Electrodeposition
and
selenization
of
brass/tin/germanium
multilayers
for
Cu
2
Zn(Sn
1-x
Ge
x
)Se
4
thin
film
photovoltaic
devices
Kwinten
Clauwaert
a,b,
Maaike
Goossens
a,
Jessica
De
Wild
c,
Diego
Colombara
c,
Phillip
J.
Dale
c,
Koen
Binnemans
d,
Edward
Matthijs
a,*
,
Jan
Fransaer
b,*
aKULeuven,FacultyofEngineeringTechnology,ClusterSustainableChemicalProcessTechnology,TechnologyCampusGent,GebroedersDesmetstraat1, B-9000Gent,Belgium
b
KULeuven,DepartmentofMaterialsEngineering,KasteelparkArenberg44,P.O.Box2450,B-3001Leuven,Belgium c
PhysicsandMaterialsScienceResearchUnit,UniversityofLuxembourg,rueduBrill41,L-4422Belvaux,Luxembourg d
KULeuven,DepartmentofChemistry,Celestijnenlaan200F,P.O.Box2404,B-3001Leuven,Belgium
ARTICLE INFO Articlehistory:
Received27January2016
Receivedinrevisedform1March2016 Accepted8March2016
Availableonline14March2016 Keywords: Electrodeposition CZTGSe Germanium Kesterite Solarcell ABSTRACT
Dense,closed,homogeneousbrass/tin/germaniummultilayershavebeenpreparedbyelectrochemical methods. The stackswere electrodeposited on molybdenum-sputtered soda-lime glass. Brasswas deposited firstfrom anaqueouspyrophosphate electrolyte. Then, tin wasdepositedfrom atin(II) pyrophosphateelectrolyte.Germaniumwasdepositedasthetopfilmfrom5vol%GeCl4inpropylene
glycol.Thecompositionofthebrass/tin/germaniumstackcouldeasilybecontrolledbyadjustmentofthe chargeofthedepositionofeachcoatingandwithspecialattentiontoexchangereactions.Thevariationin filmthicknesswasminimizedbyusingarotatingdiskelectrode(RDE)sampleholderwithacurrentthief. Soft-annealingunderanitrogenatmosphereandselenizationwithelementalseleniumresultedindense, closedcopper-poorand zinc-richCZTGSeabsorbermaterial.P-typeconductivity wasconfirmed by photoelectrochemistryandafirstphotovoltaicdeviceshowedapowerconversionefficiencyof0.6%anda bandgapnear1.2eV(withGe/(Sn+Ge)=0.38).
ã2016ElsevierLtd.Allrightsreserved.
1.Introduction
Thin filmphotovoltaicsbased onCu(In,Ga)(S,Se)2 (CIGS)and
CdTehaverecentlyattainedpowerconversionefficienciesof22.3%
[1]and21.0%[2],respectively,andaredeployedonanindustrial scale.However,bothtypesofphotovoltaicdevicesfacelong-term sustainabilityissuesbecauseofthehealthrisksofcadmiumand thescarcityofindium,galliumandtellurium[3,4].Inrecentyears, Cu2ZnSn(S,Se)4 (CZT(S,Se)) thin films are investigated as an
alternative for CIGS because of their similar optoelectronic properties and more abundant and environmentally benign elements.The crystal structure of this material is obtainedby replacinghalfof theindiumandgallium atomsinCIGSbyzinc atomsandtheotherhalfbytinatoms,whicharebothmore earth-abundantmaterials [5,6]. The CZT(S,Se) compound is a p-type semiconductorwithalargeabsorptioncoefficientofover104cm1
and a band gap between0.95eV and 1.5eV depending onthe
sulfur-to-selenium ratio [7]. The record efficiency for a CZTSSe deviceof12.6%wasobtainedbyWangetal.byahydrazine-based spin-coatingprocess[8].However,sulfurandseleniumarevolatile species,socontrollingtheexactcompositions,thusexactbandgap, isdifficult.Alternatively,thebandgapofCZT(S,Se)canbetunedby (partial)replacementoftinbygermanium(Cu2Zn(Sn1-xGex)(S,Se)4
or CZTG(S,Se)). With a concentration in the Earth’s crust of approximately1.5ppm,germaniumisslightlymoreabundantthan tin[9].Becauseitisonlyminedasaby-productofmainlyzincores andcoal,itspriceissignificantlyhigherthanthatoftin[10,11].It mustbetakenintoaccountthatonlysmallamountsofgermanium (andtin)arerequiredfortheproductionofCZTGSe.Furthermore, researchontherecoveryofgermaniumfrom(electronic)devicesis performedin ordertoincrease theelementsavailabilityand to controlanddecreaseitsprice[12].Tinandgermaniumcanbothbe electrodepositedandaresignificantlylessvolatilethansulfurand selenium. Band gaptuning of kesterite type absorber layersis thereforepreferablyperformedbyalteringthetin-to-germanium ratiobecausethisratiocanbemoreeasilycontrolled,enablingthe productionofsemiconductorswithaspecificbandgap.Theband gapforasulfur-freedeviceistunablefrom0.95eV(Cu2ZnSnSe4)
[13]to1.6eV(Cu2ZnGeSe4)[14].Amaximumbandgapof2.2eVis
*Correspondingauthors.
E-mailaddresses:[email protected](E.Matthijs),
[email protected](J.Fransaer).
http://dx.doi.org/10.1016/j.electacta.2016.03.048
0013-4686/ã2016ElsevierLtd.Allrightsreserved.
ContentslistsavailableatScienceDirect
Electrochimica
Acta
obtainedina selenium-freeand tin-freeCu2ZnGeS4 device[15].
ThepresenceofmultivalenttininCZT(S,Se)isassociatedwithdeep leveltrapsintheabsorbermaterial.Germaniumcanalsooccurin the +II and +IV oxidation state, but the +II form is unlikely. Therefore,partialreplacementoftinbygermaniumcan theoreti-callyleadtoimprovedoptoelectronicpropertiesbythereduction ofdeepleveltraps [16].Furthermore,alloyingwithgermanium appearstoenhancetheperformanceofCZTSphotovoltaicdevices by increasing minority charge carrier lifetimes and reducing voltage-dependentchargecarriercollection.Hagesetal.produced CZTG(S,Se)deviceswithoutgermaniumandwithpartial replace-mentofthetinbygermanium.Replacementof30%ofthetinby germaniumresultedina1%absoluteimprovementofthelightto electricityconversionefficiencyanda50mVincreaseoftheopen circuitvoltage[17].CZTG(S,Se)thinfilmsarecommonlyprepared by nanocrystalink doctor-blading [17–19], evaporation [14,20], sputtering[21]andspin-coating[22].TherecordCZTGSSedevice waspreparedbyananocrystalink-annealingmethodandhada powerconversionefficiencyof 9.4%withGe/(Sn+Ge)=0.3 [17]. Thisnewrecordstimulatesresearchonproductiontechniquesthat are more environmentally benign and offers perspectives for applicationonalargescale.
Inthisview,electrodepositionisapromisingtechniquebecause itisalow-costmethodthatallowsfastandcontinuousproduction onlargesubstrates withgoodcontroloverthecompositionand morphology[23].Threeapproachesarecommonlyusedtodeposit theprecursorsforCZT(S,Se)thinfilmsandareinterestingforthe developmentofCZTG(S,Se)thinfilmsaswell:successive deposi-tionofelementalorbinaryalloycoatings,co-electrodepositionof the metals and co-electrodeposition of the metals and sulfur/ selenium.After deposition of these films, an annealing step is requiredtosulfurize/selenizetheprecursorsand/orimprovethe optoelectronicpropertiesofthefilms.
