Vous êtes sur la page 1sur 42

PHI-189K Topics in Logic: Gdel's Incompleteness Theorems

Ted Shear UC Davis Winter, 2012

Chapter 1: What Gdel's Theorems say


Gdel's two incompleteness theorems @IWQHEIWQIA for rithmeti me in the wke of his ompleteness
theorem for (rst order logi @IWPWAF gonsider the lnguge of

basic arithmetic X

Basic Arithmetic:
N : 0, 1, 2, . . . ypertionsX +,
e hve n intuitive understnding of lot of opertions the might e de(ned in the lnguge of si rithmetiY howeverD it is quite nother thing to e le to formlly evlute lims in si rithmetiF e will presume tht rithmetil lims hve determinte nswers even if proofs for them re not villeF oD it is plusile suggestion tht rithmetil truth n e redued to provility in theory T F

Denition 1.1

theory T is (nite set of sentenes derivle from the xioms of T F

sntuitivelyD we would like to e le to estlish some T tht would pin down N suh tht for ny rithmetil sentene either or would e provleF

Denition 1.2
or T

T is F

(negation) complete if for every in the lnguge of the theoryD either T

en ttempt to xiomtize rithmetil truth mens to redue truth to provility in negtion omplete theory T F

Denition 1.3
pleteF

e theory T is

sound if every tht T

is trueF

Gdel's rst incompleteness theorem (FIT): xo sound rithmetil theory is @negtionA omE
I

qdel gives us reipeD whih given theory T D will llow us to (nd sentene GT @ qdel sentene for T A tht is true i' it is not provle in T F uppose tht T is sound theory for si rithmeti"tht isD it hs only true xioms nd isoltes ll nd only the true sentenes in the lnguge of rithmetiF he qdel sentene for T D GT D whih is sentene in the lnguge of rithmeti of ourse must e true @euse the theory is soundAF futD euse of the wy tht GT is onstrutedD it n e shown tht it is impossile to derive GT in T F ut rudelyD GT is true i' T n9t prove itF he rgument proeeds s followsX IF uppose tht T is soundF PF sf T were to prove GT D then GT would e flseD nd T would hve proved flseE hoodF QF yf ourseD this nnot hppenD so if T is soundD then GT is unprovle in T F RF oD GT is trueF SF his mens tht GT is flse nd tht T hoodF

GT euse it nnot prove flseE

TF hereforeD sine T is soundD ut proves neither GT nor GT D it nnot e ompleteF sn ftD the supposition tht T is sound n e wekened to the supposition tht T is onsistent for the (rst hlf of the proof to workD the seond hlf requires only E onsistenyF his result pplies to ny rithmetil theoryD so the result tht the theory onstituted y {T + GT }D D then S GS F imilrlyD onsider set theoreti interprettion of rithmetiY the result of the qdel9s ps tells us tht set theory nnot e negtionEompleteF sn the fmilir wyD suppose tht we rete wy to represent the nturl numers in set theoryX

, {}, {, {}}, . . .
e ould well xiomtize set theory in wy tht models rithmetiF oD qdel9s ps tells us tht ny suh ttempt to give omplete sound theory of rithmeti will filF

Gdel's second incompleteness theorem (SIT):


unprovility of 0 = 1F vet ConT represent T

he onsisteny of T is equivlent with the

0 = 1F

rlf of the onlusion of ps sys tht if T is onsistent then GT is unprovleX ConT GT F sf T

ConT GT D then sine T

GT D we know tht T

ConT F

ut more quiklyD ny dequte theory of si rithmeti nnot prove its own onsisE tenyF his ples interesting restritions on the results tht theories n prove out riher theories s wellF sn ddition to not eing le to prove its own onsistenyD theE ory n never prove the onsisteny of theories tht re more omplex thn themselvesY forD if theory ould prove the onsisteny of theory more omplex thn itselfD then we might think to ootstrp the onsisteny proof of the weker theory y then proving it in the stronger oneF yf ourseD it9s ovious why this won9t work outF

Chapter 2: Functions and enumerations


his hpter introdues numer of notions surrounding fetures of funtions tht will e developed further nd relied upon throughout the ookF

Denition 2.1 Denition 2.2

e funtion is total i' it mps every element of their domin to some unique orreE sponding vlue in its odominF e funtion is

partial i' it is not totlF

gonsider the totl funtion f : N ND whih hs the domin nd the rnge F

{x|f (x) exists} = {f (x) : x }

Denition 2.3 Denition 2.4 Denition 2.5 Denition 2.6

f is f is f is

surjective (onto) if x x f (x) = y injective (one-to-one) if f (x) = f (y) implies tht x = y bijective if it is surjetive nd injetiveF

qiven property P D it9s characteristic function cp is the funtion suh tht cp (x) = 0 i' P (x) cp (x) = 1 i' not P (x) xyiX xotie tht we re reking the ordinry onvention of ssigning 1 s the vlue of the hrteristi funtion of property to n element tht hs tht propE ertyF por our purposesD we will ssign 0 to the vlue of the funtion on n elementD i' tht element hs tht property nd 1 s the vlue i' the element does not hve the propertyF

Denition 2.7 Executing an algorithm

involves n entirely determinte sequene of disrete stepE yEstep proedures @where eh step is redily exeutle y very limited lulting gent or mhineAD suh thtX iF here isn9t ny room left for the exerise of imgintionD or intuitionD or fllile humn judgmentF iiF here is no reline on orlesF iiiF here is no reline on rndom methodsF

Denition 2.8 Denition 2.9

en algorithmic procedure is one tht suitly progrmmed omputer n exeuteF

e propertyGreltion is eectively decidable i' there is n lgorithmi proedure tht suitly progrmmed omputer ould use to deideD in (nite numer of stepsD whether the propertyGreltion pplies in ny given seF Q

Denition 2.10

e totl funtion is eectively computable i' there is n lgorithmi proedure tht suitly progrmmed omputer ould use for lultingD in (nite numer of stepsD the vlue of the funtion for ny given rgumentF

yf ourseD we won9t worry out prtil limittions on omputtion in our ssessment of the exeution of lgorithms nd the relted notionsF ell we re onerned with is whetherD in prinipleD suh omputtions ould e ompleted in (nite mount of timeF smportntlyD we know tht algorithmic computability is architecture independent 3

Turing's Thesis

he numeril funtions tht re e'etively omputle in the informl sense re just those funtions tht re in ft omputle y suitle uring mhineF vikewiseD the numeril questions tht re e'etively deidle in the informl sense re just those questions tht re deidle y suitle uring mhineF

Denition 2.11

e set is where f enumertes

enumerable i' is either empty or there is surjetive f : N

Theorem 2.1 Cantor's Theorem


Proof:

" here re @in(niteA nonEenumerle setsF

gonsiderD P D the powerset of N @the set whose memeers re ll the sets of numers X P i' X N xow suppose for redution tht some funtion f enumertes P D f : N P F e n onstrut the digonl set D N suh tht n D i' n f (n)F / fut sine D P nd f enumertes ll of the memers of P D then there must e numer d suh tht f (d) = DF hereforeD for every nD n f (d) i' n f (n)D so d f (d) i' d f (d)D ut this is / / ontrditionD so there n e no suh funtion s f tht enumertes P F hereforeD P is not enumerleF 2

Denition 2.12 Theorem 2.2

e set is eectively enumerable @eFeFA i' either is empty or else there is n e'etively omputle funtion tht enumertes itF he set of ll ordered pirs of numer n, m is eFeFX

Proof:
sn order to enumerte the set of ordered pirsD you list them on n rry nd zigE zg through the mtrixD enumerting one memer t timeF his proess de(nes ijetion f : N N2 D oneEtoEone orrespondene from n to the nEth pirF 2 (n+m)(n+m+1) f (n, m) N f (n, m) = +m 2

Chapter 3: Eective computability


Denition 3.1 Eective computation Denition 3.2
en is the suessful exeution of n lgorithmF

algorithm is sequene of elementry instrutionsD suh tht it isX

deterministi"tht isD hs no rndom movesY nd


R

hs no orles"tht isD hs no outside soures of informtionF

Denition 3.3 Denition 3.4

e successful execution of n lgorithm mens tht the lgorithm will eventully hlt in the proper outputF e funtion f : is eectively computable i' either is empty or there is n lgorithm suh tht for x there is suessful run of the lgorithm whose output is f (x)

sn prtieD we will restrit ourselves to numeril funtions sine other funtions tht we might e onerned with n e enoded into numeril funtionsF yf ourse there re numer of equivlent wys to speify the notion of lgorithmX turing mhinesD register mhinesD reursive funtionsD ElulusF

Denition 3.5 Denition 3.6 Denition 3.7


ome resultsX

e property or reltion is e'etively omputleF

(eectively) decidable i' its hrteristi funtion is

e set is e'etively deidle i' there is n lgorithmi proedure tht deides in (nite numer of steps whether ny given ojet is memer of F

e set of nturl numers N is e'etively deidle i' the hrtisti funtion of the property of elonging to is e'etively omputleF

Theorem 3.1

ivery (nite N is deidleF

Proof:
he lgorithm tht deides (nite set is just the lgoritm tht heks the input ginst every memer of the setF 2

Theorem 3.2

sf is deidleD then so is the omplement

Proof:
ke the hrteristi funtion for nd output I if c results in H nd H if c results in IF 2

Denition 3.8

N is eectively enumerable @eFeFA i' there is n e'etively omputle funtion f tht enumertes itF f :N
sf is deidle then it is eFeF

Theorem 3.3

Proof:
ine is deidleD c is omputleF oD c (x) = {0 : x Y or I otherwise o (nd some funtion f : N gompute the vlue of f (0)D if it is 0D then list the vlue of f (0)D it is 1 then disrd itF gontinue with this proess y omputing f (1), f (2), . . . in the sme wyF husD you hve surjetive funtion f : N nd so is eFeFF 2

Theorem 3.4

sf oth nd re oth eFeFD then is deidleF


S

Proof:
f :N s0 s1 F F F g:N t0 t1 F F F c on input x strt listing y dovetiling s0 , t1 , s1 , t1 , . . .

