ARTICLESPUBLISHED: 27 MARCH 2017 | VOLUME: 2 | ARTICLE NUMBER: 17045

Enhanced photoelectrochemical e�ciency andstability using a conformal TiO2 film on a blacksilicon photoanodeYanhao Yu1†, Zheng Zhang2†, Xin Yin1, Alexander Kvit3, Qingliang Liao2, Zhuo Kang2, Xiaoqin Yan2,Yue Zhang2,4* and XudongWang1*

Black silicon (b-Si) is a surface-nanostructured Si with extremely e�cient light absorption capability and is therefore ofinterest for solar energy conversion. However, intense charge recombination and low electrochemical stability limit theuse of b-Si in photoelectrochemical solar-fuel production. Here we report that a conformal, ultrathin, amorphous TiO2 filmdeposited by low-temperature atomic layer deposition (ALD) on top of b-Si can simultaneously address both of these issues.Combined with a Co(OH)2 thin film as the oxygen evolution catalyst, this b-Si/TiO2/Co(OH)2 heterostructured photoanodewas able to produce a saturated photocurrent density of 32.3mAcm−2 at an external potential of 1.48V versus reversiblereference electrode (RHE) in 1M NaOH electrolyte. The enhanced photocurrent relative to planar Si and unprotected b-Siphotoelectrodes was attributed to the enhanced charge separation e�ciency as a result of the e�ective passivation ofdefective sites on the b-Si surface. The 8-nm ALD TiO2 layer extends the operational lifetime of b-Si from less than halfan hour to four hours.

Converting sunlight to electricity, heat or chemical fuels is amajor pathway for renewable energy utilization1. Amongthem, hydrogen production from photoelectrochemical

(PEC) water splitting could potentially ease the dependence onfossil fuel and contribute to resolving the global warming issue2–6.To fulfil these ultimate goals, a PEC system needs to be efficient,robust and cost-effective, which strictly limits the selection ofelectrode materials. Semiconductor oxides (for example, TiO2,Fe2O3 and BiVO4) are generally low cost and fairly stable forcatalysing electrochemical reactions7–9. However, the low efficiencydue to the narrow light absorption band and poor charge transportproperty exclude them from practical hydrogen production. Onthe other hand, silicon (Si) is a feasible commercial PEC material,considering its optimum band structures, excellent crystallinityand high industrial maturity. Compared to planar Si wafers, Si witha porous surface (also known as black or nanostructured Si) couldconcurrently minimize light reflection by trapping light insidethe surface channels and increasing the active area for interfacialchemical reactions. These merits encouraged attempts to use blacksilicon (b-Si)-based materials for solar-fuel production, such ascarbon dioxide reduction and water splitting10–14. Nevertheless,the enlarged surface area also induces undesirable chargerecombination and accelerates surface corrosion/passivation duringthe electrochemical reactions15,16. These shortcomings severelyjeopardize the efficiency and durability of b-Si photoelectrodes,constraining contemporary efforts at a very early stage of proofof concept. Although decent operation lifetime is the prerequisiteof any practical solar-to-fuel conversion, promising strategies toovercome the stability obstacle have rarely been reported for b-Si

photoelectrodes. A sophisticated second etching process was foundto be able to improve the air stability of b-Si PEC photocathodes13.Nonetheless, stabilized electrode in the air is far from adequate sinceubiquitous hydrogen production requires continuous operation inharsh aqueous environments, such as acid or alkaline solutions.

To stabilize PEC photoelectrodes, the prevailing strategy is tointroduce a protecting layer to isolate the photoactive surfacefrom the corrosive electrolyte17–20. The protective layer should haveinert chemical activity, rapid charge conductivity, and extremelyhigh uniformity. Relying on its unique sequential self-limitingsurface reaction mechanism, atomic layer deposition (ALD) showsextraordinary advantages in producing conformal and pinhole-freeprotecting thin films on nanostructured surfaces. ALD Al2O3 thinfilms were first introduced to b-Si-based solar cells and showedgreat promise in suppressing charge recombination on defective Sisurfaces16,21,22. Recently, amorphous TiO2 coating by ALDwas foundexceptionally stable in alkaline solutions and highly conductivefor photogenerated holes, offering a promising solution for PECelectrode protection23–27.

Inspired by these discoveries, herein, we present a successfulimplementation of ultrathin TiO2 protective films on nanostruc-tured b-Si surfaces. With the presence of a Co(OH)2 catalytic film,this nanolayer heterostructure achieved a photocurrent density ashigh as 32.3mA cm−2 under 1 sun illumination. This impressivePEC performance was attributed to the dual function of the TiO2protective layer. It can effectively passivate the surface defective siteswithout compromising any light absorption and charge transportfeatures, resulting in a promoted charge separation efficiency andimproved photocurrent density. Meanwhile, the conformal TiO2

1Department of Materials Science and Engineering, University of Wisconsin-Madison, Madison, Wisconsin 53706, USA. 2Department of MaterialsPhysics, School of Materials Science and Engineering, University of Science and Technology Beijing, Beijing 100083, China. 3Materials Science Center,University of Wisconsin-Madison, Madison, Wisconsin 53706, USA. 4Beijing Municipal Key Laboratory of Advanced Energy Materials and Technologies,University of Science and Technology Beijing, Beijing 100083, China. †These authors contributed equally to this work. *e-mail: yuezhang@ustb.edu.cn;xudong.wang@wisc.edu

NATURE ENERGY 2, 17045 (2017) | DOI: 10.1038/nenergy.2017.45 | www.nature.com/natureenergy 1

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

ARTICLES NATURE ENERGY

780 790 800 810

100 nm

2

1

100 nm

a

b

c

d

e

f

g h

20 nm

8 nm

j l

20 nm

k

Binding energy (eV)

