HAL Id: cea-01591636
https://hal-cea.archives-ouvertes.fr/cea-01591636
Submitted on 21 Sep 2017
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of
sci-entific research documents, whether they are
pub-lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
memories
Munique Kazar Mendes, Eugenie Martinez, A Marty, Marc Veillerot, Y.
Yamashita, Rémy Gassilloud, M Bernard, Olivier Renault, N. Barrett
To cite this version:
Munique Kazar Mendes, Eugenie Martinez, A Marty, Marc Veillerot, Y. Yamashita, et al.. Forming
mechanism of Te-based conductive-bridge memories. Applied Surface Science, Elsevier, In press,
�10.1016/j.apsusc.2017.07.187�. �cea-01591636�
Contents lists available atScienceDirect
Applied
Surface
Science
j o u r n a l h o m e p a g e :w w w . e l s e v i e r . c o m / l o c a t e / a p s u s c
Full
Length
Article
Forming
mechanism
of
Te-based
conductive-bridge
memories
M.
Kazar
Mendes
a,b,∗,
E.
Martinez
a,b,
A.
Marty
a,b,
M.
Veillerot
a,b,
Y.
Yamashita
c,
R.
Gassilloud
a,b,
M.
Bernard
a,b,
O.
Renault
a,b,
N.
Barrett
daUnivGrenobleAlpes,F-38000Grenoble,France bCEA,LETI,MINATECCampus,F-38054Grenoble,France
cNationalInstituteforMaterialsScience,1-1Namiki,Tsukuba,Ibaraki305-0044,Japan dSPEC,CEA,CNRS,UniversitéParis-Saclay,CEASaclay,91191Gif-sur-Yvette,France
a
r
t
i
c
l
e
i
n
f
o
Articlehistory: Received14March2017
Receivedinrevisedform18July2017 Accepted20July2017 Availableonlinexxx Keywords: RRAM CBRAM Oxygenscavenging Interfacechemistry HAXPES
a
b
s
t
r
a
c
t
WeinvestigatedoriginsoftheresistivitychangeduringtheformingofZrTe/Al2O3based
conductive-bridgeresistiverandomaccessmemories.Non-destructivehardX-rayphotoelectronspectroscopywas usedtoinvestigateredoxprocesseswithsufficientdepthsensitivity.Resultshighlightedthereduction ofaluminacorrelatedtotheoxidationofzirconiumattheinterfacebetweenthesolidelectrolyteandthe activeelectrode.InadditiontheresistanceswitchingcausedadecreaseofZr-Tebondsandanincrease ofelementalTeshowinganenrichmentoftelluriumattheZrTe/Al2O3interface.XPSdepthprofiling
usingargonclustersionbeamconfirmedtheoxygendiffusiontowardsthetopelectrode.Afour-layer capacitormodelshowedanincreaseofboththeZrO2andAlOxinterfaciallayers,confirmingtheredox
processlocatedattheZrTe/Al2O3interface.Oxygenvacanciescreatedinthealuminahelpthefilament
formationbyactingaspreferentialconductivepaths.Thisstudyprovidesafirstdirectevidenceofthe physico-chemicalphenomenainvolvedinresistiveswitchingofsuchdevices.
©2017ElsevierB.V.Allrightsreserved.
1. Introduction
Resistive random access memories (RRAM) are interesting candidates for the next generation of non-volatile memories (NVM).Datastorageprincipleisbasedonswitchingtheresistivity betweenhighandlowresistancestates(HRSandLRS,respectively). Amongthem,conducting-bridgeresistiverandomaccess memo-ries(CBRAMs) areone optioncurrently investigated,leadingto promisingadvanceddevicesintermsofindustrialdemonstrators. Theswitchingmechanismisbasedonapplyingvoltageorcurrent pulsestoasolidelectrolyte.Twotypesofelectrolytesarecurrently investigated:sulphides(GeS2[1],Cu2S[2],Ag2S[3],etc.)andoxides
(Ta2O5[4],SiO2[5],Al2O3[6],etc.).Ionicdiffusionfromtheactive
electrodeintotheelectrolytecreatesaconductivebridgebetween thetwoelectrodes.However,ifthegeneralmechanismisnowwell understood,moredetailedinformationabouttheelectrochemical processesandmaterialchangesinvolvedintheswitchingisneeded tooptimizedevicedesignandmaterialengineering.
∗ Correspondingauthorat:CEA, LETI,MINATEC Campus,F-38054Grenoble, France.
E-mailaddresses:Munique.KAZARMENDES@cea.fr,
muniquekazar@yahoo.com.br(M.K.Mendes).
Historically,Ag-andCu-basedCBRAMsarethemoststudied[7]. Theresistiveswitchingofthesedevicesisattributedtothe forma-tionanddisruptionofconductivepathsbetweenthetwoelectrodes [8,9]. By applying a voltageto thedevice, Ag+ (Cu+)ions from
theactiveelectrodemigrateintotheelectrolyte.Theyarereduced whenreachingthebottomelectrodeandthefilamentisformedby accumulationintheelectrolyte.Inthesameway,byreversingthe polarityoftheappliedvoltage,someofthecationshavereturned totheactiveelectrodeandthefilamentispartiallydestroyed.