Theorem 3.5

is eFeF i' it is the domin of n e'etively omputle funtionF

f = {x|f (x) is de(ned}

Proof:
sf is the domin of n e'etively omputle funtionD then is eFeF vet f : N e ontoF gonsider the lgorithm omputing funtion g : on input xD list f (0), f (1), f (2), . . . sf n when f (n) = x for ny nD output U @or whteverA g(x) is de(ned i' x sf is eFeFD then it is the domin of n e'etively omputle funtionD then either @iA is emptyD or @iiA there is n e'etively omputle funtion f whih enumertes por @iAD then oviously ny lgorithm tht never produes n output will su0eF por @iiAD then we will need to show tht there is n lgorithm tht omputes f F yf ourseD this follows quite quiklyD s there is n lgorithmi proess for omputing f F

Theorem 3.6

he set W of ll eFeF sets is enumerleF

Proof:
ke the funtion f : N W de(ned y f (e) = We nd this funtion will enumerte WF

Theorem 3.7

here re nonEeFeF sets of numersF

Proof:
uppose for redutio tht W is eFeFD then W W F fut this implies tht P is eFeF futD y gntor9s heorem we know tht P is not eFeFD so W is not eFeF nd hene not e'etively deidleF

Theorem 3.8 The Basic Theorem on E.E. sets


omplement K is not eFeF

here is n eFeF set of numers K D suh tht its

Proof:
vet K = {e|e We } " the hlting prolem "D then K is not eFeFF oD K ontins one memer from eh of the sets ontined in W D whih mens tht
T

e K i' e We F his implies tht K = We for eh eF ine K is eFeF we n / dovetil omputtions of 0 on HD 1 on ID. . . F K is not deidleD euse if it ws K would e lso nd this would imply tht K would e eFeFF oD K is not eFeF ou nnot deide priori whether some omputtion will hltY orD the hlting prolem is not deidleF

Theorem 3.9

here is n eFeF set whih is not deidleF

Proof:
his ws proved in the seond hlf of the proof for

Theorem 3.8F

Chapter 4: Axiomatized formal theories (AFT)


wo resons for relying on epX

they enfore stndrd of rigorY nd they llow for ler regimenttion

Denition 4.1

en axiomatized formal theory (AFT) omprises formalized formule from the lnguge whih re axiomsD nd proof theoryF

languageD set of

Denition 4.2 Denition 4.3

en interpreted languageD LD is n ordered pir L, I where L is synttilly de(ned system of expressions nd I is the interprettion of LF

L9s syntti omponent L is mde up of severl pieesX

L's non-logical vocabulary: he symols or (nite strings of symols tht onstitute onE stntsD preditesD nd funtions of LF L's logical vocabulary: he symols or strings of symols of logi @vrilesD onnetivesD qunti(ersD identityD rketsAF L's

construction rules: he rules tht determine wellEformedness in LF

Denition 4.4

I D the interprettion of LD (xes the ontent of eh losed LEw' nd meets the following onditionsX it spei(es the domin of qunti(tionY interprets the nonElogil voulry y ssigning vlues to the domin of onstntsY gives sstisftion onditions for the preditesY nd dels with the logil voulry in the usul wyF

Denition 4.5 Denition 4.6

e proof system is dedutive pprtus tht de(nes whih (nite rrys of w's ount s proofs in the lngugeF qiven derivtion of the sentence from the xioms of the theory T using the kground logil proof systemD is theorem of the theory @T A e theory T is

Denition 4.7 Denition 4.8

sound i' every theorem of T is trueF decidable i' the property of eing T Etheorem is n
U

e theory T is eectively e'etively deidle propertyF

Denition 4.9
T F

e theory T

decides the sentene i' either T

or T

F nd if is flseD

Denition 4.10 Denition 4.11 Denition 4.12


T F

e theory T

correctly decides just whenD if is trueD T

e theory T is negation complete i' T deides every sentene of its lnguge @for every D either T or T AF e theory T is

inconsistent i' for some sentene D we hve oth T

nd

Theorem 4.1

sf T is n xiomtized forml theory then @iA the set of w's of T D @i9A the set of sentenes of T D @iiA the set of proofs onstrutile in T D nd @iiiA the set of theorems of T D n eh e e'etively enumertedF

Proof:
@iA ine T hs (nite lphetD there is n lgorithm for enumerting the strings of symols of T y strting with the strings of length ID then enumerting the strings of length P nd so onF purthermoreD sine T is xiomtizedD we know tht there is method of determining whih of these strings ount s w's of T F oD we know tht the set of w's of T n e enumerted y enumerting the strings of symols nd keeping only those tht ount s w'sF 2 @iiA his is proved in the sme wy s @iAD ut with sentenes insted of w'sF 2 @iiiA ine proof is just sequene of w's nd we know tht the set of w's is eFeFD we n enumerte ll of the proofs y enumerting ll of the possile strings of w'sF es we enumerte ll of the possile strings of w'sD we n keep only those tht egin with xioms nd then reord the (nl line of those proofsF his will produe n enumertion of ll of the T EtheoremsF 2

Theorem 4.2

eny onsistentD xiomtizedD negtionEomplete forml theory T is deidleF

Proof: Theorem 4.1 gives us n lgorithm for e'etively enumerting the theorems of T F

yf ourse then if we wnt to hek whether sentene is T EtheoremD then run the lgoritm nd if turns upD then it is theoremD nd if turns upD then it is not theoremF

Chapter 5: Capturing numerical properties


ome nottionl notesX

sans serif font will e reserved for expressions in lngugeF ordinry type fe will metElogil lngugeF qree letters will lwys e metlinguisti vrileF
e will e onsidering the lnguge LA D whih will denote the lnguge of si rithmetiF o LA = LA , IA is the lnguge of si rithmeti nd is s followsX

he

logical voulry of LA will e ordinry (rst order logiF

he non-logical voulry is {0, S, +, }D where 0 is onstntD S is one ple suessor funtionD nd + nd re twoEple funtionsF e e e

term of LA is n expression tht is uilt up of the nonElogil voulryF closed term is lnguge tht hve no vrilesF numeral of LA is term uilt from single onstnt nd just the suessfor funtionF

Atomic ws re uilt from terms nd the identity prediteF

he interprettion IA tkes N s its domin nd interprets the nonElogil symols in the ordinry wyF

IA ssigns vlues to losed terms s followsX


IF 0 is zero @val[0] = 0A PF sf is losed termD then val[S ] = val[ ] + 1F QF sf nd re losed termsD then val[( + )] = val[] + val[ ] nd val[( )] = val[] val[ ]

he

atomic sentences of LA ll tke the form = D where nd re termsF

woleulr sentenes re uilt up from the truth funtionl onnetivesF

Denition 5.1

e numeril property P is expressed y the open w' (x) with one free vrile in n rithmetil lnguge L i'D for every nD if n hs the property P D then () is trueD nd n if n does not hve the property P D then () is trueF n hether P is languageF

expressible in given theory just depends on the rihness of tht theory9s captures the property P y the open w' (x) i'D for ny nD

Denition 5.2

e theory T

if n hs the property P D then T ()D nd n if n does not hve the property P D then T ()F n hether property P n e nd proof systemF

captured y the theory depends on the rihness of its axioms

yf ourse in sound theory T D if (x) ptures P in T D then (x) expresses P F

Chapter 6: The truths of arithmetic


Denition 6.1 Theorem 6.1
en interpreted forml lnguge L is suciently expressive i' @iA it n express every e'etively omputle oneEple numeril funtionD nd @iiA it n form w's whih quntify over numersF sf lnguge L su0iently expressiveD then the set of LEtruths is not eFeF

Proof: First step:


e know from the fsi heorem out eFeF sets of numers tht there is set K D whih is eFeFD ut suh tht K is not eFeFF uppose tht the the omputle funtion f enumertes K D so n K i' x f (x) = nD where x rnges over the nturl numersF

Second step:

oD sine f is omputleD then given tht L is su0iently expressiveD there will e some w' of L tht expresses f D y the de(niton of su0iently expressive lngugeF gll tht w' F(x, y)D then f (m) = n just when F(m, n) is trueF

Third Step:

n K i' x(Nat(x) F(x, n)) is trueY nd n K i' x(Nat(x) F(x, n)) is trueF

Fourth step:
fut suppose tht the set of LEtruthsD T is eFeFD then given desription of the expresE sion FD we n enumerte T nd whenever we get sentene like x(Nat(x) F(x, n)) for some nD we list nF uh proedure will enumerte ll the memers of K

Fifth step:

fy hypothesis K is not eFeFD so T n9t e eFeF eitherF

Denition 6.2

e set of w's is axiomatizable i' there is n e'etively xiomtized forml theory T suh tht for ny w' D i' T F sive lnguge L is not xiomtizleF of true sentenes of su0iently expresE

Theorem 6.2 The incompletability theorem " he set T


Proof:

sf it wereD then it would lmost immeditely ontrdit heorem TFI euse the T Etheorems would e le to e e'etively enumertedF

Theorem 6.3

sf T is sound xiomtized theory whose lnguge is su0iently expressiveD then T nnot e xgF

Proof:
ine T is xiomtizleD the T Etheorems n e enumertedF fut sine T 9s lnguge is su0iently expressiveD the truths of T nnot e enumertedF his mens tht there re truths tht ren9t theoremsF ke one of these T Etruths tht is not T EtheoremD F e know tht is flseD so we know tht T euse T is soundF fut we lso know tht T euse is not theoremF o T is not xgF2

Chapter 7: Suciently strong arithmetics


ht we9ve shown to this point is tht for sound theoriesD we hve n inompleteness theoremD ut we n get something moreD in ftD we n rgue from mere onsisteny to inompletenessF sn order to do tht thoughD we 9ll need to ssume tht the theories under onsidertion n pture enoughD not just express enoughF

IH

Denition 7.1

e forml theory of rithmeti T is deidle numeril propertiesF iFeF either T

suciently strong i' it ptures ll e'etively

(x) or T

(x) where is n ritrry numeril propertyF

yf ourseD we n note tht there is totlly trivil wy for theory to e su0iently strong" tht isD prove ll of the rithmetil truths"if the theory is inonsistentD then oviously it will do the trikF futD tht isn9t interesting t llD we might wnt onsistentD nd furthermore deidle theoriesF elsX

Theorem 7.1
leF

xo onsistentD su0iently strongD xiomtized forml theory of rithmeti is deidE

Proof:

ine T is xiomtizleD then there is method for enumertiving the w's of T F es suhD we n onstrut the digonl property D suh tht n hs the property D i' T n ( ) n uppose tht d (x) is the w' tht ptures D nd sine T is su0iently strong T d (x) or T ( x)F o suppose tht T d (x)D then we know tht sine T is su0iently strong it n pture ll deidle numeril properties nd so it n pture D with d (x)F oD if d (x) ptures DD then for ny nD if n hs DD then T d ()D n if n doesn9t hve hD then T d ()D n o of ourseD then immeditely where n = dD we know tht if d hs DD then T d (d)D ut lso y the originl de(nition tht d hs the property D i' T d (d)F o these re oth theorems of T nd T is inonsistentF 2

Theorem 7.2
xgF

e onsistentD su0iently strongD xiomtized forml theory of rithmeti nnot e

Proof: Theorem 4.2: eny onsistent @IAD xiomtized @PAD negtion omplete @QA theory

T is deidle @RAF (1 & 2 & 3 4) Theorem 7.1: xo onsistent @IAD su0iently strong @SAD xiomtizle @PA theory T is deidle @RAF (1 & 5 & 2 4) oD oviously ny onsistentD su0iently strongD xiomtized forml theory nnot e xg (1 & 5 & 2 3) 2

Chapter 8: Two formalized arithmetics


ghpter V will explore two rithmetil theoriesX fy erithmeti @feA nd oinson erithmeti @AF foth of these theories will e no riher thn pyv nd will e le to quntify over numersD ut neither of these theories will llow us ny seond order qunt(tionF

