Recent progress in iron oxide based photoanodes for solar water splitting

Solar assisted water splitting in a PEC is an attractive concept to store solar energy as hydrogen fuel but the effective efficiency of the process is too low for it to be a serious contender for commercialization. The most important component of the PEC to achieve efficient water splitting is a photo active anode that could effectively absorb photons and deliver holes for the oxygen evolution reaction. Hematite has many attributes that make it a good candidate material for a photoanode but it also has some deficiencies. This article reviews the state-of-the-art hematite-based photoanodes, with special emphasis on attempts made by researchers to overcome its drawbacks. The numerous research reports are categorized under distinct strategies such as nanostructuring, elemental doping, surface passivation, cocatalyst application, conducting template incorporation and heterostructures which are possible pathways to improve the performance of hematite. The scientific understandings of the operating mechanisms for each strategy are systematically presented and discussed, and the improvements achieved by different approaches are compared. Some cutting-edge strategies, such as heterojunctions, could be important and hematite based heterostructures are discussed. A developing interest in an emerging material, Fe2TiO5 is also discussed and some of the important benefits of this material are presented. Finally, the importance of scaling up the technology is discussed and attention is drawn to some possible challenges on scaling up. The paper concludes with some future technology directions and prospects for hematite-based photoanodes.


Introduction
Rapid economic and demographic growth have been driving the upsurge in worldwide energy consumption. To date, fossil fuels continue to meet more than 80% of the total primary energy demand and over 90% of greenhouse gas emissions [1,2]. One of the most abundant sources of clean energy, solar energy, is freely available on earth surface, however, the main challenge is to convert and store solar energy by efficient and cost effective methods on a large scale. Solar driven water splitting is an interesting pathway to harvest solar energy and generate hydrogen gas, mimicking the first stage of the natural photosynthesis process in plants, where solar energy is captured and ultimately converted to carbohydrates evolving oxygen [3][4][5]. In a photoelectrochemical (PEC) cell, water molecules are broken into hydrogen on the surface of photocathode, and oxygen on the surface of a photoanode. Ideal photoanode/cathode materials should have a suitable energy bandgap for sunlight harvesting, appropriate band edges with water redox reactions, high absorption in visible wavelength range, good electrical conductivity, good chemical stability, nontoxicity, and economic viability [6,7]. For the purpose of optimizing the photoanode/cathode stack, the other electrode might be replaced with a counter electrode made of Platinum to form a half PEC cell. Compared to the H 2 evolution reaction at the photocathode, the four-electron transfer oxygen evolution reaction (OER) at the photoanode is the rate-limiting step in the PEC water splitting process, thus the development of high-efficiency photoanodes is of paramount importance.
After the pioneering work of Fujishima and Honda on TiO 2 photoanodes for PEC water splitting, various other photoactive materials such as Fe 2 O 3 , WO 3 , BiVO 4 , TaON, GaP, GaInP, np + -Si, and Ta 3 N 5 have been examined for anode [8][9][10][11][12][13]. Metal oxides are a common choice for photoanode material mainly due to their semiconducting nature, chemical stability in aqueous solutions and reasonably low cost [6]. However, most metal oxides have wide energy bandgaps and poor electrical properties (carrier concentrations and mobilities) when compared with traditional III-V semiconductors and silicon.
The US Department of Energy estimates the hydrogen threshold cost to be $2.00-$4.00 per gallon [14], while a recent report predicts the cost of hydrogen production via electrolysis to be $3.26-$6.62 per gallon [15]. To meet the DOE cost target, a PEC device should cost less than US$ 160 per m 2 with a solar-to-hydrogen (STH) efficiency of about10% [16] whereas the theoretical maximum efficiency of most of the metal oxide based photoanodes is still below 15% based on their energy band gap (see table 1

below).
Among metal oxide photoanodes, TiO 2 has received considerable attention due to its energy band offsets, but its high energy band gap and inefficient absorption of visible spectrum have limited its performance [26]. The performance of bismuth vanadate (BiVO 4 ) has been improved in recent years by surface and bulk modifications and has reached 80.7% of its theoretically predicted STH efficiency [11,21,27]. However its poor stability in the electrolyte environment remains to beaddressed. WO 3 is another oxide that has been investigated widely but it exhibits poor performance and is not stable in alkaline conditions frequently used in PEC [10]. np + -Si photoabsorber integrated with various co-catalysts shows excellent PEC water oxidation performance but it also requires a protection layer to guard against corrosion [12]. Oxynitrides, which can undergo both water oxidation and reduction, are also widely studied because of their suitable energy band gap but their reported quantum efficiency (~5%-6%) is still too low [28,29]. Out of all the candidate materials, hematite (Fe 2 O 3 ) seems to have the most favorable properties of high theoretical STH efficiency and stability [24,30].
Hematite (α-Fe 2 O 3 ), is a naturally occurring ore of iron which is the fourth most abundant element in earth's crust (6.3% by weight) [31,32]. It has many desirable characteristics of a good photoanode material such as good chemical stability in aqueous media, suitable energy band gap (1.9-2.2 eV) for light harvesting, low cost and environment friendliness [33]. Based on its energy bandgap, hematite can achieve a theoretical maximum STH efficiency of 15%, which exceeds the STH benchmark efficiency of 10% forecast for commercial applications [33,34]. However, in practice, its performance is limited by low carrier concentration, poor electronic properties and slow charge transfer kinetics at electrode/electrolyte interface leading to a high overpotential for water oxidation and a lower than the predicted STH efficiency [33,35]. Several reviews have been published highlighting significant achievements made in improving hematite photoanode performance using various approaches. In our previous review article, we discussed various nanostructures, doping strategies and surface modifications examined in hematite photoanodes with an aim of enhancement of charge carrier dynamics in PEC water splitting. Significant progress has been made on improving the performance of hematite photoanode since 1978 (figure 1). A recent report by Wonyong Choi's group reveals 6 mA cm −2 current density at 1.23 V versus RHE, which is ~50% of the theoretically predicted photocurrent density based on the energy band gap [30]. However, further improvement of hematite performance still has to be pursued by alternative strategies, such as by fabricating heterostructures to incorporate a second material to promote charge separation, charge collection, and surface catalysis. In this review, we summarize recent efforts to understand and realize the potential of hematite photoanode and discuss possible strategies to address the outstanding challenges.

Hematite (Fe 2 O 3 ) photoanodes
As a potential photoanode material with excellent stability, hematite has an appropriate valence band for water oxidation (figure 2) [33,34,40] even though the conduction band minima is not pertinent for water redox level [43]. Figure 2 shows the band offsets of hematite before and after interaction with an electrolyte. The use of hematite in thin film or bulk forms is not successful because of the competing requirement between its low minority carrier diffusion length (2-10 nm) and its need for high absorption depth (~120 nm at 550 nm wavelength) to convert significant amount of incident energy [33,44]. Because of the low hole diffusion length, only holes generated within a few nanometers of the hematite electrolyte-interface can be useful for water oxidation.
A good photoanode should have a high plateau current and a low onset potential. In our previous review, we argued that manipulations in nanostructures and doping strategies are necessary to increase the plateau current, while the surface properties of the semiconductor need to be improved to lower the onset potential [34]. In recent times, there has been many advancements in the nanostructure synthesis, bulk and surface modifications of hematite by forming heterostructures with various materials which overcome the obstacle of low minority carrier diffusion length and achieve improvements in photon absorption [24,36,38,45]. Synthesis of different nanostructures (nanorods, dendritic nanowires, cauliflowertype, nanoflowers etc,) of hematite photoanodes have been reported by various groups [46][47][48][49].

