and Fe 2p spectra, the Ni : Fe atomic ratio is approximately 1 : 3. Considering the analytical depth of the XPS (ca. 4 nm), the thickness of the Ni(OH) 2 layer is estimated to be 1 nm. Besides, SEM-EDX analysis revealed that Ni signal was homogenously detected on the whole surface of the a-Fe 2 O 3 layer, while there was no morphological change on the surface (Fig. S9, ESI †). In addition, no additional reection peaks from Ni-based phases were detected in the XRD pattern of the Ni(OH) 2 -modied sample (Fig. S10, ESI †). This indicates that particulate Ni(OH) 2 was not formed, while Ni(OH) 2 would be uniformly formed on the a-Fe 2 O 3 surface. These results support that catalytic activity was modied through the formation of Ni(OH) 2 /a-Fe 2 O 3 heterointerface.
Finally, for perspective, we propose that the LP-ALD process could provide a platform to create articial two-dimensional (2D) heterostructures. 2D heterostructures such as BaTiO 3 / SrTiO 3 superlattice were initially fabricated by a vacuum process, 49 and recently, hetero-assembly of 2D nanomaterials such as graphene, as well as 2D transition metal dichalcogenides and 2D oxides have attracted considerable attention for tuning functionalities by interface coupling. 50,51 Although LP-ALD, including SILARs and solution-ALD, 52 has only been employed for the deposition of inorganic layers with single components, the layer-by-layer deposition principle is applicable to produce such 2D heterostructures. The Ni(OH) 2 /a-Fe 2 O 3 heterointerface with excellent OER activity, found in the present study, partially demonstrate the above strategy. However, a true understanding of the enhancement of the heterointerface remains challenging because of the complex nature of the surface system, and thus is beyond the scope of this current study. For example, we found that enhancement of OER activity was less signicant, when Ni(OH) 2 was deposited on the as-deposited a-Fe 2 O 3 layer (Fig. S11, ESI †). Presently, we expect that the periodical 2D heterostructures, such as alternately stacked Ni(OH) 2 /a-Fe 2 O 3 layers, would serve rich chemistry, affording superior catalytic activity. This will be the target of our next study.

Conclusion
In conclusion, we established SMART for the direct solution deposition of a-Fe 2 O 3 layers on oxide substrates. This method yielded a a-Fe 2 O 3 layer with a 150 nm thickness and a crystalline size of 47.4 nm aer 90 deposition cycles. The growth rate was ca. 1.7 nm per deposition cycle, in which Fe 2+ cations in a Stern layer were oxidized by NaNO 2 to form Fe 3+ followed by consumption by crystal growth. Thus, the designed reaction route for the a-Fe 2 O 3 layer was experientially demonstrated. OH À ligands were introduced in the lattice of a-Fe 2 O 3 crystallites, probably because of the low-temperature aqueous process. The defective feature of SMART-derived a-Fe 2 O 3 activated and deactivated electrochemical and photoelectrochemical activity for water oxidation, respectively. The annealing in air introduced the Sn 4+ ions in the a-Fe 2 O 3 layer by the thermal diffusion from the substrate, which enhanced the electrocatalytic activity. Finally, we found that Ni(OH) 2 /a-Fe 2 O 3 heterointerface provided excellent OER activity, which would be crucial to develop stable, cost-effective, and environmentally-friendly OER catalysts.

Conflicts of interest
There are no conicts to declare.