Course

Home

Syllabus

Lecture Topics

Homework

Policies




Math 6000: Model Theory, Spring 2024


Lecture Topics


Date
What we discussed/How we spent our time
Jan 17
Syllabus. Text. HW. We introduced the concept of an abstract logic $(\mathcal{K},\Sigma,\models)$ and indicated that a first-order logic can be generated from an appropriate alphabet of nonlogical symbols.
Jan 19
Room change! Chem 145 flooded. We will meet in Math 350 when we return to in-person meetings.

We followed these slides, which identify the main ingredients of Model Theory. Then we started discussing this handout concerning the construction of first-order formulas. (We only covered up to the definition of atomic formulas.)

Jan 22
We will meet in Math 350 starting Wednesday January 24.

We finished these slides.

Jan 24
We defined (algebraic) closure operator and Galois connection. The main result on Galois connections is Theorem 5 of this handout.

The Galois connection that is of importance for us is the GC between syntax and semantics for a first-order language $L$. Given a set $U$ of $L$-sentences, $U^{\perp}$ is the class of all $L$-structures that satisfy $U$. Given a class $X$ of $L$-structures, $X^{\perp}$ is the set of all $L$-sentences true throughout the class $X$.

Jan 26
We briefly described the Galois Connection of Galois Theory and then the GC of Algebraic Geometry. I identified three 'projects' one has once an GC has been identified:

  • Give an internal description of Galois Closure. (E.g., Hilbert's Nullstellensatz, Gödel's Completeness Theorem)
  • Describe the lattice of Galois-closed subsets.
  • Exploit the Galois Correspondence.
  • We discussed the first four frames of these slides. We circulated (but did not yet discuss) this handout.

    Jan 29
    We got to the middle of slide 9 of these slides. This involved discussing the Deduction Theorem, Lindenbaum's Theorem, and Henkin theories.
    Jan 31
    We finished the proof of the Completeness Theorem. The proof showed that if $\Sigma$ is a consistent set of sentences in a signature of size $\kappa$, then $\Sigma$ has a model of size at most $\kappa+\omega$.
    Feb 2
    We discussed the first HW assignment briefly, then had a vocabulary review. Finally, we discussed the Compactness Theorem and some of its applications.
    Feb 5
    We discussed the 4th application of the Compactness Theorem. Then we discussed the space of complete theories in a given signature.
    Feb 7
    As an exercise, we reconstructed a finite topological space from its frame of open sets. Then we completed the slides on the space of complete theories in a given signature.
    Feb 9
    We started discussing homomorphisms and constructions of new models following these slides.
    Feb 12
    We discussed embeddings, congruences, quotient structures and product structures following these slides.
    At one point I posed the problem of finding all the quotients (up to isomorphism) of the symmetric graph $\langle V ; E(x, y)\rangle$ that is a 4-element path. We calculated in class that there are 15 equivalence relations on a 4-element set and computed some of the quotients with respect to these equivalence relations. The complete list of isomorphism types of quotients should contain 1 graph of size 4, 4 graphs of size 3, 3 graphs of size 2, and 1 graph of size 1, for a total of 9 isomorphism types of quotients.
    At another point I asked for a description of the product of two symmetric, loopless graphs, which are both 3-vertex paths. The answer is a graph with 2 connected components where one is a loopless 4-cycle and the other is a loopless graph of size 5 with one vertex of degree 4 and four vertices of degree 1.
    Feb 14
    Today we discussed ultrafilter convergence. We defined the following:
  • (principal, nonprincipal) filter, ultrafilter
  • filter convergence (a filter $\mathcal{F}$ converges to a point $p$ of a space $X$ if $\mathcal{F}$ contains the neighborhood filter of $p$)
  • reduced product, ultraproduct.
  • We discussed relevant examples, too, including (i) a space $X$ of size $\kappa$ whose open sets are the subsets $U\subseteq X$ with $|X\setminus U| <\lambda$, and (ii) a successor ordinal in the order topology. Both examples show that $\omega$-indexed sequences are insufficient to define convergence.
    Feb 16
    Today we discussed ultraproducts and Łos’s Theorem. We discussed the possible isomorphism types of an ultraproduct of the family $\{\mathbb Z_n\;|\;n\in \mathbb N^+\}$.
    Feb 19
    Today we discussed
  • why a collection of subsets of a set $I$ which has the finite intersection property generates a proper filter on $I$,
  • the fact that a proper filter on $I$ can be extended to an ultrafilter on $I$,
  • the relationship (= definitional equivalence) between the Boolean algebra of truth values, $\mathbb B_2 = \langle \{0,1\}; \wedge,\vee,\neg,0,1\rangle$ and the $2$-element ring/field $\mathbb F_2 = \langle \{0,1\}; \cdot,+,\neg,0,1\rangle$. Under this equivalence, $a\wedge b = ab$, $0=0$, and $1=1$, while $a\vee b = a+b+ab, \neg a = 1-a$, and $a+b = (a\wedge \neg b)\vee (b\wedge \neg a), -a = a$. This yields a correspondence between the Boolean algebra ${\mathcal P}(I)\cong 2^{|I|}$ and the commutative ring $\mathbb F_2^{|I|}$. Under this correspondence, ideas from commutative ring theory can be imported into Boolean algebra. For example, one may derive the Ultrafilter Lemma from the theorem of ring theory that asserts that every ideal can be extended to a maximal ideal.
  • Feb 21
    We proved that
  • A filter is an ultrafilter iff it is maximal among proper filters.
  • Any proper filter can be extended to an ultrafilter.
  • A nonprincipal ultraproduct of the form $\prod_{\mathcal{U}} \mathbb Z_p$, where $\mathcal U$ is an ultrafilter on the set of prime numbers, will be an infinite, torsion-free, divisible, abelian group.
  • I mentioned that the theory of infinite, torsion-free, divisible, abelian groups is $\kappa$-categorical for uncountable $\kappa$, but we haven’t proved that statement yet.
    Feb 23
    We proved that the group $\prod_{\mathcal U} \mathbb Z_p$ is isomorphic to $\bigoplus^{2^{\aleph_0}} \mathbb Q$ when $U$ is a nonprincipal ultrafilter on the set of primes. This involved the observation that $\prod_{\mathcal U} \mathbb Z_p$ is a torsion-free, divisible, abelian group of cardinality $2^{\aleph_0}$ when $\mathcal U$ is nonprincipal, and that every torsion-free, divisible, abelian group is a reduct of a uniquely determined rational vector space. We then discussed why the group $\prod_{\mathcal U} \mathbb Z_n$ will also be isomorphic to $\bigoplus^{2^{\aleph_0}} \mathbb Q$ for some choices of $\mathcal U$, but for other choices of $\mathcal U$ it is not hard to see that the group $\prod_{\mathcal U} \mathbb Z_n$ will have a nontrivial torsion subgroup.

