Journal ofMaterials Chemistry A

PAPER

Publ

ishe

d on

30

Janu

ary

2017

. Dow

nloa

ded

by U

nive

rsity

of

Sout

h C

arol

ina

Lib

rari

es o

n 10

/02/

2017

13:

50:5

8.

View Article OnlineView Journal

Surface function

Department of Chemistry and Biochemistry,

SC 29201, USA. E-mail: stek@mailbox.sc.e

† Electronic supplementary informationcharacterization and PEC performance da

Cite this: DOI: 10.1039/c6ta09485f

Received 2nd November 2016Accepted 30th January 2017

DOI: 10.1039/c6ta09485f

rsc.li/materials-a

This journal is © The Royal Society of

alized atomic layer deposition ofbismuth vanadate for single-phase scheelite†

B. Lamm, A. Sarkar and M. Stefik*

Monoclinic bismuth vanadate is one of the most promising oxide photoanodes for solar – assisted water

splitting. To date, the atomic layer deposition of bismuth vanadates has relied on the catalytic

codeposition of BiPh3 with VTIP to produce vanadium-rich compounds that undergo spinodal

decomposition to multiphase mixtures upon crystallization. A surface functionalization (SF) step of

ROH/VTIP/H2O was developed to inhibit V2O5 deposition for adjustable Bi : V stoichiometry. Ethanol,

2-propanol, and methanol were each found to inhibit V2O5 deposition, in order of increasing effect.

Applying this SF step with ternary Bi–V–O depositions (ROH/VTIP/H2O/BiPh3/H2O) enabled composition

tuning. The use of methanol enabled 45.9 : 54.1 Bi : V atomic ratio as-deposited, and was crystallizable

to phase-pure scheelite, depending on the thickness. The resulting films were applied towards photo-

assisted water splitting with a hole-scavenging sulfite where films up to 60 nm thick were free from

apparent charge transport limitations. The optoelectronic properties were markedly improved by a novel

photoelectrochemical activation step.

Introduction and background

Solar photoelectrochemistry (PEC) is an attractive source forsustainable fuels such as hydrogen.1 Bismuth vanadate was rstdemonstrated as a photocatalyst for water oxidation in 1998,with the monoclinic-scheelite phase (m-BiVO4) showing thehighest efficiency.2,3 The m-BiVO4 phase has a band gap of2.4 eV,3,4 a conduction band edge near 0 V vs. the reversiblehydrogen electrode (RHE)5 and a valence band edge near 2.4 Vvs. RHE. Thus electrons in the conduction band have appro-priate potential to support the hydrogen evolution reaction (0 Vvs. RHE) at the counter electrode with a small applied biasand holes in the valence band have ample potential to supportthe oxygen evolution reaction (1.23 V vs. RHE) at the semi-conductor–electrolyte interface. Electrons and holes generatedby photoabsorption are thus able to support photoassistedwater splitting with a low onset potential (ca. 0.25 V vs. RHE) forthe photoanode. The performance of BiVO4 photoanodes hasprogressed quickly with many recent advances.4–9

For solar-fuel applications, bismuth vanadate requires anoptical thickness of �700 nm for efficient light harvesting nearthe band edge (discussed in ESI†) that is much larger than thesum of the depletion width and the limiting carrier diffusionlength.5,10 Thus, many recent devices with record photocurrentsused BiVO4 thin lms supported upon transparent, conductive

University of South Carolina, Columbia,

du

(ESI) available: Additional physicalta. See DOI: 10.1039/c6ta09485f

Chemistry 2017

scaffolds. Such “host–guest” or extremely thin absorberapproaches have been largely successful at decoupling opticalabsorption from carrier transport.11–13 Thus far, all BiVO4 host–guest reports have utilized non-uniform depositions or usedcathodic depositions that limited the use of hole blocking layersat the host–guest interface.7,8,14,15 In contrast, atomic layerdeposition (ALD) enables controlled growth onto arbitrary 3Dporous scaffolds independent of substrate electronic properties,bringing distinct advantages to the development of efficienthost–guest nanostructures. Such structures have been devel-oped for optoelectronics including a-Fe2O3,16–35 Cu2O,36–39 andother photoelectrodes.40–42

Thus far, all reports of ALD bismuth vanadates had >65at% V, leading to spinodal decomposition into BiVO4 with anadditional V2O5 phase.43 With conventional ALD, the amount ofmaterial deposited per cycle is xed due to a self-limitingreaction of each precursor at surface saturation. The deposi-tion rate is a constant for each precursor/oxidant combinationunder steady state.44 While this is a signicant benet forconformal thin-lm growth, the quantized nature of depositionlimits stoichiometry tuning. Surface-functionalized atomiclayer deposition (SF-ALD) enables further stoichiometry tuningby modifying the amount of material deposited per cycle. SF-ALD can thus reduce the amount of material deposited bypartially blocking deposition sites. For example, alcohols canundergo a condensation reaction with surface hydroxyl groupsto inhibit the deposition of the subsequent metal precursor.44–46

The subsequent oxidant pulse removes the remaining ligandsfrom themetal as well as the SF-species. Since the SF step is self-limiting, the overall deposition remains conformal and self-

J. Mater. Chem. A

Journal of Materials Chemistry A Paper

Publ

ishe

d on

30

Janu

ary

2017

. Dow

nloa

ded

by U

nive

rsity

of

Sout

h C

arol

ina

Lib

rari

es o

n 10

/02/

2017

13:

50:5

8.

View Article Online

limiting, but with more granular control over the amountdeposited. Although SF-ALD was developed to modify the extentof doping,44–46 it is also conceptually suitable for tuning thestoichiometry of multi-metal compounds.

Here, SF-ALD with alkyl alcohols (ROH) enables the most Bi-rich depositions of bismuth vanadate to date, including nearlystoichiometric BiVO4. Crystallization of the amorphous lmsled to the phase-pure photoactive form of m-BiVO4, dependingon lm thickness. In addition, a PEC activation treatment wasfound to signicantly improve PEC efficiency.

ExperimentalMaterials

Triphenyl bismuth (BiPh3, 99%) and vanadium(V) triisopropoxyoxide (VTIP, 98+%) were used as received from STREM. Deionized(DI) water was prepared using a Siemens Labostar model; wherespecied, deionized ultra-ltered (DIUF) water (Fisher) was usedin place of DI water. Methanol (ACS Grade, Fisher), 2-propanol(70% lab grade, BDH), ethanol (95%, Fisher), H3BO3 (ACS Grade,VWR Life Science), HNO3 (Fisher, Optima grade), Na2SO3 (ACSGrade, Macron), and KOH (ACS Grade, Fisher) were used as-received. TEC-15 uorine-doped tin oxide coated glass (FTO) waspurchased from Hartford Glass. The FTO substrates were cleanedextensively before use with 2-propanol and DI water before soni-cation in soapy water (Decon Contrex, 2 wt%), followed by addi-tional rinses with water and 2-propanol, followed by sonication in2-propanol. Polished n-doped silicon wafers with (100) orientationwere purchased from University Wafers, USA. The cleaned FTOsubstrates and silicon wafers were calcined using a BarnsteadThermolyne muffle furnace at 450 �C for 1 h immediately prior touse. High-temperature grade Kapton tape was purchased fromMcMaster-Carr, USA. Ultra-high purity nitrogen (99.999%) andoxygen (99.5%) were used as received from Praxair.

