19250 | Phys. Chem. Chem. Phys., 2014, 16, 19250--19257 This journal is© the Owner Societies 2014

Cite this:Phys.Chem.Chem.Phys.,

2014, 16, 19250

Nickel–silver alloy electrocatalysts for hydrogenevolution and oxidation in an alkaline electrolyte†

Maureen H. Tang,a Christopher Hahn,ab Aidan J. Klobuchar,ab

Jia Wei Desmond Ng,a Jess Wellendorff,ab Thomas Bligaardab andThomas F. Jaramillo*a

The development of improved catalysts for the hydrogen evolution reaction (HER) and hydrogen

oxidation reaction (HOR) in basic electrolytes remains a major technical obstacle to improved fuel cells,

water electrolyzers, and other devices for electrochemical energy storage and conversion. Based on the

free energy of adsorbed hydrogen intermediates, theory predicts that alloys of nickel and silver are

active for these reactions. In this work, we synthesize binary nickel–silver bulk alloys across a range of

compositions and show that nickel–silver alloys are indeed more active than pure nickel for hydrogen

evolution and, possibly, hydrogen oxidation. To overcome the mutual insolubility of silver and nickel, we

employ electron-beam physical vapor codeposition, a low-temperature synthetic route to metastable

alloys. This method also produces flat and uniform films that facilitate the measurement of intrinsic

catalytic activity with minimal variations in the surface area, ohmic contact, and pore transport.

Rotating-disk-electrode measurements demonstrate that the hydrogen evolution activity per geometric

area of the most active catalyst in this study, Ni0.75Ag0.25, is approximately twice that of pure nickel and

has comparable stability and hydrogen oxidation activity. Our experimental results are supported by

density functional theory calculations, which show that bulk alloying of Ni and Ag creates a variety of

adsorption sites, some of which have near-optimal hydrogen binding energy.

1 Introduction

The electrochemical evolution and oxidation of hydrogen arereactions of both practical importance and fundamental interest.The hydrogen oxidation reaction (HOR) is employed in polymer-electrolyte-membrane fuel cells that convert chemical energy toelectricity, while the hydrogen evolution reaction (HER) isrelevant to water and chlor-alkali electrolysis, electrodeposition,and corrosion. As early as 1957, the hydrogen evolution reactionhas also contributed to the fundamental understanding ofelectrochemistry and electrocatalysis.1

Recently, the HER has been used in exploring the funda-mental differences between electrocatalysis in acid and base.2 Ithas long been recognized that the reaction rates of the HORand the HER are often slower in basic electrolytes than acidicelectrolytes, even though the surface intermediate of adsorbed

hydrogen (Had) is independent of solution pH.3 Early funda-mental studies suggested that, at high pH, OH�competes withHad for surface sites, effectively poisoning the electrode andreducing overall rates.4,5 More recent work has focused on therecombination step between Had and OHad during hydrogenoxidation, suggesting that optimal adsorption of OH as well asH is necessary to facilitate recombination of Had and OHad toform H2O.2 In the reverse direction, protons must be removedfrom water before adsorbing onto the electrode and recombiningas hydrogen; this water-splitting step may also be influenced byOHad species on the electrode.6

While the role of adsorbed OH in the alkaline HOR and HERis still unclear, the effect of hydrogen adsorption strength ismuch less ambiguous. Fig. 1 plots previously published valuesof measured hydrogen evolution activity under basic conditionsversus the calculated free energy of hydrogen surface binding(DGHad

). The activity follows a volcano-type relation in which thebest catalysts exhibit hydrogen binding free energy closest tothe optimum value of zero.7 As in acid, platinum and palladium arethe most active elemental catalysts, and nickel is the most activenon-precious elemental electrocatalyst. Thus, from a catalyst-development point of view, appropriately tuning the hydrogenbinding energy is viewed as a necessary, but not sufficient,requirement for an active catalyst in alkaline environments.

a Department of Chemical Engineering, Stanford University, Stanford, CA 94305,

USA. E-mail: jaramillo@stanford.edub SUNCAT Center for Interface Science and Catalysis, SLAC National Accelerator

Laboratory, Menlo Park, CA 94025, USA

† Electronic supplementary information (ESI) available: Detailed information onthe synthesis procedure and experimental characterizations, including stabilitytesting, high-resolution XPS spectra, XRD fitting results, and OH adsorptionexperiments. See DOI: 10.1039/c4cp01385a

Received 30th March 2014,Accepted 30th June 2014

DOI: 10.1039/c4cp01385a

www.rsc.org/pccp

PCCP

PAPER

Publ

ishe

d on

06

Aug

ust 2

014.

Dow

nloa

ded

by S

tanf

ord

Uni

vers

ity o

n 14

/11/

2017

22:

53:4

1.

View Article OnlineView Journal | View Issue

This journal is© the Owner Societies 2014 Phys. Chem. Chem. Phys., 2014, 16, 19250--19257 | 19251

Recent work on non-noble electrocatalysts for alkaline hydrogenoxidation has also attributed extremely high activity to optimalhydrogen binding.8 Computational screening is a useful tool tomake predictions for new materials, which helps in guiding thedevelopment of active and stable catalysts. Previous theoreticalcalculations of DGHad

predicted that surface alloys of silver andnickel bind hydrogen with an energy very close to optimum, asindicated by the arrows in Fig. 1.9 The optimal binding energy ofNiAg surface alloys suggests that bulk NiAg alloys may be similarlyactive and motivates our further experimental and theoreticalinvestigations into these catalysts.

