Evaluation of Ba deficient NdBaCo2O5+δ oxide as cathode material for IT-SOFC

Evaluation

of

Ba

de

ficient

NdBaCo

2

O

5+

d

oxide

as

cathode

material

for

IT-SOFC

A.

Donazzi

a,

*,

R.

Pelosato

b,

**,

G.

Cordaro

b

,

D.

Stucchi

b

,

C.

Cristiani

b

,

G.

Dotelli

b

,

I.

Natali

Sora

c

a

DipartimentodiEnergia,PolitecnicodiMilano,ViaLambruschini4,20156Milano,Italy

b

DipartimentodiChimicaMaterialieIngegneriaChimica“GiulioNatta”,PolitecnicodiMilano,PiazzaLeonardodaVinci32,20133Milano,Italy

c

INSTMR.U.andUniversitàdiBergamo,DipartimentodiIngegneriaeScienzeApplicate,VialeMarconi5,24044Dalmine,BG,Italy

Received8June2015

Receivedinrevisedform2September2015

Accepted20September2015

Availableonline28September2015

1.Introduction

The quest for electrode materials capableof attaining good electrochemicalperformancesinIntermediateTemperatureSolid Oxide Fuel Cells (IT-SOFCs) is a key research topic. Recently, attention has been paid to a family of cobalt-based layered perovskitecompoundsthatcanbegroupedinthegeneralformula AA0_B

2O5+d,whereAisarareearth,A0analkalineearth,commonly

Ba,whiletheBsiteismostoftenoccupiedbyCoions.Manyrelated compoundswithdifferentsubstitutionsonA0andBsites,orboth, havebeeninvestigated[1,2].Theessentialcharacteristicofthese compounds is the ordering of the A-sitecations in alternating layers,A-O,B-O,A0-O,B-Oalongthec-axisoftheperovskitelattice

[3–5].Severaleffortshavebeendevotedtostudytheeffectsthat thesubstitutionshaveonthecrystalstructure,theoxygencontent, thethermalexpansion,theoxygensurfaceexchange,theelectrical

conductivityandtheelectrochemicalactivity.Veryrecently,itwas found that the defect chemistry of these compounds can be tailored by introducing small changes in the stoichiometry of eitherAorA0site:inthislattercase,partiallysubstitutingtheA0 -site Ba cations withSris generally understoodtoenhance the electrical conductivity[6–8].Doping theB-site Cocationswith transitionmetals,especiallyFe,Cu,NiorMn,isadoptedtolimitthe thermalexpansion[6,9–11];though,highlevelsofsubstitutionof theCocationcandramaticallyaffecttheelectrochemicalactivity, evenconferringanodicpropertiesforuseinsymmetricSOFCs[12]. A0 _{siteunder-stoichiometrywasfirst investigated:Badeficiency}

(x=0
0.08)inPrBa1-xCo2O5+dcausedashrinkageofthecrystal

cellsize,slightlyaffectedthethermalexpansioncoefficient,and affectedboththeelectricalconductivityandtheelectrochemical activity[13–15].AsimilarstudywasperformedonLaBa1-xCo2O5+d:

this compound tolerated a large Ba under-stoichiometry (x=0.0
0.15) without significant modification of the lattice parameters.Amonotonicdecreasebothintheelectrical conduc-tivityandintheASRwasfoundbyincreasingthedeficiencyupto 10%[16].Theresultsoftheseinvestigationscouldberationalized based on the dominant charge compensation mechanism. Introducingdifferentamountsofdeficiencygeneratesmetalion

* Correspondingauthor.Tel.:+39223998651;fax:+3922399

3318. **Correspondingauthor.Tel.:+39223993232;fax:+3927063

8173. mailaddresses:[email protected](A. Donazzi),

vacancies ðv00

BaÞ in the structure; this loss of charge must be

compensated,bygenerationofeithernewelectronicdefectsðCo

CoÞ

orionicdefectsðv

OÞ.Giventhatthetwomentioneddefectsarethe

chargecarriersoftheelectronicand theionicconductivity,the mechanism of charge compensation affects both the electronic conductivityandtheionicconductivityofthe under-stoichiomet-riccompounds.Therefore,itisinprinciplepossibletoachievea highcathodicperformancebyoptimizingtheamountofdefectsin thestructure of thematerial, thus minimizing thepolarization losseswithouthinderingtheelectronicconductivity.Amongthe cobaltbasedlayeredperovskiteoxides,NdBaCo2O5+dappearstobe

muchlessinvestigatedthanotheroxides,eventhoughitpossesses remarkableoxygensurfaceexchangeproperties[3,8,17–20].Inthis paper,wefocusontheeffectsthatacationde_ficiencyonA0_-sitein

the range 0.0x0.2 has on the physico-chemical and the electrochemicalpropertiesofNdBa1-xCo2O5+d.Foreachcompound,

thepowderswerechemicallycharacterizedviaX-RayDiffraction, SEM,TG-DTAanalysesandcerimetrictitration.The electrochemi-calproperties wereevaluated bymeasuringthe total electrical conductivity, via 4-probe method on sintered bars, and the polarizationbehavior,viaElectrochemicalImpedance Spectrosco-py (EIS) on symmetric cell configuration. For the first time, a mechanisticinsightontheeffectof Badeficiencyin theoxygen reductionactivityofthesecathodematerialsisoffered,basedon thequantitativeanalysisoftheEISspectra.

2. MaterialsandMethods

NdBa1-xCo2O5+dsampleswithx=0,0.05,0.10and0.20(named

after the Ba de_ficiency NBC0, NBC5, NBC10 and NBC20) were preparedviasolidstatemixingandfiringmethod.Nd(NO3)3



6H2O, Ba(NO3)2,Co(OCOCH3)2



xH2Osaltswereusedasprecursors.The Nd and Co precursors were converted into the corresponding oxides(Nd2O3andCo3O4)byproperthermaltreatment.Ba(NO3)2 was heated up at 280C to remove any absorbed water. The precursorswereball-milledinisopropylalcoholfor24h,firedat 800_{Cfor12hinair,milledagain24hinisopropyl alcoholand}

finally fired twiceat 1100C for 12h in air with intermediate regrinding.Thecoolingrampsofthehightemperaturetreatments werekeptat1_{C/mintoallowfortheequilibrationoftheoxygen}

contentwiththesurroundingatmosphere.Thephasecomposition ofthepowderswas checkedbyX-RayDiffraction(XRD)usinga BrukerD8AdvanceDiffractometer,withgraphitemonochromated Cu-K

_a

radiation.The diffractionpatterns werecollectedin the

range10–80₂

_u

_{withastepof0.02}₂

_u

_{andacountingtimeof1s}

perstepforthephaseanalysesand12sperstepforthestructural analyses.TheXRDdatawereanalyzedusingroutinesavailablein the GSAS package, including the Rietveld method [21]. XRD analyseswerecarriedoutalsoonmechanicalmixtures(50/50wt%) oftheNBCsampleswithGDC(Ce0.9Gd0.1O2,SigmaAldrich)and LSGM(La0.8Sr0.2Ga0.8Mg0.2O3,FuelCellMaterials),toevaluatethe occurrenceofchemicalinteractionsbetweentheelectrolyteand thecathodematerials,afterfiringupto1100CinthecaseofGDC and1000CinthecaseofLSGM.

ThermogravimetricAnalyses(TG-DTA)wereperformedonca. 40mg of the NBC samples by a simultaneous TG-DTG Seiko 6300 instrument. In a TG-DTA experiment, the powders were heatedinairfromroomtemperatureto850Cat3C/min,and thencooledagaintoroomtemperaturewiththesamerate.

Theoxidationstateofcobaltandtheoxygencontentofallthe compounds were estimated by cerimetric redox titrations: ca. 50mgofpowderedsamplewasdissolved ina1MHCl solution (about50ml)containingaknownexcessofFe2+_(FeCl

2



4H2O)and heldafewminutesforthefollowingreactiontooccur:

Conþ_{þðn 2ÞFe}2þ_!Co2þ_{þðn 2ÞFe}3þ _ð1Þ

After the reaction, the remaining Fe2+ _{was titrated with a} solutionofCe(SO4)2in1MH2SO4usingferroinasindicator;the end-pointwasevaluatedvisuallybythesuddencolorchangefrom orange togreen. The concentrationof Ce(SO4)2 solution(about 0.01M) was standardized against a sodium oxalate (Na2C2O4) solution at 80C, and was itself used to determine the exact concentrationofthestartingFeCl2



4H2Osolution(about0.01M). Foreachdetermination,threeexperimentswerecarriedout.

Theelectrical conductivityand thepolarizationresistanceof thesamplesweremeasuredasafunctionoftemperaturewitha

Fig.1.XRDpatternsforthethreeNdBa1-xCo2O5+dsamples.NdCoO3(PDF

#00-025-1064)ismarkedwith*.

