Silicon electrodeposition in molten fluorides

O

pen

A

rchive

T

oulouse

A

rchive

O

uverte (

OATAO

)

OATAO is an open access repository that collects the work of Toulouse researchers and

makes it freely available over the web where possible.

Any correspondence concerning this service should be sent to the repository administrator:

staff-oatao@inp-toulouse.fr

-toulouse.fr/

This is an author-deposited version published in:

www.aaa.comhttp://oatao.univ

Eprints ID: 8619

To link to this article:

DOI:10.1016/j.electacta.2011.12.039

http://dx.doi.org/10.1016/j.electacta.2011.12.039

To cite this version:

Bieber, Anne-Laure and Massot, Laurent and Gibilaro, Mathieu and Cassayre,

Laurent and Taxil, Pierre and Chamelot, Pierre Silicon electrodeposition in

molten fluorides. (2011) Electrochimica Acta, vol. 62 . pp. 282-289. ISSN

0013-4686

Silicon

electrodeposition

in

molten

fluorides

A.L.

Bieber, L.

Massot

_{, M.}

_{Gibilaro, L.}

_{Cassayre, P.}

_{Taxil, P.}

_Chamelot

UniversitédeToulouse;INPT,UPS,CNRS;LaboratoiredeGénieChimique;118RoutedeNarbonne,F-31062Toulouse,France

Keywords: Instantaneousnucleation Silicon Electrodeposition Moltenfluorides Chronoamperometry

a

b

s

t

r

a

c

t

SiliconnucleationprocesswasinvestigatedinmoltenNaF–KF(40–60mol%)onsilverelectrodesinthe 820–950◦_{Ctemperaturerangeinordertooptimizesiliconcoatingoperatingconditions.}

Chronoam-perometricmeasurementsevidencedthatsiliconelectrodepositionprocessinvolvedaninstantaneous nucleationwithdiffusion-controllednucleigrowthwhatevertemperatureandSi(IV)ionsconcentration inthemixture.Theoverpotentialandtemperatureinfluenceonnucleationsitesnumberwasalsostudied. Silicondepositswereobtainedusingthesametemperaturerangeasnucleationstudy,fordifferent currentdensitiesonsubstrates:Ni,Ag,CgraphiteandCvitreous.Asensitiveinfluenceofthecathodicsubstrate

onthedepositadherenceandroughnesswasobservedanddiscussed.

1. Introduction

Photovoltaictechnologyisemergingasamajorsourceof electri-calenergy[1],andthemostcommonbasematerialforphotovoltaic cell issolargradesilicon (SoG-Si).Duetoa constantlygrowing demandforSoG-Si,thedevelopmentofnewprocessesallowing theproductionofcheapSoG-Siisamainissueforthesolarenergy industry.

Inthe1980s,siliconelectrodepositioninmoltensaltshasbeen consideredasanattractiveoptionforSoG-Siproduction[2].Indeed, siliconmetalismorereductivethanhydrogenmeaningthatsilicon electrodepositionhastobecarriedoutinnon-aqueouselectrolytes suchasalkalimetalhalides.However,noelectrolyticSiproduction processhas foundcommercialapplicationuptonow. Consider-ing more recent concernsabout carbon-freeenergy production andgrowingneedsforSoG-Si,suchkindofprocessisnevertheless beingreassessed,and,forthatpurpose,basicinvestigationsonthe saltsproperties,Si(IV)reductionmechanismandsilicondeposition strategiesareneeded.

In the frameof a production route basedon electrorefining (eitherfromMG–SiorfromSicuttingremains[3]),itwasdecidedto focusontheelectrodepositionprocessofdissolvedSi(IV)inmolten fluoride.Themainobjectivewastoevaluatethefeasibilityofsuch reactionandtoprovideanacademicstudyoftheSi(IV)reduction reaction,thankstospecificelectrochemicaltechniques[4,5].Atthis

∗ Correspondingauthor.Tel.:+33561558194;fax:+33561556139.

