Nanostructured r-Fe2O3 Electrodes for Solar Driven Water Splitting: Effect of DopingAgents on Preparation and Performance

Sina Saremi-Yarahmadi,† K. G. Upul Wijayantha,*,‡ Asif A. Tahir,‡ and Bala Vaidhyanathan†

Department of Materials and Department of Chemistry, Loughborough UniVersity,Loughborough LE11 3TU, United Kingdom

ReceiVed: February 1, 2008; ReVised Manuscript ReceiVed: January 18, 2009

Si-doped nanostructured hematite (R-Fe2O3) has attracted significant attention as a low-cost, high-efficiencycandidate material for photoelectrochemical water splitting. In this work, we investigated the effect of Siincorporation on the preparation and performance of R-Fe2O3 films produced by atmospheric pressure chemicalvapor deposition (APCVD). Structural, optical, electrical, and photoelectrochemical characterization of dopedand undoped hematite films was performed using XRD, FIB/SEM, Raman spectroscopy, UV-vis absorptionspectroscopy, J-V and electrochemical capacitance measurements. It was concluded from the XPS data thatSi is incorporated in the hematite structure as Si4+. The results suggest that Si-free additives as well as theuse of fluorinated transparent conducting oxide (FTO) substrates can influence the preferred orientation ofhematite films. It was also found that the incorporation of silicon at very low levels led to the formation ofdisorder in the hematite structure. Moreover, it is shown that the optical bandgap of Si-doped film increasedwith the increase of TEOS flow rate. It contributed to the reduction in the size of the hematite nanoparticlesand the size quantization effect. The observed donor densities for the doped samples seemed to be muchhigher than the true values, mainly because the total capacitance measured was higher than space chargelayer capacitance, which resulted from the surface area enhancement in the doped films. Therefore, it isconsidered that donor densities of doped films were smaller than that of the undoped hematite films.

Introduction

Nowadays, 95% of hydrogen is produced using a catalyticthermochemical conversion process which reforms natural gas.1

If hydrogen generation continues to rely on this process, whichleads to a significant amount of CO2 emission, then it bringsinto question the claim that hydrogen is an environmentallyfriendly fuel. Evidently, hydrogen generation from clean andenvironmentally friendly sources such as solar energy isrequired. Photoelectrochemical (PEC) water splitting, alsoknown as solar-driven water splitting, is one of the mostpromising means of producing hydrogen from a renewablesource. In this process, water molecules are broken intohydrogen and oxygen, using sunlight. There are several advan-tages to this approach, such as abundance of solar energy,inexpensive raw materials, and ease of mass production.2 Inaddition, the production of hydrogen by sunlight will minimizethe need for expensive infrastructure for transport in a futureprospective H2 economy.3

Since the pioneering work of Fujishima and Honda4 in 1972,there has been a huge interest in developing semiconductorcatalysts for hydrogen generation. Since then, a wide range ofmaterials have been used for the development of photoanodesin PEC cells such as TiO2,5 WO3,6 and Fe2O3.7,8 Hematite isthe material of choice, as it fulfills most of the requirements ofa good photocatalyst material such as appropriate bandgap,chemical and photelectrochemical stability, low cost, and easeof fabrication. However, solar to chemical conversion efficien-cies reported for hematite are relatively low. Poor performance

of hematite electrodes is widely attributed to its low lightabsorption, slow charge transfer kinetics, and a high rate ofelectron-hole recombination9 due to its short hole diffusionlength.10

It has been suggested that doping can increase the conversionefficiency of hematite photoelectrodes by altering the bandgap.11

In the last few decades, a wide range of doping agents such asTa,12 Mg,13,14 Zn,15 and Si14,16 has been used, and the effects ofdoping agents on either n-type or p-type characteristics ofhematite have been studied. The best performance reported sofar is that of Kay et al.16 who used atmospheric pressurechemical vapor deposition (APCVD) to make silicon-dopedR-Fe2O3 photoelectrodes with a photocurrent density of 2.2 mA/cm2 at 1.23 VRHE. This was about 65% enhancement of thephotocurrent density for hematite, which was previously reportedby the same group.17 It was suggested that one of the factorsresponsible for this significant improvement was Si-doping andsubsequent formation of an orientated dendritic nanostructure,which minimized the distance an average photogenerated holehad to travel to the electrode/solution interface while the filmwas thick enough for sufficient light absorption. Additionally,it was also proposed that a thin silica interfacial layer might beresponsible for part of the gain in the film performance.16

Glasscock et al.18 studied undoped and doped (with Si and Ti)hematite films prepared using reactive magnetron sputtering.They reported that doped hematite electrodes (in particular, Ti-doped samples) showed much higher photoelectrochemicalperformance, on the order of 0.57 mA/cm2 at 0.5 V vs SCE.Adopting a different interpretation from that of Kay et al.,16

the enhanced photocurrent was attributed to lowering therecombination losses as a result of the improved charge-transferrate coefficient at the surface of doped hematite electrodes whileno evidence of the effect of dopants on nanostructure was

* Corresponding author. K. G.Upul Wijayantha. E-mail: U.Wijayantha@lboro.ac.uk. Tel: +44(0)1509 222574. Fax: +44(0)1509 223925.

† Department of Materials.‡ Department of Chemistry.