Inte

nsity

Cometal

Co 2p3/2Co(OH)2

Co 2p1/2Co(OH)2

SatelliteSatellite

i

2 nm

(100)

(101)0.24 nm

(101)

0.27 nm(100)

10 nm

10 nm

100 nm 100 nm

100 nm 100 nm

Figure 1 | Microscopy and spectroscopy of black silicon photoanodes. a–f, Top-view (a,c,e) and cross-sectional (b,d,f) SEM images of b-Si (a,b), b-Si/TiO2(c,d) and b-Si/TiO2/Co(OH)2 (e,f). EDS results for blue dashed regions are presented in Supplementary Fig. 1c,d. g, Cross-sectional TEM image ofb-Si/TiO2/Co(OH)2. The TiO2 coating thickness was found to be 8 nm, as marked by the white bar. h–j, Cross-sectional STEM image (h) and EELSmappings of the Ti (i) and Co signals (j) for b-Si/TiO2/Co(OH)2. The white dashed box shows the regions used for EELS mapping. k, High-resolution TEMimage of the electrodeposited Co(OH)2 thin film. Inset is the corresponding FFT image. l, X-ray photoelectron spectrum of the element Co in ab-Si/TiO2/Co(OH)2 sample.

layer completely isolated the corrugated b-Si surface from thealkaline electrolyte solution, leading to a drastic improvement of theoperational stability under both in-air and in-electrolyte conditions.This work suggests that ALD TiO2 coating could bring significantperformance gain to nanostructured b-Si PEC electrodes in termsof both charge separation efficiency and catalyst stability.

Synthesis and characterization of b-Si photoanodesThe assembly of b-Si photoanode included three main steps:chemical etching of the n-type Si wafer; ALD coating of the TiO2protective layer; and electrodeposition of the Co(OH)2 catalystfrom CoCl2 solution (denote the electrode as b-Si/TiO2/Co(OH)2,see Methods for details). Figure 1a shows a scanning electron

microscopy (SEM) image of the Si surface after chemical etching,where densely arranged holes with diameters from 25 to 40 nmwere clearly observed. The cross-sectional SEM image in Fig. 1breveals the depth of these channels was about 300 nm. Thisdepth was purposely selected on the basis of maximizing lightabsorption with minimal charge recombination loss. After coating8-nm-thick amorphous TiO2, the diameters of the original holeswere narrowed to 9–24 nm (Fig. 1c). Owing to the conformalgrowth and precise thickness control of ALD, the entirety ofthe nano-channels was uniformly covered by the TiO2 film andthe original channel depth was well preserved for effective lighttrapping (Fig. 1d). The electrodeposited Co(OH)2 thin films wererandomly distributed on the surface of TiO2-coated b-Si (Fig. 1e

2

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

NATURE ENERGY 2, 17045 (2017) | DOI: 10.1038/nenergy.2017.45 | www.nature.com/natureenergy

NATURE ENERGY ARTICLES

20 40 60 80 1000.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

0.8 1.0 1.2 1.4 1.6 1.8 2.0

0

5

10

15

20

25

30

35

|J ano

dic −

J cat

hodi

c|/2

(µA

cm

−2)

E versus RHE (V)

Phot

ocur

rent

den

sity

(mA

cm

−2)

a b

c

Scan rate (mV s–1)

p+-b-Si/Co(OH)2

p+-b-Si/TiO2/Co(OH)2

p+-planar-Si/Co(OH)2

p+-planar-Si/TiO2/Co(OH)2

Planar Si/Co(OH)2

Dark

b-Si/Co(OH)2

b-Si/TiO2/Co(OH)2

Planar Si/TiO2/Co(OH)2

p+-b-Si/TiO2/Co(OH)2

+

+

+

++

− −

− −

−

hν

Co(OH)2

OH−OH− O2O2

OH−

O2

TiO2

Si

400 600 800 1,000 1,200 1,4000

20

40

60

80b-SiPlanar Si

Planar Si

b-Si

Wavelength (nm)

Refle

ctan

ce +

sca

tter

ing

(%)

db-Si/TiO2

b-Si/TiO2

Figure 2 | PEC water oxidation performance of planar silicon and black silicon-based photoanodes. a, Schematic illustration of the charge generation andoxygen evolution processes in b-Si/TiO2/Co(OH)2. b, Photocurrent density–potential (Jph–V) curves of planar Si/Co(OH)2, planar Si/TiO2/Co(OH)2,b-Si/Co(OH)2 and b-Si/TiO2/Co(OH)2 photoelectrodes, and dark current of p+-b-Si/TiO2/Co(OH)2 measured in 1 M NaOH electrolyte under one sunillumination. c, Plot showing the extraction of the double-layer capacitance (Cdl) for a series of Si/Co(OH)2 electrodes. The dashed lines are linear fittingplots, whose slopes correspond to Cdl according to the equation C = I/(dv/dt). d, Light reflectance + scattering spectra of planar Si, b-Si and b-Si/TiO2samples. Inset is the corresponding photograph, where the size of each wafer is approximately 2 cm× 2 cm.

and Supplementary Fig. 1a). Energy dispersive spectroscopy (EDS)revealed the obvious existence of the element Co everywhere, evenat the uncovered surface region and inside the narrow Si channels(Fig. 1f and Supplementary Fig. 1b–d). The large coverage ofCo(OH)2 catalyst is kinetically favourable for the oxygen evolutionreaction (OER) on the photoanode surface.