Experimentally, bimetallic NiAg systems have shown promise forhydrogen generation both electrocatalytically13 and in homogenouscatalysis.14 However, synthesis of binary NiAg alloys is challengingbecause thermodynamic instability causes Ag and Ni to phase-segregate into pure elemental phases at all temperatures.15 Thisimmiscibility, combined with a large lattice mismatch (0.352 vs.0.408 nm) and the lower surface energy of Ag, renders the Ag–Nicore–shell the most thermodynamically favorable arrangementfor nanoparticles.16,17 Circumventing the phase diagram andforming a metastable alloy require kinetically controlled materialssynthesis, such as rapid reduction or condensation at lowtemperature, where Ag and Ni atoms do not have enough kineticenergy to diffuse and segregate. Previously, various thin film andnanoparticle synthesis techniques have been applied for thispurpose, including electrodeposition, pulse laser deposition,e-beam PVD, DC sputtering, and gamma-irradiation.18–22

In this work, e-beam PVD was chosen for its ability to access awide range of alloy compositions. PVD has the additional advantageof producing flat, conformal films on a variety of substrates.Although alkaline electrolysis is an established technology for which

many materials have been tested as catalysts, many previousstudies are developed around high surface area structures, suchas RANEYs nickel, and tested as composite electrodes in a fuelcell or electrolyzer.23–26 Although such conditions most accuratelyrepresent those during device operation, distinguishing betweencatalysis, reactant transport, and electron transport effects in acomplicated device geometry is extremely difficult. For example,nickel–molybdenum and nickel–molybdenum–cobalt catalystsshow the best HER activity of all non-precious metals, but thereason for the improvement over pure nickel has been alternativelyattributed to increased microscopic porosity and increased intrinsiccatalytic activity.8,27–30 In other intermetallic systems, improvedactivity in the presence of dopant metals has been attributed to avariety of reasons, including better ohmic contact betweenparticles, better mass transport through catalyst pores, bettermass transport through electrode pores, and better catalysis atthe particle surface.26,31–33 Physical vapor deposition producesflat, conformal films of uniform composition, avoiding suchcomplications and enabling fundamental studies to provide agreater insight. Additionally, synthesizing thin films of catalystson a glassy carbon substrate permits alloys to be tested in arotating-disk configuration that separates catalytic activity fromthe surface area, ohmic contact, and pore transport.

2 Methods2.1 Computational details

All theoretical results are based on density functional theorycalculations. These are performed using Quantum ESPRESSO,34,35

a plane-wave36 pseudopotential37 code, and the Atomic SimulationEnvironment.38 We used the RPBE39 exchange–correlationfunctional and ultrasoft40 pseudopotentials from the PS Libraryproject.41 Kinetic energy cutoffs were 700 eV for wave functionsand 6500 eV for charge densities and potentials, and the surfaceBrillouin zone was sampled using a 6 � 6 � 1 Monkhorst–Pack42

k-point mesh. A Gaussian smearing of 0.2 eV was applied to theelectron occupation numbers near the Fermi level.

Periodically repeated 5-layer slabs were used to modelhydrogen adsorption onto three close-packed fcc(111) NiAgalloy surfaces (Ni3Ag, NiAg, and NiAg3) as well as onto the pureNi and Ag(111) facets. A 2 � 2 surface unit cell with one H atomper cell was applied. This resulted in a 1/4 monolayer coverageof atomic H on the transition-metal substrates and severalstructurally non-equivalent adsorption sites, particularly, onthe Ni3Ag and NiAg3 alloys. The three alloyed surfaces werealloyed throughout the bulk, and the top two layers of allsurfaces and the H adsorbate were relaxed until all residualforces acting on them were less than 0.05 eV �1. The bottomthree metal layers were fixed in the bulk structure, with latticeconstants computed using an equation of state approach.

2.2 Synthesis of NiAg alloys

Catalysts were deposited on 5 mm diameter glassy carbon disks(Hochtemperatur Werkstoffe) and 1.6 cm quartz slides (GMAssociates) using a load-locked dual e-beam evaporator equipped

Fig. 1 Plot of hydrogen evolution activity under basic conditions (measured)versus hydrogen adsorption energy (calculated). As in acid,7 the data show avolcano trend. Platinum, nickel, and palladium bind hydrogen with the energyclosest to optimum, but still offer room for improvement. All hydrogenbinding energies are from ref. 10, except for Ag and Ni values, which werecalculated in this work. Alkaline measurements are from ref. 11 (red squares,30% KOH, 80 1C) and ref. 12 (blue circles, 0.1 M KOH, 25 1C). The dashed linesare to guide the eye. Arrows mark the hydrogen binding energy to surfacealloys of (a) Ni on the Ag substrate and (b) Ag on the Ni substrate.9

Paper PCCP

Publ

ishe

d on

06

Aug

ust 2

014.