J. Phys. Chem. C 2009, 113, 4768–47784768

10.1021/jp808453z CCC: $40.75 2009 American Chemical SocietyPublished on Web 02/27/2009

observed.18 They also proposed that part of this gain waspossibly due to passivation of the grain boundaries by thedopants. They concluded that in Si-doped samples a high levelof surface states in conjunction with smaller grains increasedrecombination at the surface and at grain boundaries resultingin relatively poor photoelectrochemical activity (0.10 mA/cm2

at 0.5 V vs SCE).Although the results reported by Gratzel and co-workers for

Si-doped R-Fe2O3 photoelectrodes prepared by APCVD arecertainly very promising,16 other attempts to prepare Si-dopedelectrodes by methods such as reactive magnetron sputtering18

and spray pyrolysis19 have not lived up to expectations.Unfortunately, there is no further work on studying the effectsand the exact role of silicon on R-Fe2O3 prepared by APCVD,impeding further development. In the present work, we studiedthe role of silicon-containing as well as Si-free additives in thepreparation and performance of hematite photoelectrodes. Athorough investigation of the physical and photoelectrochemicalproperties of hematite films has been presented, and attemptshave been made to quantify the amount of silicon present indifferent films.

Experimental Section

Sample Preparation. R-Fe2O3 was deposited on 1 × 2 cm2

conductive fluorine-doped tin oxide (FTO) glass substrates. Priorto deposition, substrates were cleaned ultrasonically usingpropanol, deionized water, acetone, and absolute ethanol in thatorder. Films were deposited by APCVD where iron pentacar-bonyl (Fe(CO)5), tetraethoxysilane (TEOS), tetramethoxysilane(TMOS), and trimethyl orthoformate (TMOF) were used asprecursors (Figure 1). TOMF was selected as the Si-freeadditive, as it closely resembles the molecular structures of Siprecursors used in this work (TMOS and TEOS). Substratetemperature was measured by a K-type thermocouple and wasmaintained at about 375 °C during deposition. Argon was usedas the carrier gas and bubbled through 1 mL of each liquid atflow rates shown in Table 1. The flow rates mentioned here areof the samples with the best PEC performance for each type ofsample. Aerosols were then mixed with air at flow rate of 1.7L/min, and the mixture was delivered to the substrate using aglass tube with inner diameter of 10 mm positioned 20 mm

above the heated substrate. The deposition time for all sampleswas maintained at 3 min.

Structural Characterization. The phase and crystallinity ofdeposited films on microscope glass slides (soda glass) and FTOwere characterized using a Bruker D8 ray diffractometer (XRD)operating with monochromatic high-intensity Cu KR (λ )0.154 184 nm) radiation using a PSD detector. Nanostructure,surface morphology, and thickness of the films were studiedusing a FEI Nova 600 Nanolab FIB/SEM system.

Elemental Analysis. Iron and silicon content of differentfilms were quantified using a sector field high-resolutioninductively coupled plasma mass spectroscopy (SF-HR-ICP-MS) apparatus. Samples were prepared according to thefollowing procedure. Films were deposited on FTO substratesand were then placed in 7 mg of concentrated HCl (36% w/waqueous solution) for 8 h. Substrates were then removed fromthe sample tube, and the solution was diluted using 20 mg ofdeionized water prior to measurements. Standard stock solutionof silicon and iron containing 100 ppb of each element in high-purity 2% HNO3 were prepared and were then diluted to obtain10 different reference solutions of Si and Fe at different levelsof concentration (from 10 to 100 ppb).

Surface Characterization. XPS spectra were recorded on aVG Scientific Escalab Mk I instrument operating with an un-monochromatized Al KR X-ray source (1486.6 eV). Ramanspectroscopy was undertaken using a HORIBA Jobin YvonLabRAM HR (with 632.8 nm He-Ne laser) Raman spectro-photometer. The spectra were recorded in the range 100-1600cm-1.

Optical and Photoelectrochemical Measurements. Opticalabsorption measurements were performed using a Lambda 35Perkin-Elmer UV-vis spectrophotometer. Photocurrents weremeasured using a three-electrode configuration in 1 M NaOHelectrolyte, Ag/AgCl/KCl as the reference electrode, and aplatinum wire as the counter electrode. The potential of thephotoelectrode was controlled by a potentiostat (microAutoLab,type III) and is reported against the reversible hydrogen electrode(RHE). In the electrochemical cell, light enters through a quartzwindow and travels about a 5 mm path length in the electrolytebefore illuminating the photoelectrode. The Fe2O3 electrode wasilluminated through electrolyte side, and illumination area was0.8 cm2. The illumination source was AM 1.5 class A solarsimulator (Solar Light 16S - 300 solar simulator) equipped witha 300 W xenon lamp. The intensity of the light was calibratedat 1000 W/m2 using a class II pyranometer (PMA2144, SolarLight Co., Inc.) equipped with a digital photometer (PMA2100,Solar Light Co., Inc.).

Electrical Measurements. Electrochemical impedance spec-troscopic measurements were carried out in the dark toinvestigate electrical properties of the hematite films. Theexperiments were performed using a computer-controlled po-tentiostat (AutoLab, PGSTAT12). The electrochemical cell wasin the three-electrode configuration as mentioned above. Thesuperimposed a.c. signal was maintained at 20 mV, while thefrequency range was between 120 KHz and 10 Hz at potentialsfrom -0.8 to 0.2 V vs VAg/AgCl under dark conditions. A simpleRC circuit was used as the fitting model to determine thecapacitance(C)of thespacechargelayerat thehematite-electrolyteinterface. The values for flatband potential (Vfb) and donordensity (ND) were calculated using the Mott-Schottky relation-ship (eq 1) and taking into account the intercept and the slopeof the plot of (A/C)2 vs applied voltage.

Figure 1. Ball-and-stick representation of three different precursors:tetraethoxysilane (TEOS), tetramethoxysilane (TMOS), and trimethylorthoformate (TMOF).