Although simultaneous electrodeposition of all metals and chalcogen co-electrodeposition require only a minimum of equipment and process steps, further deployment of these techniquesishinderedbythedifficultprocesscontrol.Addition ofgermaniumasafourthorfifthelementtobeincorporated,will only further impede the process control. Electrodeposition of stackedelementaland/orbinaryalloycoatingsovercomesthese issues.Thismethodimpliesseveraldepositionandrinsingsteps, thusrequiringahigherequipmentcost.However,this disadvan-tage does not exceed the benefits of the technique. The most important advantageis thehigh degree of control of the final compositionofthestackbychangingthethicknessofeachcoating
[24].Here,thevariationinfilmthickness,responsibleforin-depth inhomogeneity, was minimized by artificially enlarging the substratewithacurrentthiefthatsurroundedthemolybdenum substrate. Each deposition step can be optimized separately, enablingthedepositionofuniform,densefilmsonlargeareasfrom stable electrolytes and short deposition times [23]. For the electrodeposition of brass and tin from aqueous electrolytes, severalwell-describedelectrolytesthatareusedonanindustrial scale,areavailable[25].Onthecontrary,theelectrodepositionof germaniumfromaqueouselectrolytesisdifficultduetothelow hydrogenoverpotentialongermanium[26].Electrolytesbasedon organicsolvents(e.g.propyleneglycol)aretypicallyused[27]and morerecently,ionicliquidsareinvestigatedforthe electrodeposi-tionofgermaniumwiththeaimtocompletelyimpedehydrogen formation[28,29].
Inthispaper,dense,closed,homogeneousbrass/tin/germanium multilayerswerepreparedbyelectrochemicalmethods.Brassand tinweredepositedfromaqueouselectrolytes,whilegermanium wasdepositedfromapropyleneglycolelectrolyte.Thedeposition conditions of each coating were tuned with special focus on the process control. The brass/tin/germanium stacks were
soft-annealed undera nitrogenatmosphere and selenizedwith elemental seleniumin order toproduce CZTGSe thin films for photovoltaic devices. With Cu/(Zn+Sn+Ge), Zn/(Sn+Ge) and Ge/(Sn+Ge) atomic ratios after selenization of 0.85, 1.10 and 0.38, respectively,thetargetedratiosforthekesterite structure werereachedbythisnovelapproach.
2.Experimental 2.1.Electrolytes
Theelectrodepositionofbrasswasperformedfromanaqueous electrolyte containing 0.012moldm3copper(II) pyrophosphate (Cu2P2O7
3H2O, Sigma-Aldrich), 0.210moldm3 zinc(II) sulfate(ZnSO4
7H2O,AcrosOrganics>99.5%)and0.550moldm3potas-sium pyrophosphate (K4P2O7, Sigma-Aldrich 97%). The pH was
adjustedto10withortho-phosphoricacid(H3PO4,CarlRoth85%
pro analysis). After dissolution and adjustment of the pH, the electrolytewasdeaeratedwithnitrogen.
Tinwasdepositedfromanaqueoustin(II)electrolytecontaining 0.01moldm3tin(II)pyrophosphate(Sn2P2O7,Sigma-Aldrich98%)
and0.20moldm3potassiumpyrophosphate.ThepHwas adjust-edto10withortho-phosphoricacid. Afterwards,theelectrolyte wasdeaeratedwithnitrogen.
Germaniumwasdepositedfrom5vol%germanium(IV)chloride (GeCl4, Umicore, Olen (Belgium) 99.99 %) in propylene glycol
(C3H8O2,AcrosOrganics99%).Germanium(IV)chloridewasused
asreceived.Propyleneglycolwasdriedinavacuumof0.6mbarat room temperature for 24h. The electrolyte was prepared and handledinanargon-filledglovebox(oxygenandwatercontent below1ppm).
2.2.Electrodeposition
Prior to the electrodeposition of brass, the molybdenum substratewas cleaned byimmersionin ethanol (Chem-Lab) for 5min at 22C, rinsing with water, immersion in a 12.5 wt% ammonia solution (Merck) for 5min at 22C and rinsing with
wateragain.Aftereachelectrodepositionprocedure,theworking electrodewasrinsedwithwaterandethanolanddriedwithair (40C).Beforethedepositionoftin,itwasimmersedinethanol for5minat22Candrinsedwithwater.Beforethedepositionof germanium,thesamplewasimmersedinethanolfor5minat22C withoutsubsequentrinsingwithwater.
Chronoamperometricelectrodepositionsofbrassandtinwere performedina1000mLthree-electrodecell.Theelectrodeposition conditionsaresummarizedinTable1.Aninertcounterelectrode consistingofatitaniumgridcoatedwithRuO2/IrO2(dimensionally
stableanode,DSA)coveredthecompletebottomoftheelectrolysis cell. A Ag|AgCl (3M KCl) electrode (+0.210 V vs. SHE; Microelectrodes, Inc. MI-402) was used as the reference and wasplacedintheimmediatevicinityoftheworkingelectrode.In this manuscript, all potentials are expressed versus SHE. Molybdenum-sputteredSLGplates(GuardianEcoGuard1Mo)of 2020mm2 were used as working electrode. The working
Table1
Electrodepositionparametersforbrass,tinandgermaniumcoatings.
Coating Cu-Zn Sn(II) Ge*
T/C 22 22 22
E/Vvs.SHE 1.29 1.19 –
j/Adm2 – – 10
v/rpm 175 250 Mildstirring
*
electrode was mounted in a RDE system (Fig.1) with a CTV 101speedcontrolunit(Radiochemie,Analysis).Thevariationin filmthicknesswasminimizedbyartificiallyenlargingthesurface ofthesubstrateusingastainlesssteelcurrentthief(Ø=40mm) thatsurroundedthemolybdenumworkingelectrode.All electro-deswereconnectedtoaBio-LogicSP-200potentiostatequipped withEC-Lab1software(V10.23).
Germaniumwaschronopotentiometricallydepositedontopof thebrass/tinstack in a 100mLtwo-electrode cellat 22C. The
depositionparametersaresummarizedinTable1.The electrode-positionconditionswerebasedontheworkofSaitouetal.[27].In theirwork,theelectrolytewasnotdriedbeforeuseandwasnot agitatedduringdeposition.However,inourexperiments,better adheringfilmsweredepositedwhentheelectrolytewasdriedas described in section 2.1 “Chemicals” and when it was mildly agitatedwithamagneticstirbarduringdeposition.Theworking electrodewasmountedverticallyinthecell.Agraphitecounter electrode with a significantly larger area than the working electrodewasused.
2.3.Cyclicvoltammetryandlinearsweepvoltammetry
Cyclicvoltammetryofthebrasselectrolytewasperformedat 22C in 0.012moldm3 Cu2P2O7 and 0.550moldm3 K4P2O7,
0.210moldm3ZnSO4and0.550moldm3K4P2O7and0.012mol
dm3Cu2P2O7,0.210moldm3ZnSO4and0.550moldm3K4P2O7.
The pH of all electrolytes was adjusted to pH=10. Cyclic voltammetryofthetinelectrolytewasperformedin0.01moldm3 Sn2P2O7and0.20moldm3K4P2O7 (22C,pH=10).House-made
disk electrodes of molybdenum (Ø=3.0 mm) and copper (Ø=2.0 mm) embedded in a PVDF tip were used for cyclic voltammetryinthebrass andtin electrolytes, respectively.The electrodes were consecutively polished with 15
m
m and 3m
m diamond suspensions and a 0.05m
malumina suspension on a polishingcloth(BASi).Next,theelectrodeswererinsedwithwaterinanultrasonicbathinordertoremoveresidualparticles.Finally, thepretreatmentdescribedinsection2.2“Electrodeposition”was performed. A DSA counter electrode and a Ag|AgCl (3M KCl) reference electrode wereused. All cyclicvoltammograms were recordedatascanspeedof15mVs1.
Linearsweepvoltammetry(LSV) was performed inorder to study the electrochemical behavior of deposited brass filmsin 0.2moldm3K4P2O7 at 22C (pH=10). Prior toLSV, brasswas
platedonmolybdenum-sputteredSLGplatesfrom0.012moldm3 Cu2P2O7, 0.210moldm3 ZnSO4 and 0.550moldm3 K4P2O7
(pH=10) at 22C, a deposition potential of 1.29V, a rotation speed of 175rpm and a charge density of 3.5 104Cm2. The
brass-platedsubstratewasusedasworkingelectrodeforLSV.A DSAcounterelectrodeandaAg|AgCl(3MKCl)referenceelectrode completed the three-electrode cell. After immersion of the workingelectrode,thepotentialwassweptfrom1.2Vto0.5V atascanspeedof50mVs1.