Dow

nloa

ded

by S

tanf

ord

Uni

vers

ity o

n 14

/11/

2017

22:

53:4

1.

View Article Online

19252 | Phys. Chem. Chem. Phys., 2014, 16, 19250--19257 This journal is© the Owner Societies 2014

with separate quartz crystal microbalances for each source(Temescal VES2550). Additional Ni films were separately depositedusing an Innotec ES26C e-beam evaporator. The total depositionrate was maintained at approximately 2 Å s�1 for all depositions.To improve substrate–catalyst adhesion, glassy carbon disks wereanodized in air at 400 1C for 2 hours before sonication in acetone,isopropanol, and water, followed by mounting on wafers fordeposition. An 80 nm Ti adhesion layer was deposited on thequartz substrates, but was found to increase delamination fromthe carbon substrates during catalyst testing, possibly due tohydrogen embrittlement of the Ti. Therefore, no Ti layer was usedon glassy carbon substrates. Ni and Ag were deposited in differentratios by separately adjusting the electron beam power for eachsource. The quartz-crystal monitors were calibrated for cross-talkbetween the metal sources at the beginning of each deposition;depending on the speed of equilibration, this calibration periodresulted in a 5–30 nm layer of pure Ag underneath the 70–85 nmNiAg alloy catalyst layer.

2.3 Physical and chemical characterizations

After deposition, catalysts were characterized by scanningelectron microscopy (SEM) using a FEI XL30 Sirion with a beamvoltage of 5.0 kV. X-ray diffraction (XRD) was performed ona PANanalytical X’Pert Pro using Cu Ka radiation (l = 1.541 Å).X-ray photoelectron spectroscopy (XPS) was performed using aPHI Versaprobe Scanning XPS using Al Ka radiation (hn = 1486.6 eV,spot size 200 mm, 451 collection angle). The roughness factor ofsamples was also measured by atomic force microscopy (AFM),using a Park Systems XE-70 in non-contact mode.

2.4 Electrochemical testing

Electrochemical tests were conducted in 0.1 M KOH (Sigma)using a rotating disk electrode (RDE) set-up (Pine Instruments) anda three-electrode potentiostat (Biologic Instruments) with ohmicresistance compensation. A cell made of corrosion-resistant quartz(Pine) was used to reduce contamination from KOH etching inconventional borosilicate glass.43 All potentials were measured withrespect to a mercury/mercuric oxide reference electrode (1.0 MKOH, Koslow Scientific), which was separated from the main cell bya potassium nitrate salt bridge. The reference electrode wascalibrated to �0.869 V versus the reversible hydrogen electrode(RHE); all potentials in this work are reported vs. RHE. The counter-electrode was Ni wire (Alfa, 99.99%). Although heat treatment ofthe glassy carbon disks improved catalyst–substrate adhesion,catalyst delamination still occurred during extended periodsof hydrogen bubbling. Longer stability tests were conducted onquartz substrates with better adhesion properties. Furtherdetails of sample preparation are given in the ESI.†

The competition between the hydrogen reactions and theformation and removal of Ni hydrides and oxides can lead toerroneous activity measurements.44 The testing protocol wasdeveloped to isolate the catalytic activity for HER and HOR fromthese competing reactions. After insertion into the electrolyte,the working electrode was activated with a sweep from opencircuit potential to �300 mV vs. RHE at 10 mV s�1. This stepreduced surface oxides while permitting minimal time for

hydrogen absorption. The electrode was then held at �20 mVfor seven minutes while purging hydrogen gas (Praxair,99.999%) through the electrolyte. Linear sweep voltammograms ofthe hydrogen evolution activity were then recorded between �300and �20 mV. Next, HOR activity was measured by chronoampero-metry steps from +50 to +200 mV. Chronoamperometry instead ofcyclic voltammetry was used to separate steady-state hydrogenoxidation from transients due to double-layer charging, oxidativeremoval of Ni hydrides, and oxidation of Ni catalysts. To demon-strate that the observed anodic current was due to hydrogenoxidation, separate samples were tested using an identical procedureunder a nitrogen atmosphere as control experiments.

3 Results and discussion3.1 Uniform deposition of metastable NiAg alloys

SEM was used to determine the morphology of the catalystssynthesized on glassy carbon substrates (Fig. 2). The imagesshow that all of the films are flat and conformal. The grain size,which is approximately 20 nm for Ni and the Ni-rich alloys,increases to about 40 nm for pure Ag. The XPS spectra, shownin Fig. 3, demonstrate the composition range of the deposi-tions. The integrated areas of the Ni2p region from 850–890 eVand the Ag3d region from 364–378 eV were used to calculate therelative surface compositions, which are plotted versus nominalcomposition in Fig. 4. Depositions on both quartz and glassycarbon show good agreement with the nominal composition. Incontrast to studies in which NiAg systems were synthesized athigh temperature or by a wet chemical method, no surfaceenrichment of Ag was observed, suggesting that atoms weredeposited with insufficient thermal energy to separate into thethermodynamically preferred arrangement.14,22 The oxygenpeak at 532 eV indicates that the catalysts form a native oxidein air, and that Ni is more oxidized than Ag, as expected; thepresence of Ni surface oxides was confirmed by high-resolutionXPS.†

XRD measurements showed that the e-beam depositionmethod was able to achieve partial metastable alloying of Niand Ag. All samples, except for the pure metals, were found tohave both Ni-rich and Ag-rich alloy phases. Diffractograms of thecatalysts deposited on quartz are shown in Fig. 5. The counts are

Fig. 2 SEM images of catalysts as synthesized on glassy carbon sub-strates. Films are flat and conformal, with Ag depositing on larger grainsthan Ni or the NiAg alloys.