TABLE 1: Carrier Gas (Ar) Flow Rate for DifferentPrecursors

argon flow rate (mL/min)

sample name Fe(CO)5 TEOS TMOS TMOF

pure pentacarbonyl 17 - - -TMOF 17 - - 20TMOS 17 - 2 -TEOS 17 55 - -high TEOS 17 90 - -

(A/C)2 ) 2(eNDε0εr)-1(|V - Vfb|-kT/e) (1)

Nanostructured R-Fe2O3 Electrodes J. Phys. Chem. C, Vol. 113, No. 12, 2009 4769

where A is the surface area of the film, εr is the dielectricconstant of hematite (taken as 1220), ε0 is the permittivity offree space, T is the temperature, and k is the Boltzmann’sconstant.

Results and Discussion

XRD results for films deposited in the presence and theabsence of different Si precursors are shown in Figure 2. In thespectra of all films apart from TMOS, the dominant hematitepeak is (110), while in the spectrum of TMOS, the (104)reflection is slightly stronger. It has previously been suggestedthat the (110) preferential orientation is the result of the presenceof TEOS during APCVD.16,21 However, XRD patterns for ourhematite films shown in Figure 2 indicate the presence of thepreferential (110) orientation in our deposition regime when thefilms were formed on FTO substrates by APCVD regardless ofthe presence or the absence of Si precursors.

The structural characteristics of doped and undoped samplesdeposited on different substrates were investigated (Figure 3).The XRD patterns of pure pentacarbonyl and pentacarbonyl/TEOS deposited on microscope glass slides clearly show adifferent orientation from that of the films deposited on FTOsubstrates. In the TEOS sample, there is no evidence of the

(110) reflection, whereas the (104) reflection is predominantwhen the substrate is soda glass. In the case of the purepentacarbonyl sample, the (104) reflection is still the dominantpeak when the substrate is soda glass. This is a clear differencefrom the predominant (110) orientation observed in similarsamples deposited on conductive FTO glass substrate (Figure4). It suggests that substrate plays a major role in determiningthe orientation of films which has also been observed by othergroups.18 For example, preferential (110) orientation wasreported when undoped hematite films were deposited onconductive ITO glasses using other preparation methods suchas spray pyrolysis.22,23 Therefore, it can be suggested that siliconalone may not be responsible for the change in the orientation;it is the effect of FTO on the substrate along with the presenceand decomposition of the second precursor (e.g., TMOS, TEOS,and TMOF), which results in the preferential orientation ofhematite films.

Crystallite size of different samples was estimated using theScherrer equation:

where d is the crystallite size, λ is the wavelength of radiation,� is the line broadening, and θ is the angle of diffraction. Itwas calculated that the crystallite sizes of all the samples weresimilar, ranging from ∼30 nm to ∼40 nm based on the (110)peak analysis.

In a previous study by Kay et al.,16 the level of Si dopingwas calculated on the basis of the weight loss of the precursors.However, this technique does not account for the amount ofsilicon which is lost during the APCVD process. Precisemeasurement of the amount of silicon present in the hematitefilm is extremely difficult and challenging. In order to makesome progress in this direction, ICP-MS measurements werecarried out for the solutions prepared by digesting films in orderto analyze the Si content in each hematite film. Three differentsamples were characterized using this method. The samplesselected were the films deposited with different argon flow ratesfor the silicon precursor, TEOS. As shown in Figure 5, the purepentacarbonyl samples which were prepared in the absence ofTEOS had the lowest amount of silicon varying from 55 to 66

Figure 2. XRD patterns of R-Fe2O3 (hematite) films prepared usingdifferent precursors. Standard powder pattern of hematite is shown bylines with plane indices in hexagonal coordinates.

Figure 3. XRD patterns of pure pentacarbonyl and TEOS samplesdeposited on soda glass slide.

Figure 4. XRD pattern of different samples deposited on FTO glass.(110) and (104) reflections of hematite are marked with an arrow. Thepeak marked with 0 belongs to SnO2 in the FTO substrate.

d ) 0.89λ� cos θ

(2)

4770 J. Phys. Chem. C, Vol. 113, No. 12, 2009 Saremi-Yarahmadi et al.

ppb. The existence of a relatively high amount of silicon inthis sample could be due to the dissolution of silicon present atthe glass surface (i.e., the back side of the FTO glass) duringthe sample preparation step or diffusion of Si into the hematitefilm from glass substrate during the APCVD process. Siliconcontent for the sample with the best PEC performance (whichis corresponding to the Ar flow rate for TEOS ) 55 mL/min)was measured in the range 61-71 ppb, which is slightly higherthan that of pure pentacarbonyl sample. In order to prepareelectrodes with higher levels of Si (called “High TEOS”), theargon flow rate of the TEOS precursor was increased from 55to 90 mL/min. The increase in the argon flow rate led to anincrease in the amount of silicon detected from ∼70 ppb to∼190 ppb. This trend was consistent over a range of samplesets (Figure 5). The results clearly confirm the presence ofsilicon in the hematite films. They also confirm that the increaseof argon flow rate leads to a higher amount of silicon in thehematite films. We believe that the determination of the exactamount of silicon in the hematite films using ICP-MS measure-ments was complicated by the fact that the films were made onFTO glass. Hence, it was difficult to prevent leaching Si intothe digested solution from the back side and edges of the glasssubstrate. However, the qualitative analysis of our ICP-MSresults indicates that generally the amount of Si in the hematitefilms could be adjusted by controlling the argon flow rate.

Once the presence of Si in hematite photoelectrodes had beenconfirmed, the investigation was extended to study the chemicalnature of the Si. XPS spectra of two different samples (purepentacarbonyl and TEOS) are shown in Figure 6. If Si wasincorporated as elemental Si, the binding energies of Si 2s andSi 2p (corresponding to the elemental Si) were expected to beobserved at 99 and 150.7 eV, respectively.24 Neither of the peakswas observed in the spectra obtained for both films. However,two peaks were observed at 102 and 154 eV in the spectra ofthe TEOS sample, which we attributed to the Si 2s and Si 2pin SiO2 structure. These results suggest that silicon is not presentas elemental silicon in hematite films. Rather, silicon isincorporated into the hematite structure as Si4+ ions. However,due to the small amount of silicon present in the electrodes,the quality of the XPS spectra could not be improved. Therefore,

it was difficult to draw any further conclusions on the oxidationstate of the silicon. We believe that this is an area where furtherstudies are required.

