lable at ScienceDirect

Journal of Power Sources 287 (2015) 119e128

Contents lists avai

Journal of Power Sources

journal homepage: www.elsevier .com/locate/ jpowsour

Efficient photoelectrochemical water splitting using threedimensional urchin-like hematite nanostructure modified withreduced graphene oxide

Andebet Gedamu Tamirat a, Wei-Nien Su b, Amare Aregahegn Dubale a, Chun-Jern Pan a,Hung-Ming Chen a, Delele Worku Ayele a, Jyh-Fu Lee c, Bing-Joe Hwang a, c, *

a NanoElectrochemistry Laboratory, Department of Chemical Engineering, National Taiwan University of Science and Technology, Taipei, 106, Taiwanb NanoElectrochemistry Laboratory, Graduate Institute of Applied Science and Technology, National Taiwan University of Science and Technology, Taipei,106, Taiwanc National Synchrotron Radiation Research Center, Hsinchu, 30076, Taiwan

h i g h l i g h t s

* Corresponding author. NanoElectrochemistryChemical Engineering, National Taiwan University ofpei, 106, Taiwan.

E-mail address: bjh@mail.ntust.edu.tw (B.-J. Hwan

http://dx.doi.org/10.1016/j.jpowsour.2015.04.0420378-7753/© 2015 Elsevier B.V. All rights reserved.

g r a p h i c a l a b s t r a c t

� rGO modified 3D a-Fe2O3 is synthe-sized using hydrothermal and spincoating methods.

� The rGO sheet act as both conductivescaffold and surface passivation layer.

� a-Fe2O3 modified with optimumamount of rGO show superior pho-tocatalytic activity.

� rGO modified a-Fe2O3 exhibits highercharge separation and charge injec-tion efficiency.

a r t i c l e i n f o

Article history:Received 20 December 2014Received in revised form29 March 2015Accepted 7 April 2015Available online 8 April 2015

Keywords:HematiteReduced graphene oxideCharge separation efficiencyCharge injection efficiency

a b s t r a c t

Herein, we present a highly photoactive photoanode for solar water oxidation using three dimensional(3D) urchin-like hematite (a-Fe2O3) nanostructures modified with ultra-thin reduced graphene oxide(rGO). rGO acts as both electron conducting scaffold and surface passivation layer. By virtue of thesecombined effects, the composite photoanode exhibits 1.47 times higher photocurrent density(1.06 mA cm�2, at 1.23 V vs. reversible hydrogen electrode (RHE)) and two-fold enhancement in thephotoconversion efficiency than that of pristine a-Fe2O3. The dual effect of rGO as both electron con-ducting scaffold and surface passivation layer is further evidenced from the 1.82 and 1.67 fold en-hancements in charge separation and charge injection efficiencies at 1.23 and 1 V vs. RHE respectively. Toget further evidence about the origin of the improved photoactivity of the rGO modified photoanode, aseries of electrochemical, photoelectrochemical and impedance spectroscopy measurements were car-ried out. Our results demonstrate the benefits of a noble metal free highly promising photoanode forphotoelectrochemical water oxidation.

© 2015 Elsevier B.V. All rights reserved.

Laboratory, Department ofScience and Technology, Tai-

g).

1. Introduction

The efficient conversion of solar energy to hydrogen throughphotoelectrochemical (PEC) water splitting, using semiconductor

A.G. Tamirat et al. / Journal of Power Sources 287 (2015) 119e128120

electrodes, is a long-standing challenge that offers the hope ofconverting solar energy into synthetic fuels [1]. The selection ofsemiconductor electrodes for PEC water splitting must satisfy anumber of requirements. Of paramount importance is the abilityto perform reactions at high efficiencies. Among various watersplitting semiconductor electrodes, hematite (a-Fe2O3) (“a-” isomitted henceforth) is one of few materials that favorably com-bines several promising properties such as; stability in aqueoussolutions, visible-light absorption, non-toxicity, abundance andlow cost [2]. With an energy band gap of 2.1 eV, hematite cantheoretically reach water oxidation current density as high as12.6 mA cm�2 under air mass 1.5 global (AM 1.5G) solar irradia-tion; thereby potentially enabling a maximum solar-to-hydrogenconversion efficiency of 15.5% [3]. However, the PEC activity ofhematite is limited by several factors such as relatively poor ab-sorption coefficient, very short excited-state lifetime (~10�12 s),poor oxygen evolution reaction kinetics, and a short hole diffusionlength (2e4 nm) [4].

To alleviate these limitations and improve the solar conversionefficiency, enormous efforts have been made. Some of these stra-tegies include tailoring of hematite nanostructures, e.g. nanorodarrays, nanowires, nanocorals, cauliflower structures, worm-likestructures etc [4e8]. Nanostructures with a very large interfacialarea between the electrolyte and the semiconductor would facili-tate charge collection by reducing the diffusion length of photo-excited holes. Doping is another strategy to improve the conduc-tivity of hematite, where the enhanced electrical conductivity canextend the lifetime of the charge carriers by reducing the recom-bination of photogenerated electronehole pairs. So far hematitehas been doped with Ge, Mn, Si, Sn, Ti, Zr as n-type [4,6,9,10] andCo, Cu, Mg as p-type dopants [11e13]. N-type doping can boost theconductivity of hematite by introducing additional majority car-riers [14]. On the other hand p-type doping enhances the utilizationof solar energy by altering band-edge energies [12]. To minimizethe large applied potential required, various catalysts have beenadded to the hematite surface including IrO2, Co�Pi, NieBi,Ni(OH)2, NiOOH etc [15e19]. Surface modification by catalysts im-proves the PEC oxygen evolution kinetics, either by loweringpotential-dependent rate constants for surface-mediated chargerecombination, or by increasing the rate constant for hole transferfrom the photo-electrode to the molecular reactant [20]. Anotherefficient approach to enhance the PEC performance of hematite isto incorporate electron conducting scaffolds in to the hematitenanostructure so as to improve the charge separation efficiency.Among the most known conducting scaffolds, graphene is widelyrecognized as an excellent electron collector and transporter thatcan efficiently hinder the recombination of photogenerated elec-tronehole pairs [21,22]. For thermodynamically feasible electrontransport from a semiconductor to graphene, the conduction bandof the semiconductor should be smaller than the work function ofgraphene. Typically, the work function of graphene is ~ 4.5 eV [23]and the conduction band edge of hematite is ~ �4 eV vs. vacuum atpH 14 [24]. Such an energy level configuration is beneficial fortransporting photogenerated electrons from the hematite surfaceto the graphene sheet. On this premise, several hematite-graphenecomposite photoanodes have been demonstrated recently for PECwater splitting. For example, Meng et al. reported that incorpo-rating hematite nanoparticles to reduced graphene oxide (rGO)nanosheet can suppress charge recombination and enhancescharge separation [25]. Similarly, He et al. reported N-doped gra-phene�Fe2O3 nanocomposite for photoelectrochemical wateroxidation and for photocatalytic colorless pollutant degradation[26]. However, many of these reports are suffering from too low PECperformance. Moreover, no direct experimental evidence has yetbeen provided to prove the charge separation and charge injection