Table1

RefinedstructuralparametersoftheNdBa1-xCo2O5+dsamples.

x=0 x=0.05 x=0.10 Spacegroup Pmmm Pmmm P4/mmm a(Å) 3.89921(9) 3.8987(1) 3.89858(4) b(Å) 7.8107(2) 7.8043(2) – c(Å) 7.6125(1) 7.6110(1) 7.6113(1) V(Å3 ) 231.840(8) 231.576(8) 115.683(3) occupancyBa – – 0.922(3) rtheor(g/cm3) 7.056 6.963 6.865 x2 1.323 1.382 1.458 Rwp(%) 19.19 20.35 20.80 Rp(%) 11.91 13.04 12.94 RB(%) 5.53 7.05 7.73

potentiostat/galvanostat(AMEL7050)equippedwithaFrequency ResponseAnalyzer (FRA). The electrical conductivity was mea-suredviaa four-electrodeDC methodonsintered bars(25mm long,5mmwide, 3mmthick,calcinedin airat 1100Cfor 4h) between 25 and 850C. The calcination temperature was purposely kept at 1100C, in order for the samples to be representativeof the cellelectrodes and toavoid volatilization andlossofCo.Though,atemperatureof1100Cwasnotsufficient toachievecompletesinterizationofthebarsandresidualporosity waspresent.Theporosityofthebarswasestimatedbycomparing thedensityvaluesmeasuredwiththebuoyancybalanceinethanol andthevaluesofthetheoreticaldensity(

_r

theor)calculatedfrom thelatticeparameters.Onaverage,35%porosityresultedforallthe samples.

EIStestswereperformedusingasymmetriccellconfiguration withGDC asthe electrolyte.TheGDC pellets (1.1cm diameter) werefabricatedviadie-pressingandcalcination(1400C,12h,air). The sinterization of each pellet was verified bycomparing the densityvaluemeasuredwiththebuoyancybalance(7.101g/cm3₎ withthetheoreticaldensity(7.159g/cm3_{).99%ofthetheoretical}

density was achieved.Aslurryof thecathodematerial (60wt% solid content) was prepared from the powders by adding

a

-terpineol,isopropylalcoholandethylcellulose(76:20:4relative weights)andstirringtheresultingmixturefor2h.Theslurrywas thenappliedoneachsideofthepellet,driedat110Cfor12hinair, andcalcinedat1100Ctoreachadhesion.Tocollectthecurrent,Ag mesheswereappliedwithsilverpaintonbothsidesofeachpellet. TheEIStestswereperformedbetween550and750C,atOpen Circuit Voltage(OCV)with10mVvoltageamplitudeand inthe frequency range0.1Hz–10kHz. Atall thetemperatures,theEIS testswerecarriedoutunderairflow(50Ncc/min)andatdifferent O2partialpressure(5%,10%and100%v/v).Experimentswerealso carriedoutbyusingHeinsteadofN2astheO2diluentat21%and 5% O2 to verify the impact of mass transfer limitations. The EquivalentCircuitModel(ECM)methodwasappliedwithZviewto evaluatethecontributionsofthephysical,chemicaland electro-chemical stepsoftheOxygenReductionReaction(ORR) mecha-nism.

Themorphologicalfeaturesofthecalcinedpowdersandofthe electrodes of the symmetric cells were assessed via Scanning

Fig.3.SEMmicrographs.PanelsAandB:NBC10(a)andNBC20(b)powdersaftercalcinationat1100_{C.InsertsareEDSspectra.PanelC:crosssectionoftheNBC10/GDC/}

ElectronMicroscopy(SEM)usingaCarlZeissEVO50VPinstrument equipped with an Energy Dispersive Spectrometer (EDS) for elementalanalysis.

3. ResultsandDiscussion

3.1.XRDcharacterizationoftheNBCpowders

TheXRDpatternsofthesamplesobtainedaftercalcinationat 1100C are shown in Fig.1.The pure NdBa1-xCo2O5+dphase is

obtainedatx=0and0.05(NBC0andNBC5),whileatraceofan impurity phase is detected in thex=0.10 sample (NBC10).The same phase is found in larger amount in the x=0.20 sample (NBC20).ThisimpuritycanbeidentifiedasNdCoO3(PDF #00-025-1064).TheformationofasecondaryphasecontainingNdandCoin the 10% deficient compound suggests that the maximum Ba deficiencyacceptedbyNdBa1-xCo2O5+disslightlylowerthan10%:

oncethisthresholdisovercome,theBa-freephaseemerges.For thisreason,thestructuralpropertieswerestudiedonlyforNBC0, NBC5andNBC10materials,andtheNBC20wasexcluded.Allthe compoundscrystallizeinanA-siteorderedperovskitestructure. ThepatternofsampleNBC0canbeindexedusingatetragonalunit cellhavinglatticeparametersa=b=ap,c=2ap(apbeingthelattice parameterofthecubicperovskite).However,Rietveldrefinements usingthetetragonalunitcell(P4/mmmspacegroup)givesomehow unsatisfactory goodness-of-fit parameters, i.e.

_x

2_{=1.516, R}

wp= 20.56%, Rp=13.44% and RB=7.24%. Closer examination of the diffractionprofile evidence thebroadening of somereflections, especiallythatatca.1.98Å,which suggeststoindexthepattern usinganorthorhombicunitcell.Moreover,thepresenceofavery weakreflection at ca. 7.8Å indicatesdoubling of the b lattice parameterduetoorderingofoxygenvacancies.Rietveldre fine-ments performed using the orthorhombic doubled unit cell ap2ap2ap(Pmmm space group) and the crystal parameters

of NdBaCo2O5.69 [22] give better results, namely

x

2=1.323, Rwp=19.19%, Rp=11.91% and RB=5.53% (Table 1). The Rietveld refinementswereperformed using thetotal oxygen occupancy obtained by the cerimetric titration (see Section 3.2). The structuralparametersofNBC5aresimilartothoseoforthorhombic NBC0, although with a smaller orthorhombic distortion. The patternofNBC10showneitheraclearpeakbroadeningat1.98Å, norsuperstructurereflection.Refinementsareperformedusinga modelonthebasisofatetragonalapap2apunitcell(P4/mmm space group). The final refined cellparameters are reportedin

Table1.AsshowninFig.2,theBadeficiencycausesashrinkageof

thestructure, mainlydue totheb parameter,and toa smaller extenttotheaparameter.Thecparameterismuchlessaffected. For theNBC10 sample,the refined occupancyof Bagave value 0.922(3),correspondingtoathresholdofca.8%.Thisresultisin agreementwiththevaluefoundbyPangetal.[15]insimilarPrBa 1-xCo2O5+dcompounds.ThesimilaritybetweentheionsizesofNd

andPrseemstovalidateavaluearound8%forthetoleratedBa deficiency in NdBa1-xCo2O5+d. Despite the similarity of the ion

sizes, however, the transition observed in NBC from the orthorhombicstructuretothetetragonaloneisnotreportedfor thePrBa1-xCo2O5+dcathodes,whosestructuremaintainsunaltered

atvaryingtheamountofBa(orthorhombicin[15]andtetragonal in[13]).Thetransitionisassociatedtoasignificantchangeinthe concentration of oxygen vacancies: a lower amount of oxygen vacanciesgivesrisetoanorthorhombicstructure,whileahigher amountresultsinatetragonalstructure.Therefore,theresultsof the XRD analysis suggest that the concentration of oxygen vacanciesprogressivelyincreasespassingfromNBC0toNBC10.

Thepowdersofthefourcompoundswerealsoanalyzedwitha scanning electron microscope toaccess the morphology. Good homogeneityisfoundforallthesamplesbothincompositionand size (Fig. 3a, the NBC10 sample is taken as a representative example):althoughtheaggregatesareslightlyirregularinshape,

Fig.4.XRDpatternsofNBC/LSGM50/50wt%mixture.LaBaGa3O7(*).NBC10(a);

the grains have similar size,in the range of few micrometers. Overall, after calcination at 1100C, the powders show a continuous structure with diffuse open porosity. Additionally, the EDS spectra are also reasonably in linewith the expected compositions,namely51.2%Co,25.3%Ndand23.5%Baonmolar basis for NBC10, estimated by neglectingthe O signaland the signals ofAu and Al,these latterbeing respectivelydue tothe platingandtotheholder.OnlyinthecaseoftheNBC20sample (Fig.3b),particles ofa Co-richphase withtracesof NdandBa (98.5%Co,1%Ndand 0.5%Baonmolarbasis) canbedetected, whicharenotobservedintheXRDandwhichconfirmthatatoo highBadeficiencypromotestheseparationofsecondaryphases. The pieces of evidence collected by the XRD and the SEM analysesallowedtodiscardtheNBC20sampleasavalidcathode. Hence,thecompatibilitybetweenthecathodesandtheelectrolyte materials was evaluated exclusively on the NBC0, NBC5 and NBC10 samples. The occurrence of chemical interactions with eitherGDCorLSGMwasassessedbypreparing50/50wt%mixtures oftheNBC powdersand theelectrolytepowders.Themixtures werefiredatincreasingtemperaturesandtherecoveredpowders wereanalyzedviaXRD.AlltheNBC/LSGMmixturesfiredat1000C for5h(Fig.4)revealtheformationofnewphases:newpeaksare detectedintheXRDpatterns,whichcanbeassignedtoBaNdGa3O7 (PDF #00-024-0110) and/or BaLaGa3O7 (PDF #00-024-0107). Traces of the same reaction products are found also at lower temperature,900Cfor12h,confirmingthefastchemicalreaction betweenLSGM and theNBCcompounds. A previousreport[3]

showednoreactivitybetweenNdBaCo2O5+dand LSGM:though,

the thermal treatments adopted in the reference were much shorterthanthepresentones(1100Cfor0.5hand1000Cfor3h), likelysuggestingthatlongertimeswouldhaveeventuallyrevealed some reactivity. In the present case, temperatures lower than

900C were not investigated as they were not suitable for obtaining good adhesion of the cathode material onto the electrolyte.These observationssuggesttoavoidtheuseof NBC compounds withLSGMelectrolyte withouttheuseof aproper barrierlayer.

