7/27/2019 La 3032489

1/13

Transition Metal Sulfide Hydrogen Evolution Catalysts forHydrobromic Acid Electrolysis

Anna Ivanovskaya,*, Nirala Singh, Ru-Fen Liu, Haley Kreutzer, Jonas Baltrusaitis,

Trung Van Nguyen, Horia Metiu, and Eric McFarland*,

Department of Chemistry and Biochemistry and Department of Chemical Engineering, University of California, Santa Barbara,California 93106, United StatesDepartment of Chemical and Petroleum Engineering, The University of Kansas, Lawrence, Kansas 66045, United StatesPhotoCatalytic Synthesis Group, MESA+ Institute for Nanotechnology, Faculty of Science and Technology, University of Twente,Meander 225, P.O. Box 217, 7500 AE Enschede, The Netherlands

*S Supporting Information

ABSTRACT: Mixed metal sulfides containing combinations of W, Fe, Mo,

Ni, and Ru were synthesized and screened for activity and stability for thehydrogen evolution reaction (HER) in aqueous hydrobromic acid (HBr).Co- and Ni-substituted RuS2 were identified as potentially active HERelectrocatalysts by high-throughput screening (HTS), and the specificcompositions Co0.4Ru0.6S2 and Ni0.6Ru0.4S2 were identified by optimization.Hydrogen evolution activity of Co0.4Ru0.6S2 in HBr is greater than RuS2 orCoS2 and comparable to Pt and commercial RhxSy. Structural andmorphological characterizations of the Co-substituted RuS2 suggest thatthe nanoparticulate solids are a homogeneous solid solution with a pyritecrystal structure. No phase separation is detected for Co substitutions below30% by X-ray diffraction. In 0.5 M HBr electrolyte, the CoRu electrodematerial synthesized with 30% Co rapidly lost approximately 34% of theinitial loading of Co; thereafter, it was observed to exhibit stable activity forHER with no further loss of Co. Density functional theory calculations indicate that the S2

2 sites are the most important forHER and the presence of Co influences the S

2

2 sites such that the hydrogen binding energy at sufficiently high hydrogencoverage is decreased compared to ruthenium sulfide. Although showing high HER activity in a flow cell, the reverse reaction ofhydrogen oxidation is slow on the RuS2 catalysts tested when compared to platinum and rhodium sulfide, leaving rhodium sulfideas the only suitable tested material for a regenerative HBr cell due its stability compared to platinum.

1. INTRODUCTION

The electrolysis of hydrohalic acids, in particular HBr and HCl,has several potential applications for the production ofhydrogen,13 production or recovery of halogens,37 and inreversible flow cells for energy storage.813 In energy-storageapplications, HBr electrolysis is particularly attractive becausethe relatively low decomposition voltage14,15 of HBr is belowthe oxygen evolution potential and the reactions at both the

hydrogen and bromine electrodes are relatively fast.13,16Further, bromine is earth-abundant, a liquid at room temper-ature, and readily separated from hydrogen.15,16

An aqueous solution of HBr and Br2 (red acid) is anextremely corrosive electrolyte, and the identification ofelectrode materials that are electrochemically active and stablein it is challenging. In previously developed electrolyticcells,14,16 the kinetics of the hydrogen evolution reaction(HER) have been shown to be limiting, and improvedhydrogen electrode materials are required. The rates can beincreased using high-surface-area metallic nanoparticulatecatalysts, but unfortunately, bromine is known to bind strongly

to metals, and the catalysts are unstable with respect tocorrosion in HBr.14 There has been considerable interest infinding materials for electrocatalysts that are sufficiently stableto allow the utilization of nanodispersed particles to maintainhigh reaction rates.17,18

There are indications that transition metal sulfides (TMSs)might provide active and stable electrocatalytic materials.Commercially, rhodium sulfide is known to be a stable and

active cathode catalyst for oxygen-reduction-assisted electrolysisof hydrochloric acid.19,20

Density functional theory (DFT) calculations have beenhelpful in assessing the HER thermodynamics on variouselectrodes. The free energy of hydrogen bound to a surfaceatom can be calculated by DFT and is commonly used as apredictor for HER activity.21 If the adsorption free energy ofhydrogen to the surface is small, the activity is high, because

Received: August 10, 2012Revised: November 9, 2012Published: December 3, 2012

Article

pubs.acs.org/Langmuir

2012 American Chemical Society 480 dx.doi.org/10.1021/la3032489 | Langmuir2013, 29, 480492

http://localhost/var/www/apps/conversion/tmp/scratch_3/pubs.acs.org/Langmuirhttp://localhost/var/www/apps/conversion/tmp/scratch_3/pubs.acs.org/Langmuir

7/27/2019 La 3032489

2/13

hydrogen (H) can adsorb well and can be removed from thesurface as molecular hydrogen (H2) without a high activationenergy.21 TMSs show activities and hydrogen adsorption freeenergies similar to the more active metal catalysts.22,23 Thehydrogen adsorption energy to MoS2 is 0.1 eV,

24 compared to0.1 eV for Pt at the same hydrogen coverage,21 predicting thatMoS2 and Pt would have comparable activities for hydrogenevolution. The hydrogen evolution rates on MoS2 nanoparticlesare slightly below those of the most-active noble metal catalysts,indicating that the theoretical predictions can be reliable and

the calculated hydrogen binding energy is a good indicator ofHER activity for metal sulfides.22

It is possible to optimize the catalytic activity of the industrialhydrodesulfurization sulfide catalysts by substituting them withCo or Ni2527 to modify the properties of the surface. It ishoped that substitution may decrease the activation barrier forhydrogen activation and change the activity for hydrogenevolution.

There are two specific families of metal sulfides that areworth focusing on for hydrogen electrocatalysts in aqueousHBr electrolyte due to activity, stability, and the ability tosynergistically interact with substituted metals. The transitionmetals (Ru and 3d transition metals from Mn to Zn) formdisulfides with pyrite structure. Mixed metal pyrites can be

synthesized in bulk as solid solutions for a large range ofcompositional variations.28,29 Pyrites are conductive semimetalsor narrow gap semiconductors,29 which is necessary forelectrocatalysis. Mo and W disulfides are also interesting.MoS2 is an active catalyst for HER, and Co and Ni promoteHER activity of nanoparticulate carbon (or SiO2) supportedMo and W disulfides.30,31

This background prompted us to investigate the stability andactivity of substituted TMSs as hydrogen evolution electro-catalysts for hydrobromic acid electrolysis. We attempt here toanswer the following questions:

(1) Can the hydrogen evolution activity of known TMSs beincreased by substitution with a second metal species,and can we use high-throughput experimentation

methods to rapidly identify candidate mixed metalsulfides?

(2) What are the quantified activities of the most-activeTMSs identified by high-throughput screening, and howdo they compare to industrially used hydrogen electrodematerials?

(3) What are the stabilities of the metal sulfides in thecorrosive conditions of HBr that would be typical of anelectrolyzer during both operation and shutdown, andwhat is the mechanism of corrosion?

(4) How does the hydrogen electrode based on TMSsperform in a H2H2 flow cell?

(5) Can DFT be helpful in understanding the experimental

observations?

2. METHODS

2.1. Preparation of Catalysts, Inks, and Electrodes. VulcanX72 high-surface-area carbon was impregnated with an aqueoussolution of metal salts (4.3 mL of 1 M solution per 1 g of carbon for allsolutions, or an equivalent volume of Mo and W precursors to load 4.3mmol of metal per gram of carbon). Wet powders were dried in the

furnace at 100 C for 1 h. The dried powders were exposed to 50%H2S (in N2) in a tube furnace reactor using a temperature profile of 10C/min ramp followed by a 3 h dwell at 300 C prior to cooling toroom temperature. Samples were synthesized in small ceramiccrucibles; 7 mg of carbon was used for one sample preparation. Upto 48 samples were prepared in the furnace simultaneously.

The metal precursors used for the synthesis of TMS catalysts (eitheras pure monometallic sulfides or as metal sulfides with substitutions)are listed in Table 1 along with the ratios of sulfur to metal determined

by energy dispersive X-ray spectroscopy (EDS) and the structuredetermined by X-ray diffraction (XRD).

Inks were made from a suspension of 3 mg of catalyst, 5 mL of a 1:1vol % mixture of 2-propanol (HPLC grade, Fisher Scientific) anddeionized water, and 17.5 L of a 5 wt % Nafion solution (Aldrich).The inks were sonicated in an ultrasonic bath at room temperature for2 h and distributed on 0.05 cm2 glassy carbon supports. For high-

throughput screening, 3 L of the inks were distributed over theelectrode substrates on glassy carbon (see details on substratepreparation in the Supporting Information); catalyst loading was 36g/cm2.

Inks were applied to a 0.178 cm2 area glassy carbon disk (PineInstrument Co.) from two 8 L aliquots and dried in ambient air aftereach application for the RDE experiments. The nominal loadings forall RDE experiments were 54 g/cm2 (geometric).

Electrodes on Toray carbon paper (Fuel Cell Store Inc.) wereprepared by dispersing catalyst inks on rectangular strips 1 cm 3 cm.Concentrated inks were prepared by dispensing 6 mg of catalyst in 1mL of a 1:1 vol % mixture of 2-propanol and deionized water, and 35L of a 5 wt % Nafion solution. An area of 2 cm2 was covered with0.164 mL of ink for a catalyst loading of 0.5 mg/cm2. The electrodesfor the O2/Ar/air samples were prepared in the same manner as thatfor the 48 h corrosion test.