E-mailaddress:massot@chimie.ups-tlse.fr(L.Massot).

pointof thestudy,thepurityof theobtaineddepositswasnot consideredasanissue.

Itwasshown,inapreviouswork[6],thatthevolatilityofSi(IV) compoundsgreatlydependsonthenatureofthefluoridemixture. TheNaF–KFeutecticmixture,whichhighlystabilizesSi(IV) com-pounds,wasthusselectedastheelectrolyteforthepresentstudy. Thearticledescribesanextstageinthepreparationofsilicon layersforsolarenergywhichisthedemonstrationofthe electrol-ysistechniquevalidityinmoltensaltstechniquetorecoverpure silicononacathode.

Forthispurpose,thenucleationprocessandthesiliconnuclei growthonsilverareinvestigated usingcyclicvoltammetryand chronoamperometryandtheseresultsarecomparedwithsuitable models[4,7]inordertostatewhetherthenucleationis instanta-neousorprogressive.Meanwhile,theovervoltageandtemperature influenceonthe nucleation mode and nucleation sitesnumber wereexamined. Then,silicondepositionrunswerecarriedwith variouselectrode substrates(Ni, Ag, Cgraphite andCvitreous),

cur-rentdensities (−20 to−200mAcm−2 _{range) and temperatures}

(820–900◦_{Crange).Theinfluenceofeachparameteronthe}

coat-ingadherenceandroughnesswasobservedbyscanningelectron microscopy(SEM).

2. Experimental

2.1. Cell

Thecellwasavitreouscarboncrucibleplacedinacylindrical vesselmadeofrefractorysteelandclosedbyastainlesssteellid cooledbycirculatingwater.Thedescriptionofthiscellhasbeen

detailedinpreviouswork[8].Theexperimentswereperformed underaninertargonatmosphere(Linde).Thecellwasheatedusing aprogrammablefurnaceandthetemperaturewasmeasuredusing achromel–alumelthermocouple.

The electrolytic bath consisted of the NaF–KF (40–60mol%) eutectic mixture (Carlo Erba 99.995%). As mentioned earlier, the choice of this specificsolvent was based on its stabilizing properties regarding to the volatility of Si(IV) compounds [6]. Themixturewasinitiallydehydratedbyheatingundervacuum (7×10−2_{mbar)fromambienttemperatureuptoitsmeltingpoint}

duringone week.Silicon ionswereintroducedintothebathin theformofsodiumhexafluorosilicateNa2SiF6(AlfaAesar99.99%)

powder.

2.2. Electrodes

Silver wires (1mm diameter), vitreous and graphite car-bon rods (3mm diameter–Mersen spectroscopic quality) and nickel plates (50mm×10mm×1mm) were used as working electrode. Thesurface areaoftheworkingelectrodewas deter-minedaftereachexperimentbymeasuringtheimmersiondepth in the bath. The auxiliary electrode was a silicon thick plate (4mm×8mm×11mm) with a large surface area (3cm2_{). The}

potentials were measured with reference to a silicon plate immersedinthemoltenelectrolyte.

2.3. Techniques

The electrochemical techniques used for the silicon nucle-ationand depositioninvestigationwerecyclicvoltammetryand chronoamperometry.Alltheelectrochemicalmeasurementswere performedwithanAutolabPGStat30potentiostat/galvanostat con-trolledbyacomputerusingtheresearchsoftwareGPES.

ThesilicondepositswereobservedbySEMcoupledwithEDX probe, aftercleaningina HCl–AlCl3–H2Omixtureat50◦Cwith

ultrasonicwaves.