Atomic layer deposition of bismuth vanadate lms

Samples were masked using Kapton tape to dene the depositionregion. Samples for composition analysis were masked on thebottom to avoid diffusion gradients along the back of the lms.Samples on FTO were masked on the back and partially on thefront to provide clean electrical contacts for PEC measurements.The BiPh3 and VTIP were each loaded into separate stainless steelcylinders in an argon glovebox. Water and the alcohol in use(ROH; ethanol, 2-propanol, or methanol) were loaded into sepa-rate stainless steel cylinders under normal atmosphere. Thecylinders were sealed and connected to and Arradiance Gemstar-8reactor. Nitrogen was used as both the carrier gas and the purginggas. Precursor dosing was controlled using Swagelok ALD valves.The BiPh3 and VTIP cylinders were heated to 130 and 45 �C,respectively. The reactor was set to 130 �C, both metal precursorshad a pulse time of 2 s, ROH had a pulse time of 25 ms, and waterhad a pulse time of 25 ms. A vapor-boosting 20 ms pulse ofnitrogen was added to the BiPh3 cylinder just prior to each pulse.The reactor chamber was isolated before each pulse to contain theprecursors for 1 s aer exposure (“exposure mode”). The excessprecursors were purged aer each exposure using 200 sccm

J. Mater. Chem. A

nitrogen for 10 s. The deposition was organized into a macrocycleof (ROH/VTIP/water/BiPh3/water)a where ROHwas used to controlcomposition and ‘a’ controlled total thickness deposited. Acomprehensive ALD protocol is provided in the ESI (Table S2†).

SF-ALD sample PEC performance was compared to theperformance of previously reported samples synthesized viaconventional ALD (Fig. S4†).43 The reported ALD procedure isidentical to the SF-ALD procedure described above, sans alcoholsurface functionalization. Conventional ALD samples of55.6 nm thickness were calcined at 450 �C for 1 h and etched in1 M NaOH to remove excess V2O5. SF-ALD samples (60 nm-8k,described in the text) were measured following the standardactivation treatment, vide infra.

In situ deposition monitoring

Quartz crystal microbalances (QCMs) assembled in an arraywere used tomonitor ALD depositions in situ across the reactionchamber. It is important to note that while data were collectedin real time, certain variables – such as precursor pulsing,varying ow rates, and proportional–integral–derivative (PID)controller cycles – caused thermal transients that obfuscatedthe very small signals for mass change during individual pulses.Accordingly, average deposition rates were determined usinga least squares t analysis for linear deposition rates. A total of500–2000 (SF)-ALD cycles were performed for each experimentto provide accurate measurements of growth per cycle (GPC).The QCM crystals were installed in a low-prole plate inside thedeposition chamber and electrical connections were fedthrough an auxiliary port. The QCM array (custom design,Colnatec) utilized a special sealing interface to eliminate theneed for inert gas purging on the back side of the crystals.Radiation-compensated (RC) cut quartz crystals were used withaluminum contacts (Colnatec). The native oxide on the Al topcontact of the crystals led to the facile establishment of linearQCM response, as shown elsewhere.43 QCM deposition datawere calibrated by SEM observations of lm thickness.

Composition analysis

Samples were digested with 4 M nitric acid (�5 mL) in Teonvessels at 180 �C for 5 h. The resulting solution was diluted to 50mL using DI water. A Finnigan ELEMENT XR double-focusingmagnetic sector eld-inductively coupled plasma-mass spec-trometer (SF-ICP-MS) was used for analyzing V (51, LR), Bi (209,LR), and internal standard Rh (103 LR). A Micromist U-seriesnebulizer (GE, Australia) was operated at 0.2 mL min�1 witha quartz torch and an injector (Thermo Fisher Scientic, USA)for sample introduction. The gas ow was set to 1.08 mL min�1.The forwarding power was 1250 W. Composition analysis wasbased on a ve-point calibration curve for both V and Bi. Thecalibration range was from 10 to 600 ppb. The R2 values for theinitial calibration curves were greater than 0.999.

Film treatments

The ALDlmswere heated to induce crystallization. Samples wereheated at 5 �Cmin�1 to 200 �C, followed by 10 �Cmin�1 to 450 �C,held constant at 450 �C for 1 h, and allowed to cool in the furnace.

This journal is © The Royal Society of Chemistry 2017

Paper Journal of Materials Chemistry A

Publ

ishe

d on

30

Janu

ary

2017

. Dow

nloa

ded

by U

nive

rsity

of

Sout

h C

arol

ina

Lib

rari

es o

n 10

/02/

2017

13:

50:5

8.

View Article Online

Diffraction

X-ray diffraction experiments were conducted using a SAXSLabGanesha at the South Carolina SAXS Collaborative. A XenocsGeniX 3D microfocus source was used with a Cu target togenerate a monochromatic beam with a 0.154 nm wavelength.The instrument was calibrated using silicon powder (NIST640e). Scattering data were processed from the scattering vectorq ¼ 4pl�1 sin q where l is the X-ray wavelength and 2q is thetotal scattering angle. A Pilatus 300 K detector (Dectris) wasused to collect the two-dimensional (2D) scattering patterns.Samples were measured at an incident angle of 8� relative to thelm plane. SAXSGUI soware was used to radially integrate the2D patterns to reduced 1D proles. The conversion of theresulting intensity versus q data was converted to 2q using theabove formula.

