A proof system is incomplete if and only if there exists a formula that is both unprovable and irrefutable.
Instance Constructor
LO.Entailment.Incomplete.mk.{u_1, u_2, u_5}
Methods
indep : ∃ φ, Independent 𝓢 φ
A deduction system \mathcal{S}
is complete iff it can prove or refute every sentence \sigma
.
Otherwise, \mathcal{S}
is incomplete.
LO.Entailment.Incomplete.{u_1, u_2, u_5} {F : Type u_1} {S : Type u_2} [Entailment F S] [LogicalConnective F] (𝓢 : S) : PropLO.Entailment.Incomplete.{u_1, u_2, u_5} {F : Type u_1} {S : Type u_2} [Entailment F S] [LogicalConnective F] (𝓢 : S) : Prop
A proof system is incomplete if and only if there exists a formula that is both unprovable and irrefutable.
LO.Entailment.Incomplete.mk.{u_1, u_2, u_5}
indep : ∃ φ, Independent 𝓢 φ
Let T
be a \Delta_1
-definable arithmetic theory, stronger than \mathsf{R}_0
and \Sigma_1
-sound.
variable
(T : ArithmeticTheory)
[T.Δ₁] [𝗥₀ ⪯ T] [T.SoundOnHierarchy 𝚺 1]
Let P
be a r.e. set. There exists a formula \varphi_P(x)
such that n \in P \iff T \vdash \varphi_P(\overline{n})
for all n \in \mathbb{N}
.
LO.FirstOrder.Arithmetic.re_complete {T : ArithmeticTheory} [𝗥₀ ⪯ T] [T.SoundOnHierarchy 𝚺 1] {A : ℕ → Prop} (hp : REPred A) {x : ℕ} : A x ↔ T ⊢ (codeOfREPred A)/[↑x]LO.FirstOrder.Arithmetic.re_complete {T : ArithmeticTheory} [𝗥₀ ⪯ T] [T.SoundOnHierarchy 𝚺 1] {A : ℕ → Prop} (hp : REPred A) {x : ℕ} : A x ↔ T ⊢ (codeOfREPred A)/[↑x]
Weak representation of a r.e. predicate
Define a set D_T
, which satisfies \ulcorner\phi\urcorner \in D_T \iff T \vdash \lnot\phi(\ulcorner\phi\urcorner)
.
D_T
is a r.e. set, since the provability predicate of T
is \Sigma_1
-definable,
and the substitution function is \Delta_1
-definable.
def D : ℕ → Prop := fun φ : ℕ ↦
IsSemiformula ℒₒᵣ 1 φ ∧
T.Provable (neg ℒₒᵣ <| subst ℒₒᵣ ?[numeral φ] φ)
lemma D_spec (φ : Semisentence ℒₒᵣ 1) :
D T ⌜φ⌝ ↔ T ⊢ ∼φ/[⌜φ⌝] := T:ArithmeticTheoryinst✝²:Theory.Δ₁ Tinst✝¹:𝗥₀ ⪯ Tinst✝:T.SoundOnHierarchy 𝚺 1φ:Semisentence ℒₒᵣ 1⊢ D T ⌜φ⌝ ↔ T ⊢ ∼φ/[⌜φ⌝]
All goals completed! 🐙
lemma D_re : REPred (D T) := T:ArithmeticTheoryinst✝²:Theory.Δ₁ Tinst✝¹:𝗥₀ ⪯ Tinst✝:T.SoundOnHierarchy 𝚺 1⊢ REPred (D T)
have : 𝚺₁-Predicate fun φ : ℕ ↦
IsSemiformula ℒₒᵣ 1 φ ∧
T.Provable (neg ℒₒᵣ <| subst ℒₒᵣ ?[numeral φ] φ) := T:ArithmeticTheoryinst✝²:Theory.Δ₁ Tinst✝¹:𝗥₀ ⪯ Tinst✝:T.SoundOnHierarchy 𝚺 1⊢ REPred (D T)
All goals completed! 🐙
All goals completed! 🐙
By the representeation theorem, there exists a \Sigma
-sentence \delta_T(x)
,
such that n \in D_T \iff T \vdash \delta_T(n)
for all n \in \mathbb{N}
.
noncomputable def δ : Semisentence ℒₒᵣ 1 :=
codeOfREPred (D T)
lemma δ_spec (n : ℕ) : D T n ↔ T ⊢ (δ T)/[↑n] := T:ArithmeticTheoryinst✝²:Theory.Δ₁ Tinst✝¹:𝗥₀ ⪯ Tinst✝:T.SoundOnHierarchy 𝚺 1n:ℕ⊢ D T n ↔ T ⊢ (δ T)/[↑n]
All goals completed! 🐙
Define sentence \pi_T
, defined as \delta_T(\ulcorner\delta_T\urcorner)
noncomputable def π : Sentence ℒₒᵣ := (δ T)/[⌜δ T⌝]
By a simple inference, it can be shown that the provability of \pi_T
and the provability of its negation are equivalent.
