Screeethan

Julien Ca Departmenb Departmen

a r t i c

Article histoReceived 13Received inAccepted 29Available on

Keywords:IridiumAlloysFuel cellEthanol oxiDensity fun

1. Introd

The wcrucial toefficient aare needeincludingmental fohave largsupplied

ited by thstorage custorage m

∗ CorrespE-mail a

http://dx.do0926-860X

Applied Catalysis A: General 483 (2014) 85–96

Contents lists available at ScienceDirect

Applied Catalysis A: General

jou rn al hom ep age: www.elsev ier .com/ locate /apcata

ning iridium-based bimetallic alloys as catalysts for directol fuel cells

ourtoisa, Wenxin Dub, Emily Wongb, Xiaowei Tengb, N. Aaron Deskinsa,∗

t of Chemical Engineering, Worcester Polytechnic Institute, 100 Institute Road, Worcester, MA 01609, United Statest of Chemical Engineering, University of New Hampshire, Durham, NH 03824, United States

l e i n f o

ry: May 2014

revised form 23 June 2014 June 2014line 7 July 2014

dationctional theory

a b s t r a c t

Current ethanol oxidation catalysts in direct ethanol fuel cells (typically platinum-based) suffer fromlow conversion and are susceptible to CO poisoning. We therefore determined to find viable alternativecatalysts for ethanol oxidation based on iridium using density functional theory to model bimetallic alloy(1 1 1) surfaces. Iridium was alloyed with another transition metal (from groups 4 through 11), M, in anoverlayer (one layer of metal M on top of bulk iridium) or subsurface configuration (M inserted underthe first layer of iridium). Segregation energies were calculated and the subsurface configuration wasfound to be most stable configuration in the majority of alloy cases under vacuum conditions. Selectalloys (including Ir–Pt and Ir–Rh) showed a reversal of preferred alloy state (overlayer or underlayer)upon adsorption of CO. CO adsorption was modeled and lower CO adsorption energies were found formany alloys compared to pure Pt, indicating less CO poisoning. Complete oxidation of ethanol is typicallylimited by the breaking of strong C C bonds, so any catalyst must lower the barriers for C C bond break-ing. Activation energies for C C bond breaking (by considering a representative intermediate moleculeCHCO) were lowered for the vast majority of the alloys used in an underlayer structure, reinforcing thesignificance of the underlayer structures or “subsurface” alloys. Finally, we found, based on CO adsorptionenergy, activation energy for C C breakage reaction, and metal cost, three important catalyst descriptors,

a number of promising catalysts for the ethanol oxidation reaction, such as alloys with M = Zr, Nb, Mo, W,Ru, Os, or Re. Experimental verification (cyclic voltammetry and chronoamperometry for ethanol oxida-tion) on Ir, Ir–Ru, and Ir–Rh catalysts confirmed the superiority of these Ir-based catalysts compared toPt. Our results demonstrate that alloys of Ir show promise as more affordable substitutes for platinumgroup metals.

© 2014 Elsevier B.V. All rights reserved.

ryog. Cheis dif

suchce beity [relatrtan

low

e ca

uction

orld’s energy demand is continually growing, and it is find alternatives to current fossil fuels. Indeed, morend environmental friendly energy production processesd. Numerous solutions are currently being investigated

fuel cells, which may produce energy with low environ-otprint. Polymer electrolyte membrane (PEM) fuel cellse theoretical efficiency when a fuel, such as hydrogen isto the device, much more than combustion, which is lim-

or a c[2,3]but

holschoidensalso

impoS

anod

e Carnot cycle efficiency [1]. Nevertheless, the problem ofrrently hinders the use of hydrogen as a fuel. For example,ay require high pressure tanks operating at 200–700 bar,

onding author. Tel.: +1 508 8315445.ddress: nadeskins@wpi.edu (N.A. Deskins).

(DEFCs).

oxidationreleases otion howthe availreactions

CH3CH2O

i.org/10.1016/j.apcata.2014.06.029/© 2014 Elsevier B.V. All rights reserved.

enic tanks working at low temperatures (such as −223 ◦C)mical storage of hydrogen has also being investigated,

ficult. Other fuels have been considered, including alco- as ethanol or methanol. Ethanol is an attractive fuelcause it is easily stored as a liquid and has high energy

1,4–6], similar to gasoline’s energy density [7]. Ethanol isively non-toxic. Furthermore, ethanol can be produced int quantities from biomass [5].kinetics for the ethanol oxidation reaction (EOR) at thetalyst limits the efficiency of direct-ethanol fuel cellsFor example, over current Pt-based catalysts incomplete

of ethanol produces acetaldehyde and acetic acid whichnly 2 and 4 electrons, respectively [1,4]. Complete oxida-

ever forms CO and releases 12 electrons [1,4], increasing

2able electrical energy, as indicated by the following:

H + 3H2O → 12H+ + 12e− + 2CO2(totaloxidation) (1)

86 eral 4

2CH3CH2

+ CH3C

CurrenCO poisocatalysts

oxidationtoward Csively stuhave beealloys orcarbides

another

[5], has

ferent alto near s[12,13].

The bbe the brC C scissoxidationtion doescommonstudies h[14–17].

catalysts

(BEP) corsible intemost traing step

separatedDFT studcleavage

scission

(CHCO).Platin

catalytic

as Pt, PdEOR. Manthe literaRh [3] aducted wreported

catalyst,

the EOR

Ir (in thcandidateity compcore–sheIr–Ru alloor Ir.

Motivreactivityusing cortransitiona potentitifying pmodelingfor potenenergies

relation

alloys toals (surfaunder whwhich all

lyticaies.

etho

Com

. Simll caPBE

Gausities

tal w. Core) ps

ples

lt in

rathd. Wertierizedted tnly 1latio

. Suhe Ir

6 su repehborom laationisted

metacingr strMavce aslab.11 (

Ir.dsorula:

rption

re Ea

ace isnergH, a

bondn reCO

the Pher (d to

rptiothe mtionsrptiodsor

k alsorenc

J. Courtois et al. / Applied Catalysis A: Gen

OH + H2O → 6H+ + 6e−

OOH + CH3CHO(partialoxidation) (2)

t Pt-based catalysts are expensive and also suffer fromning. It is therefore necessary to find new electro-that could increase the overall kinetics of the ethanol

reaction and increase the selectivity of the reactionO2 formation [3]. DEFC anode catalysts have been exten-died, with much focus on platinum. Various attemptsn employed to reduce the Pt loading, such using

carbon supports or even using other materials likeand oxides as anode catalysts. Alloying, by adding

element [3,5,8–10] or by even adding a third elementbeen a popular method of catalyst development. Dif-loy structures have been studied, from metal dopingurface alloys (NSA) [11] or even core/shell structures

ottleneck for efficient ethanol oxidation is believed toeaking of strong C C bonds [3]. Various C H, C O, orion steps may occur at the catalyst surface, but complete

involves C C bond breaking, while incomplete oxida- not. Density functional theory (DFT) calculations are

ly used to identify new potential catalysts and numerousave been performed for ethanol oxidation/decompositionFerrin et al. [18] screened various transitions metals asfor ethanol decomposition using Brønsted–Evans–Polanyirelations to determine the activation barriers for pos-rmediates and they have been able to show that, for

nsitions metals, including iridium, the rate determin-for C C cleavage involves a CHCO intermediate being

into two molecules, CH and CO. Another previousy led by Alcala et al. [15] investigated C O and C Cpathways over Pt for ethanol and concluded that C Cwas also likely to proceed through a ketenyl species

um group metals are well known for their exceptionalproperties. Ferrin et al. [18] reported that metals such, Ru, Ir or Rh could be interesting candidates for they experimental or theoretical studies can be found inture concerning Pt [10,19–21], Pd [6], Ru [22–24] ands catalysts for the EOR but less work has been con-ith Ir. Recent experimental work with Ir and Sn was[25] and showed comparable results to that of a Pt3Sn/Ccurrently one of the most efficient anode catalysts for[10,16]. More recently, Du et al. [4] also showed that

e form of Ir–Sn–SnO2) could be a very good catalyst for the EOR, obtaining a high electrochemical activ-

arable to a commercial PtSn catalyst. They observed all structure for the Ir alloy. The same group developed ay [26] which showed enhanced activity compared to Pt

ated by previous experimental and theoretical work, the of Ir-based catalysts for the EOR has been investigated,e–shell-like bimetallic structures made of Ir and another

metal. Emphasis was placed on C C bond breaking asal rate determining step (CHCO → CH + CO) and in iden-otential catalysts to replace Pt. Our efforts focused on

a large number of alloys and screening these alloystial ethanol oxidation activity. Accordingly, activationwere calculated using a previously established BEP cor-[27]. We also examined the resistance of the studied

catastud

2. M

2.1.