Photoelectrochemical and electrochemical measurements

Linear sweep voltammograms were measured using a three-electrode potentiostat (BioLogic SP-150) with a Ag/AgCl/KCl(saturated) reference electrode (Pine Instruments) and a plat-inum wire counter electrode (Pine Instruments). Samples onFTO substrates were clamped with a titanium sheet to providean ohmic contact. The electrodes were placed into a cell made ofpolyether ether ketone (PEEK) with a fused-silica window.Simulated sunlight was generated using a 75 W xenon lamp(OBB, Horiba) that passed through a water infrared lter (OBB,Horiba), a KG-3 lter (317–710 nm pass, Edmund Optics), anda BG-40 lter with an antireective coating (335–610 nm pass,Thorlabs). This combination of lters removed much of the UVlight where the Xe lamp has the most spectral mismatch fromthe AM 1.5 spectrum. The transmitted light was collimatedusing a fused-silica lens (Thorlabs) and passed through anengineered diffuser with a top–hot prole to provide a homo-geneous intensity prole with a slight 10� divergence. Thetransmitted light was corrected for brightness in the 335–610 nm spectral range to generate a photocurrent identicalto AM 1.5 sunlight. The illumination intensity was measuredusing a calibrated UV-enhanced silicon photodiode (Thorlabs)equipped with a neutral reective lter (optical density 1.0,Thorlabs) to maintain a linear and calibrated photodioderesponse. This calibration practice provides accurate solarsimulation in terms of both spectral distribution and bright-ness with a minimal correction factor.47

PECmeasurements were performed in 1 M potassium boratewith or without 0.2 M sodium sulte (Na2SO3) as hole scavengerat pH 9.35. It has been well established that potassium phos-phate buffers dissolve BiVO4 at working pH values;9,48,49

however, BiVO4 photoanodes are stable in alkaline boratebuffers. The potassium borate solution was prepared byadjusting the pH of 1MH3BO3 in DIUF water to 9.35� 0.02 withKOH as conrmed by a calibrated Thermo Scientic OrionStarA211 pH meter. The sulte acted as a hole scavenger to providequantitative charge injection from the semiconductor to theelectrolyte for the measurement of lm performance withoutcatalysts. The samples were scanned from �0.600 to 0.650 V vs.Ag/AgCl reference electrode at 10 mV s�1. Multiple scans were

This journal is © The Royal Society of Chemistry 2017

completed at each condition to conrm reproducibility and thesecond scan results were reported. Film stability was measuredby chronoamperometry (CA) at 0.6 V vs. RHE under simulatedAM 1.5 illumination to measure the photocurrent stability of SFALD 30 nm-4k lm over 17 h. The lm was freshly calcined priorto the stability measurement.

Quantum efficiencies were calculated based on CA measure-ments made with monochromatic light while using the samepotassium borate buffer described above. Illumination wasgenerated using a 150 W xenon lamp (OBB, Horiba) that passedthrough an 180� monochromator with a 5 mm slit width and1200 grates per mm diffraction grating (OBB, Horiba). Trans-mitted light was collimated using a fused-silica lens (Thorlabs)and passed through an engineered diffusor with a top-hat proleto provide a homogenous intensity prole with a slight 10�

divergence. CA measurements were recorded at �0.153 V vs. Ag/AgCl reference unless otherwise noted. All electrochemicalpotentials E were reported versus the reversible hydrogen elec-trode (RHE) using the formula E (vs. RHE) ¼ E (vs. Ag/AgCl) +Eref(Ag/AgCl) + 0.059 V � pH where Eref ¼ 0.197 V in this case.

For both CA and LSV measurements, representative samplesof a given set were plotted with the mean (indicated as ‘x’) andbars for the error of the mean.

Prior to LSV and CAmeasurements, lms were exposed to anexternal bias of 0.6 V vs. RHE (�0.153 V vs. Ag/AgCl) under AM1.5 illumination for 1 hour. To compare different aspects of lmactivation (Fig. 8), freshly calcined samples were measured forUV-vis absorptance and CA data (Control). Following this, thesesame samples were either allowed to soak in potassium boratesolution for 1 h (Soak), applied 0.6 V vs. RHE for 1 h in dark (EC-Only), exposed to AM 1.5 solar simulator for 1 h (Photo-Only), ortreated with the standard PEC activation treatment (PEC) beforemeasuring LSV and CA once more (except in the case of Soak,where only CA was measured). Additionally, another set offreshly calcined samples were soaked in electrolyte for 1 h fol-lowed by LSV and CA (LSV-Only, Fig. S4†) to observe the effectsof LSV measurement, which is a combination of photo- andelectrochemical treatments. It is important to note that for thisseries of measurements, samples all came from the same 4000-cycle SF-ALD lm; additionally, for each test condition sampleswere selected from a variety of locations within the lm to avoidlocal thickness or crystal quality variance within the lm.

Optoelectronic properties

The optical response of thin lms was measured using a Shi-madzu UV-2450. A sandwich conguration of FTO-water-fusedquartz was used to minimize light scattering differencesbetween the blank measurement of bare FTO and samplescoated onto FTO. Identical measurements on FTO were used toestablish the baseline for the measurement of the opticalproperties of the deposited lms alone.

X-ray photoelectron spectroscopy

A Kratos Axis UltraDLD instrument equipped with a mono-chromated copper Ka X-ray source (wavelength of 0.154056nm�1, operating at 10 kV and 15 mA). A hybrid lens mode was

J. Mater. Chem. A

Table 1 Growth rates of V2O5 with different alcohols used for surfacefunctionalization

Samplename

Depositioncycle

Growth rate(A per cycle)

ALD-V2O5 [VTIP-W]x 0.22

Journal of Materials Chemistry A Paper

Publ

ishe

d on

30

Janu

ary

2017

. Dow

nloa

ded

by U

nive

rsity

of

Sout

h C

arol

ina

Lib

rari

es o

n 10

/02/

2017

13:

50:5

8.

View Article Online

employed during analysis, with an analysis area of approxi-mately 700 mm � 300 mm. The XPS spectra for all samples weretaken at a photoemission angle of 0�, relative to the surfacenormal. A Kratos charge neutralizer system was used on allsamples with a lament current of between 1.8–2.1 A anda charge balance of 3.6 V.

EtOH-V2O5 [EtOH-VTIP-W]x 0.14iPrOH-V2O5 [iPrOH-VTIP-W]x 0.08MeOH-V2O5 [MeOH-VTIP-W]x 0.06

Morphology

A Zeiss Ultra Plus scanning electron microscope (SEM) wasoperated at 5 kV using an in-lens secondary electron detector toobserve the lm surface and cross-sectional acquisition. ALDgrowth rates were calculated based upon cross-sectional SEMimaging of lms in the 50–100 nm thickness range.

Results and discussionEffect of alcohol on surface functionalization

The ALD growth of bismuth vanadates (BVO) has not yetenabled phase pure scheelite due to limited stoichiometrycontrol. We briey note that bismuth vanadates are a class ofcompounds with various Bi : V ratios. The prior report of ALDBVO utilized triphenyl bismuth (BiPh3), vanadium triisopropoxyoxide (VTIP), and water. To the best of our knowledge, thisremains the only prior reported ALD of BVO. Here, the lmstoichiometry was >65 at% V-rich and required a post-treatmentetch of V2O5 to achieve phase-pure m-BiVO4 lms.43 Increasingthe pulse ratio to favor Bi had limited benets since the Bi-deposition is catalytically dependent on V–OH species.43 SF-ALD provides a novel and time-efficient alternative to improvethe Bi : V stoichiometry by reducing the amount of VTIPdeposited in each cycle. Starting from a prior ALD protocol forV2O5,43,50 a pulse of different alcohols were used to inhibit thesubsequent VTIP deposition with a [ROH-VTIP-W]x depositioncycle. The specic alcohols investigated included ethanol, 2-propanol, and methanol (EtOH, iPrOH, and MeOH, respec-tively). Starting from a V2O5 growth-per cycle (GPC) of 0.22 A percycle, the alcohols were found to have increasing inhibitioneffect in order of EtOH, iPrOH, and MeOH (Fig. 1 and Table 1).The alcohol pulse length was found to saturate growth inhibi-tion well within 25 ms (Fig. S1†). MeOH had the largest effect onVTIP deposition with the lowest GPC. This deviates from priorreported SF-ALD trends with these alcohols applied towardsother materials.44,45 Many aspects determine the efficacy of

Fig. 1 (SF)-ALD growth rates of V2O5 using different alcohols forsurface functionalization with deposition cycles of [ROH-VTIP-W]x.

