Take a
language L and
L-
theory T and assume for all
structures M,N we have
M⊆N∧N⊨T→M⊨T.
We want to find a universal axiomatization for
T. Let’s try to do this by letting
Γ be the
set of all
universal formulas implied by
T. Conceptually, we are defining
Γ to be the largest axiomatization that perhaps might work.
Explicitly,
Γ:={ϕ a sentence∣ϕ universal ∧T⊨ϕ}
By construction
T⊨Γ. For
Γ to be an axiomatization of
T, then, we also need to show that
Γ⊨T. To do that, we must show that each
M modelling Γ also
models T.
So take some
M⊨Γ. Now we are going to take some unmotivated steps.
First we will show that
T∪Diag(M) is
finitely satisfiable.
1 Take
finite Δ⊆T∪Diag(M). Then we can decompose
Δ as
Δ=T0∪Φ0 with
T0 and
Φ0 both
finite and
T0⊆T and
Φ0⊆Diag(M). Since
Φ0 is a
finite, we can form the conjunction
ϕ=ϕ(a) of all
sentences in
Φ0. Then
ϕ is also in
Diag(M),
It’s worth noting that
T is a
set of
L-
sentences but
Diag(M) is a
set of
LM-
sentences.
so
M⊨ϕ(m) for some
m∈M<ω
so
M⊭∀v ¬ϕ(v)
so
Γ⊭∀v ¬ϕ(v) since
M⊨Γ
so
T⊭∀v ¬ϕ(v) since
T⊨Γ (isn’t this wrong?????)
so
T∪{∃v ϕ(v)} is
satisfiable
so
T∪{ϕ(a)} is
satisfiable, by choosing interpretations for elements of
a to be witnesses of
∃v ϕ(v)
so
T0∪Φ0 is
satisfiable
so
Δ is
satisfiable
Hence
T∪Diag(M) is
finitely satisfiable.
By
compactness,
T∪Diag(M) is also
satisfiable; let
N be a
model. Then
N⊨T and
N⊨Diag(M). By properties of
diagrams, since
N⊨Diag(M) then there exists an
embedding j:M→N−, where
N− is the
L-
reduct of
N. Hence
j(M)⊆N−⊨T, so discharging the proof assumption we know
j(M)⊨T. And since
M≅j(M) then
M⊨T. Done!