42.02.07 · mathematical-logic / model-theory

Strongly Minimal Sets, Morley Rank, and Stability

shipped3 tiersLean: none

Anchor (Master): Marker 2002 *Model Theory: An Introduction* (Springer GTM 217) Ch. 5-6 (Morley rank and degree, ω-stable theories, strongly minimal sets and the algebraic-closure pregeometry, the dimension theory and its role in Baldwin-Lachlan, the Zilber trichotomy); Tent and Ziegler 2012 *A Course in Model Theory* (Lecture Notes in Logic 40, Cambridge) Ch. 5-8 (stability, the order property, counting types, forking and independence, canonical bases, the stable forking calculus); Pillay 1996 *Geometric Stability Theory* (Oxford Logic Guides 32) (pregeometries, modularity, the trichotomy, one-based and locally modular theories); Shelah 1990 *Classification Theory and the Number of Non-isomorphic Models* 2e (North-Holland) (the main gap, the dividing lines stable/superstable/NDOP/NOTOP); Hrushovski 1996 *The Mordell-Lang conjecture for function fields* (J. AMS 9) and Hrushovski 1993 *A new strongly minimal set* (Ann. Pure Appl. Logic 62)

Intuition Beginner

The earlier units in this chapter learned to build worlds, to count the behaviours an element could have, and to rewrite every property as a plain check with no hidden search. This unit asks the deepest structural question of all: how do you put a notion of dimension on the describable parts of a world, the way a line is one-dimensional, a plane two-dimensional, and space three-dimensional — and what does having such a dimension tell you about the world?

Start with the simplest possible world: a line so plain that every describable part of it is either a short finite list or everything-except-a-short-list. Nothing in between. This is the model-theoretic version of a one-dimensional line, and it is the seed from which a whole dimension theory grows. The key move is to decide when a new point is genuinely new: a point is new relative to some chosen points when it is not pinned down to a short finite list by them. Choosing points until no more new ones can be added gives a smallest pinning-down set, and the number of points in it is the dimension.

This is exactly how you measure dimension elsewhere. In ordinary space, three independent directions pin down every other; a fourth direction is forced, not new. In a field of numbers, you can keep picking numbers that satisfy no equation built from the earlier ones until, eventually, every further number is forced — and the count of genuinely-new ones is the dimension.

The surprising payoff is that counting describable behaviours, rather than building one model, is what separates a tame world from a wild one. A tame world has few behaviours and a clean dimension; a wild world, like ordinary arithmetic, has so many describable behaviours that no clean dimension exists. This counting is the engine of the deepest structural theory in logic.

Visual Beginner

The picture shows the one idea this unit is built on: choose genuinely-new points one at a time until the rest are forced, and count how many you chose.

   THE SIMPLE LINE
   every describable part is a short list OR all-but-a-short-list

   describable parts:   { 2, 5, 9 }        (a short list)
                        everything except { 2, 5, 9 }   (all-but-a-short-list)
                        nothing in between is describable

   BUILDING A DIMENSION BY CHOOSING NEW POINTS

   start empty .................... dimension so far: 0
   pick a point p1 (genuinely new). dimension so far: 1
   pick p2, not pinned by p1 ...... dimension so far: 2
   pick p3, not pinned by p1,p2 ... dimension so far: 3
   every further point is now FORCED (pinned to a short list)
                                    final dimension: 3

   p1 ---- p2 ---- p3 ---- (p4 forced) ---- (p5 forced) ---- ...
   |<------ chosen, genuinely new ------>|<--- forced --->|

word	plain meaning	everyday echo
simple line	every part is a short list or all-but-a-short-list	a perfectly clean ruler
genuinely new	not pinned to a short list by earlier points	a fresh independent direction
pinning set (basis)	a smallest set that forces all the rest	three directions in space
dimension	how many genuinely-new points you could pick	the count of independent directions

Read top to bottom: pick new points until the rest are forced, and the count you reach is the dimension of the world.

Worked example Beginner

We measure dimension in a small, concrete way using a field of numbers where you may add, subtract, multiply, and divide, and where every equation that can have a solution does. We pick numbers one at a time and decide which are genuinely new.

Step 1. Start with nothing chosen. The numbers that are forced from nothing are the ones satisfying a fixed equation with plain whole-number coefficients — for instance $x \times x = 4$ forces $x$ to be one of $2$ or $- 2$ , a short list of two. These forced numbers are not genuinely new.

Step 2. Pick a first number $t$ that satisfies no such equation at all — a genuinely new number, not on any short list forced from nothing. The dimension so far is $1$ . Think of $t$ as a fresh independent direction.

Step 3. Now look at $t \times t$ , that is $t^{2}$ . Is it genuinely new relative to $t$ ? No: it is pinned to a one-element short list by the equation "this number equals $t$ times $t$ ." So $t^{2}$ is forced, not new. The same is true of $t + 5$ , of $3 \times t$ , and of every number you can build from $t$ by the four operations.

Step 4. Pick a second number $s$ that satisfies no equation built from $t$ — genuinely new relative to $t$ . The dimension is now $2$ . Then $t + s$ , $t \times s$ , and so on are all forced by $t$ and $s$ together; none is genuinely new.

Step 5. Count. We chose $t$ and $s$ as genuinely new, and everything else we looked at was forced. If the whole world is built from two such independent choices, its dimension is $2$ .

What this tells us: dimension is just the count of genuinely-new points you can pick before the rest are forced. A number is forced when a fixed equation pins it to a short list; it is new when no such equation does. This single test — forced versus new — is what builds dimension on a clean describable world, and it matches the everyday count of independent directions.

Check your understanding Beginner

Exercise (easy, multiple choice).

A describable world is a "simple line" in this unit's sense when which condition holds?

A. It has finitely many points.

B. Every describable part of it is either a short finite list or everything-except-a-short-list.

C. Every point satisfies some equation.

D. It can be put in order.

Hint

The defining feature is that nothing describable sits "in between" — no describable part is both endless and missing endlessly many points.

Answer

B. A simple line is one where every describable part is a short finite list or all-but-a-short-list, with nothing in between. Feedback-correct: that "finite or all-but-finite" cleanness is exactly the condition, and it is what lets a dimension be defined. Feedback-wrong: A is false since these lines are infinite; C and D describe other features that simple lines may or may not have and are not the defining test.

Formal definition Intermediate+

Work in a complete theory $T$ with infinite models and a fixed large saturated model (a monster) $M ⊨ T$ ; "definable" means definable with parameters from $M$ , and types are taken over small parameter sets. The diagram and type-space machinery is that of 42.02.02 pending and 42.02.03 pending, and the strong minimality of $ACF$ established by quantifier elimination in 42.02.05 pending is the running example.

A definable set $D$ is strongly minimal when $D$ is infinite and, in every elementary extension, every definable subset of $D$ is finite or cofinite in $D$ ; $T$ is strongly minimal when its home sort is ^{[Marker Ch. 6]}. Equivalently, $D$ is minimal — no definable subset of $D$ is infinite and co-infinite — with minimality preserved under elementary extension. The motivating instances: $ACF$ (the line of an algebraically closed field, strongly minimal because by QE every one-variable definable set is a boolean combination of polynomial zero sets, hence finite or cofinite, cross-ref 42.02.05 pending); an infinite vector space over a fixed field in the language with the scalar-action functions; the pure infinite set; and an infinite set with one fixed-point-free bijection.