J. Mater. Chem. A

a particular SF-species, including the density and type ofremaining active sites, side reactions between precursor andinhibitor, and inhibitor displacement by precursor.45

The SF-ALD of BVO was examined with the same series ofalcohols examined above. The SF-ALD deposition sequence was[ROH-VTIP-W-BiPh3-W]x. The use of EtOH, iPrOH, and MeOHfor SF-ALD were found to result in increasingly Bi-rich deposi-tions (Fig. 2 and Table 2), following a consistent trend with GPCfor SF-ALD of V2O5. The samples prepared using MeOH hadstoichiometries closest to the 1 : 1 BiVO4 compound, with 45.9at% Bi and 54.1 at% V determined by mass-spectrometry (MS).For the remainder of this paper we thus focus exclusively onternary BVO produced by SF-ALD with MeOH (MeOH-BVO).Cross-sectional and top-view SEM conrmed conformalcomposition on both Si and FTO substrates (Fig. 4a–d).Measurements from the at Si substrate were used to determinethe growth rate of 7.5 nm for every 1000 cycles of SF-ALD. Thiscorresponds to a GPC of 0.075 A per cycle that is remarkablyquite similar to 0.07 A per cycle from conventional ALD.43 Thecombination of a similar GPC with enhanced Bi contentsuggests that the SF step not only inhibits VTIP deposition, butalso that the resulting surface, aer hydrolysis, enhances thefollowing BiPh3 deposition. Such mechanistic changes could beattributed to a combination of steric effects with the spatialdistribution of active V–OH surface sites for BiPh3 catalyticdeposition.

The crystallization of MeOH-BVO lms was examined bygrazing incidence wide-angle X-ray scattering (GI-WAXS) andSEM aer high temperature calcination. The as-deposited lmswere amorphous, similar to the prior reported bismuth vana-date lms by conventional ALD (Fig. 3).43 Aer heating to 450 �Cfor 1 h, the MeOH-BVO lms exhibited strong reectionsconsistent with nearly phase-pure m-BiVO4. In contrast, the

Fig. 2 Compositions for (SF)-ALD of BiVxOy prepared with differentalcohols for surface functionalization using deposition cycles of [ROH-VTIP-BiPh3-W]x.

This journal is © The Royal Society of Chemistry 2017

Table 2 Composition of (SF)-ALD BVO films with different alcoholsused for surface functionalization

Sample name Deposition cycle at% Bi at% V

ALD-BVO [VTIP-W-BiPh3-W]x 33.2% 66.8%EtOH-BVO [EtOH-VTIP-W-BiPh3-W]x 34.6% 65.4%iPrOH-BVO [iPrOH-VTIP-W-BiPh3-W]x 40.2% 59.8%MeOH-BVO [MeOH-VTIP-W-BiPh3-W]x 45.9% 54.1%

Fig. 3 GI-WAXS of MeOH-BVO films, both as-made and calcined. Forcomparison, a calcined film from ALD-BVO is presented. All sampleshad x ¼ 4000 deposition cycles. Graphs were offset vertically forclarity and fit data correspond to PDF#14-0688, 83-1812, and 89-0611, respectively. The calcination step was to 450 �C at 10 �C min�1

with a 60 min hold.

Fig. 4 SEM of bismuth vanadates prepared by MeOH-BVO on Si (a, c,e) and FTO substrates (b, d, f). As-made films are shown in cross-section (top row, a, b) and top view (middle row, c, d). Crystallized filmsheated to 450 �C are shown in top view (bottom row, e, f). Thedeposition cycle was [MeOH-VTIP-W-BiPh3-W]x where x was either3000 (b, e, f) or 6000 (a, c, d).

Paper Journal of Materials Chemistry A

Publ

ishe

d on

30

Janu

ary

2017

. Dow

nloa

ded

by U

nive

rsity

of

Sout

h C

arol

ina

Lib

rari

es o

n 10

/02/

2017

13:

50:5

8.

View Article Online

ALD-BVO sample exhibited a pattern consistent with a mixtureof m-BiVO4, tetragonal bismuth vanadate (t-BiVO4), and ortho-rhombic vanadium oxide (o-V2O5), consistent with the signi-cant excess of vanadium detected by MS. The MS data alsoindicated a 4.1 at% V excess for the MeOH-BVO samples. Thisslight excess of vanadium did not result in an observable V2O5

phase for lms #30 nm (Fig. S2†). Additional GI-WAXS data forMeOH-BVO lms on FTO substrates matched that of lms on Si,albeit with additional substrate peaks (Fig. S3†). Film calcina-tion also resulted in minor morphology changes from crystalliteformation. Films calcined on FTO substrates remained fairlyuniform whereas lms on Si substrates underwent dewetting(Fig. 4e and f). MeOH-BVO samples enabled the crystallizationof phase-pure and nearly stoichiometric m-BiVO4.

PEC performance vs. thickness

The calcined MeOH-BVO lms were applied towards solar-assisted PEC water splitting with thickness ranging from 7.5to 75 nm. With a GPC of 0.075 A per cycle, this corresponds tosamples made using 1000–10 000 SF-ALD cycles. Here, thethickest samples examined were commensurate to the reportedelectron diffusion length of ca. 70–75 nm in BiVO4.5,8,51,52 ThePEC performance is a product of several terms, including thelight harvesting efficiency (LHE), the charge separation effi-ciency, and the charge injection efficiency (eqn (1)), where JPEC

This journal is © The Royal Society of Chemistry 2017

is the measured photocurrent density, Jabs is the photonabsorption rate expressed as current density (determined fromLHE), fsep is the carrier separation efficiency, and finj is thecharge injection efficiency.5,53–55

JPEC ¼ Jabs � fsep � finj (1)