Strategies to overcome bulk property limitations of hematite
The low inherent electrical conductivity of bulk hematite is one limitation that needs to be addressed [50,51]. Undoped hematite has a very low electrical conductivity of 10 −14 Ω −1 cm −1 , and electron mobility of 10 −2 cm 2 V −1 s −1 [52][53][54][55]. These characteristics could be improved by adding suitable dopants to hematite. When tetravalent dopants, such as Ti, Zr, Sn, Mn, Ge and Si are added to replace some of the Fe 3+ cations, the extra electrons donated to the crystal will increase its conducting electron concentration and retain its n-type conductivity [24,47,[56][57][58]. When divalent dopants, such as Mg, Cu and Ni are added to replace Fe 3+ cations in the crystal, hole carriers are created making the hematite p-type [59][60][61]. A list of dopants investigated in hematite and the resulting PEC performances are given in table 2. Elemental doping enhance the PEC performance of hematite by; (1) increasing the intrinsic conductivity, (2) generating a shallow donor level below the conduction band minima, which alter the absorption, (3) large cation substitution induces the strain within lattice and thus recombination centers. Fe 3+ cation has an ionic radii of 55 pm and substituting a cation with larger ionic radii than 55 pm will create additional defects. In our previous study, Mn 4+ cation were doped in hematite, which has similar ionic radius (53 pm) as Fe 3+ (55 pm). Mn doping improves the PEC performance of hematite by increasing the carrier conductivity without creating the additional defects (similar UV-vis spectra with and without doping) and also enhance the OER kinetics at the electrode/electrolyte interface because of its good OER catalytic property [47].
One of the most commonly used dopants in hematite is Sn which has bigger ionic radii (69 pm) than Fe 3+ . Annamalai and coworkers shows that Sn 4+ doping improves the photocurrent of hematite and discussed that the Sn 4+ doping induced a shallow donor level below the conduction band of hematite, which does not contribute to increase electrical conductivity because of its localized nature [62]. Additionally, Sn 4+ introduction induce local micro-strain and a decreased Fe-O bond ordering. Sn 4+ donor dopants in hematite lattices introduce electrons to Fe 3+ sites and reduce to Fe 2+ . These Fe 2+ sites can thus improve the electrical conductivity of hematite photoanodes through the polaron hopping [62,63]. Sn can be added intentionally. Alternatively, it could get introduced into hematite by diffusion from FTO glass which is frequently used as the conducting substrate in anodes, during its post annealing treatments [24,56,57,64,65]. In a study, it was shown that Sn 4+ doping in a hematite nanorod anode improved its photoactivity, and its carrier concentration increased from 7 × 10 19 to 2.45 × 10 20 cm −3 [64]. Our group recently demonstrated a dual effect of Sn doping [24]. Sn doping was achieved by depositing a thin overlayer of SnO x via atomic layer deposition (ALD) on FeOOH nanorods followed by annealing at 650 °C. This Sn doped hematite exhibited 3.1 mA cm −2 photocurrent density at 1.23 V versus RHE and also showed a cathodic shift in the onset potential when compared to the undoped hematite which was also annealed at 650 °C ( figure  3). High annealing temperatures usually damage the FTO and increase its series resistance and hence, the lower annealing temperature used in this study was beneficial. The annealing temperature could be further reduced by employing two layers of SnO 2 , an underlayer and an overlayer, to facilitate Sn diffusion from both sides during annealing [24,66,67]. Si 4+ has also been shown as an effective dopant for hematite photoanodes. Gratzel et al reported Si doped, cauliflower structured hematite with a high donor density of 10 20 cm −3 , which could enhance its PEC performance [68]. This cauliflower type, Si doped hematite photoanode exhibited a current density of 2.2 mA cm −2 at 1.23 V versus RHE. A large area (80 cm 2 active area), Si doped hematite synthesized by a dip coating method, exhibited ~1 mA cm −2 photocurrent density at 1.23 V versus RHE [69].
Ti is another element investigated for doping hematite [50,56,[70][71][72]. In 1978, Kennedy et al reported that Ti doped hematite increased the carrier concentration to a level > 10 19 cm −3 [40]. They also studied the effect of different electrolytes on the performance of hematite and found a higher photoconversion efficiency and a lower onset potential in 2 M NaOH (pH 13.8). Jiujun et al, also synthesized Ti doped hematite nanostructures by a hydrothermal method [56] which showed an urchin-like morphology, with enhanced effective surface area compared to undoped nanostructures. They obtained a remarkable plateau photocurrent density of 3.76 mA cm −2 for the Ti doped nanostructures under AM 1.5G in 1M NaOH electrolyte, which was 2.5 times higher than that obtained for undoped nanostructures (1.48 mA cm −2 ) [56]. Recently Cho et al demonstrated a facile flame Ti doping method which reduced the onset potential (380 mV) and enhanced the PEC performance of hematite photoanodes [73]. This onset potential reduction is attributed to decreased bulk and surface recombination and a dense Fe 2 O 3 under layer that reduced the back recombination. Oxalic acid and FeOOH surface treatment of Ti doped Fe 2 O 3 helps to further improve the quality of electrode/electrolyte interface. Moreover, the Ti doped sample shows about two times higher photocurrent density of 1.58 mA cm −2 at 1.23 V RHE and 30% IPCE at 300 nm [73].
Even though it is not frequently reported, it is worth noting that another group IV element, Ge, has also been examined as dopant in hematite. A Ge-doped hematite film has been prepared by Liu et al using a hydrothermal approach with highly reactive Ge colloidal solutions used as dopant sources [74]. Ge-doped hematite nanosheet arrays showed a photocurrent density of 1.4 mA cm −2 at 1.23 V versus RHE, which was more than 50 times that of undoped hematite nanorod arrays. This improvement is thought to originate from the two orders of magnitude higher donor density of Ge-doped hematite than that of the undoped sample, which enhanced the electrical conductivity.
Ru doped nanoporous hematite nanorod photoanode was reported by Xueli et al which was synthesized via a doctor blade method followed by 700 °C annealing at ambient atmosphere [36]. The optimized Ru doped hematite photoanode displayed a record photocurrent density of 5.7 mA cm −2 1.23 V versus RHE and an onset potential of 0.7 V versus RHE. This excellent PEC performance is probably due to the enhanced charge carrier concentration and mobility [36].
To summarize, introducing optimum doping concentrations has been shown to be beneficial for altering the inherent electronic properties of hematite, which in turn enhances the plateau photocurrent of PEC cells. It should be emphasized that the actual effectiveness of any of these intentional elemental dopants is sometimes difficult to isolate because of the effects of contaminant Sn, which unintentionally diffuses from FTO substrates during the high temperature, post annealing (750 °C-800 °C) of hematite [24,63,65,75]. In addition to the issue of low carrier concentration, one needs to address the interfacial reaction limitation of hematite photoanodes also to improve the photoanode performance. Surface treatment with co-catalysts or depositing surface passivation layers are possible approaches to reduce the charge recombination at electrode/electrolyte interface. This will be discussed in the next section [76][77][78][79][80].

