Achieving Novel Magnetic States in Oxides · bulk. The PCO films in tension are ferromagnetic, similar to the LaCoO3 system. Thus, epitaxial strain dominates the effects of chemical

AchievingNovelMagneticStatesinPerovskiteOxidesthroughHeteroepitaxyADissertationsubmittedtotheFacultyoftheGraduateSchooloftheUniversityof

California,Berkeley

ViratVasavMehta

InpartialfulfillmentoftherequirementsforthedegreeofDoctorofPhilosophyin

TheCollegeofEngineering‐

DepartmentofMaterialsScienceandEngineering

AndaDesignatedEmphasisIn

NanoscaleScienceandEngineering

IntheGraduateDivisionoftheUniversityofCalifornia,Berkeley

Committeeincharge:

ProfessorYuriSuzuki,ChairProfessorOscarDubonProfessorJoelMoore

Spring2012

AchievingNovelMagneticStatesinPerovskiteOxidesthroughHeteroepitaxy

Copyright©2012ByViratMehta

1

Abstract

AchievingNovelMagneticStatesinPerovskiteOxidesThroughHeteroepitaxy

ByViratVasavMehta

DoctorofPhilosophyinTheCollegeofEngineering‐

DepartmentofMaterialsScienceandEngineeringAndaDesignatedEmphasisin

NanoscaleScienceandEngineering

UniversityofCalifornia,BerkeleyProfessorYuriSuzuki,Chair

This dissertation is focused on controlling the spin state and long‐rangemagnetic order in cobaltites by heteroepitaxial thin‐film growth. I explore thegrowth of two different cobaltite materials, LaCoO3 and PrCoO3, on lattice‐mismatched substrates to determine the role of epitaxial strain in giving rise tolong‐rangemagnetic order.Thismagnetic order is not found in thebulk cobaltitematerial and warrants the detailed investigations carried out in this work. Iinvestigate changes in structure and stoichiometry that influence the electronicstructureandthelong‐rangemagneticorderinthesematerials. In theLaCoO3system, Iexplore thechanges instructure in the filmsundertensilestrainandcompressivestrainbygrowthonSrTiO3,LaSrAlTaO3,andLaAlO3substrates and film growth between 8 nm ‐133 nm thick. Substrate‐dependentoxygen vacancy ordering in the films is found using microstructuralcharacterization,presumablyrelatedtotheamountofstressineachofthefilms.Bycarrying out a study of the effects on the film structure from the oxygen growthpressure,Ifindanoverallincreaseintheout‐of‐planelatticeparameterwithloweroxygengrowthpressures. These structural and stoichiometry changes in the LaCoO3 films to trendsappear to be related to the stabilization of long‐range magnetic order. Highestmomentisfoundinthefilmsintension(whichalsohavethemostdefects)onSrTiO3andLaSrAlTaO3substratesandthelowestmomentisfoundinfilmsincompressiononLaAlO3.Element‐specificX‐rayabsorptiontechniquesrevealcontributionsfromCo in different spin and valence states. I show how strain affects the electronicstructureanddistributionofthesedifferentstatesandrelatetheseobservationstotrends observed in the magnetism. Strained films in tension have the highestamount of high spin Co3+and high spin Co2+, while relaxed films appear to havemostly low spin Co3+ at 25 K. I present some scenarios to explain how thesedifferentCoionscombinetogiverisetolong‐rangeferromagneticorderinLaCoO3films.

2

In thePrCoO3 system, I explorewhether long‐rangemagneticorder canbeobserved using heteroepitaxial synthesis similar to the efforts in the LaCoO3 thinfilmsystemdespitePrCoO3havingamorestablelowspinstateconfigurationinthebulk. The PCO films in tension are ferromagnetic, similar to the LaCoO3 system.Thus, epitaxial straindominates the effectsof chemicalpressurewhich stabilize alowspinstate.Thestrained filmshavemorehighspinCo3+.The implicationofCosublattice ordering on the ordering of the Pr sublattice is explored using X‐raymagnetic circular dichroism. A rare ordering of the Pr ions anti‐parallel to theorientation of themoments on the Co sublattice appears to occur in this system.These studies demonstrate the power of heteroepitaxial synthesis to give rise tonewmagneticfunctionalityinperovskiteoxidesystems.

i

ForKanubhai,Savitaben,Madhukantbhai,andKavitaben

ii

TableofContents

Chapter1:Designingnewroutestowardsnovelmagneticfunctionality.........................................1

1.1.Researchinmultifunctionalspintronicmaterials.........................................................................2

1.2.Developmentofnovelfunctionalityincomplexoxidesperovskites.....................................2

1.2.1.Physicalstructureofperovskites.................................................................................................3

1.2.2.Electronicstructureofperovskites.............................................................................................4

1.2.3.Magneticexchangeinperovskites...............................................................................................9

1.3.Investigationofcomplexoxidesusinghetero‐epitaxy.............................................................11

1.4.Exploitationofspinstatetransitionsincobaltites.....................................................................13

1.5.Goalsofthisdissertation.......................................................................................................................15

Chapter2:ExperimentalTechniques............................................................................................................17

2.1.SampleFabrication..................................................................................................................................18

2.1.1.Pulsed‐LaserDeposition(PLD)..................................................................................................18

2.2.SampleCharacterization.......................................................................................................................23

2.2.1.AtomicForceMicroscopy(AFM)...............................................................................................23

2.2.2.X‐rayDiffraction(XRD).................................................................................................................24

2.2.3.RutherfordBackscatteringSpectrometry(RBS)................................................................25

2.2.4.Magnetometry...................................................................................................................................26

2.2.5.X‐rayAbsorptionSpectroscopy(XAS)....................................................................................27

2.2.6.ScanningTunnelingElectronMicroscopy/ElectronEnergyLossSpectroscopy(STEM/EELS).................................................................................................................................................31

Chapter3:StructuralEffectsfromEpitaxy.................................................................................................34

3.1.GrowthandstructuralcharacterizationofepitaxiallystrainedLaCoO3films................35

3.2.Stabilizingtetragonally‐distortedLaCoO3filmsusingcoherentepitaxialstrain...........35

3.3.OxygenvacancyorderinginSTEM/EELS.......................................................................................40

3.4.OxygenDependencestudies................................................................................................................43

3.5.Summary......................................................................................................................................................45

Chapter4:InducingferromagneticexchangeinepitaxialLaCoO3....................................................47

4.1.Substrateandstraindependencestudies......................................................................................48

4.2.Micro‐structure/Electronicstructure............................................................................................50

iii

4.2.1.STEM/EELS........................................................................................................................................51

4.2.2.SpectroscopicCharacterization.................................................................................................52

4.3.Oxygengrowthpressureeffectsonferromagnetism................................................................57

4.4.CospinstatesandvalencestatesinLaCoO3films......................................................................59

4.5.Theongoingexplorationoftheferromagneticexchangemechanism...............................62

Chapter5:ExtensionstoPr‐basedCobaltite:inducingA‐siteordering.........................................64

5.1.Motivationforexploringothercobaltites......................................................................................65

5.2.GrowthandstructuralcharacterizationofepitaxiallystrainedPrCoO3films................66

5.3.Exploringlong‐rangemagneticorderinPrCoO3filmsonSrTiO3.........................................68

5.3.1.Magnetismandepitaxialstrain..................................................................................................68

5.3.2.Element‐specificmagneticorder...............................................................................................69

5.4.NovelandemergentPr‐sublatticeordering..................................................................................72

5.5.Conclusions.................................................................................................................................................74

DissertationSummaryandOutlook...............................................................................................................75

References................................................................................................................................................................76

iv

Acknowledgments First and foremost, I would like to thank my advisor, Professor Yuri Suzuki, formakingtheworkinthisdissertationpossible.Withouttheincrediblenumberoftoolsandresources she has made available to me in the lab, none of this work could have beenaccomplished.Shehasalsoprovidedaccess toanesteemednetworkof collaborators thatwereeasytoreachouttoinordertohelpanswersomeofthemorepressingquestionsinthis research and push new projects forward. She has created a generous space forindependent exploration of new materials and exciting new research ideas in herlaboratory, while still providing a sense of direction and encouragement to the aspiringresearcher. IwouldalsoliketothanktheU.S.DepartmentofEnergyforgraciouslyfundingmywork throughout my tenure at Berkeley. Aside from a nanotech researcher fellowship Ireceivedinmyfirsttwotermshere,thefundingprovidedbytheU.S.DepartmentofEnergyunder theOfficeofBasicEnergySciences,MaterialsSciencesDivision, andScientificUserFacilitiesDivisionhavebeencriticaltothecompletionofthiswork. I would also like to show my appreciation to the members of my dissertationcommittee for theirpatienceanddiligence inprovidingmewithvaluable timelyguidancewiththisdissertation. IamalsogratefultothenumerouscollaboratorsIhavehadthepleasureofworkingwith throughout my years at Berkeley. Even in the instances that didn’t result in apublication,thetechniquesIlearnedandthevariousapproachesIencounteredhavehelptomakememorewell‐roundedandadeptatexperimentalwork.Iespeciallywouldliketothethank: Dr. Jeff Kortright at Lawrence Berkeley National Laboratory and Dr. Frank (Bud)BridgesatUCSantaCruzfortheircollaborationsthattaughtmeagreatdealaboutbeamlinescienceandthenecessarycareneededwhenanalyzingbeamlinedata,Dr.ElkeArneholzattheAdvancedLightSourcewhohasworkedwithmethroughouttheyearsandalwaysbeensupportiveofourlab,Dr.MariaVarelawhohastaughtmeagreatdealaboutthevaluesandlimitsofelectronmicroscopy,andProfessorChrisLeightonattheUniversityofMinnesotawhohasservedasaknowledgeableresourceonallsciencerelatedtothecobaltites. Iwouldalsoliketothemanyformerlabmembersthathavemoldedmyhabitsandtaughtmethebestwaystoapproachresearch.IespeciallywanttothankRajeshChopdekarand Marco Liberati who served as mentors and role‐models throughout my graduateexperience. They taughtme the importance of beingunabashedwhen it comes to askingquestions and to take a participatory role in the operation and upkeep of laboratoryequipment.ThehandsonapproachtoresearchthattheypushedisincrediblyvaluableforanexperimentalistandIappreciatetheireffortstotrytodevelopthesehabitsinme.ThankstoJoannaBettingerforteachingmetoputtheprioritiesofthelabandourlabmatesfirst.ToBrittanyNelson‐Cheeseman for her infectious enthusiasm to new scientific ideas and forpushingmetotalktomorescientiststhatresultedinnewcollaborations.ToFranklinWongalsoforhisenthusiasmandforalwaysbeingapositiveforceinthelab.Heisalwaysreadytodiscuss scientific ideas and help hash out new solutions to open questions. He alwayspushesthosearoundhimtoexcel. IalsoamgratefulforthecurrentandfuturelabmembersIhavehadthepleasureofworkingwithduringthiswork.Special thanksto Jodi Iwatawhohasgreatlyenrichedtheexperience.Herenthusiasmandhumilityhavealwaysmadethelabawelcomingplacetodoresearch. Thanks also to Alex Grutter for his cheery outlook, and to Urusa Alaan for herindustriousnessandhumor.Thankstoallofyouandgoodluck!

v

I alsowould like to thankRachelRose forherpatience and friendshipduringmygraduatecareerand forherhelpwith findinggrammaticalerrors in thisdissertation.Shewasalways there to liftmy spirits, keepmemotivated, and celebratemy successes. I ameternallygrateful tomybrothers,VishalandVishes,withwhomIhavehad thechance toengage indeeplyphilosophicalandchallengingdiscussions throughout thisprocess.Theyhavealwaysbeenaninspiringforceinchangingmyoutlookwhenitwasmostneeded.Ialsoextendthankstoallmyfamilymemberswhohavebeenincrediblysupportive,especiallytomyfamilyinCaliforniawhohavealwaysprovidedmewithahomeawayfromhomeincaseIweretoneedanything. Finally,Iwouldliketothankmymotherandfatherfortheircontinualsupportandbeliefinme.Thewaysthattheyhavecontributedtothecompletionofthisworkisbeyondwords.Butnevertheless,Iwanttothankthemforteachingmethevalueofeducationandperseverance,andforalwaysremindingmethat“wherethereisawill,thereisaway.”

1

Chapter1:Designingnewroutestowardsnovelmagneticfunctionality

1.1.Researchinmultifunctionalspintronicmaterials

1.2.Developmentofnovelfunctionalityincomplexoxidesperovskites

1.2.1.Physicalstructureofperovskites

1.2.2.Electronicstructureofperovskites

1.2.3.Magneticexchangeinperovskites

1.3.Investigationofcomplexoxidesusinghetero‐epitaxy

1.4.Exploitationofspinstatetransitionsincobaltites

1.5.Goalsofthisdissertation

AbstractIn this chapter the driving forces behind this body of research are presented. Ageneralbackgroundregardingthephysicsofcomplexperovskiteoxidesisgiventoestablishtheirrelevancetotheinvestigationofnovelfunctionality.Specifically,wediscuss the concepts of spin states and crystal field splitting in transition metaloxides. The use of thin film growth techniques to study these phenomena is alsointroduced. Background regarding cobaltite oxides is given to show the variousroutesbywhichwecancontrolthepossiblespinstatesinthismaterial.Attheendofthischapter,anoverviewofthetopicstobecoveredintheremainingchaptersisgiven.

2

Chapter1:Designingnewroutestowardsnovelmagneticfunctionality

1.1.Researchinmultifunctionalspintronicmaterials In our current information and computer technology fueled era, there is agrowing demand for new materials that exhibit a versatile mix of magnetic,electronic, and structural properties. The current technologies, in trying to satisfyMoore’sLaw,aredesignedtoapproachfundamentalnanoscalesizelimits inorderto meet the increasing demand for high‐capacity, fast‐paced data storage andprocessing applications. The results include an increase in world‐wide electricityconsumption[1]anddetrimentalenvironmentalimpactsintheformofwasteheatandresourcedepletion.Todecreasenegativeenvironmentaleffects, scientistsandengineers have begun to seek out new multifunctional materials able to lowerpowerconsumptionwhilestillextendingthecapabilitiesofcurrenttechnology. Spintronics [2], one highly touted avenue still in its early stages, takesadvantageofmultifunctionalmaterialsthatexploitboththespinandchargeoftheelectrontocreatenewpossibilities forcurrent technology.By takingadvantageofthenonvolatileelectronspintoperformlogicanddatastorageoperations,wecouldreduceenergyconsumptionfrominformationtechnologysources.Atthesametimeexploiting the spin to perform these operationswould result in faster processingtimes.Manynewtechnologiesrelyingonspintronicsarealreadybeingincorporatedinto current devices including giant magneto‐resistance/tunneling magneto‐resistance readheads, spin valves, spin transistors, sensors,MRAM, andquantumcomputation. However, increasing the reach of this new technology into moredevicesandmoreapplicationsreliesheavilyonthedevelopmentofmultifunctionalmaterials in which spin can be easily controlled via external sources such as anapplied electric field, appliedmagnetic field, or electromagnetic radiation. Fromascientificperspective,agreaterunderstandingofthecomplexinteractionsamongamaterial’s charge, spin, and orbital degrees of freedom is fundamental toengineeringnewmultifunctionalmaterialsanddevices.

1.2.Developmentofnovelfunctionalityincomplexoxidesperovskites Theuniquepropertiesofcomplexoxidesmakethemidealcandidatesfortheinvestigation and further development of multifunctionality. These materials areknown to exhibit a vast array of properties including high‐Tc superconductivity,colossal magneto‐resistance, metal‐insulator transitions, piezoelectricity,multiferroic behavior, spin‐state transition, and other correlated electronphenomenathatrelyonthecomplexinteractionsbetweencharge,spin,andlattice(orbitals) degrees of freedom. Many of these complex oxides have rich phasediagrams that often show that access to these various properties requires only asmall amount of chemical doping, temperature change, or structural distortion.Researchfocusingonthesesystemswillenhanceourunderstandingofthecomplexelectron‐electron interactions that give rise to these properties and demonstrate

3

theirpotentialuse inspintronicapplications[3]. Inorder toappreciatewhythesesystems are ideal for studying correlated electron behavior, it is important todescribetherelationshipbetweenthephysicalandelectronicstructure,aswellastherelationshipbetweentheelectronicstructureandmagneticpropertiesofthesematerials. In theupcomingsectionswewill giveabriefdescriptionof thegeneralpropertiesofthesecomplexperovskiteoxidesbasedoninformationgatheredfromavarietyofsources.

1.2.1.Physicalstructureofperovskites Perovskite oxides represent a class of materials characterized by thechemical formula ABO3, where A is typically a rare‐earthmetal or alkaline earthmetal cation, B is typically a transitionmetal cation, and O represents an oxygenanion (Figure 1). The structure of these materials varies, but, in general, can becharacterizedbyanoctahedronofoxygenanionssurroundingthetransitionmetalcationthatcollectivelysitinsideofasimplecubiccageofrare‐earth/alkaline‐earthmetal cations.Theresult isa12‐foldcoordinatedA‐siteand6‐foldcoordinatedB‐site,withbothcationssharing theircoordinationwith theoxygenanions.Since inmost perovskite oxides the interesting electronic interactions take place betweentheoxygenoctahedralcageandthetransitionmetal,itmaybemoreusefultothinkoftheperovskitestructureasanetworkofcorner‐sharingoctahedralcomplexes.

Figure1CubicperovskiteunitcellshowingpositionsofA‐site,B‐site,andoxygenionsinthelattice.A‐sitecorresponds to therare‐earthmetaloralkalineearthmetal ion.TheB‐sitecorresponds toatransitionmetalcation.(ImagecourtesyA.Grutter)

Perovskites are known to sustain large changes from the inclusion ofdifferentelementalspeciesbychangingbond‐lengthsoroctahedraltiltswithinthecrystal.Deviationsoftheperovskitestructurefromthecubiccase(t=1)areeasilycalculatedbythetolerancefactor[6,7], ≡

√,where(A‐O)and(B‐O)referto

4

thesumoftheA‐site/B‐sitecationandoxygenanionradii.Typicallyfort>1wefindhexagonal perovskites, which we will not consider here. For t <1 the structureadjusts by decreasing the B‐O‐B bond angle from 180˚. This is seen from thecoordinatedtiltingofthecorner‐sharingoctahedrainalattice.Thesymmetryofthelatticemay changebased on the cubic axis aboutwhich the octahedra rotate. Forexample, for rotations about [111] axis we get rhombohedral symmetry, and forrotations about [110]we get orthorhombic symmetry. The rotation is greater forsmallervaluesoft.Exploringhowtheelectronicstructuremaychangeforvariouslattice changes is one of the goals of this dissertation, and so it is necessary toprovideaqualitativedescriptionoftheelectronicstructureofperovskites.

1.2.2.Electronicstructureofperovskites Thebonding in theseoxides isbest thoughtofas ionic,although thebondsbetween the cations and the oxygen anions often adopt some covalent charactershowing strong hybridization between cation and anion orbitals. Thus, for aqualitativeunderstandingoftheperovskiteelectronicstructureclosesttotheFermilevel (valence and conduction band energy levels), it is useful to think about thecationsandanionsinthissystemintermsofamolecularorbitalpictureinthelinearcombinationofatomicorbitalsapproximation[8,9,10].Thisapproachincorporatesvalence bond theory and group theory considerations and provides an intuitiveunderstandingofhowtheelectronic structure isaffectedbychanges in symmetryassociated with distortions and coordinated rotations of these transition metal‐oxygenoctahedra. In this ionic framework we consider the bonding and anti‐bondinginteractionsbetweens‐,p‐,andd‐orbitalsoftheA‐siteorB‐siteatomicorbitalsandtheoxygenatomicorbitals.Figure2showsexamplesofthestronger‐bondingandweaker‐bondinginteractionsthatcantakeplacebetweeninteractingorbitals.Wehave also shown examples of non‐bonding interactions which do not contributeenergy levels to the overall electronic structure. The lobes of the atomic orbitalwavefunctionsaredrawnwithphases(indicatedby“+”and“–“)appropriatefortheangular momentum of each orbital (s, p, d). When two overlapping lobes are inphasetheresultisconstructiveinterferenceoftheatomicorbitalwavefunctionsanda bonding interaction. When these phases are opposite the result is destructiveinterference of the interacting orbitals and an antibonding interaction. Thenonbonding examples show the interactions that are symmetry‐forbidden, andresultinnonetinterferenceorbondinginteraction.

5

Figure2Schematicshowingorbitalinteractionsforpossible‐bonding,‐bonding,andnon‐bondinginteractions between s‐, p‐, and d‐orbitals in a molecular orbital description that uses the linearcombinationofatomicorbitals.

Fortunately, theA‐site4forbitals(if theyareoccupied)arehighly localizedand are typically not heavily involved in the bonding interactionwith the oxygenanion, sowe have not considered themhere. Instead the s‐ andp‐ orbitals of thecationsareheavilyinvolvedinthebonding(andantibonding)interactionswiththeoxygen s‐ andp‐ orbitals and thusmake up the low‐lying (bonding orbitals) corelevels of the electronic structure and the unoccupied high‐lying (antibondingorbitals)levelswellabovetheFermilevel.Theweakerbondinginteractionsofthetransitionmetal3dorbitalwiththeoxygen2sand2porbitalsaretheoriginformostof the electronic structure nearest to the Fermi level. Because themajority of thecontributiontothesestatescomesfromthetransitionmetal3d‐orbital,thesestatesareoftendepictedwiththesamesymmetryasthetransitionmetal3d‐orbitalsandare labeledasthetwo‐folddegenerateeg*andthree‐folddegeneratet2g levels.Theeg* levels are made up of the more strongly interacting ‐antibonding molecularorbitals and the t2g aremadeup of theweaker‐ antibondingmolecular orbitals.Theenergyspacingbetweenthesetwosetsofdegeneratelevelsiscalledthecrystal‐fieldsplittingenergy(∆CF).

6

Usingtheseconsiderations,wecancarefullyconstructthemolecularorbitalsthatmakeuptheperovskiteoxide.Byonlyconsideringthetransitionmetalatomicorbitals in‐ bonding to theO6octahedrawhich forma set ofmolecularorbitals,Burdettetal.[11]hasconstructedanenergyleveldiagramthatcontainsmostoftherelevant details for the perovskite electronic structure. A more precise depictionmightconsider‐bondinginteractionsaswell.Figure3showsanadaptationofthisconstruction[11]illustratingamolecularorbitaldiagramforageneralMO6cluster.Theenergylevelsmayshiftupordownbasedonthechoiceoftransitionmetalandrare‐earth/alkaline‐earthmetal cation, which could influence the Fermi level andthe relative contribution from the different elements’ atomic orbitals to themolecularorbitalsinthediagram.

Figure 3 Rudimentary molecular orbital diagram for a general MO6 cluster showing crystal fieldsplittingbetweentransitionmetal‐likeantibondingt2gandegstates(red).From[11].

Although this description portrays the perovskite electronic structure asbeingcomposedofdiscreteatomic‐likeenergylevels,arealcrystallatticeisbetterdescribedusingbandtheory,wherethediscretelevelsarereplacedwithelectronic

7

bandsthathavesomeenergybandwidth.However,sinceourdescriptionisbasedinthe group theoretical approach, it proves useful when considering how theelectronic structure is affected under changes of symmetry or bond‐length fromvolumechanges, latticedistortions,oroctahedral rotations.Thisdescriptionhelpstoestablishthestronglinkbetweenthestructureoftheperovskiteunitcellandtheoverallelectronicstructure. Figure 4 shows the qualitative change in crystal field splitting energy thatmight occur from a fictional isotropic lattice contraction. The bond‐lengths in thecompressedstructurearedecreasedandtheresultisanincreaseinthecrystalfieldsplitting. Thereason for thischange isexplainedby thechange in thestrengthofthe bonding interaction under such a compression. The transition metal orbitalspointedalongthex,y,andz(dx2‐y2anddz2)directionshaveamuchstrongerbondinginteractionwith the oxygen anions, and as a result the eg*orbitals have a higherenergy (becomemore anti‐bonding). Similarly the t2g orbitals also have a higherenergy,thoughtheincreaseismuchsmallersincethebondinginteractionisalreadymuchweakerfortheseorbitals.

Figure 4 The effect on the electron energy levels nearest the Fermi level under the action of ahydrostaticcompressionontheperovskitelattice

Wecancontinueexploringthisexamplebyconsideringtheselatticeeffectsinthepresenceofelectronsoccupyingtheseenergylevels.Thiswillhelpillustratetheconceptof spin states in transitionmetaloxides,whicharehighly relevant in this

8

dissertation.First,weconsiderthegroundstateelectronconfigurationinanatomicd‐orbital. We follow Hund’s rules which state [12]: (i) the electron configuration(termsymbol)withthelowestenergyhasmaximummultiplicity(2S+1,totalspin);(ii)Withinmaximummultiplicity,theelectronconfigurationwithmaximumL(totalangularmomentum) has the lowest energy; and (iii) for atomswith amore thanhalf‐filled shell the valuewith a highest J (L+S) has the lowest energy. Themostimportantoftheserulesisthefirst,whichgovernstheAufbauprinciplestatingthatunoccupiedorbitalswillbefilledbeforeoccupiedorbitalswhenfillingelectronsinadegenerate set of orbitals. Hence, for five degenerate d‐orbitals, the filling of tenelectronsresultsinS=½,1,3/2,2,5/2,2,3/2,1,½,andthen0. The electron filling is straightforward when all the orbitals in a shell aredegenerate.However,whenthetransitionmetalisbondedtotheoctahedraloxygencage,thed‐orbitaldegeneracyissplitbythecrystalfieldenergyintothedegenerateeg*andt2gsubshells.Tounderstandhowtheelectronsfillthisnon‐degeneratesetoflevels,we can think of the principles governing electron filling as an intra‐atomicexchange(∆ex)energycostthatispaidwhenspinsarepairedonagivenorbital[4].Figure5showsaschematicexampleforthefillingofsix3delectronsbondedintheperovskitestructure.Thepresenceofelectronsintheselevelscreatesacompetitionbetween the ∆ex, which drives the alignment of all the spins in both eg* and t2gsubshells,and∆CF,whichdrivesthefillingoft2glevelsbeforefillingeg*levels.Inthefigure there is an additional energy cost, ∆ex, for populating spin‐up electronscomparedtospin‐downelectronsinthed‐orbitals.