Baby Arithmetic (BA):


he lnguge of BA is the lnguge of rithmetiD ut without the logil pprtuses of qunti(ers nd vrilesF e9ll ll the lnguge of BAD LB D suh tht LB = LB , IB F he lnguge LB is the ordinry lngugeD exept it lks qunti(ers nd vrilesD while the interprettion IB will lso e the stndrd oneF purthermoreD the proof system will e the stndrd dedutive system of propositionl logi with onnetivesF II

Denition 8.1

en

equation is w' of the form = D where nd re losed termsF

ine BA does not hve qunti(ersD ut we still need to express generl fts s xiomsD we will o'er the following shemt where 9 nd 9 stnd for numerlsF

Schema Schema Schema Schema Schema Schema

1 2 3 4 5 6

0 = S S = S = +0= + S = S( + ) 0=0 S = ( ) +

yf ourseD BA gives us the ility to derive ll of the results out the ddition nd multiplition of two numersD euse of the undne of xioms generted y eh shemF

Theorem 8.1

BA is negtion ompleteF

Proof:
he proof for this will ome in two prtsY in order to show tht fe is negtion ompleteD we will prove the following fts out ny losed terms nd X IF sf = is trueD then BA = F PF sf = is flseD then BA = F

Proof of (1):
ine our xiom shems will prove ny true simple eqution of the forms + = m or = mD we re le to prove more omplex identities y j k j k (rst evluling the vlues of terms on the innerEmost rkets nd then performing sustitutions sed on vFvF es suhD we will e le to prove the truth of ny true sentene ontining losed termsF 2

Proof of (2):
sf two losed terms of the lnguge nd hve vlues m nd nD suh tht m = nD then we will e le to derive pir of w's of the form = m, = nF fut then we know tht BA m = n. 2 es suhD BA is negtion omplete euse it deides every senteneF

Robinson Arithmetic (Q):


Q will e the extension of BA tht hs the full expressive power of pyv"in prtiulrD Q lso llows us to use (rst order qunti(ersD whih mens tht we9ll e le to void using the xiom shems of BA @tht stnd in for in(nite numer of xiomsA nd reple them with singulr generlized xiomsF

Axiom Axiom Axiom Axiom Axiom

1 2 3 4 5

x(0 = Sx) xy(Sx = Sy x = y) x(x = 0 y(x = Sy)) x(x + 0 = x) xy(x + Sy = S(x + y))
IP

Axiom 6 Axiom 7 Theorem 8.2

x(x 0 = 0) xy(x Sy = (x y) + x) Q is not negtion ompleteF

Proof:
Q is lnguge very strong on prtiulrs nd so while for ny nD Q 0 + n = nD it is not strong enough to prove the ommutitivity of dditionY s resultD it nnot prove the universlly generlized sentene x(0 + x = x) @or its negtionAF es suhD Q is not negtion ompleteF 2
ht sidD Q is su0iently strongD s it n prove the w's out the deidle properties of individE uls numersD even though it isn9t strong enough to pture generl truthsF e will move forwrd to show eventully tht interestingly Q is strong enough to pture ll of the pFrF propertiesF

Chapter 9: What Q can prove


his hpter will disuss numer of the results tht Q is strong enough to prooveF

Denition 9.1

he less-than-or-equal-to reltion is ptured y the w' v(v + x = y) in QD for ny numers m, n if m nD then Q v(v + m = n) nd otherwise Q v(v + m = n)F

Proof:
ke some m nD therefore for some k 0D k +m = nF o oviously Q + m = n k nd y iFqFD we know tht Q v(v + m = n)F elterntivelyD if we suppose tht m nD then of ourse Q v(v + m = n)F qiven the ove resultD we n dd the symol 9 to LA D the lnguge of QF es suhD writing will e short for v(v + = )F xow tht we9ve introdued the reltion into the lngugeD we n introdue notion of ounded qunti(tion in the following wyX

Denition 9.2 Denition 9.3 O1 O2 O3 O4 O5 O6 O7 O8 O9


T

he w's with bounded quantiers ( )() nd ( )() revite the w's ( ()) nd ( ()) respetivelyF e theory T is

order adequate if the following propositionsD yIEyWD holdF

x(0 x)F x({x = 0 x = 1 . . . x = n} x n) x(x n {x = 0 x = 1 . . . x = n})F (0), T (1), . . . , T ()D then T n (x n)(x) (x n)(x) (0)D or T (1), . . . , or Q ()D then Q n

por ny nD T por ny nD T por ny nD if T por ny nD if T por ny nD T por ny nD T por ny nD T

x(x n x S) n x( x ( = x S x)) n n n x(x n n x) (x n 1)(x) (x n)(x = n (x))

por ny n > 0D T

IQ

Theorem 9.1

Q is orderEdequteF

Proof:
yf ourse it will su0e to show tht

O1

O1-O9 hold in QX
exiom R iq I he(nition of P2

Proof:
IF Q PF Q QF Q

x(0 x)F x(x + 0 = x) xz(z + 0 = x) x(0 x) x({x = 0 x = 1 . . . x = n} x n)

O2

Proof:

por ny nD Q

O3

uppose tht a = 0 a = 1 . . . a = nF prom eh of these disjunts we prove tht a n nd soD y ses we know tht a nF hereforeD we9ve shown tht for n ritrry aD proves (a = 0 a = 1 . . . a = n)D so for ny nD T x({x = 0 x = 1 . . . x = n} x n) 2 por ny nD Q x(x n {x = 0 x = 1 . . . x = n})F he proof will e y informl indutionD if we show tht the x(x n {x = 0 x = 1 . . . x = n}) is provle for n = 0 nd tht if it is provle for n = k D then it is provle for n = k + 1D then we will hve our desired resultF por the se where n = 0X IF Q a 0 a = 0 uppose for ontrdition PF Q a = 0 a = Sa exiom Q QF Q b + a = 0 he(nition of nd I RF Q b + Sa = 0 vv P nd Q SF Q S(b + a ) = 0 exiom S TF Q a 0 a = 0 fy ontrdition with exiom I nd S 2 por ny nD if T (0), T (1), . . . , T ()D then T (x n)(x) n IF PF QF RF

Proof:

O4

Proof:
Q Q Q Q

(0) (1) . . . ( k) x(x k)(x) ((0) (1) ( k)) (0) (1) . . . ( x n)(x) k) (0)D or Q (1), . . . , or Q

O5

uppose essume for ontrdition sD P gontrdition I E Q 2

Proof:
IF PF QF RF

por ny nD if Q

()D then Q n

(x n)(x)
essume pyv I iq P g IEQ 2

O6

Q (0)D or Q (1), . . . , or Q () n nD Q ( por some k k) Q x n(x) por ny nD if Q (0)D or Q (1), . . . , or Q then Q (x n)(x) x(x n x S) n

() n

Proof:
IF Q PF Q

por ny nD Q

an a = 1 ... a = n
IR

essume yQD I

O7

Proof:

QF Q a = 1 . . . a = n a = S n RF Q a S n SF Q x(x n x S) n por ny nD Q x( x ( = x S x)) n n n

pyvD P yPD Q gD qixD IER

O8

IF Q n a PF Q n = a QF Q n = 0 . . . n = a RF Q n = . . . n = a 1 SF Q n = . . . S = a n TF Q n a n = a S a n UF Q (S a) n VF Q n = 0 . . . n = a WF Q (S = 0 . . . S = a) n n IHF Q n = a IIF Q n a S a n = a n IPF Q n a ( = a S a) n n IQF Q x( x ( = x S x)) n n n por ny nD Q x(x n n x)

essume essume for g yQD I pyvD PDQ pyvD R gD IES essume for g yQD I yQD U fy gses VD W gD UEIH pyvD IEII qen IP 2

O9

Proof:
IF PF QF RF SF TF UF

he proof will e y informl indutionF pirstD we will prove tht yV holds for the se where n = 0F yf ourseD this immeditely followsD for if n = 0D then y O1D we know tht 0 aY ndD s suhD 0 a a 0F xextD suppose tht the result holds where n = k D to show tht the result holds for the se where n = k + 1F IF Q a a k k sr ak+1 PF Q a k O6 a ( = a k + 1 a) QF Q k k O7 ak+1 RF Q a = k hef of SF Q a k + 1 k + 1 a pyvD IER TF Q x((x x) (x k + 1 k + 1 x)) k k qixD S 2 por ny n > 0D Q (x n 1)(x) (x n)(x = n (x))

Proof:

Q Q Q Q Q Q Q

(x n 1)(x) essume (0) . . . (n 1) sD I (x n)(x = n (x)) essume for ontrdition (x n)(x = n (x) pyvD Q (0 = n (0)) . . . ( = n (n)) n he(nition ((0 = n (0)) . . . ( = n (n))) n pyvD P edutio nd g 2 (x n 1)(x) (x n)(x = n (x))

Denition 9.4
termsF

en

atomic 0 w is w' of one of the forms = , D where nd re


IS

Denition 9.5

he

full class of 0 ws is mde up of only the following types of w'sX

IF ivery tomi 0 w'Y PF sf nd re 0 w'sD so re D ( )D ( )D ( )D nd ( )F QF sf is 0 w'D so re ( ) nd ( )D where is ny vrile free in D nd is numerl or vrile distint from F

Denition 9.6

e w' is strictly 1 i' it is of the form . . . D where is 0 D nd , , . . . , re one or more distint vriles free in F

Denition 9.7 A w is 1 i' it is logilly equivlent to stritly 1 w'F Denition 9.8


e w' is strictly 1 i' it is of the form . . . D where is 0 D nd , , . . . , re one or more distint vriles free in F

Denition 9.9 A w is 1 i' it is logilly equivlent to stritly 1 w'F


vet us puse now to note few interesting results out the rithmetil hierrhyX

Fact 9.1

he negtion of 0 w' is lso 0 F

Proof:

his result is trivilD s merely dding negtion to formul with no qunti(ers does not hnge nything out the omplexity of tht formulF 2

Fact 9.2 Fact 9.3

he negtion of 1 w' is 1 nd the negtion of 1 w' is 1 F

Proof:

his is lso trivil given the equivlenes etween nd F e 0 w' is lso oth 1 nd 1 F

Proof:

his is lso quite oviousD s given ny 0 sentene D ( z = z) is lso 0 D so then z( z = z) is stritly 1 nd the sme follows for the se with the universl qunti(erF yf ourseD sine the qunti(ed sentenes re extensionlly equivllent with D they re no more omplex nd is 1 nd 1 F

Fact 9.4

he truth or flsity of ny losed 0 w' y n lgorithmi lultionF

Proof:

he proof is y indution on the omplexity of the w' nd is pretty strightforwrdD so s won9t other to reherse itF e n now return to explin some further wys in whih theories intert with the rithmetil hierrhyF

Denition 9.10 Denition 9.11

e theoryD T D is -sound i'D for ny Esentene D if T

D then is trueF F

e theoryD T D is -complete i'D for ny Esentene D if is trueD then T

e know tht Q n prove ll of the true 1 sentenes euse it proves ll of the true prtiulr sentenes ndD s suhD will prove those sentenes fter pplying iFqF ndD s suhD we n showX

