If a0 is a fixed element of and t is a term of L, then
Let a1,...,an be any elements of . We will show that
(recall that F (a1,...,an) = and IIF = IIi1 ... inF).8 Since for every m N, fm fm+1, and for every a , there exists m such that a ran(fm), it is evident that there exists m such that a1,...,an ran(fm). Suppose that m is the least number such that a1,...,an ran(fm). We now observe that the terms of F (a1,...,an) are a1,...,an, the free variables of IIF (i.e., the elements of J0) and the terms generated by the fleeing indices of IIF when i1,...,in take the values a1,...,an, respectively. Thus, with the exception of a1,....,an, all the terms of F (a1,....,an) belong to dom(fm+1).
Let v0 be an assignment in D such that
(the construction of the sequence f0, f1,... guarantees the existence of V0). Clearly,
It easy to see that the assignments v0 and agree on the terms of F (a1,...,an). Let t be a term of F (a1,...,an). If t ,
if t dom(fm+1),
By Lemma A.4,
This show that
In order to conclude the proof, it remains only to show that is denumerable. Obviously, is countable because it is the union of a denumerable sequence of finite sets. If were finite, IIF would be finitely satisfiable, contradicting the hypothesis of the theorem. Thus, is denumerable.
1If the variables on which the fleeing term depends are existentially quantified, as is the case for example in ΣiA(i, Ki), the problem of correctness does not arise, because this type of formula is the result of putting the universal quantifiers in front of the existential quantifiers:
2Rv(t1) ... v(tn) is the sentence that results from replacing in Rt1 ...tn each m (1 ≤ m ≤ n) with V(tm).
3Observe that if A is a quantifier-free formula, then
4Recall that (3.11) states the following:
ΣkiIIiA(i, ki) is true with S in D iff there is an indexed family (Ka | a D) of elements of D such that for all a D, A[a, ka] is true with S in D.
5In fact, when in the previous chapters I speak of terms generated by a fleeing index Ki1...in, I mean the terms of this set. I have introduced a more general definition than is required by the proof of Löwenheim's theorem for the sake of the lemmas in the next subsection.
6The existential quantifiers occurring in the scope of a universal quantifier are removed because quantification on fleeing terms is not allowed in L.
7The sequence f0, f1, f2,... corresponds to the sequence of formulas Q1,Q2,Q3,... constructed by Löwenheim, but my proof of the existence of the sequence is different from Löwenheim's. If the levels were finite (as is the case in Löwenheim's proof), it would be enough to prove that for every n, Vn , in order to be sure that at each level there must exist a partial assignment having infinitely many extensions. This fact would enable us to show (reproducing the proof of the infinity lemma) the existence of a sequence f0, f1, f2,... such that for every n, fn Vn and fn fn+1. Essentially, this is how Löwenheim proceeds in his construction of the sequence of formulas.
It can be proven that for every n N, fn fn+1, and perhaps it would be closer to Löwenheim's argument to do so, but then my proof would become a little repetitive. Rather schematically, to show that for every n N, fn fn+1, we have to begin by observing that if there exists n such that fn = fn+1, then for every m > n, fm = fn. This means that ran(fn) is a closed finite subdomain, i.e., a subdomain of D such that the terms generated by the fleeing indices of IIF in ran( fn) takes values in ran(fn). We would prove then that IIF is satisfiable in ran(g), contradicting the hypothesis of the theorem. This fact is proven in the same way in which I will show that IIF is satisfiable in .
8The assignment may not treat the fleeing terms of L functionally. In addition, the formulas of Löwenheim's sequence may not be true under (, ). However, if j1,...,jm are the free variables of IIF and tx is the fleeing index replaced by the variable jx in the construction of Löwenheim's sequence, the assignment for L in defined by|
is such that (, v2) satisfies all the formulas of Löwenheim's sequence. This proves that if IIF is satisfiable, then for every n N, Pn is satisfiable.
3.148.144.100