On a strongly minimal $D$ , model-theoretic algebraic closure $acl (A) = {b \in D : b lies in a finite A -definable set}$ is a pregeometry: it is reflexive ( $A \subseteq acl (A)$ ), monotone, idempotent ( $acl (acl (A)) = acl (A)$ ), of finite character, and satisfies the Steinitz exchange property — if $a \in acl (A b) ∖ acl (A)$ then $b \in acl (A a)$ ^{[Marker Ch. 6]}. Exchange makes " $A$ is independent" (no $a \in A$ lies in $acl (A ∖ {a})$ ) a matroid notion, so maximal independent subsets — bases — all have the same cardinality, the dimension $dim (D)$ . For $ACF$ this dimension is transcendence degree; for vector spaces it is linear dimension.

The Morley rank $RM (φ)$ of a formula or definable set is the Cantor-Bendixson rank of the type lattice: $RM (φ) \geq 0$ if $φ$ is consistent; $RM (φ) \geq α + 1$ if there are infinitely many pairwise-inconsistent $ψ_{i} ⊢ φ$ with $RM (ψ_{i}) \geq α$ ; and $RM (φ) \geq λ$ at a limit $λ$ when $RM (φ) \geq β$ for all $β < λ$ . The Morley degree is the largest number of disjoint rank- $RM (φ)$ pieces. Strong minimality is precisely $RM = 1$ with Morley degree $1$ , and on a strongly minimal set Morley rank coincides with the pregeometry dimension.

A complete $T$ is $ω$ -stable (totally transcendental) when every formula has ordinal Morley rank, equivalently $∣ S_{n} (A) ∣ \leq ∣ A ∣ + ℵ_{0}$ for all parameter sets $A$ . More generally $T$ is $λ$ -stable if $∣ S_{n} (A) ∣ \leq λ$ whenever $∣ A ∣ \leq λ$ , stable if $λ$ -stable for some $λ$ , and superstable if $λ$ -stable for all $λ \geq 2^{∣ T ∣}$ . By Shelah's theorem $T$ is unstable exactly when it has the order property: a formula $φ (\overset{x}{ˉ}, \overset{y}{ˉ})$ and tuples $(\overset{a}{ˉ}_{i})_{i < ω}$ with $⊨ φ (\overset{a}{ˉ}_{i}, \overset{a}{ˉ}_{j}) ⟺ i < j$ ^{[Tent-Ziegler Ch. 8]}. Then $ACF$ , modules, and (by Sela) free groups are stable; $(R, <)$ and $(N, +, \cdot)$ are unstable.

Counterexamples to common slips Intermediate+

"Minimal and strongly minimal are the same." Minimality (no infinite, co-infinite definable subset) in a single model can be lost in an elementary extension; strong minimality demands it in all extensions. A definable set can be minimal in $M$ yet acquire an infinite/co-infinite definable subset once new parameters appear, so the two notions genuinely differ for non-saturated $M$ .
"Stable means few models." Stability counts types, not models. $ACF$ is $ω$ -stable yet has many non-isomorphic models of each uncountable size; what stability forbids is the order property and the resulting $2^{λ}$ types over a parameter set of size $λ$ . The model-counting consequences (the main gap) are downstream of, not identical to, the type-counting definition.
"Morley rank is additive without hypotheses." Additivity $RM (ab / A) = RM (a / A b) + RM (b / A)$ is a theorem of Lascar for $ω$ -stable theories and can fail as equality outside that setting, where only the Erimbetov-style inequalities survive. The clean dimension calculus is a feature of total transcendence, not of every theory carrying a rank.

Key theorem with proof Intermediate+

The signature result is the exchange property for algebraic closure on a strongly minimal set, the fact that turns the bare "finite or cofinite" condition into a genuine dimension theory. It is the engine of Baldwin-Lachlan (cross-ref 42.02.06 pending) and the foundation on which Morley rank, forking, and the whole stability geometry are built.

Theorem (Exchange on a strongly minimal set). Let $D$ be strongly minimal and $A \subseteq D$ . If $a, b \in D$ with $a \in acl (A b) ∖ acl (A)$ , then $b \in acl (A a)$ ^{[Marker Ch. 6]}.

Proof. Suppose $a \in acl (A b)$ , witnessed by a formula $φ (x, b)$ over $A$ with $⊨ φ (a, b)$ and $φ (x, b)$ having exactly $n$ solutions for some finite $n$ (take $n$ least, so $φ (x, b)$ defines a set of size $n$ containing $a$ ). Consider the set $Y = {y \in D : φ (a, y) holds and φ (x, y) has \leq n solutions}$ , an $A a$ -definable subset of $D$ containing $b$ . By strong minimality $Y$ is finite or cofinite. If $Y$ is finite, then $b \in acl (A a)$ and we are done.

Assume for contradiction $Y$ is cofinite, so $D ∖ Y$ is finite, say of size $m$ . For each $y \in Y$ the formula $φ (x, y)$ has at most $n$ solutions and $a$ is among them. Count the pairs $(x, y)$ with $⊨ φ (x, y)$ , $y \in Y$ , and $x$ ranging over the (at most $n$ ) solutions: as $y$ ranges over the cofinite $Y$ , the element $a$ is a solution of $φ (x, y)$ for cofinitely many $y$ , so the $A a$ -definable set ${y :⊨ φ (a, y)}$ is cofinite, hence infinite. But then by symmetry of the counting, the set ${x : φ (x, y) holds for infinitely many y \in Y}$ is a definable subset of $D$ over $A$ (the parameter $a$ enters only through $Y$ 's defining data, which is $A$ -definable once we quantify $y$ out), and it is nonempty and finite — it has at most $n$ elements because each $φ (x, y)$ admits $\leq n$ solutions and an averaging bound caps the over-represented $x$ . Strong minimality forces this $A$ -definable finite set to contain $a$ , so $a \in acl (A)$ , contradicting $a \in / acl (A)$ . Hence $Y$ is finite and $b \in acl (A a)$ . $□$

Corollary (well-defined dimension). Exchange makes independence a matroid notion on $D$ : any two maximal independent subsets of a definable set have the same size, the dimension $dim$ , and $dim$ agrees with Morley rank. For $ACF$ this is the equality of model-theoretic dimension with transcendence degree, the input Baldwin-Lachlan feeds into the categoricity theorem (cross-ref 42.02.06 pending).

Bridge. The exchange property is the foundational reason a strongly minimal set carries a dimension at all: without it, "independent" would not be a matroid notion and bases could have different sizes, so the whole dimension calculus would collapse. It builds toward the Morley-rank additivity and the forking independence of the Advanced results, where exchange reappears as the symmetry of nonforking, and it appears again in the categoricity theory of 42.02.06 pending, where the dimension of a strongly minimal set is exactly the isomorphism invariant of an uncountably categorical model. This is exactly the Steinitz exchange lemma of linear and field-theoretic dependence, lifted from vectors and transcendence bases to definable sets, and it generalises the back-and-forth matching of 42.02.01 pending from "match points across a gap" to "match independent points up to a dimension count." The central insight is that strong minimality plus exchange is a synthetic axiomatization of dimension, and putting these together, the symmetry of dependence ( $a$ depends on $b$ over $A$ iff $b$ depends on $a$ over $A$ ) is the bridge from a one-variable finiteness condition to the full geometric stability theory.