Theorem 9.2

Q is 1 EompleteF
IT

Proof:
e will show this y proving the following three ftsX IF Q orretly deides every tomi 0 senteneF PF Q orretly deides every 0 senteneF QF Q proves every true 1 senteneF he proofs of these three fts re so ovious tht s won9t other to mention themF

Theorem 9.3
with QAF

e 1 sentene is true i' is not logilly deduile from Q @iFeF i' is onsistent

Proof:
pirst the leftEtoErightX sf 1 sentene is trueD then is not logilly deduile from QF uppose otherwiseD then Q would prove flse 1 Esentene D ut sine Q is 1 EsoundD we know tht this n9t hppenF xow the right to leftX sf the 1 sentene is not logilly deduile from QD then the 1 sentene is trueF his follows from the 1 Eompleteness of QF 2

Denition 9.12

en interpreted lnguge L2

includes the lnguge L1 i'

every L1 Ew' is lso w' of L2 D perhps llowing for some de(nitionl extensions of L2 Y nd the opies of L1 Ew's in L2 hve the sme truth onditions s their originlsF

Denition 9.13 Theorem 9.4

en theory T2 extends T1 if T2 9s lnguge inludes T1 9s lnguge nd T2 n prove t lest ll of T1 9s theoremsF sf T extends QD then T is onsistent i' it is 1 EsoundF

Proof:
T is 1 Esound i' every 1 sentene tht T proves is true s sentene of rithmetiF xowD suppose tht T proves flse 1 sentene F will e then true 1 senteneD oviouslyF sf tht is the se thoughD then sine T extends QD whih is 1 EompleteD T will prove nd will e inonsistentF fy ontrpositionD if T is onsistentD it proves no flse 1 senteneD so is 1 EsoundF elterntivelyD if T is inonsistentD we n derive nything in T D inluding flse 1 sentenesD so T isn9t 1 soundF 2

Chapter 10: First-order Peano Arithmetic


P A is the extension of Q ut with the inlusion of the indution xiomsF e good portion of this hpter is devoted to explining mthemtil indutionD so tht we will e le to develop P AF
oD wht is mthemtil indutionc e use mthemtil indution s tool to prove fts tht hold out sequenesD y ssuming sis @some property P holds of some miniml unit in the sequeneAD nd n indutive hypothesisD for the purpose of showing tht on the sis of the indutive IU

step it holds of ll of the memers of the sequene @ssume P holds of some element in the sequene nD then the indutive step will e used to show tht it holds of n + 1AF roweverD in P AD we don9t hve ny wy to perform seondEorder qunti(tionD so we will not e le to give univol nd generl xiomti sttement of thisD so we9ll do it on the sis of n induction schema "where eh instne of the shem ounts s n xiom of P AF

Denition 10.1

eny instne of our induction schemaD ({0 x((x) (Sx))} x(x))D ounts s n xiomF

yf ourseD indution n e pplied to prove things not only out mondi propertiesD ut lso out reltionsF oD we n perform indution on reltionl properties if we ssume tht ll ut one of the rgument slots re (lled with ritrry elements of our dominY s suhD we will hve sentene tht is the universl losure of sentene with just one free vrileF he following limsD listed under ixerises9D re provle using just Q9s xioms nd indution for 0 w'sF

Exercises:
IF x(x = Sx)

Proof:
(x) = (x = Sx) fseX (0) @A x(0 = Sx) @A 0 = S0 sndutive stepX (n) (Sn) @A n = S(n) @A xy(x = y Sx = Sy) @A n = Sn Sn = SSn @dA Sn = SSn
PF x(0 + x = x) exiom I s I 2 sr exiom PD gontrposition sD P wD ID Q 2

Proof:
(x) = (0 + x = x) fseX (0) @A x(x + 0 = x) @A 0 + 0 = 0 sndutive stepX (n) (Sn) @A 0 + n = n @A xy(x + Sy = S(x + y)) @A 0 + Sn = S(0 + n) @dA 0 + Sn = Sn
QF xy(Sx + y = S(x + y)) exiom R sD I 2 essume exiom S sD P gD ID Q

Proof:
(x) = Sa + x = S(a + x)
IV

fseX (0) @A S(a + 0) = S(a + 0) @A S(a) = S(a + 0) @A Sa + 0 = S(a + 0) sndutive stepX (n) (Sn) @A Sa + n = S(a + n) @A Sa + n = a + Sn @A S(Sa + n) = S(a + Sn) @dA Sa + Sn = S(a + Sn)
RF xy(x + y = y + x)

pyv @PAD I exiom R 2 sr exiom SD I vvD P exiom SD Q 2

Proof:
(x) = x + a = a + x fseX (0) @A 0 + a = a + 0 @A 0 + a = a @A 0 + a = a + 0 sndutive stepX (n) (Sn) @A n + a = a + n @A S(n + a) = S(a + n) @A Sn + a = S(a + n) @dA Sn + a = a + Sn
SF xyz(x + (y + z) = (x + y) + z) sr @PAD I exiom RD P 2 sr vvD I @QAD P exiom SD Q 2

Proof:
(x) = x + (a + b) = (x + a) + b fseX (0) @A a + b = a + b @A 0 + (a + b) = a + b @A 0 + (a + b) = (0 + a) + b sndutive stepX (n) (Sn) @A n + (a + b) = (n + a) + b @A S(n + (a + b)) = S((n + a) + b) @A Sn + (a + b) = S((n + a) + b) @dA Sn + (a + b) = S(n + a) + b @eA Sn + (a + b) = (Sn + a) + b
TF xyz(x + y = x + z y = z) g @PAD I @PAD P 2 sr vv @QAD P @QAD Q @QAD R 2

Proof:
(x) = x + a = x + b a = b fseX (0) @A 0 + a = 0 + b
IW essume for g

@A a = 0 + b @A a = b @dA 0 + a = 0 + b a = b sndutive stepX (n) (Sn) @A n + a = n + b a = b @A Sn + a = Sn + b @A S(n + a) = S(n + b) @dA n + a = n + b @eA a = b @fA Sn + a = Sn + b a = b UF xy(x y y x)

@PAD I @PAD P g IEQ 2 sr essume for g @PAD P exiom PD Q wD sr g IES 2

Proof:
(x) = x a a x fseX (0)
sn order to prove 0 a a 0D we n unpk the reltion in the (rst disjunt in the stndrd wy nd rrive t v(v + 0 = a)Y ut of ourseD we know tht this is true euse it is just n instne of exiom RD where v = a. 2

sndutive stepX (n) (Sn)


sndutive rypothesisX n a a n uppose tht n aF fut this just mens tht for some cD c + n = aY furthermoreD we know y veiniz9s vw tht S(c + n) = Sa nd y exiom S tht c + Sn = SaF pinllyD y exiom QD we know tht either c = Sc c = 0F uppose the (rstX then we know tht Sc + Sn = Sa nd immeditely tht S(c + Sn) = Sa nd c + Sn = aF elterntivelyD suppose the seondX if c = 0D then 0 + Sn = Sa ndD therefore Sn = SaD y @PAF iither wyD if n aD then Sn a a Sn he seond disjunt entils tht for some cD c + a = nF hereforeD we know tht S(c + a) = SnD ut then Sc + a = Sn y @QA nd the seond disjunt of the onsequent of the indutive step holdsF oD if a nD then Sn a a SnF fut this mounts to proof of the indutive step nd we re (nishedF 2 VF xy((x y y x) x = y)

Proof:
(x) = x a a x x = a fseX (0) @A 0 a a 0 @A a = 0 @A a = Sa @dA a 0 @eA Sa 0 @fA Sa = 0 @gA
PH essume uppose for redutio exiom Q gD I gDR yQ gontrdition QD exiom I

@hA a = 0 edutio IEU @iA 0 a a 0 a = 0 g IEV2 sndutive stepX (n) (Sn) @A Sn a a Sn uppose for g @A (Sn = 0 Sn = 1 . . . Sn = a) (a = 0 a = 1 . . . a = Sn) yQD I @A Sn = a smmedite y ses nd P @dA Sn a a Sn Sn = a g IEQ 2 WF xyz((x y y z) x z)

Proof:
(x) = (x a a b) x b fseX (0) @A 0 = 0 @A 0 = 0 0 = 1 . . . 0 = b @A 0 b sndutive stepX (n) (Sn) @A Sn a a b @A Sn = 0 Sn = 1 . . . Sn = a @A a = b @dA Sn b @eA a < b @fA Sn = 0 Sn = 1 . . . Sn = a . . . Sn = b
g gD I yPD P 2 uppose for g yQD I uppose g I uppose g P 2

Theorem 10.1

sn ny theory whih extends Q nd hs indution for 0 w'sD 1 w' strting with n > 1 unounded existentil qunti(ers is provly equivlent to 1 w' strting with just single ounded qunti(erF

Proof:
e n show tht the following two sentenes re equivlentX IF xy(x, y) PF w(x w)(y w)(x, y) essume @IAD nd suppose tht for some a, b, (a, b)F esult @UA shows tht a b b aY soD y sesD we9ll show tht we n derive tht (a b b b (a, b)) (b a a a (a, b))F yf ourse eh disjunt implies the existentil lim nd so @IA implies @PAF essume @PAD ut then just unpking the qunti(ers gives you @IAF his rgument generlizes niely nd we hve our desired result 2 xow we will onsider P AD the lnguge of (rst order eno erithmetiX

Denition 10.2 First-order Peano Arithmetic


felow re the xioms of P AX

is the (rstEorder theory whose lnguge is LA nd whose xioms re those of Q plus the universl losures of ll instnes of the indution shemF

Axiom 1

x(0 = Sx)
PI

Axiom 2 xy(Sx = Sy x = y) Axiom 3 x(x = 0 y(x = Sy)) Axiom 4 x(x + 0 = x) Axiom 5 xy(x + Sy = S(x + y)) Axiom 6 x(x 0 = 0) Axiom 7 xy(x Sy = (x y) + x) Induction Schema ({(0) x((x) (Sx))} x(x))
gonsider the lnguge of resurger rithmetiD LP D whih hs ll of P A9s xioms for suessor nd ddition nd the universl losure of ll instnes of the sndution hemD ut does not ontin multiplitionF es it turns outD P is negtion ompleteF es historil noteD prior to qdel9s IWQI resultsD wojesz resurger proved the ompleteness of P Y of ourseD this ment tht qdel9s result z ws espeilly surprisingF BA ws known to e inompleteD while P ws shown to e ompleteD nd so it ws espeilly surprising when P A turned out to e inompleteF

Chapter 11: Primitive recursive functions


sn this hpter we will e onerned with the lss of primitive reursive @pFrFA funtionsF he lss of pFrF funtions onsist of those funtions tht n e de(ned y pir of lusesD the (rst of whih de(nes the vlue of the funtion for the zero seD nd the seond de(nes the vlue for the n + 1 seF oD of ourse in the fmilir wyD we n de(ne dditionD or multiplition reursivelyX

x+0=x x + Sy = S(x + y) x0=0 x Sy = (x y) + x

Denition 11.1

e twoEluse de(nition of funtion where the (rst luse (xes the vlue for P (0) nd the seond luse de(nes the vlue for P (Sx)D given P (x)D is denition by primitive recursionF he composition of funtions ours when the omposite funtion tkes the outE put@sA from one or more funtions nd tkes them s inputs for nother funtionF e chain of de(nitions y reursion nd funtionl omposition is hin suh tht working from the ottom upD eh line is de(ned in terms of the funtions ppeled to on the previous lineF