2.2. High-Throughput Screening for Hydrogen EvolutionActivity. Prior to taking images, the array of electrocatalysts(library) was tested in 0.5 M HBr solution, purged with argon for10 min, and conditioned by performing five cycles of potential sweepsfrom 0 to 0.4 and back to 0 V versus a standard hydrogen electrode(SHE). Images of the library were taken using a color CCD cameraSPOT insight color digital camera (Diagnostic Instruments, Inc.)controlled by LabVIEW software developed by the authors. Anoverhead projector was used as a light source to enhance the digitalimages of the library. The potential of the working electrode wascontrolled relative to a Ag/AgCl reference electrode during themeasurements (later recalculated to the potential relative to a SHE)and varied linearly from 0 to 0.325 V at a rate of 5 mV/s, confirmed

by a calibration performed immediately prior to the test. A diagram of

Table 1. Precursors, Atomic Ratios of Sulfur to Metal ([S]/[M]) in Synthesized TMSs as Determined by EDS, and Structure asDetermined by XRD for Metal Sulfides Prepareda

metal precursor [S]/[M] structure metal precursor [S]/[M] structure

Cr Cr(NO3)39H2O Cu CuCl2

Mn MnCl2 Mo H3PMo12O40 2.0 H

Fe FeCl36H2O 1.9 P Ru RuCl3 1.9 P

Co CoCl26H2O 1.7 P Rh RhCl3(H2O)x

Ni NiCl26H2O 2.5 P W H3PW12O40 2.3 HaStandard deviation of atomic ratios determined by EDS varied from 5% to 20% of the average values. In the structure type column, pyrite andhexagonal type structures are denoted by P and H, respectively.

Langmuir Article

dx.doi.org/10.1021/la3032489 | Langmuir2013, 29, 480492481

7/27/2019 La 3032489

3/13

the experimental system along with a more detailed explanation of thesetup is shown in the Supporting Information, Figure S1.

2.3. Electrochemical Test Methods. A Pine Instrument MSRXrotating disk electrode was used to measure the activity of the metalsulfides in a two-compartment electrochemical cell with cathode andanode compartments separated by a silica frit diaphragm and a Nafionmembrane. The counter electrode was a platinum mesh, and thereference electrode was Ag/AgCl. Cyclic voltammetry was performed

using a EG&G 270A potentiostat in argon-saturated 0.5 M HBrsolution.2.4. H2H2 Regenerative Cell Test. A 2.25 cm

2flow battery

consisting of Al end plates, stainless steel current collectors, graphiteflow distributors, and a six-layer (cathode-GDL/CL/membrane/CL/MPL/GDL-anode) membrane electrode assembly (MEA) made byTVN Systems Inc. was used with a Nafion 112 membrane. Thecathode contained 0.75 mg of Pt catalyst/cm2 loading (30% Pt on

Vulcan 72X), and the anode contained 0.75 mg of catalyst/cm2

loading of RuS2, Co0.3Ru0.7S2, RhxSy, or Pt. Interdigitated flow fieldswere used to improve mass transport to the electrodes.32 The cell wastested at room temperature (22 C). Hydrogen at 3 psig wascontinuously circulated through the cell both at the anode and cathodeat 2206 mL/min, so that the oxidation reaction was hydrogenoxidation and the reduction reaction was hydrogen evolution.Changing the polarity of the electrodes reversed which electrode

was acting as the cathode and which was acting as the anode. Apotentiostat/galvanostat (Arbin Instruments) was used to control thecell potential. The data reported represent average values collectedover 2 min.

2.5. Analysis of Corrosion Products. 2.5.1. Measurement ofDissolution Products. Dissolution product concentrations weremeasured by inductively coupled plasma atomic emission spectroscopy(ICP-AES) calibrated using standards obtained from High-PurityStandards (cat. no. 100013-2 for Co and no. 100046-2 for Ru).

2.5.2. Measurement of Gaseous Products. A Stanford ResearchSystems residual gas analyzer (SRS RGA) model 200 massspectrometer was used to measure the concentrations of gas due tocorrosion by injecting 0.251 mL by volume of gas into thespectrometer and measuring the response of m/z 34.

2.5.3. Qualitative Chemical Test for Hydrogen Sulfide andSulfate Ions. To test whether corrosion produced hydrogen sulfide,

350 L of the 6 M HBr solution was neutralized with 350 L of 6 MNaOH; then 23 mg of lead nitrate (corresponding to 0.1 M leadnitrate in solution) was added.

A standard analytic test33 for detection of sulfate ions in the testsolutions was used. Ba(NO3)2 powder (8 mg) was added to 500 L ofthe test solution, and the presence of sulfate ions (as HSO 4

) wouldbe confirmed by the formation of white precipitate (BaSO4). A controlsolution with sulfate ions verified the test, and a control solution ofH2S did not form a white precipitate, indicating selectivity of the testfor sulfate. Ba(NO3)2 may also react with sulfite (SO3

2) to form awhite precipitate, but SO3

2 is unstable in acidic media and forms SO2.2.6. Structure, Composition, and Morphology Character-

ization. XRD data were collected on an XPert powder diffractometer(PANalytical, Inc.) with a Cu K source (corresponding to a photon

wavelength of 1.54 ). For the measurement of lattice parameters, theposition of the (200) diffraction peak of RuS2 was preciselydetermined against a known Si(111) peak position from Si mixed

with the sample.Scanning and transmission electron microscopy (SEM and TEM) of

the powder electrocatalysts was performed with a FEI XL40 SirionFEG digital scanning microscope with an EDS system and a FEI T20electron microscope operating at 200 keV. TEM samples wereprepared by suspending the sample in ethanol and then using amicropipet to deposit the sample on a polycarbon grid.

High-resolution and survey XPS scans were collected using a Kratosaxis ultra spectrometer (Kratos Analytical, Manchester, UK) equipped

with a monochromated 1486.6 eV aluminum K source having a 500mm Rowland circle silicon single-crystal monochromator. All bindingenergies (BE) were referenced to carbon black (conductive carbon) C1s at 284.4 eV. Data processing and quantification was done using

commercially available CasaXPS.34 Further details on the XPSmeasurements are included in the Supporting Information.

2.7. Computation Methods. For the DFT35 calculations, we usedspin-polarized, generalized gradient approximation with the PerdewBurkeErnzerhof (PBE) functional.36 The core electrons weredescribed by the projector augment-wave (PAW) method37

implemented by Kresse and Joubert38 in the VASP 4.6 program.The energy cutoff for plane-wave expansion was set to 350 eV. No

symmetry was imposed during structure relaxation. A correction to thetotal energy to remove artificial dipole effects was included.39 The slabsused in the calculations consist of three stoichiometric layers. Theatomic positions in the bottom layer were fixed at the bulk positionsthat were calculated by using a 5 5 5 k-grid for the cubic pyrite

bulk structure. In all calculations, the positions of the adsorbates and ofthe atoms in the top two stoichiometric layers were obtained byminimizing the total energy without symmetry constraints. In all cases,

we examined several spin states and report here the ones that have thelowest energy. The geometry optimization was considered satisfactory

when the largest force on an individual atom was less than 20 meV/.

3. RESULTS

3.1. Synthesis of Mixed Metal Sulfides Based on W,Fe, Mo, Ni, and Ru. 3.1.1. Synthesis of Pyrite Type TMSs.

The sulfides synthesized from Ru, Fe, and Ni precursors have apyrite structure (Supporting Information,Figures S2S4). EDSanalysis shown in Table 1 indicates that the sulfur-to-metalmolar ratios in Ru and Fe sulfides are c lose to thestoichiometric value of two for disulfides. Ni sulfide is sulfur-rich.

The synthesis reaction starting from Ru (or Fe) chloride is asfollows:

+ + +RuCl 4H S 2RuS 6HCl H3 2 2 2

Experimental observation suggested that ruthenium sulfidesynthesis is very sensitive to the synthesis temperature:increasing the temperature from 300 to 350 C resulted intwo-phase crystalline compounds (verified by XRD): RuS2 and

Ru metal. Analysis of the reactive gas mixture by differentialmass spectrometry showed the presence of about 6% H2 in theH2S/N2 gas source cylinder at room temperature. Rutheniumchloride may react with hydrogen according to the followingequation:

+ +RuCl 3H 2Ru 6HCl3 2

Gas product analysis by differential mass spectrometry did notshow chlorine gas in the mixture.

The synthesis reaction starting from Ni chloride is as follows:

+ + +NiCl 2H S NiS 2HCl H2 2 2 2

3.1.2. Morphological and Structural Change with

Substitution Concentration in Pyrites. Substitution with Feand Ni resulted in formation of pyrite structures with no othercrystalline phases present in XRD. All pyrite peak positionswere shifted to the higher (Fe) or lower (Ni) 2angles with anincrease in substitution concentration, in agreement with thelattice parameter variation in mixed pyrite solid solution(labeled FexRu1xS2 theoretical and NixRu1xS2 theoretical,respectively, in the Supporting Information, Figure S5). Similarto Vegards law for alloys, pyrites can form a continuous rangeof solid solutions where the lattice parameter changes linearlywith the substitution concentration.29 Solid solutions wereformed for FexRu1xS2 and NixRu1xS2 materials for 0 < x < 0.6with no phase separation.