Here, we investigated subquantum CBRAMs, operating with reducedcurrentsandthusverypromisingforlow-power applica-tions[10].ThecurrentrequiredtoprogramaCBRAMcellisdirectly relatedtotheconductanceofa1-atomfilament(G1atom).Instead
ofusingafilamentcontainingametal(Ag,Cu),thesenew mem-orycellsarebasedonthediffusionofTeinthesolidelectrolyte.For metalsG1atomiscomparabletothequantumG0=2e2/h,whereasfor
asemiconductorsuchasTe,itissub-quantum(0.03G0)enabling
operationatlowcurrents.ThedevicewasbasedontheuseofaZrTe alloyastheactiveelectrode,whichissupposedtoreleaseTe dur-ingelectricalbiasing.Theconductivebridgeofsuchmemorycells wasattributedtothemigration oftelluriumacrossthealumina [10].Jamesonatal.suggested thatapossiblemechanismforTe releasemightbetheoxidationofZr,yieldingtosub-stoichiometric alumina.Oxygenvacancies(VO)arethoughttocreate
preferen-http://dx.doi.org/10.1016/j.apsusc.2017.07.187
2 M.K.Mendesetal./AppliedSurfaceSciencexxx(2017)xxx–xxx
Fig. 1. Schematic of the setup for electrical characterization of the TaN/ZrTe/Al2O3/Tastack.
tialpathsforTemigrationacrossthealumina.Theseassumptions werebased ona detailedanalysisofelectricalcharacterizations butnodirectproofhasbeenreportedyetintheliterature.Herewe bringinformationabouttheelectrochemicalmechanismsinvolved inresistiveswitching.Inparticular,weconfirmedthekeyroleof theZrTe/Al2O3interfacechemistry.
Characterizing a device is challenging because of the small amountofnetchangeduetothefilamentformation.Moreover,the changesmostlikelyoccurinathinlayerburiedunderathickertop electrode.Thisrequiresnon-destructivecharacterizationmethods abletoprobethroughthickcappinglayerswithhighsensitivity.The filamentcompositionandmorphologyhasbeenstudiedinthepast usingtomographicatomicforcemicroscopy(T-AFM)[11], trans-missionelectronmicroscopy(TEM)[12],andX-rayphotoelectron spectroscopy(XPS)[13].XPSprofilinggivesdepthsensitive chem-icalinformation about cation diffusionand oxidation-reduction (redox)reactionsproducedduringtheformingprocess[14].The formingprocessisacrucialstepduringwhichmostofthe chemi-calandstructuralmodificationsoccurandwhichdefinesthefinal propertiesofthedevices.
Here, we have used hard X-ray photoelectron spectroscopy (HAXPES)toprobetheburiedinterfacebetweentheZrTeactive electrodeandtheAl2O3layer.Measurementswereperformedon
as-depositedsamplesandafterex-situforming,labeledas-grown andformed.UsinghardX-rays,theelectronmeanfreepathis con-siderablyincreasedprovidingsufficientdepthsensitivitytoreach theburiedinterfaceandthusprovideinformationabouttheredox processesoccurringthere.WehavealsousedXPSdepthprofiling toinvestigatetheoxygenmigration.
2. Experiment
Thememorycellswerecomprisedofanultrathin(5nm)Al2O3
layersandwichedbetweentwo metallicelectrodes. Thebottom electrodewasa200nm-thickTalayer,depositedona200mmSi (100)wafer,followedbythealuminadeposition.A15nm-thick ZrTemetallicalloywasdepositedthroughamask,toobtain2mm diameteractive,topelectrodesasshowninFig.1.Theseelectrodes andAl2O3layerweredepositedbyphysicalvapordeposition.The
ZrTegrowthprocesswasoptimizedtoreachaZr/Testoichiometry of40/60priortothedeposition.Thebasepressureinthe deposi-tionchamberwas5×10−4mbar.Finally,a5-nmthickTaNlayer wasdepositedbyreactivesputteringtopreventoxidationofZrTe whenexposedtoair.ThisTaNlayerisacappinglayertoreproduce theoperatingconditionsofanintegratedmemorystack.
FormingwasperformedinambientatmosphereusingaKeithley 2635B.TheTabottomelectrodewasgrounded.Alinearpositive voltagesweepbetween0and+4VwasappliedbyanAutipwith
minimumcontactforceontheTaN(seeFig.1)at0.1V/s.Standard electricalconnectionscannotbeusedhereduetotheultra-thin aluminalayerwhichcanbeeasilyshort-circuitedbymechanicalor thermalstress.Thecompliancecurrentof50mA(0.16mAcm−2) waschoseninordertoavoidpermanentbreakdownoftheoxide whilstinducingsignificantionicdiffusion,facilitatingthedetection [15].