Denition 11.2

Denition 11.3

Denition 11.4

e primitive recursive function is one tht n e simply hrterized using hin of de(nitions y reursion or ompositionF

e n onsider the generl form of oneEple de(nition y reursionX

f (0) = g f (Sy) = h(y, f (y))


elterntivelyD onsider the generl sheme for de(ning twoEple funtionsX PP

f (x, 0) = g(x) f (x, Sy) = h(x, y, f (x, y))

Denition 11.5

he initial functions onsist of the suessor funtion S D the zero funtion Z(x) = 0D nd the k Eple identity funtionsD Iik (x1 , x2 , . . . , xk ) = xi for eh k D nd for eh i, 1 i kF

Denition 11.6 sf we suppose tht the following holdsD then we will sy tht f is dened from g and h by primitive recursion X
f (, 0) = g() x x , Sy) = h(, y, f (, y)) f( x x x Denition 11.6 sf g() nd h(, u, ) re funtions ! with nd possily emptyD then f is y x z x z , , ) = h(, g(), )F dened by composition by substituting g into h just if f ( x y z x y z
e will onsider funtions extensionllyD mening tht we will ount two funtions s the sme funtion i'D for eh nD f (n) = g(n)F oD we know tht ll pFrF funtions re omputle"sine the pFrF funtions re omposed out of the initil funtionsD nd omputility is losed under ompositionD this shouldn9t e surprisingF roweverD interestinglyD not ll omputle numeril funtions re primitive reursiveF

Theorem 11.1
siveF

here re e'etively omputle numeril funtions whih ren9t primitive reurE

Proof:
ine the set of pFrF funtions funtions s followsX H I P f0 f0 (0) f0 (1) f0 (2) f1 f1 (0) f1 (1) f1 (2) f2 f2 (0) f2 (1) f2 (2) f3 f3 (0) f3 (1) f3 (2) F F F F F F F F F F F F is eFeFD we know tht we n enumerte the pFrF Q f0 (3) f1 (3) f2 (3) f3 (3) F F F

... ... ... ... ... FF F

xow we will de(ne the digonl funtion (n) = fn (n) + 1D whih tkes the nEth funtion on our enumertion nd lls the vlue of tht funtion on the numer n nd dds 1F ht isD di'ers from every one of the pFrF funtions in t lest one pleFD howeverD is surely omputleD s it n e onstruted entirely mehnillyF oD we hve omputle funtionD whih is not pFrF 2

Denition 11.7

he characteristic function of the numeril property P is the oneEple funtion cP suh tht if m is P D then cP (m) = 0D nd if m isn9t P D then cP (m) = 1F he hrteristi funtion of two ple numeril reltion R is the two ple funtion cR suh tht if m is R to nD then cR (m, n) = 0 nd if m isn9t R to nD then cR (m, n) = 1F

Denition 11.8

e numeril property is totl nd e'etively omputleF

eectively decidable just if its hrteristi funtion is

PQ

Denition 11.9 e p.r. property is property with pFrF hrteristi funtionD p.r. relation is reltion with pFrF hrteristi funtionF Denition 11.10
he suh tht . . . 9F

nd likewise

minimization opertor x9 is now introduedD whih is red the lest x

e lso know tht the following fts hold out funtionGreltion uildingX

@eA sf f () is n n ple pFrF funtionD then the orresponding reltion expressed y f () = y is x x n n + 1Eple pFrF reltionF

Proof:
uppose tht f is oneEple funtionF he hrteristi funtion of the reltion expressed y f (x) = y is given y c(x, y) = sg(|f (x) y|)F yf ourseD sine ll of the omponents of the hrteristi funtion re pFrF nd re omined ompositionllyD so c is pFrFF xturllyD similr onstrution will hold for ny nEple funtionF @fA rimitive reursion is losed under the truthEfuntionl onnetivesF

Proof:
uppose tht p(x) is the hrteristi funtion of the property P F e know tht sg(p(x)) is the hrteristi funtion of the property P sine sg )ips the vluesF yf ourseD sine sg(p(x)) is just omposition of pFrF funtionsD it is pFrFF oD if P is pFrF propertyD so is P F eginD suppose tht p(x) nd q(x) re the hrteristi funtions of pFrF properties P nd Q respetivelyF husD p(n) q(n) tkes the vlue 0 if either of them re HD so p(n) q(n) mounts to the disjuntive property of eing either P or QF yf ourseD this is ll y pFrF ompositionD so the result is pFrF nd if two properties re pFrF then the disjuntive property of P or Q is tooF pinllyD given tht negtion nd disjuntion re shown to e pFrF nd we know tht pyv llows us to generte ll of the other foolen opertors with just negtion nd disjuntionD we know tht primitive reursion is losed under the onnetivesF 2 @gA eny property or reltion de(ned from pFrF reltion or property y ounded qunti(tion is lso pFrF

Proof:
his should e immeditely pprentD given tht there is purely mehnil proess for heking in (nite numer of steps whether n existentil or universl lim is true over (nite numer of intervlsF 2 @hA sf P is pFrF propertyD then the funtion f (n) = (x n)P (x) is pFrF nd more generllyD if g(n) is pFrF funtion nd P is pFrF propertyD then f (n) = (x g(n))P (x) is lso pFrF

Proof:
uppose tht p is the hrteristi funtion of property P D nd k is the hrtersti funtion of the sentene tht is expressed y n instne of ounded qunti(tionF xow we onsider the funtionD whih is de(ned reursively s followsX

f (0) = 0 f (n) = k(n 1) + k(n 2) + . . . + k(1) + k(0) for n > 0F


PR

futD k(i) = 1 for eh i tht isn9t P D nd k(i) = 0 s soon s we (nd j tht is P F o f (n) = (x n)P (x)F end the rest of this rgument follows in the sme wy s the previous onesF 2 @iA eny funtion de(ned y ses from other pFrF funtions is lso pFrF

Proof:
tust note tht

f (n) = sg(c0 (n))f0 (n)+sg(c1 (n))f1 (n)+. . .+sg(ck (n))fk (n)+c0 (n)c1 (n) . . . ck (n)a
sine sg(ci (n)) = 1 when Ci (n) nd is otherwise zeroD nd the produt of ll the ci (n) = 1 just in se none of the Ci (n) re trueD nd is zero otherwiseF 2 e n now prove tht some reltions re pFrFX

R1

he reltions m = nD m < nD nd m n re pFrF

Proof:
yf ourseD the hrteristi funtion of m = n is sg(|m n|) nd the hrterE isti funtions of m < n nd m n re sg(monus(Sm, n) nd sg(monus(m, n)) respetivelyF 2

R2

he reltion m|n tht holds when m is ftor of n is pFrF

Proof:
e know thtX

m|n (y n)(0 y 0 < m m y = n)F


oD the reltion is pFrF

R3

vet P rime(n) e true just when n is prime numerY then P rime is pFrF property

Proof:
his follows pretty trivilly s wellD forX

P rime(n) n = 1 (u n)(v n)(x v = n (u = 1 v = 1))


eginD this is just uilt from simple omposition of truthEfuntionl onnetives nd pFrF reltionsF 2

R4

vist the primes s 0 , 1 , 2 , . . . Y then the funtion (n) whose vlue is n is pFrF

Proof:
yf ourseD the funtion n D whose vlue is the nEth prime is pFrF"onsider the de(nitionX

0 = 2 Sn = (x n! + 1)(n < x P rime(x))


o oviously this is pFrF

R5

vet exp(n, i) e the ! possily zero ! exponenet of the prime numer i in the ftoriztion of nY then exp is pFrF funtionF

Proof:
yf ourseD this funtion is well de(nedX PS

x+1 x exp(n, i) = (x n){(i |n) (i |n)}F

oD we9re doneF

R6

vet len(0) = len(1) = 0Y nd when n > 1D let len(n) e the length of n9s ftoriztionD iFeF the numer of distint prime ftors of nY then len is pFrF funtionF o onsider the funtion

p(m, n) = sg(pf (m, n))F


o then p(m, n) = 1 just when m is prime ftor of n nd H otherwiseF oD

len(n) = p(0, n) + p(1, n) + . . . + p(n 1, n) + p(n, n)F


herefore to give pFrF de(nition of lenD we n de(neX

l(x, 0) = p(0, x) l(x, Sy) = (p(Sy, x) + l(x, y))


nd let len(n) = l(n, n)F

Proof:
(P rime(m) m|n) holds when m is prime ftor of nF xturllyD this is pFrF reltion nd hs pFrF hrteristi funtionD tht we9ll ll pf (m, n)F

Chapter 12: Capturing p.r. functions


e re uilding towrds notion of pFrF dequte theory @de(ned elowA nd will show tht Q nd P A re pFrF dequteF his hpter will e devoted to noting some importnt notions surrounding pturing funtionsF

Denition 12.1

e p.r. adequate theory of rithmeti is theory tht n ppropritely pture ll of the pFrF numeril funtionsD propertiesD nd reltionsF e oneEple numeril funtion is expressed L just if for ny m, nD if f (m) = nD then (m, n) is trueD if f (m) = nD then (m, n) is trueF

Denition 12.2

by (x, y) in n rithmetil lnguge

Denition 12.3
m, nD

e oneEple numeril funtion is

captured by (x, y) in theory T just if for ny

if f (m) = nD then T if f (m) = nD then T

(m, n)D (m, n)F

xow tht we9ve de(ned wht it mens for funtion to e ptured in theoryD we will further de(ne the further reltion tht holds etween funtion f nd w' (m, n)D where ptures f as a functionF

Denition 12.4

he oneEple funtion f is

captured as a function y (m, n) theory T just if

IF for every mD T

!y(m, y)D
PT

nd for ny m, n, PF if f (m) = nD then T QF if f (m) = nD then T

y(m, n)D y(m, n)F

Denition 12.5
just if IF T

he oneEple funtion f is

fully captured as a function y (x, y) in theory T

x!y(m, y)D

nd for ny m, n, PF if f (m) = nD then T QF if f (m) = nD then T

y(m, n)D y(m, n)F

ith those notions on the tleD we n sy some interesting things out their reltionsX iF sf fully ptures f s funtion in T D then ptures f s funtionF iiF sf ptures f s funtion in T D then ptures f in the wek senseF iiiF sf ptures f in the wek sense in T D then there is losely relted w' whih ptures f s funtion in T D so long s T extends QF o show thisD suppose tht ptures the one ple funtion f in T D whih extends QF xow de(ne s followsX