Strategies to overcome the surface property limitations of hematite
An additional shortcoming of hematite photoanode is the need for an overpotential which is attributed to the sluggish hole transport across the semiconductor/electrolyte interface [78,[91][92][93]. The overpotential required for hematite can be reduced either by lowering the potential-dependent rate constant for surface-mediated charge recombination, or by increasing the rate constant for hole transfer from the photoelectrode to the molecular reactant [73,77,94]. The former can be accomplished by passivating the localized electron trap states on the surface, e.g. by deposition of overlayers or by chemical treatments [76,91]. The rate constant for hole transfer across the interface could be improved by deposition of an OER catalysts. In this section we will discuss the recent attempts reported to lower the overpotential of hematite photoanodes for PEC water oxidation [80,94].
OER catalysts loaded on the surface of hematite facilitate water oxidation reactions usually through the oxidation of the metallic element of the catalyst, which inject holes from the hematite surface into the electrolyte [78,95]. Various OER catalysts like CoPi [94,96], IrO 2 [42,78], CoO x [44,97], MnO [95] and FeNiOx [80,98] have been extensively studied on hematite photoanodes (table 3). Nocera et al developed the CoPi OER catalyst, which they integrated with various photoanodes, including hematite, as a cocatalyst [96,99], and studied its working mechanism by XANES and EPR techniques. They concluded that the metallic element Co in CoPi oxidized from Co(II) to Co(III) and Co(IV), leading to the formation of high-valence Co(IV)-O intermediates [100,101]. Since the development of CoPi, it has gained a lot of acceptance because of its low overpotential of only 0.41 V required to oxidize water at pH 7 [102]. IrO 2 is a well-known OER catalyst    [118] which shows lower over potentials of 250 mV and 300 mV in alkaline and neutral pH conditions respectively for water oxidation [103,104]. Tilley et al demonstrated that an IrO 2 nanoparticle coated hematite photoanode could generate an excellent photocurrent density of 3.0 mA cm −2 at 1.23 V versus RHE and a 200 mV reduction in overpotential [42]. IrO 2 nanoparticles on the surface of hematite catalyzed the transfer of photogenerated holes resulting in lowering of the required overpotential while a higher photocurrent was achieved by nanostructuring of the hematite which ensured lower charge carrier recombination.
Recently, FeNiO x OER catalyst is extensively used because of its earth abundance, high stability in alkaline solution and easy integration with any photoanode [105][106][107]. Xile Hu's group integrated FeNiO x cocatalyst with cauliflower type hematite via PEC deposition. This FeNiO x -coated nano Fe 2 O 3 sample shows ~250 mV and 100 mV cathodic shifts in onset potential as compared to bare hematite and CoPi treated hematite photoanodes, respectively [80]. Dunwei Wang's group also reported similar FeNiO x coated hematite with an impressive cathodic shift of ~550 mV in onset potential where a simple regrowth mechanism has been used to synthesize the hematite photoanode, and FeNiO x cocatayst was deposited subsequently by a drop casting method [93]. Recently, our group demonstrated that an ultrathin MnO could be an effective cocatalyst with hematite anode for OER [95]. Ultrathin MnO nanoparticles integrated with hematite photoanode exhibits a ~200 mV cathodic potential shift and improved PEC performance, which is comparable to CoPi (figure 4). The detail findings of this study will be discussed in the next section [95].
Another common strategy to overcome the sluggish hole injection from hematite surface to electrolyte is to reduce the surface recombination rate by using thin overlayers of wide band gap metal oxides. A few surface passivation strategies such as TiO 2 [77], Al 2 O 3 [76], Ga 2 O 3 [91], and SnO x [24] have been reported so far.
Recently, Wonyong Choi's group reported good stability for over 100 h in a hematite photoanode with photocurrent density of 6 mA cm −2 at 1.23 V versus RHE under 1 sun illumination [30]. This excellent performance of hematite nanorod photoanode could be attributed to the combined effects of H 2 treatment, TiO 2 and CoPi overlayers. Hydrogen treatment enhances the electrical conductivity, ultrathin TiO 2 overlayer reduces the surface charge recombination, and CoPi enhances the rate of reaction at the electrode-electrolyte interface [30]. Mahmoud et al also studied a simple solution grown TiO 2 as a surface passivation layer on nanostructured hematite surface and showed a cathodic shift of 190 mV in onset potential and 4.5 times improvement in photocurrent density [77]. Our group showed that an ultrathin TiO 2 layer deposited on hematite nanorods by ALD exhibits 1.9 mA cm −2 photocurrent at 1.23 V and ~200 mV cathodic shift in onset potential [108]. More recently, Piangjai et al discussed the combined effect of plasmonic Ag nanoparticles and CoPi catalyst with hematite photoanode [38]. This Ag and CoPi nanoparticles modified hematite showed an excellent photocurrent density of 4.68 mAcm −2 at 1.23 V versus RHE which is attributed to the presence of plasmonic Ag nanoparticles which improved light harvesting and facilitated charge transfer while Co-Pi works as a cocatalyst for water oxidation and helps to reduce the surface recombination [38].
We demonstrated that the recombination sites on hematite surface can be reduced by passivating the nanorods with another layer of hematite [109]. A core-shell type architecture of hematite nanorod photoanode was produced where a thin hematite shell was sprayed on hematite nanorods which has proven to partially passivate these surface effects. Dunwei Wang's group also reported similar effect by regrowth of a thin FeOOH layer over hematite followed by annealing [93]. This regrowth strategy resolves the near surface disorder and hence improves the photovoltage generated by hematite (figure 5) [93].

Surface charge transfer mechanism of hematite
As discussed in section 2.2, the sluggish hole transfer from the electrode to electrolyte at the interface can be enhanced by employing cocatalysts or by noncatalytic passivation layers. These two techniques are now understood to have fundamentally different operating mechanisms in assisting the hole transport. Surface passivation (a non-catalytic process) implies the reduction of hematite surface defects which reduces the hole-electron recombination rate, while a cocatalyst usually implies the oxidation of the metallic element of the cocatalyst which enhances the rate of hole transfer from the hematite surface to the electrolyte [43,95,119].
Many attempts have been made to shed the light on the operating charge transfer mechanisms through systematic studies using intensity-modulated photocurrent spectroscopy (IMPS) [43,95,120]. IMPS allows the charge carrier dynamics at the electrode/electrolyte interface to be de-convoluted into contrib utions from charge transfer and surface recombination. It is an attractive technique for understanding the surface carrier dynamics at the semiconductor/electrolyte interface. In our previous study, we reported the surface effects of manganese (Mn) doping, and of a thin passivation (shell) layer of highly crystalline Fe 2 O 3 on hydrothermally grown hematite nanorods [121]. Our study revealed that although the photocurrent enhancements achieved by both modifications are similar, the operating mechanisms are fundamentally different. Mn doped hematite shows an enhancement in PEC performance due to increase in hole transfer constant, while the core-shell hematite shows a decrease in charge recombination rate constant (figure 6) [121].
This observation indicates that the core-shell architecture is effective in improving the photocurrent by suppressing surface recombination, rather than by increasing charge transfer. The finding that Mn doping improves the surface hole transfer kinetics was rather unexpected but a similar observation has been reported in Sn-doped hematite by Dunn et al [120] who demonstrated that Sn-enrichment at the surface improved the catalytic properties at the surface (figure 7).
We have also shown by IMPS in an ultrathin MnO nanocrystals decorated hematite and CoPi treated hematite (figure 8) that the MnO loaded hematite sample shows an increased hole transfer rate constant compared to bare hematite. The CoPi treated hematite, on the other hand, shows a decrease in charge recombination rate, while hole transfer rate was similar to that in bare hematite [95]. Peter et al also reported a similar observation on Sn doped hematite [120]. The increased transfer efficiency brought about by Sn-doping hematite was attributed to enhanced catalysis of the OER rather than to surface passivation. While it may not be clear at present how the Sn dopant beneficially impacts on the OER kinetics, the insight that dopants can speed up the sluggish OER is an important milestone in the optimization of photoanodes for water oxidation [120]. Dunwei Wang's group also studied NiFeO x coated hematite photoanode by IMPS technique [122]. They showed that the best performing hematite, in terms of photocurrent onset potential, exhibited the slowest water oxidation rate constants. When amorphous NiFeO x , a water oxidation catalyst, was present, the rate of surface holes transfer actually slowed down; resulting in low recombination rate at the hematite surface (figure 9). They concluded that the NiFeO x primarily serves as a passivation layer rather than a catalytic layer [122].
Finally, the need for overpotential in water splitting is attributed to two causes: (i) the position of the valance band of hematite, which is 0.4-0.5 V too positive to allow water splitting and, (ii) the sluggish hole transport at the semiconductor/water interface. This opens a new direction of search for a better, alternative iron-based material for highly efficient water splitting which must be robust, stable and must have good carrier conductivity. More importantly, its energy band gap should not be more than 2.2 eV because it limits the maximum STH conversion efficiency.

