Research
Paper
In
fluence
of
electrochemical
cycling
on
the
rheo-impedance
of
anolytes
for
Li-based
Semi
Solid
Flow
Batteries
A.
Narayanan
a,*
,
D.
Wijnperlé
a,
F.
Mugele
a,
D.
Buchholz
b,c,
C.
Vaalma
b,c,
X.
Dou
b,c,
S.
Passerini
b,c,
M.H.G.
Duits
aa
PhysicsofComplexFluidsgroup,UniversityofTwenteandMESA+Institute,POBox217,7500AEEnschede,TheNetherlands
b
Helmholtz-InstituteUlm,Helmholtzstraße,11,89081Ulm,Germany
c
KarlsruheInstituteofTechnology(KIT),P.O.Box3640,76021Karlsruhe,Germany
ARTICLE INFO
Articlehistory: Received15May2017
Receivedinrevisedform3August2017 Accepted5August2017
Availableonline12August2017
Keywords:
SemiSolidFlowBattery SolidElectrolyteInterface CarbonBlack
Rheology
ImpedanceSpectroscopy
ABSTRACT
TherecentlylaunchedconceptofSemi-SolidFlowBatteries(SSFBs)showsastrongpotentialforflexible
energystorage,buttheliquid-dispersedstateoftheelectrodematerialsintroducesseveralaspectsof
whichascientificunderstandingislacking.Westudiedtheeffectofelectrochemicalcyclingonthe
rheologicalandelectricalpropertiesofaSSFBanolytecontainingLi4Ti5O12(LTO)andKetjenBlack(KB)
particles in EC:DMC solvent with 1M LiPF6, using an adapted rheometer that allows in situ
electrochemicalcyclingandelectricalimpedancespectroscopy.Charging(lithiation)causedareduction
intheelectronicconductivity,yieldstressandhighshearviscosityofthefluidelectrode.Formildly
reducingvoltages (1.4V),thesechangeswerepartiallyreversed ondischarging.Formorereducing
voltages thesechangeswerestrongerand persistent.Thefindingof comparabletrendsforafluid
electrodewithouttheLTO,lendssupporttoasimplisticinterpretation,inwhichalltrendsareascribedto
theformationofasurfacelayeraroundtheconductiveKBnanoparticles.ThisSolidElectrolyteInterphase
(SEI)insulatesparticlesandreducesthevanderWaalsattractionsbetweenthem.SEIlayersformedat
less reducing voltages, partially dissolve during thesubsequent discharge. Those formed at more
reducingvoltages,arethickerandpermanent.Astheselayersincreasetheelectronicresistanceofthe
fluidelectrodeby(morethan)anorderofmagnitude,ourfindingshighlightsignificantchallengesdueto
SEIformationthatstillneedtobeovercometorealizeSSFBs.
©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense
(http://creativecommons.org/licenses/by/4.0/).
1.Introduction
Semi-Solid FlowBatteries (SSFBs),as recentlyintroducedby Dudutaetal.[1],compriseapromisingadditiontothespectrumof rechargeable battery systems. The advantages of SSFBs over conventionalbatteriesliein thedecouplingofpower(cellsize) and energy (tank size), and the potential for adjusting the chemistry of the system during operation. In particular non-aqueousSSFBsystemsareinteresting,sincetheyoffermuchhigher energydensitiesascomparedtomoreconventionalaqueousredox flowsystems[1].Theabilityofnon-aqueousSSFBstoprovideand storeenergyinaflexiblewaymakesthemparticularlypromising forgridapplications.
However,akeyaspectinwhichSSFBsareyetunprovenistheir performanceafterrepeatedelectrochemicalcycling.Whilemany SSFBs usethe samematerials [1,2]as conventional lithium-ion batteries,theymaypotentiallydegradeindifferentwaysduetothe dispersedstateofthesolid matter.InSSFBs (de)lithiationtakes placeinelectrochemicallyactiveparticleswhiletheelectronsare transportedtothecurrentcollectorsviaconductivenanoparticles (CNPs).Theoccurrenceofboth particlesinthe(sub)micronsize rangehasseveralconsequences:i)thesurface-to-volumeratiois relatively large, and ii) Brownian motion and interparticle interactions nowplay a role.Electrochemistry inducedchanges can therefore manifest themselves in different ways. They can directly affect individual particles (e.g. electronic conductivity, lithium uptake)but alsocollective effects arepossible,because colloidal particles show a tendency for self-assembly into a microstructure. The colloidal interactions, which drive this assembly,arelikelytobeaffectedbytheelectrochemicalcycling, and since the structure is kept dynamic by Brownian motion
*Correspondingauthor.
E-mailaddress:a.narayanan@utwente.nl(A. Narayanan).
http://dx.doi.org/10.1016/j.electacta.2017.08.022
0013-4686/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense(http://creativecommons.org/licenses/by/4.0/).
ContentslistsavailableatScienceDirect
Electrochimica
Acta
and/orshearflow,themicrostructuremayadaptto electrochemi-calchanges.