3. Resultsanddiscussion

3.1. Si(IV)ionsreductionmechanism

Thesiliconionsreductioninfluoridesmediawasfoundtobea singlestepprocessexchanging4electrons(asalreadypresentedin Ref.[6]):

Si(IV)+4e−= Si (1)

Cyclic voltammetry was carried out on a silver electrode in NaF–KF–Na2SiF6 system in the 820–950◦C temperature

range. Fig. 1 presents a typical cyclic voltammogram of the NaF–KF–Na2SiF6 (0.24molkg−1)systemat850◦Consilicon

elec-trode.Thiscyclicvoltammogramexhibitsonlyonereductionpeak at−0.15VvsSianditsassociatedreoxidationpeakat0.15VvsSi. ThelinearrelationshipbetweenSi(IV)reductionpeakintensityand thesquarerootofthescanratehasbeenverified(seetheinsetof

Fig.1),provingthattheelectrochemicalreductionprocessis con-trolledbytheSi(IV)ionsdiffusioninthebath(Berzins–Delahay[9]

validforareversiblesoluble/insolublesystem): Ip=−0.61nFSc0



RT



v

where n is the number of exchanged electrons, F is the Fara-day’sconstant(Cmol−1_{),Sistheelectrodesurfacearea(m}2_),c0

isthesoluteconcentration(molm−3_{),Disthediffusioncoefficient}

(m2_s−1_{),Tisthetemperature(K)and}

_v

(Vs−1_).

Furthermore,thediffusioncoefficientDcanbecalculatedusing Eq.(2)and theresultsobtained fromtheinsetofFig.1.Atthe temperatureof850◦_{C,itsvalueis2.9×10}−9_m2_s−1_.The

temper-aturedependenceofthediffusioncoefficient,in the820–950◦_C

temperaturerange,isgiveninEq.(3): D=_1.13×10−5exp





Fig. 1. Cyclic voltammogram of NaF–KF–Na2SiF6 (c0=0.24molkg−1) at 850◦C; working electrode: Ag; auxiliary electrode: Si; reference electrode: Si; Scan

Fig.2. Cyclicvoltammogramhighlightingthenucleation cross-overforsilicondepositioninNaF–KF–Na2SiF6 (c0=0.24molkg−1)at850◦C;workingelectrode:Ag

(S=0.31cm2_{);auxiliaryelectrode:Si;referenceelectrode:Si;Scanrate=50mVs}−1_.

whereDisthediffusioncoefficient(m2_s−1_{)andTthetemperature}

(K).

3.2. Siliconelectrocrystallisationprocess 3.2.1. Cyclicvoltammetry

Fig.2showsacyclicvoltammogramwherethecathodiclimit potentialishigherthantheSi(IV)ionsreductionpeakpotential observedonFig.1.Thisfigurehighlightsthemainfeaturesofan electrocrystallisationprocess:

•_{Theobservationofacross-overofdirectandreversescanning} curves,whichistheconsequenceoftheirreversiblenucleation stepduringthereductionscan[9,10].

•_{Theasymmetricalshapeofthereoxidationpeak,typicalofthe} redissolutionofasolidphasedepositedduringthecathodicrun (strippingpeak).

•_{Thelowcathodicbackgroundcurrentobservedbetween0.02V/Si} and−0.03V/Sicanbeattributedtotheformationofdissolved siliconintothecathodicsubstrate(Ag).

3.2.2. Chronoamperometry

Accordingtopreviousworksinmoltensalts[11,12], chronoam-perometry is a suitable technique to investigate nucleation phenomena.Thetransientcurvesshape(I=f(t))atthebeginningof thecathodicpolarizationallowstodeterminethenucleiformation mode,growthandgeometry[7,13,14].Aseriesof chronoamper-ogramsarepresentedinFig.3,atdifferentovervoltagesonsilver electrode,inNaF–KF–Na2SiF6 (c0=0.24molkg−1),at850◦C.This

techniquewasnotfoundsuitablefornucleationstudiesoncarbon substrates,sincethiselementalloyswithsiliconatthefirststage oftheSielectrodepositionaccordingtothebinarydiagram[15].