Fe 2 TiO 5 as a photoanode material
In previous sections, we discussed the strategies employed to improve Fe 2 O 3 based PEC devices by doping and by surface treatments with thin overlayers and cocatalysts. However, these actions do not address the underlying limitations of hematite namely small minority carrier diffusion length and poor charge transport properties. Although, these limitations have been mitigated somewhat by nanostructuring to increase the space charge layer; surface charge transfer kinetics and poor OER of hematite are still significant impediments for the realization of commercial PEC cells. This makes it inevitable that we explore new systems with better charge transport properties, longer diffusion length of minority carriers and improved surface reaction kinetics. It has been noted that some complex metal oxides may have desirable properties for consideration as an alternative to hematite. A report on such complex metal oxides has been recently compiled by Abdi et al which uncovers a range of materials and their development for application in solar water splitting [123]. One way to improve the intrinsic properties of Fe 2 O 3 is to amalgamate with another material, such as TiO 2 , which inherently has good carrier conductivity and large charge diffusion length. A scrutiny of the Fe-Ti-O ternary system might give indications for an alternative material with the best combination of the desired properties of both Fe 2 O 3 and TiO 2 . This is corroborated by the evidence that Ti is an effective dopant to tune the band level positioning in Fe 2 O 3 . When exploring a novel material for photoabsorber, it is imperative to understand both its bulk and surface characteristics. Surface characteristics are often correlated with the morphology of films and post deposition treatments, but bulk characteristics depend on the intrinsic electron band and crystal structures.
One candidate material in the Fe-Ti-O ternary system that has received considerable attention in recent times is Fe 2 TiO 5 pseudobrookite which possesses an orthorhombic structure with a Cmmm space group [124]. The cations in the Fe 2 TiO 5 crystal are located in two different octahedral sites M1 and M2 but the actual distribution of Fe 3+ and Ti 4+ ions in these sites has not been conclusively determined. Fe 2 TiO 5 comprises abundant and non-toxic elements and can be produced simply by the oxidation of ilmenite ore at temperatures above 1000 °C with rutile as the byproduct [125]. In the form of a film, it has a good thermodynamic and aqueous stability in a wide pH range. These properties make it a potential candidate as a photocatalyst for solar water splitting.
Some early reports on Fe 2 TiO 5 have been on its crystal growth and its characterization [127][128][129]. It is an n-type semiconductor with band gap of around 1.9-2.1 eV which covers a large part of solar spectrum. In a recent report, Guillaume et al presented a crystal structural analysis for Fe 2 TiO 5 pseudobrookite composition (shown in figure 10(a)) [126]. Through various coupled techniques like 57 Fe Mossbauer spectroscopy, XRD etc, they deduced that the optical properties and the phase purity strongly depend on the thermal history and the chemical composition which influences the cationic vacancy concentration. It was also noted that obtaining the pure phase Fe 2 TiO 5 by solid state reaction route is very challenging because a hematite-rich phase is usually obtained in this synthesis. This issue has been tackled by adding an excessive amount of titania to the desired Fe 2 TiO 5 stoichiometry (i.e. Fe +3 /Ti +4 < 2) in the starting reactants.
There have been reports on the dielectric properties and electrical conductivity of Fe 2 TiO 5 bulk samples [130][131][132]. Sharma et al investigated the complex dielectric and impedance properties of Fe 2 TiO 5 and reported that its non-Debye type dielectric relaxation follows a thermally activated process [131]. This was attributed to the polaron hopping type conductivity mechanism which was confirmed by AC conductivity analysis. Nikolic et al reported changes in electrical conductivity in accordance with Jonscher's power law for nanopowder and bulk samples [130]. They also concluded that the quantum mechanical-tunneling model for the case of small polaron hopping can be applied at higher temperatures which outlines the conductivity in Fe 2 TiO 5 .
The pseudobrookite, Fe 2 TiO 5 phase has exhibited diverse characteristics like the gas sensing ability, magnetic properties and photocatalytic properties which have been exploited previously for various applications [133][134][135][136][137][138]. Very recently, a novel yellow pigment, Fe 2 TiO 5 decorated mica composites was evaluated for its near infrared reflectance ability which makes it a good candidate as solar reflective coating [139]. The near infrared solar reflectance of this composite was demonstrated to be as high as 80.3% which could reduce the interior temperature in a building by about 3 °C. This characteristic could be utilized for thermal insulation of solar devices also in the large scale.
Many researchers have reported synthesis of nano and micro particles of Fe 2 TiO 5 and examined their potential in a variety of applications. Growth of nanocrystalline Fe 2 TiO 5 perovskite thin films assisted by microwave irradiation was studied by Phani et al [140] while the synthesis of Fe 2 TiO 5 nanoparticles was reported by Min et al who considered them for Li-ion electroactivity [141]. Synthesis of porous microparticulates of Fe 2 TiO 5 has been explored by Guo et al for its electrochemical activity as anode material [142]. Ku et al used Fe 2 TiO 5 as an oxygen carrier for chemical looping process. Its high equilibrium constants resulted in high syngas conversion in the fixed bed reactor [143].
An efficient photoanode must absorb a significant amount of visible light. The energy band gap of a material indicates the range of the wavelengths that could be absorbed from the solar spectrum, but it is the absorption coefficient that signifies the actual number of photons absorbed. The absorption coefficient of Fe 2 TiO 5 is found to be around 4.6 × 10 4 cm −1 at a wavelength of 500 nm which translates into a required thickness of 650 nm to absorb 95% of incident light (=3/α where α is absorption coefficient). This implies that the photoelectrode film should be about 650 nm thick to absorb most of the photons at 500 nm wavelength. This makes Fe 2 TiO 5 thin films vulnerable for insufficient photon absorption in compact or thin film forms. The important properties such as conductivity, mobility and carrier lifetime for Fe 2 TiO 5 have been largely unexplored. Since these properties contribute to the charge transport in bulk and significantly impact on their performance in PEC cell, it would be important to characterize them carefully.
We demonstrated the synthesis of pure phase Fe 2 TiO 5 by a solvothermal technique, using isopropanol as the solvent for the precursors [25]. Nanoporous thin films were synthesized on FTO substrates by this technique and coated with a SnO x overlayer for surface passivation to investigate its PEC performance. It yielded a photocurrent density of 0.35 mA cm −2 at 1.23 V versus RHE under 1 sun irradiation. Ultraviolet photoelectron spectroscopy (UPS) measurements showed that its band levels straddle the water redox levels with a work function of around 4.77 eV ( figure 10(b)). This was later confirmed by Deng et al who reported a similar valence band of around −6.31 eV for a Fe 2 TiO 5 layer synthesized by Ti treatment of a hematite layer [144].
Although we confirmed that our layer consisted pure phase of Fe 2 TiO 5 , the photocurrent density was still lower than those we obtained in Fe 2 O 3 based films which were of the order of 1 mA cm −2 . This is attributed to the low Fe 2 TiO 5 film thickness of 150 nm used which is insufficient for significant photon absorption. To realize the full potential of Fe 2 TiO 5, it is necessary to improve its quantum efficiency. Recently, An et al reported inverse opal Fe 2 TiO 5 prepared using polystyrene (PS) photonic crystals as templates [145]. Since a photonic crystal confines photons and allows coherent multiple scattering, it can potentially increase the overall absorption compared to a thin film. The performance of a Fe 2 TiO 5 photonic crystal synthesized from 250 nm PS spheres (measured using 250 mW cm −2 light source) showed higher photocurrent density when compared to the disordered sample. Therefore, photonic crystal certainly improved absorption which could be beneficial, in general, for other Fe 2 TiO 5 based PEC systems.
Very recently, Zhang et al used an iron titanate nanotube array with a TiO 2 underlayer, and modified it by hydrogen treatment and FeNiO x cocatalyst to improve its PEC activity [146]. The nanotube photoanode achieved a photocurrent density of 0.93 mA cm −2 at 1.23 V RHE under 1 sun (100 mW cm −2 ) irradiation with a cathodic shift in photocurrent onset potential of ~280 mV relative to the pristine nanotube array electrode ( figure 11). This is one of the highest performances achieved for a modified iron titanate based photoanode under standard conditions. The reports enumerated above catapulted the research activity on Fe 2 TiO 5 , especially on Fe 2 O 3 /Fe 2 TiO 5 heterojunctions and were later modified/integrated with other layers to realize higher performance PEC cells which would be discussed in the next section.  Rate constants extracted from IMPS data at different applied potentials. Rate constants of recombination are shown as black symbols; rate constants for surface charge transfer are shown as green symbols. The error bar is the standard deviation between different samples (typically 4-5 samples are measured for each group of data). Acronyms aH, sdH and rgH denote atomic layer deposited, solution grown and regrowth treated hematite respectively. Reproduced from [122]. CC BY 3.0.