TheXRDpatternsoftheNBC/GDCmixtures,treatedunderthe sameconditionsastheLSGM(12hat900_Cand5hat1000_C),

show neither the appearance of new peaks, nor any visible alterationinthepositionofthepeaks.Moreover,themixturesfired at1000Cwerefurtherheatedat1100Cfor5hours:alsointhis case,novisiblechangesarepresentintheXRDpatterns(Fig.5). Theseresultsclearlyshowthataverygoodchemicalcompatibility exists between the prepared NBC compounds and the GDC electrolyte. The GDC electrolyte was therefore chosen for the electrochemicalcharacterization,inordertoavoidtheformation ofsecondaryphasesduringthemeasurements.Overall,theXRD spectra of both the LSGM- and the GDC-based mechanical mixturesalsoindicatethattheBadeficiencydoesnotaffectthe interactionswiththetwoelectrolytesinameasurableextent.

Fig.6.PanelA:TGcurvesoftheNBCsamples.Operatingconditions:airflow,Tramp=3C/min,TMAX=850C.PanelB:variationintheoxygencontentwithtemperatureunder

airflow.

Table2

MeanoxidationstateofcobaltCon+_,Co3+_andCo4+_{ionfraction,oxygencontent(5+}

d)

andoxygenvacancyv

O in NdBa1-xCo2O5+d. Sample Con+ Co3+ Co4+ 5+d vO NBC0 3.14 1.72 0.28 5.64 0.36 NBC5 3.17 1.66 0.34 5.62 0.38 NBC10 3.20 1.60 0.40 5.60 0.40

Fig.7.TotalelectricalconductivitymeasurementsonNBCsamplesasafunctionof

3.2.Measurementofconductivityandoxygen-exchangingproperties The results of the cerimetric titration are summarized in

Table2.ThetitrationallowstodirectlymeasuretheamountsofCo3

+

and Co4+ ions present in each NdBa1-xCo2O5+d sample, and

thereforetheaverageoxidationstateoftheCoisdetermined.The oxygencontent is calculated fromthe stoichiometry, giventhe oxidation states of Ndand Ba (Nd3+_{and Ba}2+_{).The content of} oxygen vacanciesis estimated as the difference of the oxygen contentfromthestoichiometricvalueof6.UpondecreasingtheBa content,theaverageoxidationstateofcobaltandthenumberof oxygen vacancies both increase, while the oxygen content decreases. A high number of vacancies is beneficial for the electrochemicalactivityof thematerial,which isrelated tothe capability of introducing, transporting and desorbing oxygen insideandoutsidethestructure;aswell,thefractionofCo4+_ions mustbesufficientlyhigh,beingtheinclusionandtransportation mechanismassociatedtotheredoxcoupleCo4+/Co3+.Theincrease intheBadeficiencyincreasesboththeseparameters,leadingtoa favorableandsynergisticeffect.Itisthusinterestingtoexplorethe relationshipoftheBacontentwiththecapabilityofthematerialto exchangeoxygen.

This capability was first investigated via thermogravimetric analysisinair,upto850C.Foreachsample,themeasuredweight change(Fig.6a)is directlyrelatedtothechangeintheoxygen contentandtotheoxidationstateofcobalt(Fig.6b).Atincreasing temperature,alltheNBCformulationsexperiencefirstamoderate lossof oxygen between50 and 250C, which is followedby a steady,lineardecreaseupto850C.Apeakofoxygengain,and weightgain, was always detected around 275C: ananalogous observationwasreportedintheTGanalysesofNBCsamplesby Kimand Irvine[17].Afterreaching ofstableweightconditions, furthercyclingrevealedfullyreproduciblecharacteristicsforall the three samples. Upon decreasing the oxygen content, the oxidation state of cobalt decreases, as a consequence of the reductionoftheCo4+_ionstoCo3+_{.Onlyattemperatureshigher} than 800–850_{C, well beyond the limit for IT-SOFCs, further}

reductionoftheCo3+_toCo2+_{ionscausestheoxidationstatetofall} belowtheaveragestateof3,indicatingthecompletedepletionof Co4+_{ionsandtheactivationoftheCo}3+_/Co2+_{redoxcouple.Inthe} intermediatetemperaturebranch(300–500_{C),theNBC10sample}

experiences the sharpest loss of oxygen, followed by that of NBC5 and thatof NBC0. In thehightemperaturebranch(500– 800C), the NBC10 sample maintains the highest loss rate, followedbythatofNBC0,whilethatofNBC5 moderates.Inthe wholetemperaturerange,theNBC10sample loses 0.25 oxygen atoms per unit formula, 0.18 the NBC5 sample and 0.16 the NBC0and0.18.Consistently,upto700C,theNBC10sampleloses thehighestweightfraction,followedbyNBC5andNBC0.These results clearly indicate that the increase in the Ba deficiency progressively increases alsothe capability of desorbing oxygen fromthelattice.

Thedesorptionoflatticeoxygenaffectsthestructuralandthe electricalpropertiesofthematerial.Theelectricalbehaviorofthe samples was then investigated in terms of total electrical conductivity as a function of temperature, under air flow conditions.Theresultsofthetotalelectricalconductivity measure-mentsbetween50and 850CarereportedinFig.7.Thevalues reported are referred to the dense structure of the bar by normalizingthemeasuredvaluesofconductivity

_s

mbythesolid fractionofthebar(1

e

void),accordingtotheformula[23]:

s

¼

s

m

1

e

void ð2Þ

Theequationcanbeappliedonlyincasetheinter-particlenecks are well-developed and the porosity of the sintered bar is sufficiently low (

e

void<40%): as a matter of fact, Virkar and coauthorsstudiedtheeffectofthemorphologyontheconduction though porous Samaria doped Ceria (a MIEC material) and validated this equation by measuring the conductivity of bars with a porosity of 33.5% [24]. Considering that NBC is a MIEC material and that thebars havea porosity of 35% (Section 2), Equation(2)isapplicable.

Onaverage,theconductivityvaluesofallthesampleslargely meettherequirementsforapplicationinIT-SOFCs(over100S/cm at600–700_{C),inlinewithliteratureresultsobtainedwithsimilar}

double-layered perovskite oxides [3,15,16,25,26] and with the measurements reported for NBC samples [8,27,28]. In the low temperatureregion,atT<250_{C,amoderatelossofconductivity}

isobservedforNBC0andNBC10,whiletheconductivitymaintains almostconstantinNBC5.Attemperatureshigherthan250C,all thesamplesexperienceadecreaseofconductivitywithincreasing

temperature,showingametal-typebehavior.Acomparisonwith theresultsofFig.6brevealsastrongsimilarityoftheconductivity curves with those of the oxygen loss, with respect both to temperaturethresholdsanddetailfeatures(e.g.thelocalmaxima ataround275C).Thisanalogyiswellexplainedbyconsideringthe followingoxygendesorptionstoichiometry,expressedinKröger– Vinknotation: 2Co_CoþOx O$ 1 2O2þ2Co x Coþ

n

 O ð3Þ

In theequation, the reduction of Coions (i.e. Co4+,positive chargecarriers,CoCo)attheexpensesoflatticeoxygenionsðO

x OÞ

generatesmolecularoxygen,positiveoxygenvacanciesðv

OÞandCo

ionswithnosurpluscharge(Co3+_{).Uponincreasingthe} tempera-ture,thesamplelosesoxygen,decreasingitsoxidationstate:this, inturn,generatesvacanciesanddecreasesthenumberofp-type electronicchargecarriers,whichareresponsiblefortheelectrical conductivityofthesample.

