ARTICLESPUBLISHED ONLINE: 14 SEPTEMBER 2015 | DOI: 10.1038/NMAT4410

E�cient hydrogen evolution catalysis usingternary pyrite-type cobalt phosphosulphideMiguel Cabán-Acevedo1, Michael L. Stone1, J. R. Schmidt1, Joseph G. Thomas1, Qi Ding1,Hung-Chih Chang2, Meng-Lin Tsai2, Jr-Hau He2 and Song Jin1*

The scalable and sustainable production of hydrogen fuel throughwater splitting demands e�cient and robust Earth-abundantcatalysts for the hydrogen evolution reaction (HER). Building on promising metal compounds with high HER catalytic activity,such as pyrite structure cobalt disulphide (CoS2), and substituting non-metal elements to tune the hydrogen adsorption freeenergy could lead to further improvements in catalytic activity. Here we present a combined theoretical and experimentalstudy to establish ternary pyrite-type cobalt phosphosulphide (CoPS) as a high-performance Earth-abundant catalyst forelectrochemical and photoelectrochemical hydrogen production. Nanostructured CoPS electrodes achieved a geometricalcatalytic current density of 10mAcm−2 at overpotentials as low as 48mV, with outstanding long-term operational stability.Integrated photocathodes of CoPS on n+–p–p+ silicon micropyramids achieved photocurrents up to 35mAcm−2 at 0V versusthe reversible hydrogen electrode (RHE), onset photovoltages as high as 450mV versus RHE, and the most e�cient solar-driven hydrogen generation from Earth-abundant systems.

Hydrogen is a sustainable energy carrier that promises anenvironmentally friendly alternative to meet future globalterawatt energy demands1,2. The production of hydrogen by

means of water splitting, preferably solar-driven, requires highlyefficient and robust catalyst materials2. Noble metals such asplatinum are at present the most active catalysts for the hydrogenevolution reaction (HER). Unfortunately, their application in large-scale hydrogen production is limited by high cost and low elementalabundance3,4. Motivated by this challenge, the search for cost-effective, Earth-abundant materials with both high HER activityat low overpotentials and excellent stability has recently attractedsignificant research interest and become an important pursuittowards enabling a hydrogen economy. Various classes of Earth-abundant transitionmetal compounds4, such asMoS2 (refs 5,6),WS2(refs 7,8), amorphous MoSx (ref. 9), CoS2 (ref. 10), CoSe2 (ref. 11),CoP (refs 12,13), Ni2P (ref. 14), FeP (ref. 15), MoP (refs 16,17),MoP|S (ref. 18), and Ni–Mo alloys19, have been recently identifiedas promising HER electrocatalysts. Furthermore, it could be evenmore challenging to integrate catalyst materials in solar-drivenphotoelectrochemical cells (PECs; ref. 20).

Among the various Earth-abundant HER catalysts recentlydiscovered4, the pyrite-type transition metal dichalcogenides (MX2,where M = Fe, Co, or Ni and X = S or Se) have emergedas an interesting family of low-cost materials with high catalyticactivity towards the HER (refs 10,11,21–23). The family of metalpyrites includes semiconducting and metallic compounds that arevery abundant as minerals or in sedimentary deposits withinEarth’s crust, which makes them appealing for a variety of energy-conversion-related applications. For example, iron pyrite (cubicFeS2; fool’s gold) is a semiconductor that has been extensivelystudied as a promising cost-effective solar absorber24,25, and hasrecently been studied as a HER catalyst22,23. Moreover, metalliccobalt pyrite (CoS2; cattierite; Fig. 1a) has recently been found

to exhibit high catalytic activity towards the HER (ref. 10),polysulphide and triiodide reduction reactions21,22.

At present, there is no direct evidence for the active site(s)and mechanism responsible for the high HER catalytic activityof pyrite compounds. However, strong similarities between theactive metal centres of Fe-only hydrogenase26,27 and the reducedcoordination environment of the intrinsic {100} pyrite surface24suggest that the active sites are the square-pyramidal surfacemetal centres bridged by dichalcogenide dumbbells (Fig. 1b). Suchsimilarities further suggest that the HER mechanism of pyritecompounds could involve a distorted octahedral metal hydride.Thus, its stability and reactivity (hydricity) can be influenced by theelectron-donating character of the chalcogen ligands (X; ref. 28).Therefore, we predicted that the HER catalytic activity of metalchalcogenides could be improved through tuning the hydrogenadsorption free energy by changing the atomic components of theX2

2− dumbbells, while still preserving the pyrite structure. Given theweaker electronegativity of phosphorus in comparison to sulphur,and the absence of thermodynamically stable metal diphosphidesunder ambient conditions29, we investigate pyrite-type cobaltphosphosulphide (CoPS) as a novel ternary catalyst for the HER.Cobalt phosphosulphide was identified in the 1960s to be a pyritecompound analogous to the naturally occurring mineral cobaltite(CoAsS; ref. 30), but its properties have yet to be investigated. Thecubic crystal structure of CoPS can be visualized similarly to that ofCoS2 (Fig. 1a), but instead of Co2+ octahedra and S22− dumbbells,CoPS has Co3+ octahedra and dumbbells with a homogeneousdistribution of P2− and S− atoms (Fig. 1c), and a smaller latticeconstant (a=5.422Å versus 5.538Å for CoS2). In contrast to CoS2,which is a half-metal, CoPS has one less valence electron, making ita semiconductor isoelectronic to iron pyrite. As the Co octahedra inCoPS contain P2− ligands with higher electron-donating characterthan S− ligands, this ternary pyrite compound could have higher

1Department of Chemistry, University of Wisconsin–Madison, 1101 University Avenue, Madison, Wisconsin 53706, USA. 2Division of Computer, Electricaland Mathematical Sciences and Engineering, King Abdullah University of Science and Technology, Thuwal 23955-6900, Saudi Arabia.*e-mail: jin@chem.wisc.edu

NATUREMATERIALS | VOL 14 | DECEMBER 2015 | www.nature.com/naturematerials 1245

© 2015 Macmillan Publishers Limited. All rights reserved

ARTICLES NATUREMATERIALS DOI: 10.1038/NMAT4410

Free

ene

rgy

(eV

)

Co site(CoS2)Co site(CoPS)

Co site withP−H∗ (CoPS)

Reaction coordinate

P site(CoPS)

1/2 H2(g)H+ + e−

CoS2

Co site H∗ (CoS2) Co site H∗ (CoPS)

P site H∗ (CoPS)Co site H∗ with

H∗ at P site (CoPS)

Co2+

Open sites

L

{100} Pyrite-typesurface species

CoPS

Activemetal centre

Bulk-likecoordination

S−

Co3+

Co2+

Co3+

Co4+

S−

P2−

H−

H+

S−/P2−

Pa3, a = 5.538 ÅPa3, a = 5.422 Å

a

−0.4

−0.2

0.0

0.2

0.4

b c

e f

g h

d

Figure 1 | Design of the ternary pyrite CoPS catalyst and the results from DFT calculations. a–c, Crystal structure of cobalt pyrite (CoS2) (a), intrinsic{100} pyrite surface coordination environment (b), and crystal structure of pyrite-type cobalt phosphosulphide (CoPS) (c). d–h, Free-energy diagram (d)and schematic structural representations for hydrogen (H∗) adsorption at the Co site on the {100} surface of CoS2 (e), and at the Co site (f), P site (g), andCo site after H∗ at P site (h) on the {100} surface of CoPS. Note that DFT calculations for CoPS were performed for ordered P–S dumbbells for simplicity.