properties of graphene composite electrodes. Herein, we combineda 3D urchin-like nanostructured Fe2O3 with ultra-thin rGO sheetusing a facile synthetic approach to use as an efficient watersplitting photoanode. At optimum amount, the ultra-thin rGO sheetacts as a conducting scaffold to capture photogenerated electronsfrom the host Fe2O3 photocatalyst and thereby reduce electro-nehole recombination and as a surface passivation layer to improvethe sluggish water oxidation reaction. To investigate the dual effectof rGO on the PEC performance of Fe2O3, we carried out PECmeasurements in the presence of 0.5 M hydrogen peroxide (H2O2)hole scavenger, which gives direct evidence on the charge separa-tion and injection properties.

2. Experimental section

2.1. Synthesis of 3D urchin-like a-Fe2O3 photoanode

The 3D urchin-like hematite nanostructures were prepared via asimple solution-based method followed by two-step in situannealing as described in literature with minor modification [27].First, an aqueous solution containing 1 mmol of FeCl3$6H2O and1 mmol of Na2SO4 was prepared. In the meantime, a piece offluorine doped tin oxide (FTO) coated glass was cleaned ultrason-ically with EXTRAN MA 02 (neutral liquid detergent) and washedwith de-ionized (DI) water and isopropyl alcohol followed byexcess DI water. Then the clean FTO coated glass was put intoTeflon-lined stainless steel autoclave. 20 mL precursor solutionwasadded to the autoclave, sealed and heated at 120 �C. After 6 h re-action time, the autoclave was cooled down naturally. Then theFeOOH coated substrate was taken out of the Teflon, washed withabsolute ethanol and DI water separately then air dried. A uniformyellow FeOOH film was coated on the glass substrate. The FeOOHfilmwas changed to Fe2O3 via two-step in situ annealing, initially at500 �C for 3 h then at 800 �C for 20 min. The first annealing stage isresponsible for converting the FeOOH phase to hematite phasewhile the later helps to enhance the crystallinity and diffusion of tinfrom the FTO to hematite bulk resulting tin doped hematite(SneFe2O3).

2.2. Synthesis of graphite oxide (GO)

Graphite oxide was prepared from natural graphite precursor(purity >99.8%, Alpha Aesar) according to a modified Hummer'smethod [21]. In brief, a 9:1 mixture of concentrated H2SO4/H3PO4(360:40 mL) was added to a mixture of graphite flakes (3.0 g, 1 wt.equiv.) and KMnO4 (18.0 g, 6 wt. equiv.), producing a slightexothermic reaction (35e40 �C). The reaction mixture was thenheated to 50 �C and stirred for 12 h and then cooled to roomtemperature and poured in to ice water (~400 mL) containing 30%H2O2 (3 mL). The dispersionwas centrifuged (4000 rpm for 30 min)and the supernatant was decanted. The remaining solid materialwas washed in succession with 200 mL of DI water, 200 mL of 30%HCl, and 200 mL of ethanol (each two times). After successivewashing, themixturewas centrifuged at 4000 rpm for 1 h and driedovernight at 60 �C under vacuum.

2.3. Synthesis of reduced graphene oxide (rGO) modified 3D urchin-like Fe2O3

First, dried GO was exfoliated in a mixture of DI water andethanol (1:1 volume ratio) using ultrasonic treatment for 30 min toform a colloidal suspension (2 mg mL�1). Then 50 mL of GO sus-pensionwas deposited into 1 cm2 Fe2O3 film and allowed towet thesurface for 1 min. Subsequently, the substrate was allowed to spinat 800 rpm for 30 s to create uniform spreading of the solution on

A.G. Tamirat et al. / Journal of Power Sources 287 (2015) 119e128 121

the substrate, followed by another 30 s spinning at 3000 rpm to drythe film. The Fe2O3-GO filmwas dried inside an oven at 80 �C for 1 hbefore any reducing treatments. The film was thermally heated at350 �C for 2 h under continuous flow of N2 and allowed to cool toroom temperature. The sample is represented as Fe2O3-rGO1(Fe2O3 modified with reduce graphene oxide). Annealing helps toreduce GO to rGO and assist to improve adhesion between the rGOsheet and the Fe2O3 nanostructures. To determine the optimalamount of rGO, two, three, and four spin-coating cycles were alsoperformed followed by appropriate thermal treatment. The corre-sponding samples are represented as Fe2O3-rGO2, Fe2O3-rGO3, andFe2O3-rGO4 respectively.

2.4. Photoelectrochemical measurements

PEC measurements were performed using a three-electrodeconfiguration in 1 M NaOH (pH 13.6) aqueous solution, withFe2O3 and Fe2O3-rGO photoanodes as the working electrode, Ag/AgCl as the reference electrode, and Pt wire as the counter elec-trode. All photoanodes were scanned from 0.4 to 1.6 V vs. RHEduring the JeV analysis. The potential is reported relative to thereversible hydrogen electrode potential (RHE). In all PEC mea-surements the potential in Ag/AgCl was changed to the reversiblehydrogen electrode (RHE) scale according to the Nernst equation:

ERHE ¼ EAg=AgCl þ 0:059pHþ EoAg=AgCl (1)

where, ERHE is potential in RHE, EoAg/AgCl ¼ 0.1976 V at 25 �C, andEAg/AgCl is the experimentally measured potential against Ag/AgClreference. A 300 W xenon arc lamp (6258, Newport) coupled withan AM 1.5 G filter was used as the light source. Its intensity wascalibrated at 100 mW cm�2 (1 sun illumination) using a silicondiode (Newport). N2 gas was continuously bubbled through thesolution before and during the experiment to remove any dissolvedO2 and therefore suppress the reduction of O2 at the counterelectrode.