    During the lecture we made reference to the main result of

    Saharon Shelah,
    On the cardinality of ultraproducts of finite sets
    Journal of Symbolic Logic, 1970.

    One consequence of this paper is the fact that the class of ultraproducts of finite sets is not elementary. This implies that the process of generating an elementary class is more complicated than simply closing under ultraproducts.

    Feb 26
    We reviewed past material (definition of a logic, the associated Galois connection, definition of a frame, the space of complete $L$-theories is a compact Hausdorff zero-dimensional space, ultrafilter convergence in the space of complete $L$-theories). Then we discussed the first three slides about the topological structure of the space of complete $L$-theories.
    Feb 28
    We proved the Cantor-Bendixson Theorem (existence of decomposition only) following these slides.
    March 1
    Ultraproducts roundup!. During this discussion, we referred to $\kappa$-complete ultrafilters and to regular ultrafilters.
    March 4
    We discussed elementary diagrams, the upward Lowenheim-Skolem Theorem, and the Los-Vaught Test for completeness.
    March 6
    During the meeting we outlined the proof that every algebraically closed field $\mathbb F$ of characteristic zero has the form $\overline{\mathbb Q(X)}$ where $X$ is a transcendence base for $\mathbb F$ over the prime subfield $\mathbb Q$. Moreover, $|\overline{\mathbb Q(X)}|=|X|+\aleph_0$.

    This shows that the theory of algebraically closed fields of characteristic zero is $\kappa$-categorical for all uncountable $\kappa$, but is not $\aleph_0$-categorical. It also shows that a complete first-order axiomatization of the field $\mathbb C$ of complex numbers is given by:

  • axioms defining fields,
  • axioms saying the characteristic is zero, and
  • axioms saying that the field is algebraically closed.

    Key points.