PCCP Paper

Publ

ishe

d on

06

Aug

ust 2

014.

Dow

nloa

ded

by S

tanf

ord

Uni

vers

ity o

n 14

/11/

2017

22:

53:4

1.

View Article Online

This journal is© the Owner Societies 2014 Phys. Chem. Chem. Phys., 2014, 16, 19250--19257 | 19253

corrected for the quartz substrate and shown to be normalizedto the maximum intensity. For pure Ni, the Ni(111) reflection isobserved at a = 0.353 nm, which is consistent with the literaturevalue. As the Ag deposition rate increases, incorporation oflarger Ag atoms into the Ni lattice shifts the peak to smallervalues of 2y. Alloying Ag also decreases the crystallinity andcrystallite size, as evidenced by the smaller peak size and anincrease in FWHM. Similarly, the peak position of the Ag(111)reflection increases from that of pure Ag (a = 0.409 nm),showing incorporation of the smaller Ni atoms in Ag-rich films.The shifting of the Ag(111) peak can only be observed visuallyfor the Ag-rich films in Fig. 5 because the (002) reflection of theTi adhesion layer overlaps for Ni-rich films with lower intensityAg(111) reflection.

To quantitatively determine the extent of alloying betweenNi and Ag, Gaussian curves were fit to the Ni(111) and Ag(111)peaks. The peak location and FWHM were used to calculate thealloy lattice constant and grain size for each of the two phasesaccording to Bragg’s law and the Scherrer equation, respec-tively; calculations and results are shown in the ESI.† Grainsizes were calculated between 7 and 14 nm for alloys, slightlysmaller than indicated by the SEM images. Inhomogeneities incomposition may explain the slight disparity between the twomethods. The relative fractions of Ag and Ni as computed byVegard’s law are shown in Fig. 4. Fig. 4 indicates that themaximum Ag content introduced into the Ni-rich phase by thismethod is about 13%, at a bulk ratio of Ni0.25Ag0.75. The Ag-richphase reaches a maximum solubility of approximately 6% Ni, ata bulk ratio of Ni0.125Ag0.875. At a nominal composition ofNi0.125Ag0.875, the solubility of Ag in the Ni-rich phase decreases.This may be because, at higher concentration, Ag atoms need todiffuse shorter distances in order to segregate into a separatephase. Similar solubility trends in metastable NiAg thin-filmshave been observed previously.19,20

3.2 Electrochemical performance of alloys

The combined materials characterization results of Fig. 2 through 5show that the films synthesized by e-beam codeposition are flat,uniform, and composed of a Ni-rich phase and an Ag-rich phase.

Fig. 3 XPS survey spectra of the vapor-deposited NiAg samples (glassycarbon substrate). The trends in Ni2p and Ag3d peaks show the composi-tional variation of the depositions.

Fig. 4 Nominal versus measured Ni concentration. Bulk concentrationsof the Ni-rich and Ag-rich alloy phases are calculated from Vegard’s Law.Average surface concentrations are measured using XPS. Closed symbolsare for quartz substrates, and open symbols are for glassy carbon. Themaximum concentration of Ag in the Ni-rich alloy phase is about 13 at%.

Fig. 5 XRD diffractograms recorded on quartz substrates. The curves arenormalized to the maximum intensity and corrected for the substratediffraction. Although the metals segregate into two phases, shifts in theNi(111) and Ag(111) peaks demonstrate that each phase is a metastablealloy. Also shown are reference peaks for Ni (JCPDS 4-0850), Ag (JCPDS4-1783), and Ti (JCPDS 44-1294).

Paper PCCP

Publ

ishe

d on

06

Aug

ust 2

014.

Dow

nloa

ded

by S

tanf

ord

Uni

vers

ity o

n 14

/11/

2017

22:

53:4

1.

View Article Online

19254 | Phys. Chem. Chem. Phys., 2014, 16, 19250--19257 This journal is© the Owner Societies 2014

To relate the chemical composition to hydrogen evolutionactivity, linear sweep voltammetry was performed in an RDEconfiguration. Fig. 6 and 7 show the HER and HOR activities ofNi, Ag, and selected NiAg alloys. Fig. 6 shows that, as expected,the hydrogen evolution activity of pure Ag is negligible. Pure Nidisplays an activity consistent with the literature,12 while theNi0.75Ag0.25 alloy shows the highest HER activity for thesesamples (Fig. 6(a)). The current density at�300 mV overpotentialis shown in Fig. 7, demonstrating a broad maximum in HERactivity at an alloy composition of Ni0.75Ag0.25. The HOR activityat constant potentials is shown in Fig. 6(b) for Ni, Ni0.75Ag0.25,and Ni0.5Ag0.5. The Ni0.75Ag0.25 alloy shows higher current than