Thecurvesaredividedintothreeparts(cf.insetofFig.3): •_{PartI:Duringthispart,correspondingtothedouble-layer}

charg-ing,thefirstnucleiformontheelectrode.Thetimedependence onthenumberofnucleigeneratedisgivenbythePoisson’slaw

[7]:

N=N₀(1−e−At) (4)

whereNisthenucleidensity (cm−2_),N

0 thetotal numberof

nucleationsites(cm−2_{),Athenucleationconstantforasite(s}−1₎

andtthetime(s).

-IfthevalueofAishigh,thusN=N0andthenucleationis

instan-taneousmeaningthatallthenucleiappearatthebeginningof thepolarization.

-ForlowvaluesofA,N=AN0t,meaningthatthenucleationis

progressive.

•_{PartII:Thecurrentrises,duetoanactivesurfaceareaincrease,} owingtothenucleigrowth.Correlatively,theelectrolyte diffu-sionpromotesthecurrentdecrease.Thecompetitionbetween thesetwoeffectsleadstoamaximumpeakintensityandits coor-dinates(tm;Im)areusedforthenucleationmodeling.Forshort

times,thecurrentinthissecondpartallowsthenucleationmode determinationandobeysthefollowingrelationaccordingto[7]:

j=_˛tx (5)

where˛andxdependonthegrowthcontrol,thenucleigeometry andthenucleationmode.Inthecaseofhemispherical3D nucle-ationlimitedbydiffusion,xisequalto1/2or3/2forinstantaneous orprogressivenucleationrespectively.Eq.(5)hasbeenplottedin

Fig.4andtheproportionalitybetweenjandt1/2_{evidencesthat}

aninstantaneoussiliconnucleationoccurs.

•_{Part III: Si(IV) ions diffusion control is predominant, and a} slowcurrentdecreaseisobservedfollowing theCottrell’s law

[8,9],whichimpliesa linearrelationshipbetweencurrentand

t−1/2_{.Scharifker and Hills [7] established a model based on a}

dimensionalintensityandtime:(I/Im)2=f(t/tm).Foran

instan-taneousnucleation,thecurveparametersaredescribedbythe Eqs.(6)–(9)as:





n

h



i

o





Fig.3.ChronoamperogramsonsilverelectrodeatvariousovervoltagesinNaF–KF–Na2SiF6(c0=0.24molkg−1)at850◦C;auxiliaryelectrode:Si;referenceelectrode:Si/Inset:

Chronoamperogramonsilverelectrodeat=−0.04VinNaF–KF–Na2SiF6(c0=0.24molkg−1)at850◦C.

I2

mtm=0.1629(nFc₀)2D (9)

wherenisthenumberofexchangedelectrons,FtheFaraday’s constant(Cmol−1_),c

0thesoluteconcentration(molm−3),Mthe

molarmass(kgmol−1_{),thedensity(kgm}−3_{)andDthediffusion}

coefficient(m2_s−1_).

Using thechronoamperogramsatdifferenttemperatures,the experimentalcurves(I/Im)2=f(t/tm)arecomparedtothetheoretical

onesinFig.5.

The experimental data are in agreement withEq. (6), prov-ing that the instantaneous nucleation mode is predominant at theconsideredtemperature,overvoltageandconcentrationrange. BasedonSEMobservations,similarconclusionswereobtainedby

Fig.4. Plottingsofj=f(t1/2_{)ofthechronoamperogramsecondpartatvariousovervoltagesinNaF–KF–Na}

2SiF6(c0=0.24molkg−1)at850◦C;workingelectrode:Ag;auxiliary

Fig.5. ComparisonofthedimensionlessexperimentaldataatvarioustemperatureswiththetheoreticalmodelsinNaF–KF–Na2SiF6;workingelectrode:Ag;auxiliary

electrode:Si;referenceelectrode:Si.