2.1.1A

The

the

densorbi[32](GTHsamresuwasmizeproppolaculaby ocalcu

2.1.2T

a 6 ×wasneigbottfigurconstionrepllayeand

surfathe

and

withA

form

Eadso

wheEsurfthe eCO, CC Cdatioof CHoverto otfounadsoare

culaadsoCH awordiffe

CO adsorption. The preferred location of alloyed met-ce or sub-surface) was examined and we determinedat conditions such alloys are stable. Our analysis indicatesoys would potentially be economically feasible and also

[26,38]. Fgeometryposition

used was

83 (2014) 85–96

lly active, suggesting directions for future experimental

dology

putational methodology

ulation parameterslculations were performed using the CP2K code [28,29].exchange correlation function was used [30]. CP2K usessian and Plane Wave (GPW) [31] method so electron

were treated by plane wave functions and molecularere represented by double-zeta Gaussian basis functions

electrons were represented by Goedecker–Teter–Huttereudopotentials [33,34]. The current version of CP2K onlyreciprocal space at the � point, which could potentiallysimulations errors. Consequently, the slab that we useder large so any errors from k-point sampling were mini-ith the exception of Ni, Co, and Fe, which display magnetics, all bimetallic structures were modeled as non-spin, similar to previous work [35]. For instance, we cal-he CO adsorption energy over a Au–Ir alloy to differ0−4 eV between spin polarized and non-spin polarizedns.

rface model (1 1 1) surface was modeled using the slab approach, withper cell under periodic boundary conditions. Each slab

ated periodically with at least 30 A of vacuum betweening metal slabs and was three atomic layers thick. Theyer of the slab was kept frozen. Two different alloys con-s were used as shown in Fig. 1: the first configuration

of replacing the top layer of Ir with another transi-al (overlayer structure), while the second consisted of

the middle Ir layer with another transition metal (under-ucture). Similar alloy structures were used by Greeleyrikakis [11] or by Su et al. [8] in their study of near-lloys. Adsorption was allowed only on the top layer of

Transition metals between groups 4 (Ti, Zr, and Hf)Cu, Ag, and Au) of the periodic table were all alloyed

ption energies were determined according the following

= Eadsorbate/surface-Esurface − Eadsorbate in gas phase (3)

dsorbate/surface is the energy of the surface/adsorbate system, the energy of the clean surface, and Eadsorbate in gas phase isy of the gas-phase molecule. We modeled adsorption of

nd CHCO over the various alloys. As previously mentioned, breaking is typically the hardest step in the ethanol oxi-

action. Work by Alcala et al. [15] showed that C C breakinghad the lowest barrier of various possible intermediatest (1 1 1) surface. Later work by Ferrin et al. [18] expanded

1 1 1) surfaces, including Ir. The rate-determining step wasalso involve CHCO C C scission. Hence, we modeled CHCOn and its derivatives. CO was adsorbed on top sites, whichost preferred sites over Ir (1 1 1) according to our cal-

(see Fig. 3) and previous experimental testing [36]. CHn calculations by Abild-Pedersen et al. [37] reported thatbs preferentially on threefold sites (hcp or fcc). Previous

indicated that CH over Pt (1 1 1) prefers hollow sites, withes in energies between hpc and fcc sites less than 0.1 eV

ig. 2 shows CO and CH geometries. The adsorbed CHCO

had the CH group near a hcp site, so only CH in the hcpwas modeled. The initial geometry of adsorbed CHCO we

based on the work of Alcala et al. [15]. This geometry

J. Courtois et al. / Applied Catalysis A: General 483 (2014) 85–96 87

Fig. 1. Ir (1 er strurespectively

however

several athese alloor hollowpresented

2.1.3. MoCP2K

culated scomparabown testsour adsor[36]. Thein a 2 × 2the metaon one sfixed in tbasis setresults fofrom thismethod.

2.1.4. LinCalcul

can be veeral attemdetermining correlcorrelatioa functionstate of aity of thes[15,16,18scaling toCHCO. Wgives the

ETS = �Ed

� =

Ediss iefereiss is

= EC

CO-su

urfacee enphas

Expe

. Synarboaredrfac

pical mm

Mall a prG sopletef Ir4RydroAesae-necu23,8.0 murso170 ◦

plete acti

1 1) alloyed surface showing a Pt/Ir overlayer structure (top) and Pt/Ir underlay, gray and blue.

did not necessarily converge to a stable configuration onlloys, so we modeled several other initial geometries forys, which involved the two C atoms either in top, bridge,

sites. The energies for the most stable configurations are in the current paper.

del verificationhas previously been used to model Pt [39], where cal-urface and bulk properties of platinum were foundle to previous literature results. We also conducted our

in order to validate our approach. In Fig. 3, we compareption results over Ir to those obtained by Krekelberg et al.y used a four-layer Ir (1 1 1) slab, periodically repeated

unit cell with five equivalent layers of vacuum betweenl slabs. Similar to our work, adsorption was only allowedide of the slab and the first layer of metal atoms washe bulk positions. Krekelberg et al. used a plane wave

and the PW91 exchange correlation functional0]. Ourr adsorption energies do not differ by more than 10%

previous work, confirming the adequacy of our simulation

ear scaling correlationsating activation energies from an initial and final statery computationally time-consuming, and therefore sev-pts have been made to develop empirical correlations to

e activation barriers. The general idea of the linear scal-ations (often referred to as Brønsted Evans Polanyi or BEPns) is to express the energy of the transition state (ETS) as

of the difference in energy between the final and initial reaction, or reaction energy �Erxn. The accuracy and util-e correlations have been demonstrated by several authors,27,41]. For this work we took advantage of these linear-

calculate activation energies for C C bond breaking of

withand

are rso Ed

Ediss

EECO-sis thgas-

2.2.

2.2.1C

prepof sua ty(0.08(EG,intoof Ecomsis oof hAlfa

threIr77Rand

precEG (comwith

e used the parameters published by Wang et al. [27], which following equation:

iss + �, (4)

1 h, washvacuum alowed bysurfactan

cture (bottom). Side and front views are shown. Pt and Ir atoms are,

0.85 eV and � = 2.05 eV. ETS is the transition state energys the dissociation energy (reaction energy) for CHCO. Bothnced relative to the reactant molecule in the gas-phase,

defined by the following formula:

O-surface + ECH-surface − 2Esurface-ECHCO-gas. (5)

rface is the energy of the surface with adsorbed CO,is the energy of the surface with adsorbed CH, Esurface

ergy of the bare surface, and ECHCO-gas is the energy of ae CHCO molecule.

rimental methodology

thesis of Ir, Ir80Rh20, Ir77Ru23 nanoparticlesn supported Ir, Ir80Rh20, and Ir77Ru23 nanoparticles were

via the Polyol method with or without the presencetant Polyvinylpyrrolidone (PVP) as stablizing agent. In

synthesis of Ir nanoparticles, 28.2 mg of hydrous IrCl3ol, Alfa Aesar, 99.9%) was dissolved in 3 ml ethelyne glycolinckrodt Chemicals) upon sonication, and then injectede-heated (170 ◦C) 50 ml three-neck flask containing 8 mllution. The reaction proceeded for 30 min to allow the

reduction of Ir3+ into Ir metal. Similarly, for the synthe-h1 nanopariticles, a 4 ml EG mixture containing 28.2 mg

us IrCl3 (0.08 mmol) and 8.6 mg of K3RhCl6 (0.02 mmol,r, 99.99%) was injected into the pre-heated (170 ◦C) 50 mlk flask containing 7 ml of EG solution. In the synthesis of

31.8 mg of hydrous IrCl3 (0.09 mmol, Alfa Aesar, 99.9%)g of hydrous RuCl3 (0.03 mmol, Alfa Aesar, 99.9%) metal

rs were dissolved in 4 ml of EG and injected into 7 ml of hotC) which contained 55 mg of PVP. We allowed 30 min to

the reaction. These resulting particles were then mixedve carbon (Vulcan, XC-72R) at room temperature (RT) for

ed with copious ethanol and DI water, and dried out undert RT. The Ir77Ru23/C was further treated at 300 ◦C in air fol-

H2 flow at 100 ◦C using a tube furnace to remove residualt [26].