Alongwiththegeneralevolutionoftheconductivitycurves,itis importanttofocusonthedifferencesobservedatvaryingtheBa deficiency.TheNBC5sampleshowsthehighestconductivityinthe

wholetemperaturerange,followedbyNBC0andNBC10(insert). TheeffectofBaontheNBCconductivitycanberationalizedby considering that the cationic deficiency on Ba sites generates negative v00Ba defects in the structure; these defects can be

compensatedeitherbygenerationofanequalnumberofoxygen vacanciesv_O,orbyoxidationoftwoB-sitecobaltionsðCo

CoÞ.The

first mechanism hinders theelectrical conductivity, although it enhancestheoxygentransportcapability.Incontrast,thesecond mechanismbenefitstheelectricalconductivitybyincreasingthe number of electronic charge carriers, but hinders the oxygen transportfunction.Theexperimentalmeasurementssuggestthat thelattermechanismisactivepassingfromx=0.05tox=0.10,and that the high Ba deficiency of the NBC10 sample could be accompaniedbyamorefavorableoxygenexchangeactivity,asfirst suggested by the results of the thermogravimetric analyses (Fig.6a).Additionally,tobetterelucidatetheelectricalbehavior oftheNBCsamples,acomparisonwithliteratureofPrBa1-xCo2O5+d

cathodes is addressed. Avarietyof trendsare reportedfor the conductivity of PBC cathodes. A maximum (at x=0.05) in the conductivityatdecreasingBacontent(x=0–0.10)wasreportedby Dongetal.[13];inanalogousexperiments,Pangetal.[15]founda

Fig.9.ComplexplaneimpedancespectraoftheNBC10/GDC/NBC10symmetricalcellmeasuredbetween700and550Cinflowingair(100Ncc/min).Operatingconditions:

minimumintheelectricalconductivityatvaryingtheBadeficiency (x=0–0.08),locatedatx=0.03;finally,Jiangetal.[14]foundan increaseoftheconductivityupto0.10Badeficiency.Differently fromNd,Prisreportedtoexhibitboththe3+oxidationstateand the4+oxidationstate,asshownviaXPSinRefs.[9,14,29]fordouble perovskites(upto20%Pr4+_and80%Pr3+_{)andinRefs.[30,31]for} disorderedperovskites;inthecaseofNd,the4+oxidationstatehas neverbeenreportedinperovskiteoxidesandexclusivelythe3+is present.Thisisakeydifference:inPr-basedcompounds,bothCo andPrcanbeoxidizedtocompensatetheBavacancies,hencethe electronic charge carriers can be generated by Co and Pr, in different extent. Depending upon the preparation history and calcinationconditions (e.g.maximum temperatureand heating/ cooling rate), Pr4+ _{and Co}4+ _{are present in different amounts,} thereforeresultinginadifferentinfluenceonthecompensation mechanism.InthecaseofNd,theoccurrenceofasingleoxidation specieguaranteesthatthecompensationmechanismonlyinvolves theCo3+/Co4+coupleandtheoxygenvacancies.Inordertoclarifyif thecompensationmechanismofNBCcathodesactsprimarilyvia the oxygen vacancy formation or via the Co oxidation, it is

convenient toanalyzethe electrochemical ORR activity via EIS investigation.

3.3.EIScharacterizationoftheNBCsamples

Theindividuationoftheoptimalcathodicperformanceisthe resultofabalance.Ahighconductivitymustbeaccompaniedbya favorableactivityasoxygenreducingcatalyst:thesetwo character-istics can benefit from the lack of lattice Ba, although excess deficiencymaybecomedetrimental.Toelucidatethegeneraleffect ofBaonNBC,theORRactivitywasfirstevaluatedviaEIStestson symmetricNBC/GDC/NBCcells,inthetemperaturerangebetween 750and550C,atOCVandunderairflow.Experimentswerealso carriedoutbyusingHeinsteadofN2astheO2diluenttoverifythe impactof internal masstransfer limitationand individuate the correspondingeffect on the impedance arcs.For each cell, the microstructurewasanalyzedwithSEMafterthemeasurement.A sectionofthecellbasedontheNBC10cathodeisreportedinFig.3c, andalmostidenticalpicturesweretakenfortheothersamples.For each ofthecells, adetail of thecathodeelectrolyte/interfaceis

Fig. 10.ComplexplaneimpedancespectraoftheNBC10/GDC/NBC10symmetricalcellmeasuredbetween700and550_{CatvaryingtheO2partialpressurebetween100%and}

providedinthepanelsdtof,whereinthethicknessofelectrodes andtheparticlesizearebetterappreciated.Thepicturesconfirm that nosignificant differenceis foundin the microstructureat increasingtheBadeficiency:afaircontactisrealizedbetweenthe NBCparticlesandtheGDCelectrolyte.Althoughthecontinuitycan beimprovedbyusingafinergrainsize,agoodporosityisachieved, whichguaranteesa lowimpactofgasdiffusivelimitations.The imagesalsoshowthatalltheelectrodeshaveanaveragethickness of 30

m

m. The porosity was determined from the electrode weight and thickness via the density values (Table 1), and amounted to 45%. Overall, these results confirm that the symmetriccellshavealmostequalmorphologicalfeatures(layer thickness, porosity, grain size and connectivity) and therefore allowforafaircomparisonoftheEISresults.

TheASRvaluesestimatedfromtheEIStestsperformedwithair aresummarizedinFig.8a.Thebestperformanceisachievedbythe NBC10sampleforboththepolarizationresistance(0.1

V

*cm2_at 700C)andtheapparentactivationenergy(1.18eV),followedby the NBC5 sample (1.28eV and 0.25

_V

*cm2 _{at 700}_{C). The}

stoichiometricNBC0samplehasthelowestactivityandthelarger activation energy (1.31eV and 0.99

V

*cm2 _{at 700}_{C). The}

NBC10sampleis suitablefortheapplicationasa cathodein IT-SOFCs, being the polarization resistance lower than the target 0.15

V

*cm2_at700_{C,suggestedbytheguidelinesofRef.[32].A}

comparisonwiththemostrelevantliteratureresultsobtainedon NBCcathodes(Fig.8b)confirmsthattheactivationenergiesofthe three samples are well within the range extracted from the publisheddata,whichspansbetween0.9eVand1.42eVwithmost ofthevaluesbeingcloseto1.18eV.Withrespecttotheactivity, althoughadirectcomparisoncanbeonlyregardedasqualitative duetothedifferencesinmicrostructureandpreparationmethod, theASRoftheNBC10sampleisingoodagreementwiththatofthe work by Kim and Irvine [17], while the ASRs of NBC5 and NBC0reachhigherlevels.Thisresultcouldbeaconsequenceofthe firingtemperaturechosentoachievetheadhesionofthe electro-des in the symmetric cell, which is higher (1100C) than the commonvalue(1000C)adoptedintheworksand whichcould haveledtoalowerextensionoftheactivearea.

WithreferencetotheeffectoftheBacontent,theASRdecreases tenfoldpassingfromtheNBC0sampletotheNBC10,suggesting thattheoptimalconditionsaremetatthehighestBadeficiency. This result is widely in line with the resultsfound on similar perovskitesbasedonrareearths[14–16,25,33]:forthesematerials, itisgenerallyunderstoodthatacertainamountofBadeficiencyis beneficial,thankstotheintroductionofoxygenvacanciesinthe lattice. On Pr and La perovskites, a steady decrease of the polarizationresistanceistypicallyidentifiedasafunctionofthe Ba deficiency ( 0.1 for Pr [15] and 0.1–0.15 for La [16]): nonetheless,areductionof10timesofthepolarizationresistance atdecreasingtheBacontentislikelyauniquecharacteristicof Nd-basedcathodesand constitutesa mostsigni_ficant result,which distinguishestheuseofNdfromotherrareearths.Onanalogous Pr-based cathodes the polarization resistance almost halves passingfrom0to0.1Bade_ficiencyin[14],whereasthevariation reported in Refs. [13] and [15] is lower, 1.2 and 1.6 times respectively.ThesmallereffectreportedforPrcomparedtoNdcan berationalizedbasedonthepresenceoftheadditionalroutefor thecompensationoftheBavacanciesgivenbydoublevalenceofPr cations,asideoftherouteassociatedtotheCocations:thefirst possibilityisabsentinthecaseofNdandamoremarkedeffectof theBaunderstoichiometryisobserved.Additionally,aprogressive decrease of the apparent activation energy is expected upon decreasingtheBacontent.Inthepresentcase,thedecreaseofthe apparentactivationenergiesiswellwithintherangereportedfor double-perovskites:Pangetal.[15]foundaprogressivedecrease from1.19eVto1.05eVinPrBa1-xCo2O5+dsamplesuponvaryingthe

Bacontentfrom0to0.15;avariationbetween1.17eVand1.20eV wasfoundbythesameauthorswhenpassingfrom0to0.15Ba deficiency in LaBa1-xCo2O5+d cathodes. The observation of the

lowest polarization resistance at the highest Ba deficiency is consistentwiththeresultsoftheTGAanalyses(Fig.6),whichshow ahighercapabilityofNBC10inexchangingmolecularO2,aswellas withtheresultsoftheXRDanalyses(Fig.1)andofthecerimetric titrations(Table2),whichindicatetheoccurrenceofatransition fromtheorthorhombictothetetragonalstructureaccompaniedby anincreaseoftheoxygenvacancies.Apictureemergeswhereinthe tenfoldreductionofthepolarizationresistanceisconnectedtoand enhancedbythestructuraltransition.Ontheonehand,theresults suggestthatthemaximumactivityforNdBa1-xCo2O5+dcathodesis

almostachievedataround10%;ontheotherhand,theyindicate thatthecompensationmechanismactiveinthe5–10%deficiency rangeistheonewhereinthevacanciesarepreferablyformed.