2.4.Soft-annealing,selenizationandphotovoltaicdevicefabrication Thebrass/tin/germaniumstacksweresoft-annealedinatube furnaceunderanitrogenatmosphere.Afterplacingthesamplesin theoven,thetemperaturewasincreasedfrom22Cto350Cat 1Cs1and held at350Cfor 60min.Aftercoolingtoambient temperatureunderanitrogenatmosphere,theyweretransferred toagraphiteboxtogetherwithelementalseleniumpellets(99.999 +%).Theboxwasplacedinanitrogen-filledtubefurnace,heated from22Cto500Cat1Cs1,heldconstantfor40minandcooled to ambient temperature under a nitrogen atmosphere. After optimizationoftheconsecutiveetchingwithKCNandHCl(cfr.2.5
“Characterizationofthefilms”),thesampleswereprocessedinto complete photovoltaic devices with standard CdS/i-ZnO/Al:ZnO bufferlayersandNi/Al grid.Thedepositionoftheselayerswas performedasdescribedbyAidaetal.[30].
2.5.Characterizationofthefilms
Theglobalcompositionsweredeterminedbyanalysisof the copper,zinc,tinandgermaniumcontentbyinductivelycoupled plasma optical emission spectroscopy(ICP-OES, Varian 720 ES) afterdissolvingthesamplesin aquaregia.Local determinations andanalysisofthemorphologyweredonewithenergy-dispersive X-ray spectroscopy (EDX, EDAX Genesis 4000) combined with scanning electron microscopy (SEM, Philips XL 30 FEG). X-ray diffraction(XRD,Seifert3003)wasusedtocharacterizethecrystal structureofas-deposited,soft-annealedandselenizedfilms.
Photoelectrochemical characterization (PEC) of selenized sampleswereperformedwithaset-updescribedin[31]usinga 0.2moldm3 europium nitrate (Eu(NO3)3, Alfa Aesar 99.9%)
aqueouselectrolyteastheelectronscavengingspecies.Thefilm wassubjecttoseveralconsecutiveetchingtreatmentswith5wt% potassiumcyanide(KCN,Sigma-Aldrich98%)and5wt% hydro-chloricacid(HCl,Sigma-Aldrich36.5–38.0wt%solutioninwater) solutions. It must be noted that potassium cyanide and hydrochloric acid should never be used simultaneously in the samefumehoodasthiswillleadtoformationofhighlypoisonous hydrogencyanidegas.Chronoamperometricmeasurementswere performed under the flashed light of a 530nm LED at an illumination intensity equivalent to 2% of AM1.5, with the photoelectrodesheldat0.59VvsSHE.
Finishedphotovoltaicdeviceswerecharacterizedbycurrent– voltagemeasurementsusingafourpointprobeconfigurationwith an AM1.5 intensity-equivalentillumination (1000Wm2, 25C) calibratedwitha13.00.2%efficientmonocrystallinesiliconsolar cell(certifiedbyVLSIStandardsInc.).Externalquantumefficiency (EQE)spectrawererecordedbyilluminationofthephotovoltaic
Fig.1.Sampleholderfortheelectrodepositionofbrassonmolybdenum-sputtered SLG(2020mm2
deviceswithmonochromaticlightofvariablewavelengthoptically choppedat130Hz.Thephotocurrentwasmeasuredwithadigital lock-in amplifier (Stanford Research SR 850). The system was calibratedwithstandardizedsiliconandindium-gallium-arsenide photodiodes.
3.Resultsanddiscussion 3.1.Brasscoating
Inpreviouswork,weinvestigatedtheelectrochemicalbehavior of molybdenum in a 0.5moldm3 potassium pyrophosphate solutionatpH=10[32].Underthesecircumstances,molybdenum has the tendency to form a surface film of MoO2 [33], which
impedes good adhesion of the deposited metals. Therefore, a pretreatmentofthesubstratewasdeveloped.First,immersionin anammoniasolutionremovedMoO2fromthesurface[34].After
rinsingthesubstratewithwater,itwasimmersedintheplating bathandapotentialof0.11Vwasappliedfor15s,resultinginan anodiccurrent.Adherentbrassfilmscouldonlydepositedonthe molybdenumsubstrateafterperformingthispretreatment proce-dure.Withoutthepretreatment,partsofthebrassfilmspeeledoff fromthemolybdenumsubstrate.
BrasswasdepositedfromapyrophosphateelectrolyteatpH=10. ThispHwaschosenempiricallybecauseofthedifficultytodeposit dense,closedfilmswiththedesiredcompositionatlowerpHvalues. AthigherpHvalues,insolublespeciesdestabilizetheelectrolyteby theformation of precipitates. At pH=10, [Cu(P2O7)2]6 and [Zn
(P2O7)2]6arethedominantmetalpyrophosphatecomplexes[32].
Thesehighlynegativecomplexesdonotexistassuch,butassociate withpotassiumionsasKxMLn(4x2x)complexes[35].Thestandard
reductionpotentialsofthedominantcomplexeswerecalculatedby applicationofHess’slawandtherelationbetweentheGibbsfree energy, complex formation equilibrium constants (logðKCuðP
2O7Þ2
½ 6Þ=12.45 and logðK
ZnðP2O7Þ2
½ 6Þ=11.00 [36]) and standardreductionpotentialoftheuncomplexedmetal.Next,the reductionpotentialsof[Cu(P2O7)2]6and[Zn(P2O7)2]6underthe
circumstancespresentedinthiswork,werecalculatedbyusingthe Nernst equation and were found to be 0.02V and 1.05V, respectively.Becauseofthelargedifferencebetweenthereduction potentials, electrodeposition of high-quality brass was not expected.However,thereductionrateofcopper(II)iskinetically hinderedduetoadsorptionofpyrophosphateonthesurfaceofthe electrode[37–39].Johannsenet al.confirmed thisphenomenon duringelectrochemicalquartzmicrobalanceexperiments(EQCM) byobservingadecreaseinthecrystal'sfrequencyaround0.65V vs. SCEwithout measuringa current [39]. Theyattributed this observationtotheadsorptionofnon-electroactivepyrophosphate species.Becausepyrophosphateionsarenegativelycharged,the degreeofadsorptiondecreaseswithdecreasingpotential,enabling appreciable copper deposition currents at potentials that are rathernegativecomparedtothethermodynamicpotential.Cyclic voltammetry in a copper(II) pyrophosphate electrolyte on a molybdenumdiskworkingelectrodeconfirmedtheslowkinetics (Fig.2(Cu)).Atpotentialsmorenegativethan0.29V,abroadpeak wasrecordedduetotheslowreductionofcopper(II).Thecopper depositionrateincreasedat0.78V,henceappreciablecurrents wereonlyachievableatsignificantlymorenegativepotentialsthan the calculated 0.02V. Because of the kinetic hindrance, the differencebetweenthereductionpotentialsofcopper(II)andzinc (II) in the pyrophosphate electrolyte decreased, making this electrolyteasuitablebrassplatingbath.Afterreversingthescan direction,anucleationloopwasrecordedfollowedbythe cross-overpoint at 0.24V. Becausethe strippingof copperand the
oxidationofmolybdenumoverlapped,nocopperstrippingpeak wasrecorded.
Thecyclicvoltammogramofazinc(II)pyrophosphate electro-lyteisshowninFig.2(Zn).Zincwasdepositedatpotentialsmore negative than 1.15V and was accompanied by hydrogen gas formation.Anucleationloopwasrecordedafterreversingthescan direction. The cross-over point was reachedat 1.09V, which correspondswelltothecalculated1.06V.Thestrippingofthe zinc coating was recorded as two oxidation peaks. After the negative scan, the scan directionwas reversed at potentials at whichsignificanthydrogengasformationoccurred.Thereduction of water tohydrogengas is accompaniedby theproduction of hydroxideions.Becausetheelectrolytewasnotagitatedduring cyclicvoltammetry,thehydroxideionssignificantlyincreasedthe pHneartheelectrodesurfaceandcouldresultintheprecipitation of Zn(OH)2. During the reversed scan, the oxidation could be
disruptedbythepresenceofthisprecipitatedZn(OH)2layer,which
couldleadtotwooxidationpeaks.Itisalsopossiblethat,dueto hydrogen gas formation, the morphology of the zinc deposit differedsignificantlyfromthezincdepositedatpotentialsatwhich significantlyless hydrogen gasformation occurred, resulting in differentdissolutionkinetics.Asimilarbehaviorwasobservedby Syllaetal.fortheelectrodepositionofZn-Mnalloys[40].