Figure5Schematic illustrationof theelectronicconfigurationunderdifferentcrystal fieldsplittingenergies for Co3+with 6 d‐orbital electrons. Also shown is the intra‐atomic exchange splitting forspin‐upandspin‐downlevelsoftheelectronicstructure.ByconsideringtherelativestrengthofthecrystalfieldandtheHund’scoupling,differentspinstatesarepossibleintheperovskitestructure.

9

Nowletusrevisittheexampleofthelatticecontractiondescribedabove.Inthis kind of lattice change, we might observe a dramatic change in the spinconfiguration fromahigh spin state toa lowspin state. In thehigh spin state thecrystalfieldsplittingenergyislowerthanthedrivingforceforHund’scouplingandso theHund’s rulesareobeyedover the fulld‐orbital. For this casewe follow therulesforfillingusedinthefive‐folddegenerated‐orbitalcaseandgetthesamenetspinmomentontheionasshownabove.Inthelowspinstate,thelatticecontractionhasincreasedthecrystalfieldsplittingenergysuchthatthesystemratherpaystheintra‐atomicexchangeenergy(∆ex)costandcompletely fills thet2gshellwithspin‐up and spin‐down electrons before occupying the higher eg levels. In thisconfiguration,thehighestnetmomentthattheionadoptsisS=3/2.Thus,basedontherelativestrengthsofthe∆CFand∆exwecanobservealargedifferenceinthenetspinmomentoftheion. Fromtheexamplesabove,wecanseehowsmalllatticechangesmightalterthe electronic structure of a material. If these levels were partially filled withelectrons, as they would be in most transition metal perovskites, the changes incrystal field could have profound effects on various magnetic and electronicproperties of the material such as spin state and bandgap. If the pressure wereapplied anisotropically, it is possible that the lattice might change through acombinationofchangesinbondlengthandbondangles.Tounderstandthechangesin electronic structure under these anisotropic conditions, we would have toconsiderbothanisotropicbond‐lengthchangesandanisotropicbond‐anglechanges[13].Whilewecanqualitativelytreatanisotropicbond‐lengthchangesusingsimilararguments that were used in the previous example, without knowing specificallyhowthebond‐anglesandbond‐lengthschangewecannotconcludeanythingfurtherabouttheelectronicstructureunderthesehypotheticalanisotropiclatticechanges.

1.2.3.Magneticexchangeinperovskites The long‐rangemagnetic order in oxides occurs through indirect exchangemechanisms suchas superexchange (SE) [14] inmost ionic insulators anddoubleexchange (DE) [15] in materials that exhibit more conductive behavior. Thesemechanisms are called indirect (instead of direct) because they couple next‐to‐nearest neighbor cations often through an intermediary non‐magnetic nearest‐neighbor anion. Both superexchange and double exchange rely on an electrontransfer process (real and resulting in metallicity in the case of DE; virtual andresulting in localized electrons in the case of SE) that conserves spin angularmomentumandmediatesthespin‐spininteractionsbetweencationscarryinganetspin. ThetypeandstrengthofthemagnetismthatarisesfromsuperexchangearegovernedbytheGoodenough‐Kanamori[12]rules.Theserulesgovernthespin‐spininteraction between half‐filled and half‐filled, half‐filled and filled, and half‐filledandemptyorbitalsonnext‐to‐nearestneighborcations.ExamplesofsuperexchangepossibilitiesforthesecasesareshowninFigure6betweentwodx2‐y2‐orbitalsandanintermediate py‐orbital. The qualitative general principle is that the symmetry‐

10

allowedoverlappingorbitalsbetweenthecationsandtheintermediateanionhaveexcitedstatesinwhichanelectronfromoneionspendssometimeintheorbitaloftheneighboring ion, i.e. thebondsarenotperfectly ionic.Theelectron transfer insuperexchangeisconsideredvirtualbecauseweconsiderthelowestenergyexcitedstates(thestatesinvolvingelectrontransfer)inordertodeterminethegroundstatemagneticconfiguration. IneachscenarioofFigure6twoexcitedstatesareshown:(a)chargetransferfromtheaniontothecation;(b)twochargetransfers‐‐onetotheanionandone fromtheanion.The lowestenergyexcitedstate is(b), theone thatleaves the anion filled and nonmagnetic. Hund’s rules governing the net spin foreachcationcreatesanenergypenaltyassociatedwith the transferofa spin‐downelectrontoacationwithmajorityspin‐upstates.Also,thePauliExclusionPrincipleforbidsthetransferofaspin‐upelectrontoanorbitalalreadyoccupiedwithaspin‐upelectron.Thus,asshownincase(2)inFigure6,evenbetweenanemptycationorbital and a half‐filled orbital ferromagnetic exchange is possible if the orbitalsbelow the empty level in the cation are half‐filled with spins parallel to theneighboringcation.Dependingon thestrengthof thebonding interaction(or)between the overlapping orbitals, different ferromagnetic and antiferromagneticsuperexchange interactions can compete toaffect theoverall strengthand typeofmagneticinteractionoftheoxide.Thestrengthisalsogovernedbythemagnitudeofthe net moment on the cations and the cation‐anion‐cation bond angle.Superexchangemostoftenoccursinantiferromagnetsandinsulatingferromagnets.

Figure 6 A few scenarios for superexchange interaction involving excited state electron transfersfromtransitionmetaldx2‐y2orbitalsandoxygenpyorbitalsthatusuallyresultinantiferromagnetism

11

(1). In some scenarios (2, 3) between a partially filled orbital and an empty of filled orbitalferromagnetismcanoccur.

Scenarios do exist inwhich ferromagnetism via superexchange is possible,though they tend to be weaker and often dominated by competingantiferromagnetic interactions inoxides.Theyalso typicallyoccur in systems thatshowstaticorbitalorderingorlattice‐coordinatedJahn‐Tellerdistortions,inwhichthed‐orbitaldegeneracybetweenneighboringsitesalternates.Intheeventofsuchorbital‐latticeeffects,insulatingferromagnetismhasbeenknowntooccur. Indoubleexchange theelectron transfer is real (notvirtual) as theexcitedstates are low enough in energy that they tend to occur easily [4]. In doubleexchange the electron transfers from next‐to‐nearest neighbor cations via thesymmetry‐allowed orbital on the intermediate anion. Since these cations tend tohavedifferentorbitaloccupations (differentnumberofvalenceelectrons)and theelectronsareitinerant,electrontransfersfromtheaniontotheavailablestateintheneighboringcationorbitalaswell.Thisresultsinthecoincidenceofferromagnetismandconductivebehaviorinthesetypesofmaterials.Thistypeofexchangeisusuallyfoundinmixed‐valentconductingferromagnets. These examples give just a flavor of the types of indirect exchangemechanisms possible in magnetic oxides. We have left out theories for 90˚superexchange between orthogonal non‐overlapping orbitals which result inferromagnetism and are more relevant to a discussion of spinels. With thisqualitativeunderstandingofexchangeinoxideswecanunderstandhowmagnetismin transition metal oxides is a highly correlated electron phenomenon. In thesematerials all the electrons play an important role and interact with the otherelectronstogeneratefascinatingfunctionalbehavior. Hopefully, the preceding discussion has highlighted the value in theinvestigationof transitionmetaloxideperovskites.Smallchanges in latticeappeartobe intimately related to thechanges in theelectronicandmagneticbehavior inthese materials. The unique versatility of perovksites motivates structural andchemicalinvestigationswhichoftenresultinnumerousemergentproperties.Thereisalsovalue in the fact that theunderlyingmechanisms for thevariousbehaviorscanoftentimesbe intuitivelyexplainedbasedonanunderstandingof thephysicalstructure (and symmetry) and its effect on electronic interactions between thedifferentionsinthematerial.Bystudyingthesesystemswecanhopetogainbetterunderstandingandcontrolover spinandelectrondegreesof freedom toengineerusefulspintronicdevices.

1.3.Investigationofcomplexoxidesusinghetero‐epitaxy In this dissertation, we will use thin film epitaxy to affect changes in theperovskite structure. The advances in growth and characterization techniques inrecent years have led to the increasing use of epitaxial synthesis of bilayerheterostructures and superlattices to study novel material phenomena. Modern

12

growthtechniquescombinedwith in‐situmonitoringtoolssuchasreflectionhigh‐energyelectrondiffraction(RHEED),secondaryionmassspectroscopy(SIMS),orX‐raytechniquesallowforatomiclayerprecisioninthegrowthofmaterialsandahighdegreeofchemicalcontrolofeachconstituent layer.Materialengineering throughhetero‐epitaxy has thus been revisited as a viable means for developing newfunctionalityinmaterials,especiallyindesigningnanoscalefunctionalpropertiesinoxideswhichrelyonstrongspin‐orbit coupling,or changes inelectronicbehaviorinducedbyminutestructuralandchemicalmodifications. Therearetwomainroutes inwhichhetero‐epitaxyhasbeendemonstratedto be a useful materials engineering tool: (i) generating novel interfacialphenomena;and(ii)generatingnovelgroundstatesusingepitaxialstrain.Thefirstrouteusestheideaofestablishingelectroniccontrolbybringingamaterialincloseproximitytoanothermaterialtocreateanewinterface.Oneoftencitedexampleofthis is the recent discovery of a highly‐confined, conducting two‐dimensionalelectrongasformedatthe interfacebetweeninsulatingLaAlO3andSrTiO3[16]. Inthissystem, it is thought thata “polarcatastrophe”develops,wherebytheelectricfield at the surface of the ionic LaAlO3 layers diverges. It is only resolved by thecontributionof½anelectronacrosstheinterfacefromtheLaAlO3sidetothefinalTiO2 layer inSrTiO3.Thedetailsof theexplanationof this interfacearestillunderintenseinvestigation,butitissufficienttosaythatthisisanexcitingnewapproachtodiscoveringthelatentpropertiesinoxidematerials[17].Interfaceengineeringisavaluabletoolthathashadgreatsuccessinuncoveringnewpropertiesthatarenotexistentintheparentmaterials,butitwillnotbeconsideredinthisdissertation. The second route entails manipulating a material’s bulk equilibriumstructure throughcoherentepitaxialgrowthona lattice‐mismatchedsubstrate.Aswe have already alluded, this route can result in unexpected changes in theelectronic behavior of a material, especially in complex oxides where orbital(lattice) degrees of freedom are strongly coupled to charge and spin degrees offreedomoftheelectrons.Arecentexampleofthismethodisillustratedbytheuseofepitaxial strain on insulating LaTiO3to achieve metallic transport behavior [18].This material is a Mott insulator with a small charge gap of 0.1 eV in the bulk.Through a suppression of the orthorhombic distortion from epitaxial growth onSrTiO3,thebandwidthneartheFermilevelisbroadenedandthet2glevelsplittingisreduced, thus giving rise tometallicity in thin film form.The result demonstrateshowepitaxialstraincaninduceachangeintheelectronictransportbehavior[19].Inthisdissertation,wefocusonasimilarapproachtoachievenewmagneticpropertiesthroughstructuralmodificationsofperovskitecobaltites. There have been some previous attempts to tune magnetic properties inoxides using heteroepitaxy. These rely on the use of strain to affect the sign andstrength of spin‐orbit coupling and magnetocrystalline anisotropy or the use ofcomplementarynonmagneticmultilayerstogeneratemagnetismataninterface.Inthis work we present an alternative approach that focuses on tuning the localmoment on an ion via the spin state to generate amagnetic responsewhichwasabsentpriortothesubjectiontostrain.Byexperimentingwithcontroloverthelocalspin state,we can aim to control both themagnitude and presence of long‐range

13

magnetic order in thesematerials, thus giving rise to a unique form ofmagneticfunctionality.

1.4.Exploitationofspinstatetransitionsincobaltites The cobaltites are a unique class of materials that can assume a range ofdifferent spin state configurations at different temperatures. In particular, thesechanges in spin state in the bulk have given rise to anomalies in magneticsusceptibility and electron transport [20, 21]. As we have suggested above, byfindingaroutetoaccessthesespinstatesindependentoftemperature,wecanhopetocreatenewfunctionalityforoxide‐basedelectronicsthroughthemanipulationofmagneticmomentandelectronicbehaviorofamaterial.Onewellstudiedexampleofthesecompounds,LaCoO3,undergoesaspinstatetransitionatlowtemperature.Thereisaconsensusthatabove25KsomeCoionsinLaCoO3begintotransitiontosomehigher spin state froma lowspin state (LS, S=0).There is still considerabledebateastowhetherthehighspin(HS,S=2)orintermediatespinstate(IS,S=1)ofCoismorestableathighertemperatures.Figure7showsthepossiblespinstatesforCo suggested for thismaterial.Becauseof the strong competitionbetween crystalfieldsplittingenergy(∆CF)andHund’scoupling(∆ex) inthissystem,theLS, IS,andHSstateshavebeenshowntobecloseinenergy[22,23]andare,therefore,easilyaccessiblewithtemperature.Inanycase,bulkcobaltitesdonotshowanylong‐rangemagnetic ordering or indication of substantial HS or IS Co population at lowtemperatures.However,thismaterialholdstremendouspromiseforourpurposes,asithasthepotentialtoshowadramaticchangeinthemagnitudeofthe(spin‐only)magneticmomentfromnonetmoment(S=0)to4µB/Co(S=2).

Figure7Possible spin state configurations forLaCoO3. Low spin, high spin, and intermediate spinelectron configurationshave all beenproposed.TheLaCoO3 systemadoptsan evolvingmixtureofthesespinstateswithincreasingtemperature.

Numerous efforts have beenmade to control the spin state through otherexternalparametersasidefromtemperature.ThemostlikelyroutestothiscontrolmustfocusonmanipulatingthecrystalfieldthroughmodificationsoftheCo‐ObondlengthandCo‐O‐Cobondangle.Previouseffortstoenhancethestabilityofthelowspin,lowtemperaturestateusinghydrostaticpressurehavebeensuccessful[24,25,26].Hydrostaticpressureincreasescrystalfieldsplittingenergyduetotheincrease

14

inorbitaloverlapthatcompeteswiththeHund’scouplingenergy.This increase incrystalfieldsplittingstabilizesthelowspinstateofCoionsinasimilarwayasthescenariopresentedearlierinthischapter. Another approach,whichwill be explored in greater detail in Chapter 5 ofthis dissertation, is bandwidth engineering bydoping theA‐site cation, La,with asmaller rare‐earth metal, i.e. Pr, to alter the Co‐O‐Co bond angles and egand t2gbandwidths related to the Co‐O orbital overlap [27]. This method of so‐called“chemicalpressure”usessubstitutionofisovalentcationsintheA‐site.ThisdoesnottypicallyresultindramaticchangesintheCo‐Obond‐length,butinsteadachangeintheCo‐O‐Cobondangle,whichdeviatestosmallerangles from167˚[28](inLCO).The result is an increase in octahedral tilts as the overall symmetry becomesorthorhombic. This change has been closely associatedwith a reduction in the egbandwidth resulting in amore stable low spin statewithdecreasingA‐site cationsize.Thespinstatetransitiontemperatureisshiftedfrom~25KinLCOto200KinPCO,andashighas~540KforLuCoO3(Co‐O‐Coangle~146˚)[27]. The only well‐established success in destabilizing the low spin state andenhancingthestabilityofahigherspinstateinbulkLCOandPCOhasbeenthroughthe use of aliovalent cation substitutions on the A‐site to introduce a mixture ofvalence states for the Co ion. For example in La1‐xSrxCoO3and Pr1‐xSrxCoO3, Sr2+resultsinthepresenceofsmaller,intermediatespin,intermediatevalence,Coions.Withx>0.2and fullpercolationof themixed ion/spin clusters, thismixtureofCovalence and spin states results in a metallic eg electron transport behavior andferromagnetism[29,30]. Recently, therehasbeen considerableeffort touse epitaxial strain todrivetheCoionsinLCOtoahigherspinstate[31,32,33,34,35,36].Whileepitaxialfilmsof LCO were found to be ferromagnetic, an actual understanding of the strain‐derivedCospinstates in these filmshasnotbeenestablished. Inaddition there iscontroversy in the literature as to whether it is in fact strain [33, 37] orstoichiometricdefects[35,36,38]inducedbythegrowthwhichcanbecitedastheprimary cause forapparent changes in theCo spin state, and thus responsible fortheobservedferromagneticorderinthesefilms. TheoriginsofthechangesinCospinstateandtheferromagnetismhavenotbeen the only sources of controversy in LaCoO3 films.While there is a consensusthat films in tension inevitably result in ferromagnetism, during the course ofinvestigations by numerous groups, both ferromagnetic and non‐ferromagneticbehavior have been reported for films under biaxial compression. Many authorswhohaveperformedsubstratedependentinvestigationsobservedferromagnetisminfilmsgrownincompressiononLAOandSLAO[33,37].AtransmissionelectronmicroscopeinvestigationintofilmsonLAOobserveddichroismintheEELSspectraatlowtemperaturerelatedtoferromagnetism[39].However,Parketal.performedamagnetic forcemicroscopystudyshowing thatwhile ferromagnetism is likely tobe strongly associated with epitaxy in tension, ferromagnetism on LAO(compression) appears to arise mostly from chemical inhomogeneity [35]. Inchapter4wewill reporton thicknessdependenceof ferromagneticproperties forfilms on LAO, and will show that strained films on LAO are not magnetic while

15

relaxed films show ferromagnetism. Collectively, these results suggest that bothstrainanddefectscangiverisetolong‐rangeferromagneticorderinLaCoO3. By investigating cobaltites in detail using hetero‐epitaxial synthesis on aseries of different substrates, we hope to better understand the unexplainedferromagnetism in this system. In addition, by studying the magnetic propertiesunder chemical pressure from the substitution of Pr in theA‐site,we have foundepitaxialstraindominatesovertheinfluenceofchemicalpressureingoverningthespinstatespresentinthePCOfilm.Remarkably,inthissystemwehavefoundthatthePrionsshowlong‐rangeorderantiparalleltotheCosublatticeinthepresenceofHSferromagneticCoions.

1.5.Goalsofthisdissertation In the chapters to come we will explore the novel ferromagnetic groundstatesincobaltitematerialsaccessedthroughhetero‐epitaxialsynthesis.Inchapter2, we will review the pulsed‐laser deposition growth technique and variouscharacterizationtechniquesthatareusedthroughoutthisdissertation.InChapter3and 4 films grown on LAO, LSAT and STO substrates of LCO will be explored indetail.Chapter3willfocusonthestructuralcharacterizationusingX‐raydiffractionand scanning tunneling electron microscopy techniques to explore the effects ofgrowth on a lattice‐mismatched substrate and the effects of relaxation withincreasing thicknessof the films.Chapter4will connect thesedetails to trends inmagnetometry data and the electronic structure properties found from X‐rayabsorption techniques. In Chapter 5 these characterization tools will be used toinvestigate strained filmsofPCOand toexplore theCoandPrsublatticeorderingunderchemicalpressureeffectsthatcompetewiththeepitaxialstrain.Inchapter6we will summarize the results presented in this work, and provide a sense ofdirectionforfutureexplorationofnovelphenomenainthesematerials. This dissertation explores the effects of using epitaxial strain to engineernovel ordered magnetic behavior through the manipulation of the spin states incobaltites. This represents a new use of epitaxial strain to control the magneticbehaviorofoxides.Weshowthattheferromagnetisminthefilmshascontributionsfrom both the structural distortions and the defects generated during growth. Byexploring the effects from lattice distortions and epitaxial growth in great detailusing modern microscopic and spectroscopic tools, we are able to discern therelevant electronic configurations and lattice distortions that make up theferromagnetism in these films.Finally,weshowthat thisuseofepitaxial strain toengineernovelmagneticstatesispowerfulenoughthatitcanbeextendedtoothercobaltitessuchasPrCoO3,wherethehighspinstateis lessaccessible. Inthisrare‐earthcobaltite,wedemonstrateanemergentnovelferrimagnetismthatoccursduetothestabilizationofHSstate‐mediatedmagneticorderintheCosublattice.Thus,bymanipulatingthespinstateintheCoionsweareabletoinduceananti‐parallellong‐rangeorderingofthemomentsintheotherwiseparamagneticPrsublatticeinPrCoO3.Weareonlyat thebeginningstagesof thedevelopmentofcomplexoxidematerials for industrial spintronic applications. Much more work beyond this

16

dissertation still remains to better our understanding of the profound effects ofchemical dopants, oxygen vacancies, phonons, electromagnetic fields, andmechanicalstressesontheuniquepropertiesdisplayedbythesematerials.

17

Chapter2:ExperimentalTechniques2.1.SampleFabrication

2.1.1.PLD2.2.SampleCharacterization

2.2.1.AtomicForceMicroscope/Microscopy

2.2.2.X‐rayDiffraction

2.2.3.RutherfordBack‐Scattering

2.2.4.Magnetometry

2.2.5.X‐rayAbsorptionSpectroscopy(XAS)

2.2.5.1.X‐rayMagneticCircularDichroism

2.2.5.2.X‐rayNaturalandMagneticLinearDichroism

2.2.6.TEMandSTEM/EELS

AbstractIn thischapter the techniquesused forgrowthandcharacterizationofoxide filmsarebrieflydiscussed.Asurveyofthetechniquesusedinthisworkandsomespecificconsiderationsusedwhen carryingout experiments arehighlighted. Inparticular,important precautions for oxide growth using pulsed‐laser deposition areaddressed. Importantprecautionsnecessary toensuremeasurementaccuracyandproperanalysistakenduringthepost‐growthcharacterizationtechniqueswillalsobeexplained.

18

Chapter2:ExperimentalTechniques2.1.SampleFabrication Themostimportantaspectofathinfilminvestigationisproperattentionandcareduringsamplefabrication.Whilenumeroustechnologicaladvanceshavemadethe synthesis of single crystalline thin films with precise atomic layer controlrelatively routine, diligence is still required at each step. In the first part of thischapter, we will explain the thin film growth technique known as pulsed‐laserdeposition (PLD) thatwasused in the courseof this investigation.Thedepositionprocessofthisgrowthtechniqueaswellastheimportantconsiderationstobearinmindwhencarryingoutthinfilmsyntheseswillbediscussed.

2.1.1.Pulsed‐LaserDeposition(PLD) Pulsed‐laserdeposition(PLD)hasbecometheworkhorseforoxideresearch,due to its ability to create high quality crystalline films quickly and easily fromablation of a stoichiometric pressed ceramic target. However, despite the relativeeasewithwhichthinfilmsmaybefabricated,numerouscomplexinteractionstakeplaceduringthegrowth.Aschematicof thePLDsystemusedfor thesegrowths isshowninFigure8.ThePLDprocessisusuallythoughtofasafourstageprocess[1,2]:(i)thelaserbeaminteractswiththetargetandgeneratesaplume;(ii)theplumeinteractswith thebackgroundgason itswayto thesubstrate; (iii)constituentsoftheplumesticktotheheatedsubstrate;and, finally,(iv)theconstituents fromtheplumenucleateandgrowintolayersoffilmmaterial. Thefirststageinvolvesthetransferofenergyfromahighpoweredlasertoatargetmaterial(stage1).This(typically)stoichiometrictargetissuperheatedbyitsabilitytoabsorbthephotonsfromtheincidentlaser,andresultsinarapidthermalevaporationandsimultaneousionizationofitsconstituentatoms.Theresultofthisinteractionisahighlyenergeticplumeofneutralatoms,ions,electrons,and,insomecases,moleculesthatinteractwiththegasesinthedepositionpathandaredirectedtowardsaheatedsubstrate(stage2).Thespeciesintheplumethenlandandstickonthesubstrate(stage3)mostlybasedonthestickingcoefficientontothesubstrateforthematerialbeingdeposited.Finally,throughtheaidofremnantkineticenergyfromtheplumegenerationprocessandaddedheatfromthesubstrateheater,theydiffuse and nucleate into (ideally) highly ordered atomic layers (stage 4). Thisnucleationandgrowthisgovernedbythedensityoftheincidentplumespeciesonthesubstrate,theoverallannealtimeandtemperature,andtheproperties(thermaldiffusivity,thickness,surfaceenergy)oftheunderlyinglayers. At each step in this process, numerous parameters can be tuned to obtainhigh quality, uniform, stoichiometric growth to investigate novel oxide propertieswith confidence. The parameters can be separated into those affecting the laser‐target interactions and plume‐substrate interactions. On the laser‐target side, wewillconsider laserenergy(density),spotsize,pulse frequency, targetdensity,andtargetsurfacemorphology.Ontheplume‐substrateside,wewillexploretheroleof

19

oxygen background pressure, substrate temperature, substrate distance, andsubstratepreparation.

Figure 8 Schematic of a pulsed‐laser deposition configuration and an image taken during a highoxygenpressuregrowthat700˚C

The first stage in this growthprocess reliesheavily on the various tunableproperties of the laser. In fact one of the aspects thatmakes PLDwell suited foroxide filmsynthesis is theadvances inexcimer laser technology thatprovidehighenergy (and low average power), coherent beams at the suitablewavelengths forabsorption bymost oxidematerials. Excimer lasers are now available at energiesfrom157nm (~7.91 eV, F2 laser) to 351 nm (~3.53 eV, XeF laser). Thus, for thegrowth of complex oxides, choosing a laser of the appropriatewavelength that isabsorbedbythetargetmaterial is important. It isforthissamereasonthatPLDisvirtuallyimpossibleformetallicmaterials.Thehighreflectivityofthesematerialsinthese wavelengths prevents sufficient energy transfer per pulse to the targetmaterial.BychoosingaKrFexcimerlaserthatemitsquick(10‐30ns)248nm(5eV)pulses,weareabletogrowwithconsiderablespeedandcanaccommodatealargenumberofcomplexoxidematerials.