the pure Ni sample at overpotentials of 150 mV or less, but thedifference in current disappears at 200 mV overpotential. Thischange may be caused by a different HOR mechanism at higheroverpotentials, or because oxidation of the catalyst at thesepotentials changes the catalytic properties of the surface. Similarbehavior in other Ni-based systems was ascribed to catalystoxidation.8,24,27 Noise in the data is caused in part by bubblesof hydrogen gas contacting the electrode. Also shown in Fig. 6(b)are oxidation steps for a separate Ni sample tested in a nitrogen-saturated atmosphere. The difference in current shows that theanodic current in Fig. 6(b) is due to HOR and not simply becauseof catalyst oxidation. While the uncertainty associated withmeasuring small currents in the presence of vigorous bubblingmakes the HOR trend less obvious, the Ni-rich alloys have anactivity comparable to, if not better than, the pure Ni films, asshown in Fig. 7. Stability testing also shows a similar behaviorbetween the NiAg alloys and pure Ni.†

The plot of the current density at �300 mV vs. nominal Niconcentration in Fig. 7 shows that, for HER, the optimumcatalyst concentration is in between 50 and 85% Ni. The peakalloy activity is about 9.4 mA cm�2, approximately twice that ofpure Ni on a geometric-area basis. Examination of Fig. 6(a)shows that obtaining a current density of 2 mA cm�2 onelemental Ni requires about �260 mV of overpotential, slightlyhigher than the overpotential for Ni as plotted in Fig. 1. Con-sideration of surface roughness may account for slight differ-ences in the measured activity from previously published values.Fig. 7 also shows that the HER activity of Ni0.5Ag0.5 exceeds thatof pure Ni by about 60%, but that the HOR activity of this alloy isnegligible. This difference suggests that the hydrogen bindingenergy for the NiAg alloy is stronger than optimum, so thatadsorbate–adsorbate repulsions between hydrogen atoms canimprove the HER activity but not the HOR.45

It is challenging to separate the intrinsic catalytic activity from theeffects of the microscopic electrode surface area for non-nobleelectrocatalysts. Because of the low oxidation potential of Ni, com-monly used methods for measuring the surface area, includingunderpotential deposition of H, Cu, and Pb, are not applicable toNi-based systems. The low overpotential for the HER also eliminatesthe possibility of finding a non-faradaic region of the voltage windowin which the double-layer capacitance is measured. In this work,however, the vacuum deposition synthetic approach creates flat,uniform films with negligible porosity. AFM images, shown in Fig. 8,show negligible differences in the surface area among the differentfilms. Therefore, improvements in HER and HOR activity werecaused by intrinsic effects on catalysis and not by the increasedsurface area. Fig. 8 shows that for both samples, grains are only a fewnm tall, yielding calculated surface roughness factors of 1.04 and1.03, respectively, for the Ni0.825Ag0.175 and Ni samples. Measure-ments of OH adsorption also suggested that the NiAg catalysts havesimilar or smaller areas of electrochemically active Ni (ECSANi) thanfilms of pure Ni.†

3.3 Comparison with theoretical values

Although the AFM image in Fig. 8 shows negligible surfaceroughness, determining the intrinsic catalytic activity for the

Fig. 6 Electrochemical activity measurements of Ni, Ag, and selected NiAgalloys on glassy carbon substrates in 0.1 M KOH/hydrogen gas at 900 rpm.(a) Cathodic HER sweep at 10 mV s�1. (b) HOR at potential steps from +50to +200 mV. No oxidation is seen during a control experiment in nitrogen.

Fig. 7 Current density at�300 mV (left axis, blue diamonds) and +150 mV(right axis, red squares) vs. nominal Ni concentration. The best-performingcatalysts have nominal compositions between 50 and 85% Ni. Error barsrepresent two standard deviations, based on measurements of at leastthree samples per composition.

PCCP Paper

Publ

ishe

d on

06

Aug

ust 2

014.

Dow

nloa

ded

by S

tanf

ord

Uni

vers

ity o

n 14

/11/

2017

22:

53:4

1.

View Article Online

This journal is© the Owner Societies 2014 Phys. Chem. Chem. Phys., 2014, 16, 19250--19257 | 19255

NiAg alloys is still not straightforward because the active site ofthe alloys is not known. To develop a qualitative insight into theactive site for HER and HOR catalysis, DFT calculations wereused to determine the hydrogen binding energy at differentsites of Ni, NiAg3, NiAg, Ni3Ag, and Ag. The results of all of thecalculated sites are shown in Table 1, and selected bindingenergies and geometries are shown in Fig. 9. The theoreticalhydrogen evolution current density was calculated using themodel of Skulason, et al.46 and normalized to the experimentallymeasured current density for pure Ni. Because the model wasdeveloped for acidic electrolyte, not basic electrolyte, it can beused only for qualitative interpretation in studying trends. Aspreviously discussed, the effects of hydroxide ions on alkalineHER and HOR are not entirely understood, and a quantitativemicrokinetic model that incorporates their effects is beyond thescope of this work.

