MathematicalLogic
OnNumbers,Sets,Structures,andSymmetry
RomanKossak
CityUniversityofNewYork
NewYork,NY,USA
SpringerGraduateTextsinPhilosophy
ISBN978-3-319-97297-8ISBN978-3-319-97298-5(eBook) https://doi.org/10.1007/978-3-319-97298-5
LibraryofCongressControlNumber:2018953167
©SpringerInternationalPublishingAG,partofSpringerNature2018 Thisworkissubjecttocopyright.AllrightsarereservedbythePublisher,whetherthewholeorpartof thematerialisconcerned,specificallytherightsoftranslation,reprinting,reuseofillustrations,recitation, broadcasting,reproductiononmicrofilmsorinanyotherphysicalway,andtransmissionorinformation storageandretrieval,electronicadaptation,computersoftware,orbysimilarordissimilarmethodology nowknownorhereafterdeveloped.
Theuseofgeneraldescriptivenames,registerednames,trademarks,servicemarks,etc.inthispublication doesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexemptfromtherelevant protectivelawsandregulationsandthereforefreeforgeneraluse.
Thepublisher,theauthorsandtheeditorsaresafetoassumethattheadviceandinformationinthisbook arebelievedtobetrueandaccurateatthedateofpublication.Neitherthepublishernortheauthorsor theeditorsgiveawarranty,expressorimplied,withrespecttothematerialcontainedhereinorforany errorsoromissionsthatmayhavebeenmade.Thepublisherremainsneutralwithregardtojurisdictional claimsinpublishedmapsandinstitutionalaffiliations.
ThisSpringerimprintispublishedbytheregisteredcompanySpringerNatureSwitzerlandAG Theregisteredcompanyaddressis:Gewerbestrasse11,6330Cham,Switzerland
Preface
Logicandsermonsneverconvince, Thedampofthenightdrivesdeeperintomysoul. (Onlywhatprovesitselftoeverymanandwomanisso, Onlywhatnobodydeniesisso.)
WaltWhitman,LeavesofGrass
In WhyIsTherePhilosophyofMathematicsatAll? [8],IanHackingwrites:
Yetalthoughmostmembersofourspecieshavesomecapacityforgeometricaland numericalconcepts,veryfewhumanbeingshavemuchcapacityfordoingoreven understandingmathematics.Thisisoftenheldtobetheconsequenceofbadeducation,but althougheducationcansurelyhelp,thereisnoevidencethatvastdisparityoftalent,oreven interestin,mathematics,isaresultofbadpedagogy...Aparadox:wearethemathematical animal.Afewofushavemadeastonishingmathematicaldiscoveries,afewmoreofus canunderstandthem.Mathematicalapplicationshaveturnedouttobeakeytounlockand disciplinenaturetoconformtosomeofourwishes.Butthesubjectrepelsmosthuman beings.
Itallringstruetoanyonewhohasevertaughtthesubject.WhenIteach undergraduates,Isometimessay,“Mathdoesnotmakeanysense,right?”towhich Ihearinunison“Right.”Myownschoolexperiencehasnotbeenmuchdifferent. EventhoughIwasconsidered“goodatmath,”itonlymeantthatIcouldfollow instructionsanddohomeworktoasatisfyingresult.OnlyoccasionallyI’dhave momentsIwasproudof.Inmyfirsthighschoolyear,Ihadamathematicsprofessor aboutwhomlegendsweretold.Everyoneknewthatinordertosurviveinhisclass, onehadto understand.Atanytimeyoucouldbecalledtotheblackboardtobe askedapenetratingquestion.Wewitnessedhumiliatingmomentswhensomeone’s ignorancewasruthlesslyrevealed.Once,whenthetopicwassquarerootsofreal numbers,Iwascalledandasked:
–Doeseverypositivenumberhaveasquareroot?
–Yes.
–Whatisthesquarerootoffour?
–Two.
–Whatisthesquarerootoftwo?
–Thesquarerootoftwo.
–Good.Whatisthesquarerootofthesquarerootoftwo.
–Thesquarerootofthesquarerootoftwo.
–Good.Sitdown.
Theprofessorsmiled.Helikedmyanswers,andIwashappywiththemtoo. ItseemedIhadunderstoodsomethingandthatthisunderstandingwassomehow complete.Thetruthwassimple,andonceyougraspedit,therewasnothingmore toit,justplainsimpletruth.Perhapssuchraresatisfyingmomentswerethereason Idecidedtostudymathematics.
InmyfreshmenyearattheUniversityofWarsaw,ItookAnalysis,Abstract Algebra,Topology,MathematicalLogic,andIntroductiontoComputerScience.I wascompletelyunpreparedforthelevelofabstractionofthesecourses.Ithought thatwewouldjustcontinuewiththekindofmathematicswestudiedinhigh school.Iexpectedmoreofthesamebutmorecomplicated.Perhapssomeadvanced formulasforsolvingalgebraicandtrigonometricequationsandmoreelaborate constructionsinplaneandthree-dimensionalgeometry.Instead,thecoursein analysisbeganwith defining realnumbers.Somethingwetookforgrantedandwe thoughtweknewwellnowneededadefinition!Myanswersattheblackboardthat Ihadbeensoproudofturnedouttoberathernaive.Thetruthwasnotsoplain andsimpleafterall.Andtomakemattersworse,thecoursefollowedwithaproof ofexistenceanduniquenessoftheexponentialfunction f(x) = e x ,andinthe process,wehadtolearnaboutcomplexnumbersandthefundamentaltheoremof algebra.Insteadofadvancedapplicationsofmathematicswehavealreadylearned, weweregoingback,askingmoreandmorefundamentalquestions.Itwasrather unexpected.Thealgebraandtopologycourseswereevenharder.Insteadofthe alreadyfamiliarplanes,spheres,cones,andcylinders,nowwewereexposedto generalalgebraicsystemsandtopologicalspaces.Insteadofconcreteobjectsone couldtrytovisualize,westudiedinfinitespaceswithsurprisinggeneralproperties expressedintermsofalgebraicsystemsthatwereassociatedwiththem.Ilikedthe prospectoflearningallthat,andinparticularthepromisethatintheend,after allthishigh-levelabstractstuffgotsortedout,therewouldbeareturntomore down-to-earthapplications.Butwhatwasevenmoreattractivewastheattemptto gettothebottomofthings,tounderstandcompletelywhatthiselaborateedificeof mathematicswasfoundedupon.Idecidedtospecializeinmathematicallogic.
Myfirstencounterswithmathematicallogicweretraumatic.Whilestillinhigh school,IstartedreadingAndrzejGrzegorczyk’s OutlineofMathematicalLogic, 1 atextbookwhosesubtitlepromised FundamentalResultsandNotionsExplained withAllDetails.Grzegorczykwasaprominentmathematicianwhomadeimportant contributionstothemathematicaltheoryofcomputabilityandlatermovedto philosophy.ThankstoGoogleBooks,wecannowseealistofthewordsandphrases mostfrequentlyusedinthebook:axiomschema,computablefunctions,concept, emptydomain,existentialquantifier,false,finitenumber,freevariable,andmany
1 AnEnglishtranslationis[17].
more.Ifoundthisallattractive,butIdidnotunderstandanyofit,anditwasnot Grzegorczyk’sfault.Intheintroduction,theauthorwrites:
RecentyearshaveseentheappearanceofmanyEnglish-languagehandbooksoflogicand numerousmonographsontopicaldiscoveriesinthefoundationsofmathematics.These publicationsonthefoundationsofmathematicsasawholeareratherdifficultforthe beginnersorreferthereadertootherhandbooksandvariouspiecemealcontributionsand alsosometimestolargelyconceived‘mathematicalfolklore’ofunpublishedresults.As distinctfromthese,thepresentbookisaseasyaspossiblesystematicexpositionofthe nowclassicalresultsinthefoundationsofmathematics.Hencethebookmaybeuseful especiallyforthosereaderswhowanttohavealltheproofscarriedoutinfullandallthe conceptsexplainedindetail.Inthissensethebookisself-contained.Thereader’sability toguessisnotassumed,andtheauthor’sambitionwastoreducetheuseofsuchwordsas evidentandobviousinproofstoaminimum.Thisiswhythebook,itisbelieved,maybe helpfulinteachingorlearningthefoundationofmathematicsinthosesituationsinwhich thestudentcannotrefertoaparallellectureonthesubject.