Fig. 2(Cu-Zn) shows thecyclic voltammogram in thebinary electrolyte.Duringthenegativescan,threereductionpeakswere recorded.Thefirstwasabroadpeakstartingat0.16Vandwas relatedtotheslowreductionofcopper(II)inthepotentialregion wherepyrophosphateadsorptionoccurs.At0.77V,thecopper(II) reduction rate increased and a second reduction peak was observed. Thereductivecurrent increasedagainat 1.05Vdue tothesimultaneousreductionofzinc(II),followedbyhydrogengas formation at potentials more negative than 1.29V. After
Fig.2.Cyclicvoltammograms(scanspeed15mVs1)of0.012moldm3Cu2P2O7+ 0.550moldm3 K
4P2O7; 0.210moldm3 ZnSO4+0.550moldm3 K4P2O7 and 0.012moldm3 Cu
reversing the scan direction, the cross-over was reached at 0.75V.Thisvalueissituatedbetweenthecalculatedreduction potentials of copper(II) and zinc(II) (0.02V and 1.05V, respectively) which suggested the deposition of a copper-zinc alloyduringthenegativescan.Thecurrentdidnotfurtherincrease until0.11V,followedbystrippingofthealloy,oftheunderlying coppercoatingandformationofoxygengas.Afterremovingthe diskelectrodefromtheelectrolyte,aresidualdepositwasobserved ontheelectrodesurface,henceafractiondidnotdissolveduring thepositivescan.
Scraggetal.observedsignificantvariationsinfilmthickness overthesurfaceofelectrodepositedcopper/tin/zincstacks,leading to inhomogeneities on the micron to millimeter range after sulfurization[41]. Kurihara et al. reduced the variation in film thickness by performing electrodeposition experiments under hydrodynamiccontrolwithaRDE[42].Inourwork,a stainless steelcurrentthief,surroundingthemolybdenumelectrode,was addedtotheRDEinordertominimizevariationinfilmthickness. Withoutthecurrentthief,thethicknessofthebrassfilmsvaried fromapproximately800nmneartheedgesto450nminthecenter. Usingthecurrentthief,auniformbrassfilmwithathicknessof 64024nmwasobtained.Mirror-bright,yellowishcoatingswith 60at%copperweredeposited(Fig.3a).SEManalysisshowedthat fine-grained, dense, closed, adherent coatings were obtained (Fig. 3b,c). Furthermore, electrodeposition of brass coatings at differentcharge densities showed that the thicknessincreased linearlyasafunctionofthechargedensity(R2=0.98)andthatthe
composition of the coatings did not significantly vary with increasing thickness (copper content=57.50.7 at%) (Fig. 4). Thehighdegreeofcontroloverthecompositionandthicknessof thecoating,themicroscopicuniformityandgoodadhesiontothe molybdenumsubstratefacilitated the subsequentdepositionof othercoatings.
3.2.Tincoating
Tinwasdepositedontopofthebrasscoatingsfroma tin(II) pyrophosphateelectrolyte.Thereductionpotentialofthe domi-nant complex [Sn(P2O7)3]10 was calculated by the method in
section3.1“Brasscoating”(logðK½SnðP2O7Þ310Þ=18.40[43])andwas foundtobe0.66V.
Thecyclicvoltammogramofthetinplatingbathisshownin
Fig.5.Cyclicvoltammetryinthetin(II)pyrophosphateelectrolyte showedthattindepositionoccurredatpotentialsmorenegative than1.00V.Afterreversingthescandirection,thecross-overwas reached at 0.66V, which corresponded to the calculated potential.Next,theoxidationpeakofthetincoatingwasrecorded. Because of incomplete stripping (Qanodic/Qcathodic=37%), no
increaseoftheanodiccurrentduetotheoxidationoftheelectrode materialwasrecorded.AccordingtothePourbaixdiagram,tinhas thetendencytoformapassivefilmofSnO2atitssurfacebetween
pH1.0and12.5 [44].Thisfilmprotectedthesubstratematerial fromoxidizing.Afterremovalfromtheelectrolyte,asilverygray coatingwasobservedontheelectrode,whichdemonstratedthe incompletedissolutionofthetincoating.
Acopper/tin/zincstackingorderistypicallyusedfor electro-chemicalpreparationofCZTSsolarcellsbytheSELapproach.Other stackingordersoftenleadtometalexchangereactionsandpartial strippingofthecoatingduringconsecutiveplatingstepsbecauseof the relative positions of the standard reduction potentials of copper, tin and zinc [24,45]. In order toavoid metal exchange reactions,electrodepositionproceduresshouldbedevelopedthat enable the deposition of adherent, uniform coatings at rather negativepotentials.Ideally,theelectrolytesleavetheunderlying material unharmed. Therefore, the suitability of the tin(II) pyrophosphate was studied by performing anodic LSV on the brassdepositsintin(II)-freepyrophosphate electrolytes(Fig.6).
Fig.4.Thickness(*)andcompositionofthebrasscoating(&)asafunctionof depositionchargedensity.Depositionfrom0.012moldm3 Cu2P2O7, 0.210mol dm3 ZnSO
4 and0.550moldm3K4P2O7, Edep.=1.29Vvs.SHE,v=175 rpm, T=22C.
Fig.3. Image(a),cross-section(b)andtop-viewSEMimage(c)ofabrasscoatingon molybdenum-sputteredSLG(2020mm2
)obtainedfrom12mmoldm3Cu2P2O7, 210mmoldm3ZnSO
4and550mmoldm3K4P2O7(pH=10)at22C.Edep.=1.29V vs.SHE,v=175rpm,Q=2.5104Cm2.
Fig.5.Cyclicvoltammogram(scanspeed15mVs1)of0.01moldm3Sn2P2O7+ 0.20moldm3 K
Thecross-overwasreachedat0.99Vandfastdissolutionofthe brasscoatingoccurredatpotentialsmorepositivethan0.14V. Becausethereductionoftin(II)isthermodynamicallyexpectedat 0.66V,compositionalchangesofthebrasscoatingsbyexchange reactions afterimmersion in the tinelectrolyte wereexpected. As-depositedbrasscoatingswereimmersedinatin(II) pyrophos-phateelectrolytefor1to20minandthecompositionalchanges wereanalyzedwithICP-OES(Fig.7).Theamountofcopperinthe coating did not change by immersion in the electrolyte. The increasingtincontentanddecreasingzinccontentinthecoating demonstratedthe exchangereactionbetween zinc and tin. Sn2 ++Zn!Sn+Zn2+was the expectedexchange reaction, hence a
D
zinc-to-D
tinratioof1shouldbeobtained(withD
element=the massoftheelementinthecoatingafterimmersionsubtractedby themassoftheelementintheas-depositedcoating).Analysisof thecompositionshowedthatthisratiovariedbetween1.46and 4.34,hencethecoatingwas simultaneouslyaffectedbyanother reaction.Therefore,as-depositedbrasscoatingswereimmersedin a tin(II)-free pyrophosphate electrolyte for 1 to 10minutes. Between pH values of 8.5 and 10.5, zinc has the tendency to coveritselfwithafilmofZn(OH)2,whichisonlyslightlysoluble[46].However,theformationofthisfilmisinhibitedbyalloying zincwithanothermetal,allowing(slow)dissolutionofzincwith theformationofhydrogengas.Becauseofthehighsolubilityofthe [Zn(P2O7)2]6complex,thedissolutionofzincisfurtherpromoted.
Thisresultedinsignificantlossesofzincbyimmersionof brass coatingsinapyrophosphatesolutionatpH=10(Fig.7).Atshort immersiontimesintin(II)pyrophosphateelectrolytes,thelossof zincwaspredominantlycausedbythedissolutionreaction.Longer immersiontimes,thusmoreprofoundexchangereactionbetween solidzinc andSn2+ionsinsolution,enabled ahigherdegreeof
coverageoftheelectrodesurfacebysolidtinwhichinhibitedthe
chemical dissolution of zinc. This is demonstrated by the significantlysmallerlossofzincafter10minimmersioninatin (II)pyrophosphateelectrolytecomparedtothelossofzincafter 10minimmersioninatin(II)-freepyrophosphateelectrolyte.The exchangereactionwascompletelyinhibitedafterformationofatin coating of approximately 0.8
m
molcm2, i.e. approximately 130nm,ontheelectrodesurface.Inordertoavoidexchangereactionsandchemicaldissolution, thesubstratewascathodicallyprotectedbyusinghotentry,i.e.the deposition potential was applied before immersion of the substrateintheelectrolyte.AnalysiswithICP-OESofthecoatings depositedbythisprocedure showedthatnozinc and/orcopper lossesoccurred(Table2).