TheHAXPESexperimentswereperformedattheBL15XUbeam lineoftheJapanSynchrotronRadiationResearchInstitute (SPring-8).Aphotonenergyof7.9keVwithanoverallenergyresolution (beamline and spectrometer) of 243meV was employed. The inelasticmeanfreepaths(),estimatedwiththeTanuma equa-tion[16]forAl1s,Zr3p3/2andTe3d3/2photoelectronsinthestack,
were9.2,10.7and10.4nm,respectively.Thesevaluesareobtained byaveragingtheIMFPsestimatedforeach layercrossedbythe photoelectronsduringtheirtransporttowardthesurface.The aver-agewasweightedbythethicknessofeachlayer.Corresponding samplingdepths(3)of27.6,32.1and31.2nmallowprobingthe ZrTeelectrodeandaluminalayer.Theareaprobedbythex-rays was300*25m2 andphotoelectrons emittedat anangleof 80◦
withrespecttothesurfacewerecollectedwithaVGScientaR4000 hemisphericalanalyzer,withaslitsizeof0.5*25mm2.Thebinding
energywascalibratedrelativetotheFermilevelmeasuredfroma cleangoldsurface.SubtractedbackgroundsforXPSspectrawereof Shirleytype[17],andtheoxidepeaksweremodeledusinga combi-nationofLorentzian(30%)andGaussian(70%)functionswhilethe metalliccorelineswerefittedusingaDoniach-Sunjicfunction[18] usingCasaXPSv2.3software.
LaboratoryXPSdepthprofilingwasperformedusingaPHI5000 VersaProbeII(Physical Electronics)equippedwitha monochro-maticAlK␣ source(h=1486.6eV). Thepassenergywassetto 47eV,givinganoverallenergyresolutionof0.75eV,withan emis-sionangleof45.0◦.Tocompensateforsamplecharging,inparticular whenanalyzingtheAl2O3layer,adualbeamchargeneutralizerwas
used.XPSdepthprofilingwascarriedoutusinganargongascluster ionbeam(GCIB)[19]with2500Aratomspercluster,acurrentof 20nA,arasterareaof2mm*2mm,ensuringuniformsputteringof theanalyzedarea,and20kVacceleratingvoltagecorrespondingto anenergyof8eVperatom.
3. Resultsanddiscussion
3.1. Electricalcharacterization
Fig.2(a)showstheI–VcharacteristicoftheTaN/ZrTe/Al2O3/Ta
stackmeasuredunderambientatmospherefortheformingstep. Thecurrentincreases abruptlyby threeordersofmagnitudeto the50mAcompliancecurrentathighthresholdvoltages.Thisis theformingprocess,duringwhich conductivepathsarecreated throughtheinsulatingdielectriclayer.Itistriggeredbytheforming voltage(Vf),equalto3.3V.Thecurrentincreaseatlowervoltages
isprobablyduetopreliminaryredoxordiffusionphenomena[20]. Thecurrentlimitationof50mA,orcompliancecurrent,was cho-seninordertogenerateimportantchemicalmodificationsinthe insulatinglayerorattheinterfaceswiththeelectrodes,while pro-tectingthestructurefromadielectricbreakdown[15].Thedevice resistancemeasuredbeforeand afterforminggave a resistance ratioRAs-grown/RFormed=1.105,confirmingasignificantchangeof
thedevicesresistivitybytheformationofaconductivepath(or paths)intheelectrolyte.Fig.2(b)presentsacurrent–voltage(I–V) curvefortheformingand reset processesmeasuredona simi-larmemorystack.In thiscurve,itispossibletoobserve,bythe applicationofoppositepolarity,theresetoperationswitchingthe memoryfromthelow-resistancestate(LRS)tothehigh-resistance state(HRS)at−1.4V.Inthisstudy,wefocusedourattentionon
Fig.2. (a)Current–voltage(I–V)curvefortheformingoftheTaN/ZrTe/Al2O3/Tastack.(b)Current–voltage(I–V)curvefortheformingandresetoftheTaN/ZrTe/Al2O3/Ta
stack. Table1
Relativeareas(%)oftheAl1scomponentsfortheAl2O3/Tastack,as-grownand
formedsamples.
Al2O3 AlOx
Al2O3/Ta 97.8±0.1 2.2±0.3
As-grown 91.3±0.1 8.7±0.3
Formed 88.0±0.1 12.0±0.3
thecriticalformingstepwhichhasakeyroleforthedevice oper-ation.HAXPESresultsobtainedforsimilardevicesmeasuredafter formingandresetarestillunderanalysisandwillbepublishedin afurtherpaper.
