Saturation, Homogeneity, and Monster Models

Intuition Beginner

The last unit sorted behaviour reports into two piles: the compulsory ones every world must serve, and the skippable ones a carefully built world can refuse. This unit builds the opposite extreme — a single world so lavishly generous that it refuses nothing. If a behaviour report is consistent with the rules and phrased against finitely many landmarks already inside the world, that world hands you a resident matching it. Nothing consistent is left out. Such a world is called saturated.

Picture a hotel that never turns away a sensible request. Ask for a room above floor three, below floor ten, not next to the elevator — if the request breaks no house rule, a matching room already exists. A saturated world is that hotel for behaviour reports: every consistent demand sheet, written against guests already checked in, is met by an actual guest.

A second feature comes free. In such a world, any two finite groups of guests that look identical from the inside can be slid onto each other by a symmetry of the whole hotel. This is homogeneity: matching patterns are interchangeable, and the world has enough symmetry to swap them. Sameness from the inside becomes actual sameness up to a relabelling of the world.

These two features pin the world down. Two saturated worlds of the same size are secretly the same world. You match them guest by guest, back and forth, each side always able to answer the other's next request because both refuse nothing. Model theorists then fix one enormous saturated world — the monster — and work inside it, the way algebra works inside the complex numbers: a single arena where everything that can happen already has.

Visual Beginner

The picture sets a thin world beside a saturated one. The thin world leaves some consistent demand sheets unfilled; the saturated world fills every one, and any two matching finite patterns can be slid onto each other by a symmetry.

   THIN WORLD                         SATURATED WORLD (the monster, in small)

   guests: a few                      guests: every consistent demand met
   demand "above all whole numbers"   demand "above all whole numbers"
        --> NO matching guest              --> a matching guest exists
   demand "between 1 and 2"           demand "between 1 and 2"
        --> a matching guest               --> a matching guest exists

   some consistent sheets unfilled    NO consistent sheet left unfilled
   few symmetries                     rich symmetry: matching finite
                                       patterns slide onto each other

word	plain meaning	consequence
saturated	every consistent demand against current guests is met	nothing consistent is missing
homogeneous	matching finite patterns slide onto each other	the world is rich in symmetry
universal	every smaller world fits inside	the saturated world contains them all
unique	two same-size saturated worlds are the same	one arena per size
monster	one huge fixed saturated world	the arena everyone works inside

Read it as the generous hotel: it never refuses a sensible request, and its symmetry lets it swap any two matching groups of guests.

Worked example Beginner

We test two ordered worlds for the generous property, using the rational numbers and the whole numbers, each with their own order and no extra landmarks beyond a few we name.

Step 1. Take the rationals with their order. Name two guests, $0$ and $1$. Consider the demand sheet "above $0$ and below $1$." It breaks no rule of a dense order, so a generous world should serve it.

Step 2. The rationals serve it: $1/2$ sits above $0$ and below $1$. Try another sheet — "above $0$, below $1$, and above $1/4$." Still consistent, still served, by $1/2$ again or by $1/3$. Every consistent betweenness demand against named rationals has a rational meeting it, because the order is dense and unbounded. In their own countable size, the rationals refuse nothing of this kind.

Step 3. Now take the whole numbers $0,1,2,3,\dots$ in order. Consider the demand "above $5$ and below $6$." It is consistent with a dense order, but the whole numbers have no element strictly between $5$ and $6$.

Step 4. So the whole numbers refuse a consistent betweenness demand. They are not the generous kind of world; they leave gaps a saturated world would fill.

Step 5. Contrast the two. The rationals met every consistent betweenness sheet against their named guests; the whole numbers left one unfilled. The first behaves like the generous hotel; the second turns guests away.

What this tells us: a world is saturated when it meets every consistent demand phrased against its current residents. The dense rationals pass this test in their size; the gapped whole numbers fail it. The single test — is every consistent demand against named guests met by a real guest? — is exactly the line between a saturated world and a thin one.

Check your understanding Beginner

Formal definition Intermediate+

Fix a complete $\mathcal{L}$-theory $T$ with infinite models and an infinite cardinal $\kappa$. For $\mathfrak{M}\models T$ and $A\subseteq M$, recall $S_n(A)$ is the space of complete $n$-types over $A$ — maximal consistent sets of ${\mathcal{L}}_A$-formulas, equivalently ultrafilters on the algebra of definable sets (42.02.03 pending).