Theelectrodepositedtincoatingsshowedgoodadhesiontothe brasscoatingandhadalustrousgraycolor(Fig.8a).Adeposition potential of1.19Vanda rotationspeed of250rpmresultedin dense, closed coatingswithgood adhesiontothebrasscoating (Fig.8b,c).Thetincoatingthicknessincreasedlinearlyasafunction
Fig.6.Anodiclinear sweepvoltammogram(scanspeed50mVs1)ofa brass coatingonmolybdenum-sputteredSLGin0.2moldm3K4P2O7at22C(pH=10). Scanfrom1.2Vto0.5V.
Table2
Compositionofbrassbeforetheelectrodepositionoftin(Cu-Zn)andcompositionof brass/tinstacksafterelectrodepositionoftin(Sn(II))from0.01MSn2P2O7+0.20M K4P2O7,pH=10,T=22Conbrass-coatedmolybdenum-sputteredSLG(ICP-OES).
Electrolyte Cu-Zn Sn(II)*
Cu/at% 54 49(54)
Zn/at% 46 41(46)
Sn/at% – 10()
*
Thevaluesbetweenbracketsarethecopperandzincfractionswithouttaking thetincontentintoaccount.
Fig.7.Tinincrease(*)andzincdecrease(&)ofelectrodeposited brassasa function of time after immersion in 0.01M Sn2P2O7+0.20M K4P2O7, pH=10, T=22C(solidline)andzincdecrease(~)afterimmersionin0.20M K4P2O7, pH=10, T=22C (dashed line). Dmetal=nafter immersionnbefore immersion with n=theamountoftherespectivemetalinmmolcm2.AnalysiswithICP-OES.
Fig.8. Image(a),top-view(b)andcross-section(c)SEMimagesoftindepositedon abrasscoatingonmolybdenum-sputteredSLG(2020mm2).Depositionfrom 0.01moldm3Sn2P2O7and0.20moldm3K4P2O7(pH=10),Edep.=1.19Vvs.SHE,
ofthedepositionchargedensity(R2=0.98),hencethecomposition ofthebrass/tinstackcouldeasilybecontrolledbycontrollingthe depositionchargedensity(Fig.9).
3.3.Germaniumcoating
The electrodeposition of germanium from 5 vol% GeCl4 in
propyleneglycolwasperformedonmolybdenum,onbrass-coated molybdenum and on brass/tin-coated molybdenum. Due to excessive hydrogen gas formation during the electrodeposition of germanium on molybdenum, the molybdenum partially delaminatedfromtheSLG,hencegermaniumcouldnotbeused asbottomlayerinthisexperimentalset-up.Next,electrodeposited brasswithacopper-to-zincratioof1.7preparedbythemethodin section2.2“Electrodeposition”,wasusedasthebottomlayerand didnotsufferfromthis problemduringgermaniumdeposition. Mirrorbright germaniumcoatingsweredeposited onthebrass coating.Analysisofthecompositionofthebrass/germaniumstack
showed that the copper/zinc ratio shifted from 1.7 to 4.2–4.6, hencethegermaniumplatingbathaffectedthecompositionofthe brass coating. Hot entry of the sample did not prevent the dissolutionofbrass,impedingthecontrolofthecompositionofthe finalstack.Inthethirdstackingorder,germaniumwasdepositedas thetoplayeronabrass/tinstackresultinginmirrorbright,dense, closed coatings (Fig. 10). Analysis of the composition by EDX showedthatnosignificantcopper,zincortinlossesoccuredduring theelectrodepositionofgermanium(Table3).Becauseofthelow current efficiency (approximately1.6%), the germaniumcoating thicknessasafunctionofthedepositionchargedensitydeviated more from linearity (R2=0.88) compared tothe brass and tin
electrolytes(Fig.11).However,thecompositionofthebrass/tin/ germanium stack was still highly controlled by the deposition chargedensity.
3.4.Soft-annealingandselenization
Thermalpre-alloying,commonly referredtoas “soft-anneal-ing”,hasbeenshowntoimprovethehomogeneityand compact-ness of kesterite films after selenization [47]. Furthermore, intermixing the precursors results in the formation of stable structures.Therefore,thebrass/tin/germaniumstackswere soft-annealedunderanitrogenatmospherefor60minat350C.The
appearanceofthesampleschangedfromlustrousgraybefore soft-annealingtomattegrayaftersoft-annealing.XRDpatternsofeach stepinthedepositionprocessandofacompletestackafter soft-annealing(withathicknessofapproximately950nm)areshown inFig.12.TheidentificationofthepeaklabelsisshowninTable4. The compositions are shown in Table 5. In all patterns, peaks relatedtothemolybdenumsubstratewererecordedat40.5,58.8 and73.6.TheXRDpatternofthebrasscoatingwascharacterized
Fig.9.Cu/Zn(*)andCu/(Zn+Sn)(&)ofthecoatingafterelectrodepositionoftin (a)andthicknessofthetincoating(b)asafunctionofdepositionchargedensity. Depositionoftinfrom0.01moldm3Sn2P2O7and0.20moldm3K4P2O7(pH=10), Edep.=1.19Vvs.SHE,v=250rpm,T=22C.
Table3
Composition of brass/tin and brass/tin/germanium stacks on molybdenum-sputteredSLG(EDX).
BeforeGedeposition AfterGedeposition
Cu/at% 60 50(57)
Zn/at% 23 20(23)
Sn/at% 17 18(20)
Ge/at% 12()
Thevaluesbetweenbracketsarethecopper,zincandtinfractionswithouttaking thegermaniumcontentintoaccount.
Fig.10. Image(a)andtop-viewSEMimage(b)ofgermaniumdepositedonabrass/ tinstackonmolybdenum-sputteredSLG(2020mm2
).Depositionofgermanium from5vol%GeCl4inpropyleneglycolinaglovebox(waterandoxygenbelow 1ppm),j=10Adm2,mildstirringwithmagneticstirbar,T=22C.
by one additional peak at 42.6 (CuZn(110)). No peaks were attributedtoapurecopperorzincphase.Additionofatincoating tothestack,resultedinthedetectionofseveralpeaksrelatedto pure tin. A peak attributed to a copper-tin binary phase (Cu6.26Sn5(102)) was detected at 43.2. Copper-tin phases are
known to form at room temperature due to the substantial diffusion rate of copper in tin [48]. The XRD pattern of the (as-deposited)brass/tin/germaniumstackwascharacterizedbyan additionalsmallpeakat45.0,identifiedasGe(220).Althoughthis peakgreatlyoverlappedwiththeSn(211)peak,thepresenceof crystallinegermaniumisconfirmedbySaitouetal.,whodeposited germaniumfromasimilarelectrolyteanddetectedGe(220)asthe solegermaniumpeak[27].Aftersoft-annealingunderanitrogen atmospherefor60minat350C,asecondgermaniumpeak(Ge (111))wasdetectedat27.1.Thesoft-annealingalsoenabledthe formationofacopper-germaniumintermetalliccompound,which resulted in the detection of the Cu5Ge2(222) and Cu5Ge2(400)
planesat53.6 and62.8,respectively.Noelementalcopperand zinc were detected. Copper was only detected as copper-germanium, copper-tin and copper-zinc alloy phases and zinc onlyascopper-zincalloyphases.Thepresence ofelementaltin couldnot be excluded withcomplete certainty because alltin
peaksgreatlyoverlappedwithcopper-zincandcopper-tinpeaks. No peakscouldbeattributedtobinaryzinc-tin phasesbecause these elements do not form intermetallic compounds at these temperatures.