Langmuir Article

dx.doi.org/10.1021/la3032489 | Langmuir2013, 29, 480492482

7/27/2019 La 3032489

4/13

The lattice parameter of CoxRu1xS2 was found to decreasewith the molar ratio x for x < 0.3 to 0.4. The linear dependenceis consistent with the theoretical lattice contraction of mixedCoRu pyrite solid solution (labeled CoxRu1xS2 theoreticalin the Supporting Information, Figure S5). At higherconcentrations (x > 0.30.4), the lattice parameter becomesindependent of concentration. At high concentrations of cobalt,specificallyx = 0.5 and 0.6, an additional phase was observed byXRD and identified as the pentlandite structure Co9S8. Finally,material synthesized from a pure cobalt precursor (x = 1) wasfound to crystallize in the pyrite structure of CoS2.

At low Mn and Cu concentrations, the data suggest thatthere was substitution into the crystal lattice of RuS2 asobserved by the pyrite lattice expansion at x < 0.2 and x < 0.3,respectively. Mn-substituted RuS2 was found to experiencephase separation with growth of a second phase of Mn2S(alabandite structure) at x 0.4. Cu-substituted RuS2 did notshow any other crystalline phases except pyrite due to eitherlow degree of crystallinity or high degree of significantdispersion of a second phase.

The morphology of Co-substituted ruthenium sulfidematerials with different concentrations of Co substitution wasexamined by TEM and SEM. The average sizes of RuS 2, NiS2,and FeS2 crystalline domains were estimated using fits to theScherrer equation of the XRD peaks.40 A summary of thesecrystalline domains is shown in Table 2 calculated using a shapefactor k = 0.8 with an instrumental broadening of 0.1.

TEM studies confirmed the nanoparticulate morphology ofthe RuS2 sample with particle sizes of 10 nm (SupportingInformation, Figure S6a). The RuS2 nanoparticle size was notaffected by the introduction of small amounts of Co or Ni(20%), as can be seen from the TEM image (SupportingInformation, Figure S6b). The morphology of RuS2 with 60%cobalt content (x = 0.6) was examined by SEM. Largecrystallites (0.5 m) of Co9S8 were observed by SEM(Supporting Information, Figure S6c). The morphology of the100% Co sample by SEM showed that the material has a lowdegree of dispersion and consists of submicrometer particlessupported on carbon (Supporting Information, Figure S6d).

Examination of NiS2 and FeS2 samples by SEM did notreveal particulates that were distinguishable from the 50 nmcarbon particles suggesting that the sulfide particles were either

too small to visualize or, more likely, of an indistinguishable sizecompared to the carbon (sulfide sizes were estimated from theScherrer equation, see Table 2).

The characterization of Mo and W sulfides is explained indetail in the Supporting Information (Figures S9S12).

3.2. Screening by Bubble Evolution or Gas Detection.Reactivity screening of a collection of the synthesized TMSmaterials for hydrogen evolution was performed by observinghydrogen-bubble growth at the surface of the catalysts duringcyclic voltammetry scans. A library of 228 different catalystsamples based on mixed W, Fe, Mo, Ru, and Ni disulfides withthe substituted metal molar concentrations ranging from 5% to60% (for RuS2) and 10% to 50% (for WS2, FeS2, MoS2, and

NiS2) was prepared and screened. The layout of the librarytogether with a HER activity screening image under an appliedpotential of0.33 V versus NHE are shown in Figure 1.

In the image, hydrogen bubbles are visualized by the high-intensity reflections. Based on the relative density of bubblesforming on the various samples, the substituted RuS2 sampleswere the most active catalysts in the library (Figure 1). Apartfrom substituted ruthenium sulfides, WS2 substituted with Ruand Rh, FeS2 substituted with Rh, and MoS2 substituted withRu and Rh also showed significant gas evolution activity.Although TMSs substituted with Rh (WS2, FeS2) showedsignificant activity, we excluded them from our choice of thecatalysts due to the high cost of Rh. The Co-substituted RuS 2library showed sufficient activity to be considered as promising

Table 2. Crystallite Size of Synthesized Metal Sulfides onCarbon Measured by XRD (Supporting Information, FiguresS1S3) and Calculated by Scherrer Equation

metal sulfide RuS2 NiS2 FeS2

crystallite size (from Scherrer equation), nm 1316 40 23

Figure 1. TMS catalyst libraries during the test using the systemdepicted in the Supporting Information, Figure S1. On the layout map:(A) substituted RuS2 catalysts with the substituted metal molarconcentrations ranging from 5% to 60% (left to right) and substituted

WS2, FeS2, MoS2 catalysts with substituted metal molar concentrationsranging from 10% to 50% (left to right). On the image: hydrogenevolution in 0.5 M HBr under a potential of0.33 V vs NHE. Theamount of bubble evolution is used to identify high-activity materials.(B) Nickel sulfide catalyst library evolving hydrogen in 0.5 M HBrunder a potential of0.33 V vs NHE. The density of the bubbles isused to identify high-activity materials. The dopant concentration

ranges from 10% to 50%.

Langmuir Article

dx.doi.org/10.1021/la3032489 | Langmuir2013, 29, 480492483

7/27/2019 La 3032489

5/13

cost-effective catalyst candidates for the HER. Amongsubstituted NiS2 catalysts (Figure 1b), the Ru-substitutedsamples showed the highest relative activity for the HER underthese conditions.

3.3. Activity of TMSs for the HER. 3.3.1. Optimization asa Function of Substitution Metal Concentration. Materialsidentified as active catalysts by high-throughput screening werefurther tested for hydrogen evolution activity by cyclicvoltammetry. Co-substituted RuS2 samples with differentsubstituted metal concentrations show that the logarithm ofthe current varies linearly with the potential in the potentialwindow between 0.13 and 0.05 V (Supporting Information,Figure S13). This linear region was attributed to a regime ofkinetically controlled current and was used for the determi-nation of the Tafel slopes. It has been shown 41 that, under theassumption of fast proton adsorption (Volmer step), a Tafelslope varying between 40 mV/decade (mV/dec) and 118 mV/dec (values are dependent on the surface coverage) indicatesthat the HER is likely accomplished through the reaction of ahydrogen atom with a proton, that is, the VolmerHeyrovskymechanism. The Tafel slopes determined for the Co-substituted RuS2 samples, with substitution concentrations

varying from 5% to 60% are about 6575 mV/dec. The Tafelslope for the pure RuS2 sample is approximately 107 mV/decand is distinct from that of the Co-substituted samples. Theseresults indicate that the reaction on Co-substituted RuS2proceeds through a VolmerHeyrovsky mechanism and thatsubstitution with Co enhances the activity of HER.

The geometric current densities at an overpotential of0.33V measured for hydrogen evolution during HBr electrolysis forsamples with different cobalt concentration are shown in Figure2. The optimal catalyst compositions were approximately 30

40% Co-substituted RuS2 and 60% Ni-substituted RuS2.Interestingly, the highest activity of Co-substituted RuS2corresponds to a Co concentration at which the material isobserved to begin to form separate phases. Therefore, poorlydispersed Co9S8 phase is not an active catalyst for the HER.Studies of morphology indicate that the size of solid pyritesolution mixed-phase nanoparticles does not change noticeablyas a function of a substitution for small concentrations (up to20%), which may indicate little change in the surface area withincreasing concentrations of Co.

Figure 3 shows the data from cyclic voltammetry of the mostactive catalysts (30% and 40% Co-substituted RuS2 and 60%

Ni-substituted RuS2) along with RuS2, Pt on carbon (of sameloading on metal basis), and Vulcan X72 carbon support withno catalyst.

The results shown in Figures 2 and 3 indicate that the activityof RuS2 is greatly improved by substitution with nickel orcobalt. The current densities of the most active Co-substitutedRuS2 samples are approximately two times higher than that ofpure RuS2 (Figure 2). At overpotentials higher than 0.35 V,40% Co-substituted RuS2 shows higher hydrogen evolutionrates than platinum under the same metal weight loading basis(Figure 3).

3.3.2. H2H2 Cell Performance. In order to evaluate theperformance of TMS catalysts as the hydrogen electrode in aregenerative H2/Br2/HBr flow cell, electrodes made of RuS2,Co0.3Ru0.7S2, commercial Rh sulfide, and Pt were tested forboth HER and hydrogen oxidation reaction (HOR) activity inthe H2H2 cell described in the Methods section. The positivevoltage portion of the currentvoltage plot corresponds to theHER occurring on the electrode specified on the plot legend(Figure 4). Results show that the HER activity of Co0.3Ru0.7S2 isgreatly enhanced compared to RuS2 in a system similar to theH2/Br2/HBr regenerative flow cell: at an overpotential of 0.15V, Co0.3Ru0.7S2 develops current about 17 times higher thanthat of RuS2. In order to drive the cell at a current density of 0.1

A/cm2, Co0.3Ru0.7S2 requires an overpotential only 110 mVhigher than that of Pt.

Although the Rh sulfide catalyst maintained high activity forboth the HER and HOR, neither pure nor Co-substituted RuS2was active as a hydrogen oxidation catalyst. Investigation ofhydrogen oxidation activity is beyond the scope of the paper.Nonetheless, these in-cell tests demonstrate the high activity ofCo-substituted RuS2 for the HER and that, unlike Rh sulfide orPt, the Co-substituted RuS2 electrocatalyst is not a suitablebipolar cata lyst for regenerative H2/Br2/HBr flow cellapplication.