It has been suggested that doping can improve PEC perfor-mance of electrodes through enhancement of conductivity.25,16

For example, it has been shown that addition of Ti4+ increased

Figure 5. The silicon content measured for three samples prepared by changing the argon flow rates for the TEOS precursor. The silicon contentdetected for four sets of samples, each set prepared at increasing argon flow rate is shown in the table.

Figure 6. XPS spectra obtained for two different electrodes. (a)Binding energies of Fe3+ and Si4+ in the Fe2O3 and SiO2 structures aremarked, respectively. (b) XPS spectra of two samples at bindingenergies close to 715 eV. The expected position of the satellite peak(characteristic of Fe2+) is marked with arrows.

Nanostructured R-Fe2O3 Electrodes J. Phys. Chem. C, Vol. 113, No. 12, 2009 4771

the conductivity of hematite.26 The mechanism suggested wasthat titanium (Ti4+) entered the hematite lattice by substitution.This would lead to the formation Fe2+ sites in order to maintainthe charge balance in the lattice. Fe2+ would act as a donorcenter with respect to the surrounding Fe3+ and can provideelectrons to Fe3+ by thermal excitation.26 Another explanationis that tetravalent ions may introduce electrons to the conductionband of the semiconductor, and under the influence of a potentialgradient, the localized electrons can produce electric current.25

It has also been shown that tetravalent ions can enter thestructure of hematite without inducing any changes in theoxidation states of Fe ions. In this mechanism, the chargecompensation is carried out by the formation of cationvacancies.27-29

In order to understand whether the incorporation of Si4+

introduced any changes in the oxidation states of iron, we triedto examine different hematite films using XPS. It is known thatthe binding energy for Fe3+ 2p3/2 is 711.2 in all iron oxides,while Fe2+ shows binding energy of approximately 709.7 eVin FexO.30-32 The presence of a satellite at ∼715 eV is alsoregarded as characteristic of Fe2+. According to our resultsshown in Figure 6b, the binding energies of 711.5 and 711.6eV were observed for TEOS and pure pentacarbonyl samples.Moreover, there is an apparent shoulder close to 715 eV, whichsuggests that Fe2+ was present in the TEOS sample. This alsoprovides indirect evidence for the presence of Si4+ in the TEOSsample.

The Raman spectra of different samples are shown in Figure7. The spectrum obtained for the pure pentacarbonyl sample istypical of hematite phase showing peaks at 221, 287, 406, 495,and 611 cm-1. The main difference between the differentsamples is the presence of a forbidden peak at 660 cm-1 in thespectrum of the TMOS and TEOS samples. It is also observedthat the intensity of this peak increased from sample TMOS toTEOS. As the intensity of the peak at 660 cm-1 increased, thePEC performance of the samples also improved. It has beenreported that this peak is the result of the breakdown of Ramansymmetry selection rules induced by short-range order such asnanodimensions or lattice defects.33 It has also been attributedto magnetite (Fe3O4).8,34

Yu et al.35 reported the appearance of forbidden modes inthe Raman spectra of SnO2 nanoparticles. It was attributed to alarge amount of defects: vacancies of oxygen, vacancy clusters,and local lattice disorder at the interface and interior surface,

which led to significant lattice distortion and an evident space-symmetry reduction and the appearance of a group of newRaman peaks. In addition to the above defects, incorporationof silicon into the structure of hematite can also result in theformation of lattice disorder. Berry et al.27-29 reported themechanism through which tetravalent ions enter the structureof hematite. They have proposed a model which describes thatthese dopant ions adopt two distinct sites in the hematitestructure; partially substituting at octahedral Fe sites andoccupying the vacant interstitial octahedral sites in R-Fe2O3.

28

In other words, two iron ions are replaced by three tetravalentions and the charge neutrality is maintained through theformation of cation vacancies. It was also found that no changein the oxidation state of Fe occurred; the charge balance wasachieved by the formation of different defect clusters asdescribed above.28

We believe the peak at 660 cm-1 observed for TEOS andTMOS samples is the result of formation of lattice disorder inthe structure of the doped hematite film as a result of Siincorporation and also changes in the nanostructure of the films.It has been observed that in a mixed compound of Fe2O3-Cr2O3

the symmetry breaking resulted in the appearance of this Ramanpeak.36 Bersani et al.37 suggested that the nonperfect stoichi-ometry and also the presence of some metal ions which wouldreplace iron were responsible for the formation of disorder andthe appearance of this peak. Glasscock et al.18 also observedthe existence of the peak at 660 cm-1 when the hematite filmswere doped with Ti, and they attributed the existence of thispeak to the disorder in the films.

SEM images of four types of hematite films showed adistinguishable difference in the surface morphology, in par-ticular, for the film made in the presence of Si precursors. Asshown in Figure 8, agglomerate/cluster structures in the purepentacarbonyl and TMOF samples were present in two majorsizes, 350 and 750 nm, while the introduction of silicon intothe electrodes resulted in a relatively uniform dispersion ofapproximately 350 nm cluster/agglomerate structures.

