Complete Download Categories for quantum theory: an introduction chris heunen PDF All Chapters

Page 1


Categories for Quantum Theory: An Introduction Chris Heunen

Visit to download the full and correct content document: https://ebookmass.com/product/categories-for-quantum-theory-an-introduction-chris-h eunen/

More products digital (pdf, epub, mobi) instant download maybe you interests ...

Quantum Bullsh*t Chris Ferrie

https://ebookmass.com/product/quantum-bullsht-chris-ferrie/

An Introduction to quantum optics and quantum fluctuations First Edition. Edition Milonni

https://ebookmass.com/product/an-introduction-to-quantum-opticsand-quantum-fluctuations-first-edition-edition-milonni/

Introduction to Quantum Field Theory with Applications to Quantum Gravity 1st Edition Iosif L. Buchbinder

https://ebookmass.com/product/introduction-to-quantum-fieldtheory-with-applications-to-quantum-gravity-1st-edition-iosif-lbuchbinder/

Quantum Bullsh*t: How to Ruin Your Life with Advice from Quantum Physics Chris Ferrie

https://ebookmass.com/product/quantum-bullsht-how-to-ruin-yourlife-with-advice-from-quantum-physics-chris-ferrie/

Rhetorical Theory: An Introduction- Inc.-2018, Ebook PDF Version

https://ebookmass.com/product/rhetorical-theory-an-introductioninc-2018-ebook-pdf-version/

Understanding Global Conflict and Cooperation: An Introduction to Theory

https://ebookmass.com/product/understanding-global-conflict-andcooperation-an-introduction-to-theory/

Combinatorial Physics: Combinatorics, Quantum Field Theory, and Quantum Gravity Models Adrian Tanasa

https://ebookmass.com/product/combinatorial-physicscombinatorics-quantum-field-theory-and-quantum-gravity-modelsadrian-tanasa/

Qualitative Theory of ODEs: An Introduction to Dynamical Systems Theory 1st Edition Henryk Zoladek

https://ebookmass.com/product/qualitative-theory-of-odes-anintroduction-to-dynamical-systems-theory-1st-edition-henrykzoladek/

Introduction to Quantum Mechanics John Dirk Walecka

https://ebookmass.com/product/introduction-to-quantum-mechanicsjohn-dirk-walecka/

OXFORDGRADUATETEXTSINMATHEMATICS

SeriesEditors

R.Cohen|S.K.Donaldson

OXFORDGRADUATETEXTSINMATHEMATICS

Booksintheseries

1.KeithHannabuss: AnIntroductiontoQuantumTheory

2.ReinholdMeiseandDietmarVogt: IntroductiontoFunctionalAnalysis

3.JamesG.Oxley: MatroidTheory

4.N.J.Hitchin,G.B.Segal,andR.S.Ward: IntegrableSystems:Twistors,Loop Groups,andRiemannSurfaces

5.WulfRossmann: LieGroups:AnIntroductionthroughLinearGroups

6.QingLiu: AlgebraicGeometryandArithmeticCurves

7.MartinR.BridsonandSimonM.Salamon(eds): InvitationstoGeometry andTopology

8.ShmuelKantorovitz: IntroductiontoModernAnalysis

9.TerryLawson: Topology:AGeometricApproach

10.MeinolfGeck: AnIntroductiontoAlgebraicGeometryandAlgebraicGroups

11.AlastairFletcherandVladimirMarkovic: QuasiconformalMaps andTeichmüllerTheory

12.DominicJoyce: RiemannianHolonomyGroupsandCalibratedGeometry

13.FernandoVillegas: ExperimentalNumberTheory

14.PéterMedvegyev: StochasticIntegrationTheory

15.MartinA.Guest: FromQuantumCohomologytoIntegrableSystems

16.AlanD.Rendall: PartialDifferentialEquationsinGeneralRelativity

17.YvesFélix,JohnOprea,andDanielTanré: AlgebraicModelsinGeometry

18.JieXiong: IntroductiontoStochasticFilteringTheory

19.MaciejDunajski: Solitons,Instantons,andTwistors

20.GrahamR.Allan: IntroductiontoBanachSpacesandAlgebras

21.JamesOxley: MatroidTheory, SecondEdition

22.SimonDonaldson: RiemannSurfaces

23.CliffordHenryTaubes: DifferentialGeometry:Bundles,Connections, MetricsandCurvature

24.GopinathKallianpurandP.Sundar: StochasticAnalysisandDiffusionProcesses

25.SelmanAkbulut: 4-Manifolds

26.Fon-CheLiu: RealAnalysis

27.DusaMcduffandDietmarSalamon: IntroductiontoSymplecticTopology, ThirdEdition

Categoriesfor QuantumTheory

AnIntroduction

chrisheunenandjamievicary UniversityofEdinburghandUniversityofBirmingham

GreatClarendonStreet,Oxford,OX26DP, UnitedKingdom

OxfordUniversityPressisadepartmentoftheUniversityofOxford. ItfurtherstheUniversity’sobjectiveofexcellenceinresearch,scholarship, andeducationbypublishingworldwide.Oxfordisaregisteredtrademarkof OxfordUniversityPressintheUKandincertainothercountries

©ChrisHeunenandJamieVicary2019

Themoralrightsoftheauthorshavebeenasserted FirstEditionpublishedin2019

Impression:1

Allrightsreserved.Nopartofthispublicationmaybereproduced,storedin aretrievalsystem,ortransmitted,inanyformorbyanymeans,withoutthe priorpermissioninwritingofOxfordUniversityPress,orasexpresslypermitted bylaw,bylicenceorundertermsagreedwiththeappropriatereprographics rightsorganization.Enquiriesconcerningreproductionoutsidethescopeofthe aboveshouldbesenttotheRightsDepartment,OxfordUniversityPress,atthe addressabove

Youmustnotcirculatethisworkinanyotherform andyoumustimposethissameconditiononanyacquirer

PublishedintheUnitedStatesofAmericabyOxfordUniversityPress 198MadisonAvenue,NewYork,NY10016,UnitedStatesofAmerica

BritishLibraryCataloguinginPublicationData Dataavailable

LibraryofCongressControlNumber:2019941499

ISBN978–0–19–873962–3(hbk.)

ISBN978–0–19–873961–6(pbk.)

DOI:10.1093/oso/9780198739623.001.0001

Printedandboundby CPIGroup(UK)Ltd,Croydon,CR04YY

Preface

Theideaforthisbookcamefromamini-courseataspringschoolin2010,aimedat beginninggraduatestudentsfromvariousfields.Wefirstdevelopedthenotesin 2012,whentheyformedthebasisforagraduatecourseattheDepartmentofComputer ScienceinOxford,whichhasruneveryyearsince.Asof2016,aversionofthiscourse hasalsorunintheSchoolofInformaticsinEdinburgh.Thetextformedthebasisfor postgraduatesummerandwinterschoolsinDalhousie,PisaandPalmseaswell,andhas improvedwithfeedbackfromallthesestudents.