Overall,whencomparingtheNBCsamplesofFig.8a,itisvery importanttonotethattheASRvaluesarerepresentativeofthe intrinsic chemical reactivity of the materials, given that mass transportlimitationswereexperimentallyproventobenegligible in the presence of air: indeed, as shown in Fig. 12b for the NBC10sample,almostnodifferenceisfoundwhenswitchingfrom N2toHe.Thesamebehaviorwasverifiedinthecaseofalltheother samples.ThislatterobservationconfirmsthatBainfluencessome stepsintheORRmechanism.Inordertogainadeeperinsightinto thechemicaleffectofBa,theNBC10andtheNBC5samplewere furtherinvestigatedindetailbyperformingtheEISmeasurements

Table4

FittingparametersfortheEIStestsonthesymmetricNBC10/GDC/NBC10cellat

varyingO2partialpressureat650C.OperatingconditionsasinFig.10.

Element 100% 21% 10% 5%

RHF(V*cm2) 0.116 0.098 0.097 0.085

QPEHF-Q(F*cm2) 1.90E-02 1.96E-02 1.45E-02 1.17E-02

QPEHF-n 0.643 0.659 0.671 0.778

CHF(F*cm2) 6.38E-04 7.71E-04 5.83E-04 1.62E-03

fHF(Hz) 2.15E+03 2.11E+03 2.81E+03 1.16E+03

RMF(V*cm2) 0.086 0.112 0.143 0.183

QPEMF-Q(F*cm2) 1.01E-01 5.50E-02 5.22E-02 5.31E-02

QPEMF-n 0.670 0.765 0.800 0.810

CMF(F*cm2) 9.68E-03 1.16E-02 1.54E-02 1.81E-02

fMF(Hz) 1.91E+02 1.23E+02 7.19E+01 4.80E+01

RLF(V*cm2) – 0.010 0.021 0.039 QPELF-Q(F*cm2) – 5.75 3.11 4.18 QPELF-n – 0.906 0.970 0.840 CLF(F*cm2) – 4.30 2.840 2.945 fLF(Hz) – 3.53 2.64 1.37 Table3

FittingparametersfortheEIStestsonthesymmetricNBC10/SDC/NBC10cellwith

airflowatvaryingtemperature.OperatingconditionsasinFig.9.

Element 750C 700C 650C 600C 550C

ROhm(V*cm2) 1.455 2.048 3.006 4.653 7.755

RHF(V*cm2) 0.026 0.047 0.098 0.196 0.400

QPEHF-Q(F*cm2) 3.86E-02 2.54E-02 1.96E-02 8.30E-03 5.47E-03

QPEHF-n 0.687 0.681 0.659 0.681 0.659

CHF(F*cm2) 1.68E-03 1.08E-03 7.71E-04 4.10E-04 2.30E-04

fHF(Hz) 3.61E+03 3.13E+03 2.11E+03 1.98E+03 1.73E+03

RMF(V*cm2) 0.029 0.055 0.112 0.358 0.976

QPEMF-Q(F*cm2) 4.87E-02 4.57E-02 5.50E-02 5.05E-02 3.92E-02

QPEMF-n 0.881 0.834 0.765 0.764 0.752

CMF(F*cm2) 2.00E-02 1.39E-02 1.16E-02 1.47E-02 1.34E-02

fMF(Hz) 2.74E+02 2.09E+02 1.23E+02 3.03E+01 1.22E+01

RLF(V*cm2) 0.010 0.010 0.010 0.012 0.012

QPELF-Q(F*cm2) 7.08 3.07 5.75 9.55 3.51

QPELF-n 0.783 0.902 0.906 0.870 0.886

CLF(F*cm2) 3.35 2.10 4.30 6.91 2.34

atvaryingO2contentbetween100%and5%between700Cand 550C.

3.4.EISinvestigationontheNBC10sample

TheresultsoftheEISexperimentscarriedoutinairandunder varying O2 content are reported in Figs. 9 and 10 for the NBC10sample. Atall thetemperatures, the impedance spectra wereanalyzed byapplication of theEquivalent CircuitMethod (ECM) based on a circuit of the type LROhm(RHFQHF)(RMFQMF) (RLFQLF).Listheinductancecausedbytheelectricalequipmentand bytheleadwires,ROhmistheohmicresistancemainlyduetothe electrolyte,whilethethreeremainingRQelementsareassociated withprocessesoccurringintheelectrodesatdifferent character-isticfrequencies.Thiscircuitiscommonlyadoptedintheanalysis ofthepropertiesofperovskiteoxides:thehighfrequencyarc(HF) isassociatedtotransferprocessesthatinvolveionicspecies,such astheinclusionof oxygenionseither attheinterface withthe electrolyteor in thebulk of the cathodestructure;the middle frequencyarc(MF)isrelatedtoelectrodeprocesses,suchasthe dissociative adsorption of oxygen, the formation of adsorbed oxygenions,orthesurfacediffusionofoxygenadatoms;thelow

frequencyarc (LF)is due tomassdiffusive transport processes, eitherinternal (intra-porous)or external(across theinterphase boundarylayer).ThechoiceofthreeRQelementsmirrorsthebasic descriptionofthewayaMIECcathodeworks,byassumingthree very general macro-classes of phenomena, each of which is expected to occur in a certain frequency range. Aside of the elementrelatedtothephysicaltransportofmolecularoxygento theelectrodesurface,themodelaccountsfortwo characteristic processes: one element represents the gas/electrode surface processesthatleadtotheformationofelectroactiveionicspecies starting from the adsorption of molecular oxygen; the other elementrepresentsthediffusionoftheelectroactiveionicspecie, inthebulkofthematerialoracrosstheelectrolyteinterface.For each of these macro-classes, the identification of one rate determiningstepisaccomplishedexclusivelyviatheestimation oftheO2reactionorder,theactivationenergy,thecapacityandthe characteristic frequency, as well as on the basis of literature comparison.Giventhatthekineticratesofthedifferentsteps,and ultimatelytheircharacteristicfrequencies,varywiththenatureof thematerial(forinstance,withtheBadeficiencyorwiththelattice structure), overlaps or separations are possible, and various equivalentcircuitmodelscanfitthecurves.Alongwiththispoint,

Fig.11.NBC10symmetriccell,T=550_C–700_{C.PanelA:dependenceoftheHFresistanceonPO2.PanelB:dependenceoftheMFresistanceonPO2.PanelC:dependenceof}

theMFresistanceonPO2.PanelD:Arrheniusplotoftheresistanceofthehighfrequency(RHF),middlefrequency(RMF)andlowfrequency(RLF)processesunderairflow.The

itisveryimportanttonotethat,inalltheanalysespresentedinthis work,thecircuitsadoptedarethesimplestandtheoneswiththe lowest number of elements and fitting parameters which guarantee a solid and physically sound interpretation of the spectra.Thischoiceallowsforflexibility:inthecasesomereaction stepshavesimilarrates,theadditionofoneormoreRQelements permitsacoherentdescriptionofthedatawithoutlossofphysical significance. In this way, the differences in the circuits can be directly related tothe intrinsic differences in the chemical or electrochemical behavior of the material (as shown later in Section3.5).Althoughbeyond thescopeof thework,it is also worthytonotethatimprovedresolutionandnarroweridenti fica-tionoftherelevantprocessescanbeachievedbyapplicationof additional experimental techniques, such as the non-linear impedance spectroscopy (NLEIS) [34], as well as of more sophisticatedinterpretativemethods,forinstancethedifferential impedanceanalysis(DIA)[35]ortheuseofdistributionfunctionof relaxation times (DRT)[36], which do not require the a priori assumptionof an equivalentcircuitand do not applytheleast squaremethodtoestimatetheparameters.

The general associations between the RQ elementsand the transport processes are explained by estimating the order of magnitudeoftheircharacteristicfrequencies.Inthecaseof the transport of oxygen ions within the bulk of the material, the followingequationcanbeapplied:

f¼ DO

d2char

ð4Þ In the equation, DO is the bulk diffusion coefficient of the oxygenion,anddcharisthecharacteristiclengthofthemotion.It can beassumed that the characteristic length travelled by the oxygenionisasmallfraction( 10%)ofthemeanparticlesize(dP 2

m

m) [37]. Diffusion coefficients about 5*107cm2_{/s are} reported in the literature at 700C for PrBaCo2O5+d double

perovskites[13,38].FortheNBCsamples,acharacteristiclength of1

_m

mcanbeassumedfortheiondiffusion,whichresultsina frequency range centered about 1kHz, in agreement with the experimentalobservations(fullsymbolsinFigs.9and10).

The electrode processes are associated to chemical steps whereinoxygenadsorbatesplayarole.Theratesoftheseprocesses

aregenerallydefinedbysurfaceexchangecoefficients,whichare representative oftheactivityofadsorptionofmolecularoxygen betweenthegasphaseandtheelectrode:globalchemicalkinetic rates areassumed, with rate constantsKchem evaluated experi-mentally via electrical conductivity relaxation methods (ECR). Kchem values of 102–5*103cm/s at 700C are reportedin Pr-baseddoubleperovskitessimilartoNBC[13,38].Whenestimating the characteristic frequency, the surface area active for the adsorption (av, specific adsorptionsurface area) must be taken intoaccount:

fKchemav ð5Þ

The specific adsorption surface area of the electrode is reasonably estimated based on the porosity and on the mean particlesize,assumingtherelationshipforpackedbedsofspheres, i.e. av=6/dP*(1

e

void)accordingto[39].Witha particlesize of 2

m

mand 45% porosity, a surface area of 1.65*104 _cm2_/cm3 _is obtained,whichcorrespondstoacharacteristicfrequencyrangeof 165–82Hz.Thisfrequencyrange islowerthanthat oftheionic transportprocessandisaddressedasmiddle.Alsothisvalueisin agreementwiththeexperimentalresultsofFigs.9and10.