The tilted topography of each film at a much highermagnification is shown in Figure 9. It is clear that thenanostructure of the pure pentacarbonyl sample is significantlydifferent from that of other samples where there is no indicationof the existence of fine nanostructure in this sample. As TMOFis introduced as the second precursor, although the formationof cluster/dendrite structure is initiated, it still consists of largeparticles whose sizes were distributed over a wide range. Afterthe addition of silicon, the cluster/agglomerate nanostructurebecame finer; however, the particle size of the TMOS sampleis larger than that of the TEOS sample. In contrast to thenanostructure of the other three electrode types, the TEOSelectrodes consisted of finer agglomerates/cluster nanostructurescomprising small individual particles uniformly distributed insize. Chu et al.38 investigated the relative stability of TEOS andTMOS in the gas phase. Their results showed a significantdifference in the decomposition of TEOS and TMOS. The maindifference was the formation of water during the decompositionof TEOS, which accelerated further decomposition of TEOSvia hydrolysis. Shimojo et al.39 and Takeguchi et al.40 studiedthe effect of water vapor on the deposition of iron oxidenanorods from iron pentacarbonyl by electron beam-induceddeposition. They reported that addition of water facilitated thedecomposition of iron pentacarbonyl, reduced the carbon contentin nanorods, and induced the formation of crystalline iron oxide.Formation of small crystallites in the presence of TEOScompared to that of TMOS can be attributed to the effect of

Figure 7. Raman spectra obtained for different samples. The positionof 660 cm-1 is marked with a line.

4772 J. Phys. Chem. C, Vol. 113, No. 12, 2009 Saremi-Yarahmadi et al.

water formed during the decomposition of TEOS on theoxidation of iron pentacarbonyl. As water is introduced intothe system by decomposition of TEOS, oxidation of ironpentacarbonyl accelerates rapidly and nucleation of iron oxideclusters increases. The increase in the nucleation sites resultsin a smaller crystallite size. However, in the case of TMOS inwhich the water formation is less favorable during the decom-position process, nucleation of iron oxide clusters takes placeat a lower rate resulting in larger size crystallites. Smallercrystallite size along with less agglomeration could facilitatethe separation of photogenerated charge carriers within theTEOS-based hematite electrodes as the size of each particle inthe electrode structure becomes more compatible with the holediffusion length of hematite.10,16

The cross section and film thickness of hematite electrodeswere studied by FIB/SEM. As shown in Figure 10, purepentacarbonyl and TMOF films had thicknesses of approxi-mately 350 and 300 nm, respectively. The thickness of bothTEOS and TMOS films are on the order of 380-450 nm wherethe TMOS sample is generally thicker than TEOS. It is apparentthat the growth of the hematite films on FTO glass follows thesurface features of the substrate and almost conforms to thepattern of the FTO substrate. The difference in the cross-sectional nanostructure of the each type of electrode is clearlyvisible in Figure 10. The film prepared by only iron pentacar-bonyl precursor (with no additive agents) is more compact innature, and the lack of porosity agrees well with the tiltedtopography in Figure 8a. Porous structures are just beginning

to appear for the hematite electrodes deposited in the presenceof TMOF precursor. The porosity substantially increases forthe thin films deposited with silicon precursors (TMOS andTEOS). The films deposited in the presence of TEOS precursorcontain the highest porosity.

Optical properties of the films were studied by UV-visspectroscopy. Absorption spectra of the films are shown inFigure 11. The light absorption is high in the short-wavelengthregion of the spectra. As the wavelength becomes longer, theabsorption is gradually decreased. As the spectra reach thecharacteristics of the optical absorption threshold for hematite(∼600 nm), the absorption drops rapidly and extends to the redregion as a plateau. The absorption observed after this thresholdcould be the result of the light scattering properties of nano-particles at longer wavelengths.

The optical bandgap of hematite films can be estimated bycalculating the intercept of the extrapolated linear fit to theexperimental data of a plot of (RhV)2 versus incident photonenergy (hν) near the absorption edge where R is the absorptioncoefficient of the film.41 The inset of Figure 11 shows (RhV)2

versus photon energy plots for each type of hematite electrode.Indirect bandgap values measured are 2.05, 2.07, 2.08, and 2.12eV for pure pentacarbonyl, TMOF, TMOS, and TEOS elec-trodes, respectively, which closely match the values reportedin the literature.42 The bandgap values determined for purepentacarbonyl, TMOF, and TMOS are scattered around 2.0 eV,whereas a slight enhancement in the bandgap was observed forthe TEOS sample.

Figure 8. SEM pictures obtained for four types of hematite films: pure pentacarbonyl (a), TMOF (b), TMOS (c), TEOS (d). Scale bars indicate1 µm.

Nanostructured R-Fe2O3 Electrodes J. Phys. Chem. C, Vol. 113, No. 12, 2009 4773

We further investigated the effect of the TEOS flow rate onoptical properties of TEOS samples. Three different types of

samples prepared at different TEOS flow rates (30, 55, and 80mL/min) were examined. As shown in Figure 12, we observed

Figure 9. SEM pictures of four types of hematite electrodes prepared in our study: pure pentacarbonyl (a), TMOF (b), TMOS (c), and TEOSsample (d).

Figure 10. Cross-sectional FIB/SEM images of four types of hematite electrodes prepared in our study: pure pentacarbonyl (a), TMOF (b), TMOS(c), and TEOS (d). The deposited Pt layer and the FTO substrate are marked at each picture. Pictures are obtained at the tilt angle of 52°.

4774 J. Phys. Chem. C, Vol. 113, No. 12, 2009 Saremi-Yarahmadi et al.

a systematic increase in the indirect bandgap as the TEOS flowrate was increased in the sample deposition process. Thisbehavior is in good agreement with the effect of particle sizequantization on indirect bandgap of hematite.43,44 Therefore, itsuggests that the quantum size effect can be partly responsiblefor the increase of the bandgap of hematite, as the increase inthe TEOS flow rate would lead to decrease in the size ofhematite particles.