Inselectingthematerial,wehavestrivedtostrikeabalancebetweentheoryand application.Applicationsareinterspersedthroughoutthemaintextandexercises,but atthesametime,thetheoryaimsatmaximumreasonablegenerality.Theemphasisis slightlyonthetheory,becausetherearesomanyapplicationsthatwecanonlyhintat them.Proofsarewrittenoutinfull,exceptahandfulthatarebeyondthescopeofthis bookandacoupleofchoresthatareleftasexercises.Wehavetriedtoassignappropriate credit,includingreferences,inhistoricalnotesattheendofeachchapter.Ofcourse,the responsibilityformistakesremainsentirelyoursandwehopethatreaderswillpointthem outtous.Tomakethisbookasself-containedaspossible,thereisazerothchapterthat brieflydefinesallprerequisitesandfixesnotation.

ThistextwouldnotexistwereitnotforthemotivationofBobCoecke.Between inceptionandpublicationofthisbook,heandAleksKissingerwroteanotheronethat targetsadifferentaudiencebutcoverssimilarmaterial[44].Thanksarealsoduetothe studentswholetususethemasguineapigsfortestingoutthismaterial.Finally,we areenormouslygratefultoJohn-MarkAllen,PabloAndresMartinez,MiriamBackens, BruceBartlett,OscarCunningham,BrendanFong,PauEnriqueMoliner,TobiasFritz, PeterHines,MarttiKarvonen,AlexKavvos,AleksKissinger,BertLindenhovius,Daniel Marsden,AlexMerry,VaiaPatta,DavidReutter,FranciscoRios,PeterSelinger,SeanTull, DominicVerdon,LindeWesterandVladimirZamdzhievforcarefulreadinganduseful feedbackonearlyversions.

ChrisHeunenandJamieVicary EdinburghandBirmingham,December2018

Introduction

Physicalsystemscannotbestudiedinisolation.Afterall,wecanonlyobservetheir behaviourwithrespecttoothersystems,suchasameasurementdevice.Thecentral premiseofthisbookisthattheabilitytogroupindividualsystemsintocompound systemsshouldbetakenseriously.Weadopttheactionofgroupingsystemstogetherasa primitivenotionandinvestigatemodelsofquantumtheoryfromthere.

Thejudiciouslanguageforthisstoryisthatofcategories.Categorytheoryteaches thatalotcanbelearnedaboutagiventypeofmathematicalspeciesbystudyinghow specimensofthespeciesinteractwitheachother,anditprovidesapotentinstrument todiscoverpatternsintheseinteractions.Noknowledgeofspecimens’insidesisneeded; infact,thisoftenonlyleadstotunnelvision.

Themethodsofcategoriesmightlooknothinglikewhatyouwouldexpectfroma treatiseonquantumtheory.Butacrucialthemeofquantumtheorynaturallyfitswith ourguidingprincipleofcompositionality:entanglementsaysthatcompleteknowledge ofthepartsisnotenoughtodeterminethewhole.

Inprovidinganunderstandingofthewayphysicalsystemsinteract,categorytheory drawscloselyonmathematicsandcomputerscienceaswellasphysics.Theunifying languageofcategoriesaccentuatesconnectionsbetweenitssubjects.Inparticular,all physicalsystemsarereallyquantumsystems,includingthoseincomputerscience.This bookappliesitsfoundationstodescribeprotocolsandalgorithmsthatleveragequantum theory.

Operationalfoundations. Anoperationalscientisttriestodescribetheworldinterms ofoperationsshecanperform.Theonlythingsoneisallowedtotalkaboutaresuch operations—preparingaphysicalsysteminsomestate,manipulatingaphysicalsystem viasomeexperimentalsetupormeasuringaphysicalquantity—andtheiroutcomes,or whatcanbederivedfromthoseoutcomesbymathematicalreasoning.

Thisisaveryconstructivewayofdoingscience.Afterall,theseoperationsand outcomesarewhatreallymatterwhenyoubuildamachine,designaprotocolorotherwise putyourknowledgeofnaturetopracticaluse.Buttraditionalquantumtheorycontains manyingredientsforwhichwehavenooperationalexplanation,eventhoughmany peoplehavetriedtofindone.Forexample,ifstatesareunitvectorsinaHilbertspace,what aretheothervectors?IfameasurementisahermitianoperatoronaHilbertspace,what dootheroperatorssignify?Whydoweworkwithcomplexnumbers,whenattheendof

thedayallprobabilitiesarerealnumbers?Categoriesofferawayoutofthesedetails,while maintainingpowerfulconclusionsfrombasicassumptions.Theyletusfocusonwhatis goingonconceptually,showcasingtheforestratherthanthetrees.

Graphicalcalculus. Whenthinkingoperationally,onecannothelpbutdrawschematic picturessimilartoflowchartstorepresentwhatisgoingon.Forexample,the quantum teleportationprotocol,whichwewillmeetmanytimesinthisbook,hasthefollowing schematicrepresentation:

Wereadtimeupwards,andspaceextendsleft-to-right.Wesay‘time’and‘space’,butwe meanthisinaveryloosesense.SpaceisonlyimportantinsofarasAliceandBob’s operationsinfluencetheirpartofthesystem.Timeisonlyimportantinthatthereisa notionofstartandfinish.Allthatmattersinsuchadiagramisitsconnectivity.

Suchoperationaldiagramsseemlikeinformal,non-mathematicaldevices,usefulfor illustrationandintuitiveunderstanding,butnotforprecisededuction.Infact,wewillsee thattheycanliterallybetakenaspiecesofformalcategorytheory.The‘time-like’lines becomeidentitymorphisms,and‘spatial’separationismodelledbytensorproductsof objects,leadingtomonoidalcategories.

Monoidalcategoriesformabranchthatdoesnotplayastarringrole,orevenany roleatall,inmoststandardcoursesoncategorytheory.Nevertheless,theyputworking withoperationaldiagramsonacompletelyrigorousfooting.Conversely,thegraphical calculusisaneffectivetoolforcalculationinmonoidalcategories.Thisgraphicallanguage isperhapsoneofthemostcompellingfeaturesthatmonoidalcategorieshavetooffer.By theendofthebook,wehopetohaveconvincedyouthatthisistheappropriatelanguage fordescribingandunderstandingmanyphenomenainquantumtheory.

Nonstandardmodels. Oncewehavemadethejumpfromoperationaldiagramsto monoidalcategories,manyoptionsopenup.Inparticular,ratherthaninterpreting diagramsinthecategoryofHilbertspaces,wherequantumtheorytraditionallytakes

place,wecaninsteadinterpretdiagramsinadifferentcategoryandtherebyexplore alternativestoquantumtheory.Anycalculationthatwasperformedpurelyusingthe graphicalcalculuswillalsoholdwithoutadditionalworkinthesenovelsettings.For example,whenwepresentquantumteleportationinthegraphicalcalculusandrepresent itinthecategoryofsetsandrelations,weobtainadescriptionofclassicalone-time-pad encryption.Thuswecaninvestigateexactlywhatitisthatmakesquantumtheory‘tick’, andwhatfeaturessetitapartfromothercompositionaltheories.