3.4. Stability under Hydrogen Evolution and duringShutdowns. Chronopotentiograms of the synthesized cata-

Figure 2. Current density at V = 0.33 mV versus NHE extractedfrom cyclic voltammetry HER scan run at 5 mV/s in 0.5 M HBr as afunction of molar ratio x (x = [M]/([M]+[Ru])) of substitutions M(M = Ni or Co). Here, x = 0 corresponds to pure RuS2, and x = 1

corresponds to pure NiS2 or CoS2.

Figure 3. Cyclic voltammetry of the HER in 0.5 M HBr (purged withargon) on Vulcan X72 carbon, platinum 30% catalyst on carbon, RuS230% on metal basis on carbon, Co0.3Ru0.7S2 30% on metal basis oncarbon, Ni0.6Ru0.4S2 30% on metal basis on carbon, and Co0.4Ru0.6S230% on metal basis on carbon. Ink loading was 0.54 g/cm2. The

voltage was cycled at a rate of 20 mV/s. The current density isnormalized to the geometric surface area. The Pt performance is mostlikely governed by ohmic resistance appearing as a result of corrosion.

Langmuir Article

dx.doi.org/10.1021/la3032489 | Langmuir2013, 29, 480492484

7/27/2019 La 3032489

6/13

lysts together with the catalysts used as a reference (rhodiumsulfide BASF catalyst, platinum 30% catalyst on carbon, andcarbon used as a support material for the synthesis) in 0.5 MHBr are shown in Figure 5. All the electrocatalysts except MoS2

were observed to be stable in 0.5 M HBr with a stable voltage atconstant current over the experimental period. The stability wasfurther verified by conducting cyclic voltammograms beforeand after the chronopotentiometry study (not shown). OnlyMoS2 showed variations between the cyclic voltammogramsbefore and after constant current testing. Although there isimprovement seen during the chronopotentiometry of theMoS2, the lack of stability (gas-phase hydrogen was detectedwhen the catalyst was exposed to HBr) eliminates this materialas a potential electrocatalyst. The current may also be comingfrom a corrosion reaction and not from H2 production. It is alsopossible that corrosion of the MoS2 particles is exposing moreof the edge sites that have been shown to be the active sites for

H2 evolution.22 Any large change in the voltage apart from the

initial change is due to changes in the electrocatalyst. The initialincrease in the magnitude of voltage is expected from thedevelopment of a concentration gradient in the cell due to localconsumption of H+ ions (the reference electrode was situated23 cm apart from the working electrode in the samecompartment). The higher increase in voltage required for Ptrelative to the TMS may be due to the effect of bromide ionpoisoning of the catalyst surface. In addition to apparentstability, the voltage shows relative hydrogen evolutionefficiencies. RhxSy and Pt are the best-performing catalystsfollowed closely by the Co0.3Ru0.7S2 electrocatalyst. To verifythat the measured current was from hydrogen production, wemeasured the evolution of hydrogen with the Co0.3Ru0.7S2 andPt catalysts, using mass spectrometry (MS). The rate ofhydrogen evolution was 1 mol of H2 per 2 mol of electrons forboth catalysts, indicating near 100% Faradaic efficiency. Therewas no H2S evolved into the gas phase measurable above thenoise level (0.037%) by differentially pumped MS.

The stability of the 30% Co-substituted RuS2 catalyst underconditions close to HBr electrolyzer operation was investigatedin extended duration testing (48 h) in more-concentrated HBr(3 M). The potential of the working electrode was measuredperiodically with respect to a Ag/AgCl reference. The potentialof the electrode and hydrogen evolution were observed to bestable over the experimental period (Supporting Information,Figure S14) and superior to platinum, which is unstable undersuch conditions (Supporting Information, Figure S15). Thevariation of voltage during the 48 h test was attributed to thesensitivity of the measurement to the position of the referenceelectrode in the cell due to the formation of a diffusion layer inthe vicinity of the cathode.

Samples of the electrolyte were extracted during the HERtest and elemental compositions of Co, Ru, and S wereanalyzed as a function of test duration. It was found that about25 7% of cobalt present in the catalyst sample dissolvedduring the initial 5 h of the HER (Figure 6). The result wascompared to measurements of the catalyst stability with noapplied electrode bias, which represents a shutdown of theelectrolytic cell. Under the no-bias condition, 34 8% of initialcobalt in the electrocatalyst dissolves into solution during thefirst 5 h of the exposure. Slower dissolution rates and smallerdissolved Co concentrations in HBr for the electrode under

Figure 4. Currentvoltage curves with specified catalyst in a H2H2cell. Catalyst loading is 0.75 mg/cm2 of listed catalyst. Hydrogenpressure is 3 psig; flow rate is 2206 mL/min. Counter electrodecatalyst was Pt loaded on carbon. The specified electrode is operatingas a hydrogen evolution catalyst at positive voltages and a hydrogenoxidation catalyst at negative voltages. Co substitution increases thehydrogen evolution activity for RuS2; however, both ruthenium sulfideand Co-doped ruthenium sulfide show low hydrogen oxidationactivity.

Figure 5. Chronopotentiometry (11 mA/cm2) using a two-compart-ment rotating disk electrode cell, with an electrolyte of 0.5 M HBrpurged with argon. The catalyst loading was 0.54 mg/cm2 on anelectrode rotating at 2500 rpm. The larger the voltage, the less efficientthe electrocatalyst is for the hydrogen evolution. Platinum perform-ance degrades over time. There is a possibility of MoS2 becomingmore active under HER conditions, possibly due to corrosion thatincreases the active surface area. However, the final MoS2 activity andinstability in HBr/Br2 is too low for use as an active electrocatalyst.

Figure 6. The fraction of the initial cobalt loading dissolved over timein 3 M HBr with and without an applied bias for hydrogen evolution.The applied bias was adjusted for a charge density of 100 mA/cm 2.The electrodes were made by loading electrocatalyst onto Toraycarbon paper.

Langmuir Article

dx.doi.org/10.1021/la3032489 | Langmuir2013, 29, 480492485

7/27/2019 La 3032489

7/13

applied cathodic bias as compared to no bias can be attributedto the protective role of the negative potential. The electric fieldfavors redeposition of positively charged metal cations that mayleave the unbiased catalyst surface.

3.5. Corrosion Reactions. The catalyst stability was furthertested by prolonged exposure to heavily concentrated HBr (6M), under argon atmosphere. In order to perform this test, 3mg of catalysts were exposed to 1 mL of 6 M HBr for 2 weeksunder continuous stirring at room temperature. The corrosionproducts were evaluated by testing the composition ofcorrosion gas, dissolved ions, and surface modification.

3.5.1. Composition of Gas-Phase Corrosion Products. Gas-phase chemical corrosion tests did not detect any H 2S formed(detection limit = 0.037%) after exposing Co0.3Ru0.7S2 andRuS2 to HBr. Similar tests did not detect any hydrogen or sulfurdioxide as gaseous corrosion products.

3.5.2. Composition of Liquid-Phase Corrosion Products.

The results of the ICP-AES tests showed dissolution of (relativeto the starting amounts) 2.3% ruthenium for RuS2 and 30%cobalt and 0.7% ruthenium for Co0.3Ru0.7S2. Background sulfurcontent in the electrolyte and from the air made exact sulfurcomposition determination unfeasible. The fraction of totalcobalt dissolved in 6 M HBr was approximately the same as thefraction of cobalt dissolved in 3 M HBr in either chemical orelectrochemical exposure. The standard deviation in thismeasurement did not exceed 10%. The absence of a blackprecipitate by the lead nitrate test verified that no H2S formedby corrosion. The presence of a white precipitate by Ba(NO3)2addition indicates the sulfur is going into acidic solution fromthe catalyst as HSO4

. The corrosion rate and amount wasfound to be independent of O2 availability (Supporting

Information, Figure S16). We believe that the Co that isdissolving is from an inactive portion of the catalyst, along withsulfate species from the catalyst surface. Once the inactivephase containing Co is dissolved, the remaining pyrite-phaseelectrocatalyst is stable without further corrosion, even afterexposure to a fresh HBr solution, with no Co in solution. Foruse as an electrocatalyst, Co-substituted RuS2 could bepretreated to remove the inactive Co, followed by recovery ofthe remaining stable electrocatalyst. Longer-term studies ofcorrosion on the time scale of months will be more helpful indetermining the extended stability of this catalyst.

3.5.3. Characterization of Catalyst Surface after Corro-sion. XRD analysis of the 30% Co-substituted RuS2 material

after 5 h of exposure to 3 M HBr showed no observed shift inthe lattice parameter of the pyrite structure or new reflections

indicating new crystalline phases.The XPS evaluation of the surface elemental composition(Supporting Information, Figure S17) reveals that the surfaceof the synthesized 30% Co-substituted RuS2 material is sulfur-rich (as compared to the expected bulk composition) with aRu/S ratio of approximately 4.2 (48.3 at. % of the total) for thefreshly prepared sample and 3.9 (53.7 at. % of the total) for thesample after HBr exposure (Supporting Information, Table S1).Oxygen present on the surface (38.1 and 31.0 at. % of total forfreshly prepared sample and the sample after HBr exposure,respectively) may originate from oxygen chemically adsorbedon carbon, water molecules physisorbed on the high-surface-area material, or oxygen coordinated to sulfur. According to thehigh-resolution O 1s spectra (Supporting Information, Figure

S18), no oxygen coordinated to metal was observed on eitherof the samples (529.2 eV for O 1s in RuO 2 or 529.6 eV for O 1sin CoO). High-resolution S 2p spectra (Supporting Informa-tion, Figures S19 and S20) reveal two distinct types of sulfurbonding on the surface: one with binding energies typical for S2p in RuS2 (labeled S2

2) and the other with energies close tothat in compounds where oxygen is bound to sulfur (labeledSO). This observation is consistent with spectroscopicanalysis of a single-crystal RuS2(100) surface

42 that also showedno indication of RuO bonds and explained the presence of SO bonds on the surface by the formation of SSO neutralspecies. The quantitative analysis of high-resolution sulfur 2pXPS spectra (last two columns in Table S1 of the SupportingInformation) indicates that the ratio of sulfur bound to metal to

the sulfur bound to oxygen increases from 76.7% to 90% as aresult of the exposure to HBr. The ratio of Ru to Co changesfrom 5.8 to 9.9 as a result of material treatment with HBr.