Current density-potential (J-V) curves of different samplesare shown in Figure 13. The electrodes made in the presenceof Si precursors show negligible dark currents (not more than2 µA/cm2) up to 1.6 VRHE, whereas the dark current for othertwo electrodes (pure pentacarbonyl and TMOF) started to riseat potentials as low as 1.4 VRHE. Pure pentacarbonyl electrodeshowed a negligible photocurrent density of 1.5 µA/cm2 at 1.23VRHE, whereas the samples prepared with TMOF had photo-current densities slightly better than that (∼6.3 µA/cm2 at thesame bias). Comparing the cross-sectional and topographicalSEM images and XRD data of pure pentacarbonyl and TMOFelectrodes, it is apparent that the porosity is slightly increasedin the TMOF electrode. Although the magnitude of the

photocurrent enhancement in TMOF electrode compared to thatof the pure pentacarbonyl electrode is small, this improvementcan be attributed to the difference in the nanostructure. The bestphotocurrent performance was obtained when TEOS was usedas the source of silicon, 1.67 mA/cm2 at 1.23 VRHE, which isless than the previously reported benchmark performance forelectrodes prepared by APCVD.16 The lower photocurrentachieved compared to the previous benchmark16 may beattributed to the differences in the amount of Si present in thesefilms, the precise control of which is rather difficult in ourcustom-built APCVD set up. When TMOS was used as the Sisource instead of TEOS, the photocurrent density decreased to0.62 mA/cm2. It is worth mentioning that the vapor pressure ofTEOS (<1 mmHg) is considerably lower than that of TMOS(173 mmHg). Therefore, obtaining the same level of Si as inthe case of the TEOS electrode requires far better precise controlover the amount of Si-containing agents reacting in the aerosol.It is apparent in SEM cross-sectional images (Figure 10c,d) thatthe TEOS electrode consisted of fine nanostructures (comparedto that of TMOS). As suggested by Kay et al.,16 the formationof fine nanostructures with the particle sizes comparable to thehole diffusion length of hematite may be responsible for thesuperior photocurrent. Furthermore, the role of the primaryparticle size cannot be neglected; a larger primary particle sizecan still be partly responsible for the drop in the photocurrent

Figure 11. Absorption spectra of four types of hematite electrodesprepared in this study: TEOS in magenta, TMOS in blue, TMOF ingreen, and pure pentacarbonyl in brown. (Inset) (ahν)2 vs wavelengthplots for corresponding electrodes. The optical bandgap was determinedby extrapolation.

Figure 12. Absorption spectra of Si-doped hematite films with threedifferent TEOS flow rates of 30, 55, and 80 mL/min. Inset shows theoptical band gap measurements of the films.

Figure 13. (a) Current density vs applied voltage (J-V) characteristicsof TEOS and TMOS samples in 1 M NaOH under AM 1.5 conditions.Inset shows the magnified J-V plot of the samples at the potentialrange of 1.1-1.6 V vs RHE. (b) Current density vs applied voltage(J-V) characteristics of pure pentacarbonyl and TMOF samples in 1M NaOH under AM 1.5 conditions.

Nanostructured R-Fe2O3 Electrodes J. Phys. Chem. C, Vol. 113, No. 12, 2009 4775

output in the TMOS sample in comparison to that of in TEOS.The chopped photocurrent pulses for the TMOS electrode (seethe inset of Figure 13b) clearly shows higher recombination atlower applied voltages in comparison to the case of TEOSsupporting our interpretation.

The capacitance data (represented by Mott-Schottky plots)for four different types of hematite electrodes are given in Figure14. The donor densities (uncorrected for the surface roughnessfactor) and flatband potentials deduced from the capacitancemeasurements are given in Table 2 along with the photocurrentonset values read from the linear fit of the intercept of J2-Vplots.

The introduction of organic additives (regardless of thepresence or absence of Si) appears to shift the photocurrent onsettoward the negative direction (Table 2). For example, hematiteelectrodes prepared using only iron pentacarbonyl alone showedphotocurrent onset potential at around 0.12 VAg/AgCl. After theaddition of TMOF, TMOS, and TEOS, the photocurrent onsetpotentials have moved to 0.07, -0.43, and -0.47 VAg/AgCl,respectively. It is clearly noticeable that, in particular, theintroduction of Si by the incorporation of Si precursors in theAPCVD process has significantly shifted the onset potentialstoward the negative direction. It is assumed that specificadsorption of decomposition products of TMOS and TEOS mayhave played a role in this shift. As far as the photoanodic process

is concerned, this apparent shift has a significant advantage forPEC water splitting device fabrication, although the exact originof this shift is yet unclear.

It is apparent that, at the potentials above -0.7 VAg/AgCl, allelectrodes are in the depletion region. Therefore, the slopes ofthe linear region between -0.7 and -0.4 VAg/AgCl of theMott-Schottky plots were used for the estimation of donordensity (uncorrected for the surface area) for each electrode andpresented in Table 2. Although the donor densities presentedin Table 2 suggest that the concentration of charge carriers doesnot vary significantly among different samples, the true donordensity for the samples made in the presence of Si precursors(in particular TEOS and TMOS samples) should be significantlysmaller than the calculated values (Figure 10 clearly shows amuch higher effective surface area for TMOS and TEOSsamples compared to that of pure pentacarbonyl and TMOFsamples which could eventually alter the estimated donor densityvalues). As a rule of thumb, if the surface area enhancement isabout 100 times, the true donor density is on the order of 1021

m-3. This value is at least 3 orders of magnitude smaller thanthe literature reported values for Fe2O3 electrodes.18,19 Therefore,we believe that the recorded total capacitance should besomewhat higher than the true space charge layer capacitance.Although the estimated donor density (after considering thesurface area enhancement of TEOS electrode) appeared to be

Figure 14. Mott-Schottky plots of four different hematite electrodes at the frequency of 100 kHz.