Thusmonoidalcategoriesprovideaunifyinglanguageforawidevarietyofphenomena drawnfromareasincludingquantumtheory,quantuminformation,logic,topology, representationtheory,quantumalgebra,quantumfieldtheoryandevenlinguistics.

Withinquantumtheory,categorieshighlightdifferentaspectsthanotherapproaches. Instrumentsliketensorproductsanddualspacesareofcourseavailableinthetraditional Hilbertspacesetting,buttheirrelevanceisheightenedhere,astheybecomethecentral focus.Howwerepresentmathematicalideasaffectshowwethinkaboutthem.

Outline. Afterthissomewhatroundaboutdiscussionofthesubject,itistimetostop beatingaboutthebushanddescribethecontentsofeachchapter.

Chapter0coversthebackgroundmaterial.Itfixesnotationsandconventionswhile verybrieflyrecallingthebasicnotionsfromcategorytheory,linearalgebraandquantum theorythatwewillbeusing.Ourrunningexamplecategoriesareintroduced:functions betweensetsrelationsbetweensets,andboundedlinearmapsbetweenHilbertspaces.

Chapter1introducesourmainobjectofstudy:monoidalcategories.Theseare categoriesthathaveagoodnotionoftensorproduct,whichgroupsmultipleobjects togetherintoasinglecompoundobject.Wealsointroducethegraphicalcalculus,the visualnotationformonoidalcategories.Thisgivesanotionofcompositionalityforan abstractphysicaltheory.Thenextfewchapterswilladdmorestructure,sothatthe resultingcategoriesexhibitmorefeaturesofquantumtheory.Section1.3investigates coherence,atechnicaltopicthatisessentialtothecorrectnessofthegraphicalcalculus, butisnotneededtounderstandlaterchapters.

TosomeonewhoequatesquantumtheorywithHilbertspacegeometry—andthiswill probablyincludemostreaders—theobviousnextstructuretoconsiderislinearalgebra. Chapter2showsthatimportantnotionssuchasscalars,superposition,adjointsandthe Bornrulecanallberepresentedinthecategoricalsetting.

Chapter3investigatesentanglementintermsofmonoidalcategories,usingthenotion ofdualobject,buildinguptotheimportantnotionofcompactcategory.Thisstructureis quitesimpleandpowerful:itgivesrisetoabstractnotionsoftraceanddimensionandis alreadyenoughtotalkaboutthequantumteleportationprotocol.

Uptothispointweconsideredarbitrarytensorproducts.Butthereisanobviousway tobuildcompoundobjectsusuallystudiedincategorytheory,namelyCartesianproducts (whichalreadymadetheirappearanceinChapter2).InChapter4,weconsiderwhat happensifthetensorproductisinfactaCartesianproduct.Theresultisanabstract versionoftheno-cloningtheorem:ifacategorywithCartesianproductsiscompact,then itmustbedegenerate.

Chapter5thenturnsthisno-cloningtheoremonitshead.Insteadofsayingthat quantumdatacannotbecopied,rather,classicaldataisviewedasquantumdatawitha copyingmapsatisfyingcertainaxioms.ThisleadsustodefineFrobeniusstructuresand thederivednotionofclassicalstructure.Infinite-dimensionalHilbertspaces,classical structuresturnouttocorrespondtoachoiceofbasis.Weestablishanormalformtheorem forFrobeniusstructuresthatgreatlysimplifiescomputations.Classicalstructuresalso allowthedescriptionofquantummeasurements,andweusethisinseveralapplication protocolssuchasstatetransferandquantumteleportation.

Oneofthedefiningfeaturesofquantummechanicsisthatsystemscanbemeasuredin incompatible—orcomplementary—ways.(Thefamousexampleisthatofpositionand momentum.)Chapter6definescomplementaryFrobeniusstructures.Therearestrong linkstoHopfalgebrasandquantumgroups.Withcomplementarityinhand,wediscuss severalapplicationstoquantumcomputing,includingtheDeutsch–Jozsaalgorithm,and somequbitgatesthatareimportanttomeasurement-basedquantumcomputing.We alsobrieflydiscusstheZXcalculus:asound,completeanduniversalwaytohandleany quantumcomputationgraphically,whichiseminentlyamenabletoautomation.

Alldiscussionssofarhavefocusedonpure-statequantumtheory.Chapter7liftseverythingtomixedquantumtheory,wherewecantakeprobabilisticcombinationsofstates andprocesses.Thisisdonebyanalyzingthecategoricalstructureofcompletelypositive maps.Theresultisaxiomatizedintermsofenvironmentstructuresanddecoherence structures,andweuseittogiveanothermodelofquantumteleportation.Thechapter endswithadiscussionofthedifferencebetweenclassicalandquantuminformationin theseterms.

ThebookfinisheswithChapter8,whichsketcheshighercategories.Whileanordinary categoryhasobjectsandmorphismsgoingbetweentheobjects,a2-categoryalsohas 2-morphismsgoingbetweenthemorphisms.Weshowhowthesestructures,alongwith thetechniquesofhigherrepresentationtheory,allowustogiveafullygeometrical descriptionofquantumteleportationasasinglegraphicalequationinahigherdimension.

Thatconcludesthemaindevelopmentofthematerial,atwhichpointyouwillhave metthebasicideasofusingcategoriesforquantumtheory.Thebookendsthere,being anintroduction,afterall.Butthisisreallyjustthebeginning!Afterthat,itisuptoyouto expeditetheexpirationdateofthisbookbystudyingthisexcitingtopicfurther.

5 FrobeniusStructures

5.1FrobeniusStructures147

5.2NormalForms157

5.3JustifyingtheFrobeniusLaw161

5.4Classification166

5.5Phases175

5.6Modules181 Exercises190

6 Complementarity 193

6.1ComplementaryStructures194

6.2TheDeutsch–JozsaAlgorithm201

6.3Bialgebras205

6.4QubitGates212

6.5ZXCalculus219 Exercises221

7 CompletePositivity

7.1CompletelyPositiveMaps225

7.2CategoriesofCompletelyPositiveMaps231

7.3ClassicalStructures241

7.4QuantumStructures244

7.5Decoherence249

7.6InteractionwithLinearStructure257 Exercises260

8 Monoidal2-Categories

0 Basics

Traditionalfirstcoursesincategorytheoryandquantumcomputingwouldprepare thereaderwithsolidfoundationsforthisbook.However,notmuchofthatmaterial istrulyessentialtogetthemostoutofthisbook.Thischaptergivesaverybrief introductiontocategorytheory,linearalgebraandquantumcomputing,enoughtoget yougoingwiththisbookifyouhavenottakenacourseinanyoftheseareasbefore,or perhapstoremindyouofsomedetailsifyouhaveforgottenthem.Everythinginthis chaptercanbefoundinmoredetailinmanyotherstandardtexts(seetheNotesatthe endofthechapterforreferences).Youcouldskipthischapterfornow,andreferbackto itwheneversomebackgroundismissing.