2.5. Structural characterization (XRD, SEM, TEM, UVevisible, XAS,Raman)

The crystalline phase of the samples was identified using a D2phaser XRD-300W diffractometer equipped with a Cu Ka(l ¼ 1.5406 Å) radiation source operating at 30 kV and 10 mA. Themorphology of the samples was determined by field-emissionscanning electron microscopy (JSM 6500F, JEOL) coupled withenergy-dispersive X-ray analysis (EDX) with an acceleratingvoltage of 15 kV. The TEM images were collected on a Philips/FEITecnai 20G2 S-Twin apparatus. For TEM analysis, the samples werescratched from an FTO substrate and dispersed ultrasonically inethanol and dropped to a carbon coated copper grid. Then, thesolvent was evaporated inside an oven at 80 �C for 6 h. Opticalabsorption measurements were performed using JASCO 560UVevisible spectrometer (UVevisible absorption spectrometer(Shimadzu, Model UV 3600)). X-Ray absorption spectra (XAS)were recorded using beam line 17C, at the National SynchrotronRadiation Research Center (NSRRC) of Taiwan. The electron storagering was operated at 1.5 GeV with a current of 360 mA. A Si (111)double-crystal monochromator was employed for energy selectionwith a resolution DE/E better than 2 � 10�4 at the Fe K-edge.Raman spectra were obtained from ProMaker confocal Ramanmicroscope system. A solid state laser operating at 532 nm wasused as the excitation source with a laser power of 20 mW tocircumvent degradation with 10 s exposure time and 15accumulations.

3. Results and discussion

3.1. Physical characterization of photoanodes

The morphology evolution of FeOOH nanostructure was exam-ined using FE-SEM analysis. Fig.1 shows the SEM images of samplesgrown at different reaction times including; 3, 3.5, 4 and 4.5 h.Initially, smaller FeOOH nanoparticles were formed through thehydrolysis of Fe3þ, which serves as a nucleus in the growth process.It can be clearly observed that smaller feature size FeOOH nanorodshave been grown after 3 h reaction time. The nanorod structuresbegin to agglomerate each other through oriented attachmentfollowing 3.5 h reaction time. After 3.5 h reaction time, the FeOOHnanorods agglomerated more into 3-D urchin-like structuresthrough oriented attachment, so as to reduce the surface energy[28]. It has been suggested from previous study that the aggrega-tion process involves the formation of larger nanoparticles byreducing the interfacial energy of the smaller particles [27]. At 4.5 hreaction time clear independent 3D urchin-like structures havebeen formed. To obtain denser nanorod attachments we continuedthe reaction time up to 6 h. The addition of sodium sulfate plays avital role to the formation and self-assembly of FeOOH nanorodsinto 3D urchin like structures. It is well known that surfactantsoften restrict crystal growth in certain directions through prefer-ential absorption, thereby promoting increased relative growth onother crystal faces. In this particular case, the sulfate ion serves as aligand to Fe3þ and is adsorbed on the facets parallel to the c-axis ofthe FeOOH nuclei by a monodentate structure (FeeOeSO3), to giveFeOOH nanorods [27]. The overall fabrication process of 3D urchin-like Fe2O3 nanostructure on an FTO coated glass substrate is illus-trated Scheme 1.

Fig. 2(a) illustrates the SEM image of the overall morphology ofthe FeOOH sample synthesized for 6 h reaction time. The magnifiedSEM image of FeOOH (Fig. 2(b)) clearly shows that individualnanorods assembled each other into 3D urchin-like structures withrod-like crystallites radiating from the center. Fig. 3(a) shows thatthese unique 3D urchin-like structures have average radius of1.4 mm. The urchin-like structures have a large surface area whichenhances the electrode-electrolyte interface which is favorable forwater splitting. The FeOOH sample has been fully transformed intohematite (as confirmed form the XRD) via two step in situ annealingat 500 and 800 �C. The initial 3D urchin-like morphology of FeOOHsample has been persisted after successive annealing stages(Fig. 2(c)). Fig. 2(d) presents Fe2O3 structures modified with ultra-thin sheets of rGO. The TEM image in Fig. 2(e) clearly shows theformation of intimate contact between the rGO sheet and the he-matite surface. The unique 3D urchin-like architecture providesspace to let the GO suspension be incorporated and made intimatecontact with the Fe2O3. Fig. 2(f) shows HR-TEM image taken at theinterface between Fe2O3 and rGO contact area. An ultrathin contactlayer (ca. 1.5 nm thickness) is formed between the interfaces whichis beneficiary to passivate surface states of hematitenanostructures.

The EDS analysis of rGOmodified Fe2O3 nanostructures revealedthe presence of 7.78% carbon, 67.33% oxygen, 19.62% iron and 5.27%tin, thereby confirming the occurrence of residual carbon species(Fig. S1). Significant amount of Sn was observed in both Fe2O3 andrGO modified Fe2O3 EDS spectra; owing to the diffusion of Sn fromthe FTO coated glass substrate to the Fe2O3 bulk during high tem-perature annealing. Ling et al. have reported that 9.9% of Sn wasdiffused into the hematite film after annealing at 800 �C [4]. Sivulaand co-workers have also confirmed unintentional diffusion of Snfrom FTO coated glass substrate to hematite bulk [29]. Theyobserved that the incorporation of Sn caused a 2-fold enhancementin the optical absorption coefficient as a result of the structural

Fig. 1. SEM images of the morphology evolution of 3D urchin like FeOOH nanostructures.

Scheme 1. Schematic illustration of the overall synthetic route of 3D urchin-like Fe2O3-rGO hybrid photoanode.