(x, y) (z y)((x, z) z = y)F


oD for given mD m, x) is stis(ed y unique nY furthermoreD we n show tht the new w' lso ptures f s funtionF ivF sf ptures f s funtion in T D then there is nother losely relted w' uppose tht ptures the one ple funtion f in T nd onsider the w' s followsX

(x, y) =df {(x, y) !y(x, y)} {y = 0 !y(x, y)}F


he upshot of ll of this is tht in rithmetil systems s strong s QD if funtion is pturle t llD it is pturle s funtion nd fully pturle s funtionF

Denition 12.6

e theory T is weakly p.r. adequate there is orresponding T tht ptures itF

(p.r. adequate) ifD for every pFrF funtionD

Chapter 13: Q is p.r. adequate


hokinglyD given the titleD this hpter will e devoted to show tht ny pFrF funtion n e canonically ptured in QF

Denition 13.1 Denition 13.2 Denition 13.3

f is 0 f is 0 f is 0

function i' it n e expressed y 0 w'Y function i' it n e expressed y 0 w'Y function i' it n e expressed y 0 w'F
PU

Theorem 13.1
Proof:

sf funtion is 1 it is lso 1 F

uppose one ple funtion f n e expressed y the stritly 1 w' (x, y)F oD f (m) = n i' z(f (m) = z z = n) yf ourseD this mens tht f is eqully well expressed y the 1 w' tht is the result of dding the universl qunti(er s oveF 2

Theorem 13.2
Proof:

Q n pture ny 0 funtion s funtionF

uppose the one ple funtion f is expressed y the 0 w' (x, y)D then if f (m) = n then (m, n) is trueF ine Q settles every 0 w'D Q (m, n) nd if f (m) = nD then (m, n) is flseF o (m, n) not only expresses f in QD ut lso ptures itF oD (m, n) ptures f s funtion in QF 2

Theorem 13.3
Proof:

Q n pture ny 1 funtion s funtionF

st will su0e to show thtX @IA eny 1 funtion is equivlent to omposition of two 0 funtionsF @PA Q n pture ny omposition of 0 funtionsF roof of @IAX roof of @PAX

Theorem 13.4
Proof:

ivery pFrF funtion n e expressed in LA F

uppose tht the following three propositions re trueX @IA LA n express the initil funtionsF @PA sf LA n express the funtions g nd hD then it n express funtion de(ned y omposition from g nd hF @QA sf LA n express the funtions g nd hD then it n express funtion f de(ned y pFrF from g nd hF roof of @IAX roof of @PAX roof of @QAX

Theorem 13.5
Proof:

ivery pFrF funtion is 1 F

PV

Theorem 13.6
Proof:

Q is pFrF dequteF

his is proved immeditely from heorems IQFQ nd IQFSF

Chapter 15: The arithmetization of syntax


his hpter will tke three prtsX IF sntrodue oding of ojet lnguge sentenes in the ojet lngugeF PF sntrodue digonliztion for qdel odesF QF ixplin how Atom, W f f, Sent, nd P rf re pFrFF e will ssign the following si odes to the symols of our ojet lngugeX

II

IQ

= IS

( IU

) IW

0 PI

S PQ

+ PS

PU

x P

y R

z T

... ...

yf ourse the ojet lnguge hs n in(nite numer of vrilesD ut they re simply ssigned si odes extending through the evensF

Denition 15.1

e symol9s basic code is the numer tht is priminirily ssigned to tht symol for the purpose of ssigning qdel numersF

Denition 15.2 vet e e the sequene of k + 1 symols ndGor vriles s0 , s1 , s2 , . . . , sk D then e9s Gdel number @gFnFA is lulted y tking the si ode numer ci for eh si D in turnD
using ci s n exponent for the i + 1Eth prime numer i D then multiplying the resultsF his will e represented sX
c c c c 00 11 22 . . . kk

e will sideEstep our revitions tht we introdued t erlier points when ssigning gFnFF es suhD we will lwys e enoding the longer versionF

Denition 15.3 e n ode strings will ll super Gdel numberF

of gFnFs using the sme method nd the result is wht we

uper gFnFs will e ruil in the enoding of proofsF oD we n de(n the proof reltion P rf s followsX

Denition 15.4

P rf (m, n) holds just if m is the super gFnF of sequene of w's tht is P A proof of the losed w' with the regulr gFnF nF T erm(n) holds when n odes for term of LA Atom(n) holds when n odes for n tomi w' of LA W f f (n) holds when n odes for w' of LA Sent(n) holds when n odes for sentene of LA T erm(n)D Atom(n)D W f f (n)D Sent(n)D nd P rf (m, n) re pFrFF
PW

Denition 15.5 Denition 15.6 Denition 15.7 Denition 15.8 Theorem 15.1

Proof:
his is very strightforwrdD s the proess of oding nd deoding is purely meE hnil nd so there should e no surprise tht these properties re pFrF e will dhere to the following nottionl onventionX

will denote the qdel numer of in our logiins9 ugmented inglish nd s well in the ojet lnguge to denoted the gFnF of F

Denition 15.9 Theorem 15.2

he

diagonalization of is y(y = )F

here is pFrF funtion diag(n) whihD when pplied to numer n whih is the gFnF of some w' yields the gFnF of tht w'9s digonliztionF

Proof:

R5 R6 R7 R8

he funtion exp(n, i) is pFrF where this returns the exponent of in the prime ftoriztion of nF he funtion len(n) is pFrFD where this returns the numer of distint prime ftors of nF here is concatenation function suh tht @iA m n is the gFnF of the expression tht results from stringing together the expression with the gFnF of m followed y the expression with the gFnF of nD nd @iiA is pFrF he funtion num(n) whose vlue is the gFnF of the stndrd numerl of n is pFrF

Proof:
he stndrd numerl for Sn for n > 0 is of the form S followed y the stndrd numerl for nD soX

num(0) = 0 = 221 num(Sx) = S num(x) = 223 num(x)F


oD num is pFrF

R9

he funtion diag(n) is pFrF he property V ar(n) whih holds when n is the gFnF of vrile is pFrF he property T ermseq(m, n) is pFrF he property T erm is pFrF he property Atom(n) whih holds when n is the gFnF of n tomi w' is pFrF

R10 R11 R12 R13

Chapter 16: PA is incomplete


his hpter will e devoted to showing how we might onstrut q sentenes nd how to use them to show tht P A is inompleteF

Denition 16.1

he reltion Gdl(m, n) holds just when m is the super gFnF for P A proof of the digonliztion of the w' with gFnF of n nd Gdl(m, n) is pFrF

QH

Proof:
Gdl(m, n) holds when P rf (m, diag(n))F he hrteristi ft of Gdl is de(nle y omposition fro the hrteristi ft of P rf nd the funtion diag D so of ourse Gdl(m, n) is pFrF 2
e will ept the following nottionl onventionsX

Gdl(x, y) stnds for 1 w' whih nonilly ptures GdlF U(y) =def xGdl(x, y) will e the senteneD whih our gol will e to onstrutF por onveE nieneD this will e revited s U G =def y(y = U U(y)) is the digonliztion of UF G is equivlent to xGdl(x, U )
yf ourseD thenD we know tht G is true i' it is unprovle in P AF his is n impressive result tht follows merely from the de(nitions nd shorthnds tht we9ve provided oveF roweverD if tht9s not enough for youD inspetX e know tht G is true i' there is no m suh tht Gdl(m U )"orD G is true i' there is no mD suh tht m is the ode for P A proof of the w' with the gFnF U D ut tht w' is just U nd its digonliztion is GF oD this ws ll just to sy tht G is true i' there is no m tht odes P A proof for GF roweverD if G is provleD then there would e suh odeF oD (nllyD we know tht G is true i' there is no proof of it in P AF 2

Theorem 16.1
nor P A

sf P A hs true xiomsD then there is LA Esentene suh tht neither P A

Proof:
sf we suppose tht P A is sound nd given wht we9ve just lerned out G in P AD we know tht P A doesn9t prove GD for if it didD then it would prove flse theoremF edditionllyD we know tht G is flseD euse G is trueD nd so P A won9t prove G eitherF e should note tht this result is sed on the ssumption tht P A is soundY howeverD we will need to show tht the inompleteness of P A n e proven without tht suppositionD ut with some muh weker ssumptionsF e know tht while G is very long nd omplexD it is lso not omputtionlly ll tht omplex"it is tully 1 X

Proof:
Gdl(x, y) is 1 nd so of ourse Gdl(x, U ) is 1 F yf ourseD we know then tht it9s negtion is 1 F hereforeD G is lso 1 D s it is equivlent to xGdl(x, U )F 2
es resultD we n see tht ny 1 sentene @like qoldh9s onjetureA is the uFgF of 0 w'D whih expresses pFrF property or reltionF oD we9ve estlished the followingX

Denition 16.2
or reltionF

e sentene is of

Goldbach type i' it is equivlent to uFgF out pFrF property

QI

yn the sis of thisD we n estlish tht euse G is 1 senteneD we know tht it is of qoldh type nd this serves to tell us thtX

Theorem 16.2

sf P A hs true xiomsD then there is n LA Esentene of qoldh type suh tht neither P A nor P A F

fy knowledging tht our hosen Gdl not only expresses GdlD ut ptures it s wellD we n show the following ftX sf P A is onsistentD then P A

Proof:
uppose tht G is provle in P AD then we know tht there is super gFnF m tht odes its proofF fy y de(nitionD G is the digonliztion of UD so gin y de(niE tion Gdl(m, U )F fut euse Gdl ptures the reltion GdlD we know tht P A Gdl(m, U )F fut sine G is provleD we know tht P A xGdl(x U )F es suhD we know tht if we ssumed tht GD we hve ontrditionD so P A Gdl(m, U )F futD if P A is onsistentD then we know tht there n9t e proof of G in P AF 2 et this pointD we re in nie position to eshew our semnti ssumptions tht P A is sound nd show thtX

sf P A is onsistentD P A

GF

Proof:
uppose tht G is provle in P AF sf G hs proofD then there is super gFnF tht odes itF futD G is the dig of the w' U nd Gdl(m, U )F fut sine Gdl ptures GdlD this implies tht sine Gdl(m, U )D we know tht P A Gdl(m, U )F roweverD sine G is equivlent to xGdl(x, U )D nd y instntition we hve ontrditionF 2

Denition 16.3

en rithmetil theory T is Eincomplete i'D for some open w' (x)D T n prove eh (m) ut doesn9t prove x(x)F en rithmetil theory T is Ecomplete i' it is not EinompleteF sf P A is onsistentD then it is EinompleteF

Denition 16.4 Theorem 16.3


Proof:

essume tht P A is onsistentF yf ourseD then we hve shown tht P A G @P A xGdl(x, U )AF ht isD there is no super gFnF for the proof of GD or there is no super gFnF for the proof of the digonliztion of UF yr ginD for ny mD it isn9t the se tht Gdl(m, U )F oD for eh mD P A Gdl(m, U ) nd this shows tht P A is EinompleteF 2

Denition 16.5

en rithmetil theory T is Einonsistent ifD for some open w' (x)D T n prove eh (m) nd T n lso prove x(x)F