Thematerialisdividedintothreesections.Section0.1givesanintroductionto categorytheory,andinparticularthecategories Set ofsetsandfunctions,and Rel of setsandrelations.Section0.2introducesthemathematicalformalismofHilbertspaces thatunderliesquantummechanics,anddefinesthecategories Vect ofvectorspacesand linearmaps,and Hilb ofHilbertspacesandboundedlinearmaps.Section0.3recallsthe basicsofquantumtheory,includingthestandardinterpretationofstates,dynamicsand measurementandthequantumteleportationprocedure.

0.1CategoryTheory

Thissectiongivesabriefintroductiontocategorytheory.Wefocusinparticularonthe category Set ofsetsandfunctions,andthecategory Rel ofsetsandrelations,andpresent amatrixcalculusforrelations.Weintroducetheideaofcommutingdiagrams,anddefine isomorphisms,groupoids,skeletalcategories,oppositecategoriesandproductcategories. Wethendefinefunctors,equivalencesandnaturaltransformations,andalsoproducts, equalizersandidempotents.

0.1.1 Categories

Categoriesareformedfromtwobasicstructures: objectsA, B, C, ...,and morphismsA f B goingbetweenobjects.Inthisbook,wewilloftenthinkofanobjectasa system,and amorphism A f → B asa process underwhichthesystem A becomesthesystem B. Categoriescanbeconstructedfromalmostanyreasonablenotionofsystemandprocess. Hereareafewexamples:

•physicalsystems,andphysicalprocessesgoverningthem;

•datatypesincomputerscience,andalgorithmsmanipulatingthem;

•algebraicorgeometricstructuresinmathematics,andstructure-preservingfunctions;

•logicalpropositions,andimplicationsbetweenthem.

Categorytheoryisquitedifferentfromotherareasofmathematics.Whileacategoryis itselfjustanalgebraicstructure—muchlikeagroup,ringorfield—wecanusecategories toorganizeandunderstandothermathematicalobjects.Thishappensinasurprisingway: byneglectingallinformationaboutthestructureoftheobjects,andfocusingentirelyon relationships between theobjects.Categorytheoryisthestudyofthepatternsformedby theserelationships.Whileatfirstthismayseemlimiting,itisinfactempowering,asit becomesagenerallanguageforthedescriptionofmanydiversestructures.

Hereisthedefinitionofacategory.

Definition0.1. A category C consistsofthefollowingdata:

•acollectionOb(C) of objects;

•foreverypairofobjects A and B,acollection C(A, B) of morphisms,with f ∈ C(A, B) written A f B;

•foreverypairofmorphisms A f B and B g C withcommonintermediateobject, a compositeA g ◦f C;

•foreveryobject A an identitymorphismA idA A.

Thesemustsatisfythefollowingproperties,forallobjects A, B, C, D,andallmorphisms

associativity

identity

Wewillalsosometimesusethenotation f : A B foramorphism f ∈ C(A, B).

Fromthisdefinitionweseequiteclearlythatthemorphismsare‘moreimportant’ thantheobjects;afterall,everyobject A iscanonicallyrepresentedbyitsidentity

morphismidA .Thisseemslikeasimplepoint,butitisasignificantdeparturefrom muchofclassicalmathematics,inwhichparticularstructures(likegroups)playa moreimportantrolethanthestructure-preservingmapsbetweenthem(likegroup homomorphisms.)

Ourdefinitionofacategoryreferstocollectionsofobjectsandmorphisms,rather thansets,becausesetsaretoosmallingeneral.Thecategory Set definedinSection0.1.2 illustratesthiswell,sinceRussell’sparadoxpreventsthecollectionofallsetsfrombeinga set.However,suchsizeissueswillnotplayaroleinthisbook,andwewillusesettheory naivelythroughout.(SeetheNotesandFurtherReadingattheendofthischapterfor moresophisticatedreferencesoncategorytheory.)

0.1.2 TheCategory Set

Themostbasicrelationshipsbetweensetsaregivenbyfunctions.

Definition0.2. Forsets A and B,a functionA f B comprises,foreach a ∈ A,achoiceof element f (a) ∈ B.Wewrite f : a → f (a) todenotethischoice.

Writing ∅ fortheemptyset,thedataforafunction ∅ A canbeprovidedtrivially;there isnothingforthe‘foreach’partofthedefinitiontodo.Sothereisexactlyonefunction ofthistypeforeveryset A.However,functionsoftype A ∅ cannotbeconstructed unless A =∅.Ingeneralthereare |B||A| functionsoftype A B,where |−| indicatesthe cardinalityofaset.

Wecannowusethistodefinethecategoryofsetsandfunctions.

Definition0.3(Set, FSet). Inthecategory Set ofsetsandfunctions:

• objects aresets A, B, C, ;

• morphisms arefunctions f , g , h, ...;

• composition of A f B and B g C isthefunction g ◦ f : a → g (f (a));thisisthe reasonthestandardnotation g ◦ f isnotintheotherorder,eventhoughthatwould bemorenaturalinsomeequationssuchas(0.5);

• theidentitymorphism on A isthefunctionidA : a → a

Write FSet fortherestrictionof Set tofinitesets.

Wewilloftenthinkofafunction A f → B inadynamicalway,asindicatinghowelements of A canevolveintoelementsof B.Thissuggeststhefollowingsortofpicture:

0.1.3 TheCategory Rel

Relationsgiveamoregeneralnotionofprocessbetweensets.

Definition0.4. Givensets A and B,a relationA R → B isasubset R ⊆ A × B.

Ifelements a ∈ A and b ∈ B satisfy (a, b) ∈ R ,weoftenindicatethisbywriting aRb,or even a ∼ b when R isclear.Sinceasubsetcanbedefinedbygivingitselements,wecan defineourrelationsbylistingtherelatedelements,intheform a1 Rb1 , a2 Rb2 , a3 Rb3 and soon.

Wecanthinkofarelation A R → B inadynamicalway,generalizing(0.3):

Thedifferencewithfunctionsisthatthispictureindicatesinterpretingarelationasakind ofnondeterministicclassicalprocess:eachelementof A canevolveintoanyelementof B towhichitisrelated.Nondeterminismentersherebecauseanelementof A canrelate tomorethanoneelementof B,sounderthisinterpretation,wecannotpredictperfectly howthesystemwillevolve.Anelementof A couldalsoberelatedto no elementsof B:we interpretthistomeanthat,fortheseelementsof A,thedynamicalprocesshalts.Becauseof thisinterpretation,thecategoryofrelationsisimportantinthestudyofnondeterministic classicalcomputing.