A.G. Tamirat et al. / Journal of Power Sources 287 (2015) 119e128122

distortion of the hematite lattice, which played a paramount role onenhancing the conductivity of thematerial and thereby contributedto enhance the charge separation efficiency.

The cross-sectional atomic distribution of rGO modified Fe2O3was further estimated by EDS line scanning (Fig. 3). The resultsrevealed that both Fe and O have the expected near-uniform dis-tribution along the cross-section of the film. However, the amountof Sn gradually increases from the top to bottom surfaces of thefilm. This confirms that Sn diffuses from the FTO conductive sub-strate to the bulk material during high temperature calcination. Inour previous work we have shown that the gradual re-distributionof tin can create gradient doping in the Fe2O3 bulk which helps toimprove the carrier separation [17]. In contrast, the amount ofcarbon that originates from rGO, gradually decreases on going fromthe top to the bottom of the electrode.

X-ray diffraction (XRD), Raman spectra and XAS measurementswere carried out to examine the phase purity and crystalline natureof each sample. The as-prepared filmwas confirmed to be a-FeOOH(JCPDS card number 29e0713) (Fig. S2(a)). The XRD pattern ob-tained after calcination of FeOOH at 500 �C revealed the successfulgrowth of pure Fe2O3 (JCPDS card number 33e0664). The absence

of FeOOH diffraction peaks indicates the complete conversion ofFeOOH to Fe2O3. In the XRD patterns of the Fe2O3-GO and Fe2O3-rGO samples no typical diffraction peaks of GO and rGO areobserved, due to low loading contents and restricted crystallization(Fig. 4(a)). Moreover no peak shift, broadening, additional peaksappeared in the XRD patterns of rGO modified Fe2O3. This impliesthat the phase purity of Fe2O3 was maintained after Fe2O3 has beenmodified with rGO, which helps to retain its photocatalytic activity.The synthesis of GO from natural graphite flakes and its completetransformation to rGO is confirmed in the XRD patterns inFig. S2(b). A characteristics intense peak at 26.2� was observed fornatural graphite powder corresponding to (002) plane [30]. Theemergence of a new peak at 10.18� with complete disappearance ofthe graphitic peak indicates the successful formation of GO. Theinterlayer distance increases from 3.40 to 8.67 Å up on conversionof natural graphite to GO suggesting the formation of oxygencontaining functional groups and incorporation of water duringchemical oxidation [22]. After 2 h calcination, the peak at 10.18�

vanishes completely and a broad peak at 24.19� appears, indicatingthe transformation of GO to rGO. After reduction, the interlayerdistance decreases to 3.67 Å slightly close to the graphite values,

Fig. 2. Structural characterization. (a) and (b) SEM image of FeOOH after 6 h reaction time, (c) SEM image of Fe2O3 annealed at 800 �C (d) SEM images of Fe2O3-rGO (e) TEM image ofFe2O3-rGO (f) HRTEM image of Fe2O3-rGO.

A.G. Tamirat et al. / Journal of Power Sources 287 (2015) 119e128 123

due to the removal of partial oxygen moieties and subsequentrestoration of the sp2 carbon network.

To get evidence on the crystal structures of Fe2O3, Fe2O3-GO andFe2O3-rGO samples, we further carried out Raman spectra mea-surements. As shown in Fig. S3(a), our samples exhibited all peakscorresponding to the typical frequencies observed for Fe2O3 whichis composed of seven phonon lines, namely two A1g modes (226and 496 cm�1) and five Egmodes (243, 291, 299, 409, and 610 cm�1)[31]. The closely overlapping peaks at 226, 243, 291 and 299 cm�1

are seen as independent peaks through close observation inFig. S3(b). The more intense peak of Fe2O3 at 1315 cm�1 corre-sponds to a 2-magnon scattering band arising from the interactionof two magnons created on anti-parallel closed spins [32]. Inaddition to the typical vibration peaks of Fe2O3, the Raman spectraof Fe2O3-GO and Fe2O3-rGO samples display two prominent peaks:at ~1328 cm�1 and ~1598 cm�1, which correspond to the well-documented first order D (the A1g symmetry mode) and G (theE2g mode of the sp2 carbon atoms) bands respectively [33]. The Dband is related to the presence of sp3 defects, while the G band isassociatedwith the in-plane vibration of sp2 carbon atoms [34]. Theintensity ratio for D/G bands of Fe2O3-rGO hybrid (1.014) is highercompared to that of Fe2O3-GO (0.98), indicating the presence ofmore localized sp3 defects within the sp2 carbon network afterthermal reduction, which enhances the conductivity of rGO sheetand thus facilitates the charge transfer.

The formation of hematite from the akagan�eite phase and theeffect of thermal treatment were also investigated by X-ray ab-sorption near edge structure (XANES). To gain a better under-standing, the entire normalized Fe K edge XANES spectra (Fig. 4(b))was treated as two regions, i.e. regions at low (R1) and high (R2)energies. The first XANES spectral region (R1) is the pre-edge (lowenergy) which presents a similar profile and intensity for all sam-ples studied. The edge energy position (7114 eV) of the first pre-edge peak corresponds to all samples having iron ions in the Fe3þ

oxidation state [35]. The second region (R2), considered as postedge (high energy), is sensitive to the local environment around theFe absorbing atom. The shape and the peak position can be used toillustrate the effect of the thermal treatment. A shoulder peak,before a maximum absorption peak, appeared for samples calcinedat 500 and 800 �C. On the other hand, the prepared sample (FeOOH)shows only one peak with maximum energies between the

shoulder peak and maximum absorption peak of the Fe2O3 sam-ples. Furthermore, there was no defined shoulder in the spectrumfor the FeOOH sample in comparison with the Fe2O3 reference.Usually, this single peak observed in the XANES spectra at ~7132 eVis attributed to a dipole allowed 1s to 4p electron transition, indi-cating the presence of Fe(III) [36] in the structure of the akagane

ite(FeOOH) phase. Similarly, it was found that the intensity of the 1s to4p transitionwas proportional to the population of Fe(III) in FeOOH,in which the oxygenwas the major atom coordinated to the centralFe atom [37]. Based on the XANES spectral analysis, we confirmedthat Fe2O3 samples calcined at both 500 and 800 �C werecompletely transformed into pure hematite. The Fourier transformof the EXAFS data in R space is shown in Fig. 4(c). The peaks at~1.5 Å and ~2.7 Å are attributed to FeeO and FeeFe bondsrespectively [38]. An obvious increase in the intensity of the FeeFebond is observed for Fe2O3 samples calcined at 500 �C and there is amore intense peak for Fe2O3 calcined at 800 �C, implying an in-crease of FeeFe bonding. This is associated with the decrease in theDebye-Waller factor, an indicator of structural disorder, suggestingenhanced crystallinity. Both the short-range sensitive XAS andlong-range sensitive XRD characterizations confirm the completetransformation of the FeOOH phase to Fe2O3 with high purity andenhanced crystallinity upon successive calcinations at 500 and800 �C.