NowthatIknowwhatGrzegorczykistalkingabout,Itendtoagree.WhenIwas readingthebookthen,Ifounditalmostincomprehensible.Itisnotbadlywritten,it isjustthatthematerial,despiteitsdeceptivesimplicity,ishard.
Inmyfreshmenyear,AndrzejZarach,wholaterbecameadistinguishedset theorist,wasfinishinghisdoctoraldissertationandhadtheratherunusualideaof conductingaseminarforfreshmenonsettheoryandGödel’saxiomofconstructibility.Thisisanadvancedtopicthatrequiressolidunderstandingofformalmethods thatcannotbeexpectedfrombeginners.Zarachgaveusafewlecturesonaxiomatic settheory,andtheneachofuswasgivenanassignmentforaclasspresentation. MinewastheLöwenheim-Skolemtheorem.TheLöwenheim-Skolemtheoremis oneoftheearlyresultsinmodeltheory.ModeltheoryiswhatIlistnowasmy researchspecialty.Fortheseminar,myjobwastopresenttheproofasgiveninthe thenrecentlypublishedbook ConstructibleSetswithApplications [23],byanother prominentPolishlogicianAndrzejMostowski.Thetheoremisnotdifficulttoprove. Incoursesinmodeltheory,aproofisusuallygivenearly,asitdoesnotrequiremuch preparation.InMostowski’sbook,theprooftakesaboutone-thirdofthepage.Iwas readingitandreadingit,andthenreadingitagain,andIdidnotunderstand.Not onlydidInotunderstandtheideaoftheproof;asfarasIcanrecallnow,Ididnot understandasinglesentenceinit.Eventually,Imemorizedtheproofandreproduced itattheseminarinthewaythatclearlyexposedmyignorance.Itwasahumiliating experience.
Ibroughtupmyearlylearningexperienceshereforjustonereason:Ireally knowwhatitisnottounderstand.Iamfamiliarwithnotunderstanding.Atthe sametime,Iamalsofamiliarwiththoseextremelysatisfyingmomentswhenone doesfinallyunderstand.Itisaveryindividualandprivateprocess.Sometimes momentsofunderstandingcomewhensmallpieceseventuallyadduptothepoint whenonegraspsageneralidea.Sometimes,itworkstheotherwayaround.An understandingofageneralconceptcancomefirst,andthenitshedsbrightlighton anarrayofsmallerrelatedissues.Therearenosimplerecipesforunderstanding. Inmathematics,sometimestheonlygoodadviceisstudy,study,study...,but thisisnotwhatIrecommendforreadingthisbook.Thereareattractiveareasof
mathematicsanditsapplicationsthatcannotbefullyunderstoodwithoutsufficient technicalknowledge.Itishard,forexample,tounderstandmodernphysicswithout asolidgraspofmanyareasofmathematicalanalysis,topology,andalgebra.HereI willtrytodosomethingdifferent.Mygoalistotrytoexplainacertainapproachto thetheoryofmathematicalstructures.Thismaterialisalsotechnical,butitsnature isdifferent.Therearenoprerequisites,otherthansomegenuinecuriosityaboutthe subject.Nopriormathematicalexperienceisnecessary.Somewhatparadoxically,to followthelineofthought,itmaybehelpfultoforgetsomemathematicsonelearns inschool.Everythingwillbebuiltupfromscratch,butthisisnottosaythatthe subjectiseasy.
Muchofthematerialinthisbookwasdevelopedinconversationswithmywife, Wanda,andfriends,whoarenotmathematicians,butwerekindandcuriousenough tolistentomyexplanationsofwhatIdoforaliving.Ihopeitwillshedsomelight onsomeareasofmodernmathematics,butexplainingmathematicsisnottheonly goal.Iwanttopresentamethodologicalframeworkthatpotentiallycouldbeapplied outsidemathematics,theclosestareasIcanthinkofbeingarchitectureandvisual arts.Afterall,everythingisorhasastructure.
IamverygratefultoBethCaspar,AndrewMcInerney,PhilipOrding,Robert Tragesser,TonyWeaver,JimSchmerl,andJanZwickywhohavereadpreliminary versionsofthisbookandhaveprovidedinvaluableadviceandeditorialhelp.
AbouttheContent
Thecentraltopicofthisbookisfirst-orderlogic,thelogicalformalismthathas broughtmuchclarityintothestudyofclassicalmathematicalnumbersystemsand isessentialinthemodernaxiomaticapproachtomathematics.Therearemany booksthatconcentrateonthematerialleadingtoGödel’sfamousincompleteness theorems,andonresultsaboutdecidabilityandundecidabilityofformalsystems. Theapproachinthisbookisdifferent.Wewillseehowfirst-orderlogicserves asalanguageinwhichsalientfeaturesofclassicalmathematicalstructurescanbe describedandhowstructurescanbecategorizedwithrespecttotheircomplexity,or lackthereof,thatcanbemeasuredbythecomplexityoftheirfirst-orderdescriptions.
Allkindsofgeometric,combinatorial,andalgebraicobjectsarecalledstructures, butforustheword“structure”willhaveastrictmeaningdeterminedbyaformal definition.Part I ofthebookpresentsaframeworkinwhichsuchformaldefinitions canbegiven.Theexpositioninthispartiswrittenforthereaderforwhomthis materialisentirelynew.Allnecessarybackgroundisprovided,sometimesina repetitivefashion.
Theroleofexercisesistogivethereaderachancetorevisitthemainideas presentedineachchapter.Newlylearnedconceptsbecomemeaningfulonlyafter we“internalize”them.Onlythen,canonequestiontheirsoundness,lookfor alternatives,andthinkofexamplesandsituationswhentheycanbeapplied,and, sometimesmoreimportantly,whentheycannot.Internalizingtakestime,soonehas
tobepatient.Exercisesshouldhelp.Tothereaderwhohasnopriorpreparationin abstractmathematicsormathematicallogic,theexercisesmaylookintimidating, buttheyaredifferent,andmucheasier,thaninamathematicstextbook.Mostof themonlyrequirecheckingappropriatedefinitionsandfacts,andmostofthemhave pointersandhints.Theexercisesthataremarkedbyasterisksareformoreadvanced readers.
Allinstructorswillhavetheirownwayofintroducingthematerialcoveredin Part I.AselectionfromChapters 1 through 6 canbechosenforindividualreading, andexercisescanbeassignedbasedonhowadvancedthestudentsintheclassare. MysuggestionistonotskipChapter 2,wheretheideaof logicalseeing isfirst introduced.Thattermisoftenusedinthesecondpartofthebook.Iwouldalso recommendnottoskipthedevelopmentofaxiomaticsettheory,whichisdiscussed inChapter 6.Itisdonethererigorouslybutinalesstechnicalfashionthanone usuallyseesintextbooksonmathematicallogic.
HereisabriefoverviewofthechaptersinPart I.Allchaptersinbothpartshave moreextensiveintroductions:
•Chapter 1 beginswithadetaileddiscussionofaformalizationofthestatement “thereareinfinitelymanyprimenumbers,”followedbyanintroductionofthe fullsyntaxoffirst-orderlogicandAlfredTarski’sdefinitionoftruth.
•Chapter 2 introducesthemodel-theoreticconceptofsymmetry(automorphism) usingsimplefinitegraphsasexamples.Theideaof“logicalseeing”isdiscussed.
•ShortChapter 3 isdevotedtotheelusiveconceptofnaturalnumber.
•InChapter 4,buildinguponthestructureofthenaturalnumbers,adetailedformal reconstructionofthearithmeticstructuresoftheintegers(wholenumbers)and therationalnumbers(fractions)intermsoffirst-orderlogicisgiven.Thischapter isimportantforfurtherdevelopments.
•Chapter 5 providesmotivationforgroundingtherestofthediscussionin axiomaticsettheory.Itaddressesimportantquestions:Whatisarealnumber, andhowcanacontinuousreallinebemadeofpoints?
•Chapter 6 isashortintroductiontotheaxiomsofZermelo-Fraenkelsettheory.
Part II ismoreadvanced.Itsaimistogiveagentleintroductiontomodeltheory andtoexplainsomeclassicalandsomerecentresultsontheclassificationoffirstorderstructures.Afewdetailedproofsareincluded.Undoubtedly,thispartwillbe morechallengingforthereaderwhohasnopriorknowledgeofmathematicallogic; nevertheless,itiswrittenwithsuchareaderinmind.
•Chapter 7 formallyintroducesorderedpairs,Cartesianproducts,relations,and first-orderdefinability.Itconcludeswithanexampleofavarietyofstructureson aone-elementdomainandanimportantstructurewithatwo-elementdomain.