The distribution, structure and crystallography of TiO2 andCo(OH)2 catalyst were further investigated by transmission electronmicroscopy (TEM). First, a cross-section of the b-Si/TiO2/Co(OH)2sample was observed under TEM to reveal the distribution ofTiO2 and Co(OH)2 inside the Si channels. As shown in Fig. 1g,a uniform TiO2 film coated along the entire channel surface witha fairly coherent film thickness of 8 nm. The bottom portion ofan individual b-Si/TiO2/Co(OH)2 channel (Fig. 1h) was furthercharacterized under scanning transmission electron microscopy(STEM) and electron energy loss spectroscopy (EELS). The Ti signalwas recorded from the entire pore area, evidencing the tubularTiO2 coating geometry (Fig. 1i). The intensity variation was a resultof Co(OH)2 coverage that blocked Ti signal collection. Strong Cosignals were obtained from the centre region (Fig. 1j), suggestingdeep infiltration of the Co(OH)2 catalyst. The disperse distributionof Co(OH)2 excluded the possibility that the b-Si channels mightbe blocked by the Co(OH)2 catalyst. A similar distribution ofCo in b-Si/Co(OH)2 was also confirmed by EELS mapping(Supplementary Fig. 2), evidencing the comparable catalyst coatingbetween b-Si/Co(OH)2 and b-Si/TiO2/Co(OH)2 electrodes.

The two-dimensional catalyst film was semi-transparent underan electron beam (Supplementary Fig. 3), implying a small filmthickness, which is preferred for OER catalysts5,28,29. The high-resolution TEM image and fast Fourier transform (FFT) patternrevealed a polycrystalline film structure with two groups oflattice fringes (Fig. 1k and inset). The lattice spacings weremeasured to be 0.24 and 0.27 nm, matching well with the(101) and (100) planes of Co(OH)2, respectively30. The elementalfeature of b-Si/TiO2/Co(OH)2 was subsequently studied by X-rayphotoelectron spectroscopy (XPS). The characteristic peaks of b-Si,TiO2 and Co(OH)2, including Si 2p, Si 2s, Ti 2p, Co 2p and O 1s,were all identified in the full survey scan (Supplementary Fig. 4).To reveal the chemical environment of the Co catalyst, the Co2p spectrum was individually analysed (Fig. 1l). The major peakcentred at 781.2 eV was the signature peak of Co2+ in Co(OH)2,corresponding to the Co 2p3/2 core level. A small peak located ata binding energy of 778.1 eV was attributed to Co 2p3/2 for Cometal, presumably from the sparsely distributed surface particles(Supplementary Fig. 1a)28. The ratio between Co(OH)2 films andCo metal particles can be tuned by controlling the concentrationof CoCl2 and electrodeposition time (Supplementary Fig. 5a–d).The optimum composition was probed by comparing their PECbehaviours (Supplementary Figs 5e and 6). Under the optimizedcondition (that is, 10mMCoCl2 and 8 s deposition at−1.2V versussaturated calomel electrode (SCE)), the majority of the Co catalystwas in the form of Co(OH)2, which has easily accessible catalytic

NATURE ENERGY 2, 17045 (2017) | DOI: 10.1038/nenergy.2017.45 | www.nature.com/natureenergy

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

3

ARTICLES NATURE ENERGY

sites, high electrocatalytic activity, and negligible absorption ofincident light30.

Photoelectrochemical performance of b-Si photoanodesPEC performance of the b-Si/TiO2/Co(OH)2 photoanode was firstevaluated in 1M sodium hydroxide (NaOH) electrolyte with anexposed photoactive area of ∼0.1 cm2 under one sun illumination(AM 1.5G). Figure 2a schematically illustrates a cross-sectionalview of the electrode architecture and the overall oxygen evolutionprocess. Upon light illumination, b-Si absorbed photons andproduced electrons and holes. The electric field in the depletionregion at the b-Si/electrolyte interface drove holemigration throughthe amorphous TiO2, oxidizing the hydroxyl group to oxygen via theCo(OH)2 catalyst. Electrons were driven to the counter electrodeto reduce protons to hydrogen. Figure 2b shows the photocurrentdensity–potential (Jph–V ) curves of planar Si/Co(OH)2, planarSi/TiO2/Co(OH)2, b-Si/Co(OH)2 and b-Si/TiO2/Co(OH)2, and thedark current of p+-Si/TiO2/Co(OH)2. The onset potentials of thephotoactive samples were located at the same position (∼1.16Vversus reversible reference electrode (RHE)), one of the mostnegative values for Si photoanodes without a buried junction5,28. Thedark current of p+-b-Si/TiO2/Co(OH)2 started to rise from 1.65Vversus RHE, resulting in an overpotential of 0.42V after subtractingthe thermodynamic potential requirement for water splitting(1.23V). This value was comparable to typical Si-Co-based OERcatalysts31,32. By comparing Jph–V curves of b-Si/TiO2/Co(OH)2and p+-b-Si/TiO2/Co(OH)2, the photovoltage generated by the n-Siwas estimated to be 0.49V, similar to the best-performing single-junction Si photoanode5. Such a high photovoltage was probablycontributed from the combined effect of Fermi level mismatchbetween b-Si and electrolyte, and the porous catalyst-induced spatialvariation of the Si/electrolyte barrier height. The porous Co(OH)2and sparsely distributed Co metal catalysts on b-Si/TiO2 surfacewere equivalent to an inhomogeneous catalyst distribution. Severalreports have shown that a porous or inhomogeneous catalystarrangement could yield a spatially non-uniform barrier height,and consequently increase the built-in potential and enhance thephotovoltage28,33,34.