3.2. HardX-rayphotoelectronspectroscopy
HAXPES measurements were performed on as-grown and formedTaN/ZrTe/Al2O3/Tastacks,andalsoonthebareAl2O3/Ta
structureoutsidethetopelectrode(seeFig.1),inordertoprovide areferencespectrumforaluminum.Eachpeakwasnormalized rel-ativetotheintensityofthebackgroundmeasuredathighkinetic energy.Thephotoelectronrecoileffects[21]wereestimatedfrom Ekin(m/M)whereEkinisthephotoelectronkineticenergy,misthe
electronmassandMtheatomic mass.TheestimationforAl1s, Zr3p3/2andTe3d3/2corelevelsshowedarecoilenergyof0.13,
0.045and0.03eV,respectively.Takingintoaccounttheuncertainty of0.1eVonthebindingenergy,wehave neglectedtheserecoil energiesintherestofthepaper.
TheAl1scorelevelspectraoutsidethetopelectrode,i.e.forthe Al2O3/Tastack,andfortheas-grownandformedstatesare
repre-sentedinFig.3togetherwiththebestfits.Thedecompositionof theseAl1sspectratookintoaccounttwocontributions:aluminum sub-oxide(labeledAlOx)atbindingenergy(BE)of1561.5eVand
alumina(Al2O3)at1562.4eV.Therelativeareasofthese
compo-nentsaregiveninTable1.
Theseresultsshowedanincreaseofthesub-oxide contribu-tion(AlOxat1561.5eV)relativetothemaincontribution(Al2O3
at1562.4eV)fortheas-grownandformedstateswithrespectto theAl2O3/Tastack.Regardingtheas-grownstate,itmeansthatthe
ZrTetopelectrodedepositiongivesrisetoaslightreductionofthe underlyingaluminalayer.Thisphenomenonispresumablylocated atorneartheZrTe/Al2O3interfaceandrelatedtooxygen
scaveng-ingbythetopelectrode[10].Positivelybiasingthisactiveelectrode, oxygenmigration,probablyintheO2−form,isenhancedleadingto
afurtherreductionoftheAl2O3andanincreaseoftheAlOx
contri-bution(+3.3%).OxygenmigrationcreatespositivelychargedVOin
thealumina,whichmayhaveanimportantroleintheformationof conductivefilamentsbyactingaseasyionicdiffusionpathsinthe solidelectrolyte[20,22,23].
Fig.3.Al1scorelevelpeaksobtainedbyHAXPESon:a)theAl2O3/Tastack,b)
as-grownandc)formedstatesoftheTaN/ZrTe/Al2O3/Tastack.
Table2
Relativeareas(%)oftheZr3p3/2componentsfortheas-grownandformedsamples.
ZrTe ZrO2
As-grown 79.0±0.2 21.0±0.4
Formed 74.4±0.2 25.6±0.4
Fig.4presentstheZr3p3/2spectraforas-grownandformed
sam-ples.Thespectrahavebeenfittedconsideringtwocontributions. Thefirstcomponentat330.2eVwasfittedusingaDoniach-Sunjic [18] functionand it ischaracteristic of theZrTealloy.The sec-ondcomponentat332.9eVwasfittedusingaLorentzian-Gaussian functionanditischaracteristicofZrO2.
Table2showstherelativeareasofthetwocomponentsofthe Zr3p3/2spectrumforas-grownandformedsamples.Anincreaseof
theZrO2area(+4.6%)withrespecttothemainpeakwasobserved
afterforming,consistentwithoxygenmigrationtowardstheactive electroderesultinginoxidationofZrattheZrTe/Al2O3interface.
4 M.K.Mendesetal./AppliedSurfaceSciencexxx(2017)xxx–xxx
Fig.4.Zr3p3/2corelevelpeaksobtainedbyHAXPESon:a)as-grownandb)formed
statesoftheTaN/ZrTe/Al2O3/Tastack.
Fig.5. Te3d3 /2 corelevelspectraofa)as-grownandb)formedstatesofthe
TaN/ZrTe/Al2O3/Tastack.
TheproposedmechanismforZroxidationandTereleaseatthe interfaceisdescribedbyZrTex+O2→ZrO2+xTe[10].
TheoxidationofZristhoughttoallowTetransportunderan appliedfieldthroughtheoxidetocreatethefilament.TheHAXPES resultsprovideadirectevidenceofthechemistryofthefirststepof theproposedmechanism,i.e.Zroxidationbyoxygenscavenging fromthealuminaand,asaconsequence,Terelease.
TheTe3d3/2spectrameasuredforas-grownandformed
sam-plesareshowninFig.5.Theseasymmetricalpeakshavebeenfitted usingaDoniach-Sunjicfunction[18]byconsideringtwo contribu-tions.Thefirstcomponentat582.6eVisrelatedwiththeZrTealloy
Table3
Relativeareas(%)oftheTe3d3/2componentsfortheas-grownandformedsamples.
ZrTe Te-Te
As-grown 90.8±0.1 9.7±0.4
Formed 86.8±0.1 13.2±0.3
Fig.6.XPSdepthprofilesofoxygenmeasuredontheTiN/ZrTe/Al2O3/Tastackfor as-grownandformedsamples.
andthesecondcomponentat583.4eVischaracteristicofelemental Te[24,25].