•Chapter 8 isdevotedtoadetaileddiscussionofdefinableelementsand,in particular,definabilityofnumbersinthefieldofrealnumbers.
•InChapter 9,typesandsymmetriesaredefinedforarbitrarystructures.The conceptsofminimalityandorder-minimalityareillustratedbyexamplesof orderingrelationsonsetsofnumbers.
•Chapter 10 introducestheconceptofgeometryofdefinablesetsmotivatedby theexampleofgeometryofconicsectionsintheorderedfieldofrealnumbers. ThechapterendswithadiscussionofDiophantineequationsandHilbert’s10th problem.
•InChapter 11,itisshownhowthefundamentalcompactnesstheoremisusedto constructelementaryextensionsofstructures.
•Chapter 12 isdevotedtoelementaryextensionsadmittingsymmetries.Aproof ofminimalityoftheorderedsetofnaturalnumbersisgiven.
•InChapter 13,formalargumentsaregiventoshowwhythefieldsofrealand complexnumbersareconsideredtameandwhythefieldofrationalnumbersis wild.
•Chapter 14 includesafurtherdiscussionoffirst-orderdefinabilityandabrief sketchofdefinabilityinhigher-orderlogics.Well-orderingsandtheMandelbrot setareusedasexamples.
•Chapter 15 isanextendedsummaryofPart II.Thereaderwhoisfamiliarwith first-orderlogicmaywanttoreadthischapterfirst.Thechapterisfollowedby suggestionsforfurtherreading.
InAppendix A,thereaderwillfindcompleteproofsofirrationalityofthesquare rootoftwo(Tennenbaum’sproof),Cantor’stheoremonnon-denumerabilityofthe setofallrealnumbers,first-ordercategoricityofstructureswithfinitedomains, existenceofproperelementaryextensionsofstructureswithinfinitedomains,anda “nonstandard”proofoftheInfiniteRamsey’stheoremforpartitionsofsetsofpairs. Appendix B containsabriefdiscussionofHilbert’sprogramforfoundationsof mathematics.
NewYork,USARomanKossak
PartILogic,Sets,andNumbers
1First-OrderLogic
2LogicalSeeing
2.1FiniteGraphs
3WhatIsaNumber?
3.1HowNaturalAretheNaturalNumbers?
3.1.1ArithmeticOperationsandtheDecimalSystem
3.1.2HowManyNumbersAreThere?
3.1.3Zero .........................................................38
3.1.4TheSetofNaturalNumbers
4SeeingtheNumberStructures
4.1WhatIstheStructureoftheNaturalNumbers?
4.1.1SetsandSetNotation
4.1.2LanguageofFormalArithmetic
4.1.3LinearlyOrderedSets
4.1.4TheOrderingoftheNaturalNumbers
4.2TheArithmeticStructureoftheNaturalNumbers
4.3TheArithmeticStructureoftheIntegers
4.3.1NaturalNumbersAreIntegers,Naturally
4.4Fractions!
4.5TheArithmeticStructureoftheRationals
4.5.1EquivalenceRelationsandtheRationals
4.5.2DefiningAdditionandMultiplicationofthe RationalNumbersFormally ................................54
4.5.3DenseOrderingoftheRationals
5Points,Lines,andtheStructureof R
5.1DensityofRationalNumbers
5.2WhatAreRealNumbers,Really?
5.3DedekindCuts
5.3.1DedekindCompleteOrderings
5.3.2Summary ....................................................64
5.4DangerousConsequences ............................................66
5.5InfiniteDecimals .....................................................68 6SetTheory
6.1WhattoAssumeAboutInfiniteSets?
PartIIRelations,Structures,Geometry
7Relations
7.1OrderedPairs
7.3WhatIsaRelationonaSet?
7.4Definability:NewRelationsfromOld
7.5HowManyDifferentStructuresAreThere?
7.5.1AVerySmallComplexStructure
8DefinableElementsandConstants .......................................97
8.2Databases,Oracles,andtheTheoryofaStructure
8.3DefiningRealNumbers ..............................................100
8.4DefinabilityWithandWithoutParameters
9MinimalandOrder-MinimalStructures ................................105
9.1Types,Symmetries,andMinimalStructures
9.2TrivialStructures .....................................................107
9.3TheOrderingoftheNaturalNumbers
9.4TheOrderingoftheIntegers .........................................109
9.5TheAdditiveStructureoftheIntegers
9.6TheOrderingoftheRationalNumbers
10GeometryofDefinableSets ................................................115 10.1BooleanCombinations ...............................................115
10.2HigherDimensions ...................................................117
10.2.1EuclideanSpaces ...........................................119
10.3ShadowsandComplexity ............................................121
10.3.1DiophantineEquationsandHilbert’s10thProblem
10.3.2TheRealsvs.TheRationals ................................127
11WhereDoStructuresComeFrom?
12ElementaryExtensionsandSymmetries
13Tamevs.Wild
14First-OrderProperties
14.2Well-OrderedSetsandSecond-OrderLogic
15SymmetriesandLogicalVisibilityOneMoreTime
First-OrderLogic
Howevertreacherousagroundmathematicallogic,strictly interpreted,maybeforanamateur,philosophyproperisa subject,ononehandsohopelesslyobscure,ontheotherso astonishinglyelementary,thatthereknowledgehardlycounts.If onlyaquestionbesufficientlyfundamental,theargumentsfor anyanswermustbecorrespondinglycrudeandsimple,andall menmaymeettodiscussitonmoreorlessequalterms.
G.H.Hardy
MathematicalProof [10]
Abstract Thisbookisaboutaformalapproachtomathematicalstructures.Formal methodsarebytheirverynatureformal.Whenstudyingmathematicallogic, initiallyoneoftenhastogritonesteethandabsorbcertainpreliminarydefinitionson faith.Conceptsaregivenprecisedefinitions,andtheirmeaningisrevealedlaterafter onehasachancetoseetheirutility.Wewilltrytofollowadifferentroute.Beforeall formalitiesareintroduced,inthischapter,wewilltakeadetourtoseeexamplesof mathematicalstatementsandsomeelementsofthelanguagethatisusedtoexpress them.
Keywords Arithmetic·Euclid’stheorem·Formalization·Vocabularyof first-order-logic·Booleanconnectives·Quantifiers·Truthvalues·Trivial structures
1.1WhatWeTalkAboutWhenWeTalkAboutNumbers
Thenaturalnumbersare0,1,2,3,....1 Anaturalnumberis prime ifitislarger than1andisnotequaltoaproductoftwosmallernaturalnumbers.Forexample, 11and13areprime,but15isnot,because15 = 3 · 5.Proposition20inBookIXof Euclid’s Elements states:“Primenumbersaremorethananyassignedmultitudeof
1 Accordingtosomeconventions,zeroisnotanaturalnumber.Forreasonsthatwillbeexplained later,wewillcountzeroamongthenaturalnumbers.
©SpringerInternationalPublishingAG,partofSpringerNature2018 R.Kossak, MathematicalLogic,SpringerGraduateTextsinPhilosophy3, https://doi.org/10.1007/978-3-319-97298-5_1
primenumbers.”Inotherwords,thereareinfinitelymanyprimenumbers.Thisisthe celebratedEuclid’stheorem.Whatisthistheoremabout?Inthebroadestsense,itis astatementabouttheworldinwhichsomeobjectsareidentifiedasnaturalnumbers, aboutaparticularpropertyofthosenumbers—primeness,andaboutinexhaustibility ofthenumberswiththatproperty.Weunderstandwhatthetheoremsays,because weunderstanditscontext.Weknowwhatnaturalnumbersare,andwhatitmeans thatthereareinfinitelymanyofthem.However,noneofitisentirelyobvious,and wewilltakeacloserlookatbothissueslater.Concerningtheinfinitudeofprimes, itoccurredtomeoncewhenIwasabouttoshowtheproofofEuclid’stheoremin myclass,toaskstudentswhattheythoughtaboutasimplertheorem:“Thereare infinitelymanynaturalnumbers.”Itwasnotafairquestion,asitimmediatelytakes usawayfromthesolidgroundofmathematicsintothemurkywatersofphilosophy. Thestudentswerebemused,andIwasnotsurprised.