The onset potential of PEC photoanode is governed by the flat-band potential (EFB). Theoretically, EFB is determined by the Fermilevel of semiconductors relative to vacuum (φSC) and the Helmholtzdouble-layer potential drop (VH) via equation: EFB(NHE) = φSC +

VH− 4.5, where 4.5 is a scale factor relating theH+/H2 redox level tovacuum9. For our photoelectrodes, the doping densities of four n-Sisubstrateswere exactly identical. Surfacemorphology variations andthe TiO2 coating were unlikely to influence the doping level orcarrier density of the underlying Si, so that their φSC was expected tobe the same.VH depends on the solution pH and the point of zero ζpotential (pHPZZP) of the solid material that contacts with electrolytefollowing the relationshipVH = 0.059 (pHPZZP–pH)9. With the samepH and equivalent Co(OH)2/electrolyte interface (SupplementaryFigs 7 and 8), VH should be identical. Therefore, similar EFBand onset potentials of Jph–V could be anticipated for these Siphotoelectrodes, whichwas further supported by theMott–Schottkymeasurements (Supplementary Fig. 9). The Mott–Schottky-probedEFB values for b-Si/Co(OH)2 and b-Si/TiO2/Co(OH)2 were bothlocated at around 0.5V versus RHE, comparable to the valuerecorded from a similar silicon/electrolyte junction5.

In contrast, Jph exhibits a remarkable difference after addingan ALD TiO2 layer to the b-Si electrodes. The saturated Jph ofplanar Si/Co(OH)2 and planar Si/TiO2/Co(OH)2 reached 29.2 and29.0mA cm−2, respectively, comparable to that of other planarSi photoanodes32,35. After converting planar Si to b-Si, the Jphslightly increased to 30.4mA cm−2, a likely result of the enhancedlight absorption. Compared to planar Si, the intense surfacerecombination caused by the enlarged surface area and poor

interface quality of b-Si substantially compensated the photocurrentimprovement originating from the enhanced light absorption,resulting in the relatively small discrepancy between planar andb-Si electrodes. The largest Jph of 32.3mA cm−2 was obtainedfrom the b-Si/TiO2/Co(OH)2 sample, meaning the ALD TiO2 wascontributing to the OER.

Understanding the improvement in photocurrent densityA series of electrochemical characterization techniques wasimplemented to uncover the TiO2 contributions to the observedphotocurrent enhancement. First, the morphology and surfacearea of Co(OH)2 were thoroughly compared before and after TiO2coating to reveal possible alterations from the Co(OH)2 catalyst. Asshown in Supplementary Fig. 10, the Co(OH)2 catalysts on b-Si andb-Si/TiO2 exhibited no observable geometrical difference from SEMobservation. To quantify the electrochemically active surface area,the double-layer capacitances (Cdl), which serves as an estimationof solid–liquid interface area, were determined from the slope ofanodic–cathodic current density difference versus scan rate (dv/dt)extracted from a series of cyclic voltammetry curves (Fig. 2c)36.In the measurement, a degenerately doped p-type Si wafer(denoted as p+) was used instead of the n-type semiconductingSi to minimize the potential (IR) loss induced by the substrate.Two planar silicon samples (that is, p+-planar-Si/Co(OH)2 andp+-planar-Si/TiO2/Co(OH)2) showed almost identical Cdl (2.83 and2.74 µF cm−2, respectively). These values were comparable to theCdl measured from other planar Si/Co(OH)2 photoelectrodes31.Cdl of both b-Si samples were more than ten times larger thanthose of planar Si electrodes owing to the dramatically enlargedsurface area. The presence of TiO2 slightly reduced Cdl of b-Si from31.17 to 30.29 µF cm−2, which was probably due to the filling ofsmall surface pores by the TiO2 coating. These data confirmed thatthe Co(OH)2 catalyst had the same activity before and after TiO2protection.

Aside from the Co(OH)2 catalyst contribution, Jph issynergistically controlled by the light absorption, charge separationand charge injection properties of the photoelectrode. Specifically,J ph = J abs× ηseparation× ηinjection, where J abs is the photocurrent densityat 100% internal quantum efficiency; ηseparation is the electron–holeseparation efficiency for producing surface-reaching holes; andηinjection is the efficiency of surface-reaching holes participatingin the OER. For these electrodes, ηinjection was unlikely to be thedominating factor for the Jph variation considering the identicalCo(OH)2 catalyst layer (that is, comparable catalytic ability for hole-to-oxygen conversion). Light absorption and charge separationcharacteristics were subsequently targeted for understanding theJph improvement.

Figure 2d and the inset show the light reflectance + scatteringspectra and photographs of planar Si, pristine b-Si and b-Si with8-nm-thick TiO2 coating. Due to the light trapping effect at thenanostructured surface, b-Si exhibited notably reduced reflectance(∼10%) compared to planar Si (over 30%) in the ultraviolet, visibleand near-infrared regions, consistent with typical b-Si substrates37,38.The reflectance + scattering spectra of b-Si and b-Si/TiO2 showedlittle difference, suggesting the TiO2 thin film did not affect thelight absorption of b-Si. Further adding Co(OH)2 did not alterthe light reflectance + scattering spectra of b-Si (SupplementaryFig. 11). The corresponding ultraviolet–visible light absorptionspectra were derived based on the reflectance + scattering spectra(Supplementary Fig. 12a). By integrating the absorbance acrossAM 1.5G solar spectrum39, J abs of planar Si, b-Si and b-Si/TiO2were calculated to be 33.3, 38.0 and 37.8mA cm−2 (SupplementaryFig. 12b–d), respectively, suggesting that enhanced light absorptionwas the main reason for the Jph improvement in b-Si over planarSi electrodes. However, the discrepancy between b-Si/Co(OH)2 andb-Si/TiO2/Co(OH)2 was barely related to light absorption.