The characteristic frequency of the non-chemical diffusive transportprocessesisgenerallylowerthanthoseofthechemical andelectrochemicalsteps,henceitiscommonlyassociatedtothe lowfrequencyRQelement.Inthepresentcase,thesmallthickness ofthelayersandthegoodporositysuggestthattheimpactofmass diffusion inside the porous electrode is limited, and that the externalmasstransportplaysarole.Itisindeedknownthatthe storagecapacityofthestagnantgasgivesrisetoanadditionalarc withlowcharacteristicfrequencywhentheboundarylayeroutside theelectrodeissignificant[40,41].Masstransferdistancesupto 300

m

m–1mmcanbereached,dependingonthemassflowrate and onthe experimental set-up, especially when the tests are conductedwithastagnationpoint_flowgeometryandwithavoid mixing-spacebetweentheelectrodeandthegasnozzles.Thisasset istypicalandwasadoptedinthepresentexperiments.Asamatter offact,theappraisalofanimpedancearcat1–10Hzassociatedto theseconditionshasbeennumericallydemonstratedbyBesslerin

[42]. Considering also the presence of the current collectors (250

m

mthick)andofthesilverpaste,acontributionfromdiffusive

Fig.12.PanelA:PlotofthephaseasafunctionofthefrequencyfortheNBC10/GDC/NBC10cell,underairflowcondition,atvaryingtemperaturebetween700and550_C.

effects is expectedand theinclusion of an RQ element in the equivalentcircuitisrigorous(althoughitsquantitativerelevanceis verymodest,asshownbelow).

In thecaseoftheNBC10sample,averysatisfactoryfittingis obtainedatalltemperaturesandallO2partialpressureswiththe 3arcsmodel.Theresultsarerepresentedaslinesinthefigures.For theexperimentsinair,Table3summarizesthevaluesofthecircuit elements fitted by the model, as well as the values of the correspondingcapacitances and relaxation frequencies.Table 4

reportsthevaluesofthesameparametersfortheexperimentat varying of the O2 content performed at 650C, which is fully representativeofalltheanalogousexperimentsatother temper-atures.Foreachprocess,theresistancevaluesobtainedfromthe bestfitsareplottedinArrheniusformasafunctionoftheinverse temperature,inordertoevaluatetheactivationenergy(Fig.11d). ThereactionordersnofO2areestimatedaccordingtotheinverse relationshipbetweentheresistanceandthereactionrate(Eq.(6)). TheyarereportedforeachtemperatureinFig.11a–c.

Ri/PnO2 ð6Þ

TheexperimentalEISspectrashowthatonesingledepressed arcisalwaysobserved.AtdecreasingthetemperatureandtheO2 concentration,thepolarizationresistanceincreases.The inspec-tionofthephasecurvesasafunctionofthefrequency(Fig.12a) revealsthatthegrowthoftheimpedancearcsinvolvesprimarily thelowfrequencyzone,whichextendsupto0.1Hzwhenmoving to 550C and 5% O2, and keeps almost unaltered in the high frequencyregionat5kHz.Thephaseanalysisalsoshowsthatthe mainvariationislocalizedinthemiddlefrequencyregion,wherea peakrisesandshiftsbackwardupondecreasingthetemperature (andthePO2).Onlyat700Cand5%O2content,asecond,smaller arc appears in the 1_–0.1Hz frequency range. In this case, as reported in Fig. 12b, the second arc can be unambiguously associated todiffusive masstransport, given thatits amplitude changesuponsubstitutingN2withHe.Thisbehavioriscommonin thecaseofMIECcathodes,whereinthewholeelectrodevolumeis involvedinthereactionprocessandaseparateddiffusivearchis seldom distinguished, except at the lowest O2 partial pressure

[40,43].Nonetheless, theimpactof thediffusionphenomenais

limited, as verified by the resistance values of Tables 3 and 4

Fig.13.ComplexplaneimpedancespectraoftheNBC5/GDC/NBC5symmetricalcellmeasuredbetween700and550CatvaryingO2partialpressurebetween100%and5%v/

(between0.012–0.039

V

*cm2_{passingfrom0.21to0.05O} 2v/vat 650C).Consistentlywiththe associationof theLFarc toslow diffusive effects, high values of capacitance (3–7F/cm2_{) and} characteristicfrequency(4–5Hz)arecalculated. Theassociation is also confirmed by the almost null activation energy (0.1eV,

Fig.11d) and by the order of reaction observed at varying O2 content,whichiscomprisedbetween0.83and0.94,verycloseto thefirstorderexpectedfromdiffusiveeffects.

Fromamechanisticviewpoint,theassociationoftheHFandMF arcs with a reaction step requires the reaction orders to be determined together with the activation energies. For the HF process, the values of the capacitance and of the relaxation frequency obtainedin thefitting (103–104_F/cm2 _{at 1–5kHz)} suggest the association with an ionic transport step. This associationisconfirmedbytheactivationenergyof1eVestimated atalltheconcentrationofO2(Fig.11d)and,aboveall,bythevery smallreaction orderfound forO2 at allthe temperaturelevels (Fig. 11a),whichiscomprisedbetween0.02and0.12.Thislatterisa strongindicationthattheHFarcisassociatedwiththeprocessof iontransportfromthebulkofthecathodephasetotheinterface withtheelectrolytephase,followedbychargetransferacrossthe interface. The following reaction steps are generally used to describetheprocess[44]:

Oxo ðNBCbulkÞ$O x o ðNBC=GDCinterfaceÞ ð7Þ Oxo ðNBC=GDCinterfaceÞ$O x o ðGDCbulkÞ ð8Þ

Other authors find very similar pieces of evidence when exploringthekineticsofdouble-layerperovskitesdopedwithrare earth.Forinstance,Aminand Karan[45] observeda zeroorder dependenceoftheHFarcthewithLaBaCo2O5+dcathodes,whileShi

etal.founda0.18dependenceonEuBaCo2O5+dcathodesat700C

[46].Similarly,Pangetal.[15]discussedazeroorderprocesson symmetrical cells with PrBa1-xCo2O5+d cathodes under almost

identical conditions compared to the present ones, estimating values of capacitancebetween 103–105_F/cm2 _{at 10to 1kHz} between650and700C.

InthecaseoftheMFarc,anactivationenergyof1.30–1.35eVis estimatedbetween100and5%O2partialpressure(Fig.11d).The reactionorderfallswithin0.26and0.27between700and600C (Fig.11b).The values ofcapacitance and relaxationfrequencies (102–103_F/cm2_,at102_{–10Hz,Tables3and4)areconsistentwith} achargetransferprocessoccurringattheelectrode.Thesepiecesof evidencesuggestthattheMFprocesscanbeidentifiedasacharge transferprocesswhereinanoxygenionisgeneratedinthecathode phase via reaction betweenan oxygen adatom and a vacancy, accordingtothefollowingstoichiometry:

O_þ2e_þv_O$Ox

O ð9Þ

Thisprocessisinlinewiththeoverallpicturethatemergesfrom Eq.(6)to(9),whereintheadsorbedOspecies,formedviafastO2 dissociativeadsorption,isfirstturnedintoanelectroactiveoxygen ion,includedintheNBC10phase(Eq.(9)),successivelytransported throughthe latticebulk up totheGDC interface(Eq. (7)), and finally transferred across the interface (Eq. (8)). Upon further decreasingthetemperatureto550C,therate-limitingstepofthe MF arc changes, as indicated by the change in the associated reaction order, which passes toa value of 0.47. This values is representativeofadissociativeadsorptionstepthatoccursonthe cathodesurface:

O2þ$2O ð10Þ

Thisstepwas recentlyfoundtolimittheORRmechanismof stoichiometricLaBaCo2O5+dcathodesbetween750and600C[47].

Averysimilarshiftfromanorderof0.25toanorderof0.47was reportedbyChenetal.[48]onPrBaCo2O5+dcathodeswhenpassing

from500to450_{C.Additionally,severalotherauthorsworkingon}

double-layered perovskites report a variation in one or more kinetic stepsupon changing eitherthe PO2 or thetemperature

[46,49].Itisworthytonotethatachangeinthedeterminingstep

shouldbeaccompaniedbyavariationintheactivationenergy:in thepresentcase,suchachangeisscarcelyseen,duetotheabsence ofadditionalexperimentsattemperaturesclosetoorlowerthan 550C. Nonetheless, the results indicate that the dissociative adsorptionstep(Eq.(10))shouldhaveahigheractivationenergy than the ion inclusion one, so that it becomes limiting upon decreasingthetemperature.