Iron oxide based heterostructures
In the last decade, hematite has been used comprehensively in heterostructures with various photoanodes. Shen et al has reviewed the charge carrier dynamics and material design of such heterojunctions in detail [147]. In the last three years, many reports, specifically on integrating Fe 2 O 3 with Fe 2 TiO 5 layers and with other iron based metal oxides, have emerged. Such integrated systems have proven to be effective in tackling both bulk and surface issues related to recombination along with the slow kinetics of water oxidation. In the following section, the Fe 2 O 3 /Fe 2 TiO 5 heterojunction systems are discussed specifically.

Fe 2 O 3 /Fe 2 TiO 5 heterojunctions
In the last section, we focussed the discussion on developments in pristine Fe 2 TiO 5 as a photocatalyst or photoanode but it is worth highlighting that the photocurrent density reported is still lower than those obtained in pristine hematite based systems. This could be primarily due to the early, exploratory nature of the investigation into these novel material; optimization of its annealing processes, film morph ology and defect densities etc are yet to be achieved. In addition, Fe 2 TiO 5 possesses lower absorption coefficient requiring thicker films to absorb a significant amount of photons which implies that charge separation could become an issue even if significant number of photons are absorbed. Despite the low absorption coefficient, Fe 2 TiO 5 may still perform better than TiO 2 due to its lower band gap which could absorb a larger wavelength proportion of the sun light. One of the early state-of-the-art TiO 2 /Fe 2 TiO 5 heterojunction based PEC device was developed by Liu et al where they obtained a remarkable photocurrent density of around 4.1 mA cm −2 at 1.23 V versus RHE with low onset potential of 0.2V [148]. The theoretical maximum photocurrent of TiO 2 itself is only around 2 mA cm −2 based on its bandgap [149]. This means the additional photocurrent in this heterojunction is contributed by Fe 2 TiO 5 . This study attributes the enhanced photocurrent density to the good band alignment with TiO 2 (as shown in figure 12(a)) and CoO x which was used as an electrocatalyst. In this structure, Fe 2 TiO 5 acted as an ultrathin absorber layer for TiO 2 nanotubes which complemented narrow absorption range of TiO 2 nanotubes,  while the beneficial charge transport property of TiO 2 complemented the poor transport exhibited by Fe 2 TiO 5 . Such combination of TiO 2 and Fe 2 TiO 5 has also been utilized as a photocatalyst for other water oxidation reactions [150,151].
The band levels measured by UPS ( figure 10(b)) shows that the good alignment of Fe 2 O 3 and Fe 2 TiO 5 bands could enhance electron/hole separation due to their favourable band level offset. Deng et al produced such a heterojunction structure by depositing a thin TiO x layer on Fe 2 O 3 photoanodes and post annealing it to convert the TiO x layer into Fe 2 TiO 5 (shown in figure 12(b)). The ultrathin Fe 2 TiO 5 overlayer was fabricated by this solid-state reaction between Fe and Ti based oxides and hence limits the achievable thickness of Fe 2 TiO 5 . The Fe 2 TiO 5 layer was amorphous as confirmed by Synchrotron-based soft x-ray absorption spectroscopy. This structure is a thin heterojunction between Fe 2 O 3 and Fe 2 TiO 5 which appears to have improved the charge separation.
Courtin et al demonstrated nano heterostructured photoanodes synthesized via a sol gel synthesis method based on Fe x -TiO 2 with varied Fe content in TiO 2 [152]. They obtained the best photocurrent for films corresponding to x = 0.2, which contained Fe-doped anatase, pseudobrookite, and traces of hematite. Since the major phase in the product was pseudobrookite, it was proposed that it forms an intermediate conduction level between TiO 2 and Fe 2 O 3 and electron transfer is facilitated from Fe 2 TiO 5 to Fe 2 O 3 . Lately, doping of Ti in Fe 2 O 3 has been tried by Lin et al where they demonstrated electrodeposition of Fe 2 TiO 5 nanostructures [153]. When employed in water oxidation, this structure gave four fold enhanced the photocurrent density relative to pristine hematite. Such enhancement was attributed to local heterojunctions of Fe 2 TiO 5 with Fe 2 O 3 which probably decreased the hole accumulation.
Aforementioned heterojunctions either consisted of nanoscale junctions or comprised of ultrathin Fe 2 TiO 5 overlayers produced by solid-state reaction. While the former possibility is lacking the necessary quantitative information of the phases for further investigations, the latter possibility implies a Fe 2 TiO 5 layer whose thickness is limited by the diffusion of Fe and Ti and may result in irregular thickness or mixed phases of oxides. Our group reported an all-crystalline heterojunction of Fe 2 TiO 5 with Fe 2 O 3 by a hydrothermal route which facilitates a distinct and tunable electrical junction [154]. This was to facilitate efficient electron and hole transport to back contact and electrolyte respectively by forming a crystalline type II band alignment (as shown in figure 13(a)). FeOOH nanorods grown on FTO substrate by a hydrothermal route were used for sequential solvothermal synthesis of Fe 2 TiO 5 nanoparticle based film on top of the nanorods followed by thermal treatment. An appreciable photocurrent density of 1.4 mA cm −2 was achieved through the heterojunction as opposed to low photocurrent density of 0.01 mA cm −2 demonstrated by the pristine Fe 2 O 3 nanorods ( figure 13(b)). Through Electrochemical Impedance Spectroscopy (EIS), values of charge transfer resistance through surface states and space charge capacitance were extracted. The coincidence of the peak of C ss and the valley of R ct,ss with the jump in photocurrent density with increasing potential for optimized heterojunction (grey curve), as shown in figure 13(c), supports the viability of the model fitted for EIS. The model suggests that the photogenerated hole transfer into the electrolyte occurs through the surface states. The schematic representation of the 2 step hole transfer mechanism is shown in figure 13(d). Blue curve shows the first step of holes getting trapped at surface states and the red curve is the subsequent step of injection of trapped holes into the electrolyte. High surface charge separation efficiency of 85% indicated the role of Fe 2 TiO 5 in enhancing hole injection into the electrolyte. The onset potential shifted by 30 mV to 0.9 V RHE and a photocurrent of 1.6 mA cm −2 was attained through coupling with CoO x cocatalyst.