4

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

NATURE ENERGY 2, 17045 (2017) | DOI: 10.1038/nenergy.2017.45 | www.nature.com/natureenergy

NATURE ENERGY ARTICLES

0.8 1.0 1.2 1.4 1.6 1.8 2.0−5

0

5

10

15

20

25

30

35Ph

otoc

urre

nt d

ensi

ty (m

A c

m−2

)

E versus RHE (V)

E versus RHE (V)

b-Si/TiO2/Co(OH)2

b-Si/Co(OH)2

Phosphate buffer (pH = 7.0)with 1 M Na2SO3

a

1.0 1.2 1.4 1.6 1.8 2.0−0.2

0.0

0.2

0.4

0.6

0.8

1.0b-Si/TiO2/Co(OH)2

b-Si/Co(OH)2

b

Char

ge s

epar

atio

n effi

cien

cy

Figure 3 | Sulfite oxidation and charge separation properties of blacksilicon-based photoanodes. a, Jph–V curves of b-Si/Co(OH)2 andb-Si/TiO2/Co(OH)2 for sulfite oxidation measured in phosphate bu�er(pH = 7) with 1 M Na2SO3 as the hole scavenger under one sunillumination. b, Charge separation e�ciencies of b-Si/Co(OH)2 andb-Si/TiO2/Co(OH)2 calculated from sulfite oxidation curves.

To understand ηseparation, sulfite was introduced as the electrolyteand hole scavenger, enabling a complete participation of surface-reaching holes in the sulfite oxidation (that is, ηinjection = 100%).Thus, ηseparation can be determined by comparing the photocurrentdensity of sulfite oxidation and the maximum photocurrentdensity calculated from light absorption (that is, ηseparation =J sulfite/J abs). As shown in Fig. 3a, for both b-Si/Co(OH)2 andb-Si/TiO2/Co(OH)2, Jph − V curves of sulfite oxidation followedthe trend of water oxidation with slightly increased photocurrentdensity (30.4 and 33.0mA cm−2, respectively). The correspondingηseparation was calculated and presented as a function of appliedbias in Fig. 3b, where substantially higher ηseparation was obtainedfrom b-Si/TiO2/Co(OH)2 at bias greater than 1.45V versus RHE.The highest ηseparation of 86.1% was obtained at 1.65V and aboveversus RHE, which was 8% higher than that of b-Si/Co(OH)2(79.6%). The ratio between ηseparation agreed well with the Jphvariation shown in Fig. 2b, implying ηseparation would be thepredominant factor responsible for TiO2-induced Jph enhancement.Comparable radii of semicircles in electrochemical impedancespectroscopy (Supplementary Fig. 13) confirmed that no extraresistance for hole migration was induced after adding amorphous

TiO2, consistent with previous reports23,24. Increased ηseparationevidenced that the ALD TiO2 layer was able to passivate defectivesurfaces and increase the minority carrier lifetime in b-Si PECphotoelectrodes. This phenomenon is analogous to ALD aluminapassivation in b-Si solar cells16,21,22, but is observed here forthe first time in electrochemical b-Si systems, to the best ofour knowledge.

Stability evaluation of black silicon photoanodeIn addition to the photocurrent gain, the conformal and pinhole-free TiO2 film is capable of stabilizing b-Si photoelectrodes bothin air and in electrolyte. By isolating the b-Si surface from theatmosphere, the TiO2 layer can prevent the formation of a thickinsulating SiO2 layer, and thus improve the in-air stability of b-Si.Figure 4a compares Jph–V curves of freshly assembled and three-month atmosphere-aged b-Si/TiO2/Co(OH)2 photoelectrodes,where nearly equivalent behaviours can be identified, evidencingthe good in-air stability. Since practical hydrogen productionrequires continuous operation in a corrosive solution, the durabilityof b-Si/TiO2/Co(OH)2 was further assessed in 1M NaOH bymeasuring the photocurrent density–time (Jph–t) characteristicsat a constant external bias of 1.65V versus RHE. As shown inFig. 4b, b-Si/Co(OH)2 without TiO2 experienced a quick decreaseof Jph within less than a half an hour. The deterioration becameeven sharper after 2.5 h, probably caused by structural damage,leading to 77.3% loss of Jph after 3 h operation. Top-view andcross-sectional SEM images in Fig. 4c and Supplementary Fig. 14aconfirmed the complete collapse of the surface channels from theb-Si/Co(OH)2 sample. The majority of Co(OH)2 was oxidized toCoOOH after the OER reaction (Supplementary Fig. 15). On theother hand, after decoupling the fragile b-Si surface with corrosiveelectrochemical reaction, 81.2% of Jph from Si/TiO2/Co(OH)2can be maintained after 4 h of continuous OER. The small losswas possibly caused by the dissolution/exfoliation of the catalystlayer in 1M NaOH solution28, as evidenced by the Co signal lossafter a four-hour reaction (Supplementary Fig. 16). Figure 4d andSupplementary Fig. 14b demonstrate the well-preserved nano-structured b-Si surface with a good coverage of fluffy Co(OH)2catalyst, reinforcing the presence of Co catalyst along the insidechannel surface. These results highlight the vital role of theALD TiO2 film in improving the functional stability of b-Si PECphotoelectrodes.