Here, the PEC performance was measured in the presence ofa hole-scavenging sulte electrolyte to pin the charge injectionefficiency at 100% (finj ¼ 1). Thus, the measured photocurrentsrepresent the voltage dependent product of light harvestingefficiency and charge separation efficiency. For this sampleseries, the photocurrent monotonically increased with samplethickness (Table 3 and Fig. 5), corresponding to the expectedtrend in light harvesting efficiency, as conrmed by UV-vismeasurements (Fig. 6a). The highest photocurrent wasmeasured with sample 75 nm-10k with 0.69 and 1.21 mA cm�2 at0.60 and 1.23 V vs. RHE, respectively. For thicknesses from 15–60 nm the photocurrent performance monotonically increaseswith one exception, vide infra, and notably all exhibited perfor-mance that was independent of the illumination direction. Incontrast, aer 60 nm of thickness there is a statistically signi-cant difference where backside illumination results in morephotocurrent than front side illumination. Since m-BiVO4 hasa lower electron diffusion length (majority carrier) than holediffusion length (minority carrier), one expects better perfor-mance when carriers are generated closer to the electron-accepting FTO contact. This observation shows the transition totransport limited performance occurs between 60 and 75 nm,similar to prior estimations.5,10 The thinnest 7.5 nm lmproduced negligible photocurrent, presumably due to poorcrystallization (Fig. S2a†) or excessive recombination due to the

J. Mater. Chem. A

Table 3 A series of films with different thicknesses were prepared using the MeOH-BVO protocol. All samples were calcined after growth via thedeposition cycle of [MeOH-VTIP-W-BiPh3-W]x with varying values for x

Sample nameFilmthicknessa (nm)

x, number of depositioncycles (k cycles)

Photocurrentb at0.6 V vs. RHE

Photocurrentb at1.23 V vs. RHE

7.5 nm-1k 7.5 1 0.098 � 0.002 0.035 � 0.00515 nm-2k 15 2 0.08 � 0.01 0.22 � 0.0122.5 nm-3k 22.5 3 0.14 � 0.02 0.34 � 0.0530 nm-4k 30 4 0.35 � 0.02 0.70 � 0.0245 nm-6k 45 6 0.30 � 0.02 0.67 � 0.0360 nm-8k 60 8 0.42 � 0.03 0.81 � 0.0375 nm-10k 75 10 0.69 � 0.04 1.21 � 0.04

a Film thickness calculated from the product of the growth-per-cycle with the number of cycles (GPC � x). b Mean back-side photocurrent(mA cm�2).

Journal of Materials Chemistry A Paper

Publ

ishe

d on

30

Janu

ary

2017

. Dow

nloa

ded

by U

nive

rsity

of

Sout

h C

arol

ina

Lib

rari

es o

n 10

/02/

2017

13:

50:5

8.

View Article Online

FTO proximity.30,56,57 Such a “dead layer” effect has been notedbefore for PEC thin lms, including Fe2O3 and BiVO4.22,30,55–58 Incontrast to prior reports, however, MeOH-BVO yields photoactivelms when >15 nm whereas a different synthesis route required>50 nm.55 A notable curiosity was the similar photocurrentresponse from both 30 nm-4k and 45 nm-6k. This exception tothe trend discussed above was conrmed with multiple samplemeasurements produced from multiple SF-ALD depositions.There is a change in the GI-WAXS patterns between these twosamples where a second phase of o-V2O5 is apparent in >30 nmlm (Fig. S2b†). This suggests a threshold thickness for accom-modation of the excess 4.1 at% V as amorphous V2O5 or withinthe m-BiVO4 lattice. At this transition, the second phase and theassociated interfaces lowered the charge separation efficiency toa similar extent as the increase in the light harvesting efficiency,resulting in comparable performance. From these results, it isconcluded that MeOH-BVO lms <75 nm are viable thin lms forhost–guest applications.

Further PEC insights were gained by examining the wave-length dependent incident photon-to-current efficiency (IPCE,Fig. 6b) and the absorbed photon-to-current efficiency (APCE,Fig. 6c). As expected, IPCE increases with thickness, albeit with

Fig. 6 The (a) absorptance, (b) incident photon-to-current efficiencyand (c) absorbed photon-to-current efficiency are shown forcalcined MeOH-BVO films of various thicknesses. The IPCE andAPCE were measured at 0.6 V vs. RHE in 1.0 M potassium borate with0.2 M Na2SO3 as hole scavenger. The film thickness is specified in thelegend followed by the corresponding number of MeOH-BVOdeposition cycles, x.

Fig. 5 Photocurrent response for calcined MeOH-BVO films asa function of film thickness and applied potential. The electrolyte was1.0 M potassium borate with a pH of 9.36 with 0.2 M Na2SO3 as holescavenger. The dark current (dotted), photocurrent with backsideillumination (dashed) and photocurrent with front side illumination(solid) are presented. The film thickness is specified in the legendfollowed by the corresponding number of MeOH-BVO depositioncycles x.

J. Mater. Chem. A This journal is © The Royal Society of Chemistry 2017

Paper Journal of Materials Chemistry A

Publ

ishe

d on

30

Janu

ary

2017

. Dow

nloa

ded

by U

nive

rsity

of

Sout

h C

arol

ina

Lib

rari

es o

n 10

/02/

2017

13:

50:5

8.

View Article Online

the same performance overlap observed in the J–V data anddiscussed above. This is consistent with the expectation ofimproved LHE with thicker lms. The APCE was calculated bynormalizing the IPCE with the LHE, resulting in the wavelengthdependent fsep for each sample. The monotonic trend of higherAPCE with thicker lms was attributed to lower recombinationrates in thicker lms (due to the distance from the FTO/BiVO4

interface). This observation suggests the future development ofhole-blocking layers to mitigate recombination at the FTOinterface.

To conrm the validity of these measurements, the photo-current at 0.6 V vs. RHE (JJ–V) was compared to the integratedproduct of IPCE with the AM 1.5 spectrum (JIPCE, Table S1†). Thephotocurrents were matched within 15% for sample 75 nm-10k.However, it was apparent that thinner lms exhibited a greaterdifference that was attributed to the varying intensity between thesimulated AM 1.5 spectrum used for J–V and the monochromatorused IPCE measurements. Previous reports indicated a signi-cant dependence of m-BiVO4 quantum efficiencies with illumi-nation intensity.10,51

The PEC performance of SF-ALD BiVO4 electrodes werecompared to BiVO4–V2O5 electrodes prepared by conventionalALD that were subsequently etched to yield a porous lm ofphase-pure scheelite.46 Please note that conventional ALDroutes do not yet exist for phase pure BiVO4 without subsequentetching. Comparing lms of similar thickness, these twosynthesis methods result in similar plateau photocurrents withthe SF-ALD BiVO4 exhibiting an improved onset potential120 mV lower (Fig. S4†). The photocurrent stability of SF-ALDBiVO4 lms was examined under continuous illumination for17 h (Fig. S5†). Aer the initial PEC activation (t ¼ 1 h), thephotocurrent decreased by 11% over this period.