After soft-annealing, the coatings were selenized in the presence of elemental selenium at 500C for 40min. After selenization,a dense,adherentfilmofcloselypacked microme-ter-sized grains was obtained(Fig.13).The composition of the coatingsafter each stepof theprocess isshown inTable 5.As mentioned above,thecompositionof thecoatingsdidnotvary significantlyafterthedepositionofsuccessivecoatings.Although otherresearchgroupsreportedthelossofgermaniumand/ortin during selenization [17,18], the composition of the coatings presentedinourwork,werenotsignificantlyinfluencedby soft-annealingandselenizationofthebrass/tin/germaniumstacks.Itis likely that the soft-annealing enabled the formation of stable (alloy) phases, which significantly hindered the formation of volatilecompoundsundertheproposedselenizationconditions.
TheXRDpatternoftheselenizedfilmisshowninFig.14.Two peaks wererelatedtomolybdenum,i.e. at40.5 (Mo(110)) and
73.6(Mo(211)).Thepeaksat31.7,55.9and66.2wereattributed to MoSe2, i.e to MoSe2(101), MoSe2(110) and MoSe2(202),
respectively. In kesterite solar cells, molybdenum is typically chosen as the back contact because of its good electrical and structuralproperties.However,thickmolybdenumselenidelayers havesignificantlylessfavorableelectricalandstructuralproperties
[49],resultingindecreasedsolarcellperformances.Becauseofthe pronouncedmolybdenumselenidepeaksintheXRDpattern,itis expected that theefficiency of theCZTGSe solar cellwould be limitedbyitspresence.Threeothermainpeaksat27.4,45.5and 53.9wererelatedtothekesterite(112),(220)and(312)planes, respectively.Itmustbenotedthatthepeakpositionswereshifted towardshigher2
u
valuescomparedtogermanium-freeCZTSeand lowervaluescomparedtotin-freeCZGSe.Bypartiallyreplacingthe tin(IV)ionsinCZTSebysmallergermanium(IV)ions,thelattice constantsdecrease,thusshiftingthediffractionpeakstohigher angles. Some additional minor peaks could be attributed to secondaryphases: copperselenide,tinselenideandgermanium selenide.Furthermore,thepresenceofzincselenidecouldnotbe excludedbecauseitsdiffractionpeaksoverlapalmostentirelywith themainkesteritepeaks.Thepresenceofthesesecondaryphases, together with molybdenum selenide, are unfavorable for the production of high efficiency CZTGSe solar cells. From this structuralstudy,itwasconcludedthatgermaniumwas success-fully incorporatedin thekesterite structure (because of the2u
shift) and that further studies are required to minimize the presenceofsecondaryphasesafterselenization.3.5.Photoelectrochemicalcharacterization
ACZTGSefilmobtainedbyselenizationat520Cfor40minwas photoelectrochemically characterized in order to obtain
Fig.12.XRDpatternsofmolybdenum-sputteredSLG,brasscoating,brass/tinstack, brass/tin/germaniumstackand brass/tin/germaniumstackaftersoft-annealing (350Cfor60min).ThepeaklabelsareidentifiedinTable4.Thecompositionofthe coatingsaftereachstepintheproductionprocessisshowninTable5.
Table5
Compositionofas-depositedbrass,brass/tinandbrass/tin/germaniumstacksand brass/tin/germaniumstack afterselenization. TheCZTGSesamples were soft-annealedunderanitrogenatmosphereat350Cfor60minandselenizedwith elementalseleniumat500Cfor40min.
Atomicratio Cu-Zn Cu-Zn/Sn Cu-Zn/Sn/Ge CZTGSe Cu/Zn 1.550.02 1.490.04 Cu/(Zn+Sn) 0.960.02 1.020.02 Cu/(Zn+Sn+Ge) 0.900.02 0.900.03 Zn/(Sn+Ge) 1.420.03 1.340.06 Ge/(Sn+Ge) 0.270.02 0.300.02 Table4
IdentificationofthepeaksrecordedintheXRDpatternsinFig.12.
Label Crystalplane Label Crystalplane
information on its optoelectronic properties, namely on its majoritycarriertypeandabilitytogenerate,collectandtransport minority carriers under illumination to gain insights on its quantum efficiency. The effect of the removal of potentially detrimentalsecondaryphasesonthePECpropertiesofthefilmhas also been investigated. For this purpose, consecutive surface etchingtreatmentshavebeenperformedwithpotassiumcyanide and hydrochloric acid solutions, which are known to remove secondaryphasesfromthesurfaceoftheabsorbermaterial[50,51].
Fig.15ashowsthephotocurrenttransientsoftheas-deposited CZTGSe film (A) upon illumination with the 530nm LED. The negativesignofthephotocurrentindicatesthat electrons were
collectedatthesemiconductor/electrolyteinterface,whichimplies thattheCZTGSebehavedasap-typesemiconductor.Thetraces(B) and (C) correspond to the same film after 60s etching with potassiumcyanideandsubsequentetchingwithhydrochloricacid. The external quantum efficiencies recorded at 530nm (i.e. differencebetweentheequilibriumphotocurrentanddarkcurrent divided by the photon flux) after sequential etching with potassium cyanide and hydrochloric acid is shown in Fig.15b. Thephotocurrentwasdoubledbyetchingwithpotassiumcyanide for60s,reachinganEQE(530nm)of5%.Thisisaverycommon phenomenonforchalcogenidesemiconductorfilms,andislikely relatedtotheremovalofconductiveCu-Sephasesfromthesurface acting as preferential paths for recombination of the charge carriers [31]. Subsequent etching steps with hydrochloric acid reducedthephotocurrent,butmodifiedthephotocurrenttransient substantially (i.e. it became more square) and generated a seemingly more ideal surface. The process seemed partially reversible, as a further potassium cyanide etching after hydro-chloricacidtreatmentwasabletorestorethephotocurrenttoa slightlyhigherlevel.Thissuggeststhathydrochloricacidpossibly alsohadtheeffectofremovingsomedetrimentalphasesfromthe surfaceofCZTGSe.Furtheriterationofetchingstepsresultedina decreaseofthephotocurrent,suggestingthattheoptimaletching conditions consist of 30s hydrochloric acid followed by 60s potassium cyanide. Furthermore, the photocurrent onset of CZTGSe has been determined to be approximately 0.24V vs. SHE,whichissimilarforthatobservedforCZTSeandCIGSe[52], suggestingasimilarflatbandpotential.However,itneedstobe mentioned that bandedges of chalcogenide semiconductorsin solutionseem tobehighlydependentonthesurfaceoxidation state[53].
Fig.13.Top-viewSEM images ofa coatingafterselenization withelemental seleniumfor40minat500C;10,000(a)and20,000(b).Beforeselenization,the brass/tin/germaniumstacks weresoft-annealed for 60min at350C under a nitrogenatmosphere.
Fig.14.XRDpatternofaCZTGSecoatingonmolybdenum-sputteredSLGafter soft-annealing(350Cfor60minunderanitrogenatmosphere)andselenizationwith elementalseleniumat500Cfor40min.Peakidentificationisshowninthefigure.
Fig.15.ChronoamperometrictestsafterconsecutiveKCNandHCletchingstepsas describedin(b)ina0.2moldm3Eu3+
3.6.Photovoltaicdeviceperformance
Afterthephotoelectrochemicalcharacterizationand optimiza-tionoftheetchingprocedure,photovoltaicdeviceswithaCZTGSe/ CdS/i-ZnO/Al:ZnOconfigurationwithaNi/Algridwereprepared. Thej-Vcharacteristicsof thephotovoltaic devicewiththebest performanceareshowninFig.16.Aftercorrectionforsurfacearea and resistance of the electrical contact, it had a solar power conversionefficiencyof0.6%withashortcircuitcurrentdensity (jsc)of7.5mAcm2,anopencircuitvoltage(Voc)of287mV,afill
factor(FF)of28%,aseriesresistance(Rs)of30
V
cm2andaparallelresistance (Rp) of 51
V
cm2. The series resistance is at least1.5ordersofmagnitudelargerthanitsvalueinefficientdevices, considerably reducing theshort circuit current density and fill factor.