Denition 16.5.1

en rithmetil theory T is Einonsistent ifD for some open w' (x)D T x (x)D yet for eh numer mD we hve T (m)F xyiX Einonsisteny is fully synttilly de(ned propertyD like ordinry onsistenyF QP

Theorem 16.4

sf T is Einonsistent theoryD then T 9s xioms n9t ll e true on stndrd rithmeti interprettionF

e n now give the (nl step in the syntti version of the rgumentX

sf P A is EonsistentD then P A

Proof:
uppose tht P A is Eonsistent nd P A GF his mens tht P A xGdl(x, U )F fut then we know tht Gdl(m, U ) is flse for eh m ndD thus P A Gdl(m, U ) for eh mF his sttement ontrdits our ssumptionD so if P A is EonsistentD then G is unprovleF 2 he results of this hpter tken together mount to proof ofX

Theorem 16.5
PA

here is n LA Esentene of qoldh type suh thtD if P A is onsistentD then Y nd if P A is EonsistentD then P A F

Chapter 17: Gdel's First Theorem


hile we hve shown up until now tht P A is inompleteD this hpter will do the work of showing us tht P A is in n interesting sense incompletable "tht isD we nnot omplete it y simply dding nother xiom y rute foreF ell Denition 4.1X

Denition 4.1

en axiomatized formal theory (AFT) omprises formalized formule from the lnguge whih re axiomsD nd proof theoryF

languageD set of

e know will need to de(ne the further notion of

p.r. axiomatized X

Denition 17.1

e theory T is p.r. axiomatized i' @iA the numeril properties of eing the gFnF of T Ew'GT Esentene re pFrF @iiA the numeril property of eing the gFnF of n xiom is pFrF @iiiA the numeril property of eing the super gFnF of properly onstruted proof is pFrF @ivA the numeril property of eing super gFnF of properly onstruted proof from T 9s xioms is pFrF tooF

Theorem 17.1

sf theory T D whose lnguge inludes LA D is pFrF xiomtizle nd is rithmetiE lly soundD then there is n LA Esentene of qoldh type suh tht T nd T F

his is just the generlized version of heorem ITFI e n get to the hert of the inompletility of rithmetil theories y relizing tht even if we ttempt to omplete n rithmetil theory T D y dding the qdel sentene for T D GT to T to rrive t the new rithmetil theory T + = {T + GT }D T + will still e inompleteF yrD there will e new qdel sentene for the new theoryF he spei( pplition of this ide tht is interesting is tht this mens tht even if we dd G to P AD then there will e some other new true GP A+G whih will show tht P A + G is inompleteF

Denition 17.2 Theorem 17.2


T

e theory T is

nice i' it is onsistentD pFrF xiomtizedD nd extends QF

sf T is nie theory then there is n LA Esentene of qoldh type suh tht nd @if T is lso EonsistentA T F QQ

mith then notes tht heorem IUFP surely doesn9t pture qdel9s (rst theorem euse Q wsn9t isolted for nother PH yersF ht sidD we were interested in Q to egin with euse it ws pFrF dequteF he following is theorem tht hs s muh historil right s ny to e lled qdel9s pirst heoremX

Theorem 17.3

sf T is pFrF dequteD pFrF xiomtized theory whose lnguge inludes LA D then there is n LA Esentene of qoldh type suh tht if T is onsistent then T nd if T is Eonsistent then T F

Chapter 18: Interlude: About the First Theorem


ghpter IV disusses the implitions of psD ut most spei(lly seeks to lrify some misonepE tions out wht is proved with this resultF oD for strtersD wht have we provedc uppose tht we hve n rithmetil theoryD T D whose lnguge inludes LA nd proves ll the truths of LA F purthermoreD it seems intuitive tht we should wnt T to e negtion ompleteF elsoD we should think tht there should e T Eproof for eh of the T theorems"tht isD it should e properly xiomtizle theoryF ps shows tht there n9t e ny suh theoryF gonsider the semnti rgument from erly in the ookX IF ivery pFrF funtion nd reltion n e expressed in LA PF P f rt (m, n) nd GdlT (m, n) re pFrF QF oD y @IA nd @PAD GdlT is expressile in LA Y nd we n onstrut the w' GT D whih is true i' unprovleF RF hereforeD sine GT is true i' unprovle nd T is soundD we n show quikly tht T GT nd T GT F elterntivelyD we n void the semnti route to inompleteness nd tke the syntti routeX I9F essume tht T extends QD then every pFrF reltion n e ptured in T F P9F uppose T is xiomtized nd tht P f rt (m, n) nd GdlT (m, n) re pFrF Q9F e n onstrut GT in the sme wyF R9F e n show tht if T is onsistentD then T T GT F ht doesn9t follow from psc @A por ny nie theory T D T n9t prove it9s qdel sentenes ut we lwys nF9 hile we might know tht GT is trueD there is oviusly no T Eproof of GT F sn ftD we only know this if we know tht T is onsistentD ut there re plusile ses in whih we might not know whether or not T is onsistentF ke theory Q + C D where C is n unproven onjetureF oD sine we don9t know if QC is onsistentD we my hve no lue whether GC 9s qdel sentene is trueF @A here re truths of LA whih re not provle in ny nie xiomtized theoryF9 ell this is oviously flseD euse of ourse y dding GT to nie theoryD T D to form the new theory T + GT will result in nie theoryD nd trivilly in T + GT D GT is provleF QR

GT nd if T is EonsistentD then

@A por some nie theories T D we n prove their qdel sentenes re true even though these truths n9t e derived in forml theory we independently ept s soundF9 his is flse for essentilly the sme reson tht the previous lim ws flseF ell tht sidD we do know tht the set of rithmetil truths is not itself rithmetil"or rithmetil truth isn9t provility in some single xiomtizle systemF

Chapter 19: Strengthening the First Theorem


his hpter will strengthen the inompleteness theorem in the following two wysX IF e n generlize the result from pplying to pFrF xiomtizle to pplying to ny forml theory tht is xiomtized in the intuitive senseF PF e will prove the qdelEosser heoremD whih shows how we n do wy with the ssumpE tion of EonsistenyF qiven qrig9s heoremD qdelin inompleteness infets ny suitle soundGnie xiomtized theE oryD whether it is pFrF xiomtized or not3

Theorem 19.1 Craig's Re-axiomatization Theorem "sf T


Proof:

is n xiomtized theoryD then there is pFrF xiomtized theory T whih hs extly the sme theoremsF sf T is formlized theory then its theorems re eFeF fut onsider the derivtive theory T de(ned s followsX @iA for ehT Etheorem j T hs n xiom (j j . . . j )D there there re n onjunts in the xiom nd n = s(j)F @iiA T 9s sole rule of inferene is elimintionF o oviously every T Etheorem is T theorem nd every T Ethorem is T theoremF qiven n ritrry w' there is proess to tell if it is T xiomF purthermoreD tht proess hs pFrF hrteristi funtion nd then T is pFrF iomtizedF 2

Theorem 19.2

sf theoryD T D whose lnguge inludes LA D is xiomtized nd soundD then there is n LA sentene of qoldh type suh tht T nd T F sf T is onsistent xiomtized theory whih extends QD then there is n LA sentene of qoldh type suh tht T ndD ssuming tht T is EonsistentD T F sf is set of w's whih n e e'etively enumertedD then there is pFrF xiomtized theory T whose theorems re extly the memers of F

Theorem 19.3 Theorem 19.4

Proof:
his is preeise s for

Theorem 19.1F

Theorem 19.5
Proof:

TA D rue fsi erithmetiD is not xiomtizleF

uppose tht T is xiomtizle nd proves extly the sme truths s TA F hen T 9s theorems re ll true so it is sound nd sine its lnguge inludes LA Theorem 19.2 ppliesD so for some LA sentene T nd T F fut of ourse one of them must e true nd so T doesn9t prove ll of the truths in TA F 2 QS

hnks to qeorg ureisel @IWSUAD we know tht we n do without Eonsisteny nd just get y with 1Eonsisteny @or equivlently 1 EsoundnessA for the seond hlf of the ps evenF purthermoreD frkley osser @IWQTA showed tht we n onstrut di'erent nd more omplex senE tene tht only needs to ssume T 9s plin old onsisteny to show tht it is undeidleF hile the qdel sentene sys something like s m unprovle in T 9D the osser senteneD RT sys something like if s m provle in T D then my negtion is lredy provleF9

Theorem 19.6 The Gdel-Rosser Theorem "sf T


of qoldh type suh tht neither T

is nie theoryD then there is n LA sentene nor T F

Proof:
e semnti rgumentX essume tht T is sound nd suppose tht RT ws theoremF hen if RT is provleD RT is lredy provle9 would e trueF purthermoreD we oviously know RT D so then we know RT F e9ve rrived t ontrdition so we9re doneF 2

Denition 19.1
T

(m)F

T is Econsistent if there is no (x) suh tht T

x(x) while for eh mD () while for eh x x

Denition 19.2
mD T

T is (m)F

1-consistent i' there is no 0 w' () suh tht T x

Denition 19.3 Lemma 19.1


Proof:

T is 1 Esound i'D for ny 1 sentene suh tht T

D is trueF

sf T is 1 sound it is onsistentF

sf T is inonsistentD then we n derive nything in T D inluding ny two ontrdiE tory 1 sentenesF o of ourseD if T is 1 soundD then it is onsistentF 2

Lemma 19.2

sf T is nieD then T is 1 Esound i' it is IEonsistentF

Theorem 19.7 Theorem 19.8

sf T is nie 1 Esound theoryD then there is n LA sentene of qoldh type whih is formlly deidle y T F

sf T is nie IEonsistent theoryD then there is n LA sentene of qoldh type whih is formlly undeidle y T F

Chapter 20: The Diagonalization Lemma


Denition 20.1 Denition 20.2 Theorem 20.1
Prf(x, y) stnds for 1 w' whih nonilly ptures P rf F Prof(x) =def v Prf(v, x)F
vet T e nie theoryF hen for ny sentene X

gIF sf T D then T ProvT ( )Y g F uppose T is EonsistentY then if T

ProvT ( )D T

QT

Proof of (C1):
his is ovious given the de(nitionsF fut stillD suppose tht T Y this mens tht there is proof of the w' with the gFnF F o P rfT (m, ) nd sine Prf T ptures P rfT D we know tht PrfT (m, ) ndD thus the existentilly generlized version of the sentene whih turns immeditely into T ProvT ( ). 2

Proof of (C):
essuming tht T is Eonsistent nd tht T ProvT ( )D we n further suppose for redutio tht T F fut then T PrfT (m, ) nd this would mke T Einonsistent ginst the hypothesisF 2 e know tht the qdel sentene for P AD G is true i' it is unprovleF qiven our new provility prediteD we n express this ft inside P AY furthermoreD we n show tht we n tully prove this ft out G in P AF fe refulD thoughD not to think tht this mens nything out the provility of G itselfF gonsider the senteneX