The amount of oxygen and sulfur in the SO bonding stateon the surface was reduced after exposure to HBr,corresponding to the SO group detected dissolving intosolution. There was no significant change in the amount ofsurface cobalt, after the initial loss of 34% of the original cobaltobserved in the ICP-AES solution experiment. This discrepancyis likely due to the error in quantification of the surface cobaltconcentrations at low amounts of cobalt. Detection of cobaltmay be inaccurate and diminished because of the energy lossassociated with its high binding energy.34

Figure 7. Supercell used in the DFT calculations. We show here two atomic layers in the side view; three atomic layers were used in the calculations.

Langmuir Article

dx.doi.org/10.1021/la3032489 | Langmuir2013, 29, 480492486

7/27/2019 La 3032489

8/13

The activity of the catalyst is unchanged when the cobaltdissolves into solution; therefore, we conclude that the solublecobalt is inactive for the HER. There was no detection of RuO bonds by high-resolution XPS, meaning that the remainingruthenium was still ruthenium sulfide.

An additional issue is the possibility of poisoning of the RuS2electrode by bromine crossing over from the HBr electrode.Our experiments show that this is not the case. DFTcalculations indicate that Br does not adsorb on the (111)surface (the free energy change is 0.2 eV) but it does bind tothe (100) surface (the free energy change is 0.63 eV). Thismeans that there are faces in the polycrystalline electrode thatwill have sites that are not blocked by Br. It is not clear how Br

adsorption on the (100) surface will affect electrochemistry.

4. COMPUTATIONAL RESULTS

4.1. Models of the RuS2 Surface. The models used in theDFT calculations for the RuS2 surface are shown in Figure 7.Figure 7a shows the 2 2 supercell (a = 11.32 ) used forcalculations for the (100) surface. When this face is cut, itexposes S2 pairs and Ru atoms that are coordinated with fivesulfur atoms each belonging to a S2 pair. All sulfur atoms in thesurface are paired, and each sulfur pair is coordinated with fiveRu atoms, four in the surface plane and one in the layer below.In bulk RuS2, the anions are S2

2, similar to peroxides (but notto oxides), and the two sulfur atoms in each S2 pair areequivalent. However, on the (100) surface, they are not: the

one labeled SP4c (Figure 7a) is coordinated to three Ru atoms(two in the surface layer and one in the layer below) and onesulfur atom. The S atom labeled SP3c is coordinated to two Ruatoms in the top layer (there is no Ru atom in the layer below)and a sulfur atom. The bond length of the sulfur pairs at thesurface is the same as in the bulk (the calculated SS bondlength in bulk is 2.21 ): this is surprising considering that, toform the surface, we had to cut SRu bonds.

The (111) face can be cut in two ways, shown in Figure 7b,cand denoted by (111)SS and (111)SS, respectively. The metalatoms on the (111)SS surface (Figure 7b) are six-coordinated,as in the bulk. This face has three kinds of sulfur atoms: the onelabeled SP2c is coordinated with one Ru atom and one S atom;

the one labeled SP3c is connected to two Ru atoms and one Satom; the one labeled SU3c needs a separate explanation.Frechard and Sautet4347 have shown that the surface formedby removing two sulfur atoms from the supercell of the (111)face results in a surface with lower energy. We have followedthis work and removed two sulfur atoms from the supercell,which now no longer has the RuS2 stoichiometry. The twosulfur atoms that were bonded to the sulfur atoms that havebeen removed are labeled SU3c. In spite of having cut the SSbond, when S atoms are removed, the SU3c atoms do not bindhydrogen as strongly as the SP2c binds hydrogen. The bondlength in the sulfur pair on the (111)SS surface (Figure 7b) isshorter by 0.14 than the same bond in the bulk (2.21 ).

This correlates with the fact that the Bader charge on the pair inthe surface layer is smaller by half than that of a pair in the bulk.To form the (111)SS surface, we cut all SS bonds, and thisface has no S2 pairs on it.

Each supercell contains three atomic layers, and the vacuumspace above the slab is 16.5 . For the (100) surface (Figure7a), a 1 1 1 k-grid gives the same results as 2 2 1; forthe (111) surface (Figure 7b,c), calculations with 2 4 1 k-grid points give the same results as do calculations with a 3 6 1 set.

4.2. Hydrogen Adsorption Free Energies on DifferentFaces of RuS2. To calculate the free-energy changes in theelectrode reaction, we use a method devised in Nrskovsgroup.21,48 The essential point is the use of the reaction taking

place at the SHE as a reference for the equilibrium calculations.The details of the derivations are given in the SupportingInformation.4951

The HER in general follows two mechanisms: the VolmerHeyrovsky pathway, which consists of the charge transfer step(Volmer reaction):

+ + * *+

eH ) H (1)

and the reaction of a proton with the adsorbed hydrogen toform gaseous H2 (Heyrovsky reaction):

* + + +

eH (H ) H2(g) (2)

Table 3. Energies of Reaction 1 Calculated from DFT and the Change in Free Energy at Zero Voltagea

a

EH* (eV) is the binding energy of a H atom to various surface sites. ZPEH* is the zero point energy of the adsorbed hydrogen atom. GH* is thefree energy of the reaction H++ e + * H* (eq4) in a 0.5 M HBr solution, at zero voltage (USHE = 0), at 298 K and a H2 pressure of 1 bar. Theupper part of the table shows the results for various binding sites on three faces of RuS 2. The empty cells in the table indicate that H does not bindon the corresponding site (for example, H does not bind on the SP4c site of the (100)SS face). The lower part of the table shows the same results forthe Co-substituted surface. The ZPEH* is not shown because it is the same as on the unsubstituted surface. The two columns having the heading M 5c(for the (100)SS face) give the values ofEH* and GH* (eV) when H binds to the Co substitution that replaced a Ru on the M5c site (these valuesare 0.16 and 0.09, respectively). The column with heading Ru (below the heading M5c) gives the lowest binding energy (and free energy) when H

binds to one of the surface Ru atoms in Co-substituted RuS2. On the (111)SS and (111)SS surfaces, the results for the substituted surface are forCo replacing the Ru atom on the M6c

t site (Figure 7b). The H coverage in all calculations is one H per eight sulfur sites.

Langmuir Article

dx.doi.org/10.1021/la3032489 | Langmuir2013, 29, 480492487

7/27/2019 La 3032489

9/13

For the VolmerTafel pathway, the second step is accom-plished by chemical recombination of two adsorbed hydrogenatoms on the surface into H2:

* 2H H2(g) (3)

We have calculated the activation barrier for the reaction in eq3 using the nudged elastic-band method52,53 and found it to be

equal to 1.72 eV on the RuS2(111)SS surface and 0.94 on the(100)SS surface. The barrier for the associative desorption is1.08 eV on the RuS2(111)SS surface and 1.14 eV on the (100)surface. This means that the Tafel mechanism is not operativein this system. This is also intuitively clear because the bindingsites of the H atoms are far apart, and dissociative adsorptionwould require stretching the H2 molecule (which has a shortbond length) by a large amount before the energy can belowered by the formation of bonds between the H atoms andthe surface. This means that the dissociative adsorption willhave a high activation energy, as the calculations show. Thisqualitative rule suggests that the barrier to dissociativeadsorption of H2 will be high for all sulfides on which thebinding sites for H are far apart.

The energies of reaction 1, calculated from DFT, (EH*, see

the Supporting Information, eq S11) are given in Table 3 alongwith the change in free energy at zero voltage. The bindingenergy of H to the SP2c site on the (111)SS surface is lowerthan the binding energy to the SU3c site by 0.14 eV, while thebinding energy to SP3c is higher than those on the other twosites. Furthermore, since the supercell has six SP2c sites and twoSU3c sites and the binding energies to them are close, we treatthese sites as being equivalent. Otherwise, we would have toconsider that the adsorbed H forms a binary lattice gas, withtwo kinds of sites and two different coverages. While this ispossible, it does not seem worthwhile given the errors inherentin the DFT calculations and the fact that our main goal is tocompare the free energy of adsorption on different faces.

The trends in the free energyGH*Ads for the Volmer step (eq

1) are the same as those in the energy EH* of the reaction1/2H2(g) H*. For example, the free energy change GH*

Ads, forthe proton neutralization and binding to the SP3c site of the(100)SS surface, is 1.02 eV, which is much higher than that forthe same reaction on the SP2c site on the (111)SS surface,which is 0.24 eV. Similarly, EH* changes from 0.78 to 0.49eV. Negative values of G indicate that a reaction isthermodynamically spontaneous: the more negative G, thegreater the driving force for the reaction. Positive values usuallyindicate that the reaction will not take place. The results arediscussed further in Section 4.3.2.