Exercises Intermediate+

Exercise 3 (medium, symbolic).

Show that an infinite vector space $V$ over a fixed field $k$ , in the language of abelian groups expanded by a unary function for each scalar, is strongly minimal.

Hint

The theory has quantifier elimination down to boolean combinations of linear equations $\sum λ_{i} x = c$ in one variable $x$ . What is the solution set of one such equation?

Answer

The theory of an infinite $k$ -vector space eliminates quantifiers to boolean combinations of equations $λ x = c$ for $λ \in k$ and $c$ a parameter (a $k$ -linear combination of named elements). A single equation $λ x = c$ with $λ \neq = 0$ has exactly one solution $x = λ^{- 1} c$ ; with $λ = 0$ it is " $0 = c$ ", either all of $V$ or empty. So every atomic one-variable definable set is a point, all of $V$ , or empty, and a boolean combination of these is finite or cofinite. As this holds in every elementary extension (the axioms of an infinite $k$ -vector space are preserved), $V$ is strongly minimal. Its acl-dimension is linear dimension over $k$ . Rubric: full credit for the QE reduction to linear equations, the one-point solution sets, and the finite/cofinite closure preserved under extension.

Exercise 4 (medium, symbolic).

Prove that $(R, <)$ is unstable by exhibiting the order property.

Hint

The order property needs a formula $φ (x, y)$ and elements $a_{i}$ with $φ (a_{i}, a_{j})$ holding exactly when $i < j$ . The order relation itself is the candidate formula.

Answer

Take $φ (x, y) \equiv x < y$ and the elements $a_{i} = i$ for $i < ω$ (any strictly increasing sequence works). Then $⊨ φ (a_{i}, a_{j})$ , i.e. $a_{i} < a_{j}$ , holds exactly when $i < j$ . This is the order property for $φ$ , so by Shelah's theorem $(R, <)$ is unstable. Concretely, over the parameter set ${a_{i} : i < λ}$ the cuts realize $2^{λ}$ distinct $1$ -types ("above $a_{i}$ for $i \in S$ , below for $i \in / S$ "), far exceeding $λ$ , which is the type-counting face of instability. Rubric: full credit for naming the formula $x < y$ , the increasing witnesses, and either the order-property or the $2^{λ}$ -types conclusion.

Exercise 5 (medium, symbolic).

Using that every formula in an $ω$ -stable theory has ordinal Morley rank, show that $∣ S_{1} (A) ∣$ is at most $∣ A ∣ + ℵ_{0}$ for countable $A$ .

Hint

Each complete type over $A$ has a Morley rank and degree; an isolated point of the type space corresponds to a formula of minimal rank in its piece. Bound the number of (rank, degree)-minimal formulas.

Answer

If some $∣ S_{1} (A) ∣ > ∣ A ∣ + ℵ_{0} = ℵ_{0}$ were uncountable, the type space $S_{1} (A)$ — a compact Hausdorff space — would have a perfect (Cantor-Bendixson rank $\infty$ ) subset, so some formula $φ$ over $A$ would have a binary tree of consistent formulas $φ_{s}$ ( $s \in 2^{< ω}$ ) with $φ_{s 0}, φ_{s 1}$ inconsistent and each implying $φ_{s}$ . This tree witnesses $RM (φ) \geq α$ for every ordinal $α$ (the splitting never terminates), so $φ$ has no ordinal Morley rank, contradicting $ω$ -stability. Hence $S_{1} (A)$ is countable, i.e. $∣ S_{1} (A) ∣ \leq ∣ A ∣ + ℵ_{0}$ . Rubric: full credit for the perfect-set/binary-tree dichotomy, the resulting failure of ordinal rank, and the contradiction with total transcendence.

Exercise 6 (medium, symbolic).

State the three cases of the Zilber trichotomy for a strongly minimal set and name the example realizing each.

Hint

The trichotomy classifies the pregeometry: disintegrated, group-like (locally modular non-degenerate), and field-like (non-locally-modular).

Answer

A strongly minimal set falls into exactly one geometric case. (1) Disintegrated: $acl (A) = ⋃_{a \in A} acl (a)$ — closure adds nothing beyond the union of one-point closures; example, the pure infinite set or an infinite set with one fixed-point-free bijection. (2) Locally modular non-degenerate (group-like): the geometry is that of a projective/affine space over a division ring, modular after localizing at one point; example, an infinite vector space, where dependence is linear dependence. (3) Non-locally-modular (field-like): the geometry is not modular even locally; example, an algebraically closed field, where dependence is algebraic dependence/transcendence. Zilber conjectured case (3) forces an interpretable algebraically closed field; Hrushovski refuted this in general. Rubric: full credit for the three cases with one correct example each and the conjecture/refutation note.

Exercise 7 (hard, symbolic).

Prove that on a strongly minimal set, if $\overset{a}{ˉ} = (a_{1}, \dots, a_{n})$ and $\overset{a}{ˉ}^{'}$ are both independent over $A$ with $acl (A \overset{a}{ˉ}) = acl (A \overset{a}{ˉ}^{'})$ , then $dim (\overset{a}{ˉ} / A) = dim (\overset{a}{ˉ}^{'} / A)$ , and deduce that dimension is well defined.

Hint

Use exchange to show any two maximal independent subsets of a definable closure have equal size, mimicking the proof that bases of a vector space have equal cardinality.

Answer

Let $B = {b_{1}, \dots, b_{p}}$ and $C = {c_{1}, \dots, c_{q}}$ be two maximal $A$ -independent subsets of a set with the same algebraic closure $acl (A B) = acl (A C) = X$ . We show $p = q$ by an exchange/Steinitz swap. Each $c_{j} \in acl (A B)$ , and by maximality each $b_{i} \in acl (A C)$ . Suppose $p < q$ . Inductively replace elements of $B$ by elements of $C$ : given that ${c_{1}, \dots, c_{k}, b_{k + 1}, \dots, b_{p}}$ is independent and has closure $X$ (true at $k = 0$ ), the element $c_{k + 1} \in acl (A c_{1} \dots c_{k} b_{k + 1} \dots b_{p})$ ; since $c_{k + 1} \in / acl (A c_{1} \dots c_{k})$ (the $c$ 's are independent), some $b_{i}$ ( $i > k$ ) satisfies $c_{k + 1} \in acl (\dots b_{i} \dots) ∖ acl (\dots b_{i} \dots)$ , and exchange swaps $b_{i}$ for $c_{k + 1}$ , keeping independence and closure $X$ . After $p$ steps all $b$ 's are replaced by $c_{1}, \dots, c_{p}$ , whose closure is $X ∋ c_{p + 1}$ , forcing $c_{p + 1} \in acl (A c_{1} \dots c_{p})$ and contradicting independence of $C$ . Hence $p \geq q$ , and symmetrically $q \geq p$ , so $p = q$ . Dimension, the common size of all maximal independent subsets, is therefore well defined. Rubric: full credit for the Steinitz swap driven by exchange, the contradiction from a strict size inequality, and the symmetric conclusion.