88 J. Courtois et al. / Applied Catalysis A: General 483 (2014) 85–96

Fig. 2. Top and side views of CH and CO

Fig. 3. Comparison of adsorption energies (in eV) for H, CH, CH3, CO and O overIr (1 1 1) surface slabs. Filled markers correspond to our work while the unfilledmarkers correspond to values extracted from Krekelberg et al. [36].

2.2.2. EleThe el

Ir77Ru23using a Cinstrumecarbon wwire coutrode wawere prewhile thewas madink was

Upon dryV0.5 wt% Nalyst layerate of thmetry (CV(99.999%The electalysts we

adsorption.

ctrochemical measurementsectrochemical properties of carbon supported Ir, Ir4Rh1,and commerical Pt/C (ETEK) catalysts were evaluatedHI 660 single channel electrochemical workstation (CHnts, Inc.). The three-electrode system composed of a glassyorking electrode (rotating disc electrode, RDE), a platinumnter electrode and an Ag/AgCl (1 M KCl) reference elec-s employed for the test. All the Ir-based electrocatalystspared in de-ionized water making a 2.5 mg/ml suspension

commercial Pt/C (ETEK, 20 wt.% of Pt) water dispersione as 1.25 mg/ml by sonication. A volume of 10 �l catalystdrop-casted on the RDE to form a thin layer of catalyst.ing in a vacuum, another 10 �l of Nafion solution (0.5%,

afion:Vwater = 0.05 ml:10 ml) was dropped on top of the cat-r and dried prior to the electrochemical tests. The rotatinge RDE was controlled at 1000 rpm. Blank cyclic voltam-

) was firstly performed in a Ar-purged 0.5 M sulfuric acid

, Sigma-Aldrich) solution recording from −0.25 V to 0.35 V.rochemically active surface areas (ECASAs) of Ir-based cat-re estimated by choosing −0.22 V vs. Ag/AgCl as the lower

eral 4

boundarythat the Iradsorptiotion of chand 210 �used in Etively.

The pments (CH2SO4/0.were recactivity tokept at 3held at 0.ments rigpotentialsurementat ambiewith respCA and CVat least sde-aerate

3. Result

3.1. Segre

The firstructuretion weretwo strucwith each

Eseg = Eo

where Eustructurenatom is thenergies

that the ution enersegregatiaction beand silveing that tseparate

positive sis thermo

Previoture, suchenergy issegregatiimpurity

presents

shown inwith our rthe valuetable, or

segregate

3.2. Adso

3.2.1. COCO is a

how resiing viable

CO

latioideri, so bgies

s andver

ig. 3um-beren1) su7 eV

k on

d onwer

d onr Ir–Cstabhat

contioned m

not fide ade rrely depreot clloy

s, lik stru

etheltivitynderave Ce therptio

of As to

undeloweindinture

, rece allod to

ly elr alteken os areRe, Oe allrptiod whs areub-s

d me CO a

n to whidsore ene

J. Courtois et al. / Applied Catalysis A: Gen

for integration of charge in the backward scan. We note77Ru23 only showed poorly defined features around the Hn/desorption region and the lower boundary for integra-arge was somewhat arbitrary. The charges of 220 �C/cm2

C/cm2 for a monolayer of H adsorption/desorption wereCSA calculations for Ir–Ru/C catalysts and Pt/C, respec-

olarization curves and chronoamperometry measure-A) of those electrocatalysts were also conducted in 0.5 M5 M ethanol solution. The first cycles in CV measurementsorded from −0.1 V to 0.35 V to determine their catalyticward ethanol oxidation. The potential sweeping rate was

0 mV/s for all the CV measurements. The potential was3 V for Ir/C, Ir4Rh1/C, and Pt/C for 1 h in the CA measure-ht after the completion of polarization curves. The applied

however, was held at 0.2 V for Ir77Ru23/C in the CA mea-. All the electrochemical measurements were performed

nt room temperature and the potentials were reportedect to Ag/AgCl (1 M KCl) unless otherwise specified. The

plots presented in this work are the averaged data afterix repeats with high reproducibility. The electrolyte wasd by bubbling Ar 30 min prior to the measurements.

s and discussion

gation of the alloy

st step of our approach was to examine which of the twos (overlayer and underlayer, see Fig. 1) under considera-

the most stable. To evaluate the relative stability of thetures, we calculated the segregation energy associated

transition metal by the following:

verlayer − Eunderlayer

natom, (6)

nderlayer is the calculated energy of the underlayer alloy, Eoverlayer is the energy of the overlayer alloy structure, ande number of atoms in the slab. The calculated segregation

are given in Table 1. Positive segregation energy indicatesnderlayer structure is preferred while negative segrega-

gy indicates that the overlayer structure is preferred. Theon energy is thus an indicator of the strength of inter-tween the metal alloyed with iridium. For instance, goldr have very strong negative segregation energies, indicat-hese elements would prefer to segregate and potentiallyfrom an iridium alloy. Many early transition metals haveegregation energies, indicating that alloying with iridiumdynamically feasible.us similar calculation results can be found in the litera-

as by Ruban et al. [35]. Our definition of the segregation slightly different from theirs since they calculated theon energy based on an iridium structure with only oneatom in the top or second layer. Nevertheless their worka good comparison with our calculations. Their results are

parenthesis in Table 1 and typically are in good agreementesults. The general trend for the segregation energy is thats decrease as one moves toward the right of the periodicthat late transition metals have a greater preference to

from iridium.

rption energies

fromcalcuconssionenertionCO oby Firidia ref(1 1

−2.1marbaseor lobaselayeto a

ing talloymizashoware

provinclusevenot rdid nthe aalloyalloyNonreacthe uals hwhiladsotiontendThe

and

CO bstrucisonIr/Ruparelargelayeweature(W,

whiladsotrenalloythe stutethatknowcles,CO ais th

potential poison for fuel cell catalysis, so understandingstant potential catalysts are to CO is crucial to identify-

Pt replacements. Any potential catalyst should not suffer

layer) accmagnitudIr/Pt), indCO than p

83 (2014) 85–96 89

poisoning problems like Pt. We therefore utilized DFTns to establish whether CO poisoning is an issue worthng for Ir-based alloys. CO is also a product of CHCO scis-inding energies for CO can be used to calculate activationfor the C C bond breaking. According to our calcula-

previous work [36], the most stable adsorption site fora pure iridium (1 1 1) slab was the top site, as indicated. We thus chose to model CO in the top site over ourased surfaces, and our results are reported in Table 2. Asce, we calculated the CO adsorption energy on a pure Ptrface (using a similar cell to those shown in Fig. 1) to be. As a first approximation, this value serves as a bench-whether a catalyst is more or less CO-resistant than Pt,

whether the adsorption energy over the alloy is higherthan this value. We note that two values in the table are

modified calculations, namely the values for the under-o and Ir–Ni alloys. These calculations did not converge

le geometry after many iterations and attempts, indicat-such a system is relatively unstable, possibly due to thefiguration. We thus froze the alloy surface during opti-

of CO adsorption similar to previous work [18] whichinimal effect due to full relaxation. These values thus

ully correct since surface relaxation is ignored, but do first estimate of CO adsorption values. We also do not

