Model Theory Cheat Sheet
Model Theory Cheat Sheet
Contents
1 Languages and Structures 1
1.1 Languages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Deductions 4
2.1 Intermission: Propositional Logic . . . . . . . . . . . . . . . . 4
2.2 Our Proof System . . . . . . . . . . . . . . . . . . . . . . . . 5
4 Changing Models 7
4.1 (Elementary) Substructures, and the Tarski-Vaught Test . . . 7
4.2 Löwenheim-Skolem Theorems . . . . . . . . . . . . . . . . . . 8
4.3 Complete Theories, Categoricity, and Back-and-Forth . . . . . 8
4.4 Denability . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
4.5 Quantier Elimination and Model-Completeness . . . . . . . . 9
4.6 Ultraproducts, o±' Theorem and Some Applications . . . . . 10
1.1 Languages
Language: A language L is a collection of symbols, which are separated into the following
categories:
1. Logical symbols
2. Punctuation symbols
Because the rst three are obligatory, you only give the symbols from the last three cate-
gories, if you have any, when specifying a language.
1
MATH406 Cheat Sheet O§ul Can Yurdakul
Term: For a given language L, the terms of that language T ermL are dened recursively as
follows:
TI If t1 , t2 , ...tn are terms and f is an n-ary function, then f (t1 , t2 , ..., tn ) is also a term.
Formulas: For a given language L, the formulas of that language F ormL are dened recursively
as follows:
- and show it for a non-atomic formula, assuming that the claim holds for the non-
atomic formula's constituents.
Freedom: Given x a variable and ϕ a formula, we say that x is free in ϕ in the following cases:
2 Contents
MATH406 Cheat Sheet O§ul Can Yurdakul
- x is not free in ϕ, or
1.2 Structures
Fix a language L.
Variable and Term Assignment Maps: Let A be a structure. A variable assignment func-
tion is a mapping s : V arsA → A.
The modied variable assignment function s[x|a] for some a ∈ A and x ∈ V arsA is dened
as (
s(v) , v 6= x
s[x|a](v) =
a ,v = a
The modied variable assignment function is useful in instantiating a variable to a specic
value.
The term assignment function s̄ : T ermL → A generated by s is dened recursively as
follows:
- ϕ is of the form t=u and s̄(t) = s̄(u), where t and u are terms.
- ϕ is of the form r(t1 , t2 , ..., tn ) and (s̄(t1 ), s̄(t2 ), ..., s̄(tn )) ∈ rA , where t1 , t2 , ..., tn are
terms.
3 Contents
MATH406 Cheat Sheet O§ul Can Yurdakul
In case a formula ϕ is satised by a structure A under any variable assignment map, then
we write A |= ϕ.
Theory: A theory is a set of sentences Σ ⊆ F ormL . In case a structure A satises every sentence
of the theory Σ, we write A |= Σ.
Theory of a Structure: Let A be a structure. Its theory T hm(A) is dened as the set of
sentences it satises.
Isomorphism of Structures: Let A and B be two structures. We say that they are isomor-
phic, A∼
= B, if there is a bijection j : A → B such that
2 Deductions
Given a formula ϕ ∈ F ormL in rst order logic, one can obtain its propositional counterpart
ϕ̃ easily.
Truth Assignment: A truth assignment map for propositional variables v : P ropV ars →
{0, 1} can be easily extended to a map for propositional formulas v̄ : P ropF orm → {0, 1}
4 Contents
MATH406 Cheat Sheet O§ul Can Yurdakul
as follows:
v(ϕ) if ϕ is atomic
1 − v(ψ) if ϕ is of the form ¬ψ
v(ψ)v(γ) if ϕ is of the form ψ∧γ
v̄(ϕ) =
d(v(ψ) + v(γ))/3e if ϕ is of the form ψ∨γ
d((1 − v(ψ)) + v(γ))/3e if ϕ is of the form ψ→γ
1 if ϕ is of the form ψ↔γ and v(ψ) = v(γ)
^
ψ̃ → ϕ̃
ψ̃∈Γ̃
Given Γ ⊆ F ormL and ϕ ∈ F ormL , one can easily extend the propositional implication
denition here, by obtaining Γ̃ and ϕ̃.
Fix a language L.