Carletonetal.forthesiliconnucleationmodeonvitreouscarbon inLiF–KF–K2SiF6at750◦C[16].TheSEMmicrographofnuclei,

pre-sentedinFig.6(t=1.1sand=−0.04VvsSi),seemstoconfirmthe instantaneousnucleationmodeandthehemisphericalgeometryof germs.

ThenucleationsitesnumberN0werecalculatedusingEqs.(6)

and(8)atvariousovervoltagesandtemperatures;thevaluesare gatheredinTable1.

Anincreaseinthenucleationsitesnumberwiththeovervoltage isobservedwhiletheproductIm2tmremainsconstant,asforeseen

in Eq.(9).AspreviouslydemonstratedinSection3.1,thenuclei growthislimitedbySi(IV)ionsdiffusion,which iseither hemi-sphericalorlinear.Inlineardiffusion,theslope˛(definedinEq.

Fig.6.SEMmicrographofsiliconnucleionsilverelectrodeinNaF–KF–Na2SiF6

(c0=0.24molkg−1)at850◦C;=−0.04VvsSi;t=1.1s.

Table1

OvervoltageinfluenceonthenucleifavourablesitesnumberinNaF–KF–Na2SiF6

(c0=0.24molkg−1)at850◦C;workingelectrode:Ag;auxiliaryelectrode:Si;

refer-enceelectrode:Si.

(V) im(Acm−2) tm(s) 10−2im2tm 10−11N0 −0.030 −0.098 1.55 1.49 2.10 −0.040 −0.158 0.680 1.70 4.79 −0.050 −0.236 0.298 1.66 10.9 −0.055 −0.274 0.218 1.64 14.9 −0.060 −0.316 0.167 1.66 19.5 −0.070 −0.393 0.116 1.79 28.1 −0.080 −0.469 0.076 1.67 42.8

(5))followsEq.(10)whereastheEq.(11)describesahemispherical diffusion: ˛=−8nFM2c₀3N₀D3/2−2−1/2 (10) a=−_nFN0(2Dc0)3/2





withtheonecalculatedfromtheslopesofj=f(t−1/2_{)(Fig.4)ineach}

differentdiffusionconditions,derivedfromEqs.(10)and(11). ForthevaluesmentionedinTable2,thiscomparisonseemsto statethatthenuclei’sgrowthisgovernedbyalineardiffusionand thusshouldleadtoadendriticgrowth.

Table3showsthetemperatureinfluenceforsiliconnucleation on a silver electrode at various overvoltages. It indicates that thenucleationsitesnumberN0increasessignificantlywithboth

Table2

Comparison of the nuclei favourable sites number in NaF–KF–Na2SiF6

(c0=0.24molkg−1)at850◦Cfordifferentdiffusionmodels;workingelectrode:Ag;

auxiliaryelectrode:Si;referenceelectrode:Si.

(V) 10−11_N 0 Lineardiffusion Eq.(10) Hemisphericaldiffusion Eq.(11) 10−11_N 0 10−8N0 −0.03 2.10 2.14 0.47 −0.04 4.79 4.44 0.98 −0.05 10.9 6.49 1.44

-0.08 -0.07 -0.06 -0.055 -0.05 -0.045 -0.04

η/V

Fig.7. LinearrelationshipbetweenthelogarithmofthenucleationfavourablesitesnumberandsquareovervoltageinverseinNaF–KF–Na2SiF6(c0=0.24molkg−1)at850◦C;

workingelectrode:Ag;auxiliaryelectrode:Si;referenceelectrode:Si.

temperature and overvoltage. In Fig. 7, a linear relationshipis evidencedbetweenln(N0)andtheinverseofsquareovervoltage

andobeysEq.(12)[13,17]: ln(N0)=A− B

2 (12)

Knowing thenucleation mode,Fig.7 allowsoptimized elec-trodepositionconditionstobeproposed.Indeed,toobtainasmooth coatingstructure,thecreationofahighnumberofnucleiis essen-tial,whichisfavouredbybothhightemperaturesandovervoltages. Neverthelessforhighovervoltage,atoohighelectrolytediffusion limitationcanpromotedendriticgrowthphenomena.Takinginto accounttheseconflictingeffects,itwasdecidedtorealisesilicon coatingsatdifferentovervoltagesandtemperatures.