Supposewehaveapairofrelations,withthecodomainofthefirstequaltothedomain ofthesecond:

Ourinterpretationofrelationsasdynamicalprocessesthensuggestsanaturalnotionof composition:anelement a ∈ A isrelatedto c ∈ C ifthereissome b ∈ B with aRb and bSc.Fortheexamplehere,thisgivesrisetothefollowingcompositerelation:

AB R
(0.4)

Thisdefinitionofrelationalcompositionhasthefollowingalgebraicform:

Wecanwritethisdifferentlyas

where ∨ represents logicaldisjunction (or),and ∧ represents logicalconjunction (and).

Comparingthiswiththedefinitionofmatrixmultiplication,weseeastrongsimilarity:

Thissuggestsanotherwaytointerpretarelation:asamatrixoftruthvalues.Forthe examplerelation(0.4),thisgivesthefollowingmatrix,wherewewrite0forfalseand1 fortrue:

Compositionofrelationsisthenjustgivenbyordinarymatrixmultiplication,withlogical disjunctionandconjunctionreplacing+and ×,respectively(sothat1 + 1 = 1).

Thereisaninterestinganalogybetweenquantumdynamicsandthetheoryofrelations. First,arelation A R B tellsus,foreach a ∈ A and b ∈ B,whetheritis possible for a to produce b,whereasacomplex-valuedmatrix H f K givesusthe amplitude for a toevolve to b.Second,relationalcompositiontellsusthe possibility ofevolvingviaanintermediate

pointthroughasum-of-pathsformula,whereasmatrixcompositiontellsusthe amplitude forthistohappen.

Theintuitionwehavedevelopedleadstothefollowingcategory.

Definition0.5(Rel, FRel). Inthecategory Rel ofsetsandrelations:

• objects aresets A, B, C, ...;

• morphisms arerelations R ⊆ A × B;

• composition of A R B and B S C istherelation

• theidentitymorphism on A istherelation {(a, a) ∈ A × A | a ∈ A}.

Write FRel fortherestrictionof Rel tofinitesets.

While Set isasettingforclassicalphysics,and Hilb (tobeintroducedinSection0.2) isasettingforquantumphysics, Rel issomewhereinthemiddle.Itseemslikeitshould bealotlike Set,butinfact,itspropertiesaremuchmorelikethoseof Hilb.Thismakes itanexcellenttest-bedforinvestigatingdifferentaspectsofquantummechanicsfroma categoricalperspective.

0.1.4 Morphisms

Itoftenhelpstodrawdiagramsofmorphisms,indicatinghowtheycompose.Hereisan example:

Wesayadiagram commutes wheneverypossiblepathfromoneobjectinittoanother isthesame.Intheexample,thismeans i ◦ f = k ◦ h and g = j ◦ i.Itthenfollowsthat g ◦ f = j ◦ k ◦ h,wherewedonotneedtowriteparenthesesthankstotheassociativity equation(0.1).Thus,wehavetwowaystospeakaboutequalityofcompositemorphisms: byalgebraicequations,orbycommutingdiagrams.

Thefollowingtermsareveryusefulwhendiscussingmorphisms.Theterm‘operator’ thatfollowscomesfromphysics.

Definition0.6(Domain,codomain,endomorphism,operator). Foramorphism A f B,its domain istheobject A,andits codomain istheobject B.If A = B thenwecall f an endomorphism or operator .Wesometimeswritedom(f ) = A andcod(f ) = B.

Definition0.7(Isomorphism,retraction). Amorphism A f B isan isomorphism when ithasan inverse morphism B f 1 A satisfying:

Wethensaythat A and B are isomorphic,andwrite A B.Ifonlytheleftorrightequation of(0.10)holds,then f iscalled left- or right-invertible,respectively.Aright-invertible morphismisalsocalleda retraction.

Lemma0.8. Ifamorphismhasaninverse,thenthisinverseisunique.

Proof. If g and g areinversesfor f ,then:

Example0.9. Letusseewhatisomorphismsarelikeinourexamplecategories:

•in Set,theisomorphismsareexactlythebijectionsofsets;

•in Rel,theisomorphismsarethegraphsofbijections:arelation A R B isan isomorphismwhenthereissomebijection A f B suchthat aRb ⇔ f (a) = b.

Thenotionofisomorphismleadstosomeimportanttypesofcategory.

Definition0.10(Skeletalcategory). Acategoryis skeletal whenanytwoisomorphic objectsareequal.

WewillseeinSection0.1.6thateverycategoryis equivalent toaskeletalcategory,which meanstheyencodeessentiallythesamealgebraicdata.

Definition0.11(Groupoid,group). A groupoid isacategoryinwhicheverymorphism isanisomorphism.A group isagroupoidwithoneobject.

Ofcourse,thisdefinitionofgroupagreeswiththeordinaryone. Manyconstructionswithandpropertiesofcategoriescanbeeasilydescribedinterms ofmorphisms.

Definition0.12(Oppositecategory). Givenacategory C,its opposite Cop isacategory withthesameobjects,butwith Cop (A, B) givenby C(B, A).Thatis,themorphisms A B in Cop aremorphisms B A in C.

Definition0.13(Productcategory). Forcategories C and D,their product isacategory C × D,whoseobjectsarepairs (A, B) ofobjects A ∈ Ob(C) and B ∈ Ob(D),andwhose morphismsarepairs (A, B) (f ,g ) (C, D) with A f C and B g D.

Definition0.14(Discretecategory). Acategoryis discrete whenallthemorphismsare identities.

Definition0.15(Indiscretecategory). Acategoryis indiscrete whenthereisaunique morphism A B foreachtwoobjects A and B.

0.1.5 GraphicalNotation

Thereisagraphicalnotationformorphismsandtheircomposites.Drawanobject A as follows:

Itisjustaline.Infact,youshouldthinkofitasapictureoftheidentitymorphism A idA A. Remember,incategorytheory,themorphismsaremoreimportantthantheobjects.

Amorphism A f → B isdrawnasaboxwithone‘input’atthebottom,andone‘output’ atthetop:

Compositionof A f → B and B g → C isthendrawnbyconnectingtheoutputofthefirst boxtotheinputofthesecondbox:

Theidentitylaw f ◦ idA = f = idB ◦ f andtheassociativitylaw (h ◦ g ) ◦ f = h ◦ (g ◦ f ) thenlooklike:

Tomaketheselawsimmediatelyobvious,wechoosetonotdepicttheidentitymorphisms idA atallandnotindicatethebracketingofcomposites.

Thegraphicalcalculusisusefulbecauseit‘absorbs’theaxiomsofacategory, makingthemaconsequenceofthenotation.Thisisbecausetheaxiomsofacategory areaboutstringingthingstogetherinsequence.Atafundamentallevel,thisconnects tothegeometryoftheline,whichisalso one-dimensional.Ofcourse,thisgraphical representationisquitefamiliar:youusuallydrawithorizontallyandcallitalgebra.