Wang et al reported a 4 to 8 times increase in the water oxidation photocurrent for Fe 2 O 3 /Fe 2 TiO 5 system with a significant cathodic shift of the onset potential of up to 0.53-0.62 V RHE [155]. The performance of this heterostructure was further improved by decoration of the surface with a SnO x layer owing to the synergistic effect of passivation of surface states by the latter. Li et al exhibited a 3D Fe 2 O 3 /Fe 2 TiO 5 heterojunction photoanode synthesized by a hydrothermal technique [156]. The titanate layer was formed using ALD on the core FeOOH followed by high temperature annealing. After coupling with FeNiO x cocatalyst, it yielded a photocurrent density of around 2.7 mA cm −2 at 1.23 V RHE with an onset potential of 0.8 V RHE . This is similar to our earlier work on ALD deposited TiO x ultrathin overlayer on FeOOH nanorods [108]. As opposed to the high temperature of 750 °C used by Li et al, we kept our annealing temperature to 650 °C to avoid the degradation of FTO and to control Sn diffusion from FTO substrate to hematite. We established that the surface passivation by the TiO x overlayer was responsible for the higher performance as compared to pristine Fe 2 O 3 nanorods. Very recently, Wang et al demonstrated a PEC device with three components Fe 2 TiO 5 /Fe 2 O 3 /Pt heterostructure with enhanced light harvesting through Pt nanoparticles [157]. This also had a synergistic charge transfer between hematite film and Fe 2 TiO 5 outer shell as well as with Pt underlayer. On the one hand, Fe 2 TiO 5 shell improved the hole transfer to the surface; on the other, Pt nanoparticles worked as an electron transfer mediator and a light trapping element. They achieved a photocurrent of 1 mA cm −2 at 1.23 V versus RHE under AM 1.5G illumination. Some other reports have used different modifications and treatments to improve the surface of Fe 2 O 3 /Fe 2 TiO 5 . Lv et al reported hydrothermal deposition of FeOOH on pre-treated FTO substrates, modified with titanium and phosphorous, which resulted in Ti and P modified hematite hollow nanostructures after annealing [158]. They demonstrated that P modified hematite with Fe 2 TiO 5 incorporated yielded the highest photocurrent density of 2.37 mA cm −2 at 1.23 V RHE . It was proposed that Fe 2 TiO 5 improves charge separation with Fe 2 O 3 while FePO 4 enhances the hole transfer on the surface. This was further improved by coupling it with Co-Pi cocatalyst to achieve 2.9 mA cm −2 at 1.23 V RHE . Deng et al demonstrated F and Rh based treatments on Fe 2 O 3 /Fe 2 TiO 5 system to lower the onset potential [159]. The photoanode system was fabricated by dip coating FeOOH nanorods in TiCl 4 solution. The treatment of fluorine and subsequent immersion in NaOH solution improved the semiconductor/electrolyte interface by forming a hydrogen bond network. The hole transfer was enhanced by partial reduction of surface Fe atoms. Coupling the system with Rh based cocatalyst resulted in a photocurrent density of 1.47 mA cm −2 at 1.0 versus RHE. More importantly, this yielded a final onset potential of 0.63 V versus RHE which is comparable to the lowest reported value for hematite. Such work has been reported by Tang et al who discussed the interfacial coupling effect in quarternary hematite composite [160]. A photocurrent of 2.2 mA cm −2 was achieved for the integrated photoanode ITO/Fe 2 O 3 / Fe 2 TiO 5 /FeNiOOH multi-layer nanowires. It was designed with ITO underlayer to improve the electron conductivity to the back contact, Fe 2 TiO 5 coating as energy level modulator and FeNiOOH nanodot catalyst to improve the interfacial hole transfer through regulating surface states energy level. The design presented by them appears to be complete in targeting different limitations of the device but further improvement by even more intense optimization is still needed.
Recently, Shuai et al demonstrated preparation of hematite and ultrathin iron titanate layer via an in situ reaction which yielded, the highest photocurrent density for Fe 2 O 3 /Fe 2 TiO 5 heterojunction systems of around 3.05 mA cm −2 at 1.23 V versus RHE after employing a CoPi catalyst on top [161]. They presented a strategy to avoid going through the route of depositing FeOOH film by using an electro-reduction (ER) technique. This was done to prevent the undesirable microstructural defects created by the complexes and the bridging ligands formed by Fe +3 species in the solution during FeOOH growth. IMPS and Mott-Schottky results corroborate that the electron migration to the back contact was better for such systems than those prepared by FeOOH route. A stable performance was shown for 60 h in the integrated system where the photocurrent density remained at 98.9% of the initial value at the completion of test. A similar Fe 2 O 3 /Fe 2 TiO 5 heterojunction, coupled further with TiO 2 , was also used for the removal of some organic pollutants through photo fenton reaction (PFR) [162]. This triple heterojunction structure demonstrated much higher and stable PFR activity for the degradation of Methyl Orange than the catalysts of Fe 2 O 3 and TiO 2 /Fe 2 O 3 .

Understanding the charge dynamics of Fe 2 O 3 /Fe 2 TiO 5 heterojunction
Since Fe 2 O 3 / Fe 2 TiO 5 heterojunction composite has proven to be a promising photoanode system with good performance and stability, it is necessary to understand its internal charge dynamics. The Fe 2 O 3 /Fe 2 TiO 5 heterojunction has been   [167] investigated using a variety of spectroscopic techniques and it seems that this junction promotes the fast decay of accumulating holes leading to better charge carrier kinetics. A detailed exploration of the charge carrier kinetics in such systems was carried out by Ruoko et al through transient absorption spectroscopy (TAS) on hematite-titania nanocomposite photoanodes, from sub-picosecond to second time-scale [163]. It was suggested that hole transport efficiency can be evaluated by observing fast electron trapping to localized oxidaized states near the conduction band edge of hematite which competes with electron hole pair recombination. Hence, longest time scale (>5 ns) resolved TAS was obtained under different bias through picoseconds-nanoseconds TAS for hematitetitania photoanodes and bare hematite for compariso n. As shown in figure 14(a), absorption feature at 580 nm remains unchanged with changing bias voltages for bare hematite and hence no electron trapping was observed. On the other hand, for the hematite-titania photoanode (shown in figure 14(b)), electron trapping to oxidized electron states is prominent at 580 nm with biases higher than onset voltage. This is attributed to increased charge separation due to the formation of the pseudobrookite-hematite heterojunction. The presence of titania affected the charge separation positively by increasing the number of long lived holes with an average accumulation lifetime of 0.4 ± 0.1 s. In comparison, long-lived holes in bare hematite photoanode were much less adversely affecting the performance. The decay of the long-lived holes is one order of magnitude faster in the composite photoanodes than previously published for doped hematite, indicative of higher catalytic efficiency. This study on different lifetime scales for recombination, electron trapping and hole transport mechanisms in hematite-titania photoanodes (shown in figure 14(c)) reaffirmed the importance of titania junction in improving the hole transport and corroborated that bare  hematite possesses poor charge dynamics that limits its performance as a photoanode. Damián et al studied the interfacial charge transfer in hematite-titania system with titania concentrations ranging from 0% to 20% (as shown in figure 14(b)) [164]. Solid state diffusion and interface reactions lead to an Fe 2 TiO 5 shell with core as Fe 2 O 3 . They observed that an optimum condition was obtained for 10% titania concentration due to an efficient transfer of holes from hematite to pseudobrookite and electrons from pseudobrookite to hematite. At higher concentration of TiO 2 , an unreacted TiO 2 layer exists which prevents efficient hole transfer by increasing recombination in the Fe 2 TiO 5 layer, and hence inhibits water photooxidation. This is referred to as a 'hole mirror' mechanism. The charge transfer mechanism revealed by EIS shows hole trapping at the surface states before transferring into water as discussed in our earlier report [154].