Comparison between the relative areas extracted from as-grown and formed spectra is presented in Table3.This result showedasignificantelementaltelluriumincrease(+3.5%)during theformingprocess.ThisincreasewassimilartothatoftheZrO2
contribution.Thisresultcouldberelatedtothedetachmentand migration ofTe [10]or simplyformationofTe-Te bondsinside thetopelectrodeasaresultsofZrO2formationfollowingoxygen
uptakefromthealumina. 3.3. XPSdepthprofiling
XPS depth profiling were carried out on the as-grown and formedsamplestoprovideinformation aboutoxygenmigration alongthestack.Theanalysiswasperformedonaslightlydifferent stack,witha15nm-thickTiNcappinglayerinsteadofthethinner 5nm-thickTaNlayerandathinner(10nm)ZrTeelectrode.
Fig.6showstheXPSatomicconcentrationsputterdepthprofile ofoxygenobtainedfromtheO1scorelevelintensity,forthe as-grownandformedsamples.
Thehighoxygensignalbeforesputteringisduetosurface oxida-tionoftheTiNcapanddoesnotinfluencethesubsequentprofiling. TheonsetoftheburiedOintensityobtainedafterformingisshifted towardsthetopelectrodecomparedtotheonemeasuredforthe as-grownsample.ThisresultindicatesOdiffusionintotheZrTelayer whichisconsistentwiththescavengingmechanismsuggestedby J.R.Jamesonandal[10].andisconsistentwithwhatisobservedon theZr3p3/2corelevelspectra,i.e.adecreaseofZrTecomponent
andanincreaseofZrO2.
4. Discussion
InthecaseofTe-basedCBRAMs,suchasZrTe/Al2O3 devices,
ourresultsshowedthataredoxprocessoccursattheZrTe/Al2O3
interface(ZrO2formationandAl2O3reduction)yieldingtoa
sub-stoichiometricaluminium oxide.Thisbehaviour canalready be observedduringthedepositionofthetopelectrodeonthealumina. 8.7%ofAlwasinaloweroxidationstatecomparedtoonly2.2%in theas-depositedAl2O3.Thissuggestsapreliminaryredoxprocess
Fig.7. Four-layercapacitormodelforthe(a)as-grownand(b)formedsamples.
attheZrTe/Al2O3interfacecreatingVOinthealumina.Indeed,the
formationofZrO2isthermodynamicallymorefavourablein
com-parisonwithAl2O3formation,asshownbytheEllinghamdiagrams
[8].VOfacilitatethecreationoflowresistancepathspresumablyvia
Tediffusioninsidetheelectrolyte,asstatedbyJamesonetal.[10]. TheVOconcentrationdeducedfromtheAlOxcomponentincreased
afterformingwith12%Alinloweroxidationstate.Furthermorethe comparisonbetweentheoxygendepthprofilesmeasuredon as-grownandformedsampleshighlightedanoxygenmigrationfrom theelectrolyteintotheactiveelectrode.Electroforminghas there-foreenhancedtheOscavengingbytheZrTeelectrodeand,aswe willsee,increasedthereleaseofelementalTe.
Afour-layercapacitormodel[26,27]wasusedtoquantifythe redoxprocessattheZrTe/Al2O3 interface.Weassumedthattwo
distinctlayers ofZrO2 and AlOx wereformedwithsharp
inter-facesyieldingthestructureshowninFig.7.Wealsoassumedthat thereleasedTestaysattheinterface,andislocatedatthesame levelastheinterfacialZrO2 layer,labelledZrO2+Te-TeinFig.7.
Thisisequivalenttoassumingaphaseseparationwithinthelayer betweenZrO2andTe.
Tocalculatetherespectivethicknesses,weconsideredthe pho-toelectronintensitiesfromeachoftheAl2O3,AlOx,ZrO2andZrTe
layers,givenbytheareasofthecorrespondingcomponentsofthe Al1sandZr3p3/2corelevelspectra.Theseintensitiesweredefined
bythefollowingequations:
IZrTe=I∞ZrTe
1−exp −d ZrTe Zr (1)IZrO2 =I∞ZrO2exp
−d ZrTe Zr[1−exp −dZrO2 Zr
(2)
IAlOx=IAlO∞xexp
−dZrTe Al exp −d ZrO2 Al 1−exp −d AlOx Al (3) IAl2O3=I ∞ Al2O3exp −d ZrTe Al exp −d ZrO2 Al exp −dAlOx Al 1−exp −dAl2O3 Al(4)
where dZrTe, dZrO2, dAlOx and dAl2O3 are the thicknesses of the
ZrTe,ZrO2,AlOxandAl2O3layersofthemodel.Forsimplicity,we
assumedthattheinelasticmeanfreepathsofAl1sandZr3p3/2
photoelectronswerethesameinallthelayers ofthestack and respectivelyequaltoAl=9.2nmandZr=10.7nm.This
approx-imation leads to an uncertainty of ±1.3nm in the ZrO2 layer
estimation.