Since the laserwavelengthandpulseduration is setby the laser, themostcriticalcontrollableparameterofthelaseristheenergydensity.Theenergydensityisdeterminedbythespotsizeandenergydeliveredbythelaser. Foragivenspotsize,thelaserenergyandconsequentlytheoverallenergydensitydeliveredtothetarget require some optimization. The plume generated during the target‐laserinteractioniscreatedbytherapidvaporizationofthetargetmaterialfromthermalablation and from the strong Coulomb repulsion of ions in the target that haveejected photoelectrons after absorbing the laser energy. Thus, a target will notvaporizeenoughmaterialtoformaplumeuntilacertainenergyablationthresholdisreached.Thisthresholdenergydensitycandependontargetdensity,atomicmassof thetargetelements,andabsorptionskindepthof thetarget.However, it isalsoimportanttonotethatdepositingwithtoohighofanenergydensitycanresultinahighlyionizedplumewhichcancontributetostoichiometricdefectsinthefilm.Formostofourcomplexoxidethinfilmgrowth,theenergydensityvaluesrangesfrom

20

0.5 – 2 J/cm2. Thin filmquality and speedof growth canbe further optimizedbyexperimentingwithinthisrange. Controlling the spotsize canalsohaveeffectson thequalityof the film. Asmallerspotsizeresults in largerplumespread[3,4]andbecomeslesspeakedintheforwarddirection.Thiseffectisunderstoodintermsofthelawsgoverninggasdynamicswheregreaterplumedivergenceoccursfromthecornersofalaserspot.As a result the growth rate is slowed considerably and a larger spatial uniform,albeit slower, plume is delivered to the substrate per pulse. A small spot size,however,introducesinconsistenciesintheenergydensityandplumestoichiometryfrom pulse to pulse arising from defects, chemical and morphologicalinhomogeneities in the target, or drift in the laser energy. On the other hand, alarger spot generates amore forward directed plume, faster growth rates, and amorerobustenergydensity,butwillleadtomoreinhomogeneitiesindirectionsoffof the axis perpendicular to the target surface. For the majority of the growthscarriedoutinthiswork(exceptwhenindicated)thespotsizewaschosentobeonthelargersiderangingfrom1.0‐1.2cm2. Anotherimportantparameterassociatedwiththelaseristhelaserpulseratewhich can be tuned to enable faster or slower growths. This can have dramaticeffectsonthesurfacequalityofthefilm.Afasterpulseratecancreateapileupofatoms to supersaturate the number of species at the growth interface on thesubstrate. On the other hand a slower pulse rate can allow more time betweenpulses for species to diffuse and the surface to recover (with the aid of thermalenergysuppliedbythesubstrateheater).Infilmmaterialsorfilmorientationsthataredifficulttogrowunderthermalequilibriumconditions,afasterpulseratemaybemoredesirabletopreventtheexcessexposuretohightemperatures.Thiscouldhelpprevent the system from relaxing to itsmost stable lowest energy state.Thechoiceofpulseratealsomayaffectthegrowthmodeofthefilmandcouldrequiresomeoptimizationtoachievethedesiredfilmproperties. Inaddition to the laser,wehave to take into consideration thequalityandconditionof the target. It isusuallydesirable toablatea targetwithhighdensitysincethedensitycanplayasignificantroleinablationthresholdsandefficiencyofthe plume generation process. The overall surface morphology of the targetimpingeduponby the lasercanaffect thestoichiometryanddensityofclusters inthe plume. Typically, a smooth surface is desired; however, over the course ofnumerous laser‐target ablation processes an unavoidable cone‐like morphologydevelops. To minimize variation and improve reproducibility from run‐to‐run, wecarry out the following. First the target surface is made smooth by sanding thetargetpriortogrowth.Thenthetargetispre‐ablatedforabout5minutesunderthedepositionconditionsbeforeexposingthesubstratetothetargetplume.Duringtheactual deposition and during the pre‐ablation process, a target rotation system isused to continually move the laser spot to a new position on the target. Thisestablishes a uniform, consistent track overwhich the laser can ablate the targetduring the actual film deposition. The pre‐ablation step helps to avoid anysignificantchangethatmightoccurinthegrowthduetothechangesinthesurface

21

morphologyof the target from the initial ablation (after the sanding step). In thiswaytheplumewill ideallyadoptacertaindegreeofuniformitythat ismaintainedthroughoutthegrowthprocess. Oncesuchauniformplumeisgeneratedbythelaser‐targetablationprocess,itisforwarddirectedwithneutralatoms,ions,electronsandmoleculesofenergiesof several hundreds of eV towards the heated substrate. In a number of oxidegrowths it is necessary to allow the highly energetic plume to interact with anoxygen gas before reaching the substrate. The ambient oxygen atmosphere canserve two purposes: to slow down and lower the energy at which the plumeconstituents approach the substrate surface (kinetics), and to enhance thestoichiometry of the film (chemistry) to correct for deficiencies in the target, orensurefulloxygenationofmorereductivematerialsystems.However,theadditionof this background gas does have some disadvantages. In general, adding thebackground oxygen will lower the mean free path of the species in the plumeresultinginmultiplebombardmentsandpossibleclusterformationwiththeoxygengas.This can result in large clusters forminganddepositingon the surfaceof thefilm.Ontheotherhand,acompleteabsenceofthebackgroundgas,withoutproperattentiontothelaserenergydensityused,canresultinanoverlyenergetic(severalhundredeV)plumethatcanablatematerialfromthesurfaceofthenewlyformingfilm. In some instances, higher oxygen pressures must be used to ensure fulloxygenationofthefilmattheexpenseofincreasedfilmroughness. Some of the issues regarding the mean‐free path of the plume species inthese ambient pressures can be affected by changing the substrate to targetdistance. Bringing the substrate closer to the target will increase the number ofplumespeciesthat impingeonthesurfaceandincreasethegrowthrate.However,bringing the substrate too far into the plume can result in re‐ablation of the filmdeposited on the substrate surface by the energetic plume causing stoichiometricandstructuraldefects.Inourstudy,weheldthesubstratetotargetdistanceat~7.6cmtoensurethattheplumedidnotengulfthefilm. Finally, the substrate temperature must also be optimized. Highertemperatures can result in higher crystallinity, butmay cause excess diffusion ofatoms in the sample, affecting the stoichiometry or homogeneity in the film andacross the substrate/film interface. Increasing the temperaturewill alsodrive thegrowth towards thermal equilibrium, andmay be undesirable in the synthesis ofsomematerialsthatrequirethekineticbenefitsofPLDgrowthtoformstablesingle‐crystallinefilms.Intheendabalancemustbestruckamongtheparametersoflaserenergy density, oxygen background pressure, pulse frequency, substrate distance,and substrate temperature to obtain optimal smooth film growth at reasonablerates. Many steps can be taken to prepare substrates for sample fabrication,ranging from subjecting the substrates to a comprehensive onslaught ofcharacterization (SQUID, AFM, XRD, etc.) to minor chemical etching, thermalannealing,andgeneralsubstratewashing.Forthegrowthsperformedinthiswork,the procedure was kept simple and the substrates were simply washed usingacetone,methanol, and isopropanol inanultrasonicbath for5‐15minutes. Some

22

samples were subjected to measurements in the atomic force microscope (todetermine substrate miscut) and others were subjected to measurements in theSQUIDmagnetometer toobtainbackgroundmagnetic responsedata that couldbeuseful in background subtractions for future film measurements. Of paramountimportance is that the substrate surface is free of dust or defects that couldadverselyaffectfilmproperties. Uptofoursubstratescanbeloadedontotheheaterusingsilverpastemakingsuretoconfinethepositionof thesubstratestowithina1cm2area inthecentralregionof theheateralong the central axisof theplume.Whenapplying thesilverpaste, great carewas taken to ensure just enoughwas applied to allow for goodthermalcontactbetweenthesubstratesandtheheater. If toomuchsilverpaste isused,thereisariskofpastestickingalongthesidesofthesubstrateorevenlandingon the surface of the substrate. To ensure uniform heating of the substrates, wemustapplypasteonlyalongthebacksurfaceofthesubstrate.Priortoinstallingtheheaterintothechamber,theheaterisheatedto~200‐300°Coutsideofthevacuumchambertodriveoffallsolventsfromthepaste.Thisaidsinmorerapidpumpdownofthechamber. Post‐growthprocessingcanbeofutmost importanceespecially inensuringproperoxygenstoichiometryandmicrostructure.Thedetailsofthesamplecoolingprocessoradditionalpost‐growthannealingstepcanbecrucialinobtaininghighlycrystalline, stoichiometric films. In general, for the thin film studies in thisdissertation,thefilmsgrownin320mTorrwerecooledinatleast1Torrofoxygen.When lower growth pressureswere used, the samplewas cooled in the pressureusedduringgrowth.Onlyinsomespecificannealingstudies,werefilmsannealedin600‐1000°Cforafewhoursinatubefurnaceexposedtoair.Ifthefilmsareoxygendeficient,thisprocessshouldbesufficienttodiffuseoxygenthroughoutthesamples. Additionally,greatcareisneededwhenhandlingthesampleswithtweezersandwhen removing silver paste. For SQUIDmeasurements, remnant silver pastewasgentlysandedoffthebacksideusingnonmagneticsandpaper.Aftersilverpasteremoval, the samples were gently rinsed in isopropanol to remove any dust orparticles that may scratch the films. Small amounts of impurities in the film ordamage to the film can adversely affectmeasurements, and so the utmost care istakenduringthesesteps. Usingthisapproach,wesynthesizedhighqualityoxidefilmsinordertocarryouttheinvestigationspresentedhere.Ourhighqualitysinglecrystallinesubstrateswere typically obtained from Crystec, GmbH. The LaCoO3 target used in thisdissertationwasobtained fromPraxair. ThePrCoO3 target used inChapter5wassintered and pressed by Shameek Bose and Dr. Christopher Leighton at theUniversityofMinnesota.Itwassynthesizedbysolidstatereactionofstoichiometricquantities of Pr6O11 (Sigma‐Aldrich, 99.9% purity) and Co (C2O4).2H20 (Sigma‐Aldrich)powders.Thereactantswerethoroughlygroundandreactedat1000˚Cinair for 7 days,with 2 intermediate grindings. The reactedpowderwas then cold‐pressedintoadiskat16,000psiandsinteredat1200˚Cinairfor96hours.Itwasthenslowcooledtoroomtemperatureat0.5˚C/min.ThephasepurityofthetargetwasconfirmedbyX‐raydiffractionpriortouseinPLDfilmsynthesis.

23

2.2.SampleCharacterization

Once samples were grown, a range of characterization techniques wereperformed to ensure high structural quality as well as to determine functionalpropertiesofthesefilms.Theremainderofthischapterisdevotedtodescribingthevariouscharacterizationtechniquesthatwereusedincarryingouttheresearchinthefollowingchapters.Whilesomecharacterizationmethodsareleftoutandnotallof these methods are used on every grown sample, we introduce many of themethods that were used in the course of this dissertation. Throughout thisdissertation, we have focused on the structural and magnetic characterization ofcomplex oxide thin films. We will cover atomic force microscopy (AFM), X‐raydiffraction(XRD),Rutherfordbackscatteringspectrometry(RBS),magnetometry,X‐ray absorption (XA) techniques, and scanning tunneling electron microscopy(STEM) techniques. RBS was performed after training under Dr. Kin Man Yu atLawrence Berkeley National Laboratory (LBNL). The XA measurements wereperformedattheAdvanceLightSourceatBeamlines6.3.1and4.0.2incollaborationwithDr.ElkeArenholz.TheSTEMandEELSmeasurementswereperformedbyDr.MariaVarelaandDr.NevenBiskupatOakRidgeNationalLaboratory.

2.2.1.AtomicForceMicroscopy(AFM) An atomic force microscope (AFM) was used to probe the surfacemorphologyofthefilms.Asiliconcantileverisoscillatedatasetfrequencyasit israsteredoverthesurfaceofthesample.Asthecantileverencountersfeaturesonthefilmsurface,thechangeintheoscillationfrequencyismonitoredbyanopticallaseron the cantilever.These frequency changes (changes inphase andamplitude) arethenoutputvisuallybycorrespondinglyvaryingpixelcontrasttoproduceanimageofthesurfaceofthefilm.TheDigitalInstrumentsDimension3100ScanningProbeMicroscopeusedinthisdissertationhasaveryhighverticalresolutionof<1Å.Ontheotherhandthelateralresolutionis<20nm.Forourpurposeofmeasuringtheaverage surface roughness over the entire sample this is sufficient since smoothfilms can be characterized by root‐mean‐squared (RMS) roughness values on theorderofaunitcellforourmaterials(~4Å).AnylargeislandsformedduringgrowthwillbeobviousfromanincreaseintheRMSroughnessvalue. Thistechniqueprovidesagoodcheckforthequalityofthesubstratepriortothegrowthofthefilm,aswellasforthefilmaftergrowth.Inidealsinglecrystallinegrowth the surface would be extremely smooth (< 3 Å) and terraced featuresevident in the bare substrate (from intentional substrate polishing at a miscutangle)wouldmaketheirappearanceonthesurfaceofthefilm.SmoothfilmstendtoattractdustandotherparticulatesovertimeandthusmeasurementofthesurfaceusinganAFMwascarriedoutimmediatelyfollowinggrowth(andpriortoanyothercharacterizationtechniqueorsamplehandling).

24

2.2.2.X‐rayDiffraction(XRD) X‐rayDiffraction(XRD)wasusedtodeterminethemajorityofthestructuralpropertiesof the film.XRDuses thephenomenaof constructive interference fromelastically scatteredX‐rays incidentona sample toprovide information regardingthe sample’s crystal structure [5]. Figure 9 shows two geometries that satisfyBragg’s Law, 2 sin or ′ ∆ , such that information regarding theinterplanar spacing in the crystal lattice could be obtained. By determining theanglesatwhichtheBraggconditionissatisfied,wecanobtaininformationaboutthelatticeparameter for filmsgrownondifferentsubstratesundervariousconditionscouldbetaken.

Figure 9 Schematic showing sample orientations relative to incident and diffracted beam for twodifferenttypesofdiffractionscan

In a standard 2 measurement, is held constant to2 /2while the2 valueissweptoveralargeanglerange(seeFigure9(a)).Inthistypeofsymmetricscan,theout‐of‐planelatticeparameterofthefilmscouldbedetermined.Additionalinformationregardingthecrystallinequalityof thefilmsandthemosaicspreadofgrains inthefilmisdeterminedbyrockingthe angleaboutthe2 /2anglewhilefixingthedetector2 valuetoanout‐of‐planepeakofthefilm.Thefullwidthathalfmaximum(FWHM)ofthe scan,or∆ ,isregardedasameasureofcrystallinity. Toobtainin‐planelatticeparametersofthefilms,weusedasymmetricscans(Figure 9(b)) with the scattering vector satisfying the Bragg condition andcorrespondingtoalatticevectorthathasin‐planeandout‐of‐planecomponents[6].In these asymmetricmeasurements, is equal to2 to cause a tilting ofthesamplesothatthescatteringvectorisnolongernormaltothefilmsurface(asitwas for the 2 measurement). Inorder todetermine therelationshipbetween

25

filmandsubstratepeaks,wetakeaseriesof 2 scansoverasmallrangeinbothand2 . In thiswayan intensitymap canbeobtainedasa functionof and2 ,

representing a two‐dimensional cut of reciprocal space which can be used todeterminetherealspacelatticespacingbetweenplanesalongin‐planeandout‐of‐planedirections.Thistypeoftwo‐dimensionalscaniscalledareciprocalspacemap(RSM). The final XRD geometry that will be discussed pertains to the usage ofgrazing incidenceX‐rayreflectivity todetermine the thicknessof individual layersgrown in single films and multi‐layered structures. To perform a reflectivitymeasurement, a symmetric scan ismeasured in the rangeof 0‐ 10°usinga set ofsourceanddetectoropticsoptimized forhigherX‐ray intensities todeterminetheangle below which total internal reflection of the X‐rays in the film occurs. Thisangle is related to the refractive index of the film. Strong interference effects areknown to occur between boundaries of different refractive index (film andsubstrate,filmandair)thatgiveriseto“fringing”ofthelowanglediffractedX‐raybeam.Using information related to the critical angleof internal reflectionand theperiod between fringes, we can determine the thickness of the film or individuallayers in a film. When the sample was between 5 nm and 100 nm thick, thethickness was determined using the X‐ray reflectivity technique. For thickersamplesRBSwasusedtomoreaccuratelydeterminefilmthickness. AllXRDworkpresented in thisdissertationwascarriedoutusingaPhilipsPanalytical four‐circleMaterialsResearchDiffractometer (MRD)atLBNL. ACuK‐edgelinesourcewasusedtogeneratetheX‐rays.ThebeamwasconditionedwithaparabolicmirrorandaGe2204‐bouncecrystalmonochromatorbefore impingingonthesampletocollimateandmonochromatetheX‐raystotheCuKαwavelengthof1.541 Å. For X‐ray reflectivity, the monochromator is removed to increase theintensity(attheexpenseofwavelengthresolution)andasetofparallelslits(sollerslits)areinsertedbeforethedetectortolimitthebeamdivergenceandenhancethelowscatteringangleresolution.

2.2.3.RutherfordBackscatteringSpectrometry(RBS) Rutherford Backscattering Spectrometry (RBS) was used to determine thestoichiometryofthesampleandthethicknessofsamplesgreaterthan100nm.ThetechniquereliesontheenergylossdetectionofbackscatteredHe2+(2‐3MeV)ionsafter impinging on the sample [7]. The energy of the detected ion is affected byscatteringwithatomsatthesurfaceandtheinteriorofthesample.Themassoftheatoms inthesampleaffect thestrengthof the interactionwiththe impinging ions.Heavier elements have higher scattering cross‐sections and stronger interactions,andasaresult,RBSismoresensitivefordetectingquantitiesofheavierelements.Ata fixed detection angle the backscattered ion will have a different energy forscatteringeventswithdifferentelements.Ionsthatpenetratefurtherintoasamplealsolosemoreenergybeforebackscatteringtothedetector.InanRBSspectrumthepeak position and intensity is governed by the scattering cross‐section of theelementscatteringtheimpingingion.Thewidth,andinthecaseofburiedelements,

26

relativeshiftofpeakpositionisdeterminedbythedepthoftheelementscatteringtheimpingingion.Inthisdissertation,RBSresolutionisworseforlighterelements,suchasoxygen,andisonlyusedtodeterminetherelativeratioofheavierelementsinthesamplesandforthicknessdeterminationofsomesamples.Themeasurementswere performed at the Ion BeamAnalysis Facility at Lawrence BerkeleyNationalLaboratory with the help of Dr. Kin Man Yu. The analysis was performed usingSIMNRAsoftwaretofitthespectra[8].

2.2.4.Magnetometry Measurements of the net magnetic response of the film were performedusingmagnetometry.Magnetometerscanbeusedtoobtainawealthofinformationregarding themagnetizationwith respect to temperature and field. They can alsoevaluate, to some extent, the anisotropy (angular/ structural) of the magneticresponse. Because of the possible contributions to the magnetic response fromvarioussources[9,10,11,12],includingthesubstrate,adhesives,magneticdustaswellasthesampleitself,samplemountingprocedures,samplehandlingprocedures,and the management of background signal arising from the different substratesneedtobecarefullymonitored. Samplesweremountedinplasticstrawsineitherin‐plane(strawparalleltosampleplane)orout‐of‐plane(strawperpendiculartosampleplane)configuration.Thesampleisplacedinthestrawusingtweezersandtypicallyslidtothemiddleofthestrawusingwoodencottonswabstogentlymovethesamplewithinthestrawwithoutscratchingorputtingunduepressureonthesample.Ininstancesofsmallerthan5mmx5mmsamplesitwasnecessarytousesecondarystraws,slitalongthelengthofthestraw,toprovideextracushioningbetweenthesampleedgesandtheinnerwallofthe(uncut)straw.Insuchinstances,thesampleismountedintheslitstrawfirst,andthenslid intotheuncutstraw.Whenmounting intheout‐of‐planeconfiguration itwasdifficult tomountsamples larger than3mmx5mm. Insuchinstances,weusedaseriesofconcentricallymountedpartiallyslitstrawswithnouncutstraw.InallinstancestheendswereusuallytapedwithKaptontapetoensurethesamplewouldnotbelostduringthecourseofthemeasurement,andtoensurethat the slit strawswould not become loose from the assembly. Using this strawmethod,withminimal use of Kapton tape, string or othermounting devices,wasimportantforminimizingunwantedmagneticimpuritiesthatcouldadverselyaffectthe measurement. Clean straws provided a uniform low signal diamagneticbackgroundthatwas,inmostinstances,easilysubtractedout For the measurements presented in this dissertation, the most effectivemethodforcomparingthemagneticresponseofsamplesreliedoncarefulhandlingandmountingofthesamplecombinedwithasimplelinearbackgroundsubtractionofthesubstratesignal.Thismethodyieldsthemostconsistentresultsandmeansforcomparing films grown at different times, on different substrates, with differentamounts of silver paste. Since there existed a linear region in all measurementspresentedinthisdissertationbetween+3T5Tand‐3T ‐5Twith identicalslopes,thisaverageslopewassubtractedfromtherawsignal.

27

Thismethodpresumesnodiamagneticorparamagneticcontribution to thesignalexists fromthe film(ifso,suchcomponentsaresubtractedoutby themainbackground subtraction). Thismethod also presumes that any components of therawsignalfromthesubstrate,remnantsilverpaste,ordustarealldiamagnetic.Onoccasion ferromagnetic impurities fromsourcesoutsideof the film (contaminatedsubstrates,poor tweezerhandling,etc.)mayappear in thesignal, inwhichcase, itbecomes necessary to repeat the entire experiment with new samples. Baresubstrateswere alsomeasured to confirm that the signal for a clean high qualitysubstrateisdiamagneticandlinear. ThefilmmagnetizationmeasurementsinthisdissertationweremadeusingaQuantum Design Superconducting Quantum Interference Device Magnetometer(SQUID)equippedwitha5TeslaDCmagnetandacryostatformeasurementsdownto4.2Kusingliquidhelium.TheDCSQUIDheadallowsforhighresolutionmagneticfluxdetectionwiththeabilitytomeasuremagneticsignalsabove~2x10‐7emu.

2.2.5.X‐rayAbsorptionSpectroscopy(XAS) X‐rayabsorption(XA)spectroscopyperformedinthisdissertationmeasuresX‐rayabsorptioninthesampleasafunctionofenergyandgivesinformationabouttheelectronicstructureofthefilms.ByusinganelementspecificenergyrangeoftheincidentX‐ray,differentinitialcorestateelectrons(arisingfromdifferentelementsinthesample)areexcitedtoafinalstateelectronconfiguration[13].Thistechniquecanbeused togarner informationabout theelectronconfiguration,hybridization,andcoordinationoftheelectronsinthefilms.XAintensityisthedifferencebetweenincident ( ) and “transmitted” ( ) X‐rays through the sample due to the energy‐dependentXA coefficient ( ), and is givenby . For very thin samples(surfaces) or for electron escape depths much smaller than the XA length, totalelectronyieldmeasurementsgiveadifferencethatisdirectlyproportionaltotheXAcoefficientgivenby~ .Thus,onecaneasilyobservechangesintheXAintensityin total electron yield by detecting changes in the current coming off the sampleusingapicoammeter,thoughoneshouldalwaysbewaryofsaturationeffects[14].These effects are a result of the decreasing X‐ray penetration depth as energy issweptthroughtheabsorptionedge.Whenthisdepthdecreasestotheorderoftheelectron escape depth, (~2‐5 nm) distortion of the spectral features can occurmakinganalysisdifficult. IngeneraltheX‐rayabsorptionprocessisafirst‐ordertransitioninducedbyan incident electromagnetic radiation between some initial state and final state.Using first‐order perturbation theory in the dipole approximation ( ≫

), we can determine the selection rules for the induced transition byconsidering ~| | | | ~| | ∙ | | where and are the initial and finalelectronic states, and is theunit photonpolarization vector and is the electronposition vector. Thematrix elements (| | ∙ | | ) for the electron states canbefurther separated into the radial and angular components while the dot productgivesthepolarizationdependentdipoleoperatorandisrelatedtothepolarization

28

ofincidentX‐rays(horizontallinear,verticallinear,rightcircular,leftcircular,etc.).The non‐zeromatrix elements have been calculated by Stohr [13] and the dipoleselectionrulescanbereadoffbyinspection.Theselectionrulesallowfortransitionswhere ∆ , ∆ 0, 1, ∆ 0, ∆ 0 and so the X‐ray absorptionspectroscopycanbeusedtomeasuretheelementspecificvalenceshellproperties.Specifically,wecanusesoftX‐raystoprobetheallowedelectrontransitionsfromO1s2ptransitions,transitionmetal2p3dtransitions,andrare‐earthmetal3d4ftransitionsinoursamples. In order to obtain a spectrum that is independent of sample geometry orpolarizationdirectionandsensitivetothetotalnumberofholesinthevalenceband,itisimportanttotakeanaveragedspectrumoverallincidenceangles(twoin‐planeand one out‐of‐plane) or over all polarizations (right circular, left circular, andlinear).Ifinformationregardingthesample’sstructuralsymmetryisknown,wecanreduce the number of scans needed. For example, for magnetic tetragonal singlecrystals with ,measurements need to be taken with the E‐field both in‐plane and out‐of‐plane if using linearly polarized light or at a constant X‐rayincidence angle using both left and right circular polarization (or, equivalently,alternatingfieldandasinglepolarization).Ifthisisdone,themeasuredspectrumisisotropic and the information it gives is void of any anisotropic magnetic orcrystallineeffects.Inthisdissertation,theXAspectraaretypicallytakenasthesumofrightandleftcircularlypolarizedX‐rays(~60%polarizedatBeamline6.3.1and~90% polarized at Beamline 4.0.2). The X‐rays are absorbed at a grazing 30˚sampleincidenceangleoranormal90˚sampleincidenceangle.