Exercise 8 (hard, symbolic).

Prove the symmetry of algebraic dependence on a strongly minimal set: for $a, b \in D$ and $A \subseteq D$ , $a \in acl (A b) ∖ acl (A)$ iff $b \in acl (A a) ∖ acl (A)$ , and explain why this is the strongly minimal shadow of forking symmetry.

Hint

One direction is the exchange property. For the symmetric statement also rule out $b \in acl (A)$ when $a \in acl (A b) ∖ acl (A)$ .

Answer

Assume $a \in acl (A b) ∖ acl (A)$ . By the exchange property $b \in acl (A a)$ . It remains to see $b \in / acl (A)$ : if $b \in acl (A)$ then $acl (A b) = acl (A)$ , so $a \in acl (A)$ , contradicting the hypothesis. Hence $b \in acl (A a) ∖ acl (A)$ . The converse is the same argument with $a, b$ swapped. Thus " $a$ depends on $b$ over $A$ " — meaning $dim (a / A b) < dim (a / A)$ , i.e. $a \in acl (A b) ∖ acl (A)$ — is symmetric in $a$ and $b$ . This is exactly the strongly minimal case of forking symmetry: in a stable theory $a ⌣ ∣_{A} b$ (forking dependence) is symmetric, and on a strongly minimal set nonforking independence is precisely algebraic independence, so exchange-symmetry is the rank- $1$ instance of the general symmetry of the forking independence relation. Rubric: full credit for the exchange step, the $b \in / acl (A)$ exclusion, the symmetric converse, and the identification with forking symmetry at rank $1$ .

Advanced results Master

Strong minimality, Morley rank, stability, and forking organize five developments: the pregeometry and its dimension on a strongly minimal set, the Cantor-Bendixson rank theory of $ω$ -stable theories with Morley-rank additivity, the type-counting stability spectrum and Shelah's order-property dichotomy, the abstract forking independence calculus generalizing algebraic and linear independence, and the geometric classification — the Zilber trichotomy, the Hrushovski construction, the Shelah main gap, and the Diophantine applications.

Theorem 1 (strongly minimal pregeometry and dimension). On a strongly minimal set $D$ , $acl$ is a pregeometry with the Steinitz exchange property, so independence is a matroid notion, bases of any definable set have a common cardinality $dim$ , and $dim$ equals Morley rank ^{[Marker Ch. 6]}. For $ACF$ this is transcendence degree and for vector spaces linear dimension; the dimension is the unique isomorphism invariant of the prime-over- $A$ strongly minimal structure, the input the Baldwin-Lachlan analysis of 42.02.06 pending turns into the categoricity theorem. A type over $A$ of dimension $0$ is algebraic; a type of full dimension is the generic type of $D$ , the unique nonalgebraic complete type over $A$ extending " $x \in D$ ", and generics are the rank-maximal points the forking calculus tracks.

Theorem 2 (Morley rank, degree, and additivity in $ω$ -stable theories). A complete $T$ is $ω$ -stable iff every formula has ordinal Morley rank iff $∣ S_{n} (A) ∣ \leq ∣ A ∣ + ℵ_{0}$ for all $A$ ^{[Tent-Ziegler Ch. 5]}. Morley rank is the Cantor-Bendixson rank of the Stone space of types, definable (the set of $\overset{ˉ}{b}$ with $RM (φ (\overset{x}{ˉ}, \overset{ˉ}{b})) \geq α$ is definable), and additive by Lascar: $RM (tp (a, b / A)) = RM (tp (a / A b)) + RM (tp (b / A))$ . Strong minimality is $RM = 1$ , degree $1$ ; finite Morley rank theories are the iterated extensions of strongly minimal sets, where rank is a genuine dimension and degree counts the components. The definability of rank is what makes uniform finiteness and definable families behave, and it underlies the indecomposability and stabilizer theory of groups of finite Morley rank.

Theorem 3 (the stability spectrum and Shelah's dichotomy). $T$ is $λ$ -stable when $∣ S_{n} (A) ∣ \leq λ$ for $∣ A ∣ \leq λ$ ; stable means $λ$ -stable for some $λ$ , superstable means $λ$ -stable for all $λ \geq 2^{∣ T ∣}$ , $ω$ -stable means $ℵ_{0}$ -stable. By Shelah, $T$ is unstable iff it has the order property, and the stable theories are exactly those where every formula is definable (the defining scheme of a type is itself definable) and no formula has the binary-tree/independence property in the unstable sense ^{[Tent-Ziegler Ch. 8]}. The spectrum sorts the model-theoretic universe: $ACF$ , modules, differentially and separably closed fields, and free groups are stable; $(R, <)$ , $Th (N, +, \cdot)$ , the random graph, and every infinite linear order are unstable. NIP (no independence property) and simplicity are the broader dividing lines beyond stability, with $ACFA$ and the random graph simple-unstable and o-minimal structures (cross-ref 42.02.08 pending) NIP-unstable, marking the modern tame landscape.

Theorem 4 (forking and the independence calculus). In a stable theory, forking defines an independence relation $a ⌣ ∣_{A} b$ : a complete type $p \in S (B)$ forks over $A \subseteq B$ if it implies a finite disjunction of formulas each dividing over $A$ (a formula $δ (x, b)$ divides over $A$ if some $A$ -indiscernible sequence $(b_{i})$ with $b_{0} = b$ has ${δ (x, b_{i})}$ inconsistent) ^[Pillay]. Nonforking is invariant, of finite character, monotone, base-monotone, symmetric, transitive, with local character (every type does not fork over a set of size $\leq ∣ T ∣$ ), existence and extension (every type has a nonforking extension), and stationarity over models — a type over a model has a unique nonforking extension to any superset, encoded by its canonical base $Cb (p)$ , the smallest definably closed set over which $p$ does not fork. Forking specializes to algebraic independence in $ACF$ (the canonical base is the field of definition of a variety) and to linear independence in modules; it is the abstract dimension theory that strong minimality exhibits at rank $1$ .

Theorem 5 (trichotomy, Hrushovski construction, main gap, applications). The Zilber trichotomy classifies a strongly minimal pregeometry as disintegrated, locally modular non-degenerate (group-like), or non-locally-modular (field-like) ^[Pillay]. Zilber conjectured the field-like case interprets an algebraically closed field; Hrushovski's construction — a Fraïssé amalgamation controlled by a submodular predimension $δ$ — builds a strongly minimal set with non-locally-modular geometry interpreting no infinite group, refuting the conjecture in general, while the positive trichotomy survives for Zariski geometries (Hrushovski-Zilber) ^{[Hrushovski 1993]}. At the macroscopic scale, Shelah's main gap theorem partitions complete theories by dividing lines (stable, superstable, NDOP, NOTOP, shallow): the number $I (T, κ)$ of models of cardinality $κ$ is either bounded well below $2^{κ}$ with a structure theory (decomposition into a tree of independent strongly-minimal-like pieces) or is the maximum $2^{κ}$ for unboundedly many $κ$ — there is no middle. Hrushovski's proof of the Mordell-Lang conjecture for function fields runs the modular/non-modular dichotomy inside separably closed and difference fields, forcing the modular case where definable subsets of a one-based group are boolean combinations of cosets, yielding the coset structure of $X \cap Γ$ ^{[Hrushovski 1996]}.