WewillformalizeEuclid’stheoreminaparticularway,andtodothiswewill havetosignificantlynarrowdownitscontext.Inaradicalapproach,thecontextwill bereducedtoabareminimum.Wewillbetalkingaboutcertaindomainsofobjects, andinthecaseofEuclid’stheoremthedomainisthesetofallnaturalnumbers.Once thedomainofdiscourseisspecified,weneedtodecidewhatfeaturesofitselements wewanttoconsider.Inschoolwefirstlearnhowtoaddandhowtomultiply naturalnumbers;andwewillfollowthatpath.WewillexpressEuclid’stheorem asastatementaboutadditionandmultiplicationinthedomainofnaturalnumbers.
Wewilltalkaboutadditionandmultiplicationusingexpressions,called formulas, inaveryrestrictedvocabulary.Wewilluse variables,twooperationsymbols: + and ·,andthesymbol = forequality.Thevariableswillbelowercaseletters x , y , z,....Forexample, x + y = z isaformulaexpressingthattheresultofaddinga number x toanumber y issomenumber z.Thisexpressionbyitselfcarriesno truth value.Itcanbeneithertruenorfalse,sincewedonotassignanyspecificvalues tothevariables.Laterwewillseewaysinwhichwecanspeakaboutindividual elementsofadomain,butfornowwewillonlyhavetheoptionof quantifying over theelementsofthedomain,andthatmeansstatingthateithersomethingholdsfor allelements,orthatsomethingholdsforsome.Forexample:
(1.1)
Thesentenceaboveexpressesthattheresultdoesnotdependontheorderinwhich thenumbersareadded.Itisanexampleofa universal statement;itdeclaresthat somethingholdsforallelementsinthedomain.
Andhereisanexampleofan existential statement,itdeclaresthatobjectswitha certainpropertyexistinthedomain:
x suchthat x + x = x.
(1.2)
Thisstatementisalsotrue.Thereisanelementinthedomainofnaturalnumbers thathastherequiredproperty.Inthiscasethereisonlyonesuchelement,zero.But ingeneral,therecanbemoreelementsthatwitnesstruthofanexistentialstatement.
Forall x andall y,x + y = y + x.
Thereisan
Forexample,
Thereisan x suchthat x x = x
isatrueexistentialstatementaboutthenaturalnumbers,andtherearetwowitnesses toitsveracity,zero,andone.
Interestingstatementsaboutnumbersofteninvolvecomparisonsoftheirsizes.To expresssuchstatements,wecanenlargeourvocabularybyaddingarelationsymbol, forexample <,andinterpretexpressionsoftheform x<y as“somenumber x is lessthansomenumber y .”Hereisanexampleofatruestatementaboutnatural numbersinthisextendedlanguage.
Forall x,y, and z, if x<y, then x + z<y + z. (1.3)
Noticethegrammaticalform“if...then....”
Thenextexampleisaboutmultiplication.Itisanexpressionwithouta truthvalue.
1 <x andforall y and z, if x = y z, then x = y or x = z. (1.4)
Instatements(1.1),(1.2),and(1.3),allvariableswere quantified byaprefix,either “forall”or“thereexists.”In(1.4)thevariable x isnotquantified,itisleft free;it doesnotassumeanyspecificvalue.
Becauseofthepresenceofafreevariable,(1.4)doesnothaveatruthvalue, neverthelessitservesapurpose.It defines thepropertyofbeingaprimenumberin termsofmultiplicationandtherelation <.Letmeexplainhowitworks.
Thinkofaprimenumber,say7,asavalueof x .IfItellyouthat7 = y z,for somenaturalnumbers y and z,withouttellingyouwhatthesenumbersare,then youknowthatoneofthemmustbe7andtheotheris1,becauseonecannotbreak downsevenintoaproductofsmallernumbers.Itistrue“forall y and z,”because forallbutacoupleofthemitisnottruethat7 = y · x ,andinsuchcasesitdoesnot matterwhattherestoftheformulasays.Weonlyconsiderthe“then”partifindeed 7 = y · z.Ifthevalueof x isnotprime,say6,then6 = 2 · 3,sowhenyouthinkof y as2and z as3,itistruethat6 = y · z,butneither y nor z isequalto6,hencethe propertydescribedin(1.4)doesnothold“forall y and z.”
Ifyouarefamiliarwithformallogic,Iamexplainingtoomuch,butifyouarenot, itisworthwhiletomakesurethatyouseehowtheformula(1.4)definesprimeness. Chosesomeothercandidatesfor x andseehowitworks.Also,noticethreenew additionstothevocabulary:thesymbol1forthenumberone;andtwoconnectives “and”and“or.”
Withtheaidof(1.4)wecannowwritethefullstatementofEuclid’stheorem: Forall w ,thereisan x suchthat w<x ,andforall y and z,if x = y · z,then x = y or x = z.
Whatisthedifferencebetweenthestatementaboveandtheoriginal“Thereare infinitelymanyprimenumbers.”?Firstofall,thenewformulationincludesthe
definitionofprimenessinthestatement.Secondly,whatismoreimportant,the directreferencetoinfinityiseliminated.Instead,wejustsaythatforeverynumber w thereisaprimenumbergreaterthanitwithsuchandsuchproperties,soitfollows thatsincethereareinfinitelymanynaturalnumbers,theremustbeinfinitelymany primenumbersaswell.Themostimportanthoweveristhatwemanagedtoexpress animportantfactaboutnumberswithmodestmeans,justvariables,thesymbols · and <,theprefixes“forevery”and“thereis,”andtheconnectives:“and,”“or”,and “if...then....”
Wehavemadethefirststeptowardsformalizingmathematics,andwedid thisinformally.Thepointwastowriteastatementrepresentingameaningful mathematicalfactinalanguagethatisasunambiguousaspossible.Wesucceeded, byreducingthevocabularytoafewbasicelements.Thiswillguideusinoursecond step,inwhichwewillformallydefineacertainformallanguageanditsgrammar. Wewillcarefullyspecifythewayinwhichexpressionsinthislanguagecanbe formed.Someofthoseexpressionswillbestatementsthatcanbeassignedtruth values—trueorfalse—wheninterpretedinparticularstructures.Theevaluationof thosetruthvalueswillalsobepreciselydefined.Someotherexpressions,those thatcontainfreevariables,willserveasdefinitionsofpropertiesofelementsin structures,andwillplayanimportantrole.Allthoseproperlyformedexpressions willbereferredtoas formulas.Mydictionaryexplainsthataformulais“a mathematicalrelationshiporruleexpressedinsymbols.”Themeaninginthisbook isdifferent.Wewilltalkaboutrelationships,andwewillusesymbols,butformulas willalwaysrepresentstatements.Forexample,theexpression b · b 4a · c isa computationalrulewritteninsymbols,butitisnotaformulainoursense,sinceit isnotastatementaboutthenumbers a , b ,and c .Incontrast, d = b · b 4a · c isa formula.Itstatesthatifwemultiply b byitselfandsubtractfromittheproductof fourtimes a times c ,theresultis d .
1.1.1HowtoChooseaVocabulary?
Intheprevioussection,weformulatedanimportantfactaboutnumbers—Euclid’s theorem—usingsymbolsformultiplicationaandtheordering(<).Thisisjustone example,buthowdoesitworkingeneral?Whatpropertiesofnumbersdowewant totalkabout?Whatbasicoperationsorrelationscanwechoose?Theanswersare verymuchdrivenbyapplicationsandparticularneedsandtrendsinmathematics.In thecaseofnumbertheory,thedisciplinethatdealswithfundamentalpropertiesof naturalnumbers,itturnsoutthatalmostanyimportantresultcanbeformulatedina formallanguageinwhichonerefersonlytoadditionandmultiplication.2 Number
2 Inourexamplewealsousedtheorderingrelation <,butinthedomainofthenaturalnumbers, therelation x<y canbedefinedintermsofaddition,sinceforallnaturalnumbers x and y , x is lessthan y ifandonlyifthereisanaturalnumber z suchthat z isnot0and x + z = y
theorymaybethemostdifficultandmysteriousbranchofmathematics.Proofs ofmanycentralresultsareimmenselycomplex,andtheyoftenusemathematical machinerythatreacheswellbeyondthenaturalnumbers.Still,abitsurprisingly, aformalsystemwithafewsymbolsinitsvocabularysufficestoexpressalmost alltheoremsofnumbertheory.Itissimilarinotherbranchesofmathematics.The mathematicalstructures,andthefactsaboutthemarecomplex,butthevocabulary andthegrammaroftheformalsystemthatwewilldiscussinthisbookaremuch simpler.