4.3. Effect of Cobalt Substitution on HydrogenAdsorption Energy on Sulfur Sites. 4.3.1. Changes inGeometry. In this section, we calculate how the free energy of

proton neutralization is affected by the substitution of a surfaceRu atom with a Co atom. The RuS bond length in bulk RuS2is 2.37 , while the CoS distance in bulk CoS2 is 2.31 .Because of this near equality in the bond lengths, we do notexpect that substituting a Ru atom with a Co atom would causea large geometry disruption. The computations show this to betrue. The (111)SS face has two nonequivalent Ru sites, labeledM6c

t and M6cs in Figure 7b. The six-coordinated M6c

s rutheniumatom is bonded to three S atoms in the second layer (with aRuS bond length equal to 2.34 ) and to three sulfur atoms inthe surface layer (with a RuS bond length equal to 2.39 ).All S atoms connected to the M6c

s Ru belong to SS pairs. TheRu atom on the M6c

t site is also coordinated to six S atoms, but

onlyfive of them belong to a S2 pair. The remaining S atom isnot paired because one of the S atoms in the pair has beenremoved when we formed the nonstoichiometric surface.4347

Only two of the six sulfur atoms are in the top sulfur layer(Figure 7b).

Substituting a Ru atom in the surface layer of RuS2(111)SSwith a Co atom causes slight, but noticeable, geometry changes.The Co placed in the M

6c

t site is shifted toward the bulk by 0.14 (as compared to the position where the replaced Ru waslocated). No metal sulfur bonds are broken when Co replacesRu. The largest CoS distance is 2.39 while the largest RuSdistance (in the surface of the unsubstituted RuS 2) is 2.47 .The shortest CoS bond is 2.24 while the shortest RuSbond in the unsubstituted RuS2 is 2.29 . For the M6c

s site, thelargest CoS distance (in Co-substituted RuS2) is 2.33 whilethe largest RuS distance (in surface of the unsubstitutedRuS2) is 2.39 . The smallest CoS distance (in Co-substitutedRuS2) is 2.29 while the smallest RuS distance (in thesurface of unsubstituted RuS2) for Ru is 2.34 . The Co atom isshifted toward the bulk by only 0.05 (as compared to theposition of the replaced Ru atom).

On the (100)SS face (Figure 7a), the five-coordinated Ruatom labeled as M5c has nonequivalent RuS bond lengths.The length of the Ru bond with the S atom from subsurface is2.23 , and it is the shortest. Among the four RuS bonds inthe plane of the surface, two have the same length as in thebulk, but the other two are changed slightly (one is shorter by0.06 and the other is longer by 0.03 than bulk bondlengths). The number of CoS bonds is the same as thenumber of RuS bonds. The length of the bond with the sulfuratom in the subsurface layer is the same as that of the RuSbond. The other four CoS bonds are slightly shorter than theRuS bonds.

When substituting RuS2 with Co, we replace a divalent cationwith another divalent atom. Work on substituted (doped)oxides, for which extensive calculations have been performed,

indicates that such a substitution does not cause a verysubstantial change in the chemical properties of theneighboring atoms.54 By analogy, we expect that Co will notsubstantially change the chemical activity of the S atoms. We donot expect that the chemical activity of the Co ion will be verydifferent from that of the Ru ion.

4.3.2. Changes in the Free Energy of the Reaction H+ + e+ * H*. We evaluated the effect of surface cobaltconcentration by computing the DFT adsorption energiesEH* for two different (111) surface models (see Section 3 inthe Supporting Information). Decreasing Co and H concen-tration does not affect the H binding energy to S. Theadsorption energies for the H atom bound to the sites near tothe cobalt dopant are found to become stronger by 0.1 eV than

to the site without the presence of Co on the surface (see theSupporting Information, Figure S21). This is a small change,but it does affect the free energy at room temperature.

The changes in free energies GH*Ads (computed from eq S10

of the Supporting Information) for reaction 1 when = 0.125and USHE = 0 are given in Table 3. The substitution increasesslightly the thermodynamic driving force for the reaction. Theeffect is particularly large for the (100)SS face where thepresence of Co changes GH*

Ads from 0.59 to 0.08 eV. Whenexamining free energies, one should keep in mind that GH*

Ads

for this reaction is affected by the voltage. Therefore, animpossible reaction (GH*

Ads > 0) can be performed if thevoltage is changed to make GH*

Ads negative.

Langmuir Article

dx.doi.org/10.1021/la3032489 | Langmuir2013, 29, 480492488

7/27/2019 La 3032489

10/13

In Figure 8, we present the free energy diagrams for theHER:

+ +

e2(H (W)) H2(g) (4)

taking place at 0.1 V, in the 0.5 M HBr solution, at roomtemperature, and a hydrogen pressure of 1 bar. Since the

thermodynamic properties depend only on free energydifferences, we take the free energy G(H2 + *) of the gaseousH2 and the clean surface * (i.e., no H adsorbed on it) to bezero. This is convenient because this state is not affected bychanges in the electrode potential. On this scale, the free energyG(H* + H+ + e) of the state in which H is adsorbed on thesurface is equal to GH*

Des (Supporting Information, eq S13),and the free energy G(2(H+ + e) + *) of the clean surfacewith H+ ions in solution and an electron e in the electrode isequal to GH*

Ads GH*

Des.Figure 8a shows the free energy diagram for the (111)SS

surface. The right-hand side panel (shaded in blue) shows thechange of the free energy of the state 2(H+ + e) + * with the

electrode potential 0.1 V. The middle panel (shaded in green)shows the free energies for the state (H+ + e) + H* where theadsorption of H on different sulfur sites at hydrogen coverages= 0.125 (blue and black) and = 0.875 (green dashed). Here= nH/nL, where nH is the number of H atoms on the surfaceand nL is the number of lattice sites. As we explained earlier,

when we calculate the coverage, we do not take into accountthat there are three different sulfur sites for H adsorption.Therefore, nL is equal to eight surface sulfur sites (six SP 2c sitesplus two SU3c sites). Strictly speaking, for one hydrogenadsorbed on the SP3c site on the (111)SS face, we should haveused nL = 6 and = 0.167 to calculate the contribution fromconfiguration entropy; however, the effect of such modificationon the free energy is very small (0.04 eV).

The free energy change for hydrogen adsorption on a surfacedepends on the coverage. We examined the hydrogenadsorption free energy at different coverages on the (111)SSand the Co-substituted (111)SS surfaces, according to thefolowing reaction:

Figure 8. Free energy diagram for the reaction (2(H+ + e) H2(g)

) in 0.5 M HBr, at 298K, 1 bar of hydrogen pressure, and at 0.1 V. In eachdiagram, the left panel (shaded in red) shows the free energy of the gas-phase molecular hydrogen and the metal sul fide surface, which is taken hereto be zero. The middle panel (shaded in green) shows the free energy of a proton with the adsorbed hydrogen, and the right panel (shaded in blue)shows the free energy of the hydrated proton in the solution and the electron in the sulfide electrode. The dots indicate the free energies at zero

voltage. The horizontal lines show the free energies at the voltage indicated in the figure. The notation G(X;V), where X = SP2c, SP3c, SU3c, M5c,indicates the hydrogen adsorption site; the notation X = SP2c-Co denotes the H-adsorption sulfur site nearest to the cobalt dopant. The site labelsare defined in Figure 7. (a) Gives the free energy diagram for the undoped RuS2(111)SS surface; (b) the same diagram for the Co-dopedRuS2(111)SS surface; (c) the diagram for the undoped RuS2(100)SS surface; (d) the diagram for the Co-doped RuS2(100)SS surface.

Langmuir Article

dx.doi.org/10.1021/la3032489 | Langmuir2013, 29, 480492489

7/27/2019 La 3032489

11/13

+ * + *n n1

2H H ( 1)H2(g) H H (5)

Our calculations show that the free energy of reaction 1 doesnot change significantly when the coverage varies from 0.125 to0.75. Because of this, we only show the free energies for =0.125 and = 0.875 in Figure 8a,b.

We need to emphasize that we are calculating a constrainedfree energy that assumes that somehow the coverage can becontrolled and fixed at a given value. Ordinary thermodynamicequilibrium calculations would determine the equilibrium valueof the coverage by setting the G for the two reactions (eqs 1and 2) equal to zero and solving these equations for . Theseconstrained calculations provide the following information: ifG for one of the reactions is positive, the reaction isimpossible at the fixed coverage; ifG < 0, then the reaction isthermodynamically possible, at the fixed coverage. Note thatthe initial state (2(H+ + e) + *) and the final state (H2 + *)are the same in all four diagrams. If the electrode potential iszero, the reaction 2(H+ + e) + * H2 + * is impossiblebecause G is slightly positive (GHER = GH*

Ads + GH*Des =

0.024 eV at 298 K, the H2 pressure of 1 bar, in a 0.5 M HBrsolution). However, a slight change to a negative electrodepotential will make the reaction possible.

We examine next each diagram to find out whichintermediate state ((H+ + e) + H*) is thermodynamicallypossible. In Figure 8a, for the face (111)SS, at zero electrodepotential, the conversion of H+ to adsorbed hydrogen ispossible for H bound to SU3c or SP2c, but the conversion to Hadsorbed on the SP3c site is not possible. Unfortunately, thedesorption of H to form gas-phase H2 is not thermodynamicallyallowed from SU3c or SP2c. Therefore, neither of the final states(H2 in gas) can be reached at zero potential. However, thereaction is possible if the voltage is more negative than 0.1 V,because this will lift up the free energies of the intermediate andthe initial state, while the free energy of the final state will not

change.The diagram in Figure 8b, which shows calculations

performed on the Co-substituted RuS2(111)SS surface, canbe interpreted in the same way as Figure 8a. At zero voltage anda coverage of= 0.125, there is at least one step with a positiveG as one goes from 2H+ to H2(g). At a voltage of0.1 V, theconversion of 2H+ to H2(g) is possible through adsorption of Hon the SP2c sites near the Co dopant at a coverage of 0.875.This may explain the increase of activity observed for Co-substituted RuS2.