Synthesis. Stability theory is the foundational reason model theory has a geometry rather than only a model count, and putting these together it threads all five developments through a single principle: dimension and independence, exhibited concretely by exchange on a strongly minimal set, abstract into the forking calculus that classifies every theory. This is exactly what Theorem 1's exchange property delivers — a synthetic dimension on the rank- $1$ line — and Theorem 2 shows the immediate dividend: Morley rank is that dimension made ordinal-valued and additive, so finite-rank theories are towers of strongly minimal pieces and the central insight is that the Cantor-Bendixson analysis of the type space is a dimension theory. Theorem 3's type-counting dichotomy is dual to the order property: a theory is stable exactly when it omits the linear order that would force $2^{λ}$ types, and stability generalises strong minimality from "one rank- $1$ line" to "every type definable, every independence relation well-behaved." Theorem 4's forking is the abstraction itself — algebraic independence in $ACF$ and linear independence in modules are the same relation, symmetric and transitive with canonical bases, and this is exactly why the strongly minimal exchange of the Key theorem is the rank- $1$ shadow of forking symmetry.

Theorem 5 is the payoff at two scales. Microscopically, the Zilber trichotomy reads a strongly minimal geometry as disintegrated, group-like, or field-like, and the Hrushovski construction shows the field-like case is wilder than conjectured, the bridge from abstract pregeometry to the Zariski geometries that carry Diophantine content. Macroscopically, Shelah's main gap is the dimension theory ascended to the count of models: a theory is either structured — decomposable into independent strongly-minimal-like dimensions whose count of models stays small — or wild, with the maximum number of models, and the dividing lines stable/superstable/NDOP are the successive refinements of "has a dimension theory." Stability replaces the coarse invariant categoricity of 42.02.06 pending with a fine geometry of independence on definable sets, the deepest structural theory in logic, and Hrushovski's Mordell-Lang is the proof that this geometry is not internal bookkeeping but a tool that settles arithmetic; the parallel order-theoretic tame world is the o-minimality of 42.02.08 pending.

Full proof set Master

Proposition 1 (exchange on a strongly minimal set). If $D$ is strongly minimal, $A \subseteq D$ , and $a \in acl (A b) ∖ acl (A)$ , then $b \in acl (A a)$ .

Proof. Let $φ (x, b)$ over $A$ witness $a \in acl (A b)$ , chosen so that $φ (x, b)$ has exactly $n$ solutions with $n$ least over all such witnesses, and $a$ is among them. The set $Y = {y :⊨ φ (a, y) \land “ φ (x, y) has \leq n solutions”}$ is $A a$ -definable and contains $b$ . By strong minimality $Y$ is finite or cofinite. If $Y$ is finite, $b \in acl (A a)$ . If $Y$ is cofinite, then the $A$ -definable set $Z = {x : {y \in Y :⊨ φ (x, y)} is infinite}$ is definable over $A$ alone (the parameter $a$ has been quantified away through $Y$ ). Each $x \in Z$ has $φ (x, y)$ holding for infinitely many $y$ with $φ (x, y)$ of $\leq n$ solutions; a double-counting over the cofinite $Y$ caps $∣ Z ∣ \leq n$ , so $Z$ is finite. Since $a \in Z$ (for cofinitely many $y \in Y$ , $a$ solves $φ (x, y)$ by definition of $Y$ ), $a \in acl (A)$ — contradiction. Hence $Y$ is finite and $b \in acl (A a)$ . $□$

Proposition 2 (well-definedness of dimension). Any two maximal $A$ -independent subsets of an $acl$ -closed definable set $X \subseteq D$ have the same cardinality.

Proof. Let $B, C$ be maximal $A$ -independent with $acl (A B) = acl (A C) = X$ . For finite subsets the Steinitz swap of Exercise 7 applies: each $c \in acl (A B)$ , and replacing elements of $B$ one at a time by elements of $C$ using Proposition 1 preserves independence and closure, forcing $∣ C ∣ \leq ∣ B ∣$ ; symmetry gives equality. For infinite $B, C$ , finite character of $acl$ reduces each dependence to a finite subset, and a back-and-forth bookkeeping over finite swaps yields a bijection $B \to C$ , so $∣ B ∣ = ∣ C ∣$ . $□$

Proposition 3 (strong minimality $\Leftrightarrow$ Morley rank $1$ , degree $1$ ). A definable infinite set $D$ is strongly minimal iff $RM (D) = 1$ and the Morley degree of $D$ is $1$ .

Proof. ( $\Rightarrow$ ) $D$ infinite gives $RM (D) \geq 1$ . If $RM (D) \geq 2$ , there are infinitely many pairwise-inconsistent $ψ_{i} ⊢ (x \in D)$ each of rank $\geq 1$ , hence each infinite; but two disjoint infinite definable subsets of $D$ already violate strong minimality (one is infinite and co-infinite). So $RM (D) = 1$ . Degree $1$ : two disjoint rank- $1$ (infinite) pieces again contradict finite-or-cofinite. ( $\Leftarrow$ ) If $RM (D) = 1$ and degree $1$ , any definable $E \subseteq D$ has $RM (E) \leq 1$ ; if $RM (E) = 1$ and $RM (D ∖ E) = 1$ then $D$ has two disjoint rank- $1$ pieces, degree $\geq 2$ , contradiction, so one of $E$ , $D ∖ E$ has rank $0$ , i.e. is finite. Thus every definable subset is finite or cofinite, in every elementary extension since Morley rank is preserved, and $D$ is strongly minimal. $□$

Proposition 4 ( $ω$ -stable $\Leftrightarrow$ ordinal Morley rank everywhere $\Leftrightarrow$ few types). For complete $T$ : every formula has ordinal Morley rank iff $∣ S_{n} (A) ∣ \leq ∣ A ∣ + ℵ_{0}$ for all $A$ iff $T$ is $ω$ -stable.

Proof. If some formula $φ$ lacks ordinal rank, the rank-defining process never stabilizes, producing a binary tree $(φ_{s})_{s \in 2^{< ω}}$ of consistent formulas with $φ_{s 0}, φ_{s 1}$ inconsistent, each branch a distinct type; over the countable parameter set naming the tree, $∣ S_{n} (A) ∣ \geq 2^{ℵ_{0}} > ∣ A ∣ + ℵ_{0}$ . Conversely if every formula has ordinal rank, each type over a countable $A$ is determined by a rank-and-degree-minimal formula it contains, and the formulas over $A$ number $∣ A ∣ + ℵ_{0}$ , bounding $∣ S_{n} (A) ∣$ by $∣ A ∣ + ℵ_{0}$ . The two outer conditions are thus equivalent, and either is the definition of $ω$ -stability. $□$

Proposition 5 (symmetry of forking at rank $1$ via exchange). On a strongly minimal set, nonforking independence coincides with algebraic independence, and is symmetric: $a ⌣ ∣_{A} b$ iff $dim (a / A b) = dim (a / A)$ iff $dim (b / A a) = dim (b / A)$ .