esults for the Ir–Zr overlayer structure since this alloy wasistorted from the model (1 1 1) surface structure and thus

sentative of a real alloy. Several adsorbed structures alsoonverge (discussed further in the paper) indicating thatmodel we used, a simplistic way to consider near surfaceely requires further refinement. Such refinement of thectures however is beyond the scope of the current work.ess, as our results will show, the trends in Ir-based alloy

provide important insight into catalyst development. Alllayer bimetallic structures involving early transition met-O adsorption energies smaller in magnitude than over Pt,

underlayer structures for later transition metals have COn energies more negative than over Pt (with the excep-

u). For the underlayer structures, the CO binding energyincrease when moving to the right of the periodic table.rlayer structures are more stable for metals of group 8r (see Table 1) and these preferred alloys all have smallerg energies than Pt, suggesting several alloys with stable

s that are more CO-resistant than Pt. As a point of compar-nt experimental results by Du et al. [26] indicated that any demonstrated superior CO anti-poisoning ability com-Pt. For the underlayer structures, the effect of alloying isectronic. The top surface layer is still Ir, but the sub-surfacers the electronic orbitals of the Ir atoms sufficiently tor strengthen CO binding. The trends in the overlayer struc-

not so apparent. Several group 4 (V, Cr, Mn) and group 6s, Ir) metals show increased CO adsorption relative to Pt,

group 5 metals having the overlayer structure have COn energies less exothermic than Pt. There is no apparenten moving across the periodic table. The top layers of the

composed of the substituted metal, so the presence ofurface Ir modifies the electronic structure of the substi-tals, but the trends in this effect are not clear. We notedsorption energies for the Au alloys are all very low. Au is

generally be inert, except typically for nano-sized parti-ch is consistent with these results. We also note that theption energy over all the preferred alloy structures, thatrgetically most stable alloys (either overlayer or under-

ording to Table 1, have CO adsorption energies lower ine than pure Pt (with the exception of pure Ir, Ir/Cu andicating that several Ir-M alloys may be more resistant toure Pt. Our DFT results show that CO poisoning is less of

90 J. Courtois et al. / Applied Catalysis A: General 483 (2014) 85–96

Table 1Calculated segregation energies as calculated according to Eq. (6). Energies are given in eV. Segregation energies calculated by Ruban et al. [35] are given in parenthesis. Darkshaded cells correspond to alloys where the underlayer structure is preferred, or the segregation energy is positive.

Ti V Cr Mn Fe Co Ni Cu

0.53 (0.29) 0.55 (0.51) 0.50 (0.35) 0.36 (0.09) 0.20 (0.11) 0.16 (0.16) 0.07 (0.12) −0.12 (−0.12)Zr Nb Mo Tc Ru Rh Pd Ag0.36 (−0.43) 0.42 (0.10) 0.43 (0.35) 0.23 (0.35) 0.09 (0.23) −0.04 (−0.08) −0.22 (−0.55) −0.30 (−1.00)Hf Ta W Re Os Ir Pt Au0.42 (−0.17) 0.50 (0.26) 0.48 (0.47) 0.25 (0.48) 0.12 (0.32) 0.00 (0.00) −0.21 (−0.58) −0.43 (−1.20)

Table 2CO adsorption energies on both under- and overlayer alloy surfaces. Shaded values correspond to alloys where CO binds more strongly than on pure platinum. Pure Pt hasa calculated adsorption energy of −2.17 eV. All energies are given in eV. Adsorption energies for the Ir–Zr overlayer structure are not given due to the extreme distortionsobserved in the Ir–Zr surface.

Ti V Cr Mn Fe Co Ni Cu

Overlayer −2.07 −2.48 −2.59 −2.73 −1.94 −1.97 −1.92 −1.10Underlayer −1.66 −1.71 −1.55 −1.70 −1.96 −1.06 −1.53 −2.43

Zr Nb Mo Tc Ru Rh Pd AgOverlayer − −1.92 −2.07 −2.01 −2.10 −1.96 −1.44 −0.39Underlayer −1.71 −1.60 −1.63 −1.99 −2.00 −2.27 −2.58 −2.61

OverlayerUnderlay

a problemlower CO

3.2.2. CHLitera

at hollowin which

over hcpCH adsorenergies

appear inunderlay(<−17 eVlikely indhave modsimplisticplicated

Ir–Zr ovestructureindicativeis simplissurface. Tor underlnificant showever,comparin

3.2.3. CHThe in

identifiedgeometryplatinumthe underuration ptop-like pFor the ofied as stathat is distructureals the seoverlayernot invol

s inble 4dsorn a ferptioh a sce,

. In these cases the CHCO adsorbate converged to a geometryre only the carbon atom linked to the hydrogen atom wasd to the surface, while the second carbon atom moved away

the surface. These surfaces are apparently resistant to CHCO

Hf Ta W Re

−1.79 −2.08 −2.30 −2.21

er −1.66 −1.66 −1.55 −1.97

for most Ir-based alloys compared to Pt as indicated by adsorption energies for the majority of the studied alloys.

ture studies strongly suggest that CH preferably adsorbs sites [37,42]. Similar to our approach with CO adsorptiontop sites were considered, we modeled adsorption of CH

sites for the alloys under consideration. Our calculatedption energies are reported in Table 3. Most adsorptionappear to be near −6 to −7 eV. A few notable exceptions

the results. Binding energies for the Cr overlayer and Mner/overlayer structures appear to be unrealistically large) and are thus not reported. These unrealistic values areicative of the unstable nature of the alloy structure. Weeled only underlayer and overlayer structures, which are

representations of the alloy surface. More stable, com-structures may exist for these alloys. In the case of therlayer alloy surface CH binding (also not reported) the alloy

distorted significantly. Again, these distortions are likely that the Ir–Zr alloy overlayer structure that we employedtic and may not be representative of an actual Ir–Zr alloyhe models we used, the inclusion of an entire overlayerayer of alloyed metals into an iridium slab, may have sig-train for some alloy combinations. These alloy models,

allow a computationally fast approach to modeling andg many alloy combinations.

COitial geometry that we used for CHCO was the geometry

by Alcala et al. [15] over a Pt (1 1 1) surface. This initial appeared to be stable when the surface was made of a

group metals, and thus all CHCO adsorption energies forlayer structures were calculated with CHCO in the config-

resented on Fig. 4a. In this geometry the CO is oriented in aosition, while the CH is oriented in a bridge-like position.

verlayer structures, two different structures were identi-ble. The first geometry was the one proposed in Ref. [15]

splayed in Fig. 4a, and was encountered with underlayer

atomin Ta

AAgaiadsoreacsurfatrieswhebounfrom

s and overlayer structures involving platinum group met-cond geometry is represented in Fig. 4b for the case of a Ti

and was more stable for all the other overlayers structureving platinum group metals. This geometry had the two C

Fig. 4. (a) CTi/Ir overlagen atoms ainterpretatto the web

Os Ir Pt Au−2.35 −2.25 −1.63 −0.63−1.98 −2.25 −2.38 −0.13

hollow-like positions. A summary of our results is found.

ption energies of CHCO varied between −1.7 and ∼−7 eV.w problems were encountered in the simulations. CHCOn over the silver and gold overlayer structures did nottable structure where both C atoms were bound to thedespite modeling a number of different initial geome-

HCO geometry on a Ru/Ir overlayer structure. (b) CHCO geometry on ayer structure. Ruthenium, titanium, iridium, carbon, oxygen and hydro-re, respectively, turquoise, light gray, blue, dark gray, red and white. (For

ion of the references to color in this figure legend, the reader is referredversion of this article.)

J. Courtois et al. / Applied Catalysis A: General 483 (2014) 85–96 91

Table 3CH adsorption energies on both under- and overlayer structures. Energies are given in eV. Adsorption energies for the Ir–Mn alloys, Ir–Cr overlayer, and Ir–Zr overlayer arenot given due to difficulties in alloy stability as discussed in the text. Pure Pt has an adsorption energy of −7.46 eV. Shaded values correspond to alloys where CH binds morestrongly than on pure platinum.