G Prov( G )F
sn ftD we n tully ell tht

prove this sentene inside e333

G =df y(y = U U)
nd

U =df xGdl(x, y)
where Gdl ptures GdlF oD y de(nitionX

Gdl(m, n) =df P rf (m, diag(n))


nd soX

Gdl(x, y) =df z(Pfr(x, z) Diag(y, z))F


his puts us in position to prove the following equivleneX I P Q R S T

U =df xGdl(x, y) U =df xz(Prf(x, z) Diag(y, z)) U zx(Prf(x, z) Diag(y, z)) U z(Diag(y, z) xPrf(x, z)) U =df z(Diag(y, z) Prof(z)) U =df U PA G Prov( G )F
he(nition of Gdl pyv pyv he(nition of Prov xew revition

Theorem 20.2
Proof:

he ove equivlene gives us the result thtX QU

PA
oX

G y(u = ( U U ) G U ( U )Y or G z(Diag( U , z) Prov(z))F G z(z = G Prov(z))Y ut this is justD G Prov( G )F GT ProvT ( GT )F

PA PA PA PA

snterestinglyD this mens thtX

Theorem 20.3
Proof:

sf T is nie theoryD then T

he proof is just s oveF

Theorem 20.4 The Diagonalization Lemma "sf


Proof:

T is nie theory nd (x) is any w' of its lnguge with one free vrileD then there is sentene of T 9s lnguge suh tht T ( )F
he proof strts the sme s the previous ses nd show tht (y) =df z(DiagT (y, z) (z))D where DiagT ptures the digonliztion funtion as a function in T F xext we digonlize nd onstrut D suh tht it is the digonliztion of (y)"or just ( )F xowD we n see the end in sightD for diagT ( ) = nd soD T z(DiagT ( , z) z = )F pinllyD T z(DiagT ( , z) (z)) z(z = (z))F fut then trivillyD T ( ). 2

Denition 20.3
onto itselfF

xed point is point on funtion where the funtion mps n rgument k

Denition 20.4

e xed point theorem is theorem tht sttesD under ertin onditionsD there is (xed point for given funtionF sf T is nie theoryD nd (x) is 1 open w' of its lnguge with one free vrileD then there is 1 (xed point for (x)F

Corrollary 20.1
Proof:

he (xed point for diagT flls out of the higonliztion vemm quite oviouslyF he 1 sentene (y) =df z(DiagT (y, z) (z)) gives us the (xed point for diagT s D or ( )F 2

Corrollary 20.2
Proof:

sf T is nie ut n unsound theoryD nd (x) is ny w' of its lnguge with one free vrileD then (x) hs flse (xed pointF ine T is unsound it will hve flse theoremD ll it F uppose then tht DiagT (x, y) ptures diagT F fut then of ourseD so does DiagT (x, y) =df [DiagT (x, y) ]F hen we rry out with the proof s ove ut onjoin our flse sentene tht we just onstruted nd will e flseF 2

QV

Chapter 21: Using the diagonalization lemma


Theorem 21.1
vet T e nie theoryD nd let e ny (xed point for ProvT (x)F hen T nd if T is EonsistentD then T F

Proof:
his is proved lmost immeditely from gI nd g from the lst hpterF

xow we n proeed to show tht we n even do wy with the ssumption of onsistenyD from our erlier proofsF gonsider the reltionD P rf T (m, n) whih holds when m numers T Eproof of the negtion of the w' with the numer nF his reltion will e pFrF euse P rfT (m, n) isD so the reltion will e pturedD if T is nieD y the w' Prf T (x, y)F xow onsiderX

Denition 21.1

he

Rosser provability predicate is de(ned s followsX

RProvT (x) =df v(PrfT (v, x) (w v)Prf T (w, x))F


he osser provility predite is stis(ed just in se there is no shorter proof of the rgument9s negtionF fut then we n pply the higonliztion vemm nd (nd out tht there is sentene RT tht is the (xed point of RProvT (x) nd ssuming tht T is nieD we know tht T RT RProvT ( RT )F he rest is just the sme old stepsF

Theorem 21.2
nd T

vet T e nie theoryD nd let e ny (xed point for RProvT (x)F hen T F xo open w' in nie theory T n pture the numeril property ProvT F

Theorem 21.3

xow we n puse for moment nd disuss rski9s heorem nd the role tht T rue predite might ply in our systemF

Denition 21.2 Denition 21.3

T rue(n) is true i' n is the gFnF of true sentene of LF

en open L Ew' T(x) is T( ) is trueF e theory T is every L sentene F

formal truth-predicate for L i' for every LEsentene D


T( ) for

Denition 21.4 Theorem 21.4


Proof:

truth-theory for L i' for some L Ew' T(x)D T

xo nie theory n de(ne truth for its own lngugeF

essume tht T de(nes truth for LY sine T is nieD higonliztion vemm pplies nd so if we pply the vemm to the negtion of T(x) we end up with lir sentene3 his is of ourse n immedite ontrdition nd so T n9t de(ne its own truthF2

Theorem 21.5

xo predite of n rithmetilly dequte lnguge L n express the numeril property T rueL F

QW

Chapter 22: Second-order arithmetics


his hpter explores wht hppens if we introdue seondEorder rithmetil priniples nd their implitions on the qdel osser heoremF st seems tht introduing seondEorder priniples might llow us to settle ll of the truths of (rstEorder eF vet us extend LA with seond order qunti(ers to get L2A nd the ompnying interprettion to get L2A

Denition 22.1

e dd the

second-order variables X, Y, Z, . . .

por the semntisD we hve these seondEorder vrile pik out numeril propertiesF xow we introdue the seond order indution xiomD this llows us to get rid of our previous indution shem into univol xiomX

Induction Axiom

X({X0 x(Xx XSx)} xXx)

fut of ourseD we9ll lso need something tht llows us to onnet the indution xiom with the instntitions for properties so we will still need omprehension shemX

Comprehension Schema Denition 22.2

Xx(Xx (x))

e neat formal system of seondEorder rithmeti is n xiomtized theory whih stis(es the following onditionsX IF sts syntx is L2A F PF sts si generl logi is stndrd dedutive system with (rstEorder logil xiomsD ut lso the seondEorder ounterprtsF QF he speil dditionl xioms re the xioms of QD plus the sndution exiomD nd the universl losure of ll the instnes of the gomprehension hemF

Denition 22.3 Second-Order Peano ArithmeticD P A2 D is the strongest net theory in the lnguge
L2A F

Theorem 22.1
Proof:

P A2 n prove every LA w' tht P A n prove nd some tht P A n9t proveF

yf ourse P A2 n prove ll of the truths of P AD ut it n lso prove P A9s qdel senteneD even though it nnot prove its ownF

Denition 22.4

uppose T is theory in the lnguge LF J is n interpretion for the theory T is n interprettion of the syntx L whih keeps the sme interprettions s in I ut interprets the nonElogil voulry of T s wellF sf J mkes ll the xioms of T trueD then J is model for T F wo interprettions re e theory is

Denition 22.5 Denition 22.6 Theorem 22.2 Theorem 22.3


Proof:

isomorphic i' they struturlly look the smeF

categorical i' it hs models ut ll its models re isomorphiF

essuming oth theories hve modelsD P A isn9t tegorilD ut P A2 isF essuming I2A is in ft model for P A2 D P A2 semntilly entils ll L2A EtruthsF

his follows pretty quikly s P A2 is tegorilF RH

fut given thisD we know tht while P A2 G2 D we tully know tht P A2 |=2 G2 "or tht P A2 9s dedutive system is inompleteF xow onsider ertin frgments of P A2 D wherein we wnt to (nd n interprettion for whih the qunti(ers no longer rnge over P(M) ut over sulsses where they stisfy the omprehension xiomF gonsiderX

M = (M, Sm , 0, s, +, )
gonsider ACA0 @ege for the rithmetil omprehension xiomAa@QCindution xiomCegeA ACA0 hs unounded qunti(tion over N ut is onstrutivist out setsF

Theorem

ACA0 is onservtive over P AF sf is (rst order nd ACA0

D then P A

Proof:
eny model of P A n e extended to model of ACA0 F o tkeX

M = (M, Sm , 0, s, +, ) where Sm aolletion of ll of the (rst order de(nle susets of M F M |= ACA0 sf P A D then P A ons hs model MD M |= ACA0 + F

xow onsider 1 CA @the omprehension xiom for 1 sentenesF 1 1 1 CA ny two ountle wellEorderings re ountleY utD 1 ACA0 ny two ountle wellEorderings re ountleF xext we might de(ne weker frgmentD RCA0 suh tht for ny 0 nd 0 : n((n) 1 1 (n)) Xn(Xn (n)) st turns out htt RCA0 hs unique miniml model where Sm = {X|X is reursive} nd tht RCA0 is onservtive over I1 F he following re some interesting results tht RCA0 provesX

RCA0 RCA0 RCA0 RCA0 RCA0 RCA0

soundness the existene of lgeri losure of (elds ompleteness for ll onsistent dedutively losed sets tht hve model reineGforge vemmX ivery overing of [0, 1] y open sets hs (nite suovering wximl rinipleX ivery ontinuous funtion over [0, 1] is ounded unig9s vemmX ivery in(nite ut (nitely rnhing tree hs n in(nite rnhF

xow we will rie)y survey numer of the interesting frgments of seond order rithmeti nd mention some of their fetures @they re listed elow in deresing order of strengthAX prgment 1 CA 1 @erithmetil trns(nite reursionA AT R ACA0 @ek unig9s vemmA W KL0 RCA0 pirstEorder prt essoited view smpreditivism edutive preditivism reditivism edutive (nitism @rilertA trit (nitismGonstrutivism

PA PA I1 I1
RI

Bolzano-Weierstrass Theorem (BW )

ivery oundedGinresing sequene of rels hs limitF

es it turns outD RCA0 n prove the following equivlenesX

RCA0 RCA0 RCA0 RCA0 RCA0 RCA0 RCA0 RCA0

ACA0 BW ACA0 KL ACA0 ivery ounded sequene of rels hs onvergent susequene ACA0 msy9s oloring theorm for [N]3 W KL0 reineGforge W KL0 ivery ontinuous funtion [0, 1] is ounded W KL0 ivery onsistent set hs model AT R0 ivery two ountle well orderings re omprle

Chapter 24: Gdel's Second Incompleteness Theorem


his hpter will disuss the sF e9ll use Prov( ) s sttement tht P A is onsistent @iFeF it doesn9t prove the plseAF oD this motivtesX

Denition 24.1

Con =df Prov( )

Con is of ourse 1 F
he ps sys tht it9s provle in P A tht if P A is onsistent then G is not provle in eF o of ourse given our new mhinery we n express tht with the followingX

PA
elsoD

Con Prov( G )F

Theorem 20.2 tells us tht the following ionditionl n e proved from inside of P AX
PA G Prov( G ) ConF

Theorem 24.1 The Second Incompleteness Theorem "sf P A is onsistentD then P A Corrollary 24.1 Corrollary 24.2 Theorem 24.2
sf T is onsistent suEtheory of P A then T sf T extends P AD then P A

ConP AF

ConT ConF

sf P A is EonsistentD then P A

RP

Vous aimerez peut-être aussi