While theprecisemicrostructureofSSFBfluidsis still tobe ascertained, the generally accepted view [1–3] is that in the absenceofflow,theCNPsassembleintoabranchedpercolating network. This network provides electronic conduction and sustains static forces, thereby resisting the sedimentation of particles.Inflow,thenetworkgetsbrokendownintoagglomerates withasizethatdependsontheshearrate[3–5].Thecontribution ofactiveparticlestothemicrostructureislessunderstood.They areexpectedtobehavelikeadisorderedfluidthatsurroundsthe CNPnetwork.Thislackoforderisinferredfromtheinsignificance of bothattractiveand long-ranged repulsiveforces;theformer sinceotherwisetheviscositywouldbeveryhigh,thelatterfrom thestrongscreeningbythedissolvedsalt[6].
ThesedifferencesbetweenSSFBsandconventionallithium-ion batteriesraisethequestion, howdegradationprocessessuchas volumeandstructuralchangesoftheactivematerialsupon(de-) lithiation[7] ortheformationofsolidelectrolyteinterface(SEI)
[8,9]affectSSFBs.Severalconsequencesofsuchprocessesforfluid electrodesareconceivable.ConsideringtheCNPnetwork,boththe natureoftheinterparticlecontactsandtheirnumberdensitycan change:theformerasaconsequenceofsurfaceprocesses,andthe latterdueto(forflowbatteriesinherent)mechanicalrejuvenation: shear-induced fragmentation of the CNP network creates a possibility for the fragments to re-assemble into a different microstructure when the fluid returns to the quiescent state3.
Macroscopically,theelectronicconductivityandtheyieldstress arelikelyaffectedbythesemicroscopicprocesses.Whilecritically relevanttoSSFBs,theabovephenomenacanpotentiallyalsoaffect othertypesofbatterysystemsthatuseself-assemblingcolloidal particles, such as polysulphide [10] and carbon free [11] flow batteries. Recentwork oncarbon slurrybased iron redoxflow batterieshasshownelectronicconductivityenhancementthrough changesintheinterparticlecontactsduetoironplating[12].
The objective of the present work is twofold: to quantify changesinrheologicalandelectricalperformanceduetorepeated electrochemicalcycling,andtogainamechanisticunderstanding ofthesemacroscopicchanges.Toachievethesegoals,a commer-cialrheometerwasextendedtoallowparallelelectricalimpedance measurements, as well as a controlled cycling of the fluid electrodesviathe inclusionof a lithiumcounter electrode.The studiedfluidelectrodeconsistsofamixtureofKetjenBlack(KB) andLi4Ti5O12 (LTO)particlesdispersedinEC:DMC1:1with1M
LiPF6. LTO haspreviously beenidentifiedas a promising active
materialfor SSFBs [13] as lithiationoccurs atabout 1.55V vs Li/Li+, within the safe operating range of the non-aqueous electrolyte[2,14,15].Toexaminetheroleofelectrochemicalstate (andhistory),wecyclethefluidelectrodetoaseriesofincreasingly reducing voltages, measuring the rheological and electrical properties before and after each charge and discharge step. Comparisonsarealsomadebeforeandaftermechanical rejuvena-tion, to probe the changes in self-assembly. To facilitate interpretationofthevariouschangeswealsomakeacomparison betweentheresultsfortheKB-LTOelectrode,andafluidelectrode withouttheLTO.
2.MaterialsandMethods 2.1.FluidElectrodePreparation
Ethylenecarbonate(EC)anddimethylcarbonate(DMC)were obtained from Sigma Aldrich (anhydrous, 99%+ purity). Binary mixturesofECandDMCwere1:1bymass.LP30(EC:DMC1:1with 1MLiPF6)wasobtainedfromBASF.KetjenBlackEC600JDpowder
(KB)wasobtainedfromAkzoNobel.Li4Ti5O12powderwasobtained
from Südchemie. Lithium foil was purchased from Alfa Aesar (99.9%).Allsamplepreparationsandexperimentswerecarriedout inanMBraunArgon-filledglovebox(O2,H2Obelow1ppm).Two
fluidelectrodeswereprepared:amixtureof1wt.%KBand5wt.% LTO,andareferencesampleat1wt.%KB.Thedryparticleswere first wetted by EC-DMC solvent for 8hours to improve their dispersibility; from an earlier study it is known that KB is colloidallyunstableincarbonatesolventwithlargeamountsofsalt
[6].TheKBreferencecontained2.9wt.%KBwhileforthemixtureit was2.6wt.%KBand13.4wt.%LTOatthisstage.Afterwards,LiPF6
salt (Alfa Aesar (98%)) was added via a concentrated solution (LP30+LiPF6)toreacha concentrationof1M(viscosity 4mPas [16]).Afteranadditional8h,thesampleswerehomogenizedby rotor stator mixing (Ultraturrax) at 15000rpm for 2min, and loadedintherheo-impedancesetup.