3.5.EISinvestigationoftheNBC5sample

The kinetic effect of Ba understoichiometryon the electro-catalytic activityof the NBC cathodes is betterappreciated by examiningtheresultsoftheEISexperiments,carriedoutunderair

Table5

FittingparametersfortheEIStestsonthesymmetricNBC5/SDC/NBC5cellwithair

flowatvaryingtemperature.OperatingconditionsasinFig.13.

Element 750C 700C 650C 600C 550C

ROhm(V*cm2) 1.628 2.286 3.341 5.127 8.415

RHF1(V*cm2) 0.034 0.059 0.190 0.492 1.270

QPEHF1-Q(F*cm2) 2.17E-03 2.55E-03 1.31E-03 7.39E-04 4.92E-04

QPEHF1-n 0.934 0.883 0.835 0.808 0.766

CHF1(F*cm2) 1.11E-03 8.00E-04 2.55E-04 1.12E-04 5.14E-05

fHF1(Hz) 4.22E+03 3.37E+03 3.29E+03 2.89E+03 2.44E+03

RHF2(V*cm2) – 0.045 0.130 0.293 0.662

QPEHF2-Q(F*cm2) – 6.92E-04 2.13E-04 1.06E-04 6.05E-05

QPEHF2-n – 0.943 0.956 0.968 0.948

CHF2(F*cm2) – 3.69E-04 1.31E-04 7.46E-05 3.05E-05

fHF2(Hz) – 9.65E+03 9.38E+03 7.28E+03 7.87E+03

RMF(V*cm2) 0.103 0.181 0.409 0.837 2.143

QPEMF-Q(F*cm2) 7.64E-03 6.43E-03 6.12E-03 6.15E-03 6.13E-03

QPEMF-n 0.826 0.796 0.764 0.765 0.787

CMF(F*cm2) 1.69E-03 1.14E-03 9.62E-04 1.22E-03 1.89E-03

fMF(Hz) 9.12E+02 7.73E+02 4.04E+02 1.57E+02 3.93E+01

RLF(V*cm2) 0.005 0.005 0.005 0.005 0.005 QPELF-Q(F*cm2) 3.009 2.883 2.675 2.738 2.719 QPELF-n 0.960 0.977 0.969 0.961 0.959 CLF(F*cm2) 2.537 2.599 2.323 2.295 2.259 fLF(Hz) 11.619 13.325 14.458 14.447 14.053 Table6

FittingparametersfortheEIStestsonthesymmetricNBC5/GDC/NBC5cellat

varyingO2partialpressureat650C.OperatingconditionsasinFig.13.

Element 100% 21% 10% 5%

RHF1(V*cm2) 0.158 0.190 0.214 0.220

QPEHF1-Q(F*cm2) 1.74E-03 1.31E-03 1.32E-03 1.45E-03

QPEHF1-n 0.793 0.835 0.833 0.859

CHF1(F*cm2) 2.04E-04 2.55E-04 2.56E-04 3.86E-04

fHF1(Hz) 4.95E+03 3.29E+03 2.90E+03 1.87E+03

RHF2(V*cm2) 0.111 0.130 0.129 0.147

QPEHF2-Q(F*cm2) 3.07E-04 2.13E-04 2.46E-04 2.40E-04

QPEHF2-n 0.936 0.956 0.952 0.947

CHF2(F*cm2) 1.51E-04 1.31E-04 1.45E-04 1.35E-04

fHF2(Hz) 9.46E+03 9.38E+03 8.47E+03 8.02E+03

RMF(V*cm2) 0.277 0.409 0.500 0.624

QPEMF-Q(F*cm2) 6.12E-03 6.12E-03 6.73E-03 7.25E-03

QPEMF-n 0.785 0.764 0.744 0.728

CMF(F*cm2) 1.07E-03 9.62E-04 9.46E-04 9.64E-04

fMF(Hz) 5.39E+02 4.04E+02 3.37E+02 2.64E+02

RLF(V*cm2) – 0.005 0.011 0.019

QPELF-Q(F*cm2) – 2.675 4.270 3.967

QPELF-n – 0.969 0.755 0.792

CLF(F*cm2) – 2.323 1.586 2.014

flowatvaryingtheO2partialpressurefrom100%to5%between 700Cand550C.TheresultsaresummarizedinFig.13:forthe sakeofclarity,theresultsofthetestsatvariablePO2arereported onlyat650C,acase thatisfullyrepresentativeofalltheother temperatures. The numerical resultscalculated with Zview are summarizedinTables5and6,withrespecttothefittingofthe circuitelements,andinFig.14a-c,intermsofordersofreaction andactivationenergy.Thefirst,fundamentaldifferencebetween theEISresultsoftheNBC10andtheNBC5sampleisthechoiceof theequivalentcircuit:inthecaseofNBC5,onlyacircuitwithfour RQelementsconsistentlyfitsthearcsandallowsforagoodmatch betweenthemodelandthedataunderalltheexploredconditions. Indeed,exclusivelywiththiskindofcircuit,thephysico-chemical consistencyofsomeparameterscanberespected,aboveallthe relaxationfrequency,whichdecreaseswithadecreaseoftheO2 partialpressure(Table6)andofthetemperature(orkeepsalmost constant,asinthecaseofthenon-activateddiffusivearc,Table5), asaconsequenceofthedropoftheintrinsicrate.

AlsointhecaseoftheNBC5sample,theEISspectrapresentonly onedepressedarc.Thelowfrequencyregionisagainassociatedto theinternaldiffusivemasstransport:consistently,theprocessis almostnonactivated,afirstorderdependenceonO2isfoundand

highcapacitiveandhighfrequencyvaluesareestimated(10–2F/ cm2at 1Hz). However,differently fromtheNBC10sample, the experimentaldataandthephaseplot(Fig.14d)showthat,upon decreasing the temperature, the EIS arcs grow in the high frequency region, while the low frequency region maintains almostconstantataround1Hz.AsclearlyseeninFig.14d,going from700to550Caphasepeakdevelopsinthehighfrequency regioninadditiontotheonethatgrowsinthemiddlefrequency region,whichisalsoobservedinNBC10.Thisisthemostrelevant differencebetweentheNBC10sampleandtheNBC5:anadditional arcis presentathighfrequency, whoseresistanceprogressively increasesupon decreasing thetemperature.Theanalysisof the reactionorderandtheactivationenergyrevealsthatafirsthigh frequencyarcispresent(HF1)whoseactivationenergyis1.2eVand whosedependencyonO2maintainscloselybetween0.10and0.12 (Fig. 14a,fullsymbols).Thecapacityvaluesareintherange2



104– 8



105F/cm2_{andthefrequencybetween5and1kHz(Tables5and}

6). The reaction order suggests that the HF1 process can be associatedtoachargetransferprocessoccurringattheinterface andinvolvingoxygenion,astheonealreadydescribedinEquation

(8)ofoxygentransferacrosstheTPBinterfacewithinclusioninthe electrolyte bulk. For the second high frequency arc (HF2), an

Fig.14.NBC5symmetriccell,T=550_C–700_{C.PanelA:dependenceoftheHF1andHF2resistancesonPO2.FullsymbolsareHF1arc,emptysymbolsareHF2.PanelB:}

dependenceoftheMFresistanceonPO2.PanelC:Arrheniusplotoftheresistanceofthehighfrequency(RHF1andRHF2),middlefrequency(RMF)andlowfrequency(RLF)

processesunderairflow.PanelD:plotofthephaseasafunctionofthefrequency,underairflowcondition,atvaryingtemperaturebetween700and550C.Theabsolutevalue

activationenergyof 1.1eVisestimated,witha reactionorder closertozero(0.06–0.08). Thecapacity valuesare intherange 104–105_F/cm2 _{and the frequency is around 10kHz. The zero} orderdependencysuggeststhatthestepdescribedinEquation(7)

canbeassociatedtothisarc:anassociationwiththecrossingof grainboundaries, asobserved on PrBaCo2O5+dcathodes [48] is

unlikely,giventhatmuch smallercapacities andlower temper-aturesareexpected.TheMFfrequencyarchasthesameassociation alreadyobservedfortheNBC10sample:anactivationenergyof 1.1–1.2eV,withadependenceonO2of0.25andcapacitiesof102– 103F/cm2_{at100–10Hz.}

Overall,theresultsoftheEISanalysissuggestthatthedecrease in the Ba content promotes the steps at high frequency, by decreasingthepolarizationresistanceassociatedtothetransferof oxygenions,eitherinsidethecathodephaseoracrosstheinterface. At5%Badeficiencythesetwostepsappeartobeslowerandmore distinguishable, whereas at 10% Ba deficiency the difference is almostabsent:thatis,inthislattercase,thestepofdiffusionof oxygen ions withinthe lattice of the NBC phase becomes less resistive.Thisresultisinfullagreementwiththepictureprovided bythestructuraltransitionfromtheorthorhombicstructureofthe sampleat5%Badeficiency,whichhasaloweramountofoxygen vacanciesandslower kinetics,tothetetragonalstructureof the sample at 10% Ba deficiency, which has a higher amount of vacanciesandthereforeafasterrateofoxygendiffusion.Theresult also agrees well with the increase of the oxygen vacancies suggestedbytheTGandbytheconductivitymeasurementsinthe range5 to10% Badeficiency.Froma theoreticalviewpoint,the comparison between the results of the numerical analyses performedonthesamplesat5%and10%Badeficiencyallowsto identify the mechanistic effect of Ba in the oxygen reduction mechanism,akey-rolebeingplayedbythechoiceoftheequivalent circuitswhichrequirethelowestnumberofparameterstofitthe arcswhilemaintainingaphysicallysoundbasis.