In order to study the optical properties of the samples, UVevi-sible absorption spectra measurements were made. Fe2O3 samplesobtained after calcination at 800 �C showed a slight red shift andhigher absorption intensity than those calcined at 500 �C(Fig. S4(a)). An additional slight red shift is also observed afterFe2O3 has been modified with rGO (Fig. 4(d)). This leads to a slightnarrowing of the band gap which favors an increase in light ab-sorption intensity. The slight narrowing of the band gap could beattributed to the formation of the FeeOeC bond between Fe2O3 andRGO. A similar phenomenon has been reported in other rGO-basedsemiconductor materials [33,39,40].

3.2. Photoelectrochemical performance measurements

First, we examined the PEC performance of Fe2O3 calcined at500 �C that yields a negligible photocurrent density of 15 mA cm�2

at 1.23 (V) vs. RHE (Fig. S5(a)). We have found that the photocurrent

Fig. 3. (a) SEM image of a single urchin-like Fe2O3 microstructure, (b) Cross-sectional SEM image, and (c) EDS line scans of rGO modified Fe2O3 with the cross-sectional distributionof iron, oxygen, tin and carbon.

A.G. Tamirat et al. / Journal of Power Sources 287 (2015) 119e128124

density gradually increases with increasing calcination tempera-ture, reaching 0.72 mA cm�2 for Fe2O3 calcined at 800 �C. Thisremarkable enhancement stem from enhanced crystallinity (evi-denced from XRD pattern) and unintentional Sn doping whichdiffuses from the FTO layer to Fe2O3 bulk at elevated temperature.Sn, as n-type dopant, introduces majority carriers to Fe2O3 crystalsystem and thereby enhances the electrical conductivity [4,17].From the EDS line scan results we confirmed that tin diffuses fromthe substrate to the bulk material non-uniformly. Relatively, highconcentration of tin lies in the near-substrate region of the Fe2O3film and gradually decreases towards the top surface of the film.This non-uniform distribution of tin might create a typical gradientdoping in the Fe2O3 bulk. This could create multi-step bandbending over the entire Fe2O3 bulk, that improves the carrier sep-aration and thereby enhance the photocurrent.

Among the rGO modified Fe2O3 samples, the Fe2O3-rGO2 sam-ple (two spin-coating cycles) results the highest photocurrentdensity of 1.06 mA cm�2 at 1.23 V which is 1.47 times higher thanthe photocurrent density of pristine Fe2O3 (Fig. 5(a)). So far, inde-pendent studies by He et al. and Kim et al. on hematite based rGOcomposite photoanodes revealed amaximum photocurrent density

of ~0.15 and 0.3 mA cm�2 at 1.23 V vs. RHE respectively [26,41].When compared to these values, we have found much higherphotocurrent under similar operation conditions. The enhancedphotoelectrochemical activity is attributed to the effective electroncollection and transportation characteristics of rGO that can effi-ciently hinder the recombination of photogenerated electroneholepairs. This in turn is due to the unique 3D urchin-like architecture ofFe2O3 that provides space for the appropriate dispersion of rGO thinsheet. This helps direct interaction between Fe2O3 and rGO sheetwhich provides a quick path way for photogenerated electron to-wards the charge collector via rGO. Increasing the number of spincoating cycles further to three and four drastically decreases thephotocurrent density to 0.2 and 0.1 mA cm�2 respectively. Thedecrease in the photocurrent is associated with the agglomeration,or restacking of rGO sheets which results low electrochemicalactive surface area of Fe2O3. Another imperative feature of the J-Vresponse is a ~60mV cathodic shift in the onset potential of the rGOmodified Fe2O3 photoanode (Fig. 5(b)). The negative shift of theonset potential indicates a smaller kinetic energy barrier for chargetransfer across the Fe2O3/electrolyte interface. This is due to anultrathin passivation layer formed between hematite and rGO

Fig. 4. (a) X-ray diffraction patterns of Fe2O3, Fe2O3-GO, and Fe2O3-rGO. (b) and (c) XANES and EXAFS spectra of FeOOH and Fe2O3 samples heat treated at 500 and 800 �C. (d)UVevisible absorption spectra of pristine Fe2O3 calcined at 800 �C and rGO modified Fe2O3.

A.G. Tamirat et al. / Journal of Power Sources 287 (2015) 119e128 125

interfaces that promotes facile hole transfer across the interface. Asimilar report by Nielander et al. have demonstrated single layer ofgraphene as both a protection and passivation layer of n type siliconthrough the formation oxide layer under illumination [42]. Simi-larly, previous studies by Meng et al., and Rai et al., reported

Fig. 5. (a), (b) and (c) Variation of photocurrent density vs. applied potential (JeV) (positivefunction of applied potential). (d) Photocurrent density vs. time plot measured at applied p

cathodic shift in onset potential when Fe2O3 photoanode wasmodified with rGO [25,43]. It has been claimed that this cathodicphotocurrent onset potential shift is due to the smaller kineticenergy barrier for charge transfer across the interface of rGO andFe2O3. The PEC water splitting efficiency of the samples was

scan, scan rate, 10 mV s�1); (The inset in (b) represents photoconversion efficiency as aotential of 1.23 V vs. RHE under illumination.