lemma paradoxical : T ⊢ π T ↔ T ⊢ ∼π T := calc
T ⊢ π T ↔ T ⊢ (δ T)/[⌜δ T⌝] := T:ArithmeticTheoryinst✝²:Theory.Δ₁ Tinst✝¹:𝗥₀ ⪯ Tinst✝:T.SoundOnHierarchy 𝚺 1⊢ T ⊢ π T ↔ T ⊢ (δ T)/[⌜δ T⌝] All goals completed! 🐙
_ ↔ D T ⌜δ T⌝ := T:ArithmeticTheoryinst✝²:Theory.Δ₁ Tinst✝¹:𝗥₀ ⪯ Tinst✝:T.SoundOnHierarchy 𝚺 1⊢ T ⊢ (δ T)/[⌜δ T⌝] ↔ D T ⌜δ T⌝ All goals completed! 🐙
_ ↔ T ⊢ ∼(δ T)/[⌜δ T⌝] := D_spec T (δ T)
_ ↔ T ⊢ ∼π T := T:ArithmeticTheoryinst✝²:Theory.Δ₁ Tinst✝¹:𝗥₀ ⪯ Tinst✝:T.SoundOnHierarchy 𝚺 1⊢ T ⊢ ∼(δ T)/[⌜δ T⌝] ↔ T ⊢ ∼π T All goals completed! 🐙
Hence, \pi_T
and its negation are both unprovable from T
, since T
is consistent.
theorem unprovable : T ⊬ π T := T:ArithmeticTheoryinst✝²:Theory.Δ₁ Tinst✝¹:𝗥₀ ⪯ Tinst✝:T.SoundOnHierarchy 𝚺 1⊢ T ⊬ π T
T:ArithmeticTheoryinst✝²:Theory.Δ₁ Tinst✝¹:𝗥₀ ⪯ Tinst✝:T.SoundOnHierarchy 𝚺 1h:T ⊢ π T⊢ False
T:ArithmeticTheoryinst✝²:Theory.Δ₁ Tinst✝¹:𝗥₀ ⪯ Tinst✝:T.SoundOnHierarchy 𝚺 1h:T ⊢ π Tthis:_fvar.2 ⊢ ∼π _fvar.2 := (paradoxical _fvar.2).mp _fvar.178⊢ False
T:ArithmeticTheoryinst✝²:Theory.Δ₁ Tinst✝¹:𝗥₀ ⪯ Tinst✝:T.SoundOnHierarchy 𝚺 1h:T ⊢ π Tthis✝:_fvar.2 ⊢ ∼π _fvar.2 := (paradoxical _fvar.2).mp _fvar.178this:Inconsistent _fvar.2 := inconsistent_of_provable_of_unprovable _fvar.178 _fvar.894⊢ False
All goals completed! 🐙
theorem irrefutable : T ⊬ ∼π T := T:ArithmeticTheoryinst✝²:Theory.Δ₁ Tinst✝¹:𝗥₀ ⪯ Tinst✝:T.SoundOnHierarchy 𝚺 1⊢ T ⊬ ∼π T
T:ArithmeticTheoryinst✝²:Theory.Δ₁ Tinst✝¹:𝗥₀ ⪯ Tinst✝:T.SoundOnHierarchy 𝚺 1h:T ⊢ ∼π T⊢ False
T:ArithmeticTheoryinst✝²:Theory.Δ₁ Tinst✝¹:𝗥₀ ⪯ Tinst✝:T.SoundOnHierarchy 𝚺 1h:T ⊢ ∼π Tthis:_fvar.1523 ⊢ π _fvar.1523 := (paradoxical _fvar.1523).mpr _fvar.2353⊢ False
T:ArithmeticTheoryinst✝²:Theory.Δ₁ Tinst✝¹:𝗥₀ ⪯ Tinst✝:T.SoundOnHierarchy 𝚺 1h:T ⊢ ∼π Tthis✝:_fvar.1523 ⊢ π _fvar.1523 := (paradoxical _fvar.1523).mpr _fvar.2353this:Inconsistent _fvar.1523 := inconsistent_of_provable_of_unprovable _fvar.2415 _fvar.2353⊢ False
All goals completed! 🐙
Therefore, we obtain the Gödel's first incompleteness theorem for \Delta_1
-theory stronger than \mathsf{R_0}
and \Sigma_1
-sound.
LO.FirstOrder.Arithmetic.incomplete (T : ArithmeticTheory) [Theory.Δ₁ T] [𝗥₀ ⪯ T] [T.SoundOnHierarchy 𝚺 1] : Incomplete TLO.FirstOrder.Arithmetic.incomplete (T : ArithmeticTheory) [Theory.Δ₁ T] [𝗥₀ ⪯ T] [T.SoundOnHierarchy 𝚺 1] : Incomplete T
Gödel's first incompleteness theorem