3.3. Siliconelectrodeposition

Variousrunsofsiliconelectrodepositionondifferentcathodic substrates were performed taking into account the conditions suggested in the previous section. The electrolyses were con-ductedunderintentiostaticconditionsasmentionedinreferences

[16,18,19].

Thecurrentefficiencywasdeterminedongraphiteandsilver electrodesusingthemassdifferencebetweenthecathodebefore andaftertheelectrolysisrun.Itwasfoundtobebetween88and96% forlowcurrentdensities(-20and-100mAcm−2_{)resultingindense}

deposits;athighcurrentdensities,thedepositswereverydendritic andbrittle,whichcausedsiliconlossesduringthewashing treat-ment:thecurrentefficiencycouldnotbeevaluated.Thesecurrent

Table3

Temperatureinfluenceonthenucleifavourablesitesnumberatdifferent

over-voltagesinNaF–KF–Na2SiF6(c0=0.24molkg−1);workingelectrode:Ag;auxiliary

electrode:Si;referenceelectrode:Si.

(V) tm(s) 10−11N0

850◦_{C 900}◦_{C 850}◦_{C 900}◦_C

−0.03 1.55 0.568 1.46 5.15

−0.04 0.680 0.336 3.41 8.70

−0.05 0.298 0.203 12.4 14.41

efficienciesarehigherthanthoseobtainedbyElwellandRao[20]

inLiF-KF-K2SiF6at750◦C:50%asanaveragecurrentefficiencyand

80%for−25mAcm−2_{currentdensity.}

3.3.1. Operatingconditionsdependence

Several silicon coatings were obtained in NaF–KF–Na2SiF6

(c0=0.47molkg−1)inthe820–900◦Ctemperaturerangeoncarbon

rod,forappliedcurrentdensitiesbetween−20and−200mAcm−2_.

After frozen salt removal, the coating microstructures were observedbySEM.Themicrographs,presentedinFig.8,suggest firstthathighertemperaturesfavoursomewhatthesurface coat-ingregularitybutalsothathighcurrentdensitiesaredetrimental forthisproperty,whichisinagreementwithstudiesfromElwellin NaF–CaF2–SiO2[19],BoenandBouteilloninLiF–NaF–Na2SiF6[18]

andalsoforothermetalsdeposition,e.g.ontantaluminmolten flu-orides[13].Forapplicationswherehighplatingratesarerequired, theuseofpulsedcurrentcouldhelppreventingthediffusion limita-tionandthusleadtomoreregularcoatingsathighcurrentdensities

[21–23].

EDXanalysisofthedepositsdidnotrevealanyothercompound thanSithatmeansthatthedepositpurityisassumedtobeatleast 99.9%.AsforthepurityoftheobtainedSi,whichisakeyissuefor thesolarindustry,recentworkfromNohiraetal.[24,25]showed that,inthecaseofprocessesinvolvingmoltensaltssolventsathigh temperatures,acarefuldesignoftheelectrorefinerhastobemade inordertoavoidpollutionfromthecellelements.

3.3.2. Substrateinfluence

Duetothestrongeffectofthecurrentdensityontheregularity ofthecoatingsurfacestateandthemoderateinfluenceofthe tem-perature,electrodepositionrunswerecarriedoutat−20mAcm−2

at820◦_{Cfor5h.Fig.9showsSEMmicrographsofsilicondeposits}

onthreeelectrodesubstrates(Ni,AgandCv).