2.2.5.1.X‐rayMagneticCircularDichroism(XMCD) XMCD probes the element specific magnetism of a material through spindependentelectronictransitions.Inordertoprobethemagnetism,XAspectroscopyisperformedwithcircularlypolarized incidentX‐rayradiation.ByusingcircularlypolarizedX‐raystheabsorptionprocessbecomesspindependentsincetheangularmomentoftheincidentphotonscanbeusedtopreferentiallyexcitespin‐uporspin‐downelectrons.BymeasuringthedifferenceinXAspectrawithincidentleftorrightcircularly polarized X‐rays in a constant magnetic field, we can determineinformationregardingthespinoftheholesinthevalenceband. Whenthephotonangular momentum, Lph, is collinear to the sample magnetization,M, the XMCDsignalismaximized.Alternatively,thedifference,ordichroism,canbemeasuredbyalternatingtheappliedmagneticfield,andhencethemagnetizationdirection,whilekeeping the photon spin direction (or angular momentum) constant. For all themeasurements presented here, the difference of the absorption spectra, oftenrepresentedby ↑↓ ↑↑,istakeninpositiveandnegativefields(thearrowsindicatetherelativeorientationofphotonspinandmagneticmoments). ThespindependentX‐rayabsorptionprocesscanbethoughtofasoccurringin two steps. In the first step, the angular momentum of the incident photon ispartially transferred through spin‐orbit coupling to the spin of the excitedphotoelectron. For transition metals this electron is excited with opposite spin

29

polarizations from the 2p3/2 and 2p1/2 spin orbit (l+s and l‐s, respectively) splitlevels. The next step involves detection (absorption) of the spin‐polarizedphotoelectron in the spin‐up and spin‐down levels of the valencebandwhich aresplitbytheexchangeenergy.Becausetheselectionrules(seeprevioussection)fortheX‐rayabsorptionprocessdoesnotallowforspin‐flipprocesses,onlytransitionsto levels with the same spin polarization in the valence band as the excitedphotoelectron are allowed. The majority and minority spins of the valence bandcontributeoppositesignstothedichroismsignalaswell.Tosummarize,thesignofthemeasureddichroismwillbeoppositefor(i)excitationsfromdifferentspin‐orbitsplitcorelevels(the2p3/2and2p1/2),(ii)theuserightcircularlyandleftcircularlypolarizedX‐rays,and(iii)valencebandlevelsfilledspin‐upelectronsandspin‐downelectrons. Also as a final note, the transfer of photon angular momentum tophotoelectron angularmomentumhas both spin and orbital components. Thus, ifthevalencebandhaseitheraspinororbitalmomentitcanbedetectedbytheXMCDeffect. For the XMCD spectroscopy used in this dissertation, samples weremeasuredusinga1.5T(Beamline6.3.1)or0.5T(Beamline4.0.2)appliedmagneticfieldandcircularlypolarizedX‐rays(~60%polarizedatBeamline6.3.1and~90%polarizedatBeamline4.0.2).BymeasuringthesamplewithX‐rayincidentalongthe90˚normalangleorata30˚grazingincidenceangletothesample,wecanmeasureelementspecific informationregardingthespinandorbitalmoment in thesamplealongthein‐planeandout‐of‐planedirections.Inaddition,itispossibletomeasurethe element specificmagnetization as a function of applied field in the sample bysweepingthemagneticfieldataconstantX‐rayphotonenergycorrespondingtothepeakintensityinthedichroismsignal.

2.2.5.2.X‐rayNaturalandMagneticLinearDichroism DichroismusinglinearlypolarizedX‐rays,withtheX‐rayelectricfieldvectorpointedalongtwodifferentcrystallographicdirectionsrelativetothesamplecanbeusedtoobtainspecificinformationrelatedtothesample’selectronicandmagneticanisotropies. Linear dichroism arises from any non‐spherical charge distribution.Thisdistributioncanbeduetoanisotropiesinbondingandsymmetryofthesamplecrystalorfromastrongspin‐orbitcouplingwhichmayleadtoadeformation(andanisotropy)ofthechargedensitywhenthematerialmagneticallyorders. TounderstandtheXAprocessforlinearlypolarizedX‐rays,weconsidertheE‐field interactionwith the initial state (core) and finals state (valence) electrons.Since the sumof the filled core states results in a spherically symmetric electrondensitydistributionthecontributiontotheintensityfromtheseinitialstatesisthesame for all X‐ray E‐vector orientations. Thus, themain source for differences inintensityofanXAmeasurementwithlinearlypolarizedlightisentirelydeterminedby thespatialdistribution(seeFigure10)of theemptyvalencebandorbitals thataccepttheexcitedphotoelectron(whichisexcitedalongtheE‐vectordirection).TheXAintensitywillbemaximumwhentheE‐vectorforthelinearlypolarizedX‐rayisalignedalongtheorbital,anditwillbezeroiftheholedensityofthevalenceshell

30

orbital oriented along the E‐vector direction is zero. This effect, absent of anymagnetic field contributions to the dichroism, is known as X‐ray natural lineardichroism (XNLD). The experimental XNLD set‐up used to determine the holedensityinthevalencebandofoursamplesisshowninFigure10.Bymeasuringthesampleina30˚grazingincidentangle,verticalandhorizontallinearlypolarizedX‐rayscanbeusedtoprobein‐planeand(predominantly)out‐of‐planeholedensity,respectively. Under this geometry the difference between vertical and horizontalpolarizationcorrespondsto:

° sin 60 sin 60

34

34

X‐ray magnetic linear dichroism (XMLD) measures the change in the chargeanisotropiesofthesampleinducedbyexchangeandspin‐orbitinteractionsrelativetoaneasymagneticaxis.Theeffectisinducedbyuniaxialspinalignmentandcanbefound in both ferromagnets and antiferromagnets below the ordering transitiontemperature.

Figure10Spatialchargedistributionford‐orbitalsinacubicoctahedralcrystalfieldandgeometryofX‐ray LinearDichroismmeasurement probing spatial chargedistribution of valence band holes ingrazingincidence

31

2.2.6. Scanning Tunneling Electron Microscopy/Electron Energy LossSpectroscopy(STEM/EELS) Sample microstructure for some samples was probed using ScanningTunneling ElectronMicroscopy (STEM) at the electronmicroscopy center at OakRidge National Laboratory (ORNL) in collaborationwith Dr. Maria Varela. In thisdissertationwe employ thismicroscopy technique to obtain valuable informationconcerning samplemicrostructure in cross sectional specimens. In addition, localelectronic and chemical information was obtained through Electron Energy LossSpectroscopy(EELS)inconjunctionwithSTEM. Conventional transmission electron microscopy (TEM) uses coherentlyscatteredBragg beams from the electron‐beam illuminated sample to produce animage.Becauseitusesthewavelengthoftheacceleratedelectrons(~300keV),withtheproperopticsandimagingtoolsthistechniqueisabletoresolvefeaturesdownto ~1‐2 Å. In general, contrast is generated by selecting for the direct beamelectrons or the scattered electrons transmitted through the sample [15]. This isdonebyusinganobjectiveaperture toselect forelectronscoming fromthedirectbeamonly (bright field imaging) or the scatteredbeamonly (dark field imaging).The contrast can come from themass of the atoms, thickness (number of atomsscatteringtheelectrons),orcrystallinefeaturesinthefilm(Braggscattering).Highmassandhigherdensityregionsinthesamplewillscatterelectronsmorestronglythan low mass, thinner regions. In bright field imaging, this will cause fewerscatteredelectronstomakeitthroughtheapertureleadingtoadarkerappearancein the image.Thereverse is true indark field imagingmode,sincetheaperture isplacedtoselectforthesehigheranglescatteringevents. In STEM/EELS an electron beam is rastered (scanned) over a very thin(“electrontransparent”)sampleandadirectbeamimageisproduced(brightfield)or a high‐angle annular dark‐field (HAA‐DF) detector is used to preferentiallycapture a large number of the incoherently scattered beams from the sample toproduceahigh resolutionZ‐contrast image [16].Unlike in conventionalTEM, thisdetector helps to minimize the contrast from coherently scattered Bragg beamsfrom the sample. In this way, the contrast in the image from the sample can beassuredtoarisefromchangesinatomicnumber(Z‐contrast)only. TheEELSdetectormeasuresthechangeintheenergyoftheelectronsafterscatteringwiththesampletoproducespectrathatareproducedfromfollowingthesamedipoleselectionrulesasXA.InEELStheenergylossisalsoameasureoftheelectronenergyabsorptionbythesample.TheinformationobtainedissimilartoXAspectra, thekeydifferencebeingtheuseofX‐raysor theuseofelectronstocausecore‐valencetransitionsinthesample.AlsoinEELSthespectracanbetakenfromaspecific column of atoms in the sample, albeit with lower energy resolutioncomparedtoXAspectra. InperformingSTEMandEELSmeasurements,oneshouldalwaysbewaryoftheimagingofsecondaryeffectsfromsamplepreparationorbeaminduceddamage[17].Duringsamplepreparationprecautionmustbetakentonotexposethesampleto undesired high temperatures or solvents that could adversely affect the

32

propertiesoftheelectron‐transparentTEMsample.Thesecanbeintroducedduringrequiredannealingstepstosetadhesivesorsolventwashingstepstoremoveexcessadhesives.Unwantedpropertiescanalsobeintroducedduringthefinalion‐millingsteps,whichdestructivelysputterthefilmwithfocusedArionsinordertoobtainacleanandelectron‐transparentsample. Beamdamage from inelastic collisions used to image the film can typicallyarisefromelectronscatteringinducedbond‐breakage,specimenheatingfrombeamexposure and sputtering or atom displacement from the surface of very thinsamples. The scattering induced bond‐breakage, or radiolysis, in some oxidematerialscanleadtoananionvacancypairedwithacationinterstitialdefectbytheKnotek‐Feibelmanmechanism.ThismechanisminvolvestheformationofejectedO+ionsaftertheAugeremissionofoxygenanionelectronsduetothedecayofacationelectrontoadeepcoreleveluponbeamirradiation. Thesecanbereducedthoughnotaltogetherpreventedthroughtheuseofhighervoltagesandthinnersamplestoreduce the chances of electron‐electron interactions (sample thickness<mean freepathofinelasticelectrons)inthefilm.Ininsulatingmaterialsbeamheatingcanalsopose a serious problem, but can be made negligible through the use of thinnersamples and higher voltages. In thiswork, precautionswere taken to ensure thatsampleswerethinenoughthatconsiderablebeamheatingdidnotoccur.Sputteringis stronglyrelated to thestrengthof thebonds in thesample,butwithhighbeamvoltagesinverythinsamplesitcanoccurreadily.However,atenergiesevenlowerthanthoseneededforsputtering,atomscanstillbemovedaroundbytheelectronbeam, especially in the presence of pre‐existing vacancies. This makes chemicalcharacterizationmoredifficult.Althoughhighvoltageandthinsamplescanreducebeamheatingproblems,theycanresultindamagedependingonthesource.Thus,inmany samples some amount of beam damage is unavoidable. In the images andspectra gathered for this work, we have taken steps to minimize these effectswheneverpossible. In summary, STEM can provide much of the same information asconventional TEM, although the images in STEM can be assured to be Z‐contrastimages. InadditionwecansimultaneouslymeasuresupportingEELSspectra fromspatiallyspecificregionsinthesample. In this chapterwe have reviewed some of the procedures for growth andcharacterization used in carrying out our oxide thin film investigations.We haveoutlinedsomeofthenecessarystepsthatneedtobeperformedtoensureconsistentPLD growth from run‐to‐run. Also, some of the known issues from thecharacterization techniques (XRD, AFM, RBS, SQUID magnetometry, XAspectroscopy,STEMandEELSthatcouldbeasourceforinaccuraciesandanomaliesinourresultshavebeenconsideredsothatwecanbeassuredofthedata’svalidity.The atomically precise growth techniques and advanced characterizationtechniquespresentedhereenableawide‐rangeof studies, andarewell‐suited forourendeavors.Thesetechniquesarequitepowerfulandwitheachnewgenerationof the instruments used for the types of growth and characterization techniques

33

presentedhere,wewillbeabletopavethewayforthecomingeraofnewadvancedmaterialswithnovelfunctionalproperties.

34

Chapter3:StructuralEffectsfromEpitaxy

3.1. Growth and structural characterization of epitaxially strained LaCoO3films

3.2. Stabilizing tetragonally‐distorted LaCoO3 films using coherent epitaxialstrain

3.3.OxygenvacancyorderinginSTEM/EELS

3.4.Oxygendependencestudies

3.5.Summary

AbstractThe goal of this chapter is to demonstrate the structural modifications possiblethrough heteroepitaxy and the effects of stoichiometric deviations from the bulkequilibriumstructure.Inparticular,weshowthesuccessfulstructuralmanipulationusingepitaxyintheLCOsystem.IdemonstratethedegreeofcontrolovertheLCOstructurethatispossibleviahetero‐epitaxyonvarioussubstratesbygrowingfilmsofdifferentthicknessesandunderdifferentgrowthconditions.Themicrostructuraldetails observed using STEM and EELS attest to the defect related changes thatstrain can cause in the film. The effects of hetero‐epitaxy on the magnetic andelectronicstructurewillbediscussedinchapters4and5.

35

Chapter3:StructuralEffectsfromEpitaxy3.1. Growth and structural characterization of epitaxially strained LaCoO3films

LCO films were grown on STO, LSAT, and LAO substrates under similarconditionsinalargerangeofthicknesses(8nm‐133nmthick)usingpulsed‐laserdepositionwitha248nmKrFlaser(~1J/cm2)pulsedat3Hz.AcomparisonofthelatticemismatchforthesefilmandsubstratematerialsisgiveninFigure11.Filmsgrown on STO and LSAT impart tension in‐plane and films on LAO impartcompressionin‐plane.FilmsonSTOhavethelargestlatticemismatchof~2.4%.

Figure11GrowthschematicshowingdifferentepitaxialstrainconditionsforLaCoO3filmsgrownonsingle crystalline oxide substrates. Percentages shown on the line are the latticemismatch valuesbetweensubstrateandfilmlatticeparameters.

The films were grown at 700˚C in 320 mTorr of O2 and cooled in an

atmosphereof1TorrofO2.ThicknessesweredeterminedusingX‐rayreflectivity,andthestructureofthefilmswasdeterminedbyX‐raydiffraction.15nmand75nmthick samples on LAO, LSAT, and STO substrates were measured at Oak RidgeNationalLaboratoryusinghighresolutionScanningTunnelingElectronMicroscopy(STEM) andElectronEnergy Loss Spectroscopy (EELS) to obtainedmoredetailedlocal structural and chemical information. Finally, a set of films was grown in arangeofoxygengrowthpressures(10mTorr‐320mTorr)onLAO,LSAT,andSTOtofurtherstudythestructuraleffectsofnon‐stoichiometryinthefilm.

3.2. Stabilizing tetragonally‐distorted LaCoO3 films using coherent epitaxialstrain Aftergrowth,detailedinformationonthefilms’crystallinityandin‐planeandout‐of‐plane lattice parameters were determined using X‐ray diffraction. The

36

crystallinequalitywasassessedbytherockingcurvemeasurements.InFigure12,the∆, or full‐widthhalfmaximum(FWHM),wasused to compare8nm‐133nmthick filmsonSTO(Figure12(a))andLAO(Figure12(b)) substrates.For filmsonSTO the crystallinity is excellent below 50 nm thick, and AFM confirms smoothsurfaces with RMS <0.3 nm. At greater thicknesses on STO the mosaic spreadincreases considerably, as films tend to crack during cooling, possibly from thethermal expansion mismatch between STO and LCO [1]. The formation of twinboundariesinLAOiswellknowntooccuratthegrowthtemperaturesusedinthisstudy(700˚C).Inthe8nmsample,twopeaksarevisibleinfortheLCOfilm.ThisisaresultoftheunderlyingLAOtwinstructure.Regardless,themosaicspreadwithineach twinned grain appears to be relatively low (∆~0.126˚). The crystallinityworsensatalowerthicknesscomparedtoSTO(broadeningat30nm,insteadof50nm),possiblyduetoincreasedformationandmigrationoftwinboundariesinLAOwithincreasedexposuretohightemperaturesduringthelongergrowthtimes.The002filmpeakoverlapswiththesubstratepeakinthe‐2measurementsforLCOfilms on LAO above 70 nm thick, making an accurate determination of thecrystallinity difficult for these thicknesses. In general, the crystalline qualitydecreaseswithincreasingfilmthickness.

Figure12rockingcurvesmeasuredforthe002filmpeaktoassessthemosaicspreadinthefilmson(a)STOand(b)LAO.Themosaicspreadincreaseswithincreasingthickness.

The‐2X‐raydiffractionmeasurementsonthe002peak(Figure13)revealthat structure evolves toward a relaxed bulk‐like LCO structure as a function ofincreasingthickness.Thethinnestfilms(6nmand8nm)showbroad,lowintensitypeaks due to the low number of diffracting planes in these samples (fewerconstructiveinterferencescatteringevents).Asaresultwehavedifficultyobtainingaccuratelatticeparametervaluesforthesethicknesses.The2 values(andlatticeparameters) in the figurewere obtained after fitting aGaussian to the filmpeaksandusingtheGaussianpeakposition.Thehighcrystallinity isconfirmedinthe15

37

nm and 33 nm films by the presence of thickness Kiessig oscillations generatingsatellitepeaksoneithersideoftheBraggpeaks.Infilms75nmandthickerthefilmpeaksareobviouslybroadenedindicatingthatthefilmstructuretakesonarangeofout‐of‐plane latticeparameters as the film tends to relax. 15nm thick films showout‐of‐planelatticeparameterssignificantlychangedrelativetobulk~3.81Å.FilmsonLAO(Figure13(c))showa largerout‐of‐plane latticeparameter~3.85Åwhilefilms on LSAT (Figure 13b)) and STO (Figure 13(a)) have out‐of‐plane latticeparametersdepressedwithrespecttobulkLCOwithvaluesof~3.80Åand~3.77Å.Thicker films on LAO and STO show out‐of‐plane lattice parameters that clearlyrelax towardbulkLCO. Intriguingly,out‐of‐planepeaks for thick filmsonLSATdonotshowanobviousshiftin2 butbroadenwithincreasingthickness,suggestingthat even with these higher thicknesses a large portion of the film maintainsepitaxialstrain.

Figure13‐2rockingcurvesaround002peakshowingevolutionofout‐of‐planelatticeparameterasafunctionofthickness.

Tobetterunderstandtheoverallstructuralchangesasafunctionofthicknessonthedifferentsubstrates,wecomparedreciprocalspacemaps(RSMs)takenatthe013 peak for ~15 nm and 95 nm thick films (Figure 14). Thin films on all threesubstrates have sharp uniform peaks showing an increased (decreased) in‐planeandadecreased(increased)out‐of‐planelatticeparameteronSTOandLSAT(LAO)with respect to bulk LCO (Figure 14 green circle), consistent with the ‐2measurements in Figure 13. Thick films all show different degrees of partialrelaxation of the film peak towards bulk LCO values, with films on LSAT (LAO)appearingleast(most)relaxed.Thispartialrelaxationsuggeststhattheappearanceof a constant out‐of‐plane lattice parameter as a function of increasing thicknessfromthe‐2measurementsonLSATismisleading.LSATfilmsareinfactrelaxing,thoughtoalesserdegreecomparedtoSTOandLAOfilms.Inaddition,allthickfilmsshow a severely broadened peak indicating increases in the mosaic spread andlateralcorrelationlengthinthefilm,consistentwithourobservationsfromscans(Figure11).

38

Figure14Reciprocalspacemapsshowing013peakreflectionsforthe~15nmand~100nmfilmsgrownonSTO,LSAT,andLAO.

The in‐plane (open squares) and out‐of‐plane (closed squares) latticeparameters for the 15 nm, coherently strained films on all three substrates aresummarized in Figure 15. The ratio of out‐of‐plane lattice parameter to in‐planelattice parameter (c/a) of the LCO film is calculated for growth on differentsubstrates. As thickness is increased, the LCO films evolve from a tetragonallydistortedlatticetooneapproachingc/a~1.Thus,throughanappropriatechoiceofsubstrateandthickness,filmscanbegrownwithvariousaspectratios.Weachievec/a~1.014withcoherentlystrainedfilmsincompressiononLAO,andc/a~0.978andc/a~0.968forthecomparablefilmsstrainedintension.Thecrossoverfromc/a>1 to c/a <1 occurs at 3.82 Åwhich is slightly higher than the bulk pseudocubicvalueofLCO~3.81Å.

39

Figure15Summaryoftetragonalstructuresstabilizedincoherentlystrained15nmthickfilms.

We also gauged whether the unit cell volume was preserved in thesedistortedfilmsbycomparisontobulkLCOunitcellvolume(55.306Å3).Table1listsaverage value for in‐plane and out‐of‐plane lattice parameters (obtained fromRSMs)forthreesetsofcoherentlystrainedfilmsgrownlessthan15nmthick.Inthefinalthreecolumnsweassessthevolumechange,tetragonaldistortion,andratioofvolume change to tetragonal distortionusing the components of the strain tensorthat correspond to these changes in in‐plane and out‐of‐plane lattice parameters.For thesecalculationswehaveassumedxx=yy, since the filmsaregrownon001substrates.We find that filmsonLSATandSTOwhichshowsignificant tetragonaldistortion (2zz‐xx‐yy,seventh column) also show somedegreeof volume change(xx+yy+zz,sixthcolumn).FilmscompressivelystrainedonLAOonlyshowalargechangeinthetetragonaldistortion,whilethevolumechangesverylittle. The final column in Table 1 gives a quantity that allows us to assess howmuchthevolumechangesforagivenquantityoftetragonaldistortion.Remarkably,wefindthatdespitethesmallertetragonaldistortionforfilmsonLSATcomparedtoSTO, the corresponding volume change is the largest. This shows that there is adifferencebetween themodesbywhich strain is accommodated in filmsonLSATcompared to filmsonSTO, and suggests that either (i) theoctahedral rotations infilmsonLSATresultinagreaterincreaseintheCo‐O‐Cobondanglescomparedto

40

filmsonSTOor(ii)thereisanincreasedpresenceofotherdefectsthatincreasethevolumeoftheaverageunitcellinthesefilmscomparedtofilmsonSTO.Table1AveragestructuralvaluesdeterminedfromX‐raydiffractionreciprocalspacemapsforthreedifferentsetsoffilms<15nmthick.Allvaluesasidefromthefirsttwocolumnsarecalculatedvalues.

Substrate(001)

IP(Å)OOP(Å)

c/a V(Å3) ∆Vorxx+yy+zz(%)

2zz‐xx‐yy(%)

xx+yy+zz(%)/2zz‐xx‐yy(%)

STO 3.899 3.767 0.965 57.256 3.53 ‐7.02 ‐0.503LSAT 3.871 3.790 0.979 56.802 2.70 ‐4.27 ‐0.632LAO 3.79 3.845 1.015 55.242 ‐0.11 2.93 ‐0.023

3.3.OxygenvacancyorderinginSTEM/EELS WhileX‐raydiffractionsuggeststhattheepitaxialstrainisaccommodatedinthe film through changes in bond‐length or coordinated octahedral rotations toexpand or contract the lattice in a given direction, scanning tunneling electronmicroscopy reveals microstructural details that suggest the strain may beaccommodated through structurally ordered changes in oxygen stoichiometry(Figure 16). Ordered linear planes of dark contrast were found on 15 nm thicksamplesonLAO,LSAT,andSTO.Thesedarkcontrastplanesappearedperpendicularto the substrate‐film interface in films on LSAT (Figure 16(b)) and STO (Figure16(c)) andparallel to this interface in filmsonLAO (Figure16(a)).Also, thedarkcontrastplanes’occurperiodicallyafterevery3‐4brightcontrastplanesonSTOandLAO, and after every 2‐3 bright contrast planes on LSAT. Similar dark contrastplanes have been observed in other Sr‐doped cobaltite films and have beenattributedtooxygenvacancies[2,3,4],cationvacancies[5],andstructuraltwinning[6,7]. Figures16(d,e,f)showtheO‐KedgeEELSlinescanspectratakenalongthedark (black line) and bright (red line) contrast planes. The spectral differencesbetweenbrightanddarkcontrastplanesareeasilydiscernibleonLAOandSTO.Bycontrast,thespectraldifferencesbetweendarkandbrightcontrastplanesonLSATaredifficult todiscernbecauseof the largenumberofdarkcontrastplanes in thissample and the limited spatial resolution of the EELS line scan. In general, thebrightcontrastplaneshavehigherO‐Kedgeintensities.FilmsonLAOandSTOshowanobviousshoulderpeakat~527eV.ThedarkcontrastplanesalsohaveahigherL3/L2 peak intensity ratio at the Co‐L edges (Figure 16(g, h, i)). These spectraldifferencesbetweenbrightanddarkcontrastplanessuggestthatthedarkcontrastplanes areoxygendeficient planes of ordered oxygen vacancies. Thus, LSAT filmscontainthemostvacancies,followedbySTO,whileLAOhasthefewestvacancies.

41

Figure 16 Scanning tunneling electron microscopy images in bright field imaging mode showingperiodicdarkcontrastplanesparallelandperpendiculartothesubstrate‐filminterfacefor~15nmthick filmsofLaCoO3.Co‐Ledge (g,h, i) andO‐Kedge (d, e, f)EELS spectra from line scans takenalong thebright (red)anddark(black) contrastplanes indark field imagingareshownbelowthecorrespondingimages.