ThebandgapoftheCZTGSeabsorberfilmwascalculatedbytwo methods:determinationoftheinflectionpoint(i.e.themaximum
ofdEQE/d
l
)(Fig.17a)andbyalinearfitof(hn
ln(1-EQE))2asafunction of h
n
(Fig.17b).The inflection point was found tobe 1010nm,whichcorrespondstoabandgapof1.23eV.InFig.17b, the linear fit crosses the hn
-axis at 1.18eV (1050nm). Both methodsgiveabandgapcloseto1.2eV,whichisconsistentwith previously reported CZTGSe photovoltaic devices with Ge/(Sn+Ge) 0.40 with a band gap of 1.15eV [18]. The band gapshift,togetherwiththepeakshiftintheXRDpatterns,leadsto theconclusionthatgermaniumwaswellincludedinthecrystal structure of the absorber material and that the proposed electrodeposition-annealing route enabled the production of CZTGSe photovoltaic devices. A structural study of the (soft-) annealingconditionsandetchingprocessareexpectedtofurther improvetheperformanceofthedevices.4.Conclusions
Inthiswork,CZTGSeabsorbermaterialforphotovoltaicdevices was prepared byan electrodeposition-annealing approach. The electrochemical approach enabled the fabrication of uniform, dense, closed brass/tin/germanium stacks without the use of hazardouscompounds,suchashydrazine.Thecompositionofthe brasscoatingwasindependentofthechargedensity.Thethickness andthecompositionofthecompletestackcouldeasilybetunedby controlofthedepositionchargeofeachcoatingandwithspecial attentionforexchangereactions.Theselenizedcoatingsconsisted ofcloselypackedmicrometer-sizedgrainstogetherwithsmaller grains.XRDandphotocurrentspectroscopyshowedthat germani-umwassuccessfullyincorporatedinthekesteritestructure.The CZTGSephotovoltaicdevicewiththebestperformanceshoweda solarpowerconversionefficiencyof0.6%withabandgapnear 1.2eV, which is consistent with Ge/(Sn+Ge) 0.40 [18]. It is thereforeconcludedthattheproposed electrodeposition-anneal-ingapproachleadstothesuccessfulincorporationofgermaniumin the crystal structure of the absorber material, enabling the production of CZTGSe absorber material with the desired copper-poorandzinc-richcompositionforphotovoltaic devices. Infurtherstudiesoftheselenizationprocedure,thepresenceofthe secondaryphasescopperselenide,tinselenide,zincselenideand germaniumselenide,togetherwithamolybdenumselenidelayer betweenthekesteriteandthemolybdenumbackcontact,should beminimizedinordertoincreasetheefficiency.
Acknowledgements
The authors would like to thank KU Leuven, Technology CampusGentforfundingtheresearchofKwintenClauwaertby BACresources.DiegoColombaraandJessicaDeWildacknowledge thefinancialcontributionoftheLuxembourgNationalResearch Fundunder theGALDOCHS ProjectC14/MS/8302176 and EATSS Project C13/MS/5898466. Marc Meuris, Guy Brammertz and Sylvester Sahayaraj from IMEC are kindly acknowledged for selenizationofthesamples.Theauthorswouldalsoliketothank Umicore(Olen,Belgium)forthegenerousgiftofgermanium(IV) chloride.
References
[1]A.Colthorpe,SolarFrontier’srecordefficiency22.3%CIScellfaces‘global marketchallenge’,Internet http://www.pv-tech.org/news/solar-frontiers-record-efficiency-22.3-cis-cell-faces-global-market-challen,December8, 2015.
[2]M.A.Green,K.Emery,Y.Hishikawa,W.Warta,E.D.Dunlop,Solarcellefficiency tables(version46),Prog.Photovoltaics23(2015)805.
[3]W.D.Cyrs,H.J.Avens,Z.A.Capshaw,R.A.Kingsbury,J.Sahmel,B.E.Tvermoes, Landfillwasteandrecycling:useofascreening-levelriskassessmenttoolfor end-of-lifecadmiumtelluride(CdTe)thin-filmphotovoltaic(PV)panels, EnergyPolicy68(2014)524.
Fig.17. EQE(solidline),dEQE/dl(dashedline)andfitofdEQE/dl(a)and(hnln (1-EQE))2
as a function of hn (solid line) with linear fit (dashed line) (b). Measurementswereperformedonthe0.25cm2
CZTGSephotovoltaicdevicewith thebestperformance(selenizedfor40minat500C).
Fig. 16.Dark(dashedline)andilluminated(solidline)j-Vcharacteristicsofthebest performing0.25cm2
[4]G.Phipps,C.Mikolajczak,T.Guckes,Indiumandgallium:long-termsupply, Renew.EnergyFocus9(2008)56.
[5]M.Jiang,X.Yan,Cu2ZnSnS4thinfilmsolarcells:presentstatusandfuture prospects,in:A.Morales-Acevedo(Ed.),Solarcells—Researchandapplication perspectives,InTech,Rijeka(Croatia),2013,pp.107.
[6]M.P.Suryawanshi,G.L.Agawane,S.M.Bhosale,S.W.Shin,P.S.Patil,J.H.Kim,A.V. Moholkar,CZTSbasedthinfilmsolarcells:astatusreview,Mater.Technol.8 (2013)98.
[7]C.Persson,ElectronicandopticalpropertiesofCu2ZnSnS4andCu2ZnSnSe4,J. Appl.Phys.107(2010)053710.
[8]W.Wang,M.T.Winkler,O.Gunawan,T.Gokmen,T.K.Todorov,Y.Zhu,D.B. Mitzi,DevicecharacteristicsofCZTSSethin-filmsolarcellswith12.6% efficiency,Adv.Energ.Mat.4(2014)1301465.
[9]M.Frenzel,M.P.Ketris,J.Gutzmer,Onthegeologicalavailabilityofgermanium, MinerDeposita49(2014)471.
[10]D.E.Guberman,Mineralcommoditysummaries2016:germanium,United StatesGeologicalSurvey,2016,pp.70.
[11]C.S.Anderson,MineralCommoditySummaries2015:tin,UnitedStates GeologicalSurvey(2015)174.
[12]B.Robertz,J.Verhelle,M.Schurmans,Theprimaryandsecondaryproduction ofgermanium:alife-cycleassessmentofdifferentprocessalternatives, Jom-Us67(2015)412.
[13]F.Luckert,D.I.Hamilton,M.V.Yakushev,N.S.Beattie,G.Zoppi,M.Moynihan,I. Forbes,A.V.Karotki,A.V.Mudryi,M.Grossberg,J.Krustok,R.W.Martin,Optical propertiesofhighqualityCu2ZnSnSe4thinfilms,Appl.Phys.Lett.99(2011) 062104.
[14]H.Matsushita,T.Ochiai,A.Katsui,Preparationandcharacterizationof Cu2ZnGeSe4thinfilmsbyselenizationmethodusingtheCu-Zn-Geevaporated layerprecursors,J.Cryst.Growth275(2005)E995.
[15]Y.B.Zhang,X.D.Sun,P.H.Zhang,X.Yuan,F.Q.Huang,W.Q.Zhang,Structural propertiesandquasiparticlebandstructuresofCu-basedquaternary semiconductorsforphotovoltaicapplications,J.Appl.Phys. 111(2012)063709. [16]K.Biswas,S.Lany,A.Zunger,Theelectronicconsequencesofmultivalent
elementsininorganicsolarabsorbers:multivalencyofSninCu2ZnSnS4,Appl. Phys.Lett.96(2010)201902.
[17]C.J.Hages,S.Levcenco,C.K.Miskin,J.H.Alsmeier,D.Abou-Ras,R.G.Wilks,M. Bär,T.Unold,R.Agrawal,ImprovedperformanceofGe-alloyedCZTGeSSe thin-filmsolarcellsthroughcontrolofelementallosses,Prog.Photovolt.Res. Appl.23(2013)376.
[18]Q.Guo,G.M.Ford,W.-C.Yang,C.J.Hages,H.W.Hillhouse,R.Agrawal,Enhancing theperformanceofCZTSSesolarcellswithGealloying,Sol.Energ.Mat.Sol. Cells105(2012)132.
[19]G.M.Ford,Q.J.Guo,R.Agrawal,H.W.Hillhouse,EarthabundantelementCu2Zn (Sn1-xGex)S4nanocrystalsfortunablebandgapsolarcells:6.8%efficientdevice fabrication,Chem.Mater.23(2011)2626.