The SF-ALD BiVO4 lms presented here are promising forapplication towards host–guest composites. Prior works havedemonstrated mesoporous lms with photocurrents over 3 mAcm�2 at potentials as low as 0.6 V vs. RHE.5,9 In the past year,photocurrents at the thermodynamic water splitting potentialof 1.23 V vs. RHE of$5 mA cm�2 have been reported with host–guest composites.8,9 These prior works used cathodic electro-deposition to produce BiVO4 lms and thus are not ideallysuited for elaboration with hole blocking materials at the FTO-BiVO4 interface. In contrast, the SF-ALD route presented here iscompatible with arbitrary substrates, independent of theirelectronic properties. We anticipate that future studies with theSF-ALD of BiVO4 onto high surface area conductive substrates(such as nanostructured Sb : SnO2)59 will lead to improvementsin overall performance.

Fig. 7 The effect of PEC activation treatment on (a) absorptance, (b)IPCE, and (c) APCE varied with film thickness. IPCE and APCE weremeasured at 0.6 V vs. RHE in 1.0 M potassium borate with 0.2 MNa2SO3 as hole scavenger.

Effects of PEC activation treatment

A pronounced effect of PEC measurement time was notedfor the performance of the above-discussed lms. In all cases,the photocurrent continuously increased with measurementtime until a plateau. Improvements as large as 380%were foundin some cases. We thus developed a PEC activation treatmentto quickly stabilize lm performance before the abovemeasurements were performed. Recent reports with other

This journal is © The Royal Society of Chemistry 2017

fabrication routes have also indicated that various pretreat-ments improve PEC performance. Pretreatments have been re-ported using controlled reduction and/or oxidation of Vcations,9,60 ultraviolet light,61 and open-circuit exposure to AM1.5 illumination.62 It was suggested that the elimination ofsurface oxygen by producing oxygen vacancies is the source ofimproved PEC performance.9,60–64

A PEC activation treatment was developed to establish stablephotocurrents before performance measurements (described inExperimental). The improved IPCE aer PEC activation treat-ment was surprising since the lms had uniformly reducedoptical absorption (Fig. 7a). The extent of change in perfor-mance was found to vary with lm thickness. For 30 nm lmsthe IPCE and APCE increased by 2.6� and 3.1� with the PECactivation treatment. For 60 nm lms the IPCE and APCE

J. Mater. Chem. A

Table 4 The effectiveness of sample activation pretreatment variedwith film thickness. All samples were calcined after growth from thesameMeOH-BVO deposition cycle: [MeOH-VTIP-W-BiPh3-W]xwherex was either 4k or 8k corresponding respectively to the 30 and 60 nmfilms

Sample name TreatmentFilmthickness

30 nm None. Used directlyfollowing calcination

30 nm

30 nm-activated 1 h at 0.6 V vs. RHEwith AM 1.5 illumination

30 nm

60 nm None. Used directlyfollowing calcination

60 nm

60 nm-activated 1 h at 0.6 V vs. RHEwith AM 1.5 illumination

60 nm

Fig. 8 The IPCE, APCE, and absorptance varied with differentcomponents of the PEC activation treatment. IPCE and APCE weremeasured at 0.6 V vs. RHE in 1.0 M potassium borate with 0.2 MNa2SO3 as hole scavenger.

Journal of Materials Chemistry A Paper

Publ

ishe

d on

30

Janu

ary

2017

. Dow

nloa

ded

by U

nive

rsity

of

Sout

h C

arol

ina

Lib

rari

es o

n 10

/02/

2017

13:

50:5

8.

View Article Online

increased by 3.8� and 4.1� with the PEC activation treatment(Table 4 and Fig. 7). We note that GI-WAXS of calcined MeOH-BVO lms presented above indicated a change from single-phase m-BiVO4 for #30 nm thick lms to a mixture of m-BiVO4 and o-V2O5 for $45 nm thick lms (Fig. S2†). Thesethicker lms with crystalline o-V2O5 exhibited the most changeduring PEC activation treatment, suggesting a dependence onthe nature of the slight V-excess with MeOH-BVO deposition.

To evaluate the efficacy of this multicomponent PEC activa-tion treatment, multiple control experiments were run toisolate the effects of each component of the treatment.Measurements were performed on photo-only treated (PhotoOnly), electrochemical-only treated (EC-Only), and electrolyte-only (Soak) treatments (Table 5, see Experimental for details).As seen in Fig. 8a and b, each aspect of the PEC activationtreatment improves both the IPCE and APCE performance ofthe lms to some degree. The largest effect was found for thecomplete treatment (Activated). All treatment conditions, exceptSoak, reduced the absorptance of the lms (Fig. 8c). Thus theimprovement with Soak may be attributed to surface changessuch as the hydrolytic formation of hydroxyl groups and is notconsistent with dissolution nor signicant internal changes tooxidation states. In contrast, all other components of the PECactivation treatment (Photo-Only and EC-Only) resulted ina concomitant reduction of absorptance with the improved

Table 5 The effect of each component of the PEC activation wereinvestigated separately. All samples are calcined MeOH-BVO filmsprepared with [MeOH-VTIP-W-BiPh3-W]4000 resulting in a thicknessof 30 nm

Sample names Treatment

Control NoneSoak Soaked in electrolyte for 1 h

(open circuit, no illumination)EC-Only Applied 0.6 V vs. RHE for 1 h

(no illumination)Photo-Only AM 1.5 illumination for 1 h

(open circuit)Activated Applied 0.6 V vs. RHE during AM

1.5 illumination for 1 h

J. Mater. Chem. A

performance. Previous reports have indicated that the V oxida-tion state in BiVO4 plays an important role with some hints ofan optimal V4+/V5+ ratio.9,60,62 It was also demonstrated that AM1.5 illumination under open circuit canmarkedly improve somelms where it was suggested that removing surface statesassisted to unpin the Fermi level.62,65 Fermi level pinning fromsurface states has been identied in other photoanode mate-rials, including Fe2O3 and TiO2.66,67 However, the changes inabsorptance observed with Photo-Only, EC-Only, and Activatedare consistent with additional redox changes.

X-ray photoelectron spectroscopy (XPS) was performed on 30nm-4k samples promptly aer calcination as well as aer thePEC activation treatment (Fig. S6;† Calcined and PEC Activated,

This journal is © The Royal Society of Chemistry 2017

Paper Journal of Materials Chemistry A

Publ

ishe

d on

30

Janu

ary

2017

. Dow

nloa

ded

by U

nive

rsity

of

Sout

h C

arol

ina

Lib

rari

es o

n 10

/02/

2017

13:

50:5

8.

