Lecture 9 (23.12.96)
Previous lecture:
Lecture 8 ,
Overview ,
Course home page ,
Material of this and subsequent lectures:
- [EFT]
H.D. Ebbinghaus, J. Flum and W. Thomas, Mathematical Logic, 2nd
ed., Springer 1994
The Lindstrom Theorems
Detailed definition of
- A logic $L = < Sent(\tau), Mean(\phi, A, z) >$
with closure conditions:
- Isomorphism condition
- Atmic formulas
- Closure under boolean operations
- Closure under existential quantification
- Closure under substitution of relation symbols
by formulas
- Closure under $\Phi^#$ where $\Phi$ is a translation
scheme using formulas from $Sent$.
- The compactnes property
- The Lowenheim-Skolem (L-S) Property for countable sets
of formulas
- Examples of logics which are not compact:
$FOL(Q_0)$ with the quantifier "there are infinitely many".
- Examples of logics without the L-S-Property:
$FOL(Q_1)$ with the quantifier "there are uncountably many".
Lindstrom's Theorem (1)
If a logic is compact and has the L-S Property,
then it is of the same expressive power as FOL.
Proof of the theorem
- Assume that each formula depends only on
finite part of the vocabulary. It can be proved that this
follows from compactness.
- Assume there is $\phi \in Sent(\tau)$ not
equivalent to a first order formula.
- Use compactness again to show that there are
$\tau$-structures
$\cA \models \phi$ and $\cB \models \neg\phi$
such that for each
n player II has a winning strategy in the FOL-game
of lenth n over $\cA$ and $\cB$.
- Build `rabbit'-structure $[\cA, \cB, \cN, E_k]$
which encodes Ehrenfeucht game and define countable
theory T_1 on this structure.
- Find countable non-standard model of T_1 (with respect to
$\bN$) $[\cA_0, \cB_0, \cN_0, E_k]$
- Construct isomorphism between
$\cA_$ and $\cB_0$ using non-standard length of the game.
- This contradicts the fact that
$\cA_0 \models \phi$ and $\cB_0 \models \neg\phi$.
Lindstrom's Theorem (2)
If a logic has a recursive set of sentences, has recursive enumerable
validities and has the L-S Property,
then it is of the same expressive power as FOL.
Without proof.
Next lecture:
Lecture 10