Oxygenvacanciesoftenresult inexpansionof the lattice[8]andmayaid inalleviationoftensilestressesinthefilmfromepitaxialstrain[3,4].TheperiodicityandorientationofthevacancyplanesindicatethatthevacanciesarearrangedinthefilmtorelievethestressesfromgrowthonSTO,LSAT,andLAO.Thusvacanciesfilltheelongateddirectionsinthefilm:in‐planeforfilmsonSTOandLSAT,out‐of‐planeforfilmsonLAO.

42

Figure17 Scanning tunneling electronmicroscopy images taken in bright field imaging for 75nmthickfilms.

ThedensityofdarkcontrastplanesdecreasesinthethickerLCOfilms,whichalsosuggeststhattheoxygenvacancyorderingisstrainrelated.STEMinFigure17showsthatinadditiontoadecreaseindensityofdarkcontrastvacancyplanes,thevacancyplaneorientation ismoredisordered.Vacancyplanesorientedalongbothin‐planeandout‐of‐planedirectionscanbeobservedin75nmfilmsgrownonLAOandSTO.Bycontrast,the75nmthickfilmonLSATdoesnotshowclearevidenceofvacancyordering.

Figure18ThepresenceofoxygenvacancyorderedplanesaftereffectsfrombeamexposureinfilmsonLAO,LSAT,andSTO.

As a final note, it may be important to mention that observation of thesedefectplanesdoesnotalwaysoccur,andappearstobeaffectedbycertainunknownvariable conditions related to sample prep, sample charging, electron beamexposure, or properties intrinsic to LCO samples. In Figure 18, films on LAO andLSAT show evidence that the defect ordering appears more pronounced afterelectronbeamexposure.Thissuggeststhatthesestoichiometricdefectsarealwayspresentinthesample,butorderunderelectronbeamirradiation.Therelativespeedwithwhichthedefectsorganizeintoplanesmayberelatedtothemobilityofsuchdefects in films grown on LAO, LSAT, and STO. This suggests that these oxygenvacanciesmigrateeasilyinfilmsonSTO(wherethedefectplanesarealmostalways

43

observedwithoutneedforincreasedbeamexposuretime),andwithsomedifficultyin films on LSAT (where the ordering is only observed after >1 hour of electronbeamexposure).Thisalsosuggeststhatthesedefectsaremigratinginthesampletoregions that are high strain or space‐allowing [3, 4], considering the larger radialsizeassociatedwiththesedefects(oxygenvacancyorCo2+).

3.4.OxygenDependencestudies Since the microstructural features observed in the films suggest a strongpresenceofoxygenvacancies,weexplored theroleofoxygendepositionpressureon the film structure.We grew~30nm thick films on LAO, LSAT, and STO in 10mTorr‐ 320mTorr of O2. Films grown in the lower oxygen pressures tend to beslightlythickerandsmoother(lowerRMSroughness)duetoreducedscatteringofthe plumewith oxygen background gas during growth.However, the crystallinitydoesnotappeartochangesignificantlywithoxygenpressure.Figure19showsthenormalized∆rockingcurvescansforfilmsgrownonLSATin20mTorr,70mTorr,and 200 mTorr of oxygen taken for the 002 film diffraction peak. Despite thechanges inoxygengrowthpressure, thecrystallinityremainshigh(∆<0.1˚)andsimilartotheotherfilmsofcomparablethickness.SimilartrendswereobservedforfilmsonSTOandLAO.

Figure19∆ for~30nmthickLaCoO3 filmsgrownonLSAT.Nochange in theFWHMisobservedovertheoxygenpressurerangestudied.

Theout‐of‐planelatticeparameterwasfoundtobesignificantlyincreasedinall films grown at 10 mTorr compared to films grown in 320 mTorr. Figure 20showstheout‐of‐planelatticeparameterchangesmeasuredforthefilmsgrownon

44

all three substrates.Above100mTorr theout‐of‐plane latticeparameter changedonly slightly showing the smallest values at 320 mTorr. Under epitaxial strainconditions fora stoichiometric film,weexpect theout‐of‐plane latticeparametersforfilmintension(LSATandSTO)tobelessthanLCObulk(~3.81Å)andforfilmsincompression(LAO)tobegreaterthanLCObulk.Thestructureoffilmswellabove75mTorrchangesaswemightexpectforstoichiometricfilmsunderepitaxialstrain.Thesmallchangestostructureunderincreasingpressurebeyond75mTorrsuggestthatevenwith320mTorrofO2thefilmsmaynotbefullystoichiometric,althoughthelatticeparameterappearstoleveloffinthisrange.Despitethesedifferences,thesimilar∆ suggests that thestructuralchangesareaccommodatedwithout loss intheoverallcrystallinequalityofthefilm. The in‐plane lattice parameters did not change between the 10mTorr and320mTorr30nmthickfilmsamples,indicatingthatallstructuralchangesresultingfromO2pressurechangesoccurintheout‐of‐planedirection.Figure21showstheRSMs for strained30nm filmson STOgrown in10mTorr and320mTorr.Apartfromthechangeintheout‐of‐planelatticeparameter,nochangeisobservedinthestructureofthefilm.RelaxedfilmsonLAOshowasimilarexpansionexclusivelyintheout‐of‐planelatticeparameterwithdecreasingoxygengrowthpressure.

Figure20Changeinout‐of‐planelatticeparameterwithchangesinoxygengrowthpressure.Theout‐of‐planelatticeparameterislowestinthe320mTorrsamples.

The LCO unit cell volume increases as a function of decreasing oxygendepositionpressure,whichmaybeexplainedbythepresenceofoxygenvacanciesatloweroxygendepositionpressures [8,9].Oxygenvacanciesmayarise topreserveequilibrium during the deposition process; the loss of the neutral oxygen atom

45

leaves extra electrons that may hop to nearby Co3+, forming larger Co2+ cationswhich in turn expand the lattice. Additional lattice expansion can occur fromdisplacements of nearby atoms to accommodate the oxygen vacancy site, withcations displacing away from the site and oxygen ions displacing towards thevacancysite[8].Theselatticeexpansionsandatomicdisplacementshavesignificantimplications in themagneticorderingdue to theireffecton theCo‐O‐Comagneticexchangemechanism.

Figure21Reciprocal spacemaps for filmsgrownonSTOandLAO in10mTorrand320mTorrofoxygen.LAO30nmfilmsarerelaxedandSTO30nmfilmsappeartobestrainedatthesethicknesses.

3.5.Summary In this chapter we have shown that the choice of a lattice mismatchedsubstrate is effective in altering the structure of thin LCO films. At thicknesses of~35 nm and beyond the film structure appears to show evidence of latticerelaxation towards bulk LCO values ~3.81 Å in both in‐plane and out‐of‐planedirections.WhileXRDsuggeststhatbondlengthsandoctahedralrotationsmayberesponsible for the changes in structure inLCO, STEMsuggests that the substrateinducedtensilestressesandstrainsmaybeaccommodatedbynucleationofordereddefectplanesalongtheelongationdirectioninthefilm.Ontheotherhand,attemptsto introduceoxygen‐relateddefects in the film throughchanges inoxygengrowthpressureshowthatthedefectscausechangesintheout‐of‐planelatticeparameter

46

only, even in relaxed films on LAO. This is surprising since in STEM we haveobservedoxygenvacanciescanbeaccommodatedinbothin‐plane(STOandLSAT)and out‐of‐plane (LAO) directions, but defects from oxygen induced growth mayonlybeoccurringintheout‐of‐planedirections.Inthenextchapterthesestructuralchanges are related to trends in the magnetic behavior of LCO films. Detailedanalysis of the electronic structure of the LCO filmswill be presented in thenextchaptertobetterillustratetheconnectionbetweenstrainandferromagnetism.

47

Chapter4:InducingferromagneticexchangeinepitaxialLaCoO3

4.1.Substrateandstraindependencestudies

4.2.Micro‐structure/Electronicstructure

4.2.1.STEM/EELS

4.2.2.SpectroscopicCharacterization

4.2.2.1.XASpectroscopy

4.2.2.2.XMCD

4.2.2.3.XNLD

4.3.Oxygengrowthpressureeffectsonferromagnetism

4.4.CospinstatesandvalencestatesinLaCoO3films

4.5.Theongoingexplorationoftheferromagneticexchangemechanism

AbstractIn this chapter the surprising observation of ferromagnetism in epitaxial LaCoO3thin filmsnotpresent inbulkLaCoO3 isdemonstrated. Inorder tounderstandthepossibleoriginsof long‐rangeorder inLCOfilms,magnetismiscorrelatedwith(i)epitaxial strain, (ii) stoichiometry, and (iii) microstructure. A simple model isdevelopedtohelpexplainhowthemodificationsofelectronicstructureinthefilms,caused by epitaxial strain and non‐stoichiometry can give rise to spin statetransitionsandinturnstabilizeaferromagneticgroundstate.Ifindthatcoherentlystrainedfilmsintensionhavethehighestmomentwhilecoherentlystrainedfilmsincompressionarenotmagnetic.Analysisofthemicrostructurerevealsthatagreaternumber of oxygen defects are found in filmswith a highermagneticmoment. XAreveals that the differences between films grown under different conditions areassociatedwithvariousfractionsofHSCo2+,HSCo3+,andLSCo3+.XMCDandXNLDshowthatthefeaturesintheelectronicstructureareassociatedwiththemagnetismand the growth under different conditions. Films grown in lower oxygen growthpressureshavesuppressedmagneticorder.

48

Chapter4:InducingferromagneticexchangeinepitaxialLaCoO3 Our efforts to manipulate the spin state through heteroepitaxial strain(Chapter3)have resulted in ferromagnetic filmsof LCO.This result suggests thatepitaxialstrainisindeedaviabletoolforcontrollingthespinstateinLCO;however,the possible role of defects induced during growth cannot be neglected. In fact,determining the precise source of ferromagnetism is still an open question incobaltitethinfilms[1,2].Manygroupscitetheimportanceofepitaxialstrain[3,4,5],whileothershaveimplicatedtheroleofchemicalinhomogeneity[1,2,6].Inallinstances,acohesiveexplanationofthepreciseexchangemechanismunderlyingtheferromagnetismbetweenhighspinandlowspincobaltorcobaltindifferentvalencestates, has yet to be presented. In particular, scientists currently have littleunderstanding regarding the specific differences between films grown underdifferent conditions that explain the differences inmagneticmoment.MoreworkneedstobedonetofirmlyestablishwhethertheCospinstateorthevalencestateismore closely involved in the ferromagnetism. To accomplish this, we carry out amore thoroughdeterminationof the structural and stoichiometricmake‐upof thedifferent magnetic and nonmagnetic LCO films. By exploring the relationshipbetweenmagnetism, strain, and stoichiometry in detail usingmicrostructure andelectronicstructurecharacterization,weuncoverthecrucialbuildingblocksneededtoexplainthelong‐rangemagneticorderinthissystem.

4.1.Substrateandstraindependencestudies Due to some of the confusion in the literature regarding the magnetismdependencefoundforfilmsgrownondifferentsubstrates[3,4,7],itisnecessarytoproperly establish the relationship between magnetism and strain. The trendsrelated tothe thinandthickLCOfilmsgrownondifferentsubstrates indicate thattheferromagnetismisaresultofepitaxialstrainintension.However,acloserlookat the growth on the different substrates and to different thicknesses show thatincreasing the strain does not necessarily equate to higher magnetic moments.Figure 22 shows the SQUIDmagnetometry results for films grown~15 nm thick(labeledas“strained”)and75nmthick(labeledas“relaxed”)onSTO,LSATandLAOsubstrates. Recalling the structural information from Chapter 3, we know thestructureofthesefilmsshowsvaryingamountsoftetragonaldistortionandvolumechangedependentonthethicknessandthelatticemismatchofthesubstrateusedtoinduce epitaxial strain. The presence of ferromagnetism is consistently found infilmsstrainedintension.Figure22showsthatcoherentlystrainedfilmsintensionon LSAT and STO both show evidence of ferromagnetic order. However, it isinterestingthat15nmfilmsonLSAT,whicharelessstrainedintension(comparedtofilmsonSTO),havethehighestsaturatedmomentof~2.15µB/Co(Figure22(b)).By increasing the tetragonal distortionandepitaxial strainby growthon STO,wefindalowersaturatedmomentof~0.85µB/Co(Figure22(a)).Ontheotherhand,15nm thick films strainedonLAOdonot showobvious indicationsof ferromagnetichysteresis (Figure 22(c)) or Curie temperature (Figure 22(d)) indicating that

49

strained films grown in compression on LAO do not show signs of long‐rangeferromagnetic order and that epitaxial strain in tension does play a role in theferromagnetism.

Figure 22Magnetization vs fieldmeasurements at 5 K for thin (dark) and thick (light) LCO filmsgrown on (a) STO, (b) LSAT, and (c) LAO substrates showing a range of saturatedmoments. Themagnetization vs temperaturemeasurements underH=5mT (FC) for (d) strained and (e) relaxedfilmsshowaCurietemperaturebelow~85KforallfilmsexceptstrainedfilmsonLAO.

Incontrast,all~75nmthickfilmsshowferromagneticbehaviorbelow~85K. For the films on LAO and STO at these thicknesses, the relaxation in thecoherently strained 15 nm film structure results in a comparatively lower netmoment.However,justasweobservedinthetrendsfor15nmthickfilms,eveninthese75nmthicksamples,LCOonSTOhasalowermoment(~0.6µB/Co)comparedtoLCOfilmsonLSAT(~0.95µB/Co).Figure23(a)showshowthesaturatedmomenton STO decreases as a function of increasing thickness. Surprisingly, while thecoherently strained films in compression on LAO did not show any magnetism,relaxationoftheepitaxialstrainonthickerfilmsresultedinferromagneticordering.In these thicker films, although the strainand tetragonaldistortion ispresumablymuch smaller, the magnetism is greater. This suggests that additional factorsbesides strain need to be considered in explaining all the details of theferromagnetism. The trend of increasing remanent moment with increasingthicknessisplottedinFigure23(b)forLCOfilmsonLAO.Theremanentmomentisplotted instead of saturated moment to thickness, due to the difficulty indeterminingthesaturatedmomentforverythinnonmagneticfilms.

50

Figure23MagnetismtrendsasafunctionofthicknessshowingthatfilmsonSTOandfilmsonLAOdisplay an opposite trend. For films on (a) STO the saturatedmoment is plotted as a function ofthickness.Forfilmson(b)LAOtheremanentmomentisplottedasafunctionofthicknessinstead.

These results show that films grown coherently strained in tension clearlyresult in long‐rangeferromagneticorderbelow~85K.However, thedetailsof therelationship betweenmagnetism and growth on different substrates suggest thatmorefilmpropertiesneedtobeinvestigatedtoexplainallthetrends.ThedecreaseinoverallmagneticbehaviorinthickfilmsonSTOandLSATmightbeexpectedduetothestructuralrelaxationtowardsbulk latticeparametervalues.This leadsustobelievethatstraingivesrisetoferromagneticorderinLCOfilms.Inthe75nmthickfilmsonSTOandLSATthelowernetmomentcanbeattributedtosomemixtureofrelaxed film and film still strained to the substrate. However, ifwe consider thatbulk LCO at these temperatures is mostly composed of diamagnetic LS Co, theevidenceofferromagnetismintherelaxedfilmsonLAOissurprising.Thissuggeststhat though the lattice parameters in the relaxed LCO film on LAO appear to besimilar tobulkLCO, thedefectsandstoichiometryarenot similarandgive rise tolong‐rangeferromagneticorder.Alsosurprisingisthepresenceofhighermomentincoherentlystrained filmsgrownonLSATcompared tomore tensily strained filmsgrownon STO. If themagnetismwerepurely strain related,morehighly strainedfilms on the larger lattice‐mismatched substrate, STO, would have the highestmoment. Since LSAT has the highest magnetic moment, additional sources mustcontributetothemagnetism.

4.2.Micro‐structure/Electronicstructure The discussion so far suggests that multiple factors contribute to thegeneration of ferromagnetism. However, from chapter 3 we also know that themicrostructureof filmsgrownonSTO,LSAT,andLAOshow intriguingdifferenceswhichmayberelatedtothechangesinmagnetism.Forgreaterunderstandingofthe

51

roleofepitaxialstrainanddefects,weexploretheconnectionamongtheelectronicstructure and the microstructure using STEM and EELS measurements, (asdescribedinChapter3)andthemagnetism.Byexaminingtheelectronicstructureindetail using XAS, XMCD and XNLD measurements, we can explore the specificelemental contributions to the electronic structure and gainmore insight into thesourcesofmagnetisminthesesamples.

4.2.1.STEM/EELS Thepresenceof ferromagnetism in tensilely strainedLCO films, andnot incompressively strained ones less than 15 nm thick, indicates that epitaxial straininduces lattice distortions which play some role in stabilizing ferromagnetism.However,thepresenceofdefectsobservedinSTEMandEELSalsoassociatedwithgrowthondifferent substrates suggests that a careful considerationof the role ofthemicrostructureisinorder.TheSTEMimagesfromchapter3showthatin15nmthickfilmstherearea largenumberofoxygenvacanciesandlikelyaconsiderablenumber of Co2+ ions distributed throughout the films. Their magnitude (assuggestedbytheirorientation)isrelatedtothesubstrateinducedstraininthefilmsandmay also have effects on themagnetic ordering.Wehave assumed that filmswithdefectsorientedparalleltothefilmplanehaveafewertotalnumberofdefectssince they only occur up to the thickness of the film. Films with defectsperpendiculartothesubstrate‐filminterfacelikelyoccurinmuchgreaternumberssincetheyextendacrosstheentirelaterallengthofthesample.Forexample,15nmfilmsonLSATwhichhavethehighestmagneticmomentalsoshowthemostdefects,asindicatedbytheiroccurrenceevery2‐3unitcellsperpendiculartothefilmplane.15nmfilmsonSTOwhichhaveslightlyfeweroxygendefectplanes(every3‐4unitcellsperpendicular to the filmplane)havea lowermoment compared to filmsonLSAT. Finally, 15 nm films on LAO which have the fewest oxygen defect planesoccurringalongthefilmplaneevery3‐4unitcellsdonotshowferromagneticorder.Thecorrelationbetweenthenumberofoxygendefectplanesandthemagnitudeofmagnetisminthe15nmthickfilmssuggeststhattheirpresenceplaysaroleinthemagnetism. Supporting this trend,we observe that the changes in abundance of defectplaneswiththicknessmayalsoberelatedtothechangesinmomentwithincreasingfilmthicknessforthefilmsincompressionandtension.Forexample,whilein15nmfilmsonLAOonlydefectplanesparalleltothesubstrate‐filminterfaceareobserved,in75nmfilmstheplanesappearbothparallelandperpendiculartothis interface,suggesting an overall increase in the total number of defects, in the thicker LAOfilms. The increased magnetism in thicker LAO film, suggests that there is aconnection between magnetism and the greater number of defects orientedperpendicular to thefilmplane. Thetrendsobservedwith increasingthickness infilmsonLSATandSTOalsosupportthisview.The75nmfilmonSTOshowfewerdefectplanesperpendiculartothefilmplaneandmoredefectplanesparalleltothefilmplane,representinganetdecreaseinthenumberofdefects.Inthe75nmfilmonLSAT,wedonot observe anydefect planes. Thus, in contrast to films on LAO,

52

thicker filmsonSTOandLSATshowadecrease indefectplanesperpendicular tothe film plane, and a net decrease in the number of defects. The increase in thenumberofdefectplanesinthickerfilmsonLAOandthedecreaseinthenumberofdefect planes in thicker films on STO and LSAT is consistent with the increasedmomentinthickerfilmsonLAOandthedecreasedmomentinthickerfilmsonSTOandLSAT. Ontheonehand,theconnectionbetweentheobservationofdefectplanesinSTEMandEELSandthetrendsobservedinmagnetismsuggeststhatthepresenceofCo2+ ionsoroxygenvacanciesplayan important role in themagnetism.However,we cannot rule out thepossibility that thepresenceof defect planes ismerely anindicator of greater local bond‐length and bond‐angle changes in the film whichcouldalsobeequallyimportanttothemagnetism.Forexample,thelackofobviousdefects in the thick ferromagnetic film on LSAT indicates that the defects are notnecessarily required to stabilizing long‐range ferromagnetic order, though theirpresencemayenhanceitinthe15nmfilms.Itisalsonoteworthythatinimagesoffilms on LAO and LSAT the defect planes often only appear after considerableelectronbeamexposure(seeChapter3).Whileitispossiblethattheelectronbeamcangenerateoxygenvacanciesinsomerarecases[8,9,10],itishighlypossiblethatbeam irradiation induces migration of oxygen vacancies throughout the sample[11]. In such instances, it would seem that the location to which these defectsmigratearemostlikelyregionswithstressinducedbond‐lengthchangesinthefilm,thustheparallelorientationin15nmfilmsonLAOandperpendicularorientationin15nmfilmsonLSATandSTO.

4.2.2.SpectroscopicCharacterization Althoughwehavefoundsomeintriguingcorrelationsbetweenthenumberofdefects and themagnetism of these films, we have still not addressed how theseproperties may influence the magnetic behavior. In order to more preciselyunderstandtheroleofoxygenvacanciesandinturntheroleofCo2+versusHSandLSCo3+inthemagnetism,thedistributionoftheseionsmustbedetermined.UsingXA spectroscopy techniques, we assess the relative number of holes in the d‐orbitals, the spin state, the valence state, and the element‐specific source of themagnetismasafunctionofthestrain.

4.2.2.1.XASpectroscopy ForthischaracterizationfilmsweregrownonNb‐dopedSTOtopreventtheinsulatingfilmsfromchargingatlowtemperaturesduringtotalelectronyieldmodemeasurements. Measurement of the films on all other substrates at lowtemperaturesresultedinsignificantsamplechargingduetotheirinsulatingnature,preventing credible analyses of the spectra. In fact, our attempts tomeasure thetransport properties of the film using a 4‐point Van der Pauw [18]method showinsulating resistance versus temperature behavior above 200 K, and are too

53

insulating for credible measurement using our measurement set‐up. In any case,film growth on Nb‐doped STO, circumvented our sample charging problem andprovided a sufficient electron source for the total electron yield XA spectroscopymeasurements. WecomparetheXAspectraofathinandthickfilmtoexploretheelectronicstructurechangesassociatedwiththeepitaxialstrainrelaxationinthefilm.Figure24showsXAspectrafor15nmand70nmthickfilms.Theintensitiesinthefigurearethesumofspectratakenunder+0.5Tand‐0.5TwithcircularlypolarizedX‐raysincidentata60˚anglefromthesamplesurfacenormal.Inthisgeometry,theappliedfield and X‐ray polarization E‐vector are pointed (mostly) along the in‐planedirectionofthefilm.AfternormalizingthespectratothepeakCoL3edgeintensity(black arrow), obvious spectral differences between the strained (15 nm) andrelaxed(70nm)thickfilmcanbediscerned(orangeandgreenarrows).Specifically,the15nmfilmshowsahigher intensityratioofpre‐edgeshoulder intensity(longorangearrow)topost‐edgeshoulderintensity(shorterorangearrow)attheL3peakcomparedtothe75nmthickfilm.AsimilarchangeisalsoobservedintheintensityandratioofpeaksattheL2edgeofthespectra.

Figure 24 (a) XA spectra for films grown 15 nm (red) and 70 nm (blue) thick on Nb‐doped STO.Arrows indicate relevant featuresdescribed in the text. (b)Asetof referencespectra forHSCo2+[19],HSCo3+[20],andLSCo3+[20]areprovidedtoanalyzethefilmspectra.

In order to interpret the differences in features between the samples, wecomparethesamplespectratoreferencespectraofCoinvariousvalencesandspinstate configurations (Figure 24(b)). The reference spectrum from CoO representsHSCo2+inanoctahedralcoordination[19],whiletheSr2CoO3ClandEuCoO3spectrarepresentpyramidalHSCo3+andoctahedralLSCo3+,respectively[20].Thepre‐edgeL3 feature at ~777 eV (green line) in the reference CoO spectrum [19] is not

54

reproducedineitherofthereferenceCo3+spectrafromHuetal.[20]whichsuggeststhatanyintensityatthisenergycanonlybeassociatedwithsomepresenceofCo2+in the sample [21]. Thus, the strained and relaxed films of LCO on STO show anindicationofsomeamountofHSCo2+presentinthefilmassuggestedbythesmallintensity shoulder near ~777 eV (Figure 24(a), green arrow). The intensitydifference at this energy between the two films indicates that the strained filmshows an increased presence of HS Co2+ compared to the relaxed LCO film. Thisresult is consistent with the accommodation of strain by the presence of oxygenvacanciesinthefilm(seeChapter3). Wealsotookacloseralookatthespectraofthe15nmand75nmLCOfilmswith reference toHS andLSCo3+spectra. Since these two spectra in Figure24(b)represent 100% HS Co3+ (Sr2CoO3Cl) and 100% LS Co3+ (EuCoO3), they serve aseffective unique “fingerprints” of the different spin states in the XA spectra.Although the spectra show Co3+ in two different coordinations (pyramidal versusoctahedral),wecanstillusethemtogetanindicationofthefeaturesassociatedwithHSandLSCo3+spinconfigurations[20].ThereferencespectraindicatethatLSCo3+has a higher post‐edge shoulder intensity at the L3 edge and higher pre‐edgeshoulderintensityattheL2edge.Incontrast,HSCo3+showstheoppositetrendwithrelatively higher pre‐edge shoulder intensity at the L3 edge and relatively higherpost‐edge shoulder intensity at the L2 edge. In the relaxed LCO film the relativeratiosneartheL3andL2edgesaresimilartotheLSCo3+.Thestrainedfilmsshowanincreasedintensityatthepre‐edgeshoulderfeatureoftheL3edgeandadecreasedintensityatthepost‐edgeshoulderfeatureoftheL3edge,thusindicatingsomeHSCo3+. TheL2 edge features of the strained films show intensity ratios that suggestsomepresenceofHSCo3+. According to the reference spectra, theHS Co2+ andHS Co3+ have featuresthat overlap in the pre‐edge region of the L3 peak (778 eV‐779 eV). One mightnaivelybelieveitispossibleforthespectralfeaturesinstrainedfilmstoarisefromHS Co2+ and LS Co3+ contributions only. However, the decreased intensity at thepost‐edge shoulder feature of the L3 edge, also associatedwith HS Co3+, suggeststhat at least some of the XA intensity for the strained films must arise fromcontributionsofHSCo3+.Thus,whencomparingstrainedandrelaxed filmsgrownonNb‐dopedSTOweconcludethat: (i) inboth filmsat leastsomeamountofCo2+exists(strainedfilmsmayhavemore),(ii) inrelaxedfilmswefindmostlyLSCo3+;and(iii)instrainedfilms,LSCo3+andatleastsomeamountofHSCo3+exist.