  • March 8
    During the meeting we reviewed the argument that the theory $\textrm{ACF}_0$ of algebraically closed fields of characteristic $0$ is $\kappa$-categorical for uncountable $\kappa$. By the Los-Vaught Test, $\textrm{ACF}_0$ is complete, and hence $\textrm{Th}(\mathbb{C})=\text{ACF}_0$. Then we introduced the theory $\text{RCF}$ of real closed fields and outlined the main ideas needed to prove that this theory has an algebraically prime model $\mathbf{P}=\overline{\mathbb{Q}}\cap\mathbb R$. This algebraically prime model is actually a prime model for the theory, because $\text{RCF}$ has quantifier elimination. The existence of a prime model is sufficient to establish the completeness of $\text{RCF}$, so $\textrm{Th}(\mathbb{R})=\text{RCF}$. Finally, we introduced the theory $\textrm{DLO}$ of dense linear orders without endpoints, and used the Back and Forth method to prove that this theory is $\aleph_0$-categorical and complete. Hence $\textrm{Th}(\langle \mathbb{Q}; <\rangle)=\text{DLO}$. We pointed out that $\textrm{DLO}$ is not $\kappa$-categorical for $\kappa=|\mathbb R|$.
    March 11
    Let $L$ be the language of one unary predicate $E(x,y)$. A graph is a structure $\langle V; E(x,y)\rangle$ in this language satisfying axioms that say that $E(x,y)$ is an irreflexive symmetric relation. For a graph $G=\langle V; E(x,y)\rangle$, if $U, W\subseteq V$ are finite sets satisfying $|U|=m, |W|=n, U\cap W=\emptyset$, then we say that $z\in V$ is correctly joined to $(U,W)$ if $z$ is adjacent to every element of $U$ and to no element of $W$. Let $\varphi_{m,n}$ be the sentence expressing that $\forall (U,V)$, if $|U|=m, |W|=n, U\cap W=\emptyset$, then $\exists z\notin U\cup W$ that is correctly joined to $(U,W)$. Let $T$ be the theory axiomatized by all $\varphi_{m,n}$, $m,n\geq 0$.

    We explained why $T$ is $\aleph_0$-categorical and has no finite models, hence $T$ is complete. We described a construction of a countably infinite model $R$ of $T$, which we called the Random Graph. We observed that every countable graph is embeddable in the Random Graph, and that any isomorphism between finite subgraphs of $R$ extends to an automorphism of $R$.

    We made references to related results that we did not explore. Let me list some of these here:

    1. The Random Graph is called random because it can be constructed by a random process: Start with vertex set $V=\mathbb N$ and choose edges randomly by flipping a fair coin for each $(i,j)\in \mathbb N^2$ and defining $E(i,j)=\top$ if the result is heads and $E(i,j)=\bot$ if the result is tails. The resulting graph satisfies any given sentence $\varphi_{m,n}$ with probability $1$.
    2. The symmetrization of a countable directed graph model of ZFC is random. (Here, if $V$ is a countable model of ZFC, then say that $a, b\in V$ are adjacent in the symmetrization of the associated directed graph model provided $a\in b$ or $b\in a$.)
    3. Let $V$ be the set of primes that are congruent to $1$ modulo $4$. Say that $p$ is adjacent to $q$ if $p$ is a nonzero square modulo $q$ (i.e., if $p\neq q$ and the Legendre symbol $\left(\frac{p}{q}\right)$ evaluates to $1$). The resulting graph is random. [In this example, I am restricting to the set of primes that are $1$ modulo $4$ to ensure that the edge relation is symmetric.]
    4. The theory of the Random Graphs satisfies a zero-one law: If $\sigma$ is a sentence in the language of graphs and $P_n(\sigma)$ is the proportion of graphs with vertex set $V=\{0,1,\ldots,n-1\}$ that satisfy $\sigma$, then $\lim_{n\to \infty} P_n(\sigma)$ exists and it is either $0$ or $1$. This limit is $1$ if $\sigma\in \textrm{Th}(R)$ and is $0$ if $\neg \sigma\in \textrm{Th}(R)$. For this reason, we call $\textrm{Th}(R)$ the almost sure theory of the class of graphs.
    5. All the results mentioned above about the Random Graph can be generalized to the class of all structures in a given finite relational signature. The results do not hold for signatures with operations. In a finite signature with at least operation of arity at least $2$, there will be a sentence $\sigma$ such that the limit $\lim_{n\to \infty} P_n(\sigma)$ does not exist. In a finite signature where all operations have arity $1$, the limit $\lim_{n\to \infty} P_n(\sigma)$ will always exist, but might be strictly between $0$ and $1$.
    March 13
    We reviewed the Back and Forth method which we applied to show that the theories of (i) the ordered set of rational numbers and (ii) the random graph are $\aleph_0$-categorical. We isolated the key element that allowed us to derive our conclusions, namely that the structures (i) and (ii) are $\aleph_0$-saturated. This led to a discussion of types following these slides.
    March 15
    Snow Day!
    March 18
    We continued discussing these slides (up to, but not including, Vaught’s Corollary).
    March 20
    We discussed Vaught’s Corollary, the Downward LS Theorem, and Skolem’s Paradox to complete these slides.
    March 22
    Read Section 3.1.