Table 1 and Fig. 9 indicate that simple alloying of Ni and Agcan obtain many energetically different adsorption sites, andthat several NiAg structures have sites where hydrogen ispredicted to bind with energy close to the optimum value of0 eV. The full 4d-shell of the Ag atoms is expected to shift theprojected d-band downwards from the pure-Ni case, leading to

weaker hydrogen binding and more active catalysts. This effectis observed for binding sites which include both Ni and Agatoms, such as the Ni hollow site on NiAg3 and the Ag hollowsite on Ni3Ag. Such binding sites generally have the DGHad

values closest to the optimum of 0 eV. Another competing effectis seen at the Ni hollow site of Ni3Ag, which binds morestrongly than the pure Ni site. Similarly, the Ag hollow site ofNiAg3 binds more weakly than the pure Ag site. This is due tothe effect of the alloyed atoms not in the binding site. Forexample, in the Ni3Ag case, there are Ni atoms in the bindingsite which interact with an Ag atom. This Ni–Ag bond has lowersaturation compared to a Ni–Ni bond, leading to a highersaturation of the Ni–H bond, and thus to stronger hydrogenbinding. The interplay of these two effects results in a variety ofsurface sites with differences in DGHad

, some of which arefavorable and some of which are unfavorable for the HER.However, due to the exponential dependence of activity onDGHad

, we expect the most active sites to dominate the overallalloy activity. Additionally, the results of Table 1 assume aperfectly ordered lattice, and a random distribution of Ag andNi may change both the geometric frequency and electronicstructure of binding sites.

Fig. 9 also shows experimental data as extracted from Fig. 7for Ni0.75Ag0.25. Because the exact surface structure is notknown, the experimentally measured current density is repre-sented as a horizontal green line instead of a discrete point. Allexperimental and theoretical values are normalized to theexperimentally measured activity on pure Ni (grey line). Thealloys contain both Ag-rich and Ni-rich phases (Fig. 5), and it isnot immediately clear in which phase the active site is located.Inspection of Table 1 shows that the most optimal sites for

Fig. 8 AFM images and line scans of Ni and Ni0.825Ag0.375 on glassycarbon. Differences in line scans appear larger than actual because ofthe axes aspect ratio.

Table 1 Calculated hydrogen binding free energies (eV) on uniformlyalloyed NiAg(111) surfaces. Selected images of the calculated surfacestructures are shown in Fig. 9

Composition Binding site DGHad(eV)

Ni Hollow �0.19On-top 0.38

Ni3Ag Hollow (Ni) �0.25Hollow (Ag) �0.07On-top (Ni) 0.34On-top (Ag) 1.04

NiAg Bridge �0.15Hollow (Ag) 0.16On-top (Ni) 0.32

NiAg3 Hollow (Ni) 0.09On-top (Ni) 0.26Hollow (Ag) 0.47

Ag Hollow 0.46On-top 0.99

Fig. 9 Above, combined theoretical-experimental volcano plot for most-active sites. Markers are simulations, and the horizontal lines are measuredcurrents for: pure Ni, geometric area (grey); Ni0.75Ag0.25, geometric area(green); Ni0.75Ag0.25, area of Ag in the Ni phase (blue). The dotted red linerepresents the predicted current.46 Below, selected images of the calcu-lated surface structures are shown.

Paper PCCP

Publ

ishe

d on

06

Aug

ust 2

014.

Dow

nloa

ded

by S

tanf

ord

Uni

vers

ity o

n 14

/11/

2017

22:

53:4

1.

View Article Online

19256 | Phys. Chem. Chem. Phys., 2014, 16, 19250--19257 This journal is© the Owner Societies 2014

hydrogen binding are Ag atoms in a Ni-rich environment, orNi atoms in an Ag-rich environment. Based on this theoreticalinsight, the measured current density for Ni0.75Ag0.25 was scaledto the area fraction of the Ag component in the Ni-rich phase.This value is represented by a blue dashed line in Fig. 9. Scalingin this manner would suggest an enhancement by 100� foreach of these sites relative to pure Ni. Because this calculationneglects the activity of all other sites, the estimated current istherefore an upper bound of the possible current per active site.Although the Ni atoms in the Ag-rich phase are expected to besimilarly active, they make up a much smaller fraction of thesurface, and scaling to this area would yield a current densitythat exceeds the theoretical limit. Other active sites mayoccur at the Ni-rich and Ag-rich grain boundaries; hydrogenadsorption at the bridge site between Ni and Ag is predicted tobe very close to optimum. Tuning the grain size by controllingsynthesis parameters such as deposition rate and temperaturemight elucidate the roles of different sites and suggest strate-gies for increasing the density of the most active sites. Finally,the effects of hydroxide adsorption cannot be ignored. Becausehydrogen evolution requires an additional step of water-splitting, a bifunctional catalyst is required.2 The calculatedvalues for OH adsorption energy differ by 1.8 eV between Ni andAg.47 This difference may have synergistic effects on the HERindependent of hydrogen adsorption. More work is necessary tounderstand fully the effect of hydroxide adsorption on alkalineHER and HOR catalysis, and the control afforded by e-beamcodeposition motivates further use of the NiAg system to probethese effects.