HER catalytic activity than CoS2 owing to a more thermoneutralhydrogen adsorption at the active sites. Despite the potentialfor tunable electronic and chemical properties, ternary or morecomplex metal compounds, especially those containing differentnon-metal elements, have rarely been explored for electrocatalysis4.Herein we report the theoretical study of the hydrogen evolutionactivity for CoPS surfaces, the synthesis of CoPS nanostructures andtheir excellent performance towards HER electrocatalysis and solar-driven hydrogen production for the first time.

Hydrogen evolution activity has been shown to be stronglycorrelated with the chemisorption energy of atomic hydrogento the electrocatalyst surface31. We used density function theory(DFT) to calculate the free energy for atomic hydrogen adsorption(1GH∗ ) on the {100} surface of CoPS in comparison to CoS2 (seeMethods for details), to evaluate if CoPS is indeed a promisingHER catalyst. The main results are shown in Fig. 1d. The 1GH∗

for hydrogen adsorption at the Co sites of CoPS (Fig. 1f) is indeedmore favourable than for CoS2 (Fig. 1e).Moreover, DFT calculationsshows that after spontaneous hydrogen adsorption at open P sites(Fig. 1d,g), the 1GH∗ at the adjacent Co sites becomes spontaneousand almost thermoneutral (Fig. 1d,h), reaching a 1GH∗ comparableto that of platinum31. This marked change in1GH∗ can be attributedto a reduction of Co3+ sites to Co2+ on hydrogen adsorption at anadjacent open P site (Fig. 1g), which then enables the oxidationof Co2+ sites back to Co3+ on subsequent hydrogen adsorption(Fig. 1h). We did not observe an analogous change in the 1GH∗ forCoS2, as adsorption at Co sites was found to bemore favourable thanopen S sites. Therefore, DFT calculations suggest that CoPS is a verypromising Earth-abundant HER catalyst.

Encouraged by these promising theoretical results, wesynthesized CoPS electrodes through conversion of cobalt-basednanostructured precursor materials at 500 ◦C in a thiophosphate(PxSy) atmosphere produced by the thermal evaporation of a1:1 stoichiometric mixture of phosphorus and sulphur elementalpowders under an argon atmosphere (see Methods for detail). Thehomogeneous reaction with both phosphorus and sulphur wasoptimized by melting the elemental powder mixture in a cruciblefirst at 200 ◦C for 5–10min under argon atmosphere and thencarrying out the thermal conversion immediately with preloadedcobalt precursors without air exposure to the highly hydroscopicpaste-like PxSy . CoPS films and CoPS nanowires (NWs) weresynthesized using thermally evaporated 100-nm-thick cobalt filmsand hydrothermally grown cobalt hydroxide carbonate hydrate[CHCH, Co(OH)(CO3)0.5 · xH2O] NWs (ref. 32) as precursormaterials on graphite disk substrates. Graphite was used as aconvenient conductive substrate; however, CoPS films and CoPSNWs can be synthesized on other substrates compatible with thethermal conversion conditions (such as borosilicate glass). Scanningelectron microscopy (SEM; Fig. 2a) shows the polycrystalline CoPSfilm with particle sizes of roughly 5 nm (Fig. 2a, inset) obtained onconversion of 100-nm-thick cobalt film on graphite. The flake-likeappearance of the CoPS film is due to the native roughness ofthe graphite substrate, which was purposely used to enhance themechanical stability and surface area of CoPS film. Conversionof the CHCH NWs at 500 ◦C results in CoPS NWs with identicalmorphology, but with a polycrystalline structure and similarparticle size to CoPS film samples (Fig. 2b). High-surface-areaCoPS nanoplates (NPls) were converted from CHCH NPls grown

1246 NATUREMATERIALS | VOL 14 | DECEMBER 2015 | www.nature.com/naturematerials

© 2015 Macmillan Publishers Limited. All rights reserved

NATUREMATERIALS DOI: 10.1038/NMAT4410 ARTICLES

CHCH NPls

CoPS film

25 30 35 40 45 50 55 60 652

∗

∗

∗

∗∗

∗

∗∗ ∗∗

∗

m (111)

(200)

(210)(211) (220) (311) (320)(222) (321)

CoPS film on glass

CoPS NPls on carbon paper

Cubic CoPS

CoS2

Inte

nsity

(a.u

.)

∗∗ ∗

200 300 400 500

Inte

nsity

(a.u

.)

Raman shift (cm−2)

Eg Ag

Coun

ts

Energy (keV)0

CoPS NPlsCoP1.02S0.97

CoPS NWsCoP0.88S1.07

CoPS filmCoP0.83S1.35

1 2 3 4 5 6 7 8

CoCo

SPCo

CSP

CoPS NPlsCoPS NWs

CoPS film

CoPS NWs on graphite

m m m

10 μm 10 μm

CoPS NPls

1 μm1 μm

CoPS NWs

10 μm

1 μm

10 μm

100 nm

a b e

c d

g

f

θ

Figure 2 | Structural characterization of the CoPS nanostructures. a–d, SEM images of as-synthesized CoPs film (a) and CoPS NWs (b) on graphite, andCHCH NPls on carbon fibre paper before (c) and after (d) thermal conversion into CoPS NPls. e,f, EDS and confocal micro-Raman spectra, respectively, forCoPS film and CoPS NWs on graphite, and CoPS NPls on carbon fibre paper. The peak positions for the Eg and Ag Raman modes for CoS2 are included asvertical dashed lines for comparison. g, PXRD patterns for various CoPS products with di�erent morphologies compared to the simulated pattern for cubicCoPS (ICSD collection code #62414) and the standard pattern for CoS2 (JCPDS #41-1471). Note that the CoPS film measured by PXRD was grown onborosilicate glass. A minor marcasite-type CoPS polymorph phase (Pmnn, orthorhombic CoPS) was observed in this sample (labelled as ‘m’). All peakscorresponding to graphite and carbon paper are labelled as ‘∗’.

on carbon fibre paper substrate (Fig. 2c,d). The CHCH NPls weresynthesized using the same procedures as for CHCHNWs (ref. 32),but using carbon paper substrates that were precoated with cobaltoxide (CoO) nanoparticles (see Methods for detail)11. Figure 2cshows that the CHCH NPls thoroughly and uniformly cover thecarbon fibres and efficiently utilize the high surface area of thecarbon paper. Similar to the case of CoPS NWs, thermal conversionat 500 ◦C results in polycrystalline CoPS NPls (Fig. 2d).

