2. If B is a regular set, then B ∩ V(A) is a set.Proof: (1): Let x, y ∈ B. Then f : 2 → B, where f(0) = x, f(1) = y. Hence, byregularity of B, the range of f is in B, that is {x, y} ∈ B.As 〈x, y〉 = {{x}, {x, y}}, 〈x, y〉 ∈ B follows from closure under unordered pairs.(ii): To see this let κ = rank(B), where the function rank is defined by rank(x) :=⋃ {rank(y)+1 : y ∈ x} with z +1 := z ∪{z}. One easily shows that <strong>for</strong> all sets x, rank(x)is an ordinal. Let Φ be the inductive definition withxΦ iff ∀u∈a (u ∈ |A| × x).aInvoking Lemma 3.4, let J be the class such that V(A) = ⋃ α J α , and <strong>for</strong> each α,J α = Γ Φ ( ⋃ β∈αJ β ).Moreover, define the operation Υ byΥ(X) := {u ∈ B : u ⊆ |A| × X}and by recursion on α setLetΥ α = Υ( ⋃ β∈αΥ β ).Υ
Theorem: 6.2 For every axiom θ of CZF + REA, there exists a closed application termt such thatCZF + REA ⊢ (t ⊩ θ).Proof: In view of theorem 5.1, we need only find a realizer <strong>for</strong> the axiom REA. Leta ∈ V(A). By REA there exists a regular set B such that a, 2, |A| ∈ B. LetA := B ∩ V(A),C := {〈0, x〉 : x ∈ A}.By Lemma 6.1, A is a set; hence C is a set. Moreover, as A ⊆ V(A), it follows C ∈ V(A)andp0i r ⊩ a ∈ C. (23)With ˜m := λx.λy.p0i r and ñ := p0(p0i r ) one realizes transitivity and inhabitedness ofC, respectively, i.e.,p ˜mñ ⊩ ∀u ∈ C ∀v ∈ u v ∈ C ∧ ∃x ∈ C x ∈ C. (24)Next, we would like to find a realizer q such thatq ⊩ Reg(C). (25)To this end, let b ∈ V(A), f ∈ |A|, and ϕ(x, y) be a <strong>for</strong>mula with parameters in V(A)satisfyingThen there exists d such thatf ⊩ b ∈ C ∧ ∀x ∈ b ∃y ∈ C ϕ(x, y). (26)〈f 0,0 , d〉 ∈ C ∧ f 0,1 ⊩ b = d, (27)f 1 ⊩ ∀x ∈ b ∃y ∈ C ϕ(x, y), (28)where f 0,0 := ((f) 0 ) 0 , f 0,1 := ((f) 0 ) 1 , and f 1 := (f) 1 . (27) and (28) yieldi ψ f 0,1 f 1 ⊩ ∀x ∈ d ∃y ∈ C ϕ(x, y), (29)where i ψ ⊩ ∀u∀v[u = v ∧ ψ(u) → ψ(v)], with ψ(u) being ∀x ∈ u ∃y ∈ C ϕ(x, y) (accordingto Lemma 4.2, i ψ is independent of C and the parameters in ϕ).Since B is closed under ordered pairs (Lemma 6.1) and |A| ∈ B, from (29) we get∀x∀e [〈e, x〉 ∈ d → ∃z ∈ B ∃y(z = 〈e, y〉 ∧ 〈( ˜fe) 0 , y〉 ∈ C ∧ ( ˜fe) 1 ⊩ ϕ(x, y))], (30)where ˜f := i ψ f 0,1 f 1 . Noting that 〈v, y〉 ∈ C entails v = 0, and utilizing the regularity ofB, there exists u ∈ B such that∀x∀e [〈e, x〉 ∈ d → ∃z ∈ u ∃y(z = 〈e, y〉 ∧ 〈0, y〉 ∈ C ∧ ( ˜fe) 1 ⊩ ϕ(x, y))]; (31)∀z ∈ u ∃x, e [〈e, x〉 ∈ d ∧ ∃y(〈0, y〉 ∈ C ∧ z = 〈e, y〉 ∧ ( ˜fe) 1 ⊩ ϕ(x, y))]. (32)19