ConclusionsIn summary, we developed a stabilized b-Si photoanode withenhanced charge separation capability for PEC water splitting.The entire channel surface of b-Si was uniformly covered by a8-nm-thick, hole-conductive and pinhole-free TiO2 protectivelayer prepared by ALD. Combined with an optimized thin layerof Co(OH)2 OER catalyst, this b-Si photoelectrode was able todeliver a saturated photocurrent density of 32.3mA cm−2 at alow external bias of 1.48 V versus RHE, distinctly outperformingplanar Si/Co(OH)2 and b-Si/Co(OH)2 counterparts. By comparingtheir electrochemically active surface area, light absorption andcharge separation properties, the high photocurrent density fromb-Si/TiO2/Co(OH)2 was found to be a consequence of the promotedcharge separation efficiency, which was attributed to the effectiveTiO2 passivation of the defective sites on the b-Si surface. Addi-tionally, this ALD-grown TiO2 film can thoroughly decouple thechemically unstable b-Si from corrosive electrochemical reactions,resulting in a significant improvement of operational stability underboth in-air (3month) and in-electrolyte (4 h continuous operation)conditions. This research suggests that ALD TiO2 protectioncould be an effective approach to simultaneously improving theperformance and durability of b-Si-based electrochemical systems,bringing new promise towards effective solar-fuel generation.

NATURE ENERGY 2, 17045 (2017) | DOI: 10.1038/nenergy.2017.45 | www.nature.com/natureenergy

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

5

ARTICLES NATURE ENERGY

1.0 1.2 1.4 1.6 1.8 2.0

0

5

10

15

20

25

30

35

DarkFresh

After 3 months

E versus RHE (V)

Phot

ocur

rent

den

sity

(mA

cm

−2)

a

100 nm 100 nm

c d

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.00

5

10

15

20

25

30

35

b-Si/Co(OH)2

b-Si/TiO2/Co(OH)2

Phot

ocur

rent

den

sity

(mA

cm

−2)

Time (h)

b

Figure 4 | Stability evaluation of black silicon-based photoanodes under in-air and in-electrolyte conditions. a, Jph–V curves of freshly assembled andthree-month atmosphere-aged b-Si/TiO2/Co(OH)2 for water oxidation measured in 1 M NaOH electrolyte under one sun illumination. b, Photocurrentdensity–time (Jph–t) curves of b-Si/Co(OH)2 and b-Si/TiO2/Co(OH)2 at a constant bias of 0.6 V versus SCE (that is, 1.65 V versus RHE).c,d, Cross-sectional SEM image of b-Si/Co(OH)2 (c) and b-Si/TiO2/Co(OH)2 (d) after 3 and 4 h PEC water oxidations, respectively.

MethodsFabrication of black silicon. A 380-µm-thick, phosphorus-doped,single-side-polished, 〈100〉-oriented, n-type Si wafer with a resistivity of1–10� cm was first cleaned sequentially in an ultrasonic bath of isopropanol anddistilled (DI) water for 20min. The wafer was then immersed in 5wt% HF for30 s to remove the native oxide. After that, the substrate was dipped in a mixedsolution of 1mM AgNO3 and 1wt% HF for 30 s. Subsequently, the wafer wasrinsed with DI water and then quickly transferred into the etching solution,which contained 50wt% HF, 30wt% H2O2 and DI water with a volume ratio of25:6:370. After etching for 80 s, the substrate was rinsed in DI water and thenimmersed in 35wt% HNO3 solution for 5min to remove the Ag nanoparticleresidue. Finally, the wafer was rinsed with DI water and dried by air blowing.Porous p+-Si samples were prepared in a similar way, except changing the n-Siwafer to a 380-µm-thick, heavily boron doped, single-side polished,〈100〉-oriented, p-type wafer with a resistivity of 0.001–0.005� cm.

Assembly of black silicon photoanodes. Prior to TiO2 and Co(OH)2 depositions,Si wafers were immersed in 5wt% HF to remove the native oxide. TiO2 coatingwas conducted in a home-made atomic layer deposition system followingreported procedures40,41. Briefly, TiCl4 and H2O vapours were pulsed into a120 ◦C reaction chamber separately with a pulse time of 0.5 s, separated by 60 sN2 purging. One deposition cycle involves 0.5 s of H2O pulse + 60 s of N2 purging+ 0.5 s of TiCl4 pulse + 60 s of N2 purging. An 8-nm-thick TiO2 coating wasdeposited after 100 ALD cycles, corresponding to a growth rate of 0.08 nm percycle. For Co(OH)2 deposition, the back sides of planar Si, b-Si and b-Si/TiO2

were first scratched with a diamond scribe, and then were coated with Ga/Ineutectic mixture and connected to a metal lead, forming an ohmic back contact.Silver paint was then used to affix the lead. After drying, the entire back side andpartial front side of the Si electrodes were encapsulated in epoxy, establishing anexposed active area of ∼0.1 cm2. Calibrated digital images and ImageJ35 wereused to determine the geometrical area of the exposed electrode surface definedby epoxy. The electrodeposition was carried out in a typical three-electrodeelectrochemical setup with Si photoanodes as the working electrode, Ptwire as the counter electrode and a saturated calomel electrode (SCE) as thereference electrode. During the deposition, a constant bias of −1.2V versusSCE was applied onto the working electrodes for 2–20 s with 1–50mM CoCl2as the electrolyte. The optimized condition was 8 s and 10mM CoCl2. After

rinsing with DI water, the electrode was ready for photoelectrochemical(PEC) measurement.

Photoelectrochemical measurement. A similar three-electrode cell (that is, Siphotoanode as the working electrode, Pt wire as the counter electrode and SCE asthe reference electrode) was implemented for PEC measurements. Two kinds ofelectrolytes were used, including 1M NaOH solution and phosphate buffersolution (pH = 7) with 1M Na2SO3. During the Jph–V and Jph–t scanning,working electrodes were illuminated by a 150W xenon lamp coupled with an AM1.5 global filter with a light intensity of 100mWcm−2 (one sun). Jph–t curves wasmeasured at a constant external bias of 1.65V versus RHE. Readings for SCEwere converted to RHE using the following relationship: E(RHE) = E(SCE) +0.244V + 0.059 × pH. Jph–V and Jph–t curves were recorded using an AutolabPGSTAT302N station. The extraction of Cdl started with the recording of CV inthe non-faradaic bias region (−0.1–0V versus SCE). By plotting half of thecurrent density difference between anodic and cathodic sweeps (that is, J anodic −J cathodic/2) as a function of scan rate, a linear trend was obtained. The linear fittingslope corresponds to Cdl. The scan rates were 20, 40, 60, 80, 100mV s−1. Theelectrochemical impedance measurements were performed in 1M NaOH solutionusing a Solartron Electrochemical Interface SI 1287. The Mott–Schottky plotswere acquired at a voltage of 50mV and a frequency of 1 kHz.