2.2.Cycling-rheo-impedancesetup
Electrochemical cycling and rheo-impedance measurements wereperformedonastresscontrolledrheo-meter(HaakeRS600) withahome-builtadaptation(Fig.1)comprisinganextensionofa previouslydescribedsystem[3].Briefly,the60mmparallelplate geometryof therheometer wasused asa base.Acopperplate attachedtotheupperrheometerrotorservedasashearingsurface, currentcollectorandelectrodeforelectricalimpedance spectros-copy(EIS).Aperforatedstainlesssteel(316)platewasusedasthe bottomshearingsurface.Duetothesmallsize(1mmradius)and fraction(<40%)oftheholes,rheologicalmeasurementscouldbe performedwithreasonableaccuracy(within 5%;testwith1.231 Pasand0.01Pascalibrationoils).Thisplatealsoservedasasecond electrodeforEIS.ACelgard2500separatorwasusedtoseparate theperforatedplatefromasecondcompartmentwithalithium foil(onatitaniumcurrentcollector).Thisallowedthefoiltobein contacttheelectrolytebutnottheparticles.Inthesolventtrapof therheometer(notshown),mercurywasusedasaworkingfluidto ensurealowfrictionandlow noiseelectricalconnectiontothe rotatinguppergeometry[3].Duringrheologicaltestsallelectrodes weredisconnected.
EIS measurements wereperformed(between theupper and perforatedplate,withS1closedandS2open)inafour-terminal configuration. Theperforated platewas excitedby a sinusoidal voltageof<50mVinthefrequencyrangefrom10MHzto0.01Hz. Amplitude sweeps ontheKB-only fluid had indicated that the responsewaslinear(andhencetheimpedancesthesame)atleast upto100mV(Note:thesevoltageswereappliedattheHF2output. Theactualvoltageacrossthesamplewastypicallymuchsmaller). Currentsweremeasuredbyatransimpedanceamplifier(HF2CA, ZurichInstruments)onthevirtuallygroundedrotor.Abuffer pre-amplifier(HF2TA, ZurichInstruments)wasusedtomeasurethe potentialdifferencebetweentheperforatedplateandtherotor.An impedancespectroscope(HF2IS,Zurichinstruments)wasusedto
Fig.1.Schematicofrheo-impedancesetup.darkgrey:uppergeometry,perforated plate,bottomcurrentcollectortranslucentgray:membrane,beige:lithiumfoilon currentcollector.EitherswitchS1orS2isclosed,toallowEISorelectrochemical cycling.
extract the complex impedance from the current and voltage signals. The lithium electrode was allowed to float (it was disconnectedfrom the external circuit) during these measure-ments.Frequencydependentparasiticimpedances ofthe setup werecalibratedoutusingthe“openshort”technique[17,18].
Sampleswereelectrochemicallycycled(withS2closed)using thelithiumfoil(>30cm2)asacounterelectrodeandtheCelgard
membraneasanionpermeablemedium.Asthemaximumcurrent waslow(<50
m
Acm2)thetotalpolarizationwasbelow50mV[19,20]. Currents were measured through a 50
V
resistor. The potential of the perforated plate was allowed to float during cycling.Galvanostaticandpotentiostaticchargingwereperformed usingtheimpedancespectroscope.CustomLabVIEWcodeswere usedtoperformcycling andEIS and tosynchronizethem with rheologicalmeasurements.2.3.ExperimentalProtocol
Allsurfacesincontactwiththesample(excludingthelithium foilandtheseparator)weresandpaperedandthoroughlycleaned outside the glovebox prior to the experiment. They were subsequentlywettedwithDMCfor15minpriortosampleloading in the glovebox. Cycling-rheo-impedance experiments were performedwithagapof250
m
mbetweenuppercurrentcollector andperforatedplate.Toavoidsamplevariationsduetodifferences inshearhistory[3],wepre-shearedeachsampleat1000s1for 200s, and subsequently allowed them 200s of rest. This ‘mechanical rejuvenation’ was applied before each cycling (charging or discharging) step. Rheo-impedance measurements werecarriedoutbothbeforeandafterthistreatment.Thelithium electrode was disconnected during rheo-impedance measure-ments.AschemeoftheprotocolisgiveninFig.2.Toindicatetheelectrochemicalhistory,wecodeoursamplesas follows: (Voltage window number). (Cycle number). (Charge (lithiation)/Discharge (delithiation) step). The pristine state is denotedasP.Forexample,code2.3.Crepresentsthestatereached inthe2ndvoltagewindowafterchargingthefluidforthe3rdtime undertheseconditions.Moreover,eachsamplehastwo mechani-calstates:beforeoraftermechanicalrejuvenation.
Sampleswere(dis-)chargedgalvanostaticallyusingacurrentof 1.5mA.AssuminganLTOconcentrationof(5wt.%=)0.121gml1 andaspecificcapacityof175mAhg1,thiscorrespondstoarateof aboutC/10.Oncethecutoffvoltagewasreached,thevoltagewas helduntilthecurrentfellbelow0.5mA.ThereafterEIS measure-mentswereperformed,takingintoaccounttheaforementioned mechanical protocol. Next the yield stress was measured by ramping up the shear stress (62s per stress decade) while measuring the strain. The log(strain) versus log(stress) curve wasfittedwithtwostraightlinesandthestressattheintersection was taken to be the yield stress [21]. The flow curve was subsequentlymeasuredbypre-shearingat1000s1for200sand
then slowly stepping theshear rate downwards from 1000s1 (at20stepspershearratedecade).Afteranequilibrationtimeof 20stheviscositywasaveragedoveronesecond.Sixchargeand dischargehalf-cycleswereperformedforeachvoltagewindow.At theendoftheexperiment,sampleswererecovered,driedat60C andthenanalyzedpostmortemoutsidetheglovebox.