0.1.6 Functors

Rememberthemottothatincategorytheory,morphismsaremoreimportantthan objects.Categorytheorytakesitsownmedicinehere:thereisaninterestingnotionof ‘morphismbetweencategories’,asgivenbythefollowingdefinition.

Definition0.16(Functor,covariance,contravariance). Givencategories C and D,a functorF : C D isdefinedbythefollowingdata:

•foreachobject A ∈ Ob(C),anobject F (A) ∈ Ob(D);

•foreachmorphism A f B in C,amorphism F (A) F (f ) F (B) in D.

Thisdatamustsatisfythefollowingproperties:

• F (g ◦ f ) = F (g ) ◦ F (f ) forallmorphisms A f B and B g C in C;

• F (idA ) = idF (A) foreveryobject A in C.

Functorsareimplicitly covariant .Therearealso contravariant versionsreversingthedirectionofmorphisms: F (g ◦ f ) = F (f ) ◦ F (g ).Wewillonlyusethiscovariantdefinition, andmodelthecontravariantversion C D asacovariantfunctor Cop D.Afunctor betweengroupsisalsocalleda grouphomomorphism;ofcoursethiscoincideswiththe usualnotion.

Wecanusefunctorstogiveanotionofequivalenceforcategories.

Definition0.17(Equivalence). Afunctor F : C D isan equivalence whenitis:

• full,meaningthatthefunctions C(A, B) D F (A), F (B) givenby f → F (f ) are surjectiveforall A, B ∈ Ob(C);

• faithful,meaningthatthefunctions C(A, B) D F (A), F (B) givenby f → F (f ) areinjectiveforall A, B ∈ Ob(C);

• essentiallysurjectiveonobjects,meaningthatforeachobject B ∈ Ob(D) thereisan object A ∈ Ob(C) suchthat B F (A)

Iftwocategoriesareequivalent,thenoneisjustasgoodastheotherforthepurposes ofdoingcategorytheory,eventhoughtheymightbedefinedinquiteadifferentway. Nonetheless,onemightbemucheasiertoworkwiththantheother,andthatisonereason whythenotionofequivalenceissouseful.

Acategory C isa subcategory ofacategory D wheneveryobjectof C isanobjectof D, everymorphismof C isamorphismof D,andcompositionandidentitiesin C arethe sameasin D.Inotherwords,theinclusion C D isafaithfulfunctor. Everycategoryhasaskeleton,asmallercategorywiththesamealgebraicstructure,that isequivalenttoit.

Definition0.18(Skeleton). A skeleton ofacategory C isasubcategory S suchthatevery objectin C isisomorphic(in C)toexactlyoneobjectin S.

Intuitively,askeletonisbuiltbyrestrictingthecategory C tocontainjustoneobjectfrom eachisomorphismclass.Thedefinitionsays,inotherwords,thattheinclusionfunctor S C isanequivalenceandthat S isskeletal.

0.1.7 NaturalTransformations

Justasafunctorisamapbetweencategories,sothereisanotionofamapbetween functors,calleda naturaltransformation

Definition0.19(Naturaltransformation,naturalisomorphism). Givenfunctors F : C D and G : C D,a naturaltransformation ζ : F G isanassignmentto everyobject A in C ofamorphism F (A) ζA G(A) in D,suchthatthefollowingdiagram commutesforeverymorphism A f B in C:

Ifeverycomponent ζA isanisomorphismthen ζ iscalleda naturalisomorphism,and F and G arecalled naturallyisomorphic. Manyimportantconceptsinmathematicscanbedefinedinasimplewayusingfunctors andnaturaltransformations,suchasthefollowing.

Example0.20. A grouprepresentation isafunctor G Vect,where G isagroup regardedasacategorywithoneobject(seeDefinition0.11.)An intertwiner isanatural transformationbetweensuchfunctors.

Thenotionofnaturalisomorphismleadstoanothercharacterizationofequivalenceof categories.

Definition0.21(Equivalencebynaturalisomorphism). Afunctor F : C D isan equivalence ifandonlyifthereexistsafunctor G : D C andnaturalisomorphisms G ◦ F idC andidD F ◦ G.

AfunctorisanequivalencebyDefinition0.21justwhenitisanequivalenceby Definition0.17,andsoweabuseterminologymildly,usingtheword‘equivalence’for bothconcepts.Itisinterestingtoconsiderthedifferencebetweenthesedefinitions:while Definition0.17iswrittenintermsoftheinternalstructureofthecategoriesinvolved, intheformoftheirobjectsandmorphisms,Definition0.21iswrittenintermsoftheir externalcontext,givenbythefunctorsandnaturaltransformationsbetweenthem.This isacommondichotomyincategorytheory,with‘internal’conceptsoftenbeingmore elementaryanddirect,whiletheassociated‘external’perspective,althoughmakinguseof moresophisticatednotions,isoftenmorepowerfulandelegant.Werevisitthisexternal notionofequivalenceinChapter8,fromtheperspectiveofhighercategorytheory.

0.1.8 Limits

Limits arerecipesforfindingobjectsandmorphismswith universalproperties,withgreat practicaluseincategorytheory.Wewon’tdescribethegeneralcasehere,butjustthe importantspecialcasesofproducts,equalizers,terminalobjectsandtheirdualnotions.

Togettheidea,itisusefultothinkaboutthedisjointunion S + T ofsets S and T . Itisnotjustabareset;itcomesequippedwithfunctions S iS S + T and T iT S + T thatshowhowtheindividualsetsembedintothedisjointunion.Andfurthermore,these functionshaveaspecialproperty:afunction S + T f U correspondsexactlytoapairof functionsoftypes S fS U and T fT U ,suchthat f ◦ iS = fS and f ◦ iT = fT .Theconcepts oflimitandcolimitgeneralizethisobservation.

Wenowdefineproductandcoproduct,andalsoterminalandinitialobject.

Definition0.22(Product,coproduct). Givenobjects A and B,a product isanobject A × B togetherwithmorphisms A × B pA A and A × B pB B,suchthatanytwomorphisms X f A and X g B allowauniquemorphism f g : X A × B with pA ◦ f g = f and pB ◦ f g = g .Thefollowingdiagramsummarizestheserelationships:

A coproduct isthedualnotion,thatreversesthedirectionsofallthearrowsinthis diagram.Givenobjects A and B,acoproductisanobject A + B equippedwithmorphisms A iA A + B and B iB A + B,suchthatforanymorphisms A f X and B g X ,thereisa uniquemorphism ( fg ) : A + B X suchthat ( fg ) ◦ iA = f and ( fg ) ◦ iB = g .

Definition0.23(Terminalobject,initialobject). Anobject A is terminal ifforevery object X ,thereisexactlyonemorphism X A.Itis initial ifforeveryobject X ,thereis exactlyonemorphism A X .

Acategorymaynothaveanyofthesestructures,butiftheyexist,theyareuniqueupto isomorphism.