We first characterized the crystal structure and stoichiometryof these nanostructures. Energy dispersive X-ray spectroscopy(EDS) revealed the presence of cobalt, phosphorus and sulphur inelemental compositions of CoP0.83S1.35, CoP0.88S1.07 and CoP1.02S0.97,respectively, for the CoPS film, CoPS NW and CoPS NPl electrodes(Fig. 2e). The apparently better stoichiometry of CoPS NPls can beattributed to the presence of excess thiophosphates (PxSy), whichwill be evident from the discussion of the powder X-ray diffraction(PXRD) patterns. Confocal micro-Raman spectra (Fig. 2f) showtwo peaks, at ∼322 cm−1 and 427 cm−1, for all CoPS samples.These Raman peaks correspond to the characteristic active modesof libration (Eg) and in-phase stretch (Ag) for the chalcogenidedumbbells in a pyrite crystal lattice. In comparison to the peaksfor CoS2 (vertical dashed lines in Fig. 2f), the Raman peaks forCoPS electrodes are blueshifted, which can be attributed to thesmaller unit cell of CoPS. The presence of only one set of Egand Ag peaks suggest a ternary CoPS compound rather than amixture of two solid phases. In contrast, two sets of Raman peakscorresponding to CoPS and CoS2 were observed when unsuccessfulthermal conversion was carried out using a 1:3 phosphorus tosulphur ratio (see Supplementary Fig. 1)33, which further supportsthat the obtained CoPS products are indeed a distinctive ternaryalloy phase. X-ray photoelectron spectroscopy (XPS)measurementsshow an absence of multiplet splitting of the cobalt 2p peak forCoPS film in comparison to CoS2 film (see Supplementary Fig. 2),which indicates a spin paired electronic configuration for cobaltand an oxidation state of Co3+ in CoPS (refs 34,35). Therefore,

XPS also supports that the obtained CoPS products are indeed adistinctive ternary alloy phase. The PXRD patterns (Fig. 2g) ofthe film, NWs and NPls can be indexed to a phosphorus-deficientpyrite-type CoPS phase, because the diffraction peaks match thesimulated pattern for cubic CoPS (space group Pa3̄, a=5.422Å,ICSD collection code #62414), but are very slightly shifted towardsthe standard pattern of CoS2 (a = 5.538Å, JCPDS #41-1471).Refinement of the lattice parameter revealed a lattice constant ofa= 5.44Å, which suggests a stoichiometry of ∼CoP0.84S1.16, basedon Vegard’s law. The PXRD patterns for CoPS film and NWs arein agreement with the phosphorus deficiency suggested by the EDSspectra, whereas the PXRD pattern for CoPS NPls indicates that itsapparent EDS stoichiometry can be attributed to the presence ofexcess PxSy . Phosphorus deficiencies in CoPS samples could be aresult of the preferential formation and stability of sulphur-rich PxSy(ref. 36) precursor species and the different diffusivity of elements atthe synthesis temperature of 500 ◦C.

We measured the electrochemical characteristics of the CoPSfilm, CoPS NW and CoPS NPl electrodes corresponding to theHER catalytic performance using 0.5M H2SO4 and a rotating diskelectrode (RDE) at a rate of 2,000 r.p.m., in comparison to platinumwire (see Methods for detail). Figure 3a shows that these CoPS elec-trodes achieved geometric current densities of−10mA cm−2 at verylow overpotentials of−128mV,−61mV and−48mV, respectively,versus the reversible hydrogen electrode (RHE). They also exhibitedoutstanding long-term operation stability beyond 36 h (Fig. 3b) andlittle change in the film morphology (Supplementary Fig. 3). Incomparison, high-density CoS2 NWs (ref. 10) on graphite and high-surface-area CoSe2 nanoparticles on carbon fibre paper11 have beenreported to achieve −10mA cm−2 at overpotentials of −145mVand −180mV versus RHE. The CoPS electrodes also show lowercatalytic overpotentials than non-pyrite structuremetal phosphides,such as CoP (−75mV at 10mA cm−2; ref. 12) and Ni2P (−115mVat 10mA cm−2; ref. 14) suggesting that ternary pyrite-type CoPS isan outstanding HER catalyst. From the extrapolation of the linear

NATUREMATERIALS | VOL 14 | DECEMBER 2015 | www.nature.com/naturematerials 1247

© 2015 Macmillan Publishers Limited. All rights reserved

ARTICLES NATUREMATERIALS DOI: 10.1038/NMAT4410

−0.20 −0.15 −0.10 −0.05 0.00−30

−25

−20

−15

−10

−5

0

J (m

A c

m−2

)

Potential (V) versus RHE

CoPS filmCoPS NWsCoPS NPlsPt wire

a

0 25 50 75 100

0

3

6

9

CoPS filmsCdl = 9.84 mF cm−2

CoPS NWsCdl = 16.9 mF cm−2

CoPS NPlsCdl = 99.6 mF cm−2

|J a −

J c|/

2 (m

A c

m−2

)

Scan rate (mV s−1)−0.5 0.0 0.5 1.0 1.5

0

50

100

150

Slope = 56 mV per decadeJ0 = 984 μA cm−2

Slope = 57 mV per decadeJ0 = 56 μA cm−2

Ove

rpot

entia

l (m

V)

log(J (mA cm−2))

Slope = 48 mV per decadeJ0 = 554 μA cm−2

RCTMa

CPE

RS

Panel f

10 20 30 40 50 60 70 800

20

40

60

80CoPS filmsCoPS NWsCoPS NPls

CoPS filmsCoPS NWsCoPS NPls

4 6 8 100

10

20

30

Rs

Ma

Mass transport

0 10 20 30−200

−150

−100

−50

0

Ove

rpot

entia

l (m

V)

Time (h)

CoPS NPlsCoPS NWsCoPS film

10 mA cm−2

Re(Z) (Ω) Re(Z) (Ω)

−Im

(Z')

(Ω)

−Im

(Z')

(Ω)

b

c d

e f

Figure 3 | Electrochemical characterization of CoPS film on graphite, CoPS NWs on graphite and CoPS NPls on carbon fibre paper electrodes for HERcatalysis. a, J–V curves after iR correction show the catalytic performance of the CoPS electrodes (film, blue open squares; NWs, red open triangles; NPls,green open diamonds) in comparison to a Pt wire (black open circles). b, Long-term stability test for CoPS electrodes at a current density of 10 mA cm−2.c, Tafel plot for the data presented in a. d, Plot showing the extraction of the double-layer capacitance (Cdl) for each type of CoPS electrode.e,f, Electrochemical impedance spectroscopy (EIS) Nyquist plots for CoPS electrodes. The data were fitted using the modified Randles circuits shown inthe insets.