3.ResultsandDiscussion
SSFB fluid electrodes conduct through ionic and electronic pathways[1,4].Asthemetalcurrentcollectors thatenclosethe fluidelectrodeareionicallyblockingbutelectronicallyreversible, these two contributions can (in principle) be separated using impedancespectroscopy [3,22].The electronicresistanceof the fluidelectrode’spercolatedparticlenetwork (withsome contri-butionsfromthecurrentcollectorinterface[3,4])correspondsto the low-frequency limit of the real impedance. Experimental timescalesdonotalwaysallowaccesstothislimit,andtherefore the low frequency realimpedance (LFRIsee Fig.3 panelB) at 0.01Hz wastaken aspracticalmeasure (suitableforidentifying trends) of theelectronicresistance. Furtherjustificationof this approachwillbepresentedinFig.4B,wherewealsofitLFRIvalues usinganequivalentelectricalcircuitmodel.
Fig.3 showstheimpedance,yieldstressandviscosity(from nowontermedtogetherasrheo-impedance)ofafluidelectrode containing1wt.%KBand5wt.%LTO,cycledbetween1.4-2.5Vand 1.0-2.5V. Most measurements (solid symbols) were performed aftermechanicalrejuvenationofthefluid;wewillfocusonthese first.
InregimeI,withalessreducingcut-offof1.4V,boththeLFRI andtheyieldstressshowanalternatingbehavioroncycling,witha higherLFRIandaloweryieldstressat1.4Vascomparedto2.5V. Thedifferencein theLFRIat thetwo statesof chargebecomes progressivelysmalleruponcycling,whilethedifferenceinyield stressremainsroughlythesame.SubsequentcyclinginregimeII, withacutoffof1.0V,leadstostrongchanges.TheLFRItriples,and furtheralternationissuppressed.Theyieldstressroughlyhalves, butherethealternationsremain(Fig.3C).Thelossofalternations inthemeasuredLFRImaybeduetoexperimentallimitations:as shownbytheNyquistplot(Fig.3B),thetimeconstantofthelow frequencyarcshowsalargeincrease,therebycompromisingthe sensitivityoftheLFRItotheelectronicresistance.
Changesintherheologicalpropertiesarereflectedinnotjust theyieldstressbutalsotheflowcurves.Inspectionofthelatter revealsthattheshear-ratedependenceof theviscosityisrather similarforallsamples(insetFig.3D).Thisallowsrepresentationof theeffectsofelectrochemicalcyclingviaaviscosityscale-factor VSF(mainpanelofFig.3D).TheyieldstressandVSFshowasimilar dependenceontheelectrochemicalstate;lithiationlowersboth quantitiesandviceversafordelithiation.ChangingtoregimeII,i.e. the cycling between2.5 V-1.0V, the yield stress and viscosity reduce.AgainthesechangescorrelatewellwiththehigherLFRI, withtheexceptionthatalternationsremainintheyieldstressand VSF.ThismaybeduetotheaforementionedissuewiththeLFRI. Interpretationoftheprogressivechangesinrheo-impedanceof thefluidaftercyclingisnottrivial.Oneaspecthereofisthatthe durationofthechargeanddischargewasnotthesameforeach cycle(seeSIFig.S1,andTableTS1).Forthisreason,onlyrelative changes caused by electrochemical cycling (and mechanical rejuvenation)willbediscussed.Furthermore,themeasurements are performed in a complex system, consisting of several instrumentalpartsandamulticomponentfluidelectrode. Analyz-ingtheelectricalandrheologicaldatainconjunction,thescopeof interpretationcanhoweverbenarroweddown.Importantly,the electricalandmechanicalsignalsoriginatefromthesamesystem,
Fig.2.Schematicofthe(repetitive)measurementprotocol.Pictogramsindicate mechanical rejuvenation andelectro-chemical (dis)charging.The rejuvenation servestocreateareferencestatebybreakingdowntheparticleagglomeratesand lettingthemre-assembleagain.SinceEISdoesnotinvolvemechanicaldeformation, measurementsbeforeandafterrejuvenationcanbecompared.
comprisingabulkfluidbetweenthesametwometalplates(the uppercurrentcollectorandtheperforatedplate).