A model $\mathfrak{M}\models T$ is $\kappa$-saturated when for every $A\subseteq M$ with $|A|<\kappa$, every type $p\in S_1(A)$ is realized in $\mathfrak{M}$. Realizing all $1$-types over every such $A$ already forces realization of all $n$-types over such $A$, by realizing coordinates successively over the growing parameter set. The model is saturated when it is $|M|$-saturated. A $\kappa$-saturated model has $|M|\ge \kappa$, since over a set $A$ of size $<\kappa$ one can write down, in many theories, $\ge |A|$-many distinct types whose realizations are distinct elements ^{[Marker §4.3]}.

A partial map $f:A\to M$ with $A\subseteq M$ is partial elementary when $\mathfrak{M}\models \phi (\bar{a})⟺\mathfrak{M}\models \phi (f\bar{a})$ for every $\mathcal{L}$-formula $\phi$ and tuple $\bar{a}$ from $A$. The model $\mathfrak{M}$ is $\kappa$-homogeneous when every partial elementary map $f$ with $|\mathrm{dom}(f)|<\kappa$ extends, for each $a\in M$, to a partial elementary map whose domain contains $a$; it is strongly $\kappa$-homogeneous when every such $f$ extends to an automorphism of $\mathfrak{M}$.

The two notions interlock. A saturated model of cardinality $\kappa$ is $\kappa$-homogeneous: to extend a small partial elementary map $f:\bar{a}\mapsto \bar{b}$ by a new $a$, the type $\mathrm{tp}(a/\bar{a})$ transported along $f$ is a type over $f(\bar{a})=\bar{b}$, a small parameter set, so saturation realizes it by some $b$, and $\bar{a}a\mapsto \bar{b}b$ stays partial elementary.

Counterexamples to common slips Intermediate+

• "Every model is saturated in its own cardinality." The ordered rationals $(\mathbb{Q},<)$ are ${\aleph }_0$-saturated, but $(\mathbb{N},<)$ is not even ${\aleph }_1$-saturated: the type "$x>n$ for every numeral $n$" over the countable parameter set $\mathbb{N}$ is consistent yet omitted. Saturation is a strong demand, not a generic property.

• "Saturated means every type of the theory is realized." It means every type over parameter sets smaller than the model is realized. A saturated model of size $\kappa$ need not realize types over parameter sets of size $\kappa$ — those belong to a larger saturated model. Saturation is always relative to a cardinal bound on the parameters.

• "Homogeneous and saturated are the same." Saturation is realization of types; homogeneity is extension of partial elementary maps. Saturation implies homogeneity, but a model can be homogeneous without being saturated — it may extend maps among the elements it has while still omitting some type. The countable saturated model is homogeneous; not every countable homogeneous model is saturated.

Key theorem with proof Intermediate+

The structural payoff of saturation is rigidity: a saturated model has no freedom left, so its cardinality determines it completely. This is the realizing-side analogue of the prime model's uniqueness (42.02.03 pending), and the back-and-forth runs the same way with realization in place of isolation.

Theorem (Uniqueness of saturated models). Let $T$ be complete with infinite models. Any two saturated models $\mathfrak{M},\mathfrak{N}\models T$ of the same cardinality $\kappa$ are isomorphic. Moreover a saturated model of cardinality $\kappa$ is universal — every model of $T$ of cardinality $\le \kappa$ elementarily embeds into it — and strongly $\kappa$-homogeneous ^{[Marker §4.3]}.

Proof. Enumerate $M=\{m_{\alpha }:\alpha <\kappa \}$ and $N=\{n_{\alpha }:\alpha <\kappa \}$. Build an increasing chain $(f_{\alpha }{)}_{\alpha <\kappa }$ of partial elementary maps $f_{\alpha }:A_{\alpha }\to N$ with $|A_{\alpha }|<\kappa$, taking unions at limits. Set $f_0=\varnothing$, partial elementary since $\mathfrak{M}\equiv \mathfrak{N}$ ($T$ complete).