Proof. On a strongly minimal set Morley rank is $dim$ , so $tp (a / A b)$ forks over $A$ iff its rank drops, iff $dim (a / A b) < dim (a / A)$ , iff $a \in acl (A b) ∖ acl (A)$ (rank drops from $1$ to $0$ in the one-variable case, with the multivariate case reducing coordinatewise by additivity, Proposition 4 of the type calculus). By Proposition 1 (exchange) and Exercise 8, $a \in acl (A b) ∖ acl (A)$ is symmetric in $a, b$ , so the rank-drop condition is symmetric, giving $a ⌣ ∣_{A} b ⟺ b ⌣ ∣_{A} a$ . Hence nonforking is algebraic independence and is symmetric, the rank- $1$ instance of the general stable forking symmetry. $□$

Proposition 6 (instability of a definable linear order). If $T$ interprets an infinite linear order then $T$ is unstable.

Proof. Let $φ (\overset{x}{ˉ}, \overset{y}{ˉ})$ define a relation that, on an infinite definable set, is a linear order $<$ with infinitely many elements $a_{0} < a_{1} < \dots$ . Then $⊨ φ (a_{i}, a_{j}) ⟺ i < j$ , the order property. Over the parameter set ${a_{i} : i < λ}$ , each subset $S \subseteq λ$ that is downward-image-consistent (a cut) yields a distinct complete type asserting $a_{i} < x$ for $i \in S$ and $x < a_{i}$ otherwise, and an infinite linear order has $\geq 2^{λ}$ cuts realized in an elementary extension (by compactness, every consistent cut is realized), so $∣ S_{1} (A) ∣ \geq 2^{λ} > λ$ for $∣ A ∣ = λ$ . Thus $T$ is $λ$ -stable for no $λ$ , i.e. unstable. $□$

Connections Master

Quantifier elimination and model-completeness 42.02.05 pending supplies the strong minimality of $ACF$ that this unit builds its dimension theory on. The proof there that every one-variable definable set in an algebraically closed field is a boolean combination of polynomial zero sets, hence finite or cofinite, is exactly the strong minimality this unit abstracts; the algebraic-closure pregeometry, with transcendence degree as dimension, is the geometric reading of that QE, and model-completeness is the rank- $0$ /rank- $1$ control on which forking specializes to algebraic independence. That unit makes definable sets simple; this unit measures them.
Categoricity and Morley's theorem 42.02.06 pending, co-produced, is the application that consumes this unit's dimension theory. Baldwin-Lachlan deduces Morley's uncountable-categoricity transfer theorem from the fact that an uncountably categorical model is prime and minimal over a strongly minimal set, controlled by its single invariant — the dimension built here from exchange and Morley rank. Where this unit constructs the pregeometry and its dimension, that unit shows the dimension is the unique isomorphism invariant, so categoricity in one uncountable power forces it in all; stability theory is the refinement that replaces the binary categoricity invariant with the full geometry of independence.
Types and the omitting types theorem 42.02.03 pending provides the Stone space of types whose Cantor-Bendixson analysis is Morley rank. The isolated types of that unit are the rank-and-degree-minimal points here, the $ℵ_{0}$ -categoricity finiteness criterion (Ryll-Nardzewski) is the countable shadow of the type-counting stability spectrum, and the prime/atomic models there are the rank- $0$ generic-free models the dimension theory builds over. This unit's stability is that unit's type-counting pushed from $ℵ_{0}$ to every cardinal, with the order property as the obstruction to a clean count.
O-minimality and the cell decomposition theorem 42.02.08 pending is the parallel tame world: where strong minimality reads finiteness off one-variable definable sets and builds a discrete acl-dimension, o-minimality reads finite-union-of-intervals off them and builds a topological dimension on an ordered — hence unstable — structure. The two are the stable and the NIP faces of model-theoretic tameness; this unit's forking and Morley rank are mirrored there by cell-decomposition dimension and the monotonicity theorem, and Hrushovski's Diophantine applications and the Pila-Wilkie counting theorem of that unit are the two arithmetic payoffs of the two tame geometries.
Structures, embeddings, and elementary equivalence 42.02.01 pending supplies the elementary-extension framework in which strong minimality must persist and the back-and-forth that the matroid swap of bases generalizes. The monster-model setting, indiscernible sequences (used to define dividing), and the elementary maps preserving Morley rank and acl-dimension are all read off the embedding hierarchy fixed there; the canonical bases and stationarity of forking are invariants under the elementary maps that unit makes precise.

Historical & philosophical context Master

Strong minimality and Morley rank entered model theory through Michael Morley's 1965 proof that a countable theory categorical in one uncountable power is categorical in all, where the Cantor-Bendixson analysis of the type space — Morley rank — and the total transcendence ( $ω$ -stability) of a categorical theory were introduced as the technical core ^{[Marker Ch. 6]}. John Baldwin and Alistair Lachlan in 1971 isolated the strongly minimal set as the carrier of the dimension that governs the categorical case, reorganizing Morley's theorem around the pregeometry of algebraic closure and its exchange property, the model-theoretic transposition of the Steinitz exchange lemma for transcendence bases.

The abstraction to the full stability spectrum is due to Saharon Shelah, who from the late 1960s through the classification-theory program characterized stability by the order property and the counting of types, defined superstability and the forking independence calculus, and proved the main gap dichotomy — that the number of models of a theory in each uncountable cardinal is either small and computed by a structure theory of independent dimensions or the maximum $2^{κ}$ , settling the spectrum problem ^{[Tent-Ziegler Ch. 8]}. Boris Zilber, studying the geometry of strongly minimal sets, conjectured the trichotomy that the only non-locally-modular geometries are field-like, a conjecture Ehud Hrushovski refuted in 1993 by a predimension amalgamation construction producing a new strongly minimal set with no interpretable group ^{[Hrushovski 1993]}. The positive trichotomy for Zariski geometries, due to Hrushovski and Zilber, became the bridge by which Hrushovski's 1996 model-theoretic proof of the Mordell-Lang conjecture for function fields carried stability theory into Diophantine geometry ^{[Hrushovski 1996]}.

Bibliography Master

@book{marker2002modeltheory,
  author    = {Marker, David},
  title     = {Model Theory: An Introduction},
  series    = {Graduate Texts in Mathematics},
  volume    = {217},
  publisher = {Springer},
  year      = {2002}
}

@book{tentziegler2012,
  author    = {Tent, Katrin and Ziegler, Martin},
  title     = {A Course in Model Theory},
  series    = {Lecture Notes in Logic},
  volume    = {40},
  publisher = {Association for Symbolic Logic / Cambridge University Press},
  year      = {2012}
}

@book{pillay1996geometric,
  author    = {Pillay, Anand},
  title     = {Geometric Stability Theory},
  series    = {Oxford Logic Guides},
  volume    = {32},
  publisher = {Oxford University Press},
  year      = {1996}
}

@book{shelah1990classification,
  author    = {Shelah, Saharon},
  title     = {Classification Theory and the Number of Non-isomorphic Models},
  edition   = {2},
  series    = {Studies in Logic and the Foundations of Mathematics},
  volume    = {92},
  publisher = {North-Holland},
  year      = {1990}
}

@article{morley1965categoricity,
  author  = {Morley, Michael},
  title   = {Categoricity in power},
  journal = {Transactions of the American Mathematical Society},
  volume  = {114},
  year    = {1965},
  pages   = {514--538}
}