4.2.2.2.XMCD Having established the presence of Co ions in various spin states via XAspectra, we now directly correlate the presence of these Co spin states with theferromagnetismusingXMCD.Thistechniqueprobesthespectralchangesassociatedwiththechangesinappliedmagneticfield.Thus,wecanqualitativelyassesstheroleoftheseconstituentCoionsinthelong‐rangemagneticorder.Figure25showstheXMCDat25Kfroman8nmstrainedfilmanda75nmrelaxedfilmgrownonNb‐doped STO measured below the magnetic ordering temperature. The largest

55

dichroism intensity appears near the peak labeled (b) in the corresponding XAspectrum.However,alltheHSCo3+andLSCo3+featuresdiscussedsofarintheXAspectrumcanbeassociatedwithfeatures inthemagneticdichroism.ThesmallHSCo2+ feature does not show a strong contribution to the magnetic dichroismintensity.ThissuggeststhattheHSCo3+andLSCo3+contributionsaretheprimarysource for magnetism in the strained LCO films on STO. In the 75 nm film thedichroismintensityislowercomparedtothe8nmfilm,consistentwiththefindingthatthenetmomentperCoionfromthe75nmfilmislowerthanthatofthe8nmfilm from SQUID magnetometry. Since the XA measurement is highly surfacesensitive, the lower netmomentmeasured from SQUIDmagnetometry cannot beassociated with a few strained layers closest to the substrate‐film interface, butmustarisefrom(partiallyrelaxed)layersatthesurfaceofthefilmaswell.

Figure25XA(toppanel)andcorrespondingXMCD(bottompanel)spectrafor8nmand75nmfilmsgrownonNb‐dopedSTO.

4.2.2.3.XNLD Room temperature XNLD was used to compare differences in electronicoccupation of the Co d‐orbitals for strained films on STO, LSAT, and LAO, and arelaxed filmonLAO.Bymeasuring the lineardichroismat room temperature,weare able tominimize charging effects from the insulating samples, andwe ensurethat theobserved lineardichroismis frombondinganisotropiesandnotmagnetic

56

anisotropy in the film. With the sample at 30˚ grazing incidence, vertical andhorizontal linearly polarized X‐rays can be used to probe in‐plane and(predominantly)out‐of‐planeholedensityoftheCoionelectronconfiguration. Figure 26 shows the orientation‐averaged XA spectra and correspondingXNLD intensity. The XA spectra (Figure 26(a)) for films grown on differentsubstratesshowsubtleintensitydifferencesattheL3andL2edgesassociatedwithHSCo2+,HSCo3+,andLSCo3+featuresdiscussedpreviously.Inparticular,LSATfilmshave the highest intensity at the pre‐edge shoulder associated with HS Co2+,followed by STO, relaxed LAO, and, finally, strained LAO. This trend is consistentwith findings of increased oxygen vacancies for LSAT films from STEM and EELSstudies.TherelativepresenceofHSCo3+ionsfollowsthissametrend,withfilmsonLSAThavingthelargestintensityatthepre‐edgeL3feature.

Figure26(a)XAandXNLD(c)spectraforfilmsgrownonLAO(red),LSAT(green),andSTO(blue).Thefigureontheright(b)isanenlargedregionoftheCo‐L3edgeshowingthespectralchangesfromgrowthondifferentsubstrates.

The XNLD also shows unique substrate‐dependent differences. The mostsignificantdifferencesareindicatedbythedashedredlinesinFigure26(c).Wefindthese differences between films grown in tension on STO and LSAT and films incompressiononLAO.FilmsrelaxedonLAOshowsimilaritiesinthelineardichroismtobothstrainedfilmsonLAOandstrainedfilmsonLSATandSTO.WhilemanyofthesedichroismdifferencesmaybeduetodifferentcontributionsfromHSCo2+,HSCo3+, and LS Co3+ [21], lattice distortions may also influence the anisotropies inchargedensity[22].Fromchapter3weknowthatfilmsgrownonLAOhavec/a>1and films grown on LSAT and STO have c/a <1. Thus, we might expect that theconstituentCoionsinthefilmadoptanadditionalanisotropicchargedensityduetothedistortionsoftheCo‐Ooctahedralcageunderepitaxialstrain.Forexample,thestrong linear dichroism feature near the L2 edge in all samples can only beaccounted for by a contribution from tetragonally‐distorted LS Co3+, since linear

57

dichroismfromHSCo2+,HSCo3+,andundistortedLSCo3+inthisenergyrangeisnotsignificant[21]. Together, XA spectroscopy, XMCD, and XNLD indicate that theferromagnetismisprimarilyassociatedwith increasedpresenceofHSCo2+andHSCo3+ in epitaxial films. XMCD shows that more HS Co3+ is associated withferromagneticorder.XNLDshowsthatthereisacleardifferenceintheanisotropichole density in the strained (and relaxed) films indicating that lattice distortionshaveasignificantimpactontheelectronicstructure.Whilemoresimulationworkisneeded toproperly fit the spectra todeterminequantitatively theamountof eachioninthevariousfilms,thesestudiessuggestthatachangeintherelativequantityofHSCo2+,HSCo3+,andLSCo3+resultsinachangeinthemagnitudeofthemoment.

4.3.Oxygengrowthpressureeffectsonferromagnetism While STEM and EELS results have suggested that the density of oxygendefects are related to themagnitude of long‐rangemagnetic order, we have alsofound an overall suppression of the ferromagnetism inLCO filmswith decreasingambient oxygenpressureduringdeposition. Figure 27 shows the change in Curietemperature for films grown on STO substrates in different oxygen pressures. Asgrowth pressure is decreased the Curie temperature decreases. Similar decreasesare observed in the remanence and saturation magnetization of the films withdecreasingoxygengrowthpressure.Thistrendsuggeststhatthepresenceofdefectscaused by growing the films in a lower background oxygen pressure results insuppressionof long‐rangemagneticbehavior.Aswefound inchapter3, the loweroxygenpressuresalsoresult inanexpansionoftheout‐of‐planelatticeparameter.In Figure 27, we see the magnetism (Curie temperature and saturated moment)decreases as the structure increases in size with decreasing oxygen growthpressure. To better gauge the electronic structure effects associated with growth inhigh and low growth pressures, we performed surface sensitive XMCDmeasurements on samples grown in 10 mTorr and 200 mTorr of O2. Figure 28showsthedifferencesinXAlineshapeandXMCDintensitybetweenasamplegrownin 10 mTorr and 200 mTorr on Nb‐doped STO. Intriguingly, the L3 pre‐edgeshoulderfeatureassociatedwithCo2+ionsandthepre‐edgefeatureassociatedwithHSCo3+ionsinthefilmhaveincreasedintensityinthe10mTorrsample,suggestingagreaterpresenceofCo2+ ionsandpossiblyHSCo3+ ions inthefilm.Recallingthereferencedata fromFigure24(b),we know these two ions have spectral featuresthatoverlap,sodeterminingtherelativeincreaseoftheseionsinthefilmgrownin10mTorrofO2isdifficult.Surprisingly,theincreaseinHSCo2+andHSCo3+ionsisnotassociatedwithan increase inmagnetism from theCo ion in this film.On thecontrary,thelargestmagneticdichroismintensityisobservedinfilmsgrowninthehighestoxygenpressures,confirmingtrendsobservedinSQUIDmagnetometry.Thissuggests that either, (i) there are some inherent shortcomings in using thisfingerprintingmethod to qualitatively interpret the relative number of Co ions inthefilm;or(ii)thatdespitetheincreasedpresenceofHSCo2+and(possibly)HSCo3+

58

(i.e. ions with moment) the mechanism of long range ferromagnetic ordering isdisrupted by the presence of oxygen growth related defects. For example, theincreasedpresenceofdefectscouldcauseasuppressionoftheexchangemechanismincertaindirectionsoranincompletepercolationofmagneticmixedCoionclustersthroughoutthefilm.

Figure 27 Curie temperature and saturated moment (left axes) as a function of oxygen growthpressure for films on STO. The out‐of‐plane lattice parameter decreases with increasing oxygenpressure(rightaxis).

The decrease in magnetism with lower oxygen pressure suggests that theoverlapofCoandOorbitals involved in the indirectexchangemechanismmaybedisruptedinthefilmsgrowninloweroxygenpressuresduetothelocationofvacantoxygensitesinthefilm.DespitetheincreasedHSCo3+andHSCo2+presenceinthesefilmsgrowninloweroxygenpressure,themagneticmomentandCurietemperaturearereduced.ThesefindingsareinapparentcontradictiontotheresultsfromSTEMandEELSwhichshowthatthehighestmomentinfilmsgrownonLSATandSTOareassociatedwithincreasedoxygenvacancyplanes.Howeveritappearsthatitisnotthemerepresenceofoxygendefectsbut the elongationof theout‐of‐plane latticeparameterthatcanbecorrelatedwithferromagnetism.Coherentnonmagneticfilmson LAO and films grown in lower oxygen pressure on any substrate have anelongatedout‐of‐planelatticeparameter.

59

Figure28XAandXMCDspectra for filmsgrownonNb‐dopedSTO in10mTorr (orange) and200mTorr(black)ofoxygen.

4.4.CospinstatesandvalencestatesinLaCoO3films Theapplicationofepitaxialstrainappearstohavemanydifferenteffectsonthe films, allofwhichmayplayan important role in suppressingorgiving rise tolong‐range magnetic order. At the very least, we can conclude that the epitaxialstrainresultsinatetragonaldistortionofthelattice.Thestraininthefilmappearsto be correlated to the presence of oxygen vacancy defects and changes in theamountofHSCo3+ionsinthefilm. Togetherthestoichiometric,structural,andmagnetictrendsofLCOfilmsinthese two chapters (3 and 4) allows us to construct a qualitative picture of theexchangemechanismthatmayberesponsible for the ferromagnetismobserved inthese LCO films. There are four key trends to consider. (1) Films strained withexpandedin‐planelatticeparametershaveamomentof~2µB/CoonLSATand~1µB/Co on STO. Films strained with contracted in‐plane lattice parameters do notshowlong‐rangemagneticorder.(2)Thehighestmomentweobserve(strainedLCOfilms on LSAT) is associated with the greatest number of oxygen vacancy planes(every2‐3unitcellperpendicular to the filmplane). Nonmagnetic films(onLAO)

60

areassociatedwiththefewestoxygenvacancyorderingplanes(every3‐4unitcellsorientedparalleltothefilmplane).(3)Inducingdefectsthatresultinanelongationof theout‐of‐planedirectionusing loweroxygengrowthpressuressuppresses theferromagnetisminthissystem.(4)Despitethevariations insaturatedmoment,allmagneticfilmsareinsulatingandhaveaCurietemperature<85K.

Figure 29 Schematic representation of Co unit cells and corresponding electron configurationsassociated with growth on STO and LSAT substrates showing (a) bulk LS Co3+, (b) tetragonally‐distortedHSCo3+,(c)tetragonally‐distortedLSCo3+,and(c)tetragonally‐distorted,oxygen‐deficientHSCo2+.

Figures 29 and 30 show a set of qualitative electron configurationmodelsbasedonsomelikelyeffectsofthelatticedistortionsandoxygenvacanciesontheCoenergy levels. The figures show illustrative schematic representations for:unstrainedpseudocubicLCOinaLSstate;HSCo3+underastrain‐inducedtetragonaldistortion;andHSCo2+alsoinatetragonally‐distortedoctahedralcrystalfield.Thisdoesnot includeallpossiblespinandvalencestateconfigurationspossible fortheCo ion under these strain conditions, but addresses the most likely scenariossuggested from the results. These scenarios are to serve as rudimentary buildingblocksforenvisioningaferromagneticexchangemechanismbetweenthemixturesofCospinandvalencestatesinthefilms.Thesediagramsneglectthefactthatinacrystalwithamixtureof ionicandcovalentbonding,theseenergylevelswouldbebetter visualized as electron bands than discrete isolated energy levels.We haveassumedapseudocubic octahedral ligand fieldwhere theCo‐Omolecular orbitalsare split into the three‐fold degenerate nonbonding t2g and two‐fold degenerateantibondingeg*shells.Forsimplicity,intheHSandLSschematics,theexchange‐splitspin‐upandspin‐downlevelshavebeendrawnonaredundantenergyscaleforalleg*andt2gorbitals,whileinamoreaccuratedepictionwemightexpecttheminority‐spin (spin‐down) electron levels in HS states to be higher in energy than themajority‐spin(spin‐up)eg*levels.

61

Figure29(a)showstheelectronicconfigurationforundistortedbulk‐likeLSCo3+ in pseudocubic LCOwhich has no net spinmoment (S=0) and behaves as adiamagnetic insulator at low temperatures in the bulk. Figures 29(b‐d) show theschematic change in symmetry of the LCOunit cell due to presumedbond‐lengthdistortions induced by epitaxial tensile strain. As a result of the decreased z‐axislength and increased x,y‐axis length, the eg*and t2gorbital degeneracy is furtherreduced and the octahedral Oh symmetry is reduced to D4h symmetry. Theantibonding and nonbonding molecular orbitals, which have mostly transitionmetal‐liked‐orbitalcharacter,shiftinrelativeenergyposition.Duetothepresumedbond length distortions from the epitaxial strain, d‐orbitals with x,y‐axiscomponents have decreased overlap with neighboring oxygen orbitals. Thedecreasedoverlapresultsinarelativedecreaseinenergyoftheseantibondingandnonbonding molecular orbitals. Another electron configuration possibilityintroducedintheseinvestigationsistheHSCo2+ion,whichmaybepresentduetothe observed oxygen vacancies in the film (Figure 29(d)). XNLD suggests that adistortedlatticefromepitaxialgrowthisalsolikelytobefoundfortheCo2+ions.TheelectronconfigurationforCo2+inaHSstateisverysimilartoCo3+inHSstatewithonemoreelectroninthet2gsubband.

Figure 30 Schematic representation of Co unit cells and corresponding electron configurationsassociatedwith growth on LAO substrates showing (a) bulk LS Co3+, (b) tetragonally‐distorted LSCo3+,(c)tetragonally‐distortedHSCo3+,and(c)tetragonally‐distorted,oxygen‐deficientHSCo2+.

The model that explains the ferromagnetic exchange mechanism in thismixed spin/mixed valence system must also account for the absence of aferromagneticmomentinstrainedfilmsonLAO.Figure30showsthesetofpossibleCospinandvalencestates fortetragonally‐distortedfilmsofLCOonLAO.ThekeydifferencebetweenthesepossibilitiesandtheonesforfilmsonLSATandSTOistheshiftinenergypositionofthedifferentd‐orbitalsduetothedifferentlatticestrain.InfilmsonLAOthepurein‐planeorbitalsdxyanddx2‐y2orbitalsarehigherinenergy

62

due to the shorter in‐plane bond lengths. This changewould cause only a subtledifferenceintheeg*electronfillingbetweenthefilmsonLAOandfilmsonLSATandSTO.However,inthet2gorbitalsthisresultsinalowerenergyforthedegeneratesetoforbitals, thedxz anddyz orbitals, compared to filmsonLSATandSTO.Thismayplayaroleinthedifferencesinmagneticbehavior,sincetheholesinthet2gshellinthefilmonLAOoccupythenon‐degeneratedxy level,whileincontrasttheholesinfilmsonLSATandSTOexistexclusivelyinthedegeneratet2gsubband,

4.5.Theongoingexplorationoftheferromagneticexchangemechanism The ferromagnetic exchange mechanism in the films on LSAT couples amixtureofHSCo3+(S=2,~4µB/Co),LSCo3+(S=0,~0µB/Co),andHSCo2+(S=3/2,~3µB/Co) to result in a net moment in the film of ~2 µB/Co. Films on STO showferromagnetismmostlikelythroughasimilarexchangemechanism,butinthecaseof STO the net moment is only ~1 µB/Co. STEM and EELS have shown that theoxygenvacancycontentsoftheseLSATandSTOfilmsaredifferent,mostlikelyduetothechanges instrainandlatticedistortionsbetweenthesetwofilms.Thus, it islikely that themixture of Co ions in these various spin and valence states is alsodifferent, accounting for the difference in thenetmoment. In a similarway, filmsgrownonLAOalsocontainauniquedistributionofCo ionsasrevealedby theXAspectraandSTEMandEELSmeasurements.InstrainedfilmsonLAO,theexchangeis suppressedamongthedifferentCo ionsand long‐rangemagneticorderingdoesnotoccur. It isstillunclearwhetherthesuppressionof long‐rangeorderiscausedbythisdifferentmixtureofCoionsorthroughsomeothermeanssuchasthelatticedistortions felt by these ions.However, in thickermore relaxed filmsonLAO, themixture,distortion,andorganizationofCoionsinthefilmchangestoallowforlong‐rangeorderingtooccur. Thus, the future forLCOfilmstudiesstillholdssomeinterestingchallengesandopenquestions.Theunderstandingofoxygenvacancyformationandmigrationinperovskiteoxidesisafieldheavilystudiedbythefuelcellcommunity[23]inthecontextofiontransportandoxidation‐reductionreactions,butitsimportanceinthestudy of thin film ferromagnetism is only beginning. Until we can accuratelymeasure oxygen content and oxygen‐related effects, the role of these defects incorrelatedelectronsystemswillalwaysbeachallenge.Tounderstandthecomplexinteractions between the numerous Co spin states in LCO films, wemust pursuetheoreticalwork.SimulationofmultipleteffectstodeconvolutetheCocontributionsto different XA features is a problem that can be tackled easily in the future. Thegreater underlying challenge is to develop a picture that can account for (i) thepresence of long‐range ferromagnetismwith a robust Tc, (ii) a variable saturatedmoment, and (iii) insulating behavior in LCO films. Thiswill require considerablymoreworkfromboththeoretical[24,25,26,27]andexperimentalperspectives. In conclusion, LCO is a fascinating system for epitaxial studies due to thedelicate balance between crystal field andHund’s energy that is easily perturbedthroughexternalforces.Inthisworkwehavedemonstratedsomeoftheinterestingnuances associatedwith theuseof epitaxial strain to change the spin stateof the

63

system at low temperatures. We have shown that the films display latticedistortions, strain‐correlated oxygen vacancies, and changes in valence and spinstate, and, as a result, show differences in magnetic moment and long‐rangeordering. The results presented here also show the difficulty in succinctlyattributing the cause of ferromagnetism in thin films of LCO to a Jahn‐Teller‐distortedsuperexchangeeffect[3]oramixed‐valentdefecteffect[2,6].Despitethelackofaconclusiveexchangemechanismtoexplainallofthephenomenaobservedinthesevariousfilms,itisclearthatepitaxialstrainofLCOcanbetunedtogiverisetouniquestructuralandmagneticproperties.

64

Chapter5:ExtensionstoPr‐basedCobaltite:inducingA‐siteordering

5.1.Motivationforexploringothercobaltites

5.2.GrowthandstructuralcharacterizationofepitaxiallystrainedPrCoO3films

5.3.Exploringlong‐rangemagneticorderinPrCoO3filmsonSrTiO3

5.3.1.Magnetismandepitaxialstrain

5.3.2.Element‐specificmagneticorder

5.4.NovelandemergentPr‐sublatticeordering

5.5.Conclusions

AbstractIn this chapter, I find that heteroepitaxial strain of PrCoO3 films grownon SrTiO3andLaSrAlTaO3isfoundtoinducelong‐rangeorderingoftheCoions.PrCoO3filmsgrownon SrLaAlO4 and LaAlO3 do not show any long‐range order. By specificallyexploringepitaxial filmsofPrCoO3grownonSrTiO3001substrates,weshowthatthis order is the result of a stabilization of a ferrimagnetic ground state, in starkcontrasttoparamagneticinsulatingbehaviorofobservedinbulk.Theferromagneticorderingof theCoO6array,which isdeducedtobepartially inahighspinstate, isaccompaniedbyorderingof thePr sub‐lattice inanantiparallelorientation to theCo. This ordering of the Pr sub‐lattice provides evidence for significant Co‐Prexchange, apparently facilitated by the presence of high spin Co. The long‐rangemagnetic order in epitaxial PrCoO3 is thus fundamentally different from theferromagnetism found in alkaline earth doped PrCoO3, as evidenced by the verydifferent electronic properties, tendencies to rare‐earth magnetic ordering, andspin‐state.

65

Chapter5:ExtensionstoPr‐basedCobaltite:inducingA‐siteordering

5.1.MotivationforexploringothercobaltitesThe spin state of cobaltites can clearly be manipulated by coherent

heteroepitaxial strain. Scientists have also used hydrostatic pressure [1] andchemicalpressureintheformofsubstitutionofLawithasmallerrareearthcation[2]toalterthestabilityofthelowspinstate.HydrostaticpressureincreasesthespingapbydecreasingtheCo‐Obondlengths(whichincreasesthecrystalfieldsplitting[3,4] anddecreasing theCo‐O‐Cobondangle (whichdecreases theeg* bandwidth[5].Unlikehydrostaticpressure,chemicalpressurestabilizesthe lowspinstatebydoinglittletosignificantlyaltertheCo‐Obondlength,butonlycausesareductioninthe Co‐O‐Co bond angle [2]. The net result of replacing the A‐sitewith a smallercation, as discussed in chapter 1, is the increase in the spin state transitiontemperaturewithdecreasingcationsize.

EpitaxialfilmsofPrCoO3,withPr3+ina4f2state,serveasamodelsysteminwhichwecanexplore thecompetingeffectsof chemicalpressureandalsomodifythemagnitudeofthespingap.BulkPCOisparamagneticanddoesnotexhibitlong‐range magnetic ordering. It has a smaller, orthorhombically‐distorted perovskitestructure (apseudocubic(pc) ~ 3.789 Å) compared to bulk LCO (rhombohedrally‐distorted,apc~3.81Å)duetothesmallerionicsizeofPr3+andadoptslowerCo‐O‐Cobondangles.ThechemicalpressurethusdecreasesCo‐O‐Cobondangles,leadingtoanarrowereg*‐derivedbandwidthandalarger“spingap”inbulk.ThispressureenhancesthestabilityofthelowspinstaterelativetoLCO,resultinginashiftoftheonsetofthespinstatetransitionfromabout30KinLCOtoabove200KinPCO[2].Epitaxially strained films in tension may oppose this stabilization and insteadpromotetheHSstate,astheobservationofferromagnetisminLCOthinfilmswouldsuggest[6,7,8,9,10,11,12,13].StudiesofPCOthinfilmscouldprovideafirststepin the direction of understanding the influence of the spin gap on ferromagneticorderinginepitaxialfilms.

ThissystemalsopresentsauniqueopportunitytoexplorePrionorderingincobaltites.Inundopedbulkperovskites,thePrionandtheCoionsdonotshowlong‐rangemagneticorder.Asmentionedabove,theCoionsareinlowspinstateandarediamagnetic.AnymagneticexchangebetweentheseCoionswiththePrionsistooweaktoorderthePrsublattice,andatbestonlyaweakparamagnetismisobservedforthePrionatlowtemperatures[14,15].InpartiallydopedPr‐cobaltites,wherethePr‐siteissubstitutedwithanaliovalentalkalineearthcation,suchasSr,Ca,orBa, theComagneticmoment ismodified throughachange in itsvalenceandspinstate. In these compounds, the Co ions order ferromagnetically and are inintermediate spin and intermediate valence states with an itinerant eg* electronmediating a double exchange interaction [16]. While the Pr ions show stronghybridizationwiththeOions[16],nolong‐rangemagneticorderinthePrsublatticeisobserved.ThesefilmspresentanewscenarioinwhichPrionsarepossiblyputinproximity toHSCo3+ ions, andbyusingXMCDwecanprobe theiralignmentwithrespecttotheCosublattice.

66

5.2.GrowthandstructuralcharacterizationofepitaxiallystrainedPrCoO3films ToprobetheroleofepitaxyonthespinstatetransitioninPCO,wegrewfilmson (001) LSAT (apc = 3.868 Å), LAO (rhombohedral, apc = 3.79 Å), and SLAO(tetragonal,apc=3.75Å)substrates25nmthickandon(001)STO(acubic=3.905Å,mismatch=)substrates8nm,25nm,and110nmthick,bypulsed‐laserdeposition(248nmKrFexcimerlaser,~1J/cm2).Allfilmsweregrownatalaserpulserateof3Hz in 320 mTorr of O2 at 650‐700˚C. After cooling in 10 Torr, structuralcharacterization was performed on a Philips Panalytical 4‐circle diffractometerusingaCuKαlinesource.Magnetometrywasperformedusinga5TQuantumDesignsuperconducting quantum interference device (SQUID) magnetometer. X‐rayabsorption(XA)andX‐raymagneticcirculardichroism(XMCD)measurementswereperformed in total electron yieldmode at the Advanced Light Source (beamlines4.0.2and6.3.1).TotalelectronyieldmodelimitstheprobedepthofXAandXMCDtotheelectronescapedepth(~5nmbelowthesurface). Substratesimpartdifferentamountsofin‐planestrainonthefilmdependingon the degree of lattice mismatch between in‐plane lattice parameter of thesubstrateandfilm.Figure31showsthereciprocalspacemapsforpartiallystrained25nmthickfilmsgrownonSTO(~3.1%mismatch),LSAT(~2.1%mismatch),LAO(~0%mismatch),andSLAO(~‐1.0%mismatch)substrates.FilmsonSTOandLSATarestrainedintensionandshowlatticeparametersthatdeviatesignificantlyfromthe bulk value of 3.79 Å. The film peak for the film grown on STO is broadened(Figure31(a)),suggestingthatwitha25nmthicknessthefilmispartiallyrelaxed.On the other hand, film peaks for films grown on LSAT (Figure 31(b)) and LAO(Figure 31(c)) appear sharp suggesting that these films are nominally strained tothe substrate at these thicknesses. The film on SLAO (Figure 31(d)) is mostlyrelaxedatthisthicknesstakingonvaluessimilartobulkPCOinbothin‐planeandout‐of‐planedirections.