Characterization. SEM observations were performed on a Zeiss LEO 1530field-emission microscope and TEM measurements were conducted on a FEITF30 microscope. Focussed ion beam cutting was carried out by means of a FIBZeiss Auriga. STEM observation and EELS mapping were performed on a FEITitan microscope with a CEOS probe aberration corrector operated at 200 keV.XPS was acquired using a Thermo Scientific K-alpha XPS instrument. Thereflectance spectra were recorded using a Jasco V-570 UV–Vis spectrophotometerwith an integrating sphere.

Data availability. The data that support the plots within this paper and otherfindings of this study are available from the corresponding author uponreasonable request.

Received 6 August 2016; accepted 23 February 2017;published 27 March 2017

6

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

NATURE ENERGY 2, 17045 (2017) | DOI: 10.1038/nenergy.2017.45 | www.nature.com/natureenergy

NATURE ENERGY ARTICLESReferences1. Lewis, N. S. Research opportunities to advance solar energy utilization. Science

351, 1920 (2016).2. Khaselev, O. & Turner, J. A. Monolithic photovoltaic-photoelectrochemical

device for hydrogen production via water splitting. Science 280,425–427 (1998).

3. Walter, M. G. et al. Solar water splitting cells. Chem. Rev. 110,6446–6473 (2010).

4. Joya, K. S., Joya, Y. F., Ocakoglu, K. & van de Krol, R. Water-splitting catalysisand solar fuel devices: artificial leaves on the move. Angew. Chem. Int. Ed. 52,10426–10437 (2013).

5. Kenney, M. J. et al.High-performance silicon photoanodes passivated withultrathin nickel films for water oxidation. Science 342, 836–840 (2013).

6. Liu, C., Colon, B. C., Ziesack, M., Silver, P. A. & Nocera, D. G. Watersplitting-biosynthetic system with CO2 reduction efficiencies exceedingphotosynthesis. Science 352, 1210–1213 (2016).

7. Wang, G. et al.Hydrogen-treated TiO2 nanowire arrays forphotoelectrochemical water splitting. Nano Lett. 11, 3026–3033 (2011).

8. Warren, S. C. et al. Identifying champion nanostructures for solarwater-splitting. Nat. Mater. 12, 842–849 (2013).

9. Kim, T. W. & Choi, K. S. Nanoporous BiVO4 photoanodes withdual-layer oxygen evolution catalysts for solar water splitting. Science 343,990–994 (2014).

10. Huang, Z., Geyer, N., Werner, P., de Boor, J. & Gosele, U. Metal-assistedchemical etching of silicon: a review. Adv. Mater. 23, 285–308 (2011).

11. Yuan, G. et al. Understanding the origin of the low performance of chemicallygrown silicon nanowires for solar energy conversion. Angew. Chem. Int. Ed. 50,2334–2338 (2011).

12. Oh, J., Deutsch, T. G., Yuan, H.-C. & Branz, H. M. Nanoporous black siliconphotocathode for H2 production by photoelectrochemical water splitting.Energy Environ. Sci. 4, 1690–1694 (2011).

13. Zhao, Y. et al. Oxidatively stable nanoporous silicon photocathodes withenhanced onset voltage for photoelectrochemical proton reduction. Nano Lett.15, 2517–2525 (2015).

14. Ali, M. et al. Nanostructured photoelectrochemical solar cell for nitrogenreduction using plasmon-enhanced black silicon. Nat. Commun. 7,11335 (2016).

15. Oh, J., Yuan, H. C. & Branz, H. M. An 18.2%-efficient black-silicon solar cellachieved through control of carrier recombination in nanostructures.Nat. Nanotech. 7, 743–748 (2012).

16. Savin, H. et al. Black silicon solar cells with interdigitated back-contacts achieve22.1% efficiency. Nat. Nanotech. 10, 624–628 (2015).

17. Chen, Y. W. et al. Atomic layer-deposited tunnel oxide stabilizes siliconphotoanodes for water oxidation. Nat. Mater. 10, 539–544 (2011).

18. Chen, L. et al. p-type transparent conducting oxide/n-type semiconductorheterojunctions for efficient and stable solar water oxidation. J. Am. Chem. Soc.137, 9595–9603 (2015).

19. Ji, L. et al. A silicon-based photocathode for water reduction with an epitaxialSrTiO3 protection layer and a nanostructured catalyst. Nat. Nanotech. 10,84–90 (2015).

20. Sun, K. et al. Stable solar-driven oxidation of water by semiconductingphotoanodes protected by transparent catalytic nickel oxide films. Proc. NatlAcad. Sci. USA 112, 3612–3617 (2015).

21. Wang, W. C. et al. Surface passivation of efficient nanotextured black siliconsolar cells using thermal atomic layer deposition. ACS Appl. Mater. Interfaces 5,9752–9759 (2013).

22. Wang, W. C., Tsai, M. C., Yang, J., Hsu, C. & Chen, M. J. Efficiencyenhancement of nanotextured black silicon solar cells using Al2O3/TiO2

dual-layer passivation stack prepared by atomic layer deposition. ACS Appl.Mater. Interfaces 7, 10228–10237 (2015).