4 Conclusions

Guided by theoretical calculations, we identified NiAg alloys aspromising electrocatalysts for the HOR and HER in base. Tomitigate the thermodynamic phase segregation of Ag and Ni,e-beam codeposition was used to deposit thin films of metalcatalysts on glassy carbon and quartz substrates. This techniquepermits finely-tuned catalyst compositions in a well-definedmorphology. Incorporation of up to 13% Ag into the Ni latticeand 6% Ni into the Ag lattice was demonstrated via XRD.Rotating disk voltammetry showed that Ni-rich NiAg sampleswere more active than pure Ni for the HER, with the most activecatalyst, Ni0.75Ag0.25, outperforming pure Ni by approximatelya factor of two. Because of the planar electrode geometry(AFM-measured roughness factors of approximately unity), theimprovement may be attributed to intrinsic catalytic activity, notto the higher surface area of Ni. The HOR activities of the Ni-richNiAg catalysts were comparable to pure Ni. The improved HERactivity was corroborated by DFT calculations, which demon-strate that alloying Ni and Ag creates a variety of adsorption sites,several with near-optimal hydrogen binding energies. Futurework to elucidate the active site(s) in the NiAg system for alkalineHER and HOR catalysis would help in enabling rational catalystdesign, which will in turn facilitate the incorporation ofadvanced catalysts into functional devices.

Acknowledgements

This work was supported by a TomKat Grant for EnergyResearch at Stanford University. We also thank the Departmentof Energy for SUNCAT funding. A.J.K. was supported by anAbbott Laboratories Stanford Graduate Fellowship. Part ofthis work was performed at the Stanford Nano Center (SNC)/Stanford Nanocharacterization Laboratory (SNL), part of theStanford Nano Shared Facilities. This work was also supportedby the facilities of the UCSB Nanotech, a member of the NationalNanofabrication Infrastructure Network partially supported bythe National Science Foundation.

References

1 B. E. Conway and J. O. Bockris, J. Chem. Phys., 1957, 26, 532.2 D. Strmcnik, M. Uchimura, C. Wang, R. Subbaraman,

N. Danilovic, D. van der Vliet, A. P. Paulikas, V. R. Stamenkovicand N. M. Markovic, Nat. Chem., 2013, 1–7.

3 S. Bagotzky and N. V. Osetrova, J. Electroanal. Chem. InterfacialElectrochem., 1973, 43, 233–249.

4 N. M. Markovic, S. T. Sarraf, H. A. Gasteiger and P. N. Ross,J. Chem. Soc., Faraday Trans., 1996, 92, 3719–3725.

5 T. Schmidt, P. N. Ross and N. M. Markovic, J. Electroanal.Chem., 2002, 525, 252–260.

6 R. Subbaraman, D. Tripkovic, D. Strmcnik, K.-C. Chang,M. Uchimura, A. P. Paulikas, V. Stamenkovic and N. M.Markovic, Science, 2011, 334, 1256–1260.

7 J. K. Nørskov, T. Bligaard, A. Logadottir, J. R. Kitchin,J. G. Chen, S. Pandelov and U. Stimming, J. Electrochem.Soc., 2005, 152, J23.

8 W. Sheng, A. P. Bivens, M. Myint, Z. Zhuang, R. V. Forest,Q. Fang, J. G. Chen and Y. Yan, Energy Environ. Sci., 2014,7, 1719.

9 J. Greeley and J. K. Nørskov, Surf. Sci., 2007, 601, 1590–1598.10 M. E. Bjorketun, A. S. Bondarenko, B. L. Abrams,

I. Chorkendorff and J. Rossmeisl, Phys. Chem. Chem. Phys.,2010, 12, 10536–10541.

11 M. H. Miles, J. Electroanal. Chem. Interfacial Electrochem.,1975, 60, 89–96.

12 W. Sheng, M. Myint, J. G. Chen and Y. Yan, Energy Environ.Sci., 2013, 6, 1509–1512.

13 T. F. Jaramillo, A. Ivanovskaya and E. W. McFarland,J. Comb. Chem., 2002, 4, 17–22.

14 H. Guo, Y. Chen, X. Chen, R. Wen, G.-H. Yue and D.-L. Peng,Nanotechnology, 2011, 22, 195604.

15 M. Singleton and P. Nash, in ASM Alloy Phase Diagrams Center,ed. P. Villars, ASM Int., Materials Park, OH, 2006, pp. 2–33.

16 L. Vitos, A. Ruban, H. L. Skriver and J. Kollar, Surf. Sci.,1998, 411, 186–202.

17 M. Gaudry, E. Cottancin, M. Pellarin, J. Lerme, L. Arnaud,J. Huntzinger, J. Vialle, M. Broyer, J. Rousset, M. Treilleuxand P. Melinon, Phys. Rev. B: Condens. Matter Mater. Phys.,2003, 67, 155409.

18 M. Schneider, A. Krause and M. Ruhnow, J. Mater. Sci. Lett.,2002, 21, 795–797.

PCCP Paper

Publ

ishe

d on

06

Aug

ust 2

014.

Dow

nloa

ded

by S

tanf

ord

Uni

vers

ity o

n 14

/11/

2017

22:

53:4

1.