Definition0.24(Cartesiancategory). Acategoryis Cartesian whenithasaterminal objectandproductsofanypairofobjects.

Thesestructuresexistinourmainexamplecategories.

Example0.25. Products,coproducts,terminalobjectsandinitialobjectstakethe followingformsinourmainexamplecategories:

•in Set,productsaregivenbytheCartesianproduct,andcoproductsbythedisjoint union,any1-elementsetisaterminalobject,andtheemptysetistheinitialobject;

•in Rel,productsandcoproductsarebothgivenbythedisjointunion,andtheempty setisboththeterminalandinitialobject.

Givenapairoffunctions S f ,g T ,itisinterestingtoaskonwhichelementsof S they takethesamevalue.Categorytheorydictatesthatweshouldn’taskaboutelements, butusemorphismstogetthesameinformationusingauniversalproperty.Thisleads tothenotionofequalizer,astructurethatmayormaynotexistinanyparticular category.

Definition0.26. Formorphisms A f ,g B,their equalizer isamorphism E e A satisfying f ◦ e = g ◦ e,suchthatanymorphism E e A satisfying f ◦ e = g ◦ e allowsaunique E m E with e = e ◦ m:

The coequalizer of f and g istheirequalizerintheoppositecategory.

Example0.27. Letusseewhatequalizerslooklikeinourexamplecategories.

•Thecategories Set, Vect and Hilb (seeSection0.2)haveequalizersforallpairsof parallelmorphisms.Anequalizerfor A f ,g B istheset E ={a ∈ A | f (a) = g (a)}, equippedwithitsembedding E e A;thatis,itisthelargestsubsetof A onwhich f and g agree.

•Thecategory Rel doesnothaveallequalizers.Forexample,considertherelation R ={(y, z) ∈ R2 | y < z ∈ R} : R R.Suppose E : X R wereanequalizerof R andidR .Then R ◦ R = idR ◦ R ,sothereisarelation M : R X with R = E ◦ M . Now E ◦ (M ◦ E) = (E ◦ M ) ◦ E = R ◦ E = idR ◦ E = E,andsince S = idX isthe uniquemorphismsatisfying E ◦ S = E,wemusthave M ◦ E = idX .Butthen xEy and yMx forsome x ∈ X and y ∈ R.Itfollowsthat y(E ◦ M )y,thatis, y < y,which isacontradiction.

Akernelisaspecialkindofequalizer.

Definition0.28. A kernel ofamorphism A f B isanequalizerof f andthe zero morphism A 0 B (seeSection2.2.)

Alastinstanceofuniversalpropertiesistheideaofsplitidempotents.

Definition0.29(Idempotent,splitting). Anendomorphism A f A iscalled idempotent when f ◦ f = f .Anidempotent A f Asplits whenthereexistanobject ˆ f andmorphisms A pf ˆ f and ˆ f if A suchthatthefollowinghold:

Givensuchasplitidempotent,theinjection ˆ f if A givesanequalizerof f andidA ,and theprojection A pf ˆ f givesacoequalizerof f andidA

0.2HilbertSpaces

Thissectionintroducesthemathematicalformalismthatunderliesquantumtheory: complexvectorspaces,innerproductsandHilbertspaces.Wedefinethecategories Vect and Hilb,anddefinebasicconceptssuchasorthonormalbases,linearmaps,matrices, dimensionsanddualsofHilbertspaces.Wethenintroducetheadjointofalinearmap betweenHilbertspaces,anddefinethetermsunitary,isometry,partialisometryand positive.WealsodefinethetensorproductofHilbertspacesandintroducetheKronecker productofmatrices.

0.2.1

VectorSpaces

Avectorspaceisacollectionofelementsthatcanbeaddedtooneanother,andscaled.

Definition0.30(Vectorspace). A vectorspace isaset V withachosenelement 0 ∈ V ,anadditionoperation + : V × V V ,andascalarmultiplicationoperation · : C × V V ,satisfyingthefollowingpropertiesforall a, b, c ∈ V and s, t ∈ C:

• additiveassociativity: a + (b + c) = (a + b) + c;

• additivecommutativity: a + b = b + a;

• additiveunit : a + 0 = a;

• additiveinverses:thereexists a ∈ V suchthat a + ( a) = 0;

• additivedistributivity: s (a + b) = (s a) + (s b)

• scalarunit :1 · a = a;

• scalardistributivity: (s + t ) · a = (s · a) + (t · a);

• scalarcompatibility: s (t a) = (st ) a.

Theprototypicalexampleofavectorspaceis Cn ,theCartesianproductof n copiesofthe complexnumbers.

Definition0.31(Linearmap,anti-linearmap). A linearmap isafunction f : V W betweenvectorspaces,withthefollowingproperties,forall a, b ∈ V and s ∈ C:

An anti-linearmap isafunctionthatsatisfies(0.18),butinsteadof(0.19),satisfies

wherethestardenotescomplexconjugation.

Vectorspacesandlinearmapsformacategory.

Definition0.32(Vect, FVect). Inthecategory Vect ofvectorspacesandlinearmaps:

• objects arecomplexvectorspaces;

• morphisms arelinearfunctions;

• composition iscompositionoffunctions;

• identitymorphisms areidentityfunctions.

Write FVect fortherestrictionof Vect tothosevectorspacesthatareisomorphicto Cn forsomenaturalnumber n;thesearealsocalled finite-dimensional,seeDefinition0.34.

Anymorphism f : V W in Vect hasakernel,namelytheinclusionofker(f ) = {v ∈ V | f (v) = 0} into V .Hence,kernelsinthecategoricalsensecoincidepreciselywith kernelsinthesenseoflinearalgebra.

Definition0.33. The directsum ofvectorspaces V and W isthevectorspace V ⊕ W , whoseelementsarepairs (a, b) ofelements a ∈ V and b ∈ W ,withentrywiseaddition andscalarmultiplication.

Directsumsarebothproductsandcoproductsinthecategory Vect.Similarly,thezerodimensionalspaceisbothterminalandinitialin Vect.

0.2.2 BasesandMatrices

Oneofthemostimportantstructuresavectorspacecanhaveisa basis.Abasisgives risetothenotionofdimensionofavectorspace,andletsusrepresentlinearmapsusing matrices.

Definition0.34(Basis). Foravectorspace V ,afamilyofelements {ei } is linearly independent wheneveryelement a ∈ V canbeexpressedasafinitelinearcombination a = i ai ei with coefficientsai ∈ C inatmostoneway.Itisa basis ifadditionallyany a ∈ V canbeexpressedassuchafinitelinearcombination.

Everyvectorspaceadmitsabasis,andanytwobasesforthesamevectorspacehavethe samecardinality.

Definition0.35(Dimension,finite-dimensionality). The dimension ofavectorspace V ,writtendim(V ),isthecardinalityofanybasis.Avectorspaceis finite-dimensional when ithasafinitebasis.