Thisstillleavesthequestion,whethercontributionsfromthe metal-fluid interfaces can be neglected or not. A significant interfacialcontributiontotherheologicalsignalwouldrequirea mechanicallyweaklayerneartherheometergeometrywalls(e.g. duetodepletionofparticlesorweakparticle-wallinteractions). Therearehowevernoindicationsforthis. First,theyield stress curves(seeSIFig.S3)indicateaninitialelasticdeformationand finitestrainatyield(alsoforsampleswithaveryhighLFRI).This corresponds well to a gap-spanning network, whereas a weak interfaciallayerwouldalreadyyield(i.e.flow)ataninfinitesimal strain. Secondly, the viscosities (at high shear rates) change appreciably with each cycling step, implying that the forces responsibleforparticleagglomerationshoulddothesame3.Thisis
only possibleif the particlesthemselves undergochanges. The absenceofaparticle-depletedlayeratthemetalplates,asinferred fromtherheology,suggeststhattheLFRIsignalisdominatedby thefluidbulk.Fromadifferentperspective,sincethegap(250
m
m) betweenthemeasuringsurfaces forEIS spansO(1000) particle diameters,thenumberofparticlecontactsinvolvedinanelectron conductionpathhastobeverylarge,ascomparedtothesingle particle-metalcontactpercurrentcollector.Itisthusappropriate toseekanexplanationoftheobservationsinFig.3 intermsof changestotheparticles(andnottheenclosingmetalsurfaces).Since SSFB electrodes are multi-component mixtures, the effects of electrochemical cycling are not limited to just one component. However, additional observations help identifythe mostdominantchanges.ExaminationoftheLTOparticles(bothin pristine state, and after cycling) revealed that no structural decompositioncouldbedetectedwithXRD(see SIFig.S4).The higherelectronicconductivityofcarbonblackscomparedtoLTO suggests that they will have a dominant influence on the suspensionelectronicconductivity.Moreover,thedifferent rheol-ogiesofsuspensionsofonlyKB(yieldstress,higherviscosity)and onlyLTO(noyieldstress,lowerviscosity)inthesamesolvent(see SIFig.S5)suggeststhatthecarbonblackparticleshaveadominant influence on the rheology. In absence of an all-encompassing model for how electronic conductivity and yield stress are generated in thefluid electrodes,it is thus very reasonable to assumethatthedominantcontributiontobothsignalscomesfrom theKB.
To demonstratethis further,weconsidera similar measure-mentonasamplethatcontainsonlyKBasaparticulatecomponent (Fig.4).Thissamplewascycledinfourvoltageranges,wherethe firsttwocorrespondtothoseoftheKB-LTOfluidelectrode(seeSI Fig.S2andTableTS1).InregimeI(1.4Vcutoff),thebehaviorofthe KBfluidelectrodeissimilartothatoftheKB-LTOfluidelectrode.In the subsequent, more reducing regimes II and III (with cutoff voltagesof1.0and0.8Vrespectively),thedifferencesbetweenthe
Fig.3.Rheo-impedanceof1wt.%KB+5wt.%LTOfluidelectrodesubjecttocyclingindifferentvoltageranges.(X):pristinestate,(D):dischargedto2.5V,(r):chargedto indicatedvoltage.Opensymbols:notmechanicallyrejuvenated,Closedsymbols:mechanicallyrejuvenated.A)LowfrequencyrealimpedancefromEISspectraobtainedat theendofchargeordischarge.B)NyquistplotsofdatapointsmarkedwitharrowsinA.ThedottedlinesconnectthelowestfrequencyimpedancetotheLFRIC)Yieldstress.D) Viscosityscalefactor:thefactorwithwhichtheviscosityat1000s1hastobemultipliedtocoincidewiththatofthepristinestate.Theinsetshowstheflowcurves(viscosity inPasvs.shearrateins1)aftermultiplicationwiththeVSFs.
‘charged’(reduced)and‘discharged’statesprogressivelybecome larger,spanningalmostanorderofmagnitudeforboththeLFRI andyieldstress.Thealternationofbothpropertiesappearstobe repeatableinregimesIIandIII.However,cyclinginregimeIVwith a cutoff of 0.6V (well outside the stability window of the electrolyte)resultsinamassiveincreaseintheLFRIanddecrease intheyieldstressandVSF.NocleartrendsareobservedintheLFRI orrheologicalpropertiesforsubsequentcycles.
Whilethevoltagerangesappearslightlyshifted,thebehaviorof theKB-onlyfluidelectrodequalitativelyagreeswiththatofthe KB-LTOmixture.Wecanthusattributethebehaviorofbothsystemsto theKBnetwork.Withinthisfocusedinterpretation,theLFRItrends inFigs.3and4canberationalizedbytheformationofaninsulating SEI layer around the KB particles upon exposure to reducing voltages,andapartialdissolutionofthislayerduringdelithiation (2.5V).Recentstudieshavedemonstratedthat thecomposition andpropertiesoftheSEIdependonthepotentialversuslithium whereitisformed[23–25].Athighervoltages,asparseandless insulatingSEIlayercomposedoforganiccompounds(moreprone todissolution)isformed.Atlowerpotentials;athicker,denser,less soluble, and more insulating layer composed of inorganic compoundsgetsformed.