Figure 8c gives the free energies for the undopedRuS2(100)SS surface. We find that none of the intermediatestates considered here (i.e., H adsorbed on M5c or on SP3c) is a

Figure 9. Free energy diagram for HOR (H2(g) 2(H+ + e)) in 0.5 M HBr, at 298 K, 1 bar of hydrogen pressure, and at voltage +0.1 V. The details

are the same as those in Figure 8.

Langmuir Article

dx.doi.org/10.1021/la3032489 | Langmuir2013, 29, 480492490

7/27/2019 La 3032489

12/13

thermodynamically possible intermediate (at = 0.125) unlessthe negative electrode potential is close to 0.3 V. Substitutingthis surface with Co is again beneficial, since the state with Hadsorbed on the Co substitution can become an intermediate ata small negative electrode potential (see Figure 8d).

Figure 9a shows the free energy diagrams for the reactionH2(g) 2(H

+ + e) for the RuS2(111)SS surface, and Figure9b shows the diagram for Co-doped RuS

2

(111)S

S

. In bothcalculations, the hydrogen coverage was 0.125, and the voltagewas +0. 1 V. The dissoc iat ive ads orp tion of H2(g) isthermodynamically allowed on both surfaces, but thedesorption of surface hydrogen atoms is not. However, at acoverage at = 0.875, for the Co-doped RuS2(111)SS surface,the reaction H2(g) 2(H

+ + e) is thermodynamically allowed.The free energy diagrams for an undoped RuS2(100)SS

surface and for a Co-doped RuS2(100)SS surface are shown inFigure 9c and d, respectively. For a voltage of 0.1 V, thereaction H2(g) 2(H

+ + e) is thermodynamically forbiddenon the undoped surface, and it is allowed on the doped one.The active site is the Co dopant (i.e., that is the favored Hadsorption site).

In summary, the DFT calculations show that doping RuS2with Co is thermodynamically favorable for both hydrogenoxidation and reduction. On the (111) surface, the Co dopanthelps by weakening the SH bond (at high H coverage), andon the (100) surface, the dopant helps by providing a bindingsite of the H atom.

Given the errors inherent in both DFT and the models usedfor the system, one should consider these conclusions to bequalitative. One should keep it in mind that, although thereactions are thermodynamically possible for both reactionpathways (HER and HOR), the calculations say nothing aboutthe kinetics especially in the case of charge transfer reactions.For the reactions that do not involve charge transfer, thecalculation of the activation energy is possible, and we providedone example that shows that H2 dissociative adsorption and H2

desorption have high barriers and therefore may be the ratelimiting steps.

5. CONCLUSIONS

We have established that Ni- and Co-substitutional dopants inthe surface of RuS2 increase the electrocatalytic activity forhydrogen evolution. The dopant is likely to affect theinteraction of the neighboring sulfur sites with hydrogen. Thedoped ruthenium sulfides were more active than pureruthenium sulfide, but their activity is lower than that ofrhodium sulfide. The doped ruthenium sulfide is less expensive($110/oz Ru)55 than rhodium sulfide ($1000/oz Rh) andmore stable than platinum, which is passivated by bromine andis unstable in HBr. The Co-doped RuS2 electrocatalyst seems to

have two kinds of Co atoms on the surface: one dissolvesrapidly in HBr and the other remains on the surface. The latteris the one that improves the activity of doped RuS2.

Co-doped ruthenium sulfide greatly enhanced the hydrogenevolution activity in a regenerative HBr flow cell compared toundoped ruthenium sulfide, although the activity was less thanthat of RhxSy. For the reverse reaction, hydrogen oxidation,cobalt-substituted ruthenium sulfide was inactive and thus not auseful bipolar electrocatalyst in a regenerative H2/Br2/HBr flowcell. The RhxSy catalyst was nearly as active as fresh platinumfor the hydrogen evolution and hydrogen oxidation reactions.

Thermodynamic calculations using DFT and a simple modelof the surface show that the role of the Co substitution is to

lower the free energy of the adsorbed hydrogen atoms. Thecalculations also suggest that the (111)SS face is less activethan the (100)SS and (111)SS faces.

ASSOCIATED CONTENT

*S Supporting InformationExperimental schematics, XRD spectra, Tafel slopes, stability

tests, XPS spectra, TEM/SEM images, ICP-AES data, CV data,and theory background. This material is available free of chargevia the Internet at http://pubs.acs.org.

AUTHOR INFORMATION

Corresponding Author

*Phone: (805) 729-3274 (A.I.); (805) 893 4343 (E.M.). E-mail: annai@seamedical.com (A.I.); ewmcfar@engineering.ucsb.edu (E.M.). Fax: (805) 893-4731 (E.M.).

Notes

The authors declare no competing financial interest.

ACKNOWLEDGMENTS

The authors would like to thank Dr. Tom Mates, Mr. YichiZhang, Dr. Arnold Forman, Dr. Young-Si Jun, and Mr. AlanDerk for helpful discussions and critical manuscript proofing.Financial support was by the National Science Foundation(EFRI-1038234) and the Air Force Office of Scientific Research(FA9550-12-1-0147). N.S. is supported by a fellowship fromthe ConvEne IGERT Program (NSF-DGE 0801627). TheMRL Central Facilities are supported by the MRSEC Programof the NSF under award no. DMR 1121053; a member of theNSF-funded Materials Research Facilities Network (www.mrfn.org). We made use of the California NanoSystems InstituteComputer Facility, funded in part by the National ScienceFoundation (CHE 0321368).

REFERENCES(1) Sivasubramanian, P.; Ramasamy, R. P.; Freire, F. J.; Holland, C.

E.; Weidner, J. W. Electrochemical hydrogen production fromthermochemical cycles using a proton exchange membrane electro-lyzer. Int. J. Hydrogen Energy 2007, 32, 463468.

(2) Simpson, M. F.; Herrmann, S. D.; Boyle, B. D. A hybridthermochemical electrolytic process for hydrogen production based onthe reverse Deacon reaction. Int. J. Hydrogen Energy 2006, 31, 12411246.

(3) Gupta, A. K.; Parker, R. Z.; Hanrahan, R. J. Solar-assistedproduction of hydrogen and chlorine from hydrochloric-acid usinghexachloroiridate (III) and (IV). Int. J. Hydrogen Energy 1993, 18,713718.

(4) Barmashenko, V.; Jorissen, J. Recovery of chlorine from dilutehydrochloric acid by electrolysis using a chlorine resistant anionexchange membrane. J. Appl. Electrochem. 2005, 35, 13111319.

(5) Kaabak, L. V.; Stepnova, N. P.; Khudenko, A. V.; Tomilov, A. P.Low-waste process for preparing ketopantolactone, with electro-chemical recovery of bromine. Russ. J. Appl. Chem. 2003, 76, 13151318.

(6) Berndt, K.; Dolle, V.; Kreutzberger, G. Production of chlorinefrom escape hydrochloric acids by electrolysis. Chem. Tech. 1969, 21,607610.

(7) Nawrat, G.; Kopyto, D.; Gonet, M. Integrated diaphragm-cathodesystems in the process of electrolysis production of chlorine. Przem.Chem. 2007, 86, 866871.

(8) Yeo, R. S.; McBreen, J.; Tseung, A. C. C.; Srinivasan, S.; McElroy,J. An electrochemically regenerative hydrogen-chlorine energy storagesystem: electrode kinetics and cell performance. J. Appl. Electrochem.1980, 10, 393404.

Langmuir Article

dx.doi.org/10.1021/la3032489 | Langmuir2013, 29, 480492491

http://pubs.acs.org/mailto:annai@seamedical.commailto:ewmcfar@engineering.ucsb.edumailto:ewmcfar@engineering.ucsb.eduhttp://www.mrfn.org/http://www.mrfn.org/http://www.mrfn.org/http://www.mrfn.org/mailto:ewmcfar@engineering.ucsb.edumailto:ewmcfar@engineering.ucsb.edumailto:annai@seamedical.comhttp://pubs.acs.org/

7/27/2019 La 3032489

13/13

(9) Yeo, R. S.; McBreen, J.; Srinivasan, S. Electrochemicallyregenerative hydrogen-chlorine energy-storage system - electrode-kinetics and cell performance. J. Electrochem. Soc. 1979, 126, C379C379.

(10) Gileadi, E.; Srinivasan, S.; Salzano, F. J.; Braun, C.; Beaufrere, A.;Gottesfeld, S.; Nuttall, L. J.; Laconti, A. B. Electrochemicallyregenerative hydrogen-chlorine energy-storage system for electricutilities. J. Power Sources 1977, 2, 191200.

(11) Tang, T. E.; Frank, S. N.; Barna, G.; Teherani, T. Regenerativehydrogen bromine fuel-cell energy-storage system. J. Electrochem. Soc.1984, 131, C323C323.

(12) Yeo, R. S.; Chin, D. T. Hydrogen-bromine cell for energy-storage applications. J. Electrochem. Soc. 1980, 127, 549555.