    We introduced quantifier elimination following these slides. Some key points were:

    1. In theories $T$ with q.e., every embedding between models of $T$ is elementary.
    2. q.e. helps to establish completeness of $T$.
    3. q.e. implies $\aleph_0$-categoricity for the complete theory of an infinite structure in a finite relational language.
    4. To establish q.e. for $T$ by brute force, it suffices to show how to eliminate a single $\exists y$ from formulas of the form $(\exists y)\varphi(\mathbf{x},y)$ where $\varphi(\mathbf{x},y)$ is a primitive q.f. formula.
    April 1
    Read Section 3.1.

    I announced that I will be at a 4-day conference, and plan to post videos for the three lectures: April 5, April 8, April 10.

    We reviewed the brute force method of showing that a theory has q.e. Then we began these slides which begin to develop alternative methods for establishing that theory has q.e. Our discussion included a discussion of atomic diagrams and a proof of the Diagram Lemma.

    April 3
    I announced that I will be at a 4-day conference, and plan to meet by Zoom on April 5 and post videos for the lectures on April 8, April 10.

    Read Sections 3.1 and 3.2.

    We completed the proof of the local characterization of q.e., and also explained why an $L$-theory $T$ has q.e. if and only if, whenever $\mathbf{A}$ is a substructure of a model of $T$, the $L_A$-theory $T\cup \textrm{Diag}(\mathbf{A})$ is complete. We introduced algebraically prime extensions, and explained why the theory of fields and the theory $\textrm{ACF}_0$ have algebraically prime extensions. (E.g., if $\mathbb D$ is a substructure of a model of $\textrm{ACF}_0$, then $\mathbb D$ is an integral domain of characteristic zero. The algebraic closure $\mathbb D^*$ of the field of fractions of $\mathbb D$ is algebraically prime over $\mathbb D$ with respect to $\textrm{ACF}_0$.)

    April 5
    Meeting ID: 946 8655 4534

    Read Section 3.2.

    We completed these q.e. slides. For applications, we established that the theory of $\mathbb C$ has q.e., while the theories of $\mathbb R$ and $\mathbb Q$ do not.

    April 8
    Meeting ID: 946 8655 4534

    Read Sections 4.1 and 4.2.

    We reviewed the concept of a type following these slides. This review included showing that any $n$-type of $\mathbf{A}$ can be realized in an elementary extension of $\mathbf{A}$. We started discussing the Omitting Types Theorem following these slides.

    April 10
    Meeting ID: 946 8655 4534

    Read Section 4.2.

    We finished these slides which discuss the proof of the Omitting Types Theorem.

    April 12
    We discussed the topological space $S_1(T)$ where $T$ is the elementary diagram of $\langle \mathbb Q; < \rangle$.
    April 15
    Read Section 4.2.

    We proved the Atomic Model Theorem.

    April 17
    Read Sections 4.2 and 4.3.

    We finished the slides on atomic models (handout form) and started the slides on saturated models (handout version).

    April 19
    Read Section 4.3.

    We got to slide 10 of saturated models (handout version).

    April 22
    Read Section 4.3.

    We continued the discussion of saturation (handout version), introducing the concepts of

    1. $\omega$-saturation,
    2. (strong and ordinary) $\omega$-homogeneity,
    3. $\omega^+$-universality.
    April 24
    Read Section 4.3.

    We discussed the existence and uniqueness of $\omega$-saturated models, and explained why an algebraically closed field of characteristic $p$ is saturated iff it has an infinite transcendence base.

    April 26
    Read Section 4.4.

    We proved the theorem of Engeler, Ryll-Nardzewski, and Engeler (handout version) providing many characterizations of $\omega$-categorical theories and models.

    April 29
    Read Section 4.4.

    We proved Vaught‘s Theorem (handout version) that proves that there is no complete theory in a countable language with exactly two isomorphism types of countable models. We then discussed Vaught‘s Conjecture, Morley‘s Theorems and the Baldwin-Lachlan Theorem.

    May 1
    We spent the hour finalizing HW solutions.