Ourinterpretationthatalayergetsformedonthecarbonduring charginganditpartiallydissolvesduringsubsequentdischarging, isfurthercorroboratedbyanexperimentusingellipsometry.Here a copper substrate coated with a sputtered carbon layer was immersedinEC:DMC1:1+1MLiPF6,andexposedtovoltagesof
1.0 and 2.5V with respect to an immersed lithium foil. Ellipsometric angles
c
andD
were measured in-situ with a WoollamM2000ellipsometer,asafunctionofwavelength.Fig.5showstheevolutionofPsi
c
andDeltaD
(foratypicalwavelength of800nm)asafunctionofelectrochemicalhistory.Thepristine sample(intheabsenceofcurrent)showsfairlyconstantc
andD
valuesthatareinagreementwitha81nmthickcarbonlayeron bulk copper. Strongand ongoing changesin both ellipsometric anglesareobservedwhenthevoltageissettoa1.0V(‘charging’) whilesettingthevoltageto2.5V(‘discharging’)resultsinapartial recoveryofbothPsiandDeltaangles.SinceCuandcarbondonot dissolveundertheseconditions,thechangesinc
andD
during exposureto1.0Vmustbeduetothedepositionofanewmaterial onthesubstrate.Thismakesitlikelythatthepartialreversalof thesechangesonexposureto2.5Vareduetoapartialdissolution ofthislayer.Adetailedquantitativeanalysisofthe(c
,D
)databy comparisontoopticalmodelsforthelayerstructureispossible,but thechoiceofanappropriatemodelinconjunctionwiththelimited additionalinformationaboutthelayer’sopticalpropertiespresent challenges.Asimplisticmodelwhichcandescribeourwavelength dependent(c
,D
)dataisa5-layerstack:(bulk)Cu-C-an interme-diate layer-SEI-(bulk)electrolyte. The intermediate layer repre-sentsalineartransitioninopticalpropertiesfromthatofcarbonto that of the SEI, accounting for intermixing. Using optical parametersfortheSEIasgiveninMcArthuretal.[26],themodel producesthethicknessesinFig.5(inset).ItcanbeseenthatSEI startstoformwhen thevoltageis switchedto 1.0Vand whenFig.4.Rheo-impedanceof1wt%KBfluidelectrodesubjecttocycling.SymbolsarethesameasFig.3.A)LFRI.B)NyquistplotsofdatapointsmarkedwitharrowsinAandfits (describedlater)usingtheequivalentcircuitinFig.7C)Yieldstress.D)Viscosityscalefactor,definedsimilarlyasinFig.3.
switchedbackto2.5Vitpartiallydissolves.Themodelseemsto overestimate theSEIthicknesseshowever (witha maximumof around43nm),whichmaybeduetotheactualrefractiveindices beinghigher.
Consistency with the observed changes in the rheological propertiesimpliesthatthelayermustalsoweakentheattractive forcesthatholdtheKBnetworktogether.Intheelectrostatically screeningenvironment of the1M salt solutionthis is possible through reduced van der Waals attractions. Assuming that the (typical)contactgeometrybetweentwostickingKBunitsremains thesame,thiswouldsuggestaloweredHamakerconstant.
ItmayseemsurprisingthatweobservestrongeffectsofaSEI layerwithinthe‘safe’operatingrangeoftheelectrolyte.Thismay
be due to an uneven current distribution caused by the inhomogeneityof fluid electrodes.Thiscouldleadtovariations inlocalparticlestatesofcharge,triggeringreductiveelectrolyte decompositionandSEIformation[27–29].
Oncyclingthefluidelectrodetolowervoltages,athickerSEI forms,withadrasticeffectontheelectronicresistance.Itshouldbe notedthattheincreaseintheelectronicresistanceduetocharging (lithiation), is larger than indicated by the LFRI. This is easily recognizedfromFigs. 3 Band 4 B, inwhich thelow-frequency semi-circle isfarfromcompleteat0.01Hz,inparticularforthe lithiated state. The large difference (more than one order of magnitude)intheLFRIbetween1.0V-2.5Vand0.8V-2.5Vofthe fluidelectrodeonlycontainingKBindicatesthatthedifferencein
Fig.5.TimedependenceoftheellipsometricanglesofacarbonthinfilmsampleinEC:DMC1:1+1MLiPF6,duringsubsequentexposureto1.0and2.5VversusaLicounter electrode.Psianglesareindicatedinred,whileDeltaanglesareshowninblue.Inset:FittedthicknessofSEI(solidLine)andInterlayer(dottedline).Thefitswereperformedin theCauchyregion(600–1000nm)wheretherewasminimaldepolarization.
Fig.6.Proposedmechanismtoexplaintheobservedcombinedeffectof(dis)chargingtohigherpotentialsandsubsequentmechanicalrejuvenationontheLFRI(i.e.the conductivityoftheKBnetwork).Redcolorindicates(inanexaggeratedway)thepresenceofanSEIlayer,whichgrowsduringcharging,andshrinksduringdischarging. Particlecontactsarelessstronglyaffectedbylayerdepositionordissolution.Note:inrealitytheprimaryKBparticlesarefractal-like,andre-assemblyaftershearleadstoa different(butstatisticallyequivalent)network.
electronicconductancemustbehuge,evenforvoltages1.0Vvs theLicounterelectrode.Atlower(morereducing)voltages,the effectof theirreversible processonthe electronicresistanceis probablyevenlarger.