Proofs and Proof Systems: Let Σ, Λ ⊆ F ormL . A nite sequence of L-formulas (ϕ1 , ϕ2 , ..., ϕn , Φ)
is said to be a deduction/proof of Φ, Σ ` Φ, if for all ϕi we have
For a dierent selection of Σ, Λ and inference rules, we obtain a dierent proof system.
Our Logical Axioms: Our logical axioms Λ are of two types:
- For all variables x1 , x2 , ... xn , y1 , y2 , ... yn and an n-ary relation symbol r, the
following is an axiom:
5 Contents
MATH406 Cheat Sheet O§ul Can Yurdakul
QR Quantier Rules: For any two formulas ϕ and ψ and variable symbol x not free
in ψ, we have
- Σ ⊆ C,
- Λ ⊆ C, and
`x=x
`x=y→y=x
` (x = y ∧ y = z) → x = z
Universal Closure: Σ ` θ if and only if Σ ` ∀xθ.
Deduction Theorem: Let Σ ⊆ F ormL and θ ∈ SentL . Then Σ ∪ {θ} ` ϕ if and only if
Σ ` (θ → ϕ).
3.1 Completeness
Consistency and Inconsistency: Let Σ ⊆ F ormL . Σ is said to be inconsistent if Σ ` (ϕ∧¬ϕ)
for some ϕ. Σ is said to be consistent if it is not inconsistent, in which case we write
6 Contents
MATH406 Cheat Sheet O§ul Can Yurdakul
Cons(Σ).
An inconsistent Σ can prove any ϕ ∈ F ormL .
Σ is said to be maximally consistent if it contains all formulas ϕ which does not make it
inconsistent, i.e. any addition to it from F ormL − Σ makes it inconsistent.
This theorem is important, as it ties the semantic notion of satisfaction to the syntactic
notion of provability. Combined with the Soundness Theorem, we obtain
3.2 Compactness
Satisability: Σ ∈ F ormL is said to be satisable if it has a model.
Finite Satisability: Σ ∈ F ormL is said to be nitely satisable if every nite subset of Σ has
a model.
4 Changing Models
- A ⊆ B,
- For every constant symbol c ∈ CL , we have cA = cB .
- For every n-ary relation symbol r ∈ RL , we have rA = rB ∩ An , that is, for every
(a1 , a2 , . . . , an ) ∈ An , rA (a1 , a2 , . . . , an ) if and only if rB (a1 , a2 , . . . , an )
- For every n-ary function symbol f ∈ FL , we have aA = f B An , that is, for every
(a1 , a2 , . . . , an ) ∈ An , f A (a1 , a2 , . . . , an )
= f B (a1 , a2 , . . . , an )
Elementary Substructure: Let A and B be two models such that A ⊆ B. We say that A is
an elementary substructure of B, written A B if for every formula ϕ(x̄) and ā ∈ A, we
have A |= ϕ(ā) if and only if B |= ϕ(ā).
- π is one-to-one,
- For all n-ary relation symbols r and ā ∈ A, we have rA (ā) if an only if rB (π(ā)),
- For all n-ary function symbols f and ā ∈ A, we have π(f A (ā)) = f B (π(ā)).
7 Contents
MATH406 Cheat Sheet O§ul Can Yurdakul
Tarski-Vaught Test: LEt A ⊆ B be two L-structures. Suppose for every formula ϕ(x̄, y) and
for every ā ∈ A and b ∈ B , if B |= ϕ(ā, b) then there is some d ∈ A with B |= ϕ(ā, d).
Then A B.
It is very dicult to make use of the Tarski-Vaught test on its own. For us, it is mostly
important in the proofs of the Löwenheim-Skolem theorems.
o±Vaught Test: Let Σ be κ-categorical for some κ ≥ max{ℵ0 , |L|} and suppose that Σ does
not have nite models. Then Σ is complete.
8 Contents
MATH406 Cheat Sheet O§ul Can Yurdakul
Back-and-forth Method: The back-and-forth method is a method of proof. Here, it was used
to prove that DLOWE is ℵ0 -categorical. In the proof, to show categoricity, one must
construct an isomorphism between two random DLOWE structures. One starts with two
singletons from each of the two models, and goes 'back and forth' between these two by
adding elements to both of them while still conserving the isomorphism.