The realnumbers willbedefinedpreciselylater.Forthemoment,youcan thinkofthemasallnumbersrepresentinggeometricdistancesandtheirnegative opposites.Thefollowingstatementiswritteninarigorous,butinformallanguage ofmathematics.Itinvolvestheconceptofone-to-onecorrespondence.Aone-to-one correspondencebetweentwosets A and B isamatchingthattoeveryelementof A assignsexactlyonelementof B insuchawaythateveryelementof B hasamatch.
Let A beaninfinitesetofrealnumbers.Theneitherthereisaone-to-onecorrespondence between A andthesetofallnaturalnumbers,orthereisaone-to-onecorrespondence A andthesetofallrealnumbers.
Thisisavariantofwhatisknownasthe ContinuumHypothesis.TheContinuum Hypothesiscanalsobestatedintermsofsizesofinfinitesets.Inthe1870s,Georg Cantorfoundawaytomeasuresizesofinfinitesetsbyassigningtothemcertain infiniteobjects,whichhecalled cardinalnumbers.Thesmallestinfinitecardinal numberis ℵ0 anditisthesizeofthesetofallnaturalnumbers.ItwasCantor’s greatdiscoverythatthesizeofthesetofallrealnumbers,denotedby c,islarger that ℵ0 .AnotherwaytostatetheContinuumHypothesisis:if A isaninfiniteset ofrealnumbers,thenthecardinalityof A iseither ℵ0 or c.Thehypothesiswas proposedbyGeorgCantorinthe1870s,andDavidHilbertputitprominentlyat thetopofhislistofopenproblemsinmathematicspresentedtotheInternational CongressofMathematiciansinParisin1900.TheContinuumHypothesisisabout numbers,butitisnotaboutarithmetic.Itisaboutinfinitesets,andaboutoneto-onecorrespondencesbetweenthem.Whatarethoseobjects,andhowcanwe knowanythingaboutthem?Whatisanappropriatelanguageinwhichfactsabout infiniteobjectscanbeexpressed?Whatprinciplescanbeusedinproofs?Precisely suchquestionsledDavidHilberttotheideaofformalizingandaxiomatizing mathematics.ThereisashorthistoricalnoteaboutHilbert’sprogramforfoundations ofmathematicsinAppendix B.
TheContinuumHypothesisisastatementaboutsetsorrealnumbersandtheir correspondences.Toexpressitformallyoneneedstoconsideralargedomain inwhichallrealnumbers,theirsets,andmatchingsbetweenthemareelements. Remarkably,itturnedoutthatthevocabularyofaformalsysteminwhichonecan talkaboutallthosedifferentelements,andmuchmore,canbereducedtological symbolsofthekindweusedforthedomainofthenaturalnumbers,andjustone symbolforthesetmembershiprelation ∈.Allthatwillbediscussedindetailin Chap. 6.
Sofornowwejusthavetwoexamplesofvocabularies,onewithsymbols+and forarithmetic,andtheotherwithjustonesymbol ∈ forsettheory.Wewillsee moreexampleslater,andourfocuswillbeonthenumberstructures.Ingeneral,for everymathematicalstructure,andforeverycollectionofmathematicalstructuresof aparticularkind,thereisachoiceofsymbolsthatissometimesnaturalandobvious, andsometimesarrivedatwithagreateffort.
Adigression:Onceformalized,mathematicalproofsbecomestringsofsymbols thataremanipulatedaccordingtowell-definedsyntacticrules.Inthisform,they themselvesbecomesubjectsofmathematicalinquiry.Onecanaskwhethersuch andsuchformalstatementcanbederivedformallyfromagivensetofpremises.The wholedisciplineknownasprooftheorydealswithsuchquestionswithremarkable successes.In1940,KurtGödelprovedthattheContinuumHypothesiscannotbe disprovedonthebasisofour,suitablyformalizedandcommonlyacceptedaxioms ofsettheory,andin1963,PaulCohenprovedthatitcannotbeprovedfrom thoseaxiomseither.Thisisallremarkable,andwasaresultofagreateffortin foundationalstudies.
1.2SymbolicLogic
Mathematicallogicissometimescalled symboliclogic,sinceinlogicalformulas ordinaryexpressionsarereplacedwithformalsymbols.Wewillintroducethose symbolsinthenextsection.HenriPoincaré,thegreatFrenchmathematician,who wasstronglyopposedtoformalmethodsinmathematics,wrotein1908:“Itis difficulttoadmitthattheword if acquires,whenwritten ⊃,avirtueitdidnot possesswhenwritten if.”3 Poincaréwasright.Nothingisgainedconceptuallyby justreplacingwordswithsymbols,buttheintroductionofsymbolsisjustafirststep. Themoreimportantfeatureisaprecisedefinitionofthegrammaroftheformalized language.Wearegoingtopaycloseattentiontotheshapeoflogicalformulas,and thelogicalsymbolswillhelp.Itisverymuchasinalgebra: x + y ,simplymeans x plus y ,and x · y means x times y ,butifyouthoughtthatourformalizedexpression forEuclid’stheoremwascomplicated,thinkhowcomplicateditwouldhavebeenif wedidnotuse + and · .
Therearemanyadvantagesofthesymbolicnotation.Itispreciseandconcise. Onenotonlysavesspacebyusingsymbols;sometimessymbolicnotationallows onetoexpresscomplexstatementsthatwouldbehardtounderstandinthenatural language.Themostcommonsymbolicsystemofmathematicsiscalledfirst-order logic.Itwillbedefinedinthissectionanditwillbeextensivelyusedintherestof thebook.
Themathematicalnotationwithallitssymbolsandabbreviationsisthelanguage ofmodernmathematicsthathastobelearnedasanyotherlanguage,andlearninga
3 InPoincaré’stime, ⊃ wasusedtodenoteimplication.
languagetakestime.Iwilltrytolimitnotationtoaminimum,butyouhavetobear withme.Thelanguagethatyouwilllearnservescommunicationamongmathematicianswell,butthefactthatsomuchofmathematicsrequiresitcreatesproblemsin writingaboutmathematicsfornon-mathematicians.Moreover,mathematicianshave createdtheirlanguageinaratherchaotichistoricalprocesswithoutparticularregard totheneedsofbeginners.AlexandreBorovikwrote[6]:“Whyarewesosurethatthe alphabet ofmathematics,asweteachit—allthatcorpusofterminology,notation, symbolism—isnatural?...Wehavenothingtocompareourmathematicallanguage with.Howdoweknowthatitisoptimal?”
Intheprevioussectionwesawhowmathematicalstatementsarewrittenusing logicalconnectivesandquantifiers.Nowwewillbewritingthemusing logical symbols.Thoseare ∧, ∨,and ¬,representing and, or,and not,respectively.The connectiveswillbeusedtogrouptogetherstatementsaboutrelations,andthose statementswillbecomposedofvariablesand relationsymbols.Theprefixesofthe form“forall x ...”and“thereisan x suchthat...”arethe quantifiers,thefirstis called universal anditwillbewritten ∀x... ;thesecondiscalled existential,andit willbewritten ∃x... (thinkof ∀ll,and ∃xist).
Therearetwowaysofintroducingrelationsymbols.Onecouldfirstdefinean infinitecollectionofsymbols,andthenforeachstructurechooseonlyparticular symbolsspecifictothestructure.Thiswouldgiveusa“onelanguage—allstructures”model.Alternatively,onecanfirstmakeachoiceofsymbolsforaparticular structure,oraclassofstructures,anduseonlythose.Thelatter“onekindof structures—onelanguage”modeldoesnotneedsomeofthesmalltechnicalities thattheformerrequires,sowewilladoptit.Sinceatfirstwewanttodiscuss numberstructures,wechoosethefollowingthreerelationsymbols: A foraddition, M formultiplication,and L forthe“lessthan”relation.Bythestandardconvention, regardlessofthechoiceofotherrelationsymbols,theequalityrelationsymbol=is alsoalwaysincludedinthevocabulary.
Foranimportanttechnicalreason,wewillneedinfinitelymanyvariables.We willindexthembynaturalnumbers: x0 , x1 , x2 , x3 ,andsoon.Eachformulawill onlyusefinitelymanyvariables,butthereisnolimitonthenumberofvariables thatcanbeused.Thisisanimportantfeatureoffirst-orderlogicsowehavetokeep allthoseinfinitelymanyvariablesinmind,andfromtimetotimetherewillbea needtorefertoallofthem.Tosimplifynotation,wewilloftendropthesubscripts, andwewilluseotherlettersaswell.
Weareusedtothinkingofadditionandmultiplicationasfunctions,oroperations onnumbers.NowIwillaskyoutothinkofthemasrelations.