Tofurtherexaminetheformationofsurfacelayers,wecompare the LFRI before (open symbols) and after (closed symbols) mechanicalrejuvenationinFigs.3and4.For bothsamplesitis clearthatrestructurationbyshear(followedbyrest),consistently increasestheLFRIaftercharginganddecreasesitafterdischarging. Assumingthatthepre-shearbreaksdownallagglomerates,and that the re-agglomeration process is not impeded by energy barriers(diffusion limitedagglomeration[30]),the micro-struc-tureafterrejuvenationwill(statistically)bethesame.However, theconductivityofinterparticlecontactswillhavechanged.Thisis becausethecontactsbetweensingleparticlesofthenetworkare lesslikelytobeaffectedbytheformationordissolutionofSEI,as theyarelessexposedtotheelectrolytesolution.Consequently,less SEI is formed at the contacts during charging (lithiation) and likewise less is dissolved during discharging (delithiation). Mechanicalrejuvenationleadstorandomizationofthecontacts. Thus, the subsequently formed contacts will contain mainly
maximallygrown(ormaximallydissolved)SEI.Thismechanistic explanation,illustratedinFig.6,thussupportsthattheformation oftheinsulatingSEIispartiallyreversibleathigherpotentials,i.e. lower state of charge. We remark here that the observed reversibility verylikely depends onthe duration that thefluid electrodepotentialisheldoutsideoftheelectrochemicalstability windowoftheelectrolyte.Also,asthereversibilityoftheSEI(and itscomposition)dependsonthepotentialatwhichitisformed, thismechanism isnolongervalidatverylowpotentials(0.6V) wherepermanentSEIisformed.
The identification ofSEI as thecauseof impedancechanges makes it interesting to extract the electronic resistances by modellingtheimpedancespectraoftheKBonlysystemusinga simplifiedequivalentcircuit(Fig.7).Theioniccontributioncanbe modeledasanionicresistanceRioninseriestoaconstantphase
elementQionthatrepresentsthedoublelayercapacitancesofthe
electrolyte interfaces. To model the electronic part, we use a resistor RKB that represents the summed KB intra-particle
resistances in series with a parallel resistor RSEI and capacitor
CSEIthataccountsforthesummedinterparticleimpedances(due
to SEI). A capacitor Cgeo in parallel to the rest of the circuit
represents the geometric capacitance of the system. The fits (Fig.4Binset)showgoodagreementwiththespectra.
WealsonotethatthemeasuredLFRIandthefittedelectronic resistance(thesumoftheKBinterandintraparticleresistances)in
Fig.8Ashowthesametrends,therebyjustifyingourearliergiven interpretationofthechangesinLFRI.Clearly,andasexpected,the measuredLFRIvaluesunderpredicttheelectronicresistancefor mostlithiatedstates.InFig.8BwedecomposethefittedLFRIvalues intothecontributionsRKBandRSEI.Thiscomparisonshowsthatthe
mostimportantchangesinthetotalelectronicresistancecomes fromtheinterparticleresistance.
4.Conclusions
Ourstudyoftheeffectofcyclingontherheo-impedanceofa LTO-basedSSFBanolytehasproducedseveralnewinsights.Two keypropertiesareverysensitivetotheelectrochemicalstateand history: lithiation causes the electronic resistance (LFRI) to increase and the yield stress to decrease, and vice versa for delithiation.Alithiationvoltagebelow1.0VvsaLielectrodecauses a drastic increase in LFRI. A suspension of only KB particles responds ina similarway toelectrochemicalstateand history, indicating that the observed effects of cycling can be largely
Fig.7.EquivalentcircuitusedtofitthedataofFig.4.
Fig.8.ResultsofthefitsofspectraofthemechanicallyrejuvenatedsamplesinFig.4A.A)ComparisonoftotalelectronicresistanceandLFRI.B)ComparisonoffitsummedKB interparticle(RSEI)andintraparticle(RKB)electronicresistances.
attributedtotheKB. Asimplemechanisticpicturethatcaptures mostfindings,isoneinwhichtheKBparticlesgetcoveredbyaSEI layer during charge (lithiation). For less reducing, i.e. higher voltages,apartofthislayerdissolvesduringdischarge.Thelayer electrically insulates the KB particles and diminishes their attractions. The contact points between the KB units are less affectedbythelayergrowth,similartoconventionalsolidlithium batteries. However, in SSFBs, a mechanical rejuvenation of the structuretakesplaceeverytimethefluidgetspumped,leadingto theincorporationofthethickerlayersintotheKBbackbone.For stronglyreducingvoltagesapermanentSEIlayerisformed.
TheimplicationsofourfindingsforSSFBsaresignificant.Alow electronic resistance is crucial to battery performance, and a sufficientyieldstressisrequiredtosuspendactivematerials.Both thesepropertiesareadverselyaffectedbySEIformationunderthe exploredexperimentalconditions.Ourfindingsshowthatfurther research into chemistries withabsolutely no (insulating) layer formationwouldberequiredtorealizeSSFBs.
Acknowledgements
WeacknowledgeProf.NieckBenesoftheFilmsinFluidsgroup andProf.WiebedeVosoftheMembraneScienceandTechnology Group at the University of Twente for granting access to the ellipsometer.The researchleading totheseresultshasreceived fundingfromtheEuropeanUnionSeventhFrameworkProgramme (FP7/2007-2013)undergrantagreementn608621.