(13) Yeo, R. S.; Hseuh, K. L.; Chin, D. T.; McBreen, J.; Srinivasan, S.Hydrogen-bromine fuel-cell for energy-storage applications. J. Electro-chem. Soc. 1979, 126, C135C135.

(14) Schuetz, G. H.; Fiebelmann, P. J. Electrolysis of hydrobromicacid. Int. J. Hydrogen Energy 1980, 5, 305316.

(15) McDaniel, N. D.; Bernhard, S. Solar fuels: thermodynamics,candidates, tactics, and figures of merit. Dalton Trans. 2010, 39,1002110030.

(16) Luttmer, J. D.; Konrad, D.; Trachtenberg, I. Electrode materialsfor hydrobromic acid electrolysis in Texas Instruments solar chemical

converter. J. Electrochem. Soc. 1985, 132, 10541058.(17) Zhang, J.; Sasaki, K.; Sutter, E.; Adzic, R. R. Stabilization ofplatinum oxygen-reduction electrocatalysts using gold clusters. Science2007, 315, 220222.

(18) Koh, S.; Leisch, J.; Toney, M. F.; Strasser, P. Structure-activity-stability relationships of PtCo alloy electrocatalysts in gas-diffusionelectrode layers. J. Phys. Chem. C 2007, 111, 37443752.

(19) Gulla, A. F.; Gancs, L.; Allen, R. J.; Mukerjee, S. Carbon-supported low-loading rhodium sulfide electrocatalysts for oxygendepolarized cathode applications. Appl. Catal., A 2007, 326, 227235.

(20) Ziegelbauer, J. M.; Gulla, A. F.; OLaoire, C.; Urgeghe, C.; Allen,R. J.; Mukerjee, S. Chalcogenide electrocatalysts for oxygen-depolarized aqueous hydrochloric acid electrolysis. Electrochim. Acta2007, 52, 62826294.

(21) Nrskov, J. K.; Bligaard, T.; Logadottir, A.; Kitchin, J. R.; Chen,

J. G.; Pandelov, S. Trends in the exchange current for hydrogenevolution. J. Electrochem. Soc. 2005, 152, J23J26.(22) Jaramillo, T. F.; Jorgensen, K. P.; Bonde, J.; Nielsen, J. H.;

Horch, S.; Chorkendorff, I. Identification of active edge sites forelectrochemical H2 evolution from MoS2 nanocatalysts. Science 2007,317, 100102.

(23) Li, Y. G.; Wang, H. L.; Xie, L. M.; Liang, Y. Y.; Hong, G. S.; Dai,H. J. MoS2 nanoparticles grown on graphene: an advanced catalyst forthe hydrogen evolution reaction. J. Am. Chem. Soc. 2011, 133, 72967299.

(24) Hinnemann, B.; Moses, P. G.; Bonde, J.; Jorgensen, K. P.;Nielsen, J. H.; Horch, S.; Chorkendorff, I.; Norskov, J. K. Biomimetichydrogen evolution: MoS2 nanoparticles as catalyst for hydrogenevolution. J. Am. Chem. Soc. 2005, 127, 53085309.

(25) Raybaud, P. Understanding and predicting improved sulfide

catalysts: insights from first principles modeling. Appl. Catal., A 2007,322, 7691.

(26) Chianelli, R. R.; Berhault, G.; Raybaud, P.; Kasztelan, S.; Hafner,J.; Toulhoat, H. Periodic trends in hydrodesulfurization: in support ofthe Sabatier principle. Appl. Catal., A 2002, 227, 8396.

(27) Chianelli, R. R. Periodic trends transition metal sulfide catalysis:Intuition and theory. Oil Gas Sci. Technol. 2006, 61, 503513.

(28) Bither, T. A.; Bouchard, R. J.; Cloud, W. H.; Donohue, P. C.;Siemons, W. J. Transition metal pyrite dichalcogenides high-pressuresynthesis and correlation of properties. Inorg. Chem. 1968, 7, 22082220.

(29) Weber, T.; Prins, H.; van Santen, R. A. Transition MetalSulphides: Chemistry and Catalysis; Kluwer Academic Publishers:Boston, MA, 1998; p 364.

(30) Bonde, J.; Moses, P. G.; Jaramillo, T. F.; Norskov, J. K.;Chorkendorff, I. Hydrogen evolution on nano-particulate transitionmetal sulfides. Faraday Discuss. 2008, 140, 219231.

(31) Sobczynski, A.; Bard, A. J.; Campion, A.; Fox, M. A.; Mallouk, T.E.; Webber, S. E.; White, J. M. Catalytic hydrogen evolution propertiesof nickel-doped tungsten disulfide. J. Phys. Chem. 1989, 93, 401403.

(32) Nguyen, T. V. A gas distributor design for proton-exchange-membrane fuel cells. J. Electrochem. Soc. 1996, 143, L103L105.

(33) Thompson, R. B. Illustrated Guide to Home ChemistryExperiments; Make:Books: Sebastopol, CA, 2008; p 432.(34) CasaXPS, 2.3.15; Casa Software, 2009.(35) (a) Hohenberg, P.; Kohn, W. Inhomogeneous electron gas.

Phys. Rev. 1964, 136, B864B871. (b) Kohn, W.; Sham, L. J. Self-consistent equations including exchange and correlation effects. Phys.Rev. 1965, 140, A1133A1138.

(36) Perdew, J. P.; Burke, K.; Ernzerhof, M. Generalized gradientapproximation made simple. Phys. Rev. Lett. 1996, 77, 38653868.

(37) Blochl, P. E. Projector augmented-wave method. Phys. Rev. B1994, 50, 1795317979.

(38) Kresse, G.; Joubert, D. From ultrasoft pseudopotentials to theprojector augmented-wave method. Phys. Rev. B 1999, 59, 17581775.

(39) Neugebauer, J.; Scheffler, M. Adsorbate-substrate and adsorbate-adsorbate interactions of Na and K adlayers on Al(111). Phys. Rev. B1992, 46, 1606716080.

(40) Jenkins, R.; Snyder, R. Introduction to X-ray PowderDiffractometry; John Wiley & Sons Inc.: New York, 1996; p 432.

(41) Gileadi, E. Physical Electrochemistry: Fundamentals, Techniquesand Applications; Wiley-VCH: Weinheim, Germany, 2011; p 394.

(42) Kelty, S. P.; Li, J.; Chen, J. G.; Chianelli, R. R.; Ren, J.;Whangbo, M. H. Characterization of the RuS2(100) surface byscanning tunneling microscopy, atomic force microscopy, and near-edge X-ray absorption fine structure measurements and electronic

band structure calculations. J. Phys. Chem. B 1999, 103, 46494655.(43) Grillo, M. E.; Sautet, P. Density functional study of the

structural and electronic properties of RuS2(111): II. Hydrogenatedsurfaces. Surf. Sci. 2000, 457, 285293.

(44) Grillo, M. E.; Smelyanski, V.; Sautet, P.; Hafner, J. Densityfunctional study of the structural and electronic properties ofRuS2(111) I. Bare surfaces. Surf. Sci. 1999, 439, 163172.

(45) Frechard, F.; Sautet, P. Hartree-Fock ab-initio study of thegeometric and electronic-structure of RuS2 and its (100)-surface and(111)-surface. Surf. Sci. 1995, 336, 149165.

(46) Frechard, F.; Sautet, P. Chemisorption of H2 and H2S on the(100) surface of RuS2: an ab initio theoretical study. Surf. Sci. 1997,389, 131146.

(47) Frechard, E.; Sautet, P. RuS2(111) surfaces: theoretical study ofvarious terminations and their interaction with H2. J. Catal. 1997, 170,402410.

(48) Nrskov, J. K.; Rossmeisl, J.; Logadottir, A.; Lindqvist, L.;Kitchin, J. R.; Bligaard, T.; Jonsson, H. Origin of the overpotential foroxygen reduction at a fuel-cell cathode. J. Phys. Chem. B 2004, 108,1788617892.

(49) Metiu, H. Physical Chemistry: Statistical Mechanics; Taylor &Francis Group, LLC: New York, 2006; p 292.

(50) Metiu, H. Physical Chemistry: Thermodynamics; Taylor & Francis

Group, LLC: New York, 2006; p 694.(51) Robinson, R. A.; Stokes, R. H. Electrolyte Solutions; CourierDover Publications: Mineola, NY, 2002; p 587.

(52) Henkelman, G.; Uberuaga, B. P.; Jonsson, H. A climbing imagenudge elastic band method for finding saddle points and minimumenergy paths. J. Chem. Phys. 2000, 113, 99019904.

(53) Henkelman, G.; Jonsson, H. Improved tangent estimate in thenudge elastic band method for finding minimum energy paths andsaddle points. J. Chem. Phys. 2000, 113, 99789985.

(54) Metiu, H.; Chretien, S.; Hu, Z. P.; Li, B.; Sun, X. Y. Chemistryof lewis acid-base pairs on oxide surfaces. J. Phys. Chem. C 2012, 116,1043910450.

(55) http://www.infomine.com/investment/metal-prices/ruthenium/

Langmuir Article

dx.doi.org/10.1021/la3032489 | Langmuir2013, 29, 480492492

http://www.infomine.com/investment/metal-prices/ruthenium/http://www.infomine.com/investment/metal-prices/ruthenium/http://www.infomine.com/investment/metal-prices/ruthenium/http://www.infomine.com/investment/metal-prices/ruthenium/