23. Hu, S. et al. Amorphous TiO2 coatings stabilize Si, GaAs, and GaP photoanodesfor efficient water oxidation. Science 344, 1005–1009 (2014).

24. Gu, J. et al.Water reduction by a p-GaInP2 photoelectrode stabilized by anamorphous TiO2 coating and a molecular cobalt catalyst. Nat. Mater. 15,456–460 (2016).

25. Qiu, J. et al.Microscopic study of atomic layer deposition of TiO2 on GaAs andits photocatalytic application. Chem. Mater. 27, 7977–7981 (2015).

26. Shaner, M. R., Hu, S., Sun, K. & Lewis, N. S. Stabilization of Si microwire arraysfor solar-driven H2O oxidation to O2(g) in 1.0M KOH(aq) using conformalcoatings of amorphous TiO2. Energy Environ. Sci. 8, 203–207 (2015).

27. Lee, C. Y., Park, H. S., Fontecilla-Camps, J. C. & Reisner, E.Photoelectrochemical H2 evolution with a hydrogenase immobilized on aTiO2-protected silicon electrode. Angew. Chem. Int. Ed. 55, 5971–5974 (2016).

28. Hill, J. C., Landers, A. T. & Switzer, J. A. An electrodeposited inhomogeneousmetal–insulator–semiconductor junction for efficient photoelectrochemicalwater oxidation. Nat. Mater. 14, 1150–1155 (2015).

29. Ng, J. W. D. et al. Gold-supported cerium-doped NiOx catalysts for wateroxidation. Nat. Energy 1, 16053 (2016).

30. Feng, J. X. et al. Co(OH)2@PANI hybrid nanosheets with 3D networks ashigh-performance electrocatalysts for hydrogen evolution reaction. Adv. Mater.27, 7051–7057 (2015).

31. Yang, J. et al. A multifunctional biphasic water splitting catalyst tailored forintegration with high-performance semiconductor photoanodes. Nat. Mater.16, 335–341 (2017).

32. Yang, J. et al. Efficient and sustained photoelectrochemical water oxidation bycobalt oxide/silicon photoanodes with nanotextured interfaces. J. Am. Chem.Soc. 136, 6191–6194 (2014).

33. Lin, F. & Boettcher, S. W. Adaptive semiconductor/electrocatalyst junctions inwater-splitting photoanodes. Nat. Mater. 13, 81–86 (2014).

34. Laskowski, F. L., Nellist, M. R., Venkatkarthickab, R. & Boettcher, S. W.Junction behavior of n-Si photoanodes protected by thin Ni elucidated fromdual working electrode photoelectrochemistry. Energy Environ. Sci. 10,570–579 (2017).

35. Sun, K. et al. Stable solar-driven water oxidation to O2(g) by Ni-oxide-coatedsilicon photoanodes. J. Phys. Chem. Lett. 6, 592–598 (2015).

36. Caban-Acevedo, M. et al. Efficient hydrogen evolution catalysis using ternarypyrite-type cobalt phosphosulphide. Nat. Mater. 14, 1245–1251 (2015).

37. Garnett, E. & Yang, P. Light trapping in silicon nanowire solar cells. Nano Lett.10, 1082–1087 (2010).

38. Branz, H. M. et al. Nanostructured black silicon and the optical reflectance ofgraded-density surfaces. Appl. Phys. Lett. 94, 231121 (2009).

39. Yang, W. et al. Ferroelectric polarization-enhanced photoelectrochemical watersplitting in TiO2-BaTiO3 core-shell nanowire photoanodes. Nano Lett. 15,7574–7580 (2015).

40. Yu, Y., Yin, X., Kvit, A. &Wang, X. Evolution of hollow TiO2 nanostructures viathe Kirkendall effect driven by cation exchange with enhancedphotoelectrochemical performance. Nano Lett. 14, 2528–2535 (2014).

41. Yu, Y. et al. Development of lead iodide perovskite solar cells usingthree-dimensional titanium dioxide nanowire architectures. ACS Nano 9,564–572 (2015).

AcknowledgementsWork on ALD-based growth and structure characterization is supported by USDepartment of Energy (DOE), Office of Science, Basic Energy Sciences (BES), underAward # DE-SC0008711. The rest of the work is supported by the National MajorResearch Program of China (No. 2013CB932602); the Program of Introducing Talents ofDiscipline to Universities (B14003); National Natural Science Foundation of China(No. 51527802, 51232001, 51602020 and 51672026); and Beijing Municipal Science &Technology Commission.

Author contributionsY.Y., Y.Z. and X.W. conceived the ideas, designed the experiments and oversaw the entireproject. Y.Y. and Z.Z. performed the device fabrication, electrochemical measurementsand data analysis. Y.Y. Q.L., Z.K. and X.Yan conducted the SEM, XPS and XRDcharacterizations. Y.Y., X.Yin and A.K. carried out the TEM, STEM and EELS mappings.Y.Y., Z.Z., Y.Z. and X.W. wrote the manuscript. All authors discussed the experiments andcommented on the manuscript.

Additional informationSupplementary information is available for this paper.Reprints and permissions information is available at www.nature.com/reprints.Correspondence and requests for materials should be addressed to Y.Z. or X.W.How to cite this article: Yu, Y. et al. Enhanced photoelectrochemical efficiency andstability using a conformal TiO2 film on a black silicon photoanode. Nat. Energy 2,17045 (2017).Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims inpublished maps and institutional affiliations.

Competing interestsThe authors declare no competing financial interests.

NATURE ENERGY 2, 17045 (2017) | DOI: 10.1038/nenergy.2017.45 | www.nature.com/natureenergy

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

7