View Article Online

This journal is© the Owner Societies 2014 Phys. Chem. Chem. Phys., 2014, 16, 19250--19257 | 19257

19 M. Stormer and H.-U. Krebs, J. Appl. Phys., 1995, 78, 7080.20 G. Weigang, H. Hecht and G. V. Minnigerode, Z. Phys. B:

Condens. Matter, 1995, 355, 349–355.21 O. Proux, J. Mimault, C. Revenant-Brizard, J. R. Regnard and

B. Mevel, J. Phys.: Condens. Matter, 1999, 11, 147–162.22 Z. Zhang, T. Nenoff, J. Huang, D. T. Berry and P. P. Provencio,

J. Phys. Chem. Lett., 2009, 113, 1155–1159.23 D. Pletcher and X. Li, Int. J. Hydrogen Energy, 2011, 36,

15089–15104.24 J. W. D. Ng, Y. Gorlin, T. Hatsukade and T. F. Jaramillo, Adv.

Energy Mater., 2013, 1545–1550.25 E. H. Yu, X. Wang, U. Krewer, L. Li and K. Scott, Energy

Environ. Sci., 2012, 5, 5668.26 G. McLean, T. Niet, S. Prince-Richard and N. Djilali, Int.

J. Hydrogen Energy, 2002, 27, 507–526.27 D. E. Brown, M. N. Mahmood, K. Turner and C. Road,

Electrochim. Acta, 1984, 29, 1551–1556.28 L. Chen and A. Lasia, J. Electrochem. Soc., 1992, 139, 3458–3464.29 L. Birry and A. Lasia, J. Appl. Electrochem., 2004, 34, 735–749.30 J. R. McKone, B. Sadtler, C. Werlang, N. S. Lewis and

H. B. Gray, ACS Catal., 2013, 166–169.31 M. Al-Saleh, S. Gultekin, A. Al-Zakri and A. A. A. Khan, Int.

J. Hydrogen Energy, 1996, 21, 657–661.32 S. Lu, J. Pan, A. Huang, L. Zhuang and J. Lu, Proc. Natl. Acad.

Sci. U. S. A., 2008, 105, 20611–20614.33 M. Horigome, K. Kobayashi and T. Suzuki, Int. J. Hydrogen

Energy, 2007, 32, 365–370.34 P. Giannozzi, S. Baroni, N. Bonini, M. Calandra, R. Car,

C. Cavazzoni, D. Ceresoli, G. L. Chiarotti, M. Cococcioni,I. Dabo, A. D. Corso, S. de Gironcoli, S. Fabris, G. Fratesi,R. Gebauer, U. Gerstmann, C. Gougoussis, A. Kokalj, M. Lazzeri,L. Martin-Samos, N. Marzari, F. Mauri, R. Mazzarello, S. Paolini,

A. Pasquarello, L. Paulatto, C. Sbraccia, S. Scandolo,G. Sclauzero, A. P. Seitsonen, A. Smogunov, P. Umari andR. M. Wentzcovitch, J. Phys.: Condens. Matter, 2009, 21, 395502.

35 Instructions for how to obtain a development version ofQuantum ESPRESSO allowing for RPBE calculations can befound here: https://github.com/vossjo/ase-espresso/wiki.

36 G. Kresse and J. Furthmuller, Comput. Mater. Sci., 1996, 6,15–50.

37 M. C. Payne, M. P. Teter, D. C. Allan, T. A. Arias andJ. D. Joannopoulos, Rev. Mod. Phys., 1992, 64, 1045–1097.

38 S. R. Bahn and K. W. Jacobsen, Comput. Sci. Eng., 2002,4, 56.

39 B. Hammer, L. B. Hansen and J. K. Nørskov, Phys. Rev. B:Condens. Matter Mater. Phys., 1999, 59, 7413.

40 D. Vanderbilt, Phys. Rev. B: Condens. Matter Mater. Phys.,1990, 41, 7892–7895.

41 A. A. Adllan and A. D. Corso, J. Phys.: Condens. Matter, 2011,23, 425501.

42 H. J. Monkhorst and J. D. Pack, Phys. Rev. B: Condens. MatterMater. Phys., 1976, 13, 5188.

43 K. J. J. Mayrhofer, G. K. H. Wiberg and M. Arenz,J. Electrochem. Soc., 2008, 155, P1.

44 D. S. Hall, C. Bock and B. R. MacDougall, J. Electrochem.Soc., 2013, 160, F235–F243.

45 B. Hinnemann, P. G. Moses, J. Bonde, K. P. Jørgensen,J. H. Nielsen, S. Horch, I. Chorkendorff and J. K. Nørskov,J. Am. Chem. Soc., 2005, 127, 5308–5309.

46 E. Skulason, V. Tripkovic, M. E. Bjorketun, S. Gudmundsdottir,G. Karlberg, J. Rossmeisl, T. Bligaard, H. Jonsson andJ. K. Nørskov, J. Phys. Chem. C, 2010, 114, 18182–18197.

47 J. Greeley and J. K. Nørskov, J. Phys. Chem. C, 2009, 113,4932–4939.

Paper PCCP

Publ

ishe

d on

06

Aug

ust 2

014.

Dow

nloa

ded

by S

tanf

ord

Uni

vers

ity o

n 14

/11/

2017

22:

53:4

1.

View Article Online