A.G. Tamirat et al. / Journal of Power Sources 287 (2015) 119e128126

quantitatively evaluated using the calculated applied bias photon-to-current efficiency (ABPE) based on the equation:

ABPEð%Þ ¼"���Jph����mA

�cm2�ð1:23� jVbjÞðVÞ

Ptotal�mW

�cm2

�#AM 1:5G

� 100

(2)

where, Vb is the applied bias vs. RHE, Jph is the photocurrent densityat the measured potential, and Ptotal is the power density of theincident light. Fe2O3-rGO achieves a maximum photoconversionefficiency of 0.102% at 1.06 V, which is 2.04 times higher than that ofthe pure Fe2O3 which has a value of 0.05% at 1.09 (Fig. 5(b), inset).The most interesting feature of this result is that the rGO modifiedFe2O3 sample possesses enhanced photoconversion capability at30 mV lower potential compared to pure Fe2O3. This clearly in-dicates the synergetic effect of ultra-thin rGO sheet both as chargeseparation and surface passivation layers. It should be noted that,for a photocatalyst to be commercially viable, stability and a highphotocatalyst activity are both indispensable. It is well known thatFe2O3 is stable in neutral and alkaline electrolytes under PECoperating conditions. The stability of Fe2O3 and Fe2O3-rGO photo-anodes was demonstrated for continuous (2000 s) irradiation. Bothelectrodes showed excellent stability without the addition ofsacrificial reagent (Fig. 5(d)). No noticeable decrease in activity wasfound in the extended time, indicating their excellent stability.

3.3. Charge transfer dynamics measurements

To understand the effect of rGO on hindering the recombinationof photogenerated electronehole pairs, we conducted chargetransport and injection efficiency measurements for both pristineFe2O3 and rGO modified Fe2O3 samples using 0.5 M H2O2 as a holescavenger added to 1 M NaOH electrolyte solution [17,44]. The keyassumption underpinning this approach is that the oxidation ki-netics of H2O2 is very fast and the charge transfer (injection) effi-ciency is 100%, or the surface charge recombination is eliminated.The water splitting photocurrent (JH2O

ph ) is a product of the rate ofphoton absorption expressed as a current density (Jabs); the chargeseparation yield of the photogenerated carriers (hsep); and thecharge injection efficiency to the electrolyte (hinj):

JH2Oph ¼ Jabs � hsep � hinj (3)

In the presence of H2O2, hinj ¼ 1; as the surface recombination isassumed to be completely suppressed. Hence, the photocurrentdensity in the presence of H2O2 (JH2O2

ph ) is given by:

JH2O2ph ¼ Jabs � hsep (4)

Therefore,

hsep ¼ JH2O2ph

.Jabs (5)

Combining equations (3) and (5):

hinj ¼ JH2Oph

.JH2O2ph (6)

Jabs (photocurrent assuming 100% absorbed-photon-to-currentconversion efficiency (APCE)) is determined by integrating theUVeVis absorption spectra (Fig. 4) of the samples with respect tothe AM 1.5G solar light spectrum [44]. Using this calculation the Jabsvalues for pristine Fe2O3 and rGO modified Fe2O3 were found to be7.25 and 7.62. The two samples have only 4.8% difference in Jabs andthus can be used for reliable estimation of the charge separation

and injection efficiency. Fig. 6 (a) and (b) show the J-V response ofFe2O3 and rGO modified Fe2O3 samples with and without H2O2. Inboth cases the saturated photocurrents in NaOH and NaOH þ H2O2electrolyte systems are nearly the same, since the addition of H2O2do not change the light absorption, pH or flat band potentials of thephotoanodes. This indicates that no current doubling occurs in thepresence of H2O2, while Jabs and hsep remain the same for JH2O2

ph andJH2Oph . Fig. 6 (c) and (d) illustrate the charge separation and injectionefficiencies respectively. The rGO modified Fe2O3 photoanodeshows 1.82 times higher charge separation efficiency at 1.23 V vs.RHE than pristine Fe2O3 (Fig. 6 (c)). This is attributed to the transferof photogenerated electrons to the rGO scaffold which suppressesthe charge recombination effectively. Hence, the recombination ofphoto-excited electronehole pairs is expected to be reduced. Tofurther examine the electron transport and sheet resistance char-acteristics of the electrode materials we conducted electrical con-ductivity measurements using a four point probe (KeithLinkTechnology Co, Ltd, Taiwan) technique. As depicted in Table S1 allthe rGO modified Fe2O3 samples showed improved electrical con-ductivity compared to pristine Fe2O3, suggesting that incorporatingrGO to Fe2O3 significantly improves the electron transport property.This result is much consistence with the hsep results. Fig. 6 (d)shows charge injection efficiency (hinj) as a function of theapplied potential. rGO modified Fe2O3 exhibits 1.67 times highercharge injection efficiency at 1 V vs. RHE which is correlated withthe cathodic shift of onset potential observed in J-V curves(Fig. 5(b)). The enhanced charge injection efficiency is further evi-dence that shows the act of ultra-thin rGO as a passivation layerwhich inhibits surface recombination. However, we have foundthat at higher potential region, the hinj of rGO modified Fe2O3 isslightly lower than that of pristine Fe2O3. This might be due to thepotential dependent nature of the region. In other words, in thehigher potential region the effect of applied potential outweighsthe effect of surface modification. Moreover, from the hsep and hinjresults we noted that the effect of rGO as electron shuttling scaffoldsurpasses its effect as surface modification. To study the interfacialcharge transfer process between the electrodes and the electrolyte;electrochemical impedance spectroscopy (EIS) experiments werefurther carried out. The EIS spectra of Fe2O3 and Fe2O3-rGO underdark and simulated solar light illumination were measured at apotential of 1.23 V vs. RHE and presented in Nyquist diagram inFig. S6. Under dark conditions, no charge transfer between theelectrode and the electrolyte was detected in both pristine Fe2O3and rGO modified Fe2O3 samples. Under irradiation, however, aNyquist plot has a semicircle shape, indicating that electronmovement in the interfacial region is strongly controlled by acharge-transfer process. It is also shown that the radius of the rGOmodified Fe2O3 sample was smaller than Fe2O3 indicating a moreeffective interfacial charge transfer across the semi-conductoreelectrolyte junction.