region of a plot of overpotential versus log J (Fig. 3c), we obtainedTafel slopes of 48, 56 and 57mVper decade for CoPS film, NW andNPl electrodes, respectively. Such Tafel slope values suggest a two-electron transfer process following a Volmer–Tafel mechanism ofbimolecular adsorption and hydrogen evolution (theoretical slopeof 45mVper decade)31. From the intercept of the linear region oftheTafel plots, exchange current densities (J0, geometrical) of 56 µAcm−2,554 µAcm−2 and 984 µAcm−2 were obtained for CoPS film, CoPSNWs andCoPSNPls electrodes, respectively. The exchange currentsfor CoPS electrodes are one to two orders of magnitude greater

than those observed for CoS2 (ref. 10) and CoSe2 (ref. 11) materi-als, which explains the better HER catalytic performance of CoPS.Furthermore, we prepared films andNWsofCoPwith aminorCoP2phase impurity (CoPx), CoPx |S, CoPx |H2S, CoS2|P and CoS2|P+H2(see Methods for details) and carried out PXRD and XPS character-ization (Supplementary Figs 6, 7 and 9) and HER electrochemicalmeasurements (Supplementary Figs 8 and 10 and SupplementaryTable 1) for comparison with the CoPS samples. These experimentsunequivocally demonstrate that the catalytic HER performance ofCoPS electrodes can be attributed exclusively to this ternary pyrite-

1248 NATUREMATERIALS | VOL 14 | DECEMBER 2015 | www.nature.com/naturematerials

© 2015 Macmillan Publishers Limited. All rights reserved

NATUREMATERIALS DOI: 10.1038/NMAT4410 ARTICLES

Table 1 | Summary of the electrochemical properties of CoPS electrodes with di�erent morphologies.

Sample η (mVversusRHE) forJ=−10mAcm−2

η (mVversusRHE) forJ=−20mAcm−2

RS(� cm2)

Tafel slope(mVper decade)

J0,geometrical(µAcm−2)

Cdl(mF cm−2)

Relativesurface area

J0,normalized(µAcm−2)

CoPS film −128 −145 2.64 57 56 9.84 1.00 56CoPS NWs −61 −75 1.30 48 554 16.9 1.72 90CoPS NPls −48 −65 1.42 56 984 99.6 10.1 38

type compound, instead of potential cobalt phosphide impurities oranion-substituted doping on the surface of CoPS. Therefore, theseCoPS nanostructures are the most catalytically active pyrite-typematerials and among the very best Earth-abundant catalysts4 thathave been reported so far for HER catalysis.

To understand the origin of the differences in the overall catalyticperformance among CoPS electrodes, we estimated their relativeelectrochemically active surface areas using cyclic voltammetrymeasurements by extracting the double-layer capacitance (Cdl)(Fig. 3d)10. The geometrical exchange current densities werethen normalized to compare HER performance relative to theelectrochemically active surface area. As shown in Table 1, therelative surface areas forCoPSNWs andCoPSNPls are, respectively,1.71 and 10.1 times larger thanCoPS film. The normalized exchangecurrent (J0,normalized) for the CoPS NW electrode is greater than theJ0,normalized for CoPS film (see Supplementary Fig. 4), indicating thatthe higher catalytic activity of CoPS NWs can be mainly attributedto the larger surface area. On the other hand, the J0,normalized of CoPSNPls is smaller than that of CoPS film, despite an approximately tentimes larger surface area and a similar Tafel slope, which suggests areduction in the turnover rate per active site. This is probably causedby a mass-transport-limited37 HER catalytic activity for CoPS NPlsat a RDE rate of 2,000 r.p.m. Circuit model fitting analysis of theelectrochemical impedance spectroscopy (EIS) for the CoPS filmandCoPSNWs (Fig. 3e) shows that both electrodes can bemodelledusing a modified Randles circuit consisting of a series resistance(RS), constant phase element (CPE), charge transfer resistance (RCT),andmodified mass-transport impedance element (Ma). In contrast,the EIS of the CoPSNPls (Fig. 3f) can be fitted using just RS andMa,which is indicative of mass-transport control caused by a negligiblecharge transfer resistance and a very high surface area. Qualitatively,we observed that the semicircular arc corresponding toCPE andRCTimpedance elements seen for the CoPS film significantly decreasesfor CoPS NWs, and completely disappears for CoPS NPls. Wealso observed a significant reduction in the overall impedancefollowing the same trend. All of these indicate that RCT decreaseswith increasing surface area. Themoderate increase inHER catalyticperformance seen in the CoPS NPl electrode in comparison tothe CoPS film electrode, despite a much larger surface area, canbe explained by a mass-transport-limited HER performance forthe CoPS NPls. In other words, the HER catalytic reaction is soefficient that the diffusion of H+ ions cannot keep up with thereaction over the large surface area of the CoPS NPl electrode,leading to a reduction in the exchange current per surface areaor active site, which benefits the long-term operational stability.Therefore, CoPS NPls represent an upper bound for the effects ofsurface area on enhancing the overall electrocatalytic performanceof CoPS.

Given the high electrocatalytic activity of CoPS towards theHER, we further studied CoPS as a catalyst together with p-typesilicon photocathodes for solar-driven hydrogen production.The direct integration of catalysts with homogeneously dopedsemiconductors places strong requirements on the intrinsicelectronic and optical properties of the catalyst20. For example,a high absorption coefficient and a very anodic surface workfunction for a catalyst can affect the light absorption and bandbending of the semiconductor. Therefore, in the fabrication of CoPS

electrocatalysts and silicon cathodes for the HER in PECs we mustconsider the interfacial properties and CoPS layer thickness. Wefirst fabricated CoPS/p-Si photocathode by thermal conversion ofa 10 nm cobalt film evaporated on planar p-type silicon. However,the J–V curve of CoPS/p-Si (Fig. 4a) shows no photoresponseand characteristics similar to those of CoPS electrodes. SuchJ–V characteristics suggest the formation of an ohmic junction,which can be attributed to an accumulation band bending inducedby the work function of CoPS or the formation of interfacialdefect states.

To eliminate the influence of the CoPS catalyst layer on thesilicon band bending, we doped the surface of p-type silicondegenerately n-type (n+) to obtain an optimized and ‘fixed’ surfacedepletion region38. The surface of the n+–p-silicon substratewas further nanostructured into micropyramids to enhance lightharvesting39,40. Furthermore, the back of the silicon substrates wasdoped degenerately p-type (p+) to improve the majority carriercollection. Figure 4b,c shows top and cross-section views of aCoPS/n+–p–p+ micropyramid silicon photocathode prepared bythermal conversion of an evaporated 10-nm-thick cobalt filmat 450 ◦C. The PEC J–V curves of various CoPS/n+–p–p+-Siphotocathodes with different initial cobalt thicknesses (Fig. 4a)show the highest catalytic onset potential obtained with CoPS.The best CoPS/n+–p–p+-Si photocathode, which was prepared bythermal conversion of a 10 nm cobalt film, achieved a photocurrentdensity of 0.5mA cm−2 at a potential of 440mV versus RHE, aJSC of 26mA cm−2, and a fill factor of 40%. This correspondsto an overall Faradaic solar-to-hydrogen production efficiency of4.7%. These are the highest photovoltage onset and efficiencyreported so far for any Earth-abundant catalyst directly integratedwith p-silicon41–43 or n+–p silicon44,45. In comparison, a 5 nmPt/n+–p–p+-Si photocathode achieved 0.5mA cm−2 at a potential of530mVversus RHE and a JSC of 36mA cm−2. The difference in onsetvoltage of ∼80mV can be attributed to the difference in catalyticactivity between CoPS and Pt, whereas the lower JSC suggests that afraction of the incoming light was blocked by the CoPS catalyst layer(see Supplementary Fig. 5 for the optical absorbance spectrum ofsuch a CoPS layer). When the thickness of the cobalt precursor filmwas reduced to 7.5 and 5 nm, the JSC of the photocathode increasedto ∼33–35mA cm−2. However, the onset voltage also decreased,which suggests a higher light intensity reaching the semiconductorbut a less complete CoPS catalyst coverage. On the other hand, whena 15 nm cobalt layer was used as precursor, both the JSC and onsetvoltage decreased whereas the rectifying behaviour was preserved,suggesting that much less light reached the semiconductor owingto an increased light absorption from the thicker CoPS film. Thesetrends clearly show that optimization of the catalyst thickness andcoverage is important to balance the light absorption and catalyticperformance to maximize the PEC performance20,42.