FromEqs.(1)and(2),wecanexpresstheintensityratioofthe ZrO2andZrTecontributionsextractedfromtheZr3p3/2corelevel:
IZrO2 I ZrTe =I ∞ ZrO2 I∞ZrTeexp
⎛
⎝
−dZrTe Zr⎞
⎠
1−exp − dZrO2 Zr⎡
⎣
1−exp⎛
⎝
dZrTe Zr⎞
⎠
⎤
⎦
FromEqs.(3)and(4),wecanexpresstheintensityratioofthe Al2O3andAlOxcontributionsextractedfromtheAl1scorelevel:
IAl2O3 IAlOx =I ∞ Al2O3 I∞AlOx exp
−dAlOx Al 1−exp−dAl2O3 Al 1−exp −dAlOx AlThese ratiosenable toestimatethethicknessof the interfa-cial AlOx and ZrO2 layers, given that dAlOx+ dAl2O3=5nm and
dZrO2+dZrTe=15nm.Wealsoassumedthattheintensitiesofthe
semi-infiniteblocksoftheAl2O3andAlOxlayerswereequal.The
sameassumptionwasmadefortheZrO2andZrTelayers.
ThethicknessesoftheZrO2andAlOxlayerscalculatedusingthis
modelfortheas-grownandformedsamplesaregiveninTable4 togetherwiththeintensityratiosoftheAl2O3,AlOx,ZrO2andZrTe
componentsinthecorelevelspectra.
Similarly,theintensityratiobetweentheTeandZrTe contri-butionsextractedfromtheTe3d3/2corelevelcanbeexpressedas
follows: ITe−Te I ZrTe = I∞Te−Te I∞ ZrTe exp
−d ZrTe Te 1−exp −d Te-Te Te 1−exp −d ZrTe TewheredTe-TeisthethicknessoftheTe-richlayerformednearthe
interfaceandTeisequalto10.4nm.Fromthislastequation,we
calculatedthethicknessoftheinterfacialTe-richlayertobe2.9 and4.0nm,fortheas-grownandformedsamplesrespectively(see
Table4).
Aftertheformingprocess,wesawanincreaseof1.1nmatthe interfacialZrO2 layer(seeFig.7(b)).TheTe-richlayershoweda
similargrowth.Thistrendagreeswiththeredoxprocessdeduced fromthedetailedanalysisoftheAl1s,Zr3p3/2andTe3d3/2peaks.
6 M.K.Mendesetal./AppliedSurfaceSciencexxx(2017)xxx–xxx Table4
ExperimentalintensityratiosandestimatedthicknessesoftheZrTe/ZrO2+Te-Te/AlOx/Al2O3stackextractedfromthequantitativemodelandtheAl1s,Zr3p3/2andTe3d3/2
core-levelcomponentsintensities.
Sample IAl2O3/IAlOx IZrO2/IZrTe ITe-Te/IZrTe dAlOx(nm) dZrO2(nm) dTe-Te(nm) dZrTe(nm)
As-grown 9.0 0.26 0.11 0.40 5.3 2.9 9.7
Formed 6.0 0.36 0.17 0.57 6.4 4.0 8.6
thealloy.NotethattheZrO2layerextendsovernearlyhalfofthe
topelectrode, showingthatoxygendiffusesdeeply insideZrTe. Thisthicknesswasprobablyoverestimateddue toanadditional ZrO2contributionontheZr3p3/2peakcomingfromZroxidation
attheupperTaN/ZrTeinterface.Indeed,theTaNcappinglayerdid notcompletelypreventoxidationofZrTe.TheTe-richlayeris thin-nerbyhalfcomparedtoZrO2.ThereleasedTeisconfinednearthe
ZrTe/Al2O3withapossibleextensioninsidethealumina.Asmall
increaseoftheinterfacialAlOxlayerwasalsoobtainedafterthe
voltageapplication,confirmingthecreationofVOinsidethe
alu-minalayer.Assumingthatthisisaninterfacialeffect,ourmodel showedthattheAlOxthicknessis0.17nmhigherafterformingbut
veryconfined(∼2monolayers).However,thisreductionprocess isprobablynotonlylocatedattheinterface,butextendsthrough thewholealuminalayer,toachievetheformationofpreferential conductivepaths.Therefore,theAlOxthicknessobtainedwithour
simplifiedmultilayermodelwasprobablyunderestimated. TheelectricalmeasurementsreportedbyJamesonetal. [10] showedthattheconductanceofsuchCBRAMscellsisequalto0.03 G0,belowthestandardG0value.Theswitchingmechanismisthus
relatedneithertoametallicfilamentnortoVO.Theconductance
ismoreconsistentwithaTe filament,suggesting thatthe pres-enceofTeinsidetheoxideisdirectlyrelatedtotheoxidationof Zr.OurstudyclearlyshowedthecorrelationofZroxidationwith thereleaseofTeattheZrTe/Al2O3 interface.However,itismore
difficulttoconcludeonthechemicalnatureofthefilament.Forthe sakeofconsistencywiththeelectricalresultsreportedin[10],we assumethatfilamentsaremadeofTe.TheHAXPESTe3d3/2
spec-trashowedthatthereleasedTeisnotionized.TheTemigration throughthealuminaisthusnotdirectlydrivenbytheelectricfield andmoredifficulttoexplain.Aplausiblemechanismisthat:a) firstTeaccumulatesattheinterfaceandthenprogressivelyinside thealumina,b)second,localTesegregationmightoccurthrough VO,actingaseasierdiffusionpathstowardsthebottomelectrode.