[20]M.Buffière,H.ElAnzeery,S.Oueslati,K.BenMessaoud,G.Brammertz,M. Meuris,J.Poortmans,PhysicalcharacterizationofCu2ZnGeSe4thinfilmsfrom annealingofCu-Zn-Geprecursorlayers,ThinSolidFilms582(2015)171. [21]S.Levcenko,R.Caballero,L.Dermenji,E.V.Telesh,I.A.Victorov,J.M.Merino,E.
Arushanov,M.Leon,I.V.Bodnar,Preparationandopticalcharacterizationof Cu2ZnGeSe4thinfilms,Opt.Mater.40(2015)76.
[22]S.Bag,O.Gunawan,T.Gokmen,Y.Zhu,D.B.Mitzi,Hydrazine-processed Ge-substitutedCZTSesolarcells,Chem.Mater.24(2012)4588.
[23]D.Colombara,A.Crossay,L.Vauche,S.Jaime,M.Arasimowicz,P.P.Grand,P.J. Dale,Electrodepositionofkesteritethinfilmsforphotovoltaicapplications: quovadis?Phys.StatusSolidiA212(2015)88.
[24]J.J.Scragg,D.M.Berg,P.J.Dale,A3.2%efficientkesteritedevicefrom electrodepositedstackedelementallayers,J.Electroanal.Chem.646(2010)52. [25]ModernElectroplating,in:M.Schlesinger,M.Paunovic(Eds.),5ed.,JohnWiley
&Sons,Hoboken(USA),2010.
[26]C.G.Fink,V.M.Dokras,Electrodepositionandelectrowinningofgermanium,J. Electrochem.Soc.95(1949)80.
[27]M.Saitou,K.Sakae,W.Oshikawa,Evaluationofcrystallinegermaniumthin filmselectrodepositedoncoppersubstratesfrompropyleneglycolelectrolyte, Surf.Coat.Technol.162(2003)101.
[28]M.X.Wu,N.R.Brooks,S.Schaltin,K.Binnemans,J.Fransaer,Electrodeposition ofgermaniumfromtheionicliquid1-butyl-1-methylpyrrolidinium dicyanamide,Phys.Chem.Chem.Phys.15(2013)4955.
[29]M.X.Wu,G.Vanhoutte,N.R.Brooks,K.Binnemans,J.Fransaer,
Electrodepositionofgermaniumatelevatedtemperaturesandpressuresfrom ionicliquids,Phys.Chem.Chem.Phys.17(2015)12080.
[30]Y.Aida,V.Depredurand,J.K.Larsen,H.Arai,D.Tanaka,M.Kurihara,S. Siebentritt,Cu-richCuInSe2solarcellswithaCu-poorsurface,Prog. Photovolt.:Res.Appl.23(2015)754.
[31]D.Colombara,A.Crossay,D.Regesch,C.Broussillou,T.G.deMonsabert,P.P. Grand,P.J.Dale,Predictionofphotovoltaicp-ndeviceshortcircuitcurrentby photoelectrochemicalanalysisofp-typeCIGSefilms,Electrochem.Commun. 48(2014)99.
[32]K.Clauwaert,K.Binnemans,E.Matthijs,J.Fransaer,Electrochemicalstudiesof theelectrodepositionofcopper-zinc-tinalloysfrompyrophosphate electrolytesfollowedbyselenizationforCZTSephotovoltaiccells,Electrochim. Acta188(2016)344.
[33]E.Deltombe,N.DeZoubov,M.Pourbaix,Molybdenum,in:M.Pourbaix(Ed.), Atlasofelectrochemicalequilibriainaqueoussolutions,NationalAssociation ofCorrosionEngineers,Houston(USA),1974,pp.272.
[34]P.P.Prosini,M.L.Addonizio,A.Antonaia,Effectofsubstratesurfacetreatment onthenucleationandcrystalgrowthofelectrodepositedcopperand copper-indiumalloys,ThinSolidFilms298(1997)191.
[35]O.E.SchuppIII,P.E.Sturrock,J.I.Watters,Astudyofthestabilityandbasicityof thecopper(II)pyrophosphatecomplexesusingthedroppingamalgam electrode,Inorg.Chem.2(1963)106.
[36]L.G.Sillen,A.E.Martell,Stabilityconstantsofmetal-ioncomplexes,Chemical Society,London(UK),1964.
[37]H.Konno,M.Nagayama,Mechanismofelectrodepositionofcopperfrom cupricpyrophosphatesolutions,Electrochim.Acta22(1977)353. [38]H.Konno,M.Nagayama,Theroleofpyrophosphateincopperdepositionfrom
cupricpyrophosphatesolutions,Electrochim.Acta23(1978)1001. [39]K.Johannsen,D.Page,S.Roy,Asystematicinvestigationofcurrentefficiency
duringbrassdepositionfromapyrophosphateelectrolyteusingRDERCE,and QCM,Electrochim.Acta45(2000)3691.
[40]D.Sylla,C.Savall,M.Gadouleau,C.Rebere,J.Creus,P.Refait,Electrodeposition ofZn-Mnalloysonsteelusinganalkalinepyrophosphate-basedelectrolytic bath,Surf.Coat.Tech.200(2005)2137.
[41]J.J.Scragg,P.J.Dale,L.M.Peter,SynthesisandcharacterizationofCu2ZnSnS4 absorberlayersbyanelectrodeposition-annealingroute,ThinSolidFilms517 (2009)2481.
[42]M.Kurihara,D.Berg,J.Fischer,S.Siebentritt,P.J.Dale,Kesteriteabsorberlayer uniformityfromelectrodepositedprecursors,Phys.StatusSolidiC6(2009) 1241.
[43]J.R.Duffield,D.R.Williams,I.Kron,Speciationstudiesofthesolubilityand aqueous-solutionchemistryoftin(II)pyrophosphateandtin(IV) pyrophosphatecomplexes,Polyhedron10(1991)377.
[44]E.Deltombe,N.DeZoubov,C.Vanleugenhaghe,M.Pourbaix,Tin,in:M. Pourbaix(Ed.),Atlasofelectrochemicalequilibriainaqueoussolutions, NationalAssociationofCorrosionEngineers,Houston(USA),1974,pp.475. [45]X.C.He,H.L.Shen,J.H.Pi,C.X.Zhang,Y.Hao,SynthesisofCu2ZnSnS4filmsfrom
sequentiallyelectrodepositedCu-Sn-Znprecursorsandtheirstructuraland opticalproperties,J.Mater.Sci.:Mater.Electron.24(2013)4578.
[46]N.DeZoubov,M.Pourbaix,Zinc,in:M.Pourbaix(Ed.),Atlasofelectrochemical equilibriainaqueoussolutions,NationalAssociationofCorrosionEngineers, Houston(USA),1974,pp.406.
[47]Y.Lin,S.Ikeda,W.Septina,Y.Kawasaki,T.Harada,M.Matsumura,Mechanistic aspectsofpreheatingeffectsofelectrodepositedmetallicprecursorson structuralandphotovoltaicpropertiesofCu2ZnSnS4thinfilms,Sol.Energ.Mat. Sol.Cells120(2014)218.
[48]H.C.Bhedwar,V.Balasubr,S.D.Kulkarni,K.K.Ray,Kirkendalleffectstudiesin copper-tindiffusioncouples,ScriptaMetall.Mater.6(1972)919.
[49]A.Jager-Waldau,M.C.Lux-Steiner,E. Bucher,L. Scandella,A.Schumacher,R. Prins, MoS2thin-filmspreparedbysulfurization,Appl.Surf.Sci.65–6(1993)465.
[50]M.Weber,R.Scheer,H.J.Lewerenz,H.Jungblut,U.Sturkel,Microroughness andcompositionofcyanide-treatedCuInS2,J.Electrochem.Soc.149(2002) G77.
[51]A.Fairbrother,E.Garcia-Hemme,V.Izquierdo-Roca,X.Fontane,F.A. Pulgarin-Agudelo,O.Vigil-Galan,A.Perez-Rodriguez,E.Saucedo, Developmentofaselectivechemicaletchtoimprovetheconversion efficiencyofZn-richCu2ZnSnS4solarcells,J.Am.Chem.Soc.134(2012) 8018.
[52]J.J.Scragg,P.J.Dale,L.M.Peter,G.Zoppi,I.Forbes,Newroutestosustainable photovoltaics:evaluationofCu2ZnSnS4asanalternativeabsorbermaterial, Phys.StatusSolidiB245(2008)1772.