In conclusion, we report pyrite-type CoPS as a novel high-performance Earth-abundant catalyst for robust electrochemicaland solar-driven hydrogen production. The electrocatalyticperformance of this novel pyrite-type ternary material can befurther enhanced through the synthesis of high-surface-area CoPSnanowires and nanoplates on conducting graphite or carbon fibrepaper substrates using a simple thermal conversion reaction.These CoPS nanostructures achieved stable catalytic performance

NATUREMATERIALS | VOL 14 | DECEMBER 2015 | www.nature.com/naturematerials 1249

© 2015 Macmillan Publishers Limited. All rights reserved

ARTICLES NATUREMATERIALS DOI: 10.1038/NMAT4410

200 nm2 µm

1 µm 200 nm

CoPS/p-Si(10 nm Co)

Dark

−0.2 −0.1 0.0 0.1 0.2 0.3 0.4 0.5

−50

−40

−30

−20

−10

0

15 nm

10 nm

7.5 nm

5 nm

3 nm

Pt film

J (m

A c

m−2

)

Potential (V) versus RHE

a

50 nm

b

c

Figure 4 | Photoelectrochemical hydrogen generation using integratedCoPS/Si photocathodes. a, J–V curves under dark and 1 Sun(100 mW cm−2, AM 1.5) illumination in 0.5 M H2SO4, obtained forthermally converted 3 nm, 5 nm, 7.5 nm, 10 nm and 15 nm cobalt filmscompared with 5 nm Pt film, on n+–p–p+ micropyramid silicon. A planarCoPS/p-silicon electrode obtained by thermal conversion of 10 nm Co onp-Si is shown in comparison. b,c, SEM images of a CoPS/n+–p–p+

micropyramid silicon photoelectrode obtained by conversion of 10 nm Cofilm, showing top (b) and cross-section (c) views of the surface.

superior to that reported for metal chalcogenides (especiallythe parent metal pyrites) and metal phosphides, and are amongthe most active Earth-abundant HER catalysts reported so far.These results illustrate that, in addition to a high concentration ofavailable catalytic sites, controlling the hydrogen adsorption energyof the active sites by tuning their electronic structure and reactivityby substituting non-metal atomic constituents in ternary or morecomplex compounds can serve as a new, general strategy to enhancethe electrocatalytic activity of transition metal compounds.Furthermore, CoPS-based photocathodes fabricated by directgrowth of CoPS on n+–p–p+ micropyramid silicon achievedcompetitive performance for solar-driven hydrogen production tothat of Pt/n+–p–p+-Si photocathodes. Our photoelectrochemicalstudies also show that both the work function and light absorptionof the surface catalyst layer in integrated devices are crucial factorsfor achieving efficient solar-driven hydrogen generation.

MethodsMethods and any associated references are available in the onlineversion of the paper.

Received 5 May 2015; accepted 30 July 2015;published online 14 September 2015

References1. Lewis, N. S. & Nocera, D. G. Powering the planet: Chemical challenges in solar

energy utilization. Proc. Natl Acad. Sci. USA 103, 15729–15735 (2006).2. Nocera, D. G. The artificial leaf. Acc. Chem. Res. 45, 767–776 (2012).3. Turner, J. A. Sustainable hydrogen production. Science 305, 972–974 (2004).4. Faber, M. S. & Jin, S. Earth-abundant inorganic electrocatalysts and their

nanostructures for energy conversion applications. Energy Environ. Sci. 7,3519–3542 (2014).

5. Lukowski, M. A. et al. Enhanced hydrogen evolution catalysis from chemicallyexfoliated metallic MoS2 nanosheets. J. Am. Chem. Soc. 135,10274–10277 (2013).

6. Benck, J. D., Hellstern, T. R., Kibsgaard, J., Chakthranont, P. & Jaramillo, T. F.Catalyzing the Hydrogen Evolution Reaction (HER) with molybdenum sulfidenanomaterials. ACS Catal. 4, 3957–3971 (2014).

7. Voiry, D. et al. Enhanced catalytic activity in strained chemically exfoliated WS2nanosheets for hydrogen evolution. Nature Mater. 12, 850–855 (2013).

8. Lukowski, M. A. et al.Highly active hydrogen evolution catalysis from metallicWS2 nanosheets. Energy Environ. Sci. 7, 2608–2613 (2014).

9. Merki, D., Fierro, S., Vrubel, H. & Hu, X. L. Amorphous molybdenum sulfidefilms as catalysts for electrochemical hydrogen production in water. Chem. Sci.2, 1262–1267 (2011).

10. Faber, M. S. et al.High-performance electrocatalysis using metallic cobalt pyrite(CoS2) micro- and nanostructures. J. Am. Chem. Soc. 136, 10053–10061 (2014).

11. Kong, D. S., Wang, H. T., Lu, Z. Y. & Cui, Y. CoSe2 nanoparticles grown oncarbon fiber paper: An efficient and stable electrocatalyst for hydrogenevolution reaction. J. Am. Chem. Soc. 136, 4897–4900 (2014).

12. Popczun, E. J., Read, C. G., Roske, C. W., Lewis, N. S. & Schaak, R. E. Highlyactive electrocatalysis of the hydrogen evolution reaction by cobalt phosphidenanoparticles. Angew. Chem. Int. Ed. 53, 5427–5430 (2014).

13. Liu, Q. et al. Carbon nanotubes decorated with CoP nanocrystals: A highlyactive non-noble-metal nanohybrid electrocatalyst for hydrogen evolution.Angew. Chem. 126, 6828–6832 (2014).

14. Popczun, E. J. et al. Nanostructured nickel phosphide as an electrocatalyst forthe hydrogen evolution reaction. J. Am. Chem. Soc. 135, 9267–9270 (2013).

15. Callejas, J. F. et al. Electrocatalytic and photocatalytic hydrogen productionfrom acidic and neutral-pH aqueous solutions using iron phosphidenanoparticles. ACS Nano 8, 11101–11107 (2014).

16. Xiao, P. et al.Molybdenum phosphide as an efficient electrocatalyst for thehydrogen evolution reaction. Energy Environ. Sci. 7, 2624–2629 (2014).