DLOWE: Let L = {<}. The theory of Dense Linear Orders Without Endpoints (DLOWE) is
a theory of this language as given below.
DLOWE clearly has no nite models. Being ℵ0 -categorical, by o±Vaught test, it must
be complete.
4.4 Denability
Denability: A be an L-structure. A subset X ⊆ An is called S -denable if there exists
Let
an L-formula ϕ(x̄, ȳ) and b̄ ∈ S ⊆ A such that X = {ā : A |= ϕ(ā, b̄)}.
Here the set S is a set of parameters to dene X . If we let S = ∅, we call such sets X as
denable without parameters or just denable.
Principle DLO Formulas: A principle DLO formula for the variables x1 , x2 , ..., xn is a formula
ψ(x1 , x2 , ..., xn ) of the form
- If n = 1, ψ(x1 ) is x1 = x1 .
9 Contents
MATH406 Cheat Sheet O§ul Can Yurdakul
n−1
V
- If n > 1, ψ(x1 , x2 , ..., xn ) is xσ(i) i xσ(i+1) where σ ∈ Sym{1, 2, ..., n} is a
i=1
permutation and i is either < or =.
A principle DLO formula basically tells the conguration of a certain number of elements.
The job of σ is to tell which two variables are related consecutively and the job of i at
each i-th step is to determine the relation between the two elements.
A Lemma: Let ψ(x̄) be a principle DLO formula and ā, b̄ ∈ Qn such that (Q, <) |=
ψ(ā) ∧ ψ(b̄). Then there exists f ∈ Aut(Q, <) with f (ā) = b̄.
This means that for the same specic conguration of certain elements (as set by ψ ),
one can transform two groups of numbers satisfying that condition to the other via some
automorphism f. Think of f as a collection of rules that stretches and translates the
rational line to `tailor' one group of number to the other.
Important: (Q, <) admits quantier elimination.
Quantier Elimination for Theories: A theory T is said to admit quantier elimination if for
every formula ϕ, there exists a quantier-free formula ψ such that T ` ϕ ↔ ψ (T |= ϕ ↔ ψ ).
Important: DLO admits quantier elimination.
- ∅ 6∈ F and X∈F
- For all A, B ∈ F , A ∩ B ∈ F
- For all A∈F and A ⊆ B ⊆ X, B ∈ F
A lter sort of `chooses' or `prefers' certain subsets of X. But we can think of a lter that
does not include both some subset Y and its complement Y C. Then is Y `preferred' by
this lter or not?
Filters to Ultralters: For every lter F ⊆ P(X), there is an ultralter U such that F ⊆ U.
Ultraproduct: Let L be a language with constant symbols CL , relation symbols RL and func-
tion symbols FL . Let I U ⊆ P(I)
be an index set and an ultralter over the index set. Let
{Mi |i ∈ I} be a set of L-structures indexed
Q by I .
Dene an equivalence relation ∼ on Mi given by
i∈I
10 Contents
MATH406 Cheat Sheet O§ul Can Yurdakul
That is to say that the sequences (ai )i∈I and (bi )i∈I are said to be related if and only
if they agree on a large portion of their indices where the bigness is dened using the
ultralter. Or in other words, they are equal if and only if they agree on the preferable
indices. Using this equivalence relation, dene the set M as
Y
M= Mi / ∼
i∈I
This means that a group of sequences are related under r if and only if the indices at
which the tuples formed by the elements of each sequence with that index, is preferred
according to U.
For each n-ary function symbol f ∈ FL , dene f M by
f M (xi )1i∈I , (xi )2i∈I , ..., (xi )ni∈I = f Mi (x1i , x2i , ..., xni )
That is to say that the image of a group of sequences under f is the series congruent to
the series made by the image of the function in each structure. This function is proved to
be well-dened.
Then nally, the L-structure
!
Y
M M M
M= Mi / ∼, {f }f ∈FL , {r }r∈RL , {c }c∈CL
i∈I
o±' Theorem: ϕ(x1 , x2 , ..., xn ) be an L-formula and (ai )1i∈I , (ai )2i∈I , ..., (ai )ni∈I ∈
Q
Let Mi / ∼.
i∈I
Then
!
Y
Mi /U |= ϕ (ai )1i∈I , (ai )2i∈I , ..., (ai )ni∈I
i∈I
if and only if
11 Contents