4. Conclusions

TheeffectoftheA-siteBa-deficiencyonthecrystalstructure and the electrochemical properties of NdBa1-xCo2O5+d

double-layeredperovskiteoxides(x=0
0.20)was investigated.TheBa deficiencyaffectsthecelldimensionsbyreductionoftheblattice parameter,andbyincreaseofthemeanoxidationstateofCoand increase of the oxygen vacancies. Although the highest total electricalconductivityismetataBadeficiencyequalto5%(750S/ cmat700C),ASRinvestigationrevealsthata10timesreductionof thepolarizationresistanceisachievedpassingfrom0to10%Ba understoichiometry,avalueatwhichtheoptimalelectrocatalytic activityintheoxygenreductionprocessisrealized(0.1

_V

*cm2_at 700C).Couplingthetotalconductivity testswith thermogravi-metricandEIStestssuggeststhatthemetalvacanciesintroduced byBaaremainlycompensatedviatheintroductionofnoveloxygen vacancies, and, to a minor extent, by the formation of charge carriersbyoxidation ofCoions.A detailed EISanalysisvia the equivalentcircuitmethodrevealsthattheeffectofBaismainly relatedtoprocessesoccurringathighfrequency,which tendto becomelessresistiveupondecreasingtheBacontent.Specifically, athighfrequencythepolarizationismainlycontributedbyoxygen iondiffusionandtransferacrosstheinterface,whereasatmiddle frequencytheformationofanoxygenionanditsinclusioninthe cathode phase are limiting. The decrease in the Ba content promotesthestepsathighfrequency,possiblyrelatedtothebulk diffusionofoxygentowardstheinterfacewithGDC.

References

[1] D.J.L.Brett,A.Atkinson,N.P.Brandon,S.J.Skinner,ChemicalSocietyReviews37

(2008)1568.

[2] E.V.Tsipis,V.V.Kharton,JournalofSolidStateElectrochemistry12(2008)

1367–1391.

[3] J.H.Kim,A.Manthiram,J.Electrochem.Soc.155(2008)B385.

[4] A.Tarancón,M.Burriel,J.Santiso,S.J.Skinner,J.A.Kilner,JournalofMaterials

Chemistry20(2010)3799.

[5] A.Tarancón,S.J.Skinner,R.J.Chater,F.Hernández-Ramírez,J.A.Kilner,Journal

ofMaterialsChemistry17(2007)3175.

[6] S.Choi,S.Yoo,J.Kim,S.Park,A.Jun,S.Sengodan,J.Kim,J.Shin,H.Y.Jeong,Y. Choi,G.

Kim,M.Liu,ScientificReports3(2013).

[7] J.H.Kim,M.Cassidy,J.T.S.Irvine,J.Bae,ChemistryofMaterials22(2010)

883–892.

[8] S.Yoo,S.Choi,J.Kim,J.Shin,G.Kim,ElectrochimicaActa100(2013)44–50.

[9] F.Jin,H.Xu,W.Long,Y.Shen,T.He,JournalofPowerSources243(2013)10–18.

[10] J.Kim,S.Choi,S.Park,C.Kim,J.Shin,G.Kim,ElectrochimicaActa112(2013)

712–718.

[11] F.Wang,D.Chen,Z.Shao,ElectrochimicaActa103(2013)23–31.

[12] S.Sengodan,S.Choi,A.Jun,T.H.Shin,Y.-W.Ju,H.Y.Jeong,J.Shin,J.T.S.Irvine,G.

Kim,NatureMaterials14(2015)205–209.

[13] F.Dong,M.Ni,Y.Chen,D.Chen,M.O.Tade,Z.Shao,JournalofMaterials

ChemistryA2(2014)20520–20529.

[14] L.Jiang,F.Li,T.Wei,R.Zeng,Y.Huang,ElectrochimicaActa133(2014)364–372.

[15] S.Pang,X.Jiang,X.Li,Q.Wang,Z.Su,JournalofPowerSources204(2012)

53–59.

[16] S.L.Pang,X.N.Jiang,X.N.Li,H.X.Xu,L.Jiang,Q.L.Xu,Y.C.Shi,Q.Y.Zhang,Journal

ofPowerSources240(2013)54–59.

[17] J.H.Kim,J.T.S.Irvine,InternationalJournalofHydrogenEnergy37(2012)

5920–5929.

[18] A.C.Tomkiewicz,M.Meloni,S.McIntosh,SolidStateIon.260(2014)55–59.

[19] W.Wang,T.Peh,S.Chan,T.Zhang,SolidOxideFuelCells11(Sofc-Xi)25(2009)

2277–2281.

[20] K.Zhang,L.Ge,R.Ran,Z.Shao,S.Liu,ActaMaterialia56(2008)4876–4889.

[21] A.C.Larson,R.B.V.Dreele,GeneralStructureAnalysisSystem(GSAS),Los

AlamosNationalLaboratoryReportLAUR86-748(2004).

[22] J.C.Burley,J.F.Mitchell,S.Short,D.Miller,Y.Tang,J.SolidStateChem.170

(2003)339–350.

[23] J.Wright,A.V.Virkar,JournalofPowerSources196(2011)6118–6124.

[24] F.Zhao,A.V.Virkar,JournalofPowerSources195(2010)6268–6279.

[25] X.Jiang,Y.Shi,W.Zhou,X.Li,Z.Su,S.Pang,L.Jiang,JournalofPowerSources

272(2014)371–377.

[26] J.Wang,F.Meng,T.Xia,Z.Shi,J.Lian,C.Xu,H.Zhao,J.-M.Bassat,J.-C.Grenier,

InternationalJournalofHydrogenEnergy39(2014)18392–18404.

[27] T.Broux,M.Bahout,J.M.Hanlon,O.Hernandez,S.Paofai,A.Berenov,S.J.

Skinner,JournalofMaterialsChemistryA2(2014)17015–17023.

[28] J.H.Kim,L.Mogni,F.Prado,A.Caneiro,J.A.Alonso,A.Manthiram,J.

Electrochem.Soc156(2009)B1376.

[29] F.Jin,Y.Shen,R.Wang,T.He,JournalofPowerSources234(2013)244–251.

[30] Y.Hinatsu,M.Itoh,N.Edelstein,J.SolidStateChem.132(1997)337–341.

[31] S.Majumdar,M.R.Lees,G.Balakrishnan,D.M.Paul,Journalof

Physics-CondensedMatter15(2003)7585–7590.

[32] B.C.H.Steele,SolidStateIon.86-8(1996)1223–1234.

[33] C.Setevich,L.Mogni,A.Caneiro,F.Prado,J.Electrochem.Soc.159(2012)B73.

[34] J.R.Wilson,D.T.Schwartz,S.B.Adler,ElectrochimicaActa51(2006)

1389–1402.

[35] D.E.Vladikova,Z.B.Stoynov,A.Barbucci,M.Viviani,P.Carpanese,J.A.Kilner,S.J.

Skinner,R.Rudkin,ElectrochimicaActa53(2008)7491–7499.

[36] H.Schichlein,A.C.Muller,M.Voigts,A.Krugel,E.Ivers-Tiffee,J.Appl.

Electrochem.32(2002)875–882.

[37] J.Fleig,JournalofPowerSources105(2002)228–238.

[38] G.Kim,S.Wang,A.J.Jacobson,L.Reimus,P.Brodersen,C.A.Mims,Journalof

MaterialsChemistry17(2007)2500–2505.

[39] J.Deseure,Y.Bultel,L.Dessemond,E.Siebert,P.Ozil,J.Appl.Electrochem.37

(2007)129–136.

[40] S.B.Adler,Chem.Rev.104(2004)4791–4843.

[41] S.Primdahl,M.Mogensen,J.Electrochem.Soc.146(1999)2827–2833.

[42] W.G.Bessler,J.Electrochem.Soc.153(2006)A1492–A1504.

[43] M.Koyama,C.J.Wen,T.Masuyama,J.Otomo,H.Fukunaga,K.Yamada,K.

Eguchi,H.Takahashi,J.Electrochem.Soc.148(2001)A795–A801.

[44] Y.H.Li,R.Gemmen,X.B.Liu,JournalofPowerSources195(2010)3345–3358.

[45] R.Amin,B.Kenney,K.Karan,J.Electrochem.Soc.158(2011)B1076–B1082.

[46] Z.Shi,T.Xia,F.Meng,J.Wang,J.Lian,H.Zhao,J.M.Bassat,J.C.Grenier,J.Meng,

FuelCells14(2014)979–990.

[47] R.Pelosato,A.Donazzi,G.Dotelli,C.Cristiani,I.NataliSora,M.Mariani,Journal

oftheEuropeanCeramicSociety34(2014)4257–4272.

[48] D.Chen,R.Ran,K.Zhang,J.Wang,Z.Shao,JournalofPowerSources188(2009)

96–105.