Infullgenerality,thelanguageoffirst-orderlogicincludesrelationsymbols andfunctionsymbols,buttoavoidsometechnicalitieswewillnotusefunction symbols.Theword“technicalities”isoneofthosetreacherousexpressionsthat oftenhidessomeimportantissuesthattheauthoristryingtosweepundertherug,so letmeofferanexplanation.Inmathematicsonestudiesbothfunctionsandrelations. Weusemathematicalfunctionstomodelprocessesandoperations.Metaphorically speaking,afunction“takes”anobjectasinputand“produces”anotherobjectasan output.Additionisatwoargumentfunction,theinputisapairofnumbers,say1
and3,andtheoutputistheirsum4.Functionsareusefulwhenchangeisinvolved; when,forexample,somequantitychangesasafunctionoftime.Relationsaremore likedatabases—theyrecordrelationships.Bothconceptshavetheirformalizations, andinmathematicalpracticethedistinctionbetweenthemisnotsharp.Arelation canevolveintime;afunctioncanbeconsideredasarelation,relatinginputsto outputs.Thetechnicalitieshintedatabovearetherulesthatmustbeobeyedwhen wecomposefunctions,i.e.whenweapplyonefunctionafteranotherinaspecified order.Thoserulesarenotcomplicated,butatthislevelofexpositiontheywould requireamorecarefultreatment.Wewillnotdothat,and,sinceeveryfunctioncan berepresentedasarelation,thepricethatwillbepaidwillnotbegreat.
Letusseehowadditionandmultiplicationcanberepresentedasrelations.As wenotedearlier,additionofnaturalnumbersisafunction.Toeachpairofnatural numbers,thefunctionassignsavaluethatistheirsum.Theinputsarepairsof numbers m, n,andtheoutputsarethesums m + n.Letusnamethisfunction f . Usingfunctionnotation,wecanwrite2 + 2 = 4as f(2, 2) = 4,and100 + 0 = 100 as f(100, 0) = 100.Wearenotconcernedherewithanyactualprocessofadding numbers,wethinkof f asadevicethatinstantlyprovidesacorrectanswerineach case.Butadditionalsodeterminesarelationshipbetweenorderedtriplesofnumbers asfollows.Wecansaythatthenumbers k , m,and n arerelatedifandonlyif4 f(k,m) = n,or,inotherwords, k + m = n.Inthissense,thenumbers2,3,and5 arerelated,andsoare100,0,and100,but0,0,and1arenot.Noticethatwemust becarefulabouttheorderinwhichwelistthenumbers.Forexample,2,2,and4 arerelated,but2,4,and2arenot.Additionasarelationcarriesexactlythesame informationasthefunction f does.
Togofurther,wemustnowdefinetherulesthatgenerateallformulasoffirstorderlogic.Aformulaisaformalexpressionthatcanbegenerated(constructed)in aprocessthatstartswithbasicformulas,accordingtopreciserules.Thedefinition itselfisanexampleofaformalmathematicaldefinition.Itisan inductivedefinition. Inaninductivedefinition,onefirstdefinesabasiccollectionofobjects,andthen describestherulesbywhichnewobjectscanbeconstructedfromthoseobjectswe alreadyhaveconstructed.Thedefinitionalsodeclaresthatonlyobjectsobtainedthis wayqualify.
Hereisanexampleofasimpleinductivedefinition.Everyoneknowswhata finitesequenceof0’sand1’sis.Itisenoughtoseeanexampleortwo.Hereis one:100011101.Hereisanother:1111111.Werecognizesuchsequenceswhenwe seethem,butnoticethatthisrestsonanintuitiveunderstandingoftheconceptof finitesequence.Theinductivedefinitionwillnotmakeanyexplicitreferencesto finiteness,insteadthefinitecharacteroftheconceptwillbebuiltintothedefinition. Thisaspectisnotjustaphilosophicalnicety,ithaspracticalconsequences.We useinductivedefinitioninaspecificwayproveresultsaboutthedefinedconcepts. Theadvantageofinductivedefinitionsisthattheygiveaninsightintotheinternal
4 Thephrase“ifandonlyif”iscommonlyusedinmathematicstoconnecttwoequivalent statements.
1.2SymbolicLogic11 structureoftheobjectstheydefine.Theyshowushowtheyaremadeinastep-bystepprocess.
Letusnowdefinesequencesof0’sand1’sinductively.Webeginbysayingthat 0and1arefinitesequences.5 Theseareourbasicobjects.Thencomestheinductive rule:if s isafinitesequencethensoare s 0and s 1.Finally,wedeclarethatthefinite sequencesof0’sand1’sareonlythoseobjectsthatareobtainedfromthebasic sequences0and1byapplyingtheinductiverule(overandoveragain).
Nowwegobacktoformulas.Recallthatwechose A, M ,and L forrelation symbols. A and M are ternary—theybindthreevariables,and L is binary—itbinds twovariables.“Binding”isatechnicalterm,andyoushouldnotputmoremeaning toitbeyondwhatiswritteninthefollowingdefinitionofbasicformulas,which wewillcall atomic.Atomicformulasareallexpressionsoftheform A(xi ,xj ,xk ), M(xi ,xj ,xk ), L(xi ,xj ),and xi = xj ,where i , j ,and k arearbitrarynatural numbers.Forexample, A(x0 ,x1 ,x2 ), M(x5 ,x3 ,x1 ), M(x1 ,x0 ,x0 ), L(x1 ,x2 ) are atomicformulas.Inotherwords,anatomicformulaisarelationsymbolfollowedby alistofthreearbitraryvariables,inthecaseof A and M ,ortwoarbitraryvariables, inthecaseof L.Noticethatsincethereareinfinitelymanyvariables,therearealso infinitelymanyatomicformulas.
Whenwediscussparticularexamplesofformulas,forgreaterreadabilitywewill usuallydropthesubscriptsanduseotherlettersforvariables,soexpressionssuchas A(x,y,z) or L(y,x) (notatypo)willalsobeconsideredatomicformulas,although inthestrictsense,accordingtothedefinitiontheyarenot.
Alongwithformalexpressions,definedaccordingtostrictrules,wewillalso useothercommonmathematicalexpressionsandthosewillalsooftenusesymbols. Thetwodifferentkindsofsymbolsshouldnotbeconfused.Wewillneednames formanydifferentobjects,includingformulasandsentencesoffirst-orderlogic. Wewillseethisinthedefinitionbelow.TheGreekcharacters ϕ (phi), ψ (psi),and other,areusedasnamesforformulasofthelanguagewedefine.Theyarenotapart oftheformalism.
Definition1.1
1.Everyatomicformulaisaformula.
2.If ϕ and ψ areformulas,thensoare (ϕ) ∧ (ψ), (ϕ) ∨ (ψ),and ¬(ϕ).
3.If ϕ isaformula,then,foreach n, ∃xn (ϕ) and ∀xn (ϕ),areformulas.
4.Therearenootherformulas.
Noticetheuseofparentheses.Theyplayanimportantrole.Theyguaranteethat everyfirst-orderformulacanbereadinonlyoneway.6
5 Wecouldactuallystartonelevellower.Wecouldsaythattheemptysequence,withnosymbols atall,isafinitesequenceof0’sand1’s.
6 Thisuniquereadabilityoffirst-orderformulasisnotanobviousfactandrequiresaproof,which isnotdifficult,butwewillnotpresentithere.
LetusseehowDefinition 1.1 works.Perclause(1.1),theexpressions x = y and A(x,z,y) (notyposhere)areformulas,becausetheyareatomicformulas.7 Per rule(1.2), ¬(x = y) and ∃z(A(x,y,z)) areformulas.Byapplying(1.2),weseethat (¬(x = y)) ∧ (∃z(A(x,y,z)))
isaformulaaswell.Letuscallthisformula ϕ(x,y) 8 Theonlydisplayedvariables are x and y ,becausetheyarefreein ϕ(x,y).Thethirdvariable z isboundbythe existentialquantifier ∃z.Thinkof A astheadditionrelationofthenaturalnumbers. Forwhatvaluesof x and y does ϕ(x,y) becomeatruestatements.Firstofallthey mustbedifferent,asdeclaredbythefirstcomponentof ϕ(x,y),butalso x mustbe lessthan y ,becauseonlythenthereisanaturalnumber z suchthat x + z = y Animportantcaveat.Accordingtotherules,wearefreetochooseanyvariables weliketoformatomicformulas,soforexample A(x,z,z),and L(x,x) arewellformedformulas.Ifthesamefreevariableisusedindifferentplacesinaformula, whenweinterprettheformulainastructure,thatvariablewillalwaysbeevaluated bythesameelement,butthisdoesnotmeanthatifthevariablesaredifferentthat theyrepresentdifferentobjects.Wearefreetoevaluateanyfreevariablebyany object,inparticularwecanusethesameobjectfordifferentvariables.