Thisstudyprovidesevidenceofthephysico-chemicalphenomena involvedduringtheelectroformingprocess.However,itcannotbe excludedthatotherconcurrentphysicalmechanismsgenerateor contributetotheresistiveswitching.
5. Conclusion
We investigated subquantum CBRAMs based on the ZrTe/Al2O3/Ta stack using HAXPES and XPS to learn about
electrochemical reactions and ionic transport involved in the electroformingprocess.HAXPESmeasurementsallowedin-depth analysesofthestackwithoutdegradingtheoriginalstructureby samplepreparation suchasion sputtering or chemicaletching. Thiswasofrealinteresttoavoiddamagingtheregionofinterest,in particulartheZrTe/Al2O3interface.Resultsshowedtheoxidation
ofZrtogetherwithacorrelatedaluminareductionduringforming. Thisredoxprocessisrelatedtooxygenscavengingbytheactive topelectrode.ThisOmigrationleadstothecreationofpositively chargedVO inthealumina,therebycreatingfavoredconductive
paths.XPS depth profiling confirmed the O diffusionfrom the Al2O3towardstheZrTetopelectrode.Afour-layercapacitormodel
showedanincreaseofboththeZrO2andAlOxinterfaciallayers,
confirmingtheredoxprocesslocatedbetweentheactiveelectrode
andtheelectrolyte.Furthermoretheresistanceswitchingcaused alsoanelementalTeincreaseattheinterfaceZrTe/Al2O3.
References
[1]E.Vianello,G.Molas,F.Longnos,P.Blaise,E.Souchier,C.Cagli,G.Palma,J. Guy,M.Bernard,M.Reyboz,G.Rodriguez,A.DeSalvo,F.Dahmani,P.Verrier, D.Bretegnier,J.Liebault,Sb-dopedGeS2asperformanceandreliability boosterinConductiveBridgeRAM,in:IEEE,2012,pp.31.5.1–31.5.4.
[2]N.Banno,T.Sakamoto,N.Iguchi,H.Sunamura,K.Terabe,T.Hasegawa,M. Aono,Diffusivityofcuionsinsolidelectrolyteanditseffectonthe performanceofnanometer-scaleswitch,IEEETrans.ElectronDevices55 (2008)3283–3287.
[3]K.Terabe,T.Hasegawa,T.Nakayama,M.Aono,Quantizedconductanceatomic switch,Nature433(2005)47–50.
[4]S.Maikap,S.Z.Rahaman,T.Y.Wu,F.Chen,M.-J.Kao,M.-J.Tsai,Lowcurrent(5 pA)resistiveswitchingmemoryusinghigh-kTa2O5solidelectrolyte,in:IEEE, 2009,pp.217–220.
[5]C.Schindler,S.C.P.Thermadam,R.Waser,M.N.Kozicki,Bipolarandunipolar resistiveswitchinginCu-dopedSiO2,IEEETrans.ElectronDevices54(2007) 2762–2768.
[6]L.Goux,K.Opsomer,R.Degraeve,R.Müller,C.Detavernier,D.J.Wouters,M. Jurczak,L.Altimime,J.A.Kittl,InfluenceoftheCu-Tecompositionand microstructureontheresistiveswitchingofCu-Te/Al2O3/Sicells,Appl.Phys. Lett.99(2011)053502.
[7]R.Waser,M.Aono,Nanoionics-basedresistiveswitchingmemories,Nat. Mater.6(2007)833–840.
[8]R.Waser,R.Dittmann,G.Staikov,K.Szot,Redox-basedresistiveswitching memories–nanoionicmechanismsprospects,andchallenges,Adv.Mater.21 (2009)2632–2663.
[9]S.Clima,K.Sankaran,Y.Y.Chen,A.Fantini,U.Celano,A.Belmonte,L.Zhang,L. Goux,B.Govoreanu,R.Degraeve,D.J.Wouters,M.Jurczak,W.Vandervorst, S.D.Gendt,G.Pourtois,RRAMsbasedonanionicandcationicswitching:a shortoverview,Phys.StatusSolidiRRLRapidRes.Lett.8(2014)501–511.
[10]J.R.Jameson,D.Kamalanathan,Subquantumconductive-bridgememory, Appl.Phys.Lett.108(2016)053505.