17. McEnaney, J. M. et al. Amorphous molybdenum phosphide nanoparticles forelectrocatalytic hydrogen evolution. Chem. Mater. 26, 4826–4831 (2014).

18. Kibsgaard, J. & Jaramillo, T. F. Molybdenum phosphosulfide: An active,acid-stable, earth-abundant catalyst for the hydrogen evolution reaction.Angew. Chem. Int. Ed. 53, 14433–14437 (2014).

19. McKone, J. R., Sadtler, B. F., Werlang, C. A., Lewis, N. S. & Gray, H. B. Ni–Monanopowders for efficient electrochemical hydrogen evolution. ACS Catal. 3,166–169 (2013).

20. McKone, J. R., Lewis, N. S. & Gray, H. B. Will solar-driven water-splittingdevices see the light of day? Chem. Mater. 26, 407–414 (2014).

21. Faber, M. S., Park, K., Cabán-Acevedo, M., Santra, P. K. & Jin, S.Earth-abundant cobalt pyrite (CoS2) thin film on glass as a robust,high-performance counter electrode for quantum dot-sensitized solar cells.J. Phys. Chem. Lett. 4, 1843–1849 (2013).

22. Faber, M. S., Lukowski, M. A., Ding, Q., Kaiser, N. S. & Jin, S. Earth-abundantmetal pyrites (FeS2, CoS2, NiS2, and their alloys) for highly efficient hydrogenevolution and polysulfide reduction electrocatalysis. J. Phys. Chem. C 118,21347–21356 (2014).

23. Kong, D. S., Cha, J. J., Wang, H. T., Lee, H. R. & Cui, Y. First-row transitionmetal dichalcogenide catalysts for hydrogen evolution reaction. Energy Environ.Sci. 6, 3553–3558 (2013).

24. Cabán-Acevedo, M. et al. Ionization of high-density deep donor defect statesexplains the low photovoltage of iron pyrite single crystals. J. Am. Chem. Soc.136, 17163–17179 (2014).

25. Ennaoui, A. et al. Iron disulfide for solar energy conversion. Sol. Energy Mater.Sol. Cells 29, 289–370 (1993).

26. Stiebritz, M. T. & Reiher, M. A unifying structural and electronic concept forHmd and [FeFe] hydrogenase active sites. Inorg. Chem. 49, 5818–5823 (2010).

27. Thauer, R. K. et al.Hydrogenases from methanogenic archaea, nickel, a novelcofactor, and H2 storage. Annu. Rev. Biochem. 79, 507–536 (2010).

1250 NATUREMATERIALS | VOL 14 | DECEMBER 2015 | www.nature.com/naturematerials

© 2015 Macmillan Publishers Limited. All rights reserved

NATUREMATERIALS DOI: 10.1038/NMAT4410 ARTICLES28. Kubas, G. J.Metal Dihydrogen and s-Bond Complexes: Structure, Theory, and

Reactivity (Springer Science & Business Media, 2001).29. Ishida, K. & Nishizawa, T. The Co–P (cobalt–phosphorus) system. Bull. Alloy

Phase Diagr. 11, 555–560 (1990).30. Hulliger, F. New compounds with cobaltite structure. Nature 198,

382–383 (1963).31. Nørskov, J. K. et al. Trends in the exchange current for hydrogen evolution.

J. Electrochem. Soc. 152, J23–J26 (2005).32. Xia, X. et al. Controllable growth of conducting polymers shell for constructing

high-quality organic/inorganic core/shell nanostructures and theiroptical–electrochemical properties. Nano Lett. 13, 4562–4568 (2013).

33. Lyapin, S. G. et al. Raman studies of nearly half-metallic ferromagnetic CoS2.J. Phys. Condens. Matter 26, 396001 (2014).

34. Moulder, J. F., Stickle, W. F., Sobol, P. E. & Bomben, K. D. Handbook of X-rayPhotoelectron Spectroscopy (Perkin-Elmer, 1992).

35. Biesinger, M. C., Lau, L. W. M., Gerson, A. R. & Smart, R. S. C. Resolvingsurface chemical states in XPS analysis of first row transition metals, oxides andhydroxides: Sc, Ti, V, Cu and Zn. Appl. Surf. Sci. 257, 887–898 (2010).

36. Okamoto, H. P–S (phosphorus–sulfur). J. Phase Equilib. 12, 706–707 (1991).37. Dai, P. et al. Solar hydrogen generation by silicon nanowires modified with

platinum nanoparticle catalysts by atomic layer deposition. Angew. Chem. Int.Ed. 52, 11119–11123 (2013).

38. Boettcher, S. W. et al. Photoelectrochemical hydrogen evolution using Simicrowire arrays. J. Am. Chem. Soc. 133, 1216–1219 (2011).

39. Wang, H.-P. et al. Realizing high-efficiency omnidirectional n-type Si solar cellsvia the hierarchical architecture concept with radial junctions. ACS Nano 7,9325–9335 (2013).

40. Kelzenberg, M. D. et al. Enhanced absorption and carrier collection in Si wirearrays for photovoltaic applications. Nature Mater. 9, 239–244 (2010).

41. Zhang, X. et al. Amorphous MoSxCly electrocatalyst supported by verticalgraphene for efficient electrochemical and photoelectrochemical hydrogengeneration. Energy Environ. Sci. 8, 862–868 (2015).

42. Ding, Q. et al. Efficient photoelectrochemical hydrogen generation usingheterostructures of Si and chemically exfoliated metallic MoS2. J. Am. Chem.Soc. 136, 8504–8507 (2014).

43. Huang, Z. et al. Ni12P5 nanoparticles as an efficient catalyst for hydrogengeneration via electrolysis and photoelectrolysis. ACS Nano 8,8121–8129 (2014).

44. Hou, Y. et al. Bioinspired molecular co-catalysts bonded to a siliconphotocathode for solar hydrogen evolution. Nature Mater. 10,434–438 (2011).

45. Benck, J. D. et al. Designing active and stable silicon photocathodes for solarhydrogen production using molybdenum sulfide nanomaterials. Adv. EnergyMater. 4, 1400739 (2014).

AcknowledgementsThis research is supported by the US Department of Energy, Office of Basic EnergySciences, Division of Materials Sciences and Engineering, under AwardDE-FG02-09ER46664. M.C.-A. thanks the NSF graduate Research Fellowship forsupport. J.R.S. is supported by the National Science Foundation Grant No. CHE-1362136for the theoretical work here. H.-C.C., M.-L.T. and J.-H.H. are supported by KAUSTbaseline fund for design and fabrication of light-harvesting Si substrates.

Author contributionsM.C.-A. and S.J. designed the experiments. M.C.-A. and M.L.S. carried out the synthesisof CoPS nanomaterials and electrochemical measurements. J.R.S. performed the densityfunctional calculations and computational modelling. M.C.-A. and J.G.T. carried out thefabrication of photocathodes and photoelectrochemical measurements. Q.D. contributedto the photoelectrochemical performance optimization. H.-C.C., M.-L.T. and J.-H.H.designed and fabricated the n+–p–p+ micropyramid silicon substrates. M.C.-A.performed the structural characterization. M.C.-A. and S.J. wrote the manuscript and allauthors commented on the manuscript.