Letusrecapitulate.Thelistofsymbolsofthefirst-orderlogicis: ∧, ∨, ¬ , ∃, ∀, and =,andthenforeachparticularstructure,inaddition,itincludesacollectionof relationsymbols.Inourcase,wechose A, M ,and L.Eachrelationsymbolhasa prescribed arity whichisgiveninthedefinitionoftheatomicformulas.Thesymbols A and M areofaritythree,and L isofaritytwo.Thismeansthat,forexample, A(x0 ,x1 ,x2 ) isawell-formedatomicformula,but A(x0 ,x1 ) and L(x0 ,x1 ,x2 ) are not,becausethenumberofvariablesdoesnotmatchthearityofthesymbol.
Theattentivereaderwillask:Butwhataboutallthoseparenthesesandcommas? Yes,theyarealsoformalsymbolsofourlanguage,andtheiruseisentirely determinedbyDefinition 1.1.Therulesforcommasarehiddeninclause(1.1). Onecoulddowithoutthem.Forexample Ax0 x1 x2 alsorepresentsauniquely recognizablestringofsymbols,andthisisallwewant,butforgreaterreadability, andtoavoidadditionalconventionsthatwewouldhavetointroduceintheabsence ofparentheses,weuseparenthesesandcommas.Oftenwewillalsouse“[,”and“],” andsometimes“{”and“}.”Theyallhavethesamestatusas“(,”and“).”
7 Thisisanexampleofmathematicalpedantry.Ofcourse,youwouldsay,theyareformulas.They areevenatomicformulas!Butwhenwedefinedatomicformulas,wedefinedaspecialkindof expression,andcalledexpressionsofthiskind“atomicformulas.”Whenwedidthat,theformal conceptofformulahadnotbeendefinedyet.Toknowwhataformulaoffirst-orderlogicisonehas towaitforaformaldefinitionofthekindwegavehere.Toavoidthiswholediscussionwecould havecalledatomicformulasatoms.Ifwedidthat,thenclause(1.1)ofthedefinitionabovewould say“Everyatomisaformula,”butsincetheterm“atomicformula”iscommonlyused,wedidnot havethatchoice.
8 Thisisanotherexampleofaninformalabbreviation.
Eachapplicationofrules(1.2)and(1.3)introducesanewlayerofparentheses. Ifwecontinuethisway,formulasquicklybecomeunreadable,butthisformalism isnotdesignedforthehumaneye.Wesacrificeeasyreading,butwegainmuchin return.Onebonusisthatitisnoweasytocheckandcorrectgrammar.Theonly grammaticalrulesarethoseinDefinition 1.1.Inparticular,ineveryformulathe numberofleftparenthesesmustbeequaltothenumberofrightparentheses.Ifitis not,thesequenceofsymbolsisnotproperlyformedanditisnotaformula.
ThemostimportantaspectofDefinition 1.1 isthatitshowshowallformulasare generatedinastep-by-stepprocessinwhichmoreandmorecomplexformulasare generated.Thisisacrucialfeature,thatopensthedoortoinvestigationsofformal languagesbymathematicalmeans.Anotheressentialfeatureisthatthesetofall formulasisgeneratedwithoutanyregardtowhatthoseformulasmayexpress.In fact,mostformulasdonotexpressanythinginterestingatall.Forexample
((x = x) ∨ (x = x) ∧ (x = x))
isaproper,butuninterestingformula,andsois ∃xL(y,z)
Whatisthepointofallowingmeaninglessformulas?Whatweareafterare formulasandsentencesthatexpresssalientpropertiesofstructuresandtheir elements,butwewouldbeatalosstryingtogiveamathematicaldefinitionof ameaningfulformula.Itismucheasiertoacceptthemall,whatevertheymay beexpressing.Thereissomethingprofoundintreatingallformulasthisway. Meaningfulnessisavagueconcept.Asentenceofnointeresttoday,mayturnoutto bemostimportanttomorrow,soitwouldmakenosensetoeliminateanyofthemin advance,butthisisnotthemainpoint.Mostmechanicallyformedformulasarenot onlyuninteresting,theyactuallymakenosenseatall.Stillwewanttokeepthem in,becauseitisthepricetopayfortheclarityofthedefinition.Moreover,thereare alsosomeunexpectedtechnicalapplications.If ϕ isaformula,then,accordingto rule(1.2),soare (ϕ) ∧ (ϕ) and (ϕ) ∧ ((ϕ) ∧ (ϕ)),and (ϕ) ∧ ((ϕ) ∧ ((ϕ) ∧ (ϕ))),andso on.Nothingnewisexpressed,buttherearesomeimportantresultsinmathematical logicthatdependinanessentialwayonexistenceofsuchstatements.
Thedefinitionofthesyntaxoffirst-orderlogiciscompleted.Nowitistime todefinethe semantics,i.e.theprocedurethatgivesmeaningandtruthvaluesto formulaswheninterpretedinastructure.Wealreadydidthatinformally,when wetalkedaboutEuclid’stheoremandinterpretationsofformulasinthenatural numbers.Fulldefinitionofsemanticsforfirst-orderlogicisbasedonAlfredTarski’s famous definitionoftruth from1933[34].Itisformulatedinaset-theoreticsetting thatwewilldiscusslater.Fornow,wewillshowhowitallworksusingexamples. Forafullformaldefinitionconsultanytextbookonmathematicallogic.Agood sourceonlineistheStanfordEncyclopediaofPhilosophy[13].
Wewillinterpretformulasinthedomainofthenaturalnumbers.Tobeginwith, foranythreenumbers m, n,and k ,weneedtoassigntruthvalues(trueorfalse)to allatomicformulas A(x,y,z) and M(x,y,z), L(x,y),and x = y ,when x , y ,and z areinterpretedas m, n,and k respectively.Wedeclare A(x,y,z) tobetrueifand onlyif m + n = k , M(x,y,z) tobetrue,ifandonlyif m · n = k , L(x,y) tobetrue
ifandonlyif m islessthan n,andfinally x = y tobetrueifandonlyif m equals n.Weareexceedinglypedantichere,andforagoodreason.Wejustdescribedthe definitionoftruthfortheatomicformulas.
Whatmakesnotationcomplicatedintheexplanationsaboveisthereference toevaluationofthevariables.Tosimplifymatters,oneistemptedtoassigntruth valuesdirectlytoexpressionssuchas A(m,n,k).Thereisaproblemwiththat.The expression A(x,y,z) isaformula.Itisjustastringofsymbolsofthelanguageof first-orderlogic.Theexpression A(m,n,k) isnotaformula.Theletters m, n,and k ,asusedhereareinformalnamesfornumbers.NoruleinDefinition 1.1 allows insertingnamesofobjectsintoformulas.Intheexpression A(m,n,k) twoworlds aremixed.Therelationsymbol A,theparenthesesandcommas,comefromthe worldofsyntax; m, n,and k arenotsymbolsoftheformallanguage,theyare informalnamesofelementsofthedomainofthestructure.
Whatisthedifferencebetweenthestatement“A(x,y,z) istrue,when x , y and z areinterpretedas m, n,and k ”andthestatement“m + n = k ”?Theformerstates thatacertain truthvalue isassignedtoacertainformulaundercertainconditions. Thelatterisastatementaboutthestateofaffairsinacertainstructure.Whilethe definitionistellingusunderwhatconditionscertainstatementsaretrue,ithas nothingtodowithwhetherwecanactuallycheckifthoseconditionsaresatisfied. Inthecaseofcheckingwhether m plus n equals k ,thinkofnumberssoincredibly largethatthereisnotenoughspacetowritethemdown.Wearenottalkingofany practicalaspectsofcomputationhere.Still,itmakessensetodefinethetruthvalues ofinterpretationsofformulasthisway.Thedefinitionisprecise,anditisexactlythis definitionthatmakesabridgebetweenthesyntaxandtheworldofmathematical objectsinwhichitisinterpreted.