Hematite based heterostructures with other metal oxides
Hematite based heterojunctions have been reported with other iron based metal oxides like spinel ferrites as well [165][166][167][168][169]. A comprehensive list of hematite based heterostructures is given in table 4 which includes those already discussed in    [165]. This film exhibited a photocurrent density of 2.9 mA cm −2 at 1.23 V RHE and is attributed to better charge conductivity as evidenced by EIS. Hou et al reported a photocurrent density of 3.34 mA cm −2 at 1.40 V RHE for a tree like branched structure which comprised of heterojunctions between a Co-doped Fe 2 O 3 nanorod array as core and MgFe 2 O 4 spinel as shell [167]. The nanorods array was deposited by a hydrothermal route on Ti mesh and was followed by coating MgFe 2 O 4 through wet impregnation. FeVO 4 has also been coupled with Fe 2 O 3 as a direct z-scheme structure for photoelectrochemistry [170]. Zhang et al recently investigated FeVO 4 based photoanodes and discussed their intrinsic optoelectronic properties [171]. In addition, Li et al used FeVO 4 /Fe 2 TiO 5 composites for the photocatalytic removal of norfloxacin [172]. Another strategy to form a heterojunction is by doping hematite to alter its properties. Lin et al doped hematite with Mg to invert its naturally n-type nature to p-type and reported a p-n homojunction fabricated by ALD [173]. Importance of Sn +4 infusion was reported by our group where photocurrent density of 2.25 mA cm −2 at 1.23 V RHE was achieved in SnO x coated Fe 2 O 3 nanorods [64]. The formation of complex Fe x Sn 1−x O 4 phase after the thermal treatment was thought to improve the charge separation in this photoanode device.

Hematite integrated with conducting scaffolds
Another efficient approach to enhance the charge separation efficiency of hematite is to incorporate electron conducting scaffolds into the hematite nanostructure (table 5). During illumination, electrons from the photo generated charge carriers can be collected by the conductive scaffolds and readily transported away to the current collector while holes are transferred to the semiconductor-liquid junction to oxidize water.
A good example of a hematite anode with an embedded conducting scaffold was demonstrated by Lin et al ( figure  15) [174] which used high surface area TiSi 2 nanonet as the conducting scaffold for efficient charge collection which also played the additional role of being a structural support to hematite. In such a composite nano structure architecture the distance from any point in the hematite to the electrode/ electrolyte interface can be made shorter than the chargediffusionlength, allowing efficient charge collection. They reported an external quantum efficiency of 46% at 400 nm wavelength without any intentional doping of hematite in a water-splitting environment.
Graphene is another material recognized as a good electron collector and transporter. Use of graphene as a charge collecting scaffold could effectively prevent recombination of photogenerated electron-hole pairs resulting in improved photocatalytic conversion efficiency [175,176]. Yoon et al demonstrated this charge collection enhancement exper imentally in a graphene coupled hematite system [177]. Meng et al also demonstrated the benefit of graphene as a charge collector in an anode where hematite nanoparticles were added onto reduced graphene oxide (rGO) nanosheets [178]. Similarly, carbon nanotubes (CNTs) have also been used as charge separation scaffolds when combined with the hematite photoanode [175]. Kim and coworkers reported hematite photoanode coupled with multi-walled CNTs which exhibited 66% boost in photocurrent relative to plain hematite photoanode [175]. Multi-walled CNT also served as a structural support to hematite nanoparticles.
Finally, more efforts in using a conducting scaffold or heterostructure with hematite coating layer could be pursued to efficiently extract the photogenerated carriers from hematite. An example of successful demonstration by this approach is the WO 3 /BiVO 4 core shell structure which gave a photocurrent density of 6.8 mA cm −2 which is 90% of the theoretical limit of photocurrent of BiVO 4 [179] .

PEC-PV tandem approach with hematite based photoanode.
Single material photoanodes cannot generate the required photovoltage (1.6-1.8 V) to split water even if all the above-mentioned modifications are successfully incorporated. Therefore, a bias voltage is always required. Here, we will discuss a different approach of tandem cell design that could generate enough photo voltage to drive water-splitting reaction without an external bias.
Soon after the discovery of tandem photovoltaic cells by Fujishima and Honda, it was realized that this approach of additive voltages could be exploited for efficient solar water splitting too [8]. The theoretically calculated efficiencies for these types of tandem approaches can be >25% depending on the illumination conditions (i.e solar concentration) and catalyst overpotentials used in the simulations [187,188]. General reviews of tandem approaches for solar to hydrogen conversion have been published recently [20,[188][189][190]. Hence, we will focus specifically on hematite based tandem devices for solar water splitting A tandem cell configuration (PEC-PV) is where the total photovoltage required to split water is generated by optical absorption in two or more series connected photoabsorbers [24,35,191]. Some of the absorbers could be photo active electrodes while a series connected solar cell could also be one. Sivula et al reported Fe 2 O 3 /2-series connected DSSC solar cells and WO 3 /2-series connected DSSC solar cells in tandem configuration for solar assisted overall water splitting (figure 16) [192]. The WO 3 /DSSC tandem device achieved a STH conversion efficiency η STH of up to 3.10% (this is ~50% of the maximum efficiency obtainable with this material: η STH,max = 6.24%, assuming Eg = 2.6 eV), whereas the Fe 2 O 3 /DSSC tandem device gave only η STH = 1.17% (less than 8% of its maximum: η STH,max = 16% with Eg = 2.1 eV) needing significant improvement [192].
The photovoltage outputs (V OC ) of the solar cells are crucial in these tandem devices. The photovoltage generated by traditional solar cells i.e. silicon, CIGS and DSSC is not enough to drive water splitting reaction solely [93,[193][194][195][196].
The relatively large open circuit voltages generated by halide perovskites could make them attractive candidates as the photovoltaic components of solar fuel devices [197][198][199]. Our group fabricated a hematite-perovskite tandem cell which gave STH conversion efficiency of 2.4% (figure 17) [35]. The main limitation in this tandem cell was found to be the high onset potential of the hematite photoanode. However, the total photopotential generated by our tandem system was 1.87 V which exceeded both the thermodynamic and kinetic requirements for water splitting (1.6 V), resulting in successful splitting without the assistance of any electrical bias [35]. Gratzel et al also demonstrated an optically transparent FeNiO x cocatalyst coated hematite and perovskite tandem cell for solar assisted over all water splitting with a STH conversion efficiency of 2% [80]. Recently, we have produced a more efficient SnO x treated hematite-perovskite tandem cell whose STH conversion efficiency reached 3.4% ( figure 18) [24]. The ALD coating of SnO x to hematite shows a dramatic enhancement in the performance of hematite. Sn acts both, as a dopant, as well as a surface passivation layer in hematite giving 3.12 mA cm −2 photocurrent density at 1.23 V versus RHE. A photoanode-photocathode tandem configuration by utilizing a hematite photoanode and a-Si photocathode was demonstrated by Dunwei Wang's group with a STH conversion efficiency of 0.91% ( figure 19) [93]. This device was stable for 10 h although its performance was limited by the low photocurrent and high onset potential. The overall performance of this configuration was very low in comparison to other PV-PEC tandem configurations.