Ti V Cr Mn Fe Co Ni Cu

Overlayer −7.36 −8.58 – – −6.71 −6.98 −6.68 −4.82Underlayer −6.51 −6.48 −7.92 – −6.64 −7.91 −6.47 −8.72

Zr Nb Mo Tc Ru Rh Pd AgOverlayer – −6.71 −7.15 −6.91 −6.96 −6.71 −5.75 −3.20Underlayer −7.06 −6.81 −5.99 −6.57 −6.72 −6.91 −7.62 −8.09

OverlayerUnderlay

adsorptiooverlayeralytic actoverlayerthe pointalloy strualready dWe thereand do nlayer struto very ushowed tmodel duthese twoof reactiofested thactivationear scalinenergy ofalloys haThese areunderlaysurfaces tor more

(most unCHCO is lnear −2 t

3.3. Effec

The repreferred[11] showin select

adsorbateing how tdesigningsegregatiof a bare

and over

rbate thatn Ir/

the uO ov8 eV

adson en

migrffectrs.

Figdifferrlay

anged by

rlay mose areegatilvindsor

ther

prefende

struds fo

effelyst s

Alloy

hererption effrent

Table 4CHCO adsorto alloys wh

OverlayerUnderlay

OverlayerUnderlay

OverlayerUnderlay

Hf Ta W Re

−6.69 −7.13 −7.45 −7.25

er −6.96 −6.83 −5.86 −6.61

n (as indicated by the lowest adsorption energies of the structures), and are therefore expected to have low cat-ivity. The adsorption energies of CHCO over the Ir–Cr

and Ir–Mn structures are all extremely exothermic to of being unrealistic (<−15 eV), which indicates that thesectures are likely unstable and/or unrealistic models, asiscussed in the section on CH adsorption (see above).fore do not consider these adsorption energies reliableot use these values in further analysis. Zr and Hf over-ctures demonstrated large buckling at the surface leadingnrealistic structures. The strongly deformed structureshat the overlayer structures were an unlikely realistice to strain so we also did not consider results from

geometries further in this paper. Generally, for a classns the more stable an adsorbed intermediate is (as mani-rough the adsorption or reaction energy), the lower the

energy. This is mathematically constructed as a lin-g relationship (see Eq. (4)). Our calculated adsorption

CHCO was found to be −4.15 eV over Pt (1 1 1). Severalve more exothermic adsorption energies compared to Pt.

overlayer structures of early transition metals and a fewer structures of late transition metals. On some of thesehe CHCO adsorption energies are more exothermic by 1eV compared to Pt. On the other hand for many alloysderlayer and late transition metal overlayer structures)ess stable than over Pt, with adsorption energies typicallyo −3 eV.

ts of adsorbates on segregation

action environment may change the thermodynamically state of the alloy. For example Greeley and Mavrikakised that hydrogen adsorption could induce segregationalloys. Other work also showed that the presence ofs can alter segregation preferences [43–45]. Determin-he reaction environment affects segregation is crucial for

adsosuchfor athatfor C−2.4CO isgatiothe

net eoccu

Inthe

undea chcateundeoverThersegr(invoCO awheRh)

the ualloytrenSuchcata

3.4.

Tadsostraidiffe

and understanding stable catalysts. We thus calculatedon energies in the presence of CO. The segregation energysurface represents the relative stability of the underlayerlayer structures in the absence of adsorption. When an

as they astrain indexample,under com

ption energies on both the under- and overlayer alloy structures. Energies are given in eV.ere CHCO binds more strongly than on pure platinum. Several values are not listed due t

Ti V Cr Mn

−5.85 −7.09 – –

er −2.74 −2.56 −4.33 –

Zr Nb Mo Tc

− −4.64 −4.76 −4.22

er −3.00 −2.58 −2.30 −2.96

Hf Ta W Re

− −5.22 −5.09 −4.26

er −2.85 −2.58 −2.17 −3.07

Os Ir Pt Au−7.22 −6.94 −6.09 −3.99−7.41 −6.94 −7.17 −8.48

is present, the energy released upon adsorption may be the segregation trend changes. This is illustrated in Fig. 5V alloy. The bare segregation energy is 0.55 eV, indicatingnderlayer structure is preferred. The adsorption energieser the overlayer and underlayer structures are −1.71 and, respectively. There is 0.77 eV more energy released whenrbed on the overlayer alloy, which is more than the segre-ergy of 0.55 eV. Consequently, CO adsorption may induceation of the sub-layer V atoms toward the surface. The

is a gain of 0.22 eV (0.77–0.55 eV) when the segregation

. 6 we have plotted the bare segregation energies versusence in CO adsorption energies between overlayer and

er structures. This plot shows which alloys may undergo in segregation preference in the presence of CO, as indi-the gray areas. The majority of bare alloys prefer to former structures (positive segregation energies). Howevert of the alloys CO prefers to bind to overlayer structures.

a number of alloys where this preference may induceon upon CO adsorption, such as Os, W, or V. Other alloysg Ru, Re, and Mo) have segregation energies close to theption energy differences, so it is difficult to determinesegregation may actually occur. A few bare alloys (Pt andr the overlayer structure, but upon CO adsorption preferrlayer structure. Our results have only considered selectctures and one adsorbate, but do show that segregation

r Ir–M alloys can change due to environmental conditions.cts may be useful for catalyst synthesis and also may affecttability.

ing and surface chemistry

are two major effects related to alloys that influencen energies and surface chemistry [46,47]. The first is theect. Because the bond lengths and lattice parameters arefor each metal, alloying may put strain on metal atoms

dapt new bond lengths to fit into the parent metal. Thisuces changes in the electronic orbitals of the metals. For

under tensile strain an upshift in the d-band occurs, whilepressive strain the d-band moves down in energy. These

Pure Pt has an adsorption energy of −4.15 eV. Shaded values correspondo problems of convergence/distortion as discussed in the text.

Fe Co Ni Cu

−3.35 −4.34 −3.90 −2.42−3.09 −3.60 −2.74 −5.22Ru Rh Pd Ag−3.21 −3.20 −2.90 −1.73−3.03 −3.64 −4.26 −4.68Os Ir Pt Au−3.59 −3.58 −3.15 −2.24−3.17 −3.58 −3.93 −5.33

92 J. Courtois et al. / Applied Catalysis A: General 483 (2014) 85–96

Fig. 5. Illus adsorenergy over ation e

shifts in dstrengtheshown aning refere

Fig. 6. Stabsegregationtive values

values for tthe underlaover the ovadsorbate-isented on tof their seg(Mn) on the

he se

tration of adsorbate-induced segregation for the case of an Ir/V alloy. The energy of the underlayer structure by 0.77 eV, which is larger in magnitude than the segreg

-band energies can weaken (downward d-band energy) or T

n (upward d-band energy) adsorption. Such effects haved discussed in the literature (see for example the follow-nces: [46,47–50]).

ility of bimetallic structures with respect to CO-induced segregation. The energies indicate the most stable structures for the bare surfaces. Posi-correspond to alloys where the underlayer surface is preferred. Positivehe difference in binding energy indicates that CO prefers to adsorb overyer structure, while negative values indicate that CO prefers to adsorberlayer structure. Alloys situated in the gray area are likely to undergonduced segregation. Alloys containing Pd, Ag, Cu and Mn are not repre-his figure for clarity purposes since they have a very high absolute valueregation energy and are situated either far right (Pd, Ag, Cu) or far left

graph in an area not influenced by adsorbate-induced segregation.

betweenof anothemetals duthe undestrain in

changed)the liganture, thestrain antypes maon cataly

We haunderlaythe samefor the unmetals toin terms

adsorptio

Fig. 7. CO agies are plotable.

ption over the overlayer structure is more negative than the adsorptionnergy, and so the segregation trend is reversed.

cond effect arises due to intrinsic electronic differences two metals and is called the ligand effect. The presencer nearby metal may alter the electronic structure of thee to new interactions between the metals. In the case of

rlayer structure, the top layer of Ir atoms are only underthe vertical direction (the surface lattice parameter is not

so the predominant change in adsorption is attributed tod, or electronic, effect. In contrast, for the overlayer struc-

top layer of surface metal atoms may experience bothd ligand effects. Thus, comparing the two alloy surfacey provide clues as to the role these two effects may havesis.ve plotted the adsorption energies for CO and CH over theer (Figs. 7 and 8) and overlayer (Figs. 9 and 10) structures;

data is presented in Tables 2 and 3. Our results show thatderlayer structures stronger adsorption occurs for alloyed

the right of the periodic. These results can be explainedof the d-band model by Hammer and Nørskov [46]. Uponn, bonding and anti-bonding states form between the

dsorption energies over the underlayer (Ir/M/Ir) alloys. Adsorption ener-tted relative to the respective position of the alloyed metal in the periodic

J. Courtois et al. / Applied Catalysis A: General 4

Fig. 8. CH Adsorption energies over the underlayer (Ir/M/Ir) alloys. Adsorption ener-gies are plotted relative to the respective position of the alloyed metal in the periodictable. The Ir–Mn alloy is not included, as discussed in the text due to stability prob-lems with CH adsorption.