View Article Online

respectively). Changes to both the Bi and V oxidation states weredetected at the sample surface as a result of the PEC activationtreatment. Following calcination, lms contained a minorcomponent of reduced Bi0 and V4+ in addition to the expectedBiVO4 states. The PEC activation treatment was found to fullyoxidize these reduced states to Bi3+ and V5+. This oxidationprocess is consistent with the observed oxidation current duringCV cycling in the dark. The ALD of bismuth titanates with BiPh3

and water was previously reported to result in a mixture of Bi3+

and Bi0.68 A recent study also identied switchable oxidationstates of V4+/5+ during electrochemical cycling below 0.5 V vs.RHE.9 In contrast, a recently reported open-circuit illuminationpretreatment identied the reduction of V from the +5 to the +4state with a corresponding increase in performance.62 Theseresults highlight the connection of synthetic route and defectchemistries that can enhance the performance of BiVO4

photoanodes.Neither crystallographic nor surface morphology changes

were observed for any of the treatments (Fig. S7 and S8†). Whilemechanistically intriguing, the detailed study of PEC activationis outside the scope of this SF-ALD study. The developed PECactivation treatment resulted in a marked improvement ofMeOH-BVO lm performance.

Conclusions

The uniform deposition of thin lm m-BiVO4 onto arbitraryhosts is necessary for the development of next-generation host–guest architectures for solar water splitting. The capability todeposit BiVO4 by ALD enables the further development inde-pendent of substrate properties. Phase pure monoclinic-scheelite BiVO4 was obtained by MeOH SF-ALD and the result-ing lms were highly photoactive. Photoactivity was improvedby applying a novel PEC activation treatment. Development ofa layer-by-layer technique for PEC-functional BiVO4 withimproved control over stoichiometry allows for the developmentof efficient multi-layer devices with advanced architectures.

Acknowledgements

We acknowledge the University of South Carolina for startupfunds. This work made use of the South Carolina SAXSCollaborative, supported by the NSF Major Research Instru-mentation program (award #DMR-1428620). We thank Arra-diance for their dedicated ALD support and Colnatec for thedesign of a custom low-prole QCM array. We thank Lijian Heand the Mass Spectrometry Center at USC for assistance withcomposition analysis.

Notes and references

1 M. G. Walter, E. L. Warren, J. R. McKone, S. W. Boettcher,Q. Mi, E. A. Santori and N. S. Lewis, Chem. Rev., 2010, 110,6446–6473.

2 A. Kudo, K. Ueda, H. Kato and I. Mikami, Catal. Lett., 1998,53, 229–230.

This journal is © The Royal Society of Chemistry 2017

3 S. Tokunaga, H. Kato and A. Kudo, Chem. Mater., 2001, 13,4624–4628.

4 F. F. Abdi, L. Han, A. H. M. Smets, M. Zeman, B. Dam andR. van de Krol, Nat. Commun., 2013, 4, 2195–2202.

5 T. W. Kim and K.-S. Choi, Science, 2014, 343, 990–994.6 Z.-F. Huang, L. Pan, J.-J. Zou, X. Zhang and L. Wang,Nanoscale, 2014, 6, 14044–14063.

7 X. Shi, I. Y. Choi, K. Zhang, J. Kwon, D. Y. Kim, J. K. Lee,S. H. Oh, J. K. Kim and J. H. Park, Nat. Commun., 2014, 5,4775–4783.

8 Y. Pihosh, I. Turkevych, K. Mawatari, J. Uemura, Y. Kazoe,S. Kosar, K. Makita, T. Sugaya, T. Matsui, D. Fujita,M. Tosa, M. Kondo and T. Kitamori, Sci. Rep., 2015, 5, 11141.

9 Y. Kuang, Q. Jia, H. Nishiyama, T. Yamada, A. Kudo andK. Domen, Adv. Energy Mater., 2016, 6, 1501645.

10 F. F. Abdi and R. van de Krol, J. Phys. Chem. C, 2012, 116,9398–9404.

11 B. M. Kayes, H. A. Atwater and N. S. Lewis, J. Appl. Phys.,2005, 97, 114302.

12 J. R. Maiolo, B. M. Kayes, M. A. Filler, M. C. Putnam,M. D. Kelzenberg, H. A. Atwater and N. S. Lewis, J. Am.Chem. Soc., 2007, 129, 12346–12347.

13 S. W. Boettcher, J. M. Spurgeon, M. C. Putnam, E. L. Warren,D. B. Turner-Evans, M. D. Kelzenberg, J. R. Maiolo,H. A. Atwater and N. S. Lewis, Science, 2010, 327, 185–187.

14 L. Zhang, E. Reisner and J. J. Baumberg, Energy Environ. Sci.,2014, 7, 1402.

15 J. Su, L. Guo, N. Bao and C. A. Grimes, Nano Lett., 2011, 11,1928–1933.

16 Y. Lin, G. Yuan, S. Sheehan, S. Zhou and D. Wang, EnergyEnviron. Sci., 2011, 4, 4862.

17 Y. Lin, S. Zhou, S. W. Sheehan and D. Wang, J. Am. Chem.Soc., 2011, 133, 2398–2401.

18 Y. Lin, Y. Xu, M. T. Mayer, Z. I. Simpson, G. McMahon,S. Zhou and D. Wang, J. Am. Chem. Soc., 2012, 134, 5508–5511.

19 M. T. Mayer, C. Du and D.Wang, J. Am. Chem. Soc., 2012, 134,12406–12409.

20 T. W. Hamann, Dalton Trans., 2012, 41, 7830.21 B. Klahr, S. Gimenez, F. Fabregat-Santiago, J. Bisquert and

T. W. Hamann, J. Am. Chem. Soc., 2012, 134, 16693–16700.22 O. Zandi, B. M. Klahr and T. W. Hamann, Energy Environ.

Sci., 2013, 6, 634–642.23 K. M. H. Young and T. W. Hamann, Chem. Commun., 2014,

50, 8727.24 O. Zandi, J. A. Beardslee and T. Hamann, J. Phys. Chem. C,

2014, 118, 16494–16503.25 O. Zandi and T. W. Hamann, J. Phys. Chem. Lett., 2014, 5,

1522–1526.26 F. Le Formal, N. Tetreault, M. Cornuz, T. Moehl, M. Gratzel

and K. Sivula, Chem. Sci., 2011, 2, 737–743.27 T. Hisatomi, F. Le Formal, M. Cornuz, J. Brillet, N. Tetreault,

K. Sivula and M. Gratzel, Energy Environ. Sci., 2011, 4, 2512.28 L. Steier, I. Herraiz-Cardona, S. Gimenez, F. Fabregat-

Santiago, J. Bisquert, S. D. Tilley and M. Gratzel, Adv.Funct. Mater., 2014, 24, 7681–7688.

J. Mater. Chem. A

Journal of Materials Chemistry A Paper

Publ

ishe

d on

30

Janu

ary

2017

. Dow

nloa

ded

by U

nive

rsity

of

Sout

h C

arol

ina

Lib

rari

es o

n 10

/02/

2017

13:

50:5

8.