Initiative towards large area hematite based photoelectrochemical cell devices
To demonstrate the commercial viability of hematite based PEC devices, it is imperative to translate the developments to large area devices. Scaling up the size of the electrodes and PEC cell leads to a different set of limitations and challenges. It is evident that with increasing illumination area, the factors like higher resistance in the back contact FTO, limitations in mass transport of ions in the electrolyte and non-uniformity of films, come into play. For small area, hematite nanostructures have reached high performance with photocurrent density as high as 6 mA cm −2 at 1.23 V versus RHE [30]. However, not many have explored the large area devices with bigger illumination area. Grätzel et al demonstrated a large area device (10 cm × 10 cm) with an illumination area of 80 cm [2] which yielded a total photocurrent of 35 mA which translates to a current density of 0.43 mA cm −2 [69]. Cao et al reported large area preparation of porous Ag 3 PO 4 photoanodes with a photocurrent density of around 4.32 mA cm −2 at 1 V versus Ag/AgCl (equivalent to 1.86 V versus RHE for pH 11.3 Na 3 PO 4 electrolyte) with a 5 h successive PEC water splitting experiment [200]. We have been developing large area hematite photoanodes on FTO substrate consisting an illumination area of up to 7 cm [2]. To observe the effect of illumination area on the performance of PEC devices, we varied the illumination area as 0.125 cm 2 , 0.25 cm 2 , 0.5 cm 2 , 1 cm 2 , 2 cm 2 and 7 cm 2 .
As can be observed in figure 20(a), the modified hematite nanorods yielded a photocurrent density of around 1 mA cm −2 at 1.23 V versus RHE and a total current of 7 mA for an illumination area of 7 cm 2 ( figure 20(b)). The smaller areas yielded higher photocurrent densities reaching as high as 1.5 mA cm −2 for an illumination area of 0.125 cm 2 . The photocurrent stability was assessed in the large area illumination (7 cm 2 ) as shown in figure 20(c). It is clear that a large drop, of the order of 10% to 0.9 mA cm −2 , is evident in the photocurrent within a few seconds from the start and the photocurrent density eventually tails off at about 0.78 mA cm −2 in 18 h. With an intermediary light on/off after 18 h, it was observed that the current followed a slightly sharper drop as compared to the start of the measurement. It is quite possible that this transient drop is due to a build-up of evolved O 2 and H 2 gases at the electrodes. A change in the performance of the photoanode was also eventually noted in subsequent PEC measurements (figure 20(d)). A slight drop in the onset behavior at early potential range of Fe 2 O 3 nanorods anode is evident although, it remained intact at 1.23 V versus RHE. These interesting observations are the preliminary indications of remaining challenges in the realization of large area, hematite based photoanodes in PEC water splitting. There are several issues with large area photoanodes those limit the performance of hematite such as; (1) ohmic losses due to poor conductivity, (2) high temperature annealing requirement, (3) non-uniform nanostructures in large area photoanodes, (4) mass transport limitation. These issues have to be addressed to improve the performance of hematite photoanode and maintain the stability at large area. Resistance losses in large area hematite can be addressed by connecting small hematite panels in series and parallel. Mass transport can be addressed by mechanically rotating the electrolyte or having a flow cell.

Summary and outlook
Recent developments in Fe 2 O 3 (hematite) based photoanodes for solar assisted water splitting in PEC devices are reviewed here. With a theoretical solar to hydrogen efficiency of ~15%, good stability in aqueous electrolyte environment and abundance, hematite has a good potential as the main material for photoanodes in solar water splitting. Its performance, however, is limited by poor charge transport properties and sluggish hole transport across the interface to electrolyte.
To compensate for the deficiencies of hematite and to further improve it, various combinations of many different approaches have been investigated by researchers over more than 10 years. These are summarized below.
(a) Nanostructuring to compensate for the poor minority carrier diffusion length which is a morphological manipulation to facilitate efficient hole diffusion to space charge region. (b) Doping hematite with a suitable multi-valent element to increase its inherent bulk conductivity. Tetravalent elements such as Sn and Si seem to be the most effective in improving the bulk properties and are widely used by researchers. Other elements such as Ti, Ge and Ru and some divalent elements have also been examined. (c) Integration with a transparent and efficient OER cocatalyst which is applied on the surface of hematite to improve the hole injection ability across the hematite-electrolyte interface. The cocatalysts IrO x , CoPi and NiFeO x are effective in this respect. (d) Use of a surface passivation layer to reduce charge recombination at the surface. TiO 2 has been found to be very effective in reducing surface charge recombination and hence has been a popular choice for a passivation layer. However, some reports indicate that TiO 2 deposited on the surface of hematite reacts to form Fe 2 TiO 5 which is the true surface passivating material. (e) Use of an electron conducting nanonet embedded in the hematite to efficiently collect electrons and transport them to the back conductor thereby reducing the chances of recombination. TiSi 2 , graphene and CNTs have shown to be effective as the conducting nanonets giving beneficial effects on the anode performance. (f) Fabricating heterostructure with hematite and another material which has appropriate conduction and valance band offsets with respect to water redox levels and the band edges of hematite. Compounds such as TiO 2 , Fe 2 TiO 5 , FeVO 4 , and spinels such as ZnFe 2 O 4 have been combined with hematite and have been shown to improve the photocurrent density. This strategy has not been adopted widely by many workers possibly because of the difficulty in synthesizing such nanocomposite materials, but this could become a very effective technique.
The exact roles of the cocatalyst and surface passivation layers in improving the anode performance is not clear yet. For example, CoPi has been reported to improve catalytic activity by many groups, but some claim that it provides surface passivation also. The inconsistency among the reports regarding the operating mechanisms could be due to different deposition techniques employed by the researchers, which lead to different intrinsic properties of the hematite and/or the CoPi layer.
The pseudobrookite phase of Fe 2 TiO 5 is another candidate material with similar electronic properties to hematite but with better surface transport characteristics so that holes could be extracted more efficiently. It could be a potential alternative to hematite. The basic electrical and optical properties of Fe 2 TiO 5 are not well known to systematically design a photoanode based on this material. Heterojunctions of Fe 2 O 3 -Fe 2 TiO 5 and nanostructures of Fe 2 O 3 core with Fe 2 TiO 5 shell have shown very encouraging preliminary results as photoanodes. Efforts in nanostructuring, doping, and using Fe 2 TiO 5 as the shell for another semiconductor, or as a conducting scaffold, might be worth exploring in future.
The significant research efforts in the hematite based photoanode developments, which shone light on the benefits of modifications such as nanostructuring + doping + surface passivation + co-catalyst has increased the photocurrent densities to ~6 mA cm −2 , which is ~50% of its theoretical limit. Given the recent success in WO 3 -BiVO 4 core-shell heterostructured photoanodes which gave about 90% of theoretical photocurrent density, we remain optimistic that with the right combination of strategies, hematite photoanodes will also be able to realize its theoretical photocurrent density in future.
Finally, the laboratory understandings and experiences must be translated into pilot scale PEC devices to demonstrate the commercial viability of the technology. Hematite photoanodes of active areas about 7 cm 2 have been fabricated and tested which show that the photocurrent density appears to decrease with increasing illumination area. The reasons for this must be understood and the phenomenon must be circumvented for successful commercial deployment of this technology.