AppendixA.Supplementarydata
Supplementarydataassociatedwiththisarticlecanbefound,in the online version, at http://dx.doi.org/10.1016/j. electacta.2017.08.022.
References
[1]M.Duduta,B.Ho,V.C.Wood,P.Limthongkul,V.E.Brunini,W.C.Carter,Y.M. Chiang,Adv.EnergyMater.1(2011)511–516.
[2]K.B.Hatzell,M.Boota,Y.Gogotsi,ChemicalSocietyReviews44(2015)8664– 8687.
[3]A.Narayanan,F.Mugele,M.H.G.Duits,Langmuir33(7)(2017)1629–1638,doi: http://dx.doi.org/10.1021/acs.langmuir.6b04322.
[4]M.Youssry,L.Madec,P.Soudan,M.Cerbelaud,D.Guyomard,B.Lestriez,Phys ChemChemPhys15(2013)14476–14486.
[5]M.Youssry,L.Madec,P.Soudan,M.Cerbelaud,D.Guyomard,B.Lestriez, JournalofPowerSources274(2015)424–431.
[6]Y.Zhang,A.Narayanan,F.Mugele,M.A.C.Stuart,M.H.Duits,Colloidsand surfacesA:Physicochemicalandengineeringaspects489(2016)461–468. [7]C.R.Fell,L.Sun,P.B.Hallac,B.Metz,B.Sisk,JournalofTheElectrochemical
Society162(2015)A1916–A1920.
[8]J.Vetter,P.Novák,M.Wagner,C.Veit,K.-C.Möller,J.Besenhard,M.Winter,M. Wohlfahrt-Mehrens,C.Vogler,A.Hammouche,Journalofpowersources147 (2005)269–281.
[9]X.Han,M.Ouyang,L.Lu,J.Li,Y.Zheng,Z.Li,JournalofPowerSources251 (2014)38–54.
[10]F.Y.Fan,W.H.Woodford,Z.Li,N.Baram,K.C.Smith,A.Helal,G.H.McKinley,W. C.Carter,Y.-M.Chiang,Nanoletters14(2014)2210–2218.
[11]Z.Qi,G.M.Koenig,JournalofPowerSources323(2016)97–106.
[12]T.J.Petek,N.C.Hoyt,R.F.Savinell,J.S.Wainright,JournalofPowerSources294 (2015)620–626.
[13]T.Ohzuku,A.Ueda,N.Yamamoto,JournaloftheElectrochemicalSociety142 (1995)1431–1435.
[14]E.Ventosa,G.Zampardi,C.Flox,F.LaMantia,W.Schuhmann,J.Morante, ChemicalCommunications51(2015)14973–14976.
[15]L.Lombardo,S.Brutti,M.A.Navarra,S.Panero,P.Reale,JournalofPower Sources227(2013)8–14.
[16]Y.R.Dougassa,J.Jacquemin,L.ElOuatani,C.c.Tessier,M.r.m.Anouti,The JournalofPhysicalChemistryB118(2014)3973–3980.
[17]E.Barsoukov,J.R.Macdonald,Impedancespectroscopy:theory,experiment, andapplications,JohnWiley&Sons,2005.
[18]L.Callegaro,Electricalimpedance:principles,measurement,andapplications, CRCPress,2012.
[19]B.Burrows,R.Jasinski,JournaloftheElectrochemicalSociety115(1968)365– 367.
[20]F. La Mantia, C. Wessells, H. Deshazer, Y. Cui, Electrochemistry Communications31(2013)141–144.
[21]G. Schramm, A practical approach to rheology and rheometry, Haake Karlsruhe,1994.
[22]J.Jamnik,J.Maier,JournalofTheElectrochemicalSociety146(1999)4183– 4188.
[23]P.Lu,C.Li,E.W.Schneider,S.J.Harris,TheJournalofPhysicalChemistryC118 (2014)896–903.
[24]S.J.An,J.Li,C.Daniel,D.Mohanty,S.Nagpure,D.L.Wood,Carbon105(2016) 52–76.
[25]J.Collins,G.Gourdin,M.Foster,D.Qu,Carbon92(2015)193–244. [26]M.McArthur,S.Trussler,J.Dahn,JournaloftheElectrochemicalSociety159
(2012)A198–A207.
[27]T.Nishi,H.Nakai,A.Kita,JournalofTheElectrochemicalSociety160(2013) A1785–A1788.
[28]J.Nanda,J.Remillard,A.O’Neill,D.Bernardi,T.Ro,K.E.Nietering,J.Y.Go,T.J. Miller,AdvancedFunctionalMaterials21(2011)3282–3290.
[29]W.C.Chueh,F.ElGabaly,J.D.Sugar,N.C.Bartelt,A.H.McDaniel,K.R.Fenton,K.R. Zavadil,T.Tyliszczak,W.Lai,K.F.McCarty,Nanoletters13(2013)866–872. [30]M.Cerbelaud,B.Lestriez,R.Ferrando,A.Videcoq,M.Richard-Plouet,M.T.