@article{baldwinlachlan1971,
  author  = {Baldwin, John T. and Lachlan, Alistair H.},
  title   = {On strongly minimal sets},
  journal = {Journal of Symbolic Logic},
  volume  = {36},
  year    = {1971},
  pages   = {79--96}
}

@article{hrushovski1993newstronglyminimal,
  author  = {Hrushovski, Ehud},
  title   = {A new strongly minimal set},
  journal = {Annals of Pure and Applied Logic},
  volume  = {62},
  year    = {1993},
  pages   = {147--166}
}

@article{hrushovski1996mordelllang,
  author  = {Hrushovski, Ehud},
  title   = {The Mordell-Lang conjecture for function fields},
  journal = {Journal of the American Mathematical Society},
  volume  = {9},
  year    = {1996},
  pages   = {667--690}
}

@article{hrushovskizilber1996zariski,
  author  = {Hrushovski, Ehud and Zilber, Boris},
  title   = {Zariski geometries},
  journal = {Journal of the American Mathematical Society},
  volume  = {9},
  year    = {1996},
  pages   = {1--56}
}

Prerequisites

42.02.05

Tier anchors

beginner: Marker 2002 *Model Theory: An Introduction* (Springer GTM 217) §6.1 (a strongly minimal set read informally as a one-dimensional world so simple that every describable part of it is either a short finite list or everything-but-a-short-list — the model-theoretic line; the picture of a notion of dimension built the way the dimension of a line, a plane, a space is built, by counting how many independent points you can place before the rest are forced; independence read as 'this point is genuinely new, not pinned down by the ones already chosen', a basis as a smallest pinning-down set, dimension as its size; the idea that counting describable behaviours, not just building one model, is what separates a tame theory from a wild one); checking by hand that on the line of an algebraically closed world the only describable subsets are finite or all-but-finite, while on the ordinary number line with order the describable subsets are far richer
intermediate: Marker 2002 *Model Theory: An Introduction* §5.1-5.3, §6.1-6.3 (strongly minimal theories and the strongly minimal sets inside ω-stable theories; algebraic closure acl as a pregeometry/matroid with the exchange property, independence, basis and dimension; Morley rank RM as the Cantor-Bendixson rank of the formula/type lattice and Morley degree; the additivity of Morley rank; ω-stable, superstable, and stable theories defined by counting types |S_n(A)| over parameter sets; the order property and Shelah's unstable/stable dichotomy; nonforking independence and its symmetry, transitivity, and existence of nonforking extensions); Tent and Ziegler 2012 *A Course in Model Theory* (ASL/Cambridge) Ch. 5-8
master: Marker 2002 *Model Theory: An Introduction* (Springer GTM 217) Ch. 5-6 (Morley rank and degree, ω-stable theories, strongly minimal sets and the algebraic-closure pregeometry, the dimension theory and its role in Baldwin-Lachlan, the Zilber trichotomy); Tent and Ziegler 2012 *A Course in Model Theory* (Lecture Notes in Logic 40, Cambridge) Ch. 5-8 (stability, the order property, counting types, forking and independence, canonical bases, the stable forking calculus); Pillay 1996 *Geometric Stability Theory* (Oxford Logic Guides 32) (pregeometries, modularity, the trichotomy, one-based and locally modular theories); Shelah 1990 *Classification Theory and the Number of Non-isomorphic Models* 2e (North-Holland) (the main gap, the dividing lines stable/superstable/NDOP/NOTOP); Hrushovski 1996 *The Mordell-Lang conjecture for function fields* (J. AMS 9) and Hrushovski 1993 *A new strongly minimal set* (Ann. Pure Appl. Logic 62)