TABLE 2: Electrical and Photoelectrochemical Properties of Four Different Types of Hematite Electrodes Deduced fromMott-Schottky and J2-V Plots

R-Fe2O3 sample name donor densitya (m-3)flatband potential,Vfb (V vs VAg/AgCl)

photocurrent onset potential,Vonset (V vs VAg/AgCl)

pure pentacarbonyl 4.04 × 1024 -0.80 0.12TMOF 1.21 × 1025 -0.78 0.07TMOS 1.56 × 1025 -0.72 -0.43TEOS 2.52 × 1025 -0.77 -0.47

a The donor density values were not corrected for the true surface area.

4776 J. Phys. Chem. C, Vol. 113, No. 12, 2009 Saremi-Yarahmadi et al.

less than the reported values, the estimated space charge widthsfor all four cases generally suggest that the depletion layershould be extended significantly further into the interior of theTEOS and TMOS electrodes leading to better harvesting ofphotogenerated charge carriers. Considering the pattern ofphotocurrent performance obtained for four different types ofFe2O3 electrodes investigated in our study, this explanationagrees well.

An important assumption of the depletion layer model at thesemiconductor-electrolyte junction is that the semiconductoris not too heavily doped; hence, the space charge layercapacitance (Csc) is much smaller than the double layercapacitance (Cdl). Under this condition, the observed totalcapacitance is very close to the Csc (eq 3).

Therefore, consideration of the relatively low surface areadonor density obtained for pure pentacarbonyl and TMOFelectrodes should reflect the true donor density values. ForTMOS and TEOS electrodes, the Csc could be affected by twoaspects. The first is the significant surface area enhancement.The second is the decrease of the width of the space chargelayer induced by the dopants. The measured total capacitancefor these two electrodes, therefore, could be substantiallyinfluenced by the Cdl. Hence, a true estimation of the extensionof space charge layer and the donor density for TMOS andTEOS electrodes is difficult. A significant implication of thisphenomenon is the frequency dispersion of capacitance mea-surements for nanostructured electrodes. A detailed impedanceanalysis is being currently conducted to deduce the Csc fromthe recorded total capacitance of nanostructured electrodes,which is beyond the scope of this work.

Conclusions

Undoped and Si-doped nanocrystalline hematite electrodeswere successfully prepared through APCVD. It was observedthat the substrate played an important role in the formation ofthe (110) orientation in the films, which is the favorableorientation with regard to electronic conduction. The bestphotocurrent density achieved was 1.67 mA/cm2 when TEOSwas used as the silicon precursor. Although it is extremelydifficult to quantify the level of silicon incorporation in the films,our qualitative results suggest an increasing trend in the amountof silicon in hematite films as the TEOS flow rate is increased.It was concluded from our XPS data that Si is incorporated inthe hematite structure as Si4+. The XPS data also indicate thepresence of Fe2+ in the doped samples which was regarded asthe indirect confirmation of the existence of Si4+. On the basisof Raman data, we believe that introduction of silicon induceslattice defects in the hematite structure. Also noticed was thata much finer nanostructure was formed as TEOS were intro-duced to the system which may be attributed to the effect ofwater formation at a molecular level during the decompositionof TEOS and the subsequent nucleation of hematite particles.Moreover, the enhancement of the optical bandgap with theincrease of TEOS flow rate appears to be a result of thereduction in the size of hematite nanoparticles and the sizequantization effect. Although our findings did not result in adefinitive and conclusive explanation to the role of siliconincorporation, it can be concluded that there is a combinationof different factors (such as silicon incorporation into the

hematite structure and formation of lattice defects, along witha nanostructure of small agglomerate/cluster enhancing holetransportation to the surface) contributing to the improvementof PEC performance in Si-doped hematite films. It is envisagedthat a better understanding of the decomposition process of Si-containing agents as well as Si-free compounds, particularly,the effect of decomposition byproduct, will be helpful inachieving the desired fine hematite nanostructured electrodesand better PEC performance. Further work is in progress to studythe decomposition pattern of precursors. Additionally, the effectsof Si additives as well as codopants on fundamental physicaland electrical properties of hematite electrodes need to beinvestigated.

Acknowledgment. Authors would like to thank MaterialsResearch School at Loughborough University for financialsupport. Authors also acknowledge Prof. Jon Binner, Directorof Department of Materials, for his continued support. We wouldalso like to thank Prof. Stephen Fletcher for helpful discussion.The assistance given by Mr. Dave Wilson, Mr. Dhinesh Asogan(Department of Chemistry), and Mr. John Bates and Dr. GeoffWest (Department of Materials) is kindly acknowledged.

References and Notes

(1) Ohi, J. J. Mater. Res. 2005, 20, 3180–3187.(2) Wijayantha, K. G. U.; Auty, D. H. In Encyclopaedia of Materials:

Science and Technology Updates; Elsevier: Oxford, 2005; pp 15.(3) Armor, J. N. Catal. Lett. 2005, 101, 131–135.(4) Fujishima, A.; Honda, K. Nature 1972, 238, 37–38.(5) Khan, S. U.; Al-Shahry, M; Ingler, W. B., Jr. Science 2002, 297,

2243–2245.(6) Wang, H.; Lindgren, T.; He, J.; Hagfeldt, A.; Lindquist, S.-E. J.

Phys. Chem. B 2000, 104, 5686–5696.(7) Khan, S. U. M.; Akikusa, J. J. Phys. Chem. B 1999, 103, 7184–

7189.(8) Duret, A.; Gratzel, M. J. Phys. Chem. B 2005, 109, 17184–17191.(9) Sartoretti, C. J.; Ulmann, M.; Alexander, B. D.; Augustynski, J.;

Weidenkaff, A. Chem. Phys. Lett. 2003, 376, 194–200.(10) Kennedy, J. H.; Frese, K. W. J. Electrochem. Soc. 1978, 125, 709–

714.(11) Aroutiounian, V. M.; Arakelyan, V. M.; Shahnazaryan, G. E.;

Hovhannisyan, H. R.; Wang, H.; Turner, J. A. Solar Energy 2007, 81, 1369–1376.