[11]U.Celano,L.Goux,A.Belmonte,A.Schulze,K.Opsomer,C.Detavernier,O. Richard,H.Bender,M.Jurczak,W.Vandervorst,Conductive-AFMtomography for3Dfilamentobservationinresistiveswitchingdevices,in:2013IEEEInt. ElectronDevicesMeet,2013,pp.21.6.1–21.6.4.
[12]Y.Yang,P.Gao,S.Gaba,T.Chang,X.Pan,W.Lu,Observationofconducting filamentgrowthinnanoscaleresistivememories,Nat.Commun.3(2012)732.
[13]W.Devulder,K.Opsomer,F.Seidel,A.Belmonte,R.Muller,B.DeSchutter,H. Bender,W.Vandervorst,S.VanElshocht,M.Jurczak,L.Goux,C.Detavernier, Influenceofcarbonalloyingonthethermalstabilityandresistiveswitching behaviorofcopper-telluridebasedCBRAMcells,ACSAppl.Mater.Interfaces5 (2013)6984–6989.
[14]S.Tappertzhofen,I.Valov,T.Tsuruoka,T.Hasegawa,R.Waser,M.Aono, Genericrelevanceofcounterchargesforcation-basednanoscaleresistive switchingmemories,ACSNano7(2013)6396–6402.
[15]D.Ielmini,Resistiveswitchingmemoriesbasedonmetaloxides:mechanisms, reliabilityandscaling,Semicond.Sci.Technol.31(2016)063002.
[16]QUASESsoftware(http://www.quases.com/products/quases-imfp-tpp2m/) usingtheTTP-2MformulafromS.Tanuma,C.J.,Powell,D.R.,Penn, Calculationsofelectroninelasticmeanfreepaths(IMFPS).IV.Evaluationof calculatedIMFPsandofthepredictiveIMFPformulaTPP-2forelectron energiesbetween50and2000eV,Surf.InterfaceAnal.20(1993)77–89. [17]D.Briggs,M.P.Seah(Eds.),PracticalSurfaceAnalysis,AugerandX-ray
PhotoelectronSpectroscopy,vol.1,edition,WileyChichester,NewYork,1996.
[18]S.Doniach,M.Sunjic,Many-electronsingularityinX-rayphotoemissionand X-raylinespectrafrommetals,J.Phys.CSolidStatePhys.3(1970)285.
[19]I.Yamada,J.Matsuo,N.Toyoda,A.Kirkpatrick,Materialsprocessingbygas clusterionbeams,Mater.Sci.Eng.RRep.34(2001)231–295.
[20]M.Saadi,P.Gonon,C.Vallée,C.Mannequin,H.Grampeix,E.Jalaguier,F. Jomni,A.Bsiesy,Onthemechanismsofcationinjectioninconductingbridge memories:thecaseofHfO2incontactwithnoblemetalanodes(AuCu,Ag),J. Appl.Phys.119(2016)114501.
[21]Y.Takata,Y.Kayanuma,M.Yabashi,K.Tamasaku,Y.Nishino,D.Miwa,Y. Harada,K.Horiba,S.Shin,S.Tanaka,E.Ikenaga,K.Kobayashi,Y.Senba,H. Ohashi,T.Ishikawa,Recoileffectsofphotoelectronsinasolid,Phys.Rev.B75 (2007)233404.
[22]Y.-L.Chung,W.-H.Cheng,J.-S.Jeng,W.-C.Chen,S.-A.Jhan,J.-S.Chen,Joint contributionsofAgionsandoxygenvacanciestoconductingfilament evolutionofAg/TaOx/Ptmemorydevice,J.Appl.Phys.116(2014)164502.
[23]G.Molas,E.Vianello,F.Dahmani,M.Barci,P.Blaise,J.Guy,A.Toffoli,M. Bernard,A.Roule,F.Pierre,C.Licitra,B.D.Salvo,L.Perniola,Controllingoxygen vacanciesindopedoxidebasedCBRAMforimprovedmemoryperformances, in:2014IEEEInt.ElectronDevicesMeet.,2014,pp.6.1.1-6.1.4.3.
[24]M.K.Bahl,R.L.Watson,K.J.Irgolic,X-rayphotoemissionstudiesoftellurium andsomeofitscompounds,J.Chem.Phys.66(1977)5526–5535.
[25]H.Neumann,M.Mast,J.Enderlein,R.D.Tomlinson,M.V.Yakushev,XPS analysisofbridgman-grownCuInTe2andofitsnativeoxide,Cryst.Res. Technol.31(1996)75–85.
[26]N.Barrett,O.Renault,J.-F.Damlencourt,F.Martin,Thermalstabilityofthe HfO2/SiO2interfaceforsub-0.1mcomplementary
metal-oxide-semiconductorgateoxidestacks:avalencebandand quantitativecore-levelstudybysoftx-rayphotoelectronspectroscopy,J. Appl.Phys.96(2004)6362–6369.
[27]A.Jablonski,J.Zemek,OverlayerthicknessdeterminationbyXPSusingthe multilineapproach,Surf.InterfaceAnal.41(2009)193–204.