Figure 31Reciprocal spacemaps around the 013 peak for 25 nm thick PCO films on (a) STO, (b)LSAT,(c)LAO,and(d)SLAO.

67

Only the films grown on STO and LSAT show long‐range magnetic order,whilefilmsonLAOandSLAOdonotshowclearevidenceofmagneticorder.Figure32(a)showsthemomentversusappliedfieldmeasurementsforthese25nmthickfilms,andFigure32(b)isanenlargedregionofthesamemeasurementtoshowthehysteresis for films on LSAT and STO. Films on LSAT have a higher saturatedmomentof~0.95µB/PrCoO3 compared to~0.17µB/PrCoO3 for filmsonSTO.ThefilmsonLSATandSTOalsoshowaclearmagneticorderingtemperatureof~53Kand62K,respectively(Figure32c).FilmsonLAOandSLAOdonotshowindicationsof long‐rangemagnetic order but exhibit large paramagnetic signal. In fact, thesefilmshavealargermagneticsignalthananyoftheotherfilmsat5T(thesignalfromthefilmonSLAOislargest);however,itisclearthattheremanentmomentiszerointhesetwosamples.

Figure32(a)Magnetizationvsfieldmeasurementstakenat5K forfilmsonSLAO,LAO,LSAT,andSTO. (b)Zoomed‐in regionofmagnetizationvs fieldmeasurement showinghysteresis for filmsonLSATandSTO.(c)Magnetizationvstemperaturemeasuredin5mTafterFC.

The larger moment and lower Curie temperature in PCO films on LSATcompared to STO is intriguing and suggests that lattice distortions affect thestrength of the exchange (lowerTc) and the density of high and low spinCo ions(highermoment).InLCOfilmshighermomentisalsofoundinfilmsgrownonLSATcompared to films grownon STO. In both cases the latticemismatch is larger forfilms on STO than on LSAT [8, 10]. In the case of LCO, we correlate the highermomentwiththegreaterpresenceofHSCo3+,aswellasLSCo3+andHSCo2+ ions(See Chapter 3, and 4). The 25 nm thick PCO films on STOmay be slightlymorerelaxedthanfilmsonLSAT,duetotheenormous(~3%)latticemismatch.However,thisdoesnotfullyaccountforthelowerthenetmomentinfilmsonSTO,sinceeveninamorefullystrained8nmfilmonSTO(discussedindetailbelow,Figure34)themomentislowerthanthemomentfor25nmthickfilmsonLSAT.ThelowerCurietemperature for films on LSAT can be explained by a decrease in Co‐O‐Co bondangle inLSAT filmscompared toSTO films.Adecrease inbondangle isknown todecreasetheCurietemperatureinmixed‐valentrare‐earth/alkaline‐earthcobaltites[15],so itappearsthatdespitethepartialrelaxationinSTOfilms, thebondanglesarestilllargerthanfilmsonLSAT. Finally,thestraindependenceofthemagnetisminthesefilmsappearsmoredistinct for PCO compared to LCO. PCO films strained in compression or fully

68

relaxed do not show long‐range magnetic order, unlike LCO films which showmagneticorderinrelaxedfilmsonLAOandSLAO[8,10].IthasbeensuggestedthatrelaxedLCOfilmsaremagneticduetopointdefectsandchemicalinhomogeneitiesinthefilm[12,13,17].InrelaxedandunstrainedPCOfilms,however,thesedefectsmaynotexist,ormaybeunsuccessfulinpromotinglong‐rangemagneticorder.

5.3.Exploringlong‐rangemagneticorderinPrCoO3filmsonSrTiO3

5.3.1.Magnetismandepitaxialstrain Intheremainingsections,wefocusspecificallyonPCOfilmsgrownonSTOsubstrates to explore the evolution of themagnetic properties and the electronicstructure as a functionof thicknessusing element specific techniques. Filmsweregrowntothicknessesof8nm,25nm,and105nmin320mTorrofO2.PCOfilms25nmandthinnergrownonSTOsubstratesexhibitgoodcrystallinitywithaFWHMoftherockingcurves(scans)ofthe002Braggpeaksof∆<0.3°.Thickerfilmsshowlargermosaic spreadwith ∆ up to 0.7° due to cracking during cooling, possiblyfrom the thermal expansion mismatch between STO and PCO. All PCO films alsoexhibit low surface roughness with typical RMS ranging from 2.5 Å to 5 Å on alaterallengthscaleof5microns.Atomicstepsfromthesubstratecanbeobservedinthethinnerfilms.Reciprocalspacemapsofthe(013)filmandsubstratepeaksshowthatthe8nmfilmiscoherentlystrainedtotheSTOsubstrate(Figure33(a)).AsthePCO film thickness is increased to 25 nm (Figure 33(b)), the in‐plane film latticeparameter shifts from the STO substrate values towards thebulk valueof 3.79Å,indicating film relaxation. At 105 nm thickness (Figure 31(c)) this relaxationbecomesextensive.

Figure 33Reciprocal spacemaps of a series of PrCoO3 films grown on SrTiO3 substrates showingincreasingrelaxationofin‐planeandout‐of‐planelatticeparameterstothepseudocubicbulkvalueofPrCoO3withincreasingthicknessfor(a)8nm,(b)25nmand(c)105nmthickfilms.

AllPCOfilmsonSTOexhibitedmagneticorderbasedonthetemperatureand

field dependence of the magnetization (Figure 34). In order to obtain accurate

69

magnetizationvalues,ahigh fieldnegativeslopebackgroundwassubtracted fromthedatatoeliminatediamagneticandparamagneticsubstratecontributionstotheoverall magnetic signal. The magnetization versus field loops show hysteresis,characteristicofmagneticorder.Figure34(a)showsthatthesaturatedmomentsdonot scalewith thickness.We find the highest saturatedmoment (~0.4 µB/PrCoO3formulaunit) on the thinnest filmsandobservea rapiddrop in themomentwithincreasingthickness,eventuallyfallingbelow0.1µB/PrCoO3.Figure34(b)showsthemagnetizationdependenceontemperature.TheplotshavebeennormalizedtotheirlowtemperaturevaluestocomparedifferencesinCurietemperature.The8nmand25nmthickfilmshavecomparableCurietemperaturesof~60K,whilethe105nmthickfilmshowsasignificantdecreaseintheCurietemperaturetoaround40K.Asmightbeexpectedfromstrainrelaxationarguments, thethickest filmsshowsomeevidence of broadened transition to the paramagnetic state. All films show amonotonic decrease in magnetization as a function of increasing temperature,indicativeofasinglemagneticorderingtemperatureforthemomentsinthefilm.

Figure34 (a)Magnetization versusmagnetic field at 5K of PCO filmswith variable thickness. (b)Magnetizationversustemperatureafterfieldcoolingandapplyingastaticfieldof5mTindicatesthevaryingTcasafunctionoffilmthickness.

5.3.2.Element‐specificmagneticorder Theferromagnetismwasprobedbyelement‐specificX‐raymagneticcirculardichroism(XMCD)at15Kinalternating0.5Tfield.Sincetransportmeasurementsrevealed PCO film samples to be insulating, similar to films on LCO [12], XMCDsamples were grown 8 nm and 110 nm thick on conductive Nb‐doped STOsubstratestopreventcharging.Filmsweremeasuredatgrazingincidencewiththemagnetic field applied in‐plane andmoment probed along this same direction toenable comparison with in‐plane SQUID measurements. While the Co and Pr XAspectra of the 8 nm and 110 nm PCO filmswere fairly similar (Figure 35(a) and5(c)), therewereconsiderabledifferencesintheXMCDspectra. Inanycase,XMCD

70

signalsarefoundatbothCoL2,3(2pto3d)andPrM4,5(3dto4f)absorptionedges,suggestingthattheferromagnetismcomesfrombothCoandPr.

Figure35XAspectra(a,c)andXMCD(b,d)measuredusingcircularlypolarizedX‐raysandappliedmagneticfieldsupto0.5TtakenattheCoL2,3edge(a,b)andthePrM4,5edge(c,d)shownalongsidereferenceandcalculatedspectrafrom[18,19,20,22,23].

TheXAlineshapes inFigure35(a) fortheCoedgeof the8nm(redopencircles)and110nm(greensoliddowntriangles)PCOsampleindicatethatthereisagreaterpresenceofHSCo3+ inthethinnerfilms.Forcomparison,weshowspectrafromHaverkortetal.[18]forLSlowtemperatureLCO(Figure35(a)darkblueline)and40%HS‐60%LSLCOathightemperatures(Figure35(a)pinkopentriangles).DespitethermalphononbroadeningeffectsofthelineshapeandtheconvolutionofHS and LS features in the 650 K LCOspectrum, this spectrum suggests that (i) aslightshift(~0.2eV)inthepeakpositionoftheL3edge(B)tohigherenergyand(ii)achangeintherelativeratiooftheL3(B)andL2(D)peakintensityareindicativeofgreaterfractionofHSstateCo.The8nmPCOfilmspectrumhasfeaturessimilarto(i)and(ii)suggestingthepresenceofsomeHSCo.Sincetheseeffectsaresubtle,wealsocomparethePCOfilmspectrawiththeCoL2,3spectraofSr2CoO3ClandEuCoO3(Figure35(a)orangesolidsquaresandblackexes)whichhave100%HSand100%LSrespectively[19].AlthoughthesetwocompoundshavedifferentCocoordination

71

(i.e.,pyramidalSr2CoO3ClandoctahedralEuCoO3),thereisacleardifferenceintheratioofthepre‐(A)andpost‐(C)L3edgeintensitieswithHSCohavingahigherpre‐edge (A) and lower post‐edge (C) intensity compared to LS Co. In the 8 nm PCOsample, there is an increased intensity at region A and a decreased intensity atregionCcomparedtothe110nmsample,consistentwiththepresenceofmoreHSCo.Inaddition,weseeadramaticchangeintheratiooftheL2peak(D)andL2post‐edge(E)featureintheLSandHSCoreferencesamples.TheLSCohasahigherD/Eintensity ratio thanHSCo.Weobserve a similar higherD/E intensity ratio in the110 nm thick PCO film sample. These comparisons suggest that the spectraldifferences between 8 nm and 110 nmPCO film samples in these specific energyregionsisconsistentwithmoreHSCointhe8nmsamplecomparedtothe110nmfilm. ThemagneticdichroismassociatedwiththeCoinPCOfromthe8nm(Figure35(b) red open circles) film is significantly stronger than the dichroism from the110nm(Figure35(b)greensoliddowntriangles)film.Theparamagneticdichroism

[18] associated with a small population of HS Co from applying 6 T on a singlecrystalsampleofLCOat60K(Figure35(b)darkblueline)showssimilarfeatures(thoughnot identical) to thedichroismofoursamples.Thedichroismsignal fromthe thicksample is significantly lower, consistentwithSQUIDmagnetometrydata,and as a result has a lower signal to noise ratio. However, since both dichroismsignals from the films, using the convention ↑↓ ↑↑, match the dichroism signchangesintheparamagneticLaCoO3reference,wecanconcludethattheComomentisalignedtotheappliedmagneticfield. OurXMCDspectraofthePrMedge(Figure35(c))arecharacteristicofPr3+,similartotheexperimentaldataforPr2O3[20].WedonotobserveanyevidenceofchargetransferorPr4+featuresassociatedwithbond‐lengthchangesandstrong4f‐2phybridization,ashasbeenobservedforbulkPr0.5Ca0.5CoO3initsinsulatingstate[21]. Calculated dichroism features at the M5 and M4 edges [22, 23] match thefeatures in the experimental spectrum in both relative magnitude and energyposition,butnot insign, confirming themagnetismarising fromthePr3+ ionsandtheir antiparallel orientation to the Co3+ moment and the applied field. The Prdichroism signal ismuch stronger for the 8 nm film (Figure 35(d) dark red solidtriangles) than for the 110 nm PCO film (Figure 35(d) blue open squares). Thisdifference is consistent with a larger overall magnetization from the 8 nm filmobtainedfromSQUIDmagnetometry. AcloserlookatthemagneticfielddependenceoftheXMCDsignalsattheCoand Pr edges provides insight into the switching behavior of Co and Pr. Figures36(a) and 36(b) show element specific hysteresis loops taken at 779.9 eV for Co(regionB at L3 peak) and at 953.1 eV forPr (M4peak) as a function of field. Thedichroism for the8nm (Figure36(a)) shows clearhysteresis indicating thatboththe Co andPr are ferromagnetic. The 110nm film (Figure 36(b)) shows no clearhysteresisandareduceddichroisminboththeCoandthePr,thusindicatingamuchweakerferromagnetism,ifany,inthethickersample.

72

Figure 36 By holding the incident photon energy to 780 eV for Co and to 952.5 eV for Pr andsweepingthemagneticfieldfrom0.5Tto‐0.5T,elementspecifichysteresisloopsweretakenforthin(a)andthick(b)filmsofPrCoO3onSrTiO3substrates.

5.4.NovelandemergentPr‐sublatticeordering Fromthestructuralandmagnetizationdata,itisclearthatthereisastrongcorrelationbetweentensilestrainandferromagnetisminthefilms.PCOfilmsundercoherent tensile strain have tetragonal crystal symmetry and show ferromagneticbehaviorwithCurietemperatures~60K.Asthefilmthicknessisincreasedandthestructure becomes more cubic and akin to bulk PCO, the magnetic moment isdiminishedand theCurie temperaturedecreases.Combinedwith theXAevidenceforhighspinCointhethinnerPCOfilms,thistrendinthemagnetismsuggeststhatweareabletostabilizelong‐rangemagneticorderandhighspincobaltviaepitaxialtensilestrain.Wespeculatethatthelatticedistortionsarisingfromthinfilmepitaxycorrespond to a change in the overall crystal field splitting andpossibly even thesequenceofelectronicenergylevelscomparedtothebulk.InbulkPCO,thelowspinstateofS=0isstableupto200K,indicatingthatthedecreasedorbitaloverlapandhybridizationdue to thesmallerCo‐O‐Cobondangles in thissystem,comparedtoLCO, aid the stabilization of the low spin state [2]. Distortion of the lattice in thetensile strained films may oppose this decrease in the Co‐O‐Co bond angles andasymmetrically alter the crystal field enough to drive the system to a higher spinground state even at low temperatures. In the bulk, higher spin can be accessedthrough thermal activation above 200 K. Given the increased stability of the lowspinCostateinbulkPCOcomparedtobulkLCO,itissurprisingthatweobserveanylong‐range magnetic order in PCO films. Epitaxial tensile strain and chemicalpressure from the smaller Pr cation represent competing influences on the spinstatestability.Thus,thelong‐rangeordersuggeststhatinducedmodificationsoftheelectronic structure due to epitaxial strain dominate the tendency of decreasingbond‐angle(from164°inbulkLCOto~157°[24] inbulkPCO)tostabilizethelowspin Co state due to Pr chemical pressure on the Co octahedra. However, the

73

reduction inCurie temperature(Tc~60K forPCOfilmscomparedto80K forLCOfilms[6]suggeststhatthesmallerPrcationdoesreducethestrengthoftheorbitaloverlapthatgivesrisetotheferromagneticexchangebetweenCoions. The most remarkable finding of this investigation is the demonstration oftwo ferromagnetic sublattices:One sublattice of Co ions aligningwith the appliedmagneticfieldandanotherPrsublatticethatisantiferromagneticallycoupledtotheCo moments. This result reveals that these films can be more appropriatelydescribedas ferrimagneticwiththeComomentdominatingthemagnetizationandinducinganopposingmomentfromthePrions.Moreover,themagnetizationversustemperaturesuggeststhatthePrsublatticehasanorderingtemperatureatornearthesametemperatureas theCosublattice(Tc~60K).Pr3+ions inbulkPCOdonotshowanyordering (spontaneousor induced)due to theabsenceof ferromagneticexchange among the Co ions and the distance between neighboring Pr sites. IndopedPr1‐xSrxCoO3,long‐rangeferromagnetismisobservedforx>0.2[16]andcanbeattributed to thedouble exchange interactionamong intermediate spin (IS)Comomentsinanintermediatevalence,mediatedbyamobileeg*electron.Featuresinthe magnetism have been associated with an abrupt change in the magneticanisotropyandadramaticstructuralchange[16,25,26),associatedwithstrongPr4f–O2porbitalhybridizationbelow120K[16,27].However,mostimportantlyinthecontextofthisstudy,Prorderingdoesnotoccurinthesebulksystems[16].InPr3+systems,despitethetendencyforasingletstate,Prsiteorderingcanstilloccurprovided the exchange in the vicinity of the ions is strong enough [28]. Inantiferromagnetic[29],PrMnO3[30]andpossiblyPrFeO3[31],Pr3+ionshaveshowntobeweaklyexchangecoupled to the transitionmetal ions. In thesesystems, thisweakexchangeisoftenmanifestedasalowermomentthanexpectedbelowtheTNforthetransitionmetalsublattice.

In epitaxially strained films, the Pr moment appears to be coupled to theHS/LSComoments.UnlikethePr1‐xSrxCoO3ISCoscenario,thelong‐rangemagneticorderinPCOfilmsiscorrelatedwithagreaterpresenceofHSCoionsandelectronsin t2gorbitals (directed at Pr sites). Thus, the stabilization of ferrimagnetism andinsulatingbehaviorinPCOfilmspresentsuniquebehaviornotobservedinotherPr‐based perovskite oxides. Since the films are insulating, the long‐range Pr‐ionmagnetic ordering cannot be due to double exchange interaction mediated byintermediatespinandvalencestates,butperhapscanbeduetoexchangeamongPrandhighspinCo.Presumably, the larger ionicsizeandthehigherspinmomentofHSCoionscontributetothePrsiteorderingbyincreasingthepossibilityforstrongPr‐Coexchange[28].However,wedonotobserveanyevidenceforincreasedPr‐OhybridizationfromtheXAspectra,unlikeinthePr1‐xCaxCoO3[21]andPr1‐xSrxCoO3systems [16]. In any case, the ordering of Pr observed here is fundamentallydifferentfromthehighTcISCosystem(alkaline‐earthdopedPrCoO3)wherenoProrderingisobserved.Moreover,theorderingoftheHSCoionsoccursinconjunctionwith the Pr ordering, as there is evidence for only onemagnetic transition in thetemperaturedependenceofthemagnetization. SincethePrmomentopposestheComoment,wecanassumethatthetwo4fspinsofPr3+opposetheCospins.SowhileinLCOtheSQUIDmomentreflectstheCo

74

momentonly[9],inPCOtheSQUIDmomentreflectstheCospinmoment,minusthePrspinmoment,pluseffectsfromorbitalmomentsorcanting.Toexplorethislatterconceptingreaterdetail,weusedtheXAspectrafromHSSr2CoO3ClandLSEuCoO‐3Hu to fit the XA from thin and thick films of PCO. We fit the relative ratios ofshouldersandpeaksintheL3andL2CoedgestocorrespondtoamixtureofHSandLS Co to estimate the amount of HS cobalt at the surface of the films. This gives~33%+/‐10%HSCo in the 8nm thick sample and~100%LSCo in the110nmthicksample.ThisincreaseinLSCocontributionanddecreaseinmagnetizationforthe thicker more relaxed sample is expected since bulk PCO is mostly LS Co(measuredatorbelow25K). Ifweassume33%HSCo in the thin film, spinonlymomentfortheCosublatticeshouldbe~1.33µB/Co.IfweassumethePrmomentopposes the Comoment, we expect the overallmoment to be significantly lowerthan this theoretical value. The observation of 0.4 µB/PrCoO3 for films on SrTiO3from SQUID magnetometry is consistent with this analysis. Here we have notincludedpossibleeffects fromorbital contributions (frombothCoandPr lattices)andeffectsfromthecantingofthesublattices.

5.5.Conclusions In summary, we have stabilized a ferrimagnetic ground state in PCO filmsthatareundercoherenttensilestrain.FilmsonLSATandSTObothshowlong‐rangemagneticorderwhileunstrainedfilmsandrelaxedfilmsundercompressivestraindo not. Ferrimagnetic thin films on STO are composed of antiparallel Co and PrferromagneticsublatticesthatsimultaneouslyorderatCurietemperaturesof 60K.ThelowerTc,comparedtoferromagneticfilmsofLCO[6],canbeattributedtotherelativechangesinoctahedralsiterotationsandresultingchangesinorbitaloverlapmediating exchange among Co ions in PCO films. The ferrimagnetic insulatingbehaviorobserved in this studysuggests that the long‐rangemagneticordering isfundamentally different from the double exchange interaction found in dopedcobaltitesandcanbeattributedtostrongexchangeamongHSCoionsandPrions.IntheepitaxialPCOfilms,latticestrainoverwhelmsthecompetingchemicalpressureeffectsfromthesmallerPrcationtostabilizealowtemperatureHSgroundstateinthe Co ion. It appears that capitalizing on the HS/LS Co‐based magnetism incobaltite filmsmayoffer anewroute tohigher temperaturemagneticorderingofrare‐earthions.

75

DissertationSummaryandOutlook In this dissertation, I have demonstrated the use of epitaxial strain tomanipulate long‐rangemagnetic order in cobaltites. In both LaCoO3 and PrCoO3 Ihave shown that epitaxial strain in tension plays an important role in (at leastpartially) altering the spin state of the Co ions and results in a ferromagneticexchangemechanismthatmayinvolveHSCo2+,HSCo3+,andLSCo3+.Thefilmsarecharacterized with a Curie temperature consistently below ~85 K and with asaturatedmomentvaryingfrom0to~2µB/Co.Thismomentappearstodependontetragonal distortion in the film, though stoichiometry likely plays a role aswell.Indeed, in LaCoO3 films there is a clear indication (from the thick film growth onLAO substrates) that defectsmay play an important role in stabilizing long‐rangemagneticorder.Theanalysisof themicrostructureshows thatorderingofoxygenvacanciesinthefilmsseemtoberelatedtothesign(tensionorcompression)andmagnitudeof the epitaxial strain. It is still unclearpreciselywhat role theoxygenvacanciesobservedinthisstudyplayinthemagnetismofthefilms,butevidenceoftheirinfluenceiscertainfromthestudyoftheelectronicstructure.Theexistenceoforderedoxygenvacanciesinthesefilmsandtheirdependenceonthestrainstateisinteresting and warrants further studies, but unfortunately access to thesemicroscopy instruments is always in high demand and getting enough time andreliablestatistics(manysamples)isaslowprocess. InPrCoO3filmsIfoundthatthestraindependenceismoreobvious,andtherole of defects is more nebulous. Regardless the largest moments are found instrainedfilmsintensionwhilerelaxedfilmsshowasignificantlydecreasedmoment,andunstrainedfilms(onLAO)andfilmsstrainedincompressiondonotshowlong‐rangemagneticorderatall.ThefindingssuggeststhatthepresenceofanorderedHS/LS Co3+ sublattice induces a novel ordering of the Pr sublattice. To myknowledge,thisisthefirstdemonstrationofPrsublatticeorderinginthecobaltitesandanexampleofProrderingtorelativelyhightemperatures(~60K). The research presented in this dissertationmotivatesmorework inmanynewdirectionspertainingtothecontrolofspinstatesincobaltitesandthepreciseunderstandingoftheferromagneticexchangemechanisminthesethinfilmsystems.AlthoughIhaveuncoveredsomeofthenecessaryelements(Coionsinvariousspinandvalencestates)thatare involvedintheferromagnetismofcobaltitethinfilms,more research is under way by numerous groups to fully explain ferromagneticexchange mechanism. Also, the demonstration of ferrimagnetism in PrCoO3 filmsopens up newavenues of exploration into other rare‐earth cobaltites,whichmayshow strongerorweaker ferrimagnetic ordering (compared toPrCoO3 thin films)withtheuseofadifferentrare‐earthelement.ItwouldbeinterestingtoseewhethertheseobservationsareuniquelyrelatedtosomepropertyofthePrions(Priswell‐knowntohybridizestronglywithOionsinthesecompounds),oriforderingcanbefoundonasublatticeofanyoftheotherlaterelementsintheseries(Nd–Tm).

76

ReferencesChapter1References

[1]M.Ellis,N. Jollands,L.Harrington,andA.Meier,Doenergyefficientappliancescostmore.InProceedingsoftheECEEE2007Conference(2007).

[2] Gregg, J. F., I. Petej, E. Jouguelet, and C. Dennis. Spin electronics—a review,JournalofPhysicsD:AppliedPhysics35,R1212002).[3]M.Bibes andA.Barthelemy,Oxidespintronics, IEEETransactionsonElectronDevices 54, 1003 (2007).