Finally, understanding the effect of rGO on the band edge po-sition of Fe2O3 is critical in determining the enhanced photo-electrochemical activity of the Fe2O3-rGO sample. Hence, cathodiclinear scan measurements were carried out to determine the con-duction band edge position of pristine Fe2O3 and rGO modifiedFe2O3 samples. This technique has been reported in many previousstudies [45e47]. Fig. 7 depicts cathodic linear potential scan mea-surements of Fe2O3 and Fe2O3-rGO samples scanned from 0.4to �1.2 V (vs. Ag/AgCl). A sudden increase in photocurrent density,due to the formation of an inversion layer [48] occurs atapproximately �0.60 V and �0.66 V in Fe2O3 and Fe2O3-rGO sam-ples respectively, which corresponds to the approximate conduc-tion band values of both samples. Based on the preceding data,including the band gap energies, derived from optical absorptionmeasurements; Scheme 2 (constructed using the lower limits of the

Fig. 6. Variation of photocurrent density vs. applied voltage (JeV) in the presence and absence of 0.5 M H2O2 of (a) Fe2O3, (b) Fe2O3-rGO. (c) Charge separation efficiency and (d)charge injection efficiency.

Fig. 7. Cathodic linear potential scan of (a) Fe2O3, (b) Fe2O3-rGO at 5 mV s�1 scan rate in 1 M NaOH solution.

Scheme 2. Schematic diagram for (a) energy band matching, (b) a photoelectrochemical cell with the Fe2O3-rGO photoanode.

A.G. Tamirat et al. / Journal of Power Sources 287 (2015) 119e128 127

A.G. Tamirat et al. / Journal of Power Sources 287 (2015) 119e128128

band gap energy) shows the energy level diagram of Fe2O3 modi-fied with rGO compared to water reduction and oxidation poten-tials. Once a photon of appropriate wavelength is absorbed by thehost Fe2O3 photoanode the photogenerated electrons and holeswill be formed. The work function of graphene and the conductionband of hematite allow the photogenerated electrons to be easilytransferred from the hematite surface to the rGO sheet.

4. Conclusions

In summary, a 3D urchin-like Fe2O3 photoanode modified withreduced graphene oxide was synthesized using facile hydrothermaland spin coating methods. Fe2O3 nanostructure modified withultra-thin rGO sheet shows enhanced photocurrent density(1.06 mA cm�2), which is 1.47 times higher than that of the pristineFe2O3 at 1.23 V vs. RHE. Moreover, a two-fold enhancement in thephotoconversion efficiency was achieved. More importantly, thehighest photoconversion efficiency was obtained at 1.06 V which is30 mV lower than that of pristine Fe2O3 (1.09 V). The enhancementis attributed to the dual functions of rGO as both electron con-ducting scaffold and surface passivation layer that improves thepoor charge separation efficiency and surface recombination inhi-bition property respectively. This synergetic effect is further evi-denced from the 1.82 times higher charge separation efficiency (at1.23 V vs. RHE) and the 1.67 times higher charge injection efficiency(at 1 V vs. RHE) of rGO modified Fe2O3 photoanode. A series ofelectrochemical, photoelectrochemical and impedance spectros-copy measurements were also carried out to confirm the origin ofthe improved photoactivity of the rGO modified photoanode. Thecurrent strategy is simple and highly scalable to other semi-conductor electrodes that have a very short excited-state lifetimeand high surface recombination.

Acknowledgments

The financial supports from the Ministry of Science and Tech-nology (MoST) (101-3113-E-011-002, 101-2923-E-011-001-MY3,100-2221-E-011-105-MY3), and the Top University Projects fromthe Ministry of Education (MoE) (100H451401), as well as the fa-cilities supports from the National Synchrotron Radiation ResearchCenter (NSRRC) and National Taiwan University of Science andTechnology (NTUST) are acknowledged.

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.jpowsour.2015.04.042.

References

[1] H. Dotan, O. Kfir, E. Sharlin, O. Blank, M. Gross, I. Dumchin, G. Ankonina,A. Rothschild, Nat. Mater. 12 (2013) 158e164.

[2] J. Brillet, M. Gr€atzel, K. Sivula, Nano Lett. 10 (2010) 4155e4160.[3] T.K. Townsend, E.M. Sabio, N.D. Browning, F.E. Osterloh, Energy & Environ. Sci.

4 (2011) 4270e4275.[4] Y. Ling, G. Wang, D.A. Wheeler, J.Z. Zhang, Y. Li, Nano Lett. 11 (2011)

2119e2125.[5] S.C. Warren, K. Voïtchovsky, Hen Dotan, Celine M. Leroy, Maurin Cornuz,

Francesco Stellacci, C�ecile H�ebert, A. Rothschild, M. Gr€atzel, Nat. Mater. 12(2013) 842e849.

[6] R. Franking, L. Li, M.A. Lukowski, F. Meng, Y. Tan, R.J. Hamers, S. Jin, Energy &Environ. Sci. 6 (2013) 500e512.

[7] Y. Hou, F. Zuo, A. Dagg, P. Feng, Angew. Chem. Int. Ed. 52 (2013) 1248e1252.[8] J.Y. Kim, G. Magesh, D.H. Youn, J.-W. Jang, J. Kubota, K. Domen, J.S. Lee, Sci.

Rep. 3 (2013) 1e8.[9] P. Liao, M.C. Toroker, E.A. Carter, Nano Lett. 11 (2011) 1775e1781.

[10] H.G. Cha, H.S. Noh, M.J. Kang, Y.S. Kang, New. J. Chem. 37 (2013) 4004e4009.[11] Y. Hou, F. Zuo, A. Dagg, P. Feng, Angew. Chem. 125 (2013) 1286e1290.[12] X.Y. Meng, G.W. Qin, S. Li, X.H. Wen, Y.P. Ren, W.L. Pei, L. Zuo, Appl. Phys. Lett.