Fig. 9. CO agies are plottable.

metal surenergies

level, andpensate fstates occhigher d-where thto our unwith latewould exis similar

Fig. 10. CHgies are plottable.

summaryinduces calloys wi

With tusing theexceptionelementsand O adbased ovand founobtainedCr, Mn, Fefrom thethe overltion enerthe lateron the subehave m(iridium)ligand efadsorptiotures, whenergies.

3.5. Activbond brea

We us[27] to caCH and Cculated thof compa[26] that

ious

rface wor

eV fos aret enece (thlmos

puregies

0.3 elts su

dsorption energies over the overlayer (M/Ir/Ir) alloys. Adsorption ener-ted relative to the respective position of the alloyed metal in the periodic

face d electrons and adsorbate electrons. Higher d-bandlead to more anti-bonding orbitals being above the Fermi

thus more anti-bonding states being unfilled. To com-or the unfilled anti-bonding states, more filling of bondingurs. Thus, more stable adsorption occurs for surfaces withband energies. Kitchin et al. [47] modeled Pt surfaces

e second layer was replaced by another metal, analogousderlayer structures. They did indeed observe that alloying

prevIr surent1.69faceexacchoi

Athanener0.1–resu

transition metals increased the d-band energy and thuspected to decrease (stabilize) adsorption energies, which

to what we observe with Ir-based underlayer alloys. In

adsorption energies over the overlayer (M/Ir/Ir) alloys. Adsorption ener-ted relative to the respective position of the alloyed metal in the periodic

facile C

tion. Themay chanFor severmay not

environmadsorbatemodynamcandidatePt. For exenergies

Our resuas a prim

3.6. Tow

In thisdifferent

potentiallysts. Thecandidate

83 (2014) 85–96 93

, our results indicate that the electronic or ligand effecthanges in overlayer structures that favor adsorption overth late transition metals.he overlayer structures, the adsorption energies obtained

different transition metals are mainly constant, with the of the metals from group 11 (Au, Ag, Cu). These later

typically bind CO weakly. Su et al. [8] investigated COsorption energies over surfaces having various platinum-erlayer structures (albeit with less surface substitution)d similar results in that the range of adsorption energies

was small (having a span of approximately only 0.1 eV for,Co, Ni, and Cu overlayer structures). In contrast to results

underlayer structures, late transition metal alloys inayer configuration demonstrated the least stable adsorp-gies, again reflecting the weak reactivity/adsorption of

transition metals, i.e. noble metals, which are exposedrface. In the case of the overlayer structures the alloysore like the dopant metal rather than the parent metal

. The exposed surface atoms experience both strain andfects but such affects give no discernable trend on then energies. This is in contrast to the underlayer struc-ere electronic effects have a clear effect on adsorption

ation energy of the ethanol oxidation reaction: C Cking

ed the transition state scaling relationship of Wang et al.lculate the activation energies for CHCO dissociation to

O over the various alloys, as displayed in Table 5. We cal-e activation energy over Pt (1 1 1) to be 1.89 eV. As a point

rison, our results compare favorably with previous workutilized a similar method, but with a four-layer slab. Thiswork calculated activation energies over Pt, Ir–Ru, ands to be 1.69, 1.05, and 1.31 eV, respectively. In the cur-

k we calculated activation energies to be 1.88, 1.42, andr the three surfaces. The trends between the three sur-

similar for both the current and previous work, but thergy differences can likely be attributed to the slab modelree versus four layers).t all underlayer structures have lower activation energies

Pt, while only a few overlayer structures have activationless than Pt. Typical activation energies are lowered byV, but may be lowered as much as 0.8 eV (Ir–Os). Theseggest that sub-layer modification of Ir may lead to moreC bond breaking, which is desirable for ethanol oxida-

issue is further complicated because the alloy structurege depending on the environment, as Section 3.3 shows.al alloys segregation occurs in the presence of CO whichoccur in an otherwise vacuum environment. The actualent will be much more complex with a large number ofs so it is difficult to predict which alloy structure is ther-ically preferred. Nonetheless, we have identified several

alloys that may oxidize ethanol with more efficiency thanample, Ir–Ru, Ir–Rh, and Ir–Os all have lower activationfor C C scission than Pt, regardless of the alloy structure.lts reinforce the possibility of using iridium-based alloysary component in EOR catalysts.

ard an economically viable catalyst

work we have determined the adsorption energies of the

reaction intermediates and the activation energies for a

rate-determining step of the EOR over various Ir–M cata-se values, or descriptors, allow the prediction of catalysts for the EOR. The notion of descriptors in combination

94 J. Courtois et al. / Applied Catalysis A: General 483 (2014) 85–96

Table 5Activation energies for the reaction step CH + CO → CHCO over different Ir/M bimetallic structures using the BEP correlation from Wang et al. [27]. Pure Pt has an activationenergy of 1.89 eV. Gray values correspond to higher activation energies than on a pure Pt. Energies are given in eV. Select values for Ir–Mn, Ir–Cr, Ir–Zr and Ir–Hf are notshown because of alloy stability problems, as discussed previously in the text.

Ti V Cr Mn Fe Co Ni Cu

Overlayer 3.76 3.66 – – 1.89 2.64 2.49 3.20Underlayer 1.68 1.48 2.20 – 1.68 1.88 1.83 1.73

Zr Nb Mo Tc Ru Rh Pd AgOverlayer – 3.20 2.83 2.54 1.42 1.73 2.65 4.44Underlayer 1.46 1.32 1.69 1.58 1.51 1.76 1.53 1.54

Os Ir Pt AuOverlayer 1.39 1.69 2.45 4.09Underlay 1.10 1.69 1.75 1.68

with modvalue of Doxidationlow CO adactivationcost of thcell’s costcell signifi

In Fig.with the avalues arunderlaythe alloyespheres ieters (COindependhave largweakly bbe more

to higherlarly, mordissociatilinear wi

1. Activation energies versus CO adsorption energy for the various studied in thmineds [52].

Fig. 12. TEMwere prepa

Hf Ta W Re

– 3.31 2.74 2.14er 1.43 1.25 1.73 1.67

eling results is described by Nørskov et al. [51] where theFT for screening catalysts is shown. A desirable ethanol

fuel cell catalyst should be resistant to CO poisoning (e.g.sorption energy) and readily break apart ethanol (e.g. low

energy for C C breaking). A third descriptor could be thee catalyst. The catalyst is a significant portion of the fuel

and cheaper catalysts may drive down the cost of the fuelcantly.

11, these three descriptors are gathered into one graphctivation energy plotted against the CO adsorption. These

e taken from the most preferred structure (overlayer orer determined by the segregation energy). The prices ofd transition metals [52] are represented by the size of the

n the plot. We note that the two surface chemistry param- adsorption energy and activation energy) are not strictlyent. A catalyst with low CO adsorption energy will likelye activation energy, since one of the final products (CO) isound to the surface and the reaction energy will thereforeendothermic. A more endothermic reaction energy leads

Fig. 1alloysdeterprice

activation energies in the BEP formalism (Eq. (4)). Simi-e stable CO binding lowers the activation energy for CHCOon. Because of this relationship, the data appears mostlyth positive slope.