(12) Aroutiounian, V M.; Arakelyan, V. M.; Shahnazaryan, G E.;Stepanyan, G. M.; Turner, J. A.; Khaselev, O. Int. J. Hydrogen Energy2002, 27, 33–38.

(13) Ingler, W. B., Jr.; Khan, S. U. M Thin Solid Films 2004, 461, 301–308.

(14) Parent, L.; Dodelet, J. P.; Dallaire, S. J. Electrochem. Soc. 1987,134, 2226–2233.

(15) Ingler, W. B., Jr.; Baltrus, J. P.; Khan, S. U. M. J. Am. Chem. Soc.2004, 126, 10238–10239.

(16) Kay, A.; Cesar, I.; Gratzel, M. J. Am. Chem. Soc. 2006, 128, 15714–17721.

(17) Cesar, I.; Kay, A.; Gonzalez Martinez, J. A.; Gratzel, M. J. Am.Chem. Soc. 2006, 128, 4582–4583.

(18) Glasscock, J. A.; Barnes, P. R. F.; Plumb, I. C.; Savvides, N. J.Phys. Chem. C 2007, 111, 16477–16488.

(19) Liang, Y.; Enache, C. S.; Van der Krol, R. Int. J Photoenergy 2008,Article Number: 739864.

(20) Lide, D. R.; Frederikse, H. P. R. CRC Handbook of Chemistry andPhysics, 75th ed.; CRC Press, Inc.: Boca Raton, 1995; pp 12-53.

(21) Campbell, A. S.; Schwertmann, U.; Stanjek, H.; Friedl, J.; Kyek,A.; Campbell, P. A. Langmuir 2002, 18, 7804–7809.

(22) Desai, J. D.; Pathon, H. M.; Min, S.-K.; Jung, K.-D.; Joo, O.-S.Semicond. Sci. Technol. 2005, 20, 705–709.

(23) Desai, J. D.; Pathon, H. M.; Min, S.-K.; Jung, K.-D.; Joo, O.-S.Appl. Surf. Sci. 2006, 252, 2251–2258.

(24) NIST XPS database: http://srdata.nist.gov/XPS (Accessed July 20,2008).

(25) Shinar, R.; Kennedy, J. H. Solar Energy Mater. 1982, 6, 323–335.(26) Morin, F. J. Phys. ReV. 1951, 83, 1005–1010.(27) Berry, F. J.; Greaves, C.; McManus, J. G.; Mortimer, M.; Oates,

G. J. Solid State Chem. 1997, 130, 272–276.(28) Berry, F. J.; Greaves, C.; Helgason, O.; McManus, J. G.; Palmer,

H. M.; Williams, R. T. J. Solid State Chem. 2000, 151, 157–162.

1C

) ( 1Csc

+ 1Cdl

) ≈ 1Csc

(3)

Nanostructured R-Fe2O3 Electrodes J. Phys. Chem. C, Vol. 113, No. 12, 2009 4777

(29) Berry, F. J.; Bohorquez, A.; Moor, E. A. J. Phys. Chem. Solids2001, 62, 1277–1284.

(30) Kuivila, C. S.; Butt, J. B.; Stair, P. C. Appl. Surf. Sci. 1988, 32,99–121.

(31) Grosvenor, A. P.; Kobe, B. A.; Biesinger, M. C.; McIntyre, N. S.Surf. Interface Anal. 2004, 36, 1564–1574.

(32) Yamashita, T.; Hayes, P. Appl. Surf. Sci. 2008, 254, 2441–2449.(33) Zoppi, A.; Lofrumento, C.; Castellucci, E. M. ; Sciau Ph. J. Raman

Spectrosc. 2008, 39, 40–46.(34) Sartoretti, C. J.; Alexander, B. D.; Solarska, R.; Rutkowska, I. A.;

Augustynski, J. J. Phys. Chem. B 2005, 109, 13685–13692.(35) Yu, K. N.; Xiong, Y.; Liu, Y.; Xiong, C. Phys. ReV. B 1997, 55,

2666–2671.(36) McCarty, K. F. Solid State Commun. 1988, 68, 799–802.(37) Bersani, D.; Lottici, P. P.; Montenero, A. J. Raman Spectrosc. 1999,

30, 355–360.

(38) Chu, J. C. S.; Breslin, J.; Wang, N. S.; Lin, M. C. Mater. Lett.1991, 12, 179–184.

(39) Shimojo, M.; Takeguchi, M.; Mitsuishi, K.; Tanaka, M.; Furuya,K. Jpn. J. Appl. Phys. 2007, 46, 6247–6249.

(40) Takeguchi, M.; Shimojo, M.; Furuya, K. Jpn. J. Appl. Phys. 2007,46, 6183–6186.

(41) Hagfeldt, A.; Bjorksten, U.; Lindquist, S.-E. Sol. Energy Mater.Sol. Cells 1992, 27, 293–304.

(42) Vayssieres, L.; Sathe, C.; Butorin, S. M.; Shuh, D. K.; Nordgren,J.; Guo, J. AdV. Mater. 2005, 17, 2320–2323.

(43) Pal, B.; Sharon, M. J. Chem. Technol. Biotechnol. 1998, 73, 269–273.

(44) Miller, E. L.; Paluselli, D.; Marsen, B.; Rocheleau, R. E. Thin SolidFilms 2004, 466, 307–313.

JP808453Z

4778 J. Phys. Chem. C, Vol. 113, No. 12, 2009 Saremi-Yarahmadi et al.