[4] J. Stohr and H. C. Siegmann, H.C.Magnetism ‐ from fundamentals tonanoscaledynamics,(Springer‐Verlag,Berlin,2006),1sted.[5] D. Feng and G. Jin, Introductiontocondensedmatterphysics, (World Scientific,2005).[6]J.B.GoodenoughandJ.–S.Zhou,Chapter1:Generalconsiderations,inLocalizedto itinerant transitions inperovskiteoxides, Structure and Bonding 98, (Springer‐Verlag,NewJersey,2001)

[7] Goodenough, J.B. Electronic and ionic transport properties and other physicalaspectsofperovskites.ReportsonProgressinPhysics,67,1915(2004).[8]R.L.DeKockandH.B.Gray,Chemicalstructureandbonding,(UniversityScienceBooks,Sausalito,1989).2nded.[9]F.A.Cotton,Chemicalapplicationsofgrouptheory,(Wiley,1990).3rded.[10]J.K.Burdett,Chemicalbondinginsolids,(OxfordUniversityPress,1995).[11] J. K. Burdett, Molecular orbital theory, AccessScience, (McGraw‐Hill, 2010.).<http://www.accessscience.com>,2012.[12] J.B.Goodenough,Magnetismandthechemicalbond, InterscienceMonographsonChemistry1,F.A.Cotton,ed(Wiley,NewYork,1963).[13]J.M.RondinelliandN.A.Spaldin,Structureandpropertiesoffunctionaloxidethin films: insights from electronic‐structure calculations, AdvancedMaterials23,3363(2011).

[14] P. W. Anderson, Antiferromagnetism. theory of superexchange interaction.PhysicalReview79,350(1950).

77

[15] C. Zener, Interaction between the d‐shells in the transiton metals. II.Ferromagnetic compoundsofmanganesewithperovskite structure, Phys.Rev.82403(1951).[16] A. Ohtomo, H. Y. Hwang, A high‐mobility electron gas at the LaAlO3/SrTiO3heterointerface,Nature427,423(2004).[17]J.MannhartandD.G.Schlom,Oxideinterfaces‐‐anopportunityforelectronics,Science327,1607(2010).[18]F.J.Wong,S.‐H.Baek,R.V.Chopdekar,V.V.Mehta,H.‐W.Jang,C.‐B.Eom,andY.Suzuki, “Metallicity in LaTiO3 thin films induced by lattice deformation,” PhysicalReviewB81,1611012010.

[19] M. Imada, A. Fujimori, Y. Tokura, Metal‐insulator transitions. Reviews ofModernPhysics,70,1039(1998).[20]R.R.Heikes,R.C.Miller,andR.Mazelsky,MagneticandelectricalanamoliesinLaCoO3,Physica30,1600(1964).[21]P.M.Raccahand J.B.Goodenough,First‐order localizedelectrontocollectiveelectrontransitioninLaCoO3,PhysicalReview155,932(1967).[22]M.A.Korotin,S.Y.Ezhov,I.V.Solovyev,V.I.Anisimov,D.I.Khomskii,andG.A.Sawatzky,Intermediate‐spinstateandpropertiesofLaCoO3,PhysicalReviewB54,5309(1996).[23]M.Haverkort,Z.Hu, J.Cezar,T.Burnus,H.Hartmann,M.Reuther,C.Zobel,T.Lorenz,A.Tanaka,N.Brookes,H.Hsieh,H.‐J.Lin,C.Chen,andL.Tjeng,SpinStateTransition in LaCoO3 Studied Using Soft X‐ray Absorption Spectroscopy andMagneticCircularDichroism,PhysicalReviewLetters97,176405(2006).[24]T.Vogt, J.Hriljac,N.Hyatt,andP.Woodward,Pressure‐inducedintermediate‐to‐lowspinstatetransitioninLaCoO3,PhysicalReviewB67,140401(2003).[25] G. Vankó, J.‐P. Rueff, A.Mattila, Z. Németh, and A. Shukla, Temperature‐ andpressure‐induced spin‐state transitions in LaCoO3, Phyical Review B 73, 024424(2006).[26]D.Kozlenko,N.Golosova,Z.Jirák,L.Dubrovinsky,B.Savenko,M.Tucker,Y.LeGodec,andV.Glazkov,Temperature‐andpressure‐drivenspin‐state transitions inLaCoO3,PhysicalReviewB75,064422(2007).

78

[27] M. Tachibana, T. Yoshida, H. Kawaji, T. Atake, and E. Takayama‐Muromachi,EvolutionofelectronicstatesinRCoO3(R=rareearth):Heatcapacitymeasurements,PhyicalReviewB77,094402(2008).[28] K. Knížek, J. Hejtmánek, Z. Jirák, P. Tomeš, P. Henry, and G. André, NeutrondiffractionandheatcapacitystudiesofPrCoO3andNdCoO3,PhysicalReviewB79,134103(2009).[29]C.Leighton,D.Stauffer,Q.Huang,Y.Ren,S.El‐Khatib,M.Torija,J.Wu,J.Lynn,L.Wang, N. Frey, H. Srikanth, J. Davies, K. Liu, and J. Mitchell, Coupledstructural/magnetocrystalline anisotropy transitions in the doped perovskitecobaltitePr1−xSrxCoO3,PhysicalReviewB79,214420(2009).[30]J.WuandC.Leighton,GlassyferromagnetismandmagneticphaseseparationinLa1‐xSrxCoO3,PhysicalReviewB67,174408(2003).[31]D.Fuchs,C.Pinta,T.Schwarz,1,P.Schweiss,P.Nagel,S.Schuppler,R.Schneider,M.Merz,G.Roth,andH.v.Löhneysen,Ferromagneticorder inepitaxiallystrainedLaCoO3thinfilms,PhysicalReivewB75,144402(2007).[32]V.V.Mehta,M.Liberati,F.J.Wong,R.V.Chopdekar,E.Arenholz,andY.Suzuki,Ferromagnetism in tetragonally distorted LaCoO3 thin films,” Journal of AppliedPhysics105,07E503(2009).[33] A. D. Rata, A. Herklotz, L. Schultz, and K. Dörr, Lattice structure andmagnetizationofLaCoO3thinfilms,EuropeanPhysicalJournalB76,215(2010).[34] A. Herklotz, A. Rata, L. Schultz, and K. Dörr, Reversible strain effect on themagnetizationofLaCoO3films,PhysicalReviewB79,092409(2009).[35]S.Park,P.Ryan,E.Karapetrova,J.W.Kim,J.X.Ma,J.Shi,J.W.Freeland,andW.Wu,Microscopicevidenceofastrain‐enhanced ferromagneticstate inLaCoO3thinfilms,AppliedPhysicsLetters95,072508(2009).[36] J.W.Freeland, J.X.Ma,andJ.Shi,Ferromagneticspin‐correlations instrainedLaCoO3thinfilms,AppliedPhysicsLetters93,212501(2008).[37] D. Fuchs, E. Arac, C. Pinta, S. Schuppler, R. Schneider, and H. v. Löhneysen,Tuning the magnetic properties of LaCoO3 thin films by epitaxial strain, PhysicalReviewB77,014434(2008).[38]V.MehtaandY.Suzuki,FerromangnetismenhancedbystructuralrelaxationofbiaxiallycompressedLaCoO3films,JournalofAppliedPhysics109,07D717(2011).

79

[39]R.F.Klie,T.Yuan,M.Tanase,G.Yang,andQ.Ramasse,Directmeasurementofferromagnetic ordering in biaxially strained LaCoO3 thin films, Applied PhysicsLetters96,082510(2010).Chapter2References[1]MiltonOhring,Materialsscienceofthinfilms,2nded.(AcademicPress,2002).[2]Chrisey,DouglasPulsedLaserDepositionofthinfilms,1sted.(Wiley,1994).[3]R.DelmdahlandR.Paetzel,RecentdevelopmentsinUVlasermicromachining,in:Proceedings of SPIE, the International Society for Optical Engineering, 68740K–1(2008).[4]A.Mele,A.G.Guidoni,R.Kelly,A.Miotello,S.Orlando,R.Teghil,andC.Flamini,Angular distribution and expansion of laser ablation plumes measured by fastintensifiedchargecoupleddevicephotographs, Nuclear InstrumentsandMethodsin Physics Research Section B: Beam InteractionswithMaterials and Atoms116,257(1996).[5]B.D.CullityandS.R.Stock,ElementsofX‐raydiffraction,3rded. (Prentice‐Hall,2001).[6]PaulFewster,X‐rayanalysisofthinfilmsandmultilayers,ReportsonProgressinPhysics59,1339(1996).[7]W.K.Chu,J.W.Mayer,M.A.Nicolet,T.M.Buck,G.Amsel,andF.Eisen,Principlesand applications of ion beam techniques for the analysis of solids and thin films,ThinSolidFilms17,1(1973).[8] E. Rauhala,N. P. Barradas, S. Fazinic,M.Mayer, E. Szilágyi, andM. Thompson,Statusofionbeamdataanalysisandsimulationsoftware,NuclearInstrumentsandMethods in Physics Research Section B: Beam Interactions with Materials andAtoms244,436(2006).[9]M.A.Garcia,E.F.Pinel,J.delaVenta,A.Quesada,V.Bouzas,J.F.Fernandez,J.J.Romero, M. S. Martin Gonzalez, and J. L. Costa‐Kraemer, Sources of experimentalerrors in theobservationofnanoscalemagnetism, JournalofAppliedPhysics105,013925(2009).[10] M. Sawicki, W. Stefanowicz, and A. Ney, Sensitive SQUID magnetometry forstudying nanomagnetism, Semiconductor Science and Technology 26, 064006(2011).

80

[11]A.Ney,T,Kammermeier,V.Ney,K.Ollefs,andS.Ye,Limitationsofmeasuringsmallmagneticsignalsofsamplesdepositedonadiamagneticsubstrate,JournalofMagnetismandMagneticMaterials320,3341(2008).[12] R. Salzer, D. Spemann, P. Esquinazi, R. Hoehne, A. Setzer, K. Schindler, H.Schmidt, and T. Butz, Possible pitfalls in search of magnetic order in thin filmsdeposited on single crystalline sapphire substrates, Journal of Magnetism andMagneticMaterials317,53(2007).[13] J. Stohr and H. C. Siegmann, Magnetism‐from fundamentals to nanoscaledynamics,(Springer‐Verlag,Berlin,2006).[14]R.Nakajima, J. Stoehr, andY. Idzerda, ElectronYield saturation effects in LedgeX‐raymagneticcirculardichroismspectraofFe,Co,andNi,PhysicalReviewB59,6421(1999).[15]D.B.WilliamsandC.B.Carter,TransmissionElectronMicroscopyIIIImaging,1sted.(PlenumPress,NewYork,1996).[16]Nellist,P.D.andS. J.Pennycook.TheprinciplesandinterpretationofAnnularDarkFieldZ‐contrastimaging.AdvancesinImagingandElectronPhysics113.147‐203.2000[17]D.B.WilliamsandC.B.Carter,TransmissionElectronMicroscopyIBasics,1sted.(PlenumPress,NewYork,1996).Chapter3references[1] D. Fuchs, E. Arac, C. Pinta, S. Schuppler, R. Schneider, and H. v. Loehneysen,Tuning the magnetic properties of LaCoO3 thin films by epitaxial strain. PhysicalReviewB77,014434(2008).[2]M.A.Torija,M.Sharma,M.R.Fitzsimmons,M.Varela,andC.Leighton,“EpitaxialLa0.5Sr0.5CoO3 thin films: Structure, magnetism, and transport,” JournalofAppliedPhysics,vol.104,no.2,p.023901,2008.[3]S.Stemmer,A.J.Jacobson,X.Chen,andA.Ignatiev,“OxygenvacancyorderinginepitaxialLa0.5Sr0.5CoO3−∂thinfilmson(001)LaAlO3,”JournalofAppliedPhysics,vol.90,no.7,p.3319,2001.[4]D.O.Klenov,W.Donner,B.Foran,andS.Stemmer,“ImpactofstressonoxygenvacancyorderinginepitaxialLa0.5Sr0.5CoO3−∂thinfilms,”AppliedPhysicsLetters,vol.82,no.20,p.3427,2003.

81

[5]R.Cacuiffo,D.Rinaldi,G.Barucca,J.Mira,J.Rivas,M.A.Senaris‐Rodriguez,P.G.Radaelli,D.Fiorani, J.B.Goodenough,StructuraldetailsandmagneticorderofLa1‐xSrxCoO3(x 3),PhysicalReviewB59,1068(1999).[6]P.E.Vullum,R.Holmestad,H.L.Lein,J.Mastin,M.‐A.Einarsrud,andT.Grande,MonoclinicFerroelasticDomainsinLaCoO3‐BasedPerovskites,AdvancedMaterials19,4399(2007).[7]P. E. Vullum, H. L. Lein, M.‐A. Einarsrud, T. Grande, and R. Holmestad, TEMobservationsof rhombohedralandmonoclinicdomains inLaCoO3‐basedceramics,PhilosophicalMagazine88,1187(2008).[8] J. Buban, H. Iddir, and S. Öğüt, Structural and electronic properties of oxygenvacancies in cubicandantiferrodistortivephasesof SrTiO3,PhysicalReviewB69,180102(2004).[9]C.Wang,B.L.Cheng,S.Y.Wang,H.B.Lu,Y.L.Zhou,Z.H.Chen,andG.Z.Yang,Effects of oxygen pressure on lattice parameter, orientation, surface morphologyanddepositionrateof(Ba0.02Sr0.98TiO3thinfilmsgrownonMgOsubstratebypulsedlaserdeposition,Thinsolidfilms485,82(2005).Chapter4References[1] V.Mehta and Y. Suzuki, Ferromagnetism enhanced by structural relaxation ofbiaxiallycompressedLaCoO3films,JournalofAppliedPhysics109,07D717(2011).[2]S.Park,P.Ryan,E.Karapetrova,J.W.Kim,J.X.Ma,J.Shi,J.W.Freeland,andW.Wu,Microscopicevidenceofastrain‐enhanced ferromagneticstate inLaCoO3thinfilms,AppliedPhysicsLetters95,072508(2009).[3] D. Fuchs, E. Arac, C. Pinta, S. Schuppler, R. Schneider, and H. v. Loehneysen,Tuning the magnetic properties of LaCoO3 thin films by epitaxial strain, PhysicalReviewB77,014434(2008).[4] A. D. Rata, A. Herklotz. L. Schultz. And K. Doerr, Lattice structure andmagnetization of LaCoO3 thin films, The European Physical Journal B 76, 215(2010).[5]A.Herklotz,A.D.Rata,L. Schultz, andK.Doerr,Reversible straineffecton themagnetizationofLaCoO3films,PhysicalReviewB79,092409(2009).[6] J.W. Freeland, J. X.Ma, and J. Shi, Ferromagnetic spin‐correlations in strainedLaCoO3thinfilms,AppliedPhysicsLetters93,212501(2008).

82

[7]R.F.Klie,T.Yuan,M.Tanase,G.Yang,andQ.Ramasse,Directmeasurementofferromagnetic ordering in biaxially strainedLaCoO3 films,AppliedPhysics Letters96,082510(2010).[8]M. R.McCartney andD. J. Smith, Epitaxial relationships in electron‐stimulateddesorption processes at transitionmetal oxide surfaces, Surface science221, 214(1989).[9]S.D.,Electronbeaminducedchanges intransitionmetaloxides,AnalyticalandBioanalyticalChemistry374,732(2002).[10] M. R. McCartney, P. A. Crozier, J. K. Weiss, and D. J. Smith, Electron‐beam‐inducedreactionsattransition‐metaloxidesurfaces,Vacuum42,301(1991).[11]D.B.WilliamsandC.B.Carter,TransmissionElectronMicroscopyIBasics,1sted.(PlenumPress,NewYork,1996).[18]L.J.vanDerPauw,Amethodofmeasuringtheresistivityandhallcoefficientonlamellaeofarbitraryshape,PhilipsTechnicalReview26,220(1958).[19] T. Burnus, Z. Hu,M. Haverkort, J. Cezar, D. Flahaut, V. Hardy, A.Maignan, N.Brookes, A. Tanaka, H. Hsieh, H.‐J. Lin, C. Chen, and L. Tjeng, Valence, spin, andorbital state of Co ions in one‐dimensional Ca3Co2O6: An x‐ray absorption andmagneticcirculardichroismstudy,PhysicalReviewB74,245111(2006).[20]Z.Hu,H.Wu,M.Haverkort,H.Hsieh,H.Lin,T.Lorenz,J.Baier,A.Reichl,I.Bonn,C.Felser,A.Tanaka,C.Chen,andL.Tjeng,DifferentLookat theSpinStateofCo3+IonsinaCoO5PyramidalCoordination,PhysicalReviewLetters92,207402(2004).[21] C. Chang, Z. Hu, H. Wu, T. Burnus, N. Hollmann, M. Benomar, T. Lorenz, A.Tanaka,H.‐J.Lin,H.Hsieh,C.Chen,andL.Tjeng,SpinBlockade,OrbitalOccupation,andChargeOrderinginLa1.5Sr0.5CoO4,PhysicalReviewLetters102,116401(2009).[22]S.Csiszar,M.Haverkort,Z.Hu,A.Tanaka,H.Hsieh,H.‐J.Lin,C.Chen,T.Hibma,and L. Tjeng, Controlling Orbital Moment and Spin Orientation in CoO Layers byStrain,PhysicalReviewLetters95,187205(2005).

[23]A.Kushima,S.Yip,andB.Yildiz,CompetingstraineffectsinreactivityofLaCoO3withoxygen,PhysicalReviewB82,115435(2010).[24] K. Gupta and P. Mahadevan, Strain driven magnetism in LaCoO3 thin film,PhysicalReviewB79,020406(2009).[25]J.RondinelliandN.A.Spaldin,Structuraleffectsonthespin‐statetransitioninepitaxiallystrainedLaCoO3films,PhysicalReviewB79,054409(2009).

83

[26]H.Hsu,P.Blaha,andR.Wentzcovitch,Ferromagneticinsulatingstateintensile‐strainedLaCoO3thinfilmsfromLDA+Ucalculations,PhysicalReviewB85,140404(2012).Chapter5References[1] Vankó, György, Jean‐Pascal Rueff, Aleksi Mattila, Zoltán Németh, and AbhayShukla, Temperature‐ and pressure‐induced spin‐state transitions in LaCoO3,PhysicalReviewB73,024424(2006).[2] Tachibana, Makoto, Takahiro Yoshida, Hitoshi Kawaji, Tooru Atake, and EijiTakayama‐Muromachi,EvolutionofelectronicstatesinRCoO3(R=rareearth):Heatcapacitymeasurements,PhysicalReviewB77,094402(2008).[3]Vogt,T.,J.A.Hriljac,N.C.Hyatt,andP.Woodward,PressureinducedintermediatetolowspinstatetransitioninLaCoO3,PhysicalReviewB67,14041(2003).[4]Kozlenko,D.,N.Golosova,Z. Jirák,L.Dubrovinsky,B.Savenko,M.Tucker,Y.LeGodec,andV.Glazkov,Temperature‐andpressure‐drivenspin‐state transitions inLaCoO3,PhysicalReviewB75,064422(2007).[5]J.‐S.Zhou,J.‐Q.Yan,andJ.Goodenough,BulkmodulusanomalyinRCoO3(R=La,Pr,andNd),PhysicalReviewB71,220103(2005).[6]Fuchs,D.,C.Pinta,T.Schwarz,P.Schweiss,P.Nagel,S.Schuppler,R.Schneider,M.Merz, G. Roth, and H. v. Löhneysen, Ferromagnetic order in epitaxially strainedLaCoO3thinfilms,PhysicalReviewB75,144402(2007).[7] D. Fuchs, L. Dieterle, E. Arac, R. Eder, P. Adelmann, V. Eyert, T. Kopp, R.Schneider,D.Gerthsen,andH.v.Löhneysen,SuppressionoftheferromagneticstateinLaCoO3filmsbyrhombohedraldistortion,PhysicalReviewB79,024424(2009).[8] D. Fuchs, E. Arac, C. Pinta, S. Schuppler, R. Schneider, and H. v. Loehneysen,Tuning the magnetic properties of LaCoO3 thin films by epitaxial strain, PhysicalReviewB77,014434(2008).[9]V.V.Mehta,M.Liberati,F.J.Wong,R.V.Chopdekar,E.Arenholz,andY.Suzuki,Ferromagnetism in tetragonally distorted LaCoO3 thin films,” Journal of AppliedPhysics105,07E503(2009).[10] A. D. Rata, A. Herklotz, L. Schultz, and K. Doerr, Lattice structure andmagnetization of LaCoO3 thin films, The European Physical Journal B 76, 215(2010).

84

[11]A.Herklotz,A.D.Rata,L.Schultz,andK.Doerr,ReversiblestraineffectonthemagnetizationofLaCoO3films,PhysicalReviewB79,092409(2009).[12] J.W.Freeland, J.X.Ma,andJ.Shi,Ferromagneticspin‐correlations instrainedLaCoO3thinfilms,AppliedPhysicsLetters93,212501(2008).[13]S.Park,P.Ryan,E.Karapetrova,J.W.Kim,J.X.Ma,J.Shi,J.W.Freeland,andW.Wu,Microscopicevidenceofastrain‐enhanced ferromagneticstate inLaCoO3thinfilms,AppliedPhysicsLetters95,072508(2009).[14] H. W. Brinks, H. Fjellvag, A. Kjekshus, and B. C. Hauback, Structure andmagnetismofPr1‐xSrxCoO3‐δ,JournalofSolidStateChemistry147,464(1999).[15]M.Paraskevopoulos,J.Hemberger,A.Krimmel,andA.Loidl,Magneticorderingandspin‐statetransitioninR0.67Sr0.33CoO3,PhysicalReviewB63,224416(2001).[16]C.Leighton,D.Stauffer,Q.Huang,Y.Ren,S.El‐Khatib,M.Torija,J.Wu,J.Lynn,L.Wang, N. Frey, H. Srikanth, J. Davies, K. Liu, and J. Mitchell, Coupledstructural/magnetocrystalline anisotropy transitions in the doped perovskitecobaltitePr1−xSrxCoO3,PhysicalReviewB79,214420(2009).[17]V.MehtaandY.Suzuki,FerromagnetismenhancedbystructuralrelaxationofbiaxiallycompressedLaCoO3films,JournalofAppliedPhysics109,07D717(2011).[18]M.Haverkort,Z.Hu, J.Cezar,T.Burnus,H.Hartmann,M.Reuther,C.Zobel,T.Lorenz,A.Tanaka,N.Brookes,H.Hsieh,H.‐J.Lin,C.Chen,andL.Tjeng,SpinStateTransition in LaCoO3 Studied Using Soft X‐ray Absorption Spectroscopy andMagneticCircularDichroism,PhysicalReviewLetters97,176405(2006).[19]Z.Hu,H.Wu,M.Haverkort,H.Hsieh,H.Lin,T.Lorenz,J.Baier,A.Reichl,I.Bonn,C.Felser,A.Tanaka,C.Chen,andL.Tjeng,DifferentLookat theSpinStateofCo3+IonsinaCoO5PyramidalCoordination,PhysicalReviewLetters92,207402(2004).[20] Z. Hu, G. Kaindl, H. Ogasawara, A. Kotani, and I. Felner, Ln‐4f/ligand‐2pcovalence inBaLnO3 andCs3LnF7(Ln = Ce, Pr, Tb), Chemical Physics Letters325,241(2000).[21]J.Herrero‐Martin,J.Garcia‐Munoz,S.Valencia,C.Frontera,J.Blasco,A.Baron‐Gonzalez,G.Subias,R.Abrudan,F.Radu,E.Dudzik,R.Feyerherm,ValencechangeofPraseodymium in PrCaCoO3 invesitigated by X‐ray absorption spectroscopy,PhysicalReviewB84,115131(2011).[22]J.B.Goedkoop,B.T.Thole,G.vanderLaan,G.A.Sawatzky,F.M.F.deGroot,andJ.C.Fuggle,Calculationsofmagneticx‐raydichroisminthe3dabsorptionspectraofrare‐earthcompounds,PhysicalReviewB37,2086(1988).

85

[23] G. van der Laan, A. Arenholz, Z. Hu, A. Bauer, E. Weschke, Ch. Schuessler‐Langeheine, E. Navas, A. Muehlig, G. Kaindl, J. B. Goedkoop, and N. B. Brookes,MagneticcirculardichroisminTb3d‐4fresonantphotoemission,PhysicalReviewB59,8835(1999).[24]Knížek,Karel,JiříHejtmánek,ZdeněkJirák,PetrTomeš,PaulHenry,andGillesAndré, Neutron diffraction and heat capacity studies of PrCoO3 and NdCoO3,PhysicalReviewB79,134103(2009).[25]M. Uchida, R.Mahendiran, Y. Tomioka, Y.Matsui, K. Ishizuka, and Y. Tokura,Changes of magnetic domain structure induced by temperature‐variation andelectron‐beam irradiation in Pr0.5Sr0.5CoO3, Applied Physics Letters 86, 131913(2005).[26] S. Hirahara, Y. Nakai, K. Miyoshi, K. Fujiwara, and J. Takeuchi, Magneticproperties,thermalexpansionandmagnetostrictionofPr0.5Sr0.5CoO3singlecrystal,JournalofMagnetismandMagneticMaterials310,1866(2007).[27]I.O.Troyanchuk,D.V.Karpinskiĭ,A.N.Chobot,D.G.Voĭitsekhovich,andV.M.Dobryanskiĭ,PhasetransformationsinPr1‐xSrxCoO3,JETPLetters84,151(2006).[28] Y. Wang, and B. Cooper, Collective Excitations and Magnetic ordering inmaterialswithsingletcrystal‐fieldgroundstate,PhysicalReview172,539(1968).[29]J.D.Gordon,R.M.Hornreich,S.Shtrikman,andB.M.Wanklyn,Magnetizationstudiesintherare‐earthorthochromites.V.TbCrO3andPrCrO3,PhysicalReviewB13,3012(1976).[30]A. A.Mukhin, V. Yu, V.D. Travkin, A.M. Balhashov,Magnetic anisotropy andgroundstateof the rare‐earth ions inPrMnO3andNdMnO3, JournalofMagnetismandMagneticMaterials226‐230,1139(2001).[31]X.Wang,S.Cao,Y.Wang,S.Yuan,B.Kang,A.Wu,andJ.Zhang,Crystalgrowthand characterization of the rare earth orthoferrite PrFeO3, Journal of CrystalGrowth,Jan.(2012).(inpress)DOI:10.1016/j.jcrysgro.2011.12.096