References

Marker, D. — Model Theory: An Introduction · Graduate Texts in Mathematics 217, Springer (2002), Chapters 5-6. A definable set D in a model M of a complete theory is strongly minimal if D is infinite and, in every elementary extension N ≽ M, every N-definable subset of D(N) is finite or cofinite in D(N); a theory is strongly minimal when the home sort is. Equivalently D is minimal (no definable subset of D in M itself is infinite and co-infinite) and minimality is preserved in all elementary extensions. The canonical examples: an algebraically closed field is strongly minimal in the ring language (every definable subset of the line is a boolean combination of zero sets of one-variable polynomials, hence finite or cofinite — cross-ref 42.02.05 for the QE that supplies this); an infinite vector space over a fixed field, in the language of the group with the scalar-multiplication unary functions, is strongly minimal; the pure infinite set (no structure) and (ℤ, succ) / an infinite set with a single bijection without finite orbits are strongly minimal. On a strongly minimal set D, model-theoretic algebraic closure acl(A) = {b ∈ D : b lies in a finite A-definable set} is a pregeometry (combinatorial geometry / matroid): it satisfies reflexivity A ⊆ acl(A), monotonicity, idempotence acl(acl(A)) = acl(A), finite character, and crucially the Steinitz EXCHANGE property — if a ∈ acl(Ab) ∖ acl(A) then b ∈ acl(Aa). Exchange makes 'independent' (a ∉ acl of the others) a matroid-independence notion, gives well-defined BASES (maximal independent subsets) and a well-defined DIMENSION dim(A) (the cardinality of any basis), and the dimension of a definable set equals its Morley rank. This dimension is exactly transcendence degree for ACF and linear dimension for vector spaces. Morley rank RM(φ) of a formula (or definable set) is the Cantor-Bendixson rank computed in the type-space lattice: RM(φ) ≥ 0 if φ is consistent; RM(φ) ≥ α+1 if there are infinitely many pairwise-inconsistent ψ_i ⊨ φ each with RM(ψ_i) ≥ α; RM(φ) ≥ λ (limit) if RM(φ) ≥ β for all β < λ. The Morley DEGREE is the maximal number of disjoint rank-RM(φ) subsets. A complete theory is ω-stable (totally transcendental) iff every formula has ordinal Morley rank iff |S_n(A)| ≤ |A| + ℵ_0 for every countable (indeed every) parameter set A; strong minimality is RM = 1, degree 1. Morley rank is ADDITIVE in ω-stable theories: RM(tp(a,b/A)) = RM(tp(a/Ab)) + RM(tp(b/A)) (Lascar; the Erimbetov/additivity inequalities in general). Stability is the type-counting dichotomy: T is λ-stable if |S_n(A)| ≤ λ whenever |A| ≤ λ; T is STABLE if λ-stable for some λ, SUPERSTABLE if λ-stable for all λ ≥ 2^{|T|}, ω-stable if ℵ_0-stable. Shelah's theorem: T is unstable iff T has the ORDER PROPERTY — a formula φ(x̄,ȳ) and tuples (ā_i)_{i<ω} with ⊨ φ(ā_i,ā_j) iff i<j; ACF, modules, separably closed fields, free groups (Sela) are stable, while (ℝ,<), (ℕ,+,·), and any infinite linear order are unstable. FORKING: a complete type p ∈ S(B) forks over A ⊆ B if p implies a disjunction of formulas each dividing over A (a formula δ(x,b) divides over A if some A-indiscernible sequence (b_i) with b_0 = b has {δ(x,b_i)} k-inconsistent). In a stable theory nonforking is a symmetric, transitive independence relation with existence and uniqueness of nonforking extensions, finite character, and the bounded-multiplicity / stationarity of types over models with their canonical bases; it generalizes algebraic independence (ACF) and linear independence (vector spaces). The ZILBER TRICHOTOMY for strongly minimal sets: the pregeometry is either disintegrated/trivial (acl(A) = ∪ acl(a)), or locally modular non-trivial (essentially a vector-space/group geometry — modular up to naming a point), or non-locally-modular ('field-like'); Zilber conjectured the third case forces an algebraically closed field interpretable on D, refuted by Hrushovski's construction of a new strongly minimal set with a non-locally-modular pregeometry interpreting no infinite group. Applications: Baldwin-Lachlan (cross-ref 42.02.06) deduces the Morley categoricity theorem from the dimension of a strongly minimal set; Hrushovski's proof of the Mordell-Lang conjecture for function fields uses the dichotomy theory of differential and difference fields; o-minimality (cross-ref 42.02.08) is the parallel order-theoretic tame world.
Pillay, A. — Geometric Stability Theory · Oxford Logic Guides 32, Oxford University Press (1996). The geometry of forking and the fine structure of stable theories. Develops the pregeometry (combinatorial geometry) attached to a strongly minimal set or, more generally, to a regular type: closure operator with exchange, dimension, and the classification of pregeometries into trivial/disintegrated, modular and locally modular, and non-locally-modular. A pregeometry is modular if for all closed X, Y, dim(X ∪ Y) = dim(X) + dim(Y) − dim(X ∩ Y); locally modular if modular after localizing at a single point; the projective/affine geometries over a division ring are the modular non-trivial examples, and the geometry of an algebraically closed field (transcendence) is the canonical non-locally-modular one. The Zilber trichotomy conjecture (every strongly minimal set is disintegrated, or locally modular and group-like, or interprets an algebraically closed field) is stated, with Hrushovski's counterexample by the amalgamation/predimension construction; the trichotomy is recovered, by Hrushovski-Zilber, for Zariski geometries, which is the form used in Diophantine applications. The book treats one-based theories (every type is based on its canonical base in a modular way — these are exactly the theories whose definable groups are abelian-by-finite and whose geometries are locally modular), the group configuration theorem (a definable group can be reconstructed from a forking configuration of points — Zilber's ladder / Hrushovski's theorem), canonical bases in stable theories, stable and superstable groups, and the geometry underlying the model theory of fields. Forking is presented through the symmetric independence relation a ⫝_A b, its characterization by nondividing, the open mapping theorem and definability of types in stable theories, and stationarity over algebraically closed parameter sets via canonical bases (the canonical base Cb(p) of a stationary type is the smallest definably-closed set over which p does not fork and is definable, the analogue of the field of definition of a variety).
Tent, K. and Ziegler, M. — A Course in Model Theory · Lecture Notes in Logic 40, Association for Symbolic Logic / Cambridge University Press (2012), Chapters 5-8. A modern self-contained development of stability theory. Defines stability by type-counting: a complete theory T is λ-stable if |S_1(A)| ≤ λ for every parameter set A with |A| ≤ λ, and proves the equivalence (over a model) of λ-stability for some λ with the failure of the order property, with the binary-tree property of formulas, and with the definability of types. Proves Shelah's theorem that the stable theories are exactly those without the order property, that a theory stable in one λ ≥ |T| is stable in all λ with λ^{|T|} = λ, and locates the stability spectrum (stable, superstable when stable in all λ ≥ 2^{|T|}, ω-stable / totally transcendental when every formula has ordinal Morley/Cantor-Bendixson rank). Morley rank and degree are developed as the Cantor-Bendixson analysis of the Stone space of types, the ω-stable theories characterized as those in which every type has ordinal Morley rank, with Lascar's additivity / inequalities for Morley rank. Forking is built abstractly: a formula δ(x,b) divides over A if there is an A-indiscernible sequence (b_i) with b_0 = b and {δ(x,b_i) : i < ω} inconsistent; a type forks over A if it implies a finite disjunction of dividing formulas; the chapter proves that in a stable theory dividing = forking is a notion of independence satisfying invariance, finite character, monotonicity, base monotonicity, symmetry, transitivity, existence and extension, local character (every type does not fork over a set of size ≤ |T|), and stationarity over models, with the canonical base of a stationary type. Includes the stable group machinery (generic types, stabilizers, the indecomposability theorem) and the equivalence of strong minimality with Morley rank 1 and degree 1, the pregeometry of acl, and the proof that an uncountably categorical theory is, after passing to a strongly minimal set, controlled by its dimension.
Hrushovski, E. — A new strongly minimal set · Annals of Pure and Applied Logic 62 (1993), pp. 147-166. The refutation of Zilber's trichotomy conjecture in its original strong form. Zilber conjectured that every non-locally-modular strongly minimal set interprets an infinite field (necessarily algebraically closed), so that the only 'field-like' strongly minimal geometries are those of algebraically closed fields. Hrushovski constructs, by a Fraïssé-style amalgamation of finite structures controlled by a PREDIMENSION function δ(A) = (number of relations) weighted against (number of points), and a self-sufficient-closure/flatness argument, a new strongly minimal set whose associated pregeometry is non-locally-modular (non-trivial geometry, genuine 'pseudo-exponential'-type dependence) yet interprets no infinite group, hence no infinite field. The predimension yields, on the generic model, a dimension d(A) = min{δ(B) : A ⊆ B finite} that is submodular; strong minimality follows because the dimension takes the place of Morley rank and forces every definable subset of the line to be finite or cofinite. The same amalgamation-with-predimension technique (the Hrushovski construction) produces a fusion of two strongly minimal theories, new ω-categorical structures, and the framework later used to build a strongly minimal set with the geometry of an algebraically closed field but the wrong characteristic behaviour; it is the central counterexample-engine of geometric stability theory. The positive trichotomy survives for Zariski geometries (Hrushovski-Zilber 1996), the form used in Hrushovski's Diophantine applications.
Hrushovski, E. — The Mordell-Lang conjecture for function fields · Journal of the American Mathematical Society 9 (1996), pp. 667-690. A model-theoretic proof, via stability and the dichotomy theory of definable sets, of the Mordell-Lang conjecture for function fields (the Manin-Mumford and function-field Mordell-Lang statements). The conjecture asserts that the intersection of a subvariety X of a semiabelian variety A (over a function field) with a finite-rank subgroup Γ ≤ A is a finite union of cosets of subgroups, unless X already contains a translate of a positive-dimensional subgroup. Hrushovski works inside the model theory of separably closed fields and of difference fields (ACFA), where the relevant subgroups become definable sets carrying a stability-theoretic dimension; the dichotomy between modular (one-based, 'group-like') and non-modular ('field-like') minimal types — controlled by the Zilber trichotomy as established for the Zariski geometries arising from these fields — forces the modular case, in which definable subsets of a one-based group are exactly the boolean combinations of cosets, yielding the coset structure of the intersection. The proof introduced the model theory of difference and differential fields as a tool for Diophantine geometry, the use of canonical bases as fields of definition, and the dichotomy/Zariski-geometry machinery (Hrushovski-Zilber) as the bridge from abstract stability to algebraic geometry; it is the principal evidence that stability theory is not only a classification of theories but a structural theory of definable sets with arithmetic consequences.

Estimated time

beginner: 24m
intermediate: 60m
master: 100m