98 (2011), 112104e112104(112103).[13] M. Gaudon, N. Pailh�e, J. Majimel, A. Wattiaux, J. Abel, A. Demourgues, J. Solid

State Chem. 183 (2010) 2101e2109.[14] B.A. Aragaw, C.-J. Pan, W.-N. Su, H.-M. Chen, J. Rick, B.-J. Hwang, Appl. Catal. B:

Environ. 163 (2015) 478e486.[15] M. Dinc�a, Y. Surendranath, D.G. Nocera, Proc. Natl. Acad. Sci. 107 (2010)

10337e10341.[16] G. Wang, Y. Ling, X. Lu, T. Zhai, F. Qian, Y. Tong, Y. Li, Nanoscale 5 (2013)

4129e4133.[17] A.G. Tamirat, W.-N. Su, A.A. Dubale, H.-M. Chen, B.-J. Hwang, J. Mater. Chem. A

3 (2015) 5949e5961.[18] M. Barroso, A.J. Cowan, S.R. Pendlebury, M. Gr€atzel, D.R. Klug, J.R. Durrant,

J. Am. Chem. Soc. 133 (2011) 14868e14871.[19] L. Badia-Bou, E. Mas-Marza, P. Rodenas, E.M. Barea, F. Fabregat-Santiago,

S. Gimenez, E. Peris, J. Bisquert, J. Phys. Chem. C 117 (2013) 3826e3833.[20] M.J. Katz, S.C. Riha, N.C. Jeong, A.B.F. Martinson, O.K. Farha, J.T. Hupp, Coord.

Chem. Rev. 256 (2012) 2521e2529.[21] A.K. Agegnehu, C.-J. Pan, J. Rick, J.-F. Lee, W.-N. Su, B.-J. Hwang, J. Mater. Chem.

22 (2012) 13849e13854.[22] A.A. Dubale, W.-N. Su, A.G. Tamirat, C.-J. Pan, B.A. Aragaw, H.-M. Chen, C.-

H. Chen, B.-J. Hwang, J. Mater. Chem. A 2 (2014) 18383e18397.[23] G. Giovannetti, P.A. Khomyakov, G. Brocks, V.M. Karpan, J. van den Brink,

P.J. Kelly, Phys. Rev. Lett. 101 (2008) 026803.[24] R. van de Krol, Y. Liang, J. Schoonman, J. Mater. Chem. 18 (2008) 2311e2320.[25] F. Meng, J. Li, S.K. Cushing, J. Bright, M. Zhi, J.D. Rowley, Z. Hong,

A. Manivannan, A.D. Bristow, N. Wu, ACS Catal. 3 (2013) 746e751.[26] L. He, L. Jing, Y. Luan, L. Wang, H. Fu, ACS Catal. 4 (2014) 990e998.[27] S. Agarwala, Z.H. Lim, E. Nicholson, G.W. Ho, Nanoscale 4 (2012) 194e205.[28] H.F. Chen, G.D. Wei, X. Han, S. Li, P.P. Wang, M. Chubik, A. Gromov, Z.P. Wang,

W. Han, J. Mater. Sci. Mater. Electron. 22 (2011) 252e259.[29] K. Sivula, R. Zboril, F. Le Formal, R. Robert, A. Weidenkaff, J. Tucek, J. Frydrych,

M. Gr€atzel, J. Am. Chem. Soc. 132 (2010) 7436e7444.[30] H. Paul, D. Mohanta, Appl. Phys. A 103 (2011) 395e402.[31] I. Cesar, K. Sivula, A. Kay, R. Zboril, M. Graetzel, J. Phys. Chem. C 113 (2009)

772e782.[32] S. Shen, P. Guo, D.A. Wheeler, J. Jiang, S.A. Lindley, C.X. Kronawitter, J.Z. Zhang,

L. Guo, S.S. Mao, Nanoscale 5 (2013) 9867e9874.[33] Y. Hou, F. Zuo, A. Dagg, P. Feng, Nano Lett. 12 (2012) 6464e6473.[34] J. Lu, J.-x. Yang, J. Wang, A. Lim, S. Wang, K.P. Loh, ACS Nano 3 (2009)

2367e2375.[35] D.K. Bora, A. Braun, S. Erat, O. Safonova, T. Graule, E.C. Constable, Curr. Appl.

Phys. 12 (2012) 817e825.[36] J. Zhao, F.E. Huggrns, Z. Feng, G.P. Huffman, Clays Clay Min. 42 (1994)

737e746.[37] L.C.C. Ferraz, W.M. Carvalho, D. Criado, F.L. Souza, ACS Appl. Mater. Interfaces

4 (2012) 5515e5523.[38] J. Deng, J. Zhong, A. Pu, D. Zhang, M. Li, X. Sun, S.-T. Lee, J. Appl. Phys. 112

(2012) 084312.[39] H. Zhang, X. Lv, Y. Li, Y. Wang, J. Li, ACS Nano 4 (2009) 380e386.[40] Q. Xiang, J. Yu, M. Jaroniec, Chem. Soc. Rev. 41 (2012) 782e796.[41] J. Young Kim, J.-W. Jang, D. Hyun Youn, J. Yul Kim, E. Sun Kim, J. Sung Lee, RSC

Adv. 2 (2012) 9415e9422.[42] A.C. Nielander, M.J. Bierman, N. Petrone, N.C. Strandwitz, S. Ardo, F. Yang,

J. Hone, N.S. Lewis, J. Am. Chem. Soc. 135 (2013) 17246e17249.[43] S. Rai, A. Ikram, S. Sahai, S. Dass, R. Shrivastav, V.R. Satsangi, RSC Adv. 4 (2014)

17671e17679.[44] H. Dotan, K. Sivula, M. Gratzel, A. Rothschild, S.C. Warren, Energy & Environ.

Sci. 4 (2011) 958e964.[45] T.-F. Yeh, S.-J. Chen, C.-S. Yeh, H. Teng, J. Phys. Chem. C 117 (2013) 6516e6524.[46] H.R. Sprunken, R. Schumacher, R.N. Schindler, Faraday Disc. Chem. Soc. 70

(1980) 55e66.[47] R. Beranek, Adv. Phys. Chem. 2011 (2011), http://dx.doi.org/10.1155/2011/

786759.[48] T.-F. Yeh, F.-F. Chan, C.-T. Hsieh, H. Teng, J. Phys. Chem. C 115 (2011)

22587e22597.