View Article Online

29 L. Steier, J. Luo, M. Schreier, M. T. Mayer, T. Sajavaara andM. Gratzel, ACS Nano, 2015, 9, 11775–11783.

30 T. Hisatomi, H. Dotan, M. Stek, K. Sivula, A. Rothschild,M. Gratzel and N. Mathews, Adv. Mater., 2012, 24, 2699–2702.

31 M. Stek, M. Cornuz, N. Mathews, T. Hisatomi,S. Mhaisalkar and M. Gratzel, Nano Lett., 2012, 12, 5431–5435.

32 S. C. Riha, M. J. DeVries Vermeer, M. J. Pellin, J. T. Hupp andA. B. F. Martinson, ACS Appl. Mater. Interfaces, 2013, 5, 360–367.

33 S. C. Riha, J. M. Racowski, M. P. Lanci, J. A. Klug, A. S. Hockand A. B. F. Martinson, Langmuir, 2013, 29, 3439–3445.

34 J. A. Klug, N. G. Becker, S. C. Riha, A. B. F. Martinson,J. W. Elam, M. J. Pellin and T. Proslier, J. Mater. Chem. A,2013, 1, 11607.

35 Y. Qiu, S.-F. Leung, Q. Zhang, B. Hua, Q. Lin, Z. Wei, K. Tsui,Y. Zhang, S. Yang and Z. Fan, Nano Lett., 2014, 14, 2123–2129.

36 A. Paracchino, V. Laporte, K. Sivula, M. Gratzel andE. Thimsen, Nat. Mater., 2011, 10, 456–461.

37 A. Paracchino, N. Mathews, T. Hisatomi, M. Stek,S. D. Tilley andM. Gratzel, Energy Environ. Sci., 2012, 5, 8673.

38 S. D. Tilley, M. Schreier, J. Azevedo, M. Stek and M. Gr€atzel,Adv. Funct. Mater., 2014, 24, 303–311.

39 J. Azevedo, L. Steier, P. Dias, M. Stek, C. T. Sousa,J. P. Araujo, A. Mendes, M. Graetzel and S. D. Tilley, EnergyEnviron. Sci., 2014, 7, 4044–4052.

40 T. Wang, Z. Luo, C. Li and J. Gong, Chem. Soc. Rev., 2014, 43,7469–7484.

41 S. Hu, M. R. Shaner, J. A. Beardslee, M. Lichterman,B. S. Brunschwig and N. S. Lewis, Science, 2014, 344, 1005–1009.

42 Z.-F. Huang, J. Song, L. Pan, X. Zhang, L. Wang and J.-J. Zou,Adv. Mater., 2015, 27, 5309–5327.

43 M. Stek, ChemSusChem, 2016, 9, 1727–1735.44 A. Yanguas-Gil, K. E. Peterson and J. W. Elam, Chem. Mater.,

2011, 23, 4295–4297.45 A. Yanguas-Gil, J. A. Libera and J. W. Elam, Chem. Mater.,

2013, 25, 4849–4860.46 H. Feng, J. A. Libera, P. C. Stair, J. T. Miller and J. W. Elam,

ACS Catal., 2011, 1, 665–673.47 C. H. Seaman, Sol. Energy, 1982, 29, 291–298.48 F. M. Toma, J. K. Cooper, V. Kunzelmann, M. T. McDowell,

J. Yu, D. M. Larson, N. J. Borys, C. Abelyan, J. W. Beeman,K. M. Yu, J. Yang, L. Chen, M. R. Shaner, J. Spurgeon,

J. Mater. Chem. A

F. A. Houle, K. A. Persson and I. D. Sharp, Nat. Commun.,2016, 7, 12012.

49 T. W. Kim and K.-S. Choi, J. Phys. Chem. Lett., 2016, 7, 447–451.

50 J. C. Badot, S. Ribes, E. B. Yous, V. Vivier, J. P. Pereira-Ramos, N. Baffier and D. Lincot, Electrochem. Solid-StateLett., 1999, 3, 485.

51 F. F. Abdi, T. J. Savenije, M. M. May, B. Dam and R. van deKrol, J. Phys. Chem. Lett., 2013, 4, 2752–2757.

52 R. A. Pala, A. J. Leenheer, M. Lichterman, H. A. Atwater andN. S. Lewis, Energy Environ. Sci., 2014, 7, 3424–3430.

53 H. Dotan, K. Sivula, M. Gratzel, A. Rothschild andS. C. Warren, Energy Environ. Sci., 2011, 4, 958–964.

54 D. K. Zhong, S. Choi and D. R. Gamelin, J. Am. Chem. Soc.,2011, 133, 18370–18377.

55 F. F. Abdi, N. Firet and R. van de Krol, ChemCatChem, 2013,5, 490–496.

56 I. Cesar, K. Sivula, A. Kay, R. Zboril and M. Gr€atzel, J. Phys.Chem. C, 2009, 113, 772–782.

57 K. Sivula, F. Le Formal and M. Gr€atzel, Chem. Mater., 2009,21, 2862–2867.

58 K. Itoh and J. O. Bockris, J. Electrochem. Soc., 1984, 131,126612–126671.

59 K. Peters, H. N. Lokupitiya, D. Sarauli, M. Labs, M. Pribil,J. Rathousky, A. Kuhn, D. Leister, M. Stek andD. Fattakhova-Rohlng, Adv. Funct. Mater., 2016, 26, 6682–6692.

60 M. D. Rossell, P. Agrawal, A. Borgschulte, C. Hebert,D. Passerone and R. Erni, Chem. Mater., 2015, 27, 3593–3600.

61 T. Li, J. He, B. Pena and C. P. Berlinguette, Angew. Chem., Int.Ed., 2016, 55, 1769–1772.

62 B. J. Trzesniewski andW. A. Smith, J. Mater. Chem. A, 2016, 4,2919–2926.

63 G. Wang, Y. Ling, X. Lu, F. Qian, Y. Tong, J. Z. Zhang,V. Lordi, C. Rocha Leao and Y. Li, J. Phys. Chem. C, 2013,117, 10957–10964.

64 H. Jiang, H. Dai, X. Meng, K. Ji, L. Zhang and J. Deng, Appl.Catal., B, 2011, 105, 326–334.

65 A. J. Bard, A. B. Bocarsly, F. R. F. Fan, E. G. Walton andM. S. Wrighton, J. Am. Chem. Soc., 1980, 102, 3671–3677.

66 C. Du, M. Zhang, J. Jang, Y. Liu, G. Liu and D. Wang, J. Phys.Chem. C, 2014, 118, 17054–17059.

67 B. Klahr and T. Hamann, J. Phys. Chem. C, 2014, 118, 10393–10399.

68 M. Schuisky, K. Kukli, M. Ritala, A. Harsta and M. Leskela,Chem. Vap. Deposition, 2000, 6, 139–145.

This journal is © The Royal Society of Chemistry 2017