Ifvectorspaces V and W havebases {di } and {ej },andwefixsomeorderonthebases,we canrepresentalinearmap V f W asthematrixwithdim(W ) rowsanddim(V ) columns, whoseentryatrow i andcolumn j isthecoefficient f (dj )i .Compositionoflinearmaps thencorrespondstomatrixmultiplication(0.7).Thisdirectlyleadstoacategory.

Definition0.36(MatC ). Intheskeletalcategory MatC :

• objects arenaturalnumbers0,1,2, ;

• morphisms n m arecomplexmatriceswith m rowsand n columns;

• composition isgivenbymatrixmultiplication;

• identities n idn n aregivenby n-by-n matriceswithentries1onthemaindiagonal, and0elsewhere.

Thistheoryofmatricesis‘justasgood’asthetheoryoffinite-dimensionalvectorspaces, madeprecisebythecategorytheorydevelopedinSection0.1.

Proposition0.37. Thereisanequivalenceofcategories MatC FVect thatsendsnto Cn andamatrixtoitsassociatedlinearmap.

Proof. Becauseeveryfinite-dimensionalcomplexvectorspace H isisomorphicto Cdim(H ) ,thefunctor R isessentiallysurjectiveonobjects.Itisfullandfaithfulsince thereisanexactcorrespondencebetweenmatricesandlinearmapsforfinite-dimensional vectorspaces.

Forsquarematrices,thetraceisanimportantoperation.

Definition0.38(Trace). Forasquarematrixwithentries mij ,its trace isthesum i mii ofitsdiagonalentries.

0.2.3 HilbertSpaces

Hilbertspacesarestructuresthatarebuiltonvectorspaces.Theextrastructureletsus defineanglesanddistancesbetweenvectors,andisusedinquantumtheorytocalculate probabilitiesofmeasurementoutcomes.

Definition0.39(Innerproduct). An innerproduct onacomplexvectorspace V isa function −|− : V × V C thatis:

• conjugate-symmetric:forall a, b ∈ V ,

• linear inthesecondargument:forall a, b, c ∈ V and s ∈ C,

positivedefinite:forall a ∈ V ,

Definition0.40(Norm). Foravectorspacewithinnerproduct,the norm ofanelement v is v = √ v | v ,anonnegativerealnumber.

Thecomplexnumberscarryacanonicalinnerproduct:

Theinducednormsatisfiesthetriangleinequality a + b ≤ a + b byvirtueofthe Cauchy–Schwarzinequality | a | b |2 ≤ a | a · b | b ,thatholdsinanyvectorspacewith aninnerproduct.Thankstotheseproperties,itmakessensetothinkof a b asthe distancebetweenvectors a and b

AHilbertspaceisaninnerproductspaceinwhichitmakessensetoaddinfinitelymany vectorsincertaincases.

Definition0.41(Hilbertspace). A Hilbertspace isavectorspace H withaninner productthatis complete inthefollowingsense:ifasequence v1 , v2 , ... ofvectorssatisfies ∞ i=1 vi < ∞,thenthereisavector v suchthat v n i=1 vi tendstozeroas n goes toinfinity.

Everyfinite-dimensionalvectorspacewithinnerproductisnecessarilycomplete.Any vectorspacewithaninnerproductcanbecompletedtoaHilbertspacebyformallyadding theappropriatelimitvectors.

ThereisanotionofboundedmapbetweenHilbertspacesthatmakesuseoftheinner productstructure.Theideaisthatforeachmapthereissomemaximumamountbywhich thenormofavectorcanincrease.

Definition0.42(Boundedlinearmap). Alinearmap f : H K betweenHilbert spacesis bounded whenthereexistsanumber r ∈ R suchthat f (a) ≤ r · a forall a ∈ H .

Everylinearmapbetweenfinite-dimensionalHilbertspacesisbounded.

Hilbertspacesandboundedlinearmapsformacategory.Thiscategorywillbethemain examplethroughoutthebooktomodelphenomenainquantumtheory.

Definition0.43(Hilb, FHilb). Inthecategory Hilb ofHilbertspacesandboundedlinear maps:

• objects areHilbertspaces;

• morphisms areboundedlinearmaps;

• composition iscompositionoflinearmapsasordinaryfunctions;

• identitymorphisms aregivenbytheidentitylinearmaps.

Write FHilb fortherestrictionof Hilb tofinite-dimensionalHilbertspaces. Thisdefinitionisperhapssurprising,especiallyinfinitedimensions:sinceeverylinear mapbetweenHilbertspacesisbounded, FHilb isanequivalentcategoryto FVect.In particular,theinnerproductsplaynoessentialrole.WewillseeinSection2.3howto modelinnerproductscategorically,usingtheideaof daggers Hilbertspaceshaveamorediscerningnotionofbasis.

Definition0.44(Basis,orthogonalbasis,orthonormalbasis). ForaHilbertspace H ,an orthogonalbasis isafamilyofelements {ei } withthefollowingproperties:

•theyare pairwiseorthogonal,thatis, ei | ej = 0forall i = j;

•everyelement a ∈ H canbewrittenasaninfinitelinearcombinationof ei ;that is,thereare coefficientsai ∈ C forwhich a n i=1 ai ei tendstozeroas n goesto infinity.

Itis orthonormal whenadditionally ei | ei = 1forall i.

Anyorthogonalfamilyofelementsislinearlyindependent.Forfinite-dimensional Hilbertspaces,theordinarynotionofbasisasavectorspace,asgivenbyDefinition0.34, isstilluseful.Hence,oncewefix(ordered)basesonfinite-dimensionalHilbertspaces, linearmapsbetweenthemcorrespondtomatrices,justaswithvectorspaces.ForinfinitedimensionalHilbertspaces,however,havingabasisfortheunderlyingvectorspaceis rarelymathematicallyuseful.

Iftwovectorspacescarryinnerproducts,wecangiveaninnerproducttotheirdirect sum,leadingtothedirectsumofHilbertspaces.

Definition0.45(Directsum). The directsum ofHilbertspaces H and K isthevector space H ⊕ K ,madeintoaHilbertspacebytheinnerproduct (a1 , b1 ) | (a2 , b2 ) = a1 | a2 + b1 | b2 .

Directsumsprovidebothproductsandcoproductsforthecategory Hilb.Hilbert spaceshavethegoodpropertythatanyclosedsubspacecanbecomplemented.Thatis,if theinclusion U → V isamorphismof Hilb satisfying u U = u H ,thenthereexists anotherinclusionmorphism U ⊥ → V of Hilb with V = U ⊕ U ⊥ .Explicitly, U ⊥ isthe orthogonalsubspace {a ∈ V |∀b ∈ U : a | b = 0}.

0.2.4 AdjointLinearMaps

Theinnerproductgivesrisetothe adjoint ofaboundedlinearmap.

Definition0.46. Foraboundedlinearmap f : H K ,its adjointf † : K H isthe uniquelinearmapwiththefollowingproperty,forall a ∈ H and b ∈ K :

Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.