Additional informationSupplementary information is available in the online version of the paper. Reprints andpermissions information is available online at www.nature.com/reprints.Correspondence and requests for materials should be addressed to S.J.

Competing financial interestsThe authors declare no competing financial interests.

NATUREMATERIALS | VOL 14 | DECEMBER 2015 | www.nature.com/naturematerials 1251

© 2015 Macmillan Publishers Limited. All rights reserved

ARTICLES NATUREMATERIALS DOI: 10.1038/NMAT4410

MethodsFree-energy calculations for atomic hydrogen adsorption (1GH∗ ).Chemisorption free energies were obtained for both CoS2 and CoPS using densityfunctional theory (DFT) calculations. All calculations were performed using theVienna Ab-initio Simulation Package (VASP; refs 46–48) using thePerdew–Burke–Ernzerhof (PBE) generalized gradient approximation (GGA)exchange–correlation functional and the projector augmented wave (PAW)method49,50 to describe the interaction between core electrons and valenceelectrons. All calculations were run with an energy cutoff of 280 eV, normalprecision, Methfessel–Paxton smearing, and accounting for spin polarization. TheBrillouin zone was sampled with a 9×9×1 Monkhorst–Pack grid. Theexperimentally reported crystal structure of CoPS consists of a S and P disorderedlattice (a=5.422Å, ICSD collection code #62414; ref. 30). A simple model of CoPSwas obtained by substituting each dumbbell within the unit cell with P and S atomsand adjusting the lattice constant to match the experimental value of CoPS.Chemisorption was modelled on the CoS2 (a=5.5388Å, ICSD collection code#86351) and CoPS {100} surfaces, while ensuring a stable surface terminationwhere the dichalcogenide dumbbells remain intact. The resulting slab modelconsists of four layers, separated by an approximately 10Å vacuum gap. For theresulting surface unit cell, adsorption of a single H atom onto Co corresponds tohalf coverage. Chemisorption energies of atomic hydrogen were calculated relativeto H2 (g) by:

1E=E(surf+nH)−

(E(surf)+

12E(H2)

)All structures were relaxed to a tolerance of less than 0.1 eVÅ−1. The associated freeenergy of chemisorption was then calculated by correcting for both the zero-pointvibrational energy and the loss of translation entropy of H2 (g) on adsorption, andneglecting the smaller vibrational entropy terms. The zero-point contributions forCoS2 and CoPS are essentially identical, yielding 1G=1E+0.29 eV.

Synthesis of cobalt precursor materials. Co films. The Co films were preparedon graphite disk substrates (6.0mm (d)× 1mm (t), Ultra Carbon, Ultra ‘F’Purity) and borosilicate glass substrates by electron-beam evaporating100 nm Co (K. J. Lesker, 99.95%) onto the substrate at a 1Å s−1deposition rate.

CHCH nanowires (NWs). The graphite substrates were first treated byannealing in air at 800 ◦C for 10min to improve the surface hydrophilicity. Thencobalt hydroxide carbonate hydrate (CHCH, Co(OH)(CO3)0.5 ·xH2O) NWs weresynthesized on these graphite substrates by following a published procedure10 withminor modifications. In a typical synthesis, 1.5mmol of Co(NO3)2 ·6H2O, 3mmolNH4F (Sigma-Aldrich, ≥98.0%), and 7.5mmol (NH2)2CO (Riedel-de Haën,99.5–100.5%) were dissolved in 50ml of distilled water. Then, 16ml of this solutionwas transferred to a 23-ml PTFE-lined stainless steel autoclave containing thegraphite substrate, which was sealed and heated at 110 ◦C for 5 h. After cooling, thesubstrate was removed, rinsed with ethanol and water, and dried under a streamof nitrogen.

CHCH nanoplates (NPls). Cobalt hydroxide carbonate hydrate(Co(OH)(CO3)0.5 ·xH2O, ‘CHCH’) nanoplates (NPls) were synthesized on carbonfibre paper (Toray Paper 060 from Fuel Cell Earth, Teflon treated, thickness0.19mm) that has been treated by annealing in air at 800 ◦C for 10min to make ithydrophilic. The NPls were synthesized in a two-step process. In the first step,cobalt oxide (CoO) nanoparticles were deposited on the substrate (1 cm× 3 cm)following a published procedure11 with slight modifications. In a typical synthesis,0.4 g cobalt (II) nitrate hexahydrate and 0.05 g polyvinylpyrrolidone (PVP) weredissolved in 0.55 g dimethylformamide (DMF) under vigorous stirring at 80 ◦C for2 h to make an ink. A treated carbon paper substrate was dip-coated in the ink andthen dried under a stream of nitrogen. The substrate was then placed in the centreof a fused silica tube (1 inch outer diameter) heated to 600 ◦C in a tube furnace(Lindberg/Blue M) at a pressure of 100mtorr and a 50 sccm argon flow rate. In thesecond step, the cobalt oxide nanoparticle-coated carbon fibre paper was used asthe substrate in the CHCH synthesis described above to result in the formation ofNPls instead of NWs.

Synthesis of CoPS nanostructures. An alumina boat containing 1 g of a 1:1mixture of sulphur (Sigma-Aldrich, 99.5–100.5%) and phosphorus (Alfa Aesar,98.9%) powders was covered with a piece of glass, then placed in the centre of afused silica tube reactor equipped with pressure and gas flow controllers, andheated in a tube furnace (Lindberg/Blue M). The Co film, CHCH NW and CHCHNPl precursor substrates were placed on an alumina plate (4mm thick) at thedownstream end of the tube, outside the heating area of the furnace. After the tubewas purged with Ar carrier gas (99.999%) at 25 sccm and 780 Torr, the furnace wasfirst heated to 200 ◦C for 10min, then opened and allowed to cool naturally tomake a thiophosphate (PxSy ) paste-like product in the alumina boat. Then, withoutopening the tube to the atmosphere, by manipulating a magnetic stir bar inside thereactor, the cover was taken off the alumina thiophosphate precursor boat, the boatwas moved to the upstream edge of the furnace, and the cobalt precursor substrates

were moved to the centre of the furnace. The furnace was then heated to 500 ◦Cfor 1 h to convert these precursors to CoPS before the furnace was cooleddown naturally.

Synthesis of CoPx , CoPx|H2S, CoPx|S, CoS2|P and CoS2|P+H2 samples. CoPx

film and NWs.We achieved the synthesis of highly catalytic CoPx material with apredominantly CoP composition and a minor CoP2 phase through thermalconversion of 100-nm-thick Co film on borosilicate glass or graphite disks andCHCH NWs on graphite disks at 500 ◦C for 45min in a phosphorus (0.3 g of redphosphorus in an alumina crucible at∼425–450 ◦C) and hydrogen gas (25 sccm H2

and 25 sccm argon gas) atmosphere.CoPx |H2S and CoPx |S film and NWs. The CoPx |H2S and CoPx |S film and NW

samples were prepared by thermal annealing of CoPx film and CoPx NWs onborosilicate glass or graphite disks at 500 ◦C for 20min in a sulphur (saturated)or hydrogen sulphide (25 sccm 10% H2S in helium and 25 sccm argongas) atmosphere.