Oncethedefinitionoftruthvaluesforatomicformulasisestablished,truth valuesformorecomplexformulasaredeterminedinawayparalleltotherules forgeneratingformulasinDefinition 1.1.Forexample, ¬(ϕ) istrueifandonlyif ϕ isfalse; (ϕ) ∧ (ψ) istrue,ifandonlyifboth ϕ and ψ aretrue;and ∃x(ϕ) istrueif andonlyifthereisanevaluationofthevariable x underwhich ϕ becomestrue.Here wetakeadvantageoftheinductiveformofDefinition 1.1.Inthesamewayinwhich themorecomplexformulasareinductivelybuiltfromsimplerones,thetruthvalues ofmorecomplexformulasareinductivelydeterminedbythetruthvaluesassigned totheirsimplercomponents,withtheatomiccaseasthebase.Aswasmentioned earlier,thefullformaldefinitionofthisprocessissomewhattechnical,andwewill omitit.
InourdiscussionofEuclid’stheorem,weincluded“if ...then...”amongthe logicconnectives.Conditionalstatementsoftheform“if ϕ then ψ ”abbreviatedby (ϕ) ⇒ (ψ),areessentialinmathematics,butDefinition 1.1 hasnoprovisionfor them.Onecouldaddanotherclausethereexplaininghow (ϕ) ⇒ (ψ) istobe interpreted,butthisisnotnecessary.Inclassicallogic,theformula (ϕ) ⇒ (ψ) is definedasanabbreviationof ¬(ϕ) ∨ (ψ),henceitsinterpretationisalreadycovered bytheDefinition 1.1.Letusseeitonanexamplewealreadydiscussed.
Theformula(1.4)definingprimenumbersintheprevioussectionincludedthe followingconditionalstatement:
Forall y andall z, if x = y z, then x = y or x = z.
Itssymbolicversionis
whichinturnisequivalentto
Convinceyourself(1 4 )istrueonlyif x isinterpretedaseither1oraprimenumber. Anothercommonlogicalconnectiveis“ifandonlyif.”Thesymbolofitis ⇐⇒, and (ϕ) ⇐⇒ (ψ) isdefinedasanabbreviationfor ((ϕ) ⇒ (ψ)) ∧ ((ψ) ⇒ (ϕ)) Foranexample,seeExercise 1.5
Afirst-orderpropertyisapropertythatcanbeexpressedinfirst-orderlogic, whichmeansitcanbedefinedbyaformulaintheformalismwejustdescribed. Thiswholebookisaboutmathematicalstructuresandtheirfirst-orderproperties. Notallpropertiesarefirst-order.Forexample,hereisapropertyofnaturalnumbers thatisnotdefinedinafirst-orderway
Everysetofnaturalnumbershasaleastelement.(1.5)
In(1.5)wequantifyoversetsofnumbers,andthatmakesthisstatement secondorder.Infirst-orderlogicwecanonlyquantifyoverindividualelementsofdomains, butwecannotquantifyoversetsofelements.Quantificationoversetsisallowedin second-orderlogicwithitsspecialsyntaxandsemantics.Thereisathird-orderlogic thatallowsquantificationoversetsofsetsofelements.Therearehigher-orderlogics, eachwithstrongerexpressivepowers.ThereismoreaboutthisinChap. 14
Wewillsticktofirst-orderlogicfortworeasons.Oneisthatevenwithits restrictions,first-orderlogicisastrongenoughformalframeworkforasubstantive analysisofmathematicalstructuresingeneral,butthereisalsoanotherappealing reason.First-orderlogicisbasedonrudimentaryprinciples.Itonlyusessimple connectives“and,”“or,”and“not,”andthequantificationonlyallowsustoask whethersomepropertyholdsforallelementsinadomain(∀),orifthereisan elementinadomainwithagivenproperty(∃).Inotherwords,itisaformalization ofthemostbasicelementsoflogic,andonecouldarguethatitalsocapturessome basicfeaturesofperception.Let’sthinkofcollectionsofelementsandsomeoftheir propertiesthatcanbevisuallyrecognized.IfIseeasetofelementshavingaproperty ϕ(x),Ialsoseeitscomplementconsistingoftheelementsthatdonothavethat property,whichisthesameashavingtheproperty ¬ϕ(x).Ifsomeelementshavea property ϕ(x),andsomehaveanotherproperty ψ(x),thenIcanseethecollection ofelementswithbothproperties,i.e.thesetofdefinedbyproperty ϕ(x) ∧ ψ(x)
SimilarlyIcanseetheelementshavingonepropertyortheother: ϕ(x) ∨ ψ(x).If allelementshaveaproperty ϕ(x),Iseethat ∀xϕ(x).Toseethatitisnotthecase, itisenoughtonoticeoneelementthatdoesnothavetheproperty,soitisenough toseethat ∃x ¬ϕ(x).Thefirst-orderapproachprovidesabasicframeworkforwhat Iwillcall logicalvisibility.Equippedwiththisframework,wewilltrytofindout whatcan,andwhatcannotbeseeninstructuresthroughtheeyesoflogic.
Hereisaroughoutlineofwhatwewilldonext.Todefineastructure,westart withacollectionofindividualobjectssharingcertainfeatures.Ineachstructure,the objectsinthecollectionarerelatedtooneanotherinvariousways.Wewillgive thoserelationsnames,andthenwewilltrytoseewhatpropertiesofthestructure anditsindividualelementsarefirst-order.Ananalysisofthecomplexityofformulas andsentencesoffirst-orderlogicwillallowustoapplygeometricintuitions,andto seegeometricpatternsinthestructure.Inthissense,theformalismwillallowusto gobacktomorenatural,unformalizedwaysofthinkingaboutthestructure,andto “logicallysee”someofitsfeatures,thatotherwisemighthavestayedinvisible.This workswellinmathematicsandwewillexaminesomeexamples.
1.2.1TrivialStructures
Thesimpleststructuresaredomainswithnorelationsonthem.Thinkofadomain withfiveobjects.Ifthoseobjectsarenotrelatedtooneanotherinanyway,this isanexampleofa trivial structure.Whatcanbesaidaboutit?Notmuchmore thanwhatwehavesaidalready,butitisgoodtokeeptrivialstructuresinmindfor furtherdiscussion.Theyareagoodsourceofexamplesandcounterexamples.Due toourconvention,theequalityrelationisalwaysamongtherelationsymbolsfor anystructure.Hence,eventhoughatrivialstructurehasnorelationsofitsown,it stillhastheequalityrelation,anditallowsustoexpressspecificfactsaboutitinthe first-orderway.Considerthesentence:
Itsaysthattherearetwodistinctelementsandanyelementinthestructuremustbe oneofthem.Inotherwords,itexpressesthatthestructurehasexactlytwoelements. Inasimilarway,foranynumber n,onecanwriteafirst-ordersentenceexpressing thatthestructurehasexactly n elements.
Itisaninterestingfact,thatfollowsfromthecompactnesstheoremforfirstorderlogic,thatwhileforeachnumber n,havingexactly n elementsisafirst-order property,havingafinitenumberofelementsisnot.Thecompactnesstheoremand anargumentshowingwhyfinitenessisnotafirst-orderpropertyarepresentedin Chap. 11.
Another random document with no related content on Scribd:
Leibl Miederstudie
Mit Genehmigung der Photographischen Gesellschaft in Berlin-Charlottenburg
Rembrandt
Frau des Tobias mit der Ziege
Abb. 8
Rembrandt Jacobs Segen
Rembrandt Jacobs Segen (Teil)
Abb. 10
Rembrandt David vor Saul
Abb. 11
Rembrandt Tobias und seine Frau
Böcklin Gefilde der Seligen Mit Genehmigung der Photographischen Gesellschaft in Berlin-Charlottenburg
Abb. 13
Triton und Nereïde
Mit Genehmigung der Photographischen Union in München
Böcklin
Abb. 14
Böcklin Die Pest Mit Genehmigung der Photographischen Union in München
Tod auf den Schienen (Vom Tode I, 8)
Verlag von Amsler & Ruthardt, Berlin
Klinger
Abb. 16
Brueghel d Ä Fall der Engel
17
Abb.
Idolino
18
Abb.
Michelangelo
Der Morgen Vom Grabmal Lorenzos
Abb. 19
Grabmal der Hegeso, Athen
Abb. 20
Orpheusrelief
Abb. 21
Michelangelo Erschaffung Evas
22
Abb.
Michelangelo Erschaffung Adams
Abb. 23
Dürer Das große Rasenstück
Grünewald
Isenheimer Altar. Antonius und Paulus in der Wüste. Teilstück
Abb. 25
Dürer Händestudie