要证明之前用的逻辑推理规则是合理的。

Theorem 1.

ϕ[xt]xϕ\phi[x\mapsto t]\models\exists x\phi

where ϕ[xt]\phi [ x\mapsto t] means substituting every free occurrence ofof xx in ϕ\phi with tt.

Example2.
Theorem1 的例子:

P(c)x(P(x))P(c)\vDash\exists x(P(x))

考虑一个更复杂一点的例子:

P(c)Q(g(c),c)x(P(x)Q(g(x),c))P(c)\wedge Q(g(c),c)\models\exists x(P(x)\wedge Q(g(x),c))

In order to prove theorem 1, we need to use the following property.
因为 Theorem1 引入了一个新的符号 ϕ[xt]\phi[x \mapsto t],这个符号表示在语法上把 ϕ\phi 中的 xx 替换成 tt。(不包括 x,x\forall x, \exists x 包括到的这些 xx

Claim 3.

ϕ[xt]J=ϕJ[xtJ]\llbracket \phi[x \mapsto t] \rrbracket_{\mathcal{J}} = \llbracket \phi \rrbracket_{\mathcal{J}[x \mapsto \llbracket t \rrbracket_{\mathcal{J}}]}

Claim3 可能有点复杂,再举两个例子:

Example 4.

  • Let ϕ=P(x)\phi = P(x) and t=ct=c. Then, P(c)J=P(x)J[xJ(c)]\llbracket P(c) \rrbracket_{\mathcal{J}} = \llbracket P(x) \rrbracket_{\mathcal{J}[x \mapsto \mathcal{J}(c)]}
  • Let ϕ=P(x)\phi = P(x) and t=f(y,c)t=f(y,c). Then, P(f(y,c))J=P(x)J[xf(y,c)J]\llbracket P(f(y,c)) \rrbracket_{\mathcal{J}} = \llbracket P(x) \rrbracket_{\mathcal{J}[x \mapsto \llbracket f(y,c) \rrbracket_{\mathcal{J}}]}.

上面的 Claim3 直接作为结论来用,不做证明。然后使用 Claim3 来证明 Theorem1:
Proof
For any J\mathcal{J}, if ϕtJ=T\llbracket \phi \mapsto t \rrbracket_{\mathcal{J}} = T, then ϕJ[xtJ]=T\llbracket \phi \rrbracket_{\mathcal{J}[x\mapsto \llbracket t \rrbracket_{\mathcal{J}}]} = T, so xϕJ=T\llbracket \exists x\phi \rrbracket_{\mathcal{J}} = T.
Qed.

再考虑一个类似的结论,同样可以用 Claim3 证明。
Theorem 5.

xϕϕ[xt]\forall x \phi \models \phi[x\mapsto t]

Proof
If J\mathcal{J} is an interpretation, xϕJ=T\llbracket \forall x\phi \rrbracket_{\mathcal{J}} = T, then for any aa in J\mathcal{J}'s domain, ϕxa=T\llbracket \phi \rrbracket_{x\mapsto a} = T.
ϕ[xt]J=ϕJ[xtJ]=T\llbracket \phi[x\mapsto t] \rrbracket_{\mathcal{J}} = \llbracket \phi \rrbracket_{\mathcal{J}\left[x\mapsto \llbracket t \rrbracket_{\mathcal{J}} \right]} = T.
Qed.

在之前的逻辑推理中,用过这样的结论:
If Φ,ψχ\Phi, \psi \vdash \chi and xx not mentioned in Φ\Phi, then Φx(ψχ)\Phi \vdash \forall x(\psi \to \chi).
上述结论可以抽象为如下结论

Theorem 7.

  • If Φψ\Phi \models \psi and xx does not freely occur in Φ\Phi, then Φxψ\Phi \models \forall x \psi

Remark: “does not freely occur” is very important.

举个例子:
Φ={P(x)},ψ=P(x)\Phi = \{ P(x) \}, \psi = P(x),那么有 Φψ\Phi \models \psi,但是 Φ⊭xψ\Phi \not \models \forall x \psi

想要证明 Theorem7,需要引入以下 Claim

Claim9
If Φ\Phi does not mention xx

ϕJ=ϕJ[xa]\llbracket \phi \rrbracket_{\mathcal{J}} = \llbracket \phi \rrbracket_{\mathcal{J}[x\mapsto a]}

接下来用 Claim9 证明 Theorem7
Proof
For any J\mathcal{J} s.t. for any ϕΦ,ϕJ=T\phi \in \Phi, \llbracket \phi \rrbracket_{\mathcal{J}} = T.
For any aa in Js\mathcal{J}'s domain, and any ϕΦ\phi \in \Phi, ϕJ[xa]=ϕJ=T\llbracket \phi \rrbracket_{\mathcal{J}[x\mapsto a]} = \llbracket \phi \rrbracket_{\mathcal{J}} = T
So ψJ[xa]=T\llbracket \psi \rrbracket_{\mathcal{J}[x\mapsto a]} = T (Φψ\Phi \models \psi), xψJ=T\llbracket \forall x \psi \rrbracket_{\mathcal{J}} = T
Qed.

Theorem7
Theorem7 还有一个对称的说法

  • If Φ,ψχ\Phi, \psi \models \chi, and xx is not mentioned in Φ\Phi and χ\chi, then Φ,xψχ\Phi, \exists x \psi \models \chi

Proof
Suppose J\mathcal{J} is an interpretation s.t. (1) For any ϕΦ\phi \in \Phi, ϕJ=T\llbracket \phi \rrbracket_{\mathcal{J}} = T; (2) xψJ=T\llbracket \exists x \psi \rrbracket_{\mathcal{J}} = T
So there exists aa s.t. ψJ[xa]\llbracket \psi \rrbracket_{\mathcal{J}[x\mapsto a]}
For any ϕΦ\phi \in \Phi, ϕJ[xa]=ϕJ=T\llbracket \phi \rrbracket_{\mathcal{J}[x\mapsto a]} = \llbracket \phi \rrbracket_{\mathcal{J}} = T
Thus χJ[xa]=T\llbracket \chi \rrbracket_{\mathcal{J}[x\mapsto a]} = T (because Φ,ψχ\Phi, \psi \models \chi)
χJ=χJ[xa]=T\llbracket \chi \rrbracket_{\mathcal{J}} = \llbracket \chi \rrbracket_{\mathcal{J}[x\mapsto a]} = T.