On nickel (a),silver substrate (b) and vitreouscarbon (c), a columnardendriticcoating,a siliconmicrowirestructure and a sponge-likestructurewereobservedrespectively.Thesmoothand densedepositobservedongraphitecarbonat820◦_C,−20mAcm−2

(Fig.8)isinagreementwithRaoetal.work[26]inLiF–KF–K2SiF6

at750◦_{C.AccordingtoCarletonetal.[16],thedifferencesbetween}

Fig.8. SEMobservationsofsiliconcoatingsoncarbonatvariouscurrentdensitiesandtemperatures.

properties leadingtosignificant differencesofnucleation active sites.

Fig.10presentsSEMobservationsofsiliconcoatingscross sec-tiononvarioussubstrates(Ni,Ag,Cand Cv)at−20mAcm−2 at

820◦_C.

•_{Nickel substrate (a): Ni–Si alloys layers are observed at the} substrate-coatingboundary,whichseemstopromoteaveryhigh adherenceofthesilicondepositonthesubstrate.SEM-EDX anal-ysis shows two silicide layers under thecolumnar layer and XRD measurement ontheNi–Si surface depositindicates the presence of Ni3Si2, NiSi and Si formation. Previousworks on

intermetallic compoundsNi–Si preparation byhigh tempera-turesoliddiffusionevidencedfournickelsilicides:Ni5Si,Ni2Si,

Ni3Si2,NiSi[27,28].Boriventetal.[27]showedthatthegrowth

ofNi3Si2proceedsbyneedles:itcouldexplainthatinthepresent

work,thesiliconcoatinggrowthoverNi3Si2needlesiscolumnar,

contrarytoothersubstrates.OurconclusionisthatNisubstrate isnotadaptedtosiliconelectrodepositionduetohigh interdiffu-sionratebetweenthecompoundswhichpromotestheformation ofintermetalliccompoundsinsteadofpureSi.

•_{Silverelectrode(b):noalloyformationisobserved,inaccordance} withtheAg–Siphasediagram[29].Thesiliconlayer,coveringthe Agsurface,becomesprogressivelymoreporousandfilamentous andtendstobreakduringthewashingtreatment.

•_{Graphitecarbon(c):adenseSiClayer(∼10mm)wasformedby} interdiffusionbetweensilicondepositandcarbonelectrode.On thisthinlayer,adenseandadherentsiliconlayerof∼100mm thicknesswasobtained.Similarresultsongraphitewereobtained byRaoetal.whoattributedthedepositcompactnesstothe coa-lescenceofsiliconnodulesinathree-dimensionalgrowth[30]. •_{Vitreouscarbon(d):aroughandpoorlyadherentdepositwas}

obtained.Thisbrittledepositwasmostprobablybrokenduring thehotpressingcarriedoutforthesampleembedding.

Fig.9.SEMobservationsofsilicondepositsurfacesat−20mAcm−2_at820◦

Fig.10.CrosssectionSEMobservationsofsilicondepositsat−20mAcm−2_at820◦_{Cfor5honvarioussubstrates:(a)nickel;(b)silver;(c)graphitecarbon;(d)vitreous}

carbon.

4. Conclusions

The silicon nucleation mechanism was studied in a NaF–KF–Na2SiF6salt mixture on silver electrode, using mainly

chronoamperometry: the silicon nuclei are generated instanta-neouslywithalineardiffusion-controlledhemisphericalgrowth. Moreover, the nucleation sites number N0 increases with the

temperatureandtheovervoltage.Then,galvanostaticelectrolyses wereperformedbyvaryingtemperatureandcurrentdensityon a carbon rod. SEM observations showedthat high temperature andlowcurrentdensityleadtoasmootherdeposit,withapurity (measuredbyEDX)higherthan99.9%.Thecathodicsubstratetests (Ni, Ag,Cand Cv)allowed concludingthatgraphitecarbon isa

well-adaptedcathodicmaterialduetothedepositadherenceand structure;onvitreouscarbon,thedepositsaredendritic;onsilver substrate,thedepositsarepoorlyadherentwhileonnickel,many alloyswereformed,hinderingpuresilicondeposition.Whenthe currentdensityincreases,thecoatingbecomesrougher. Athigh current densities, the silicon ions diffusion limitation strongly influencesthesilicongrowth,leadingtoamoredendriticdeposit. References

[1]M.A.Green,SolarEnergy76(2004)3.