Our caof −2.17sphere tosphere a

images of carbon-supported (a) Pt (ETEK), (b) Ir, (c) Ir80Rh20 and (d) Ir77Ru23 NPs alonred using surfactant-free method whereas Ir77Ru23 nanoparticles were made in the prese

eir preferred configuration (underlayer or overlayer configuration as by most stable structure). The sphere sizes are proportional to metal

lculations found that pure Pt has a CO adsorption energy eV and activation energy of 1.89 eV (indicated by the

the left), and any catalysts to the right and below thisre of interest. There are a number of potential catalysts

g with their size distribution analysis. The Ir and Ir80Rh20 nanoparticlesnce of surfactant PVP. See Section 2.2 for details.

J. Courtois et al. / Applied Catalysis A: General 483 (2014) 85–96 95

Fig. 13. (a) CV measurements (forward scan shown) of Pt/C (ETEK), Ir/C, Ir4Rh1/C along with Ir77Ru23/C sweeping from −0.1 V to 0.35 V vs. Ag/AgCl (1 M KCl) at a scan rate of30 mV/s in nts ofa 1 h reactio ents

with CO apromisinexperimeWe musteven be sdation aris needed

3.7. Expe

The colysts for enamely Isupports.as given

commercout the cIr77Ru23to be 3.2After heacarbon-suparticles

The caoxidationshows thH2SO4/0.All the Ir-current dpared to tby Ir77Ruand 0.13tively, whPt/C.

To evaalysts for0.5 M H20.3 V or 0fast but gdue to th1 h of realong termECASA—stimes higEOR was etial of 0.1of Ir, Ir–Rin EOR costill inferespecially

re wy thenol.

esultlysts.

oncl

idiumcandnol oce)

ture as idalcu

olera sciss

oxiciat

. Und in misin

rableng a

thanh theineder actlyst.ombr struationr, Tas also

R caHCOsed oterm

onditdeteoachlysts

ce dle str

0.5 M H2SO4 + 0.5 M ethanol electrolyte. (b) Chronoamperometric (CA) measuremen in the same electrolyte at room temperature. (c) Zoom-in image of CA measurem

dsorption energies between −2 and −1.5 eV that appearg: Ir–Ru, Ir–Re, Ir–Os, Ir–Hf, Ir–Ta, Ir–Nb, etc. Furtherntal work is needed to synthesize and test these catalysts.

also admit that some of the identified catalysts may nottable under reaction conditions; segregation and/or oxi-e real possibilities for some of the alloys and further work

to assess this possibility.

rimental analysis

mputational results indicate a number of viable cata-thanol oxidation, which we have tested experimentally,

r, Ir–Ru, Ir–Rh, and Pt (for comparison), all over carbon Characterization of the synthesized particles by TEM,in Fig. 12, shows that all the Ir-based nanoparticles andial Pt nanoparticles were evenly distributed through-arbon support. The average sizes of Pt, Ir, Ir4Rh1 andafter analyzing at least 120 particles were calculated

± 0.6, 2.6 ± 0.6, 2.6 ± 0.6, and 3.7 ± 0.8 nm, respectively.t treatment at 300 ◦C under open-air environment, all thepported catalysts remained in the form of nano-sized

without agglomeration.talytic activities of these Ir-based particles for the ethanol

reaction were evaluated using CV and CA. Fig. 13ae CVs (forward scan) of the electrocatalysts in 0.5 M

5 M ethanol electrolyte from −0.1 to 0.35 V vs. Ag/AgCl.based catalysts presented much enhanced ECSA-averagedensities throughout the scanned potential range com-he Pt/C electrocatalyst. In particular, the best performance23 showed ECASA-specific current densities of 0.112

1 mA/cm2 at 0.2 V and 0.3 V vs. Ag/AgCl in CV, respec-ich are 180% and 70% higher than those of commercial

luate the long-term catalytic activity of the Ir-based cat- the EOR, CA measurements were conducted in the sameSO4/0.5 M ethanol electrolyte at a constant potential of.2 V for 1 h (Fig. 13b). The current density initially droppedradually reached a quasi-steady state value. This could bee poisoning of active sites on the catalyst surface. Afterction, all the Ir-based electrocatalysts exhibited superior

activity compared to Pt/C. The Ir77Ru23 and Ir showed anpecific current density of ∼10.5 �A/cm2, which is about 10her than that of Pt/C. The superiority of Ir77Ru23 for theven under-estimated as the CA was conducted at a poten-

V lower than that of Pt/C. Although the electroactivities

Futuciallethatal rcata

4. C

Irtial

ethasurfastrucusedory cCO tC Cpletedissostepancepromfavohaviture(witobtabettcata

ClayeoxidHf–Iturean EOfor Cfocuto detal cand

apprcatasurfastab

u and Ir–Rh nanocatalysts showed dramatic enhancementmpared to commercial Pt catalysts, their performance isior to state-of-art Pt-based binary and ternary catalysts,

PtRh–SnO2 [53,54] and PtIrSn [55] reported recently.

less identfor directconfirm toxidation

above electrocatalysts at applied constant potential of 0.3 V or 0.2 V forshown in (b).

ork will evaluate other Ir-based electrocatalysts, espe- activity of Ir-based catalysts for splitting C C bonds ofOur results herein are encouraging in that experimen-s confirm the superiority of a select group of Ir-based

usions

-based bimetallic structures were investigated as poten-idates for replacing platinum as anode catalysts for thexidation reaction. Overlayer (alloyed metal directly on

and underlayer (alloyed metal in subsurface position)s made of a transition metal associated with iridium wereealized models of alloy surfaces. Density functional the-

lations were used to estimate two important descriptors:nce (e.g. CO adsorption energy) and activation energy forion. Breaking of strong C C bonds is necessary for com-dation of ethanol. Based on previous work, we modeledion of CHCO, a strong candidate for the rate-determiningerlayer structures demonstrated an enhanced CO toler-ost cases with early transition metals being the most

g alloys. CHCO bond breaking also appeared to be more on underlayer surfaces than on overlayer surfaces, always

lower CHCO activation energy on an underlayer struc- on the overlayer structure for a given transition metal

exception of Ru and Rh where comparable energies were). Moreover, underlayer structures generally exhibitedivation energy for CHCO dissociation than a pure platinum

ined with their good CO tolerance, iridium-based under-cture catalysts show excellent promise for the ethanol

reaction. We identified several alloys (Ru–Ir, Rh–Ir, Nb–Ir,–Ir, Zr–Ir, V–Ir) as promising catalysts. Overlayer struc-

demonstrated good CO tolerance but their potential astalyst was reduced by the relatively high activation energy

dissociation these alloys demonstrated. Our efforts haven screening potential catalysts, and further work is neededine whether such catalysts are stable under experimen-ions, evaluate full electrochemical environmental effectsrmine full reaction pathways. Another limitation of our

is that an ideal alloy model was used, and indeed severalthat we studied (i.e. Ir–Mn and Ir–Cr) demonstrated severeistortions upon adsorption, indicating that other, moreuctures are possible for these alloys. We have nonethe-

ified several potential alloys that warrant more attention

ethanol fuel cells and our experimental work does indeedhat Ir alloys (e.g. Ir–Rh and Ir–Ru) may be superior ethanol

catalysts to Pt.

96 eral 4

Acknowl

We wCBET-115this work

Referenc

[1] F. Vigie111–1

[2] S.a. Ki365–3

[3] F. Vigi34 (20

[4] W. DuTeng, J

[5] E. Anto[6] W. Du

(2012)[7] F. Cell,[8] H.-Y. S[9] J.-M. L

(2005)[10] W. Zho[11] J. Gree[12] S. Alay

333–3[13] A.U. N

(2010)[14] M. Li, W[15] R. Alca[16] R. Alca

Cataly[17] H. Xin

376–3[18] P. Ferr

Mavrik[19] L. Colm

Acta 5[20] L. Jiang

(2005)[21] C. Lam

(2004)[22] E. Chri

123–1[23] G. Cam

.M. S7901.L. Ca. Du

. Waneders2010). VandompuP2K

/10/2.P. Per. Lipp

. Vand. Goe. Kra

. Rub5990.P. K

. Abilon, T..C. Fo. Sanhem.