CoS2|P and CoS2|P+H2 film and NWs. The CoS2|P and CoS2|P+H2 film orNW samples were prepared by thermal annealing of CoS2 film and CoS2 NWs onborosilicate glass or graphite disks10 at 500 ◦C for 20min in a phosphorus (0.3 g ofred phosphorus in an alumina crucible at∼425–450 ◦C) or phosphorus andhydrogen gas (25 sccm H2 and 25 sccm argon gas) atmosphere.

Structural characterization. The as-prepared CoPS film and CoPS NWs ongraphite substrates, and CoPS NPls on carbon fibre paper and other comparisonsamples were characterized using a LEO SUPRA 55 VP field-emission scanningelectron microscope (SEM) with energy dispersive X-ray spectroscopy (EDS)capabilities, and a Bruker D8 ADVANCE powder X-ray diffractometer (PXRD)using Cu Kα radiation. Note that the PXRD patterns for CoPS film and variousother thin film samples were taken using borosilicate glass as substrate. Ramanspectra were taken using a Thermo Scientific DXR confocal Raman microscopeusing a 532 nm excitation laser. High-resolution XPS measurements for Co 2p weretaken using a Thermo Al Kα XPS with a 180◦ double-focusing hemisphericalanalyser and 128-channel detector (60◦ angular acceptance), which under effectiveoperating conditions had an analyser resolution of 0.8 eV. All X-ray photoelectronspectra were shifted so that the adventitious carbon C1s peak was at 284.8 eV. S 2ppeaks were fitted using doublets with a 1:0.5 (3/2p:1/2p) area ratio, 1.18 eV apart,and with the same full-width at half-maximum (FWHM). P 2p were fitted usingdoublets with a 1:0.5 (3/2p:1/2p) area ratio, 0.87 eV apart, and with thesame FWHM.

Electrochemical characterization of catalytic activity towards the HER. Allelectrochemical measurements were performed in a three-electrode configurationusing a rotating disk electrode (RDE) set-up (Bioanalytical Systems; RDE-2) andrecorded using a Bio-Logic SP-200 potentiostat. Each measurement was performedin 0.5M H2SO4 (aq) electrolyte continuously purged with H2 (g) (99.999%) using asaturated Hg/HgSO4 reference electrode (CH Instruments), a graphite rod(National Carbon, AGKSP Spectroscopic Electrode) as the counter electrode, and aCoPS film, NW, or NPl or other sample substrate affixed to a glassy carbon RDE tipusing silver paint (Ted Pella, PELCO colloidal silver) as the working electrode.Graphite paint (Ted Pella, PELCO isopropanol-based graphite paint) was furtherused to isolate the silver paint contact from the electrolyte solution. Linear sweepvoltammograms were measured from the open-circuit voltage at a scan rate of2mV s−1, while the working electrode was rotated at 2,000 r.p.m. The Hg/HgSO4

reference electrode was calibrated against the reversible hydrogen potential (RHE)using a platinum wire (K. J. Lesker, 99.99%; 0.50mm diameter) as the workingelectrode and a platinum mesh as the counter electrode after each measurement.The Pt reference trace was recorded using a Pt wire as the working electrode. Allpolarization curves were corrected for iR losses unless otherwise noted.Electrochemical impedance spectroscopy was performed in potentiostatic mode at0V versus RHE, applying a sinusoidal voltage with an amplitude of 10mV andscanning frequency from 250 kHz to 1mHz.

Fabrication of n+–p–p+ micropyramid silicon coated with CoPS films.Micropyramid silicon arrays were fabricated on both sides of 150-µm-thick p-type(100) Si wafers (dopant concentration of 5×1015 cm−3) by electrodeless chemicaletching in a solution of potassium hydroxide (KOH, 45 vol.%) and isopropylalcohol (IPA). 300 nm of n+ emitter layer (dopant concentration of 9×1019 cm−3)was formed by the thermal diffusion processes of POCl4 at 1,000 ◦C. 300 nm of p+back surface field layer (dopant concentration of 3×1020 cm−3) was fabricated byscreen-printed Al annealed at 500 ◦C. Cobalt films of varying thicknesses weredeposited on planar p-type silicon (resistivity of 1–2.5� cm, B-doped, (100)orientation, prime grade, 525 µm thickness) and n+–p–p+ micropyramid siliconsubstrates by electron-beam evaporating cobalt (K. J. Lesker, 99.95%) at a 0.1 Å s−1deposition rate. Then the cobalt films were converted to CoPS using the proceduredescribed above, except that an alumina boat containing 200mg of 1:1 mixture ofsulphur and phosphorus powders was used and, in the last step, the furnace washeated to 450 ◦C for 10min.

NATUREMATERIALS | www.nature.com/naturematerials

© 2015 Macmillan Publishers Limited. All rights reserved

NATUREMATERIALS DOI: 10.1038/NMAT4410 ARTICLESPhotoelectrochemical characterization. The photoelectrochemicalcharacteristics were measured in a three-electrode configuration undersimulated 1 Sun irradiation (100mWcm−2) supplied by a 1 kW Xe lampsolar simulator (Newport Model 91191; AM 1.5G filter) using a Bio-LogicSP-200 potentiostat. The light intensity was calibrated with a Si photodiode(Thorlabs) to generate a photocurrent equal to that at 100mWcm−2 lightintensity. All measurements were performed in 0.5M H2SO4 electrolyteconstantly purged with H2 gas (99.999%), using a graphite rod (National Carbon,AGKSP Spectroscopic Electrode) as the counter electrode, and a saturatedHg/HgSO4 reference electrode (CH Instruments). The Hg/HgSO4 referenceelectrode was calibrated against RHE using a platinum wire as the workingelectrode and a platinum mesh as the counter electrode after each measurement.The electrolyte was vigorously stirred to minimize mass-transport limitations andremove accumulated hydrogen gas bubbles on the electrode surface. The currentdensity versus potential (J–V ) curves were measured at a scan rate of 10mV s−1

and were not corrected for any uncompensated resistance losses or any otherextrinsic losses.

References46. Kresse, G. & Furthmüller, J. Efficient iterative schemes for ab initio total-energy

calculations using a plane-wave basis set. Phys. Rev. B 54, 11169–11186 (1996).47. Kresse, G. & Hafner, J. Ab initiomolecular-dynamics simulation of the

liquid-metal-amorphous-semiconductor transition in germanium. Phys. Rev. B49, 14251–14269 (1994).

48. Kresse, G. & Hafner, J. Ab initiomolecular dynamics for liquid metals. Phys.Rev. B 47, 558–561 (1993).

49. Blöchl, P. E. Projector augmented-wave method. Phys. Rev. B 50,17953–17979 (1994).

50. Kresse, G. & Joubert, D. From ultrasoft pseudopotentials to the projectoraugmented-wave method. Phys. Rev. B 59, 1758–1775 (1999).

NATUREMATERIALS | www.nature.com/naturematerials

© 2015 Macmillan Publishers Limited. All rights reserved