[2]D.Elwell,R.S.Feigelson,SolarEnergyMater.6(1982)123.

[3]T.Ulset,S.Julsrud,L.Cassayre,P.Chamelot,L.Massot,P.Taxil,T.L.Naas,PCT

Int.Appl.WO2008156372(2008).

[4]P.Chamelot,P.Taxil,B.Lafage,Electrochim.Acta39(17)(1994)2571.

[5]P.Taxil,P.Chamelot,L.Massot,C.Hamel,J.Min.Metall.B39(1–2)(2003)

177.

[6]A.L.Bieber,L.Massot,M.Gibilaro,L.Cassayre,P.Chamelot,P.Taxil,Electrochim.

Acta56(2011)5022.

[7]B.Scharifker,G.Hills,Electrochim.Acta28(1983)879.

[8] L.Massot,P.Chamelot,F.Bouyer,P.Taxil,Electrochim.Acta47(2002)1949.

[9] A.J.Bard,L.R.Faulkner,ElectrochemicalMethods:Fundamentalsand

Applica-tions,JohnWileyandSons,NewYork,2001.

[10]X.H.Xu,C.L.Hussey,J.Electrochem.Soc.139(1992)1295.

[11] L.Massot,P.Chamelot,P.Palau,P.Taxil,Electrochim.Acta50(2005)5408.

[12] L.Massot,P.Chamelot,F.Bouyer,P.Taxil,Electrochim.Acta48(2003)465.

[13]P.Chamelot,B.Lafage,P.Taxil,J.Electrochem.Soc.143(5)(1996)1570.

[14]L. Legrand,A.Tranchand,R.Messina,J. Electrochem.Soc.141 (2)(1994)

378.

[15] R.W.Olesinski,G.J.Abbaschian,Bull.AlloyPhaseDiagrams5(1984)486.

[16]K.L.Carleton,J.M.Olson,A.Kibbler,J.Electrochem.Soc.130(4)(1983)782.

[17]T.Erdey-Gruz,M.Volmer,Z.Phys.Chem.157(1931)165.

[18]R.Boen,J.Bouteillon,J.Appl.Electrochem.13(1983)277.

[19]D.Elwell,SolarEnergyMater.5(1981)205.

[20]D.Elwell,G.M.Rao,J.Appl.Electrochem.18(1)(1988)15.

[21]N.Ibl,J.C.Puippe,H.Angerer,Surf.Technol.6(4)(1978)287.

[22]J.C.Puippe,Dissertation,ETH,Zürich,Switzerland(1978).

[23]P.Chamelot,P.Taxil,D.Oquab,J.Serp,B.Lafage,J.Electrochem.Soc.147(2000)

4131.

[24]K.Yasuda,T.Nohira,R.Hagiwara,Y.Ogata,Electrochim.Acta53(2007)106.

[25] K.Kobayashi,T.Nohira,R.Hagiwara,K.Ichitsubo,K.Yamada,ECSTrans.33(7)

(2010)239.

[26]G.M.Rao,D.Elwell,R.S.Feigelson,J.Electrochem.Soc.128(8)(1981)1708.

[27]D.Borivent,J.Paret,B.Billia,J.PhaseEquilib.Diff.27(6)(2006)561.

[28]K.N.Tu,G.Ottaviana,H.Föll,J.Appl.Phys.54(2)(1983)758.

[29]P.Y.Chevalier,Thermochim.Acta130(1988)33.