.E. Su

.M. Ci11 (1.E. Ra. Ket.E. Rahem.. Ham

.R. Ki0240. Rubhem.. Ma

.G. Ch

.K. Nø (200etalP

. Kow. Liu, A.I. Frenkel, R.R. Adzic, Nat. Mater. 8 (2009) 325–330.

J. Courtois et al. / Applied Catalysis A: Gen

edgment

ish to thank the National Science Foundation (Grant no.9662) (Award number 1159662) for financial support in.

es

r, S. Rousseau, C. Coutanceau, J.-M. Leger, C. Lamy, Top. Catal. 40 (2006)21.rillov, P.E.S Tsiakaras, I.V. Romanova, J. Mol. Struct. 651–653 (2003)70.er, C. Coutanceau, A. Perrard, E.M. Belgsir, C. Lamy, J. Appl. Electrochem.04) 439–446., Q. Wang, D. Saxner, N.A. Deskins, D. Su, J.E. Krzanowski, A.I. Frenkel, X.. Am. Chem. Soc. 133 (2011) 15172–15183.lini, J. Power Sources 170 (2007) 1–12.

, K.E. Mackenzie, D.F. Milano, N.A. Deskins, D. Su, X. Teng, ACS Catal. 2 287–297.

J. Mann, N. Yao, A.B. Bocarsly, Langmuir 22 (2006) 10432–10436.u, X. Gu, X. Ma, Y.-H. Zhao, X. Bao, W.-X. Li, Catal. Today 165 (2011) 89–95.éger, S. Rousseau, C. Coutanceau, F. Hahn, C. Lamy, Electrochim. Acta 50

5118–5125.u, Z. Zhou, S. Song, W. Li, Appl. Catal., B: Environ. 46 (2003) 273–285.

ley, M. Mavrikakis, Nat. Mater. 3 (2004) 810–815.oglu, A.U. Nilekar, M. Mavrikakis, B. Eichhorn, Nat. Mater. 7 (2008)

38.ilekar, S. Alayoglu, B. Eichhorn, M. Mavrikakis, J. Am. Chem. Soc. 132

7418–7428.. Guo, R. Jiang, L. Zhao, H. Shan, Langmuir 26 (2010) 1879–1888.

la, M. Mavrikakis, J.a. Dumesic, J. Catal. 218 (2003) 178–190.la, J.W. Shabaker, G.W. Huber, M.A. Sanchez-castillo, J.A. Dumesic, P.

sts, J. Phys. Chem. B 109 (2005) 2074–2085., A. Holewinski, N. Schweitzer, E. Nikolla, S. Linic, Top. Catal. 55 (2012)90.in, D. Simonetti, S. Kandoi, E. Kunkes, J.a. Dumesic, J.K. Nørskov, M.akis, J. Am. Chem. Soc. 131 (2009) 5809–5815.enares, H. Wang, Z. Jusys, L. Jiang, S. Yan, G.Q. Sun, R.J. Behm, Electrochim.

2 (2006) 221–233., G. Sun, S. Sun, J. Liu, S. Tang, H. Li, B. Zhou, Q. Xin, Electrochim. Acta 50

5384–5389.

[24] V1

[25] B[26] W[27] S

P(

[28] JC

[29] C7

[30] J[31] G[32] J[33] S[34] M[35] A

1[36] W[37] F

s[38] D[39] G

C[41] J[42] I

3[43] G[44] M[45] G

C[46] B[47] J

1[48] A

C[49] M[50] J[51] J

1[52] M[53] A

P

y, S. Rousseau, E.M. Belgsir, C. Coutanceau, J. Léger, Electrochim. Acta 49 3901–3908.stoffersen, P. Liu, A. Ruban, H.L. Skriver, J.K. NoD/rskov, J. Catal. 199 (2001)31.ara, R. de Lima, T. Iwasita, Electrochem. Commun. 6 (2004) 812–815.

[54] W. DuJ. Mate

[55] M. Li,

Soc. 13

83 (2014) 85–96

chmidt, R. Ianniello, E. Pastor, S. González, J. Phys. Chem. 100 (1996)–17908.o, G. Sun, H. Li, Q. Xin, Fuel Cells Bull. 2007 (2007) 12–16., N.a. Deskins, D. Su, X. Teng, ACS Catal. 2 (2012) 1226–1231.g, B. Temel, J. Shen, G. Jones, L.C. Grabow, F. Studt, T. Bligaard, F. Abild-en, C.H. Christensen, J.K. Nørskov, F.A. Claus, H.C. Jens, Catal. Lett. 141

370–373.eVondele, M. Krack, F. Mohamed, M. Parrinello, T. Chassaing, J. Hutter,t. Phys. Commun. 167 (2005) 103–128.Open Source Molecular Dynamics, http://www.cp2k.org, Accessed014 (2014).dew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996) 3865–3868.ert, J. Hutter, M. Parrinello, Mol. Phys. 92 (1997) 477–487.evondele, J. Hutter, J. Chem. Phys. 127 (2007) 114105.

decker, M. Teter, J. Hutter, Phys. Rev. B: Condens. Matter 54 (1996) 1703.ck, Theor. Chem. Acc. 114 (2005) 145–152.an, H. Skriver, J. Nørskov, Phys. Rev. B: Condens. Matter 59 (1999)

–16000.rekelberg, J. Greeley, M. Mavrikakis, J. Phys. Chem. B 108 (2004) 987–994.d-Pedersen, J. Greeley, F. Studt, J. Rossmeisl, T. Munter, P. Moses, E. Skúla-

Bligaard, J. Nørskov, Phys. Rev. Lett. 99 (2007) 4–7.rd, Y. Xu, M. Mavrikakis, Surf. Sci. 587 (2005) 159–174.tarossa, A. Vargas, M. Iannuzzi, C.A. Pignedoli, D. Passerone, A. Baiker, J.

Phys. 129 (2008) 234703.tton, D.G. Vlachos, ACS Catal. 2 (2012) 1624–1634.obıca, F. Frechard, R.A. van Santen, A.W. Kleyn, J. Hafner, Chem. Phys. Lett.999) 185–192.mirez-Caballero, P.B. Balbuena, Chem. Phys. Lett. 456 (2008) 64–67.tner, W.B. Schneider, A.a. Auer, J. Phys. Chem. C 116 (2012) 15432–15438.mı, Y. Ma, R. Callejas-Tovar, P.B. Balbuena, G.E. Ramírez-Caballero, Phys.

Chem. Phys. 12 (2010) 2209–2218.mer, J.K. Nørskov, Adv. Catal. 45 (2000) 71–129.

tchin, J.K. Nørskov, M.a. Barteau, J.G. Chen, J. Chem. Phys. 120 (2004)–10246.an, B. Hammer, P. Stoltze, H.L. Skriver, J.K. NoD/rskov, J. Mol. Catal. A:

115 (1997) 421–429.vrikakis, B. Hammer, J.K. Nørskov, Phys. Rev. Lett. 81 (1998) 2819–2822.en, C.A. Menning, M.B. Zellner, Surf. Sci. Rep. 63 (2008) 201–254.rskov, T. Bligaard, J. Rossmeisl, C.H. Christensen, A.M. Design, Nat. Chem.9) 37–46.rices.com, http://www.metalprices.com, Accessed 11/20/2012 (2012).al, M. Li, M. Shao, K. Sasaki, M.B. Vukmirovic, J. Zhang, N.S. Marinkovic,

, Q. Wang, C.a. LaScala, L. Zhang, D. Su, A.I. Frenkel, V.K. Mathur, X. Teng,r. Chem. 21 (2011) 8887.

D.a. Cullen, K. Sasaki, N.S. Marinkovic, K. More, R.R. Adzic, J. Am. Chem.5 (2013) 132–141.