Forth. At an even stage, let $\bar{a}$ enumerate $\mathrm{dom}(f_{\alpha })$ and $\bar{b}=f_{\alpha }(\bar{a})$, and pick the least-indexed $m\in M$ not yet in the domain. The type $\mathrm{tp}(m/\bar{a})$ pushes forward to a type $q$ over $\bar{b}$; since $|\bar{b}|<\kappa$ and $\mathfrak{N}$ is $\kappa$-saturated, $q$ is realized by some $n\in N$. Then $\bar{a}m\mapsto \bar{b}n$ is partial elementary: for any $\phi$, $\mathfrak{M}\models \phi (\bar{a},m)$ iff $\phi (\bar{x},m)\in \mathrm{tp}(m/\bar{a})$ iff $\phi (\bar{y},n)$ holds via $q$ iff $\mathfrak{N}\models \phi (\bar{b},n)$. Extend $f_{\alpha }$ by $m\mapsto n$.

Back. At an odd stage, do the symmetric step from $\mathfrak{N}$ into $\mathfrak{M}$, using $\kappa$-saturation of $\mathfrak{M}$, to put the least-indexed missing $n\in N$ into the range. Since $\kappa$ is the common cardinality and each $A_{\alpha }$ has size $<\kappa$, every stage's parameter set is small enough for saturation to apply. The union $f={\bigcup }_{\alpha }{f}_{\alpha }$ is a partial elementary bijection with domain $M$ (forth exhausts $M$) and range $N$ (back exhausts $N$), hence an isomorphism $\mathfrak{M}\cong \mathfrak{N}$.

Universality: given $\mathfrak{K}\models T$ with $|K|\le \kappa$, run only the forth direction with $\mathfrak{K}$ in place of $\mathfrak{M}$ and the saturated $\mathfrak{M}$ in place of $\mathfrak{N}$; saturation of $\mathfrak{M}$ realizes each pushed-forward type, producing an elementary embedding $\mathfrak{K}\hookrightarrow \mathfrak{M}$. Strong $\kappa$-homogeneity: given a partial elementary $f$ of size $<\kappa$ between subsets of the saturated $\mathfrak{M}$, run back-and-forth from $f$ within $\mathfrak{M}$ itself; the union is an automorphism extending $f$. $\square$

Bridge. This uniqueness is the foundational reason the monster model is well-defined as the universal domain: fixing a cardinal makes the saturated model an isomorphism invariant of $T$, so working "inside the saturated model" is unambiguous. It builds toward the monster model of the Advanced tier, where strong homogeneity is the engine turning type-equality into automorphism-conjugacy, and it appears again in the categoricity theory of 42.02.06 pending, where Morley's theorem realizes the uncountable categorical model as a saturated (in fact unique) model in each uncountable power. The argument generalises the countable-atomic-model back-and-forth of 42.02.03 pending from isolation-supplied witnesses to saturation-supplied witnesses — this is exactly the prime/saturated duality, omitting maximally versus realizing maximally, run through the same machine. Putting these together, the central insight is that realization plus enough room to keep realizing forces uniqueness, and the bridge is that the same back-and-forth proves both the atomic model's uniqueness and the saturated model's, by swapping the source of witnesses.

Exercises Intermediate+

Advanced results Master

The saturated model organizes four developments: its existence under cardinal-arithmetic hypotheses and the special-model bypass, its uniqueness and universality as a homogeneous-universal domain, the monster model and the Galois-theoretic reading of types as automorphism orbits, and the weaker recursively-saturated and resplendent models that survive when full saturation is unavailable.

Theorem 1 (existence of saturated models). For complete $T$ and a cardinal $\kappa$ with $\kappa ={\kappa }^{<\kappa }$, $T$ has a saturated model of cardinality $\kappa$ ^{[Morley and Vaught 1962]}. The construction is a $\kappa$-stage elementary chain ${\mathfrak{M}}_0⪯{\mathfrak{M}}_1⪯\cdots$ in which ${\mathfrak{M}}_{\alpha +1}$ realizes every type over every subset of ${\mathfrak{M}}_{\alpha }$ of size $<\kappa$; the hypothesis $\kappa ={\kappa }^{<\kappa }$ keeps the number of such types at $\kappa$, so each stage stays of size $\kappa$, and the union is $\kappa$-saturated of cardinality $\kappa$. The condition holds for every regular $\kappa$ under GCH, and at every strongly inaccessible $\kappa$. In the countable case the criterion specializes: $T$ has a countable saturated model iff $T$ is small, meaning $S_n(A)$ is countable for every $n$ and every finite $A$ — the realizing dual of the prime-model density condition, and the same condition that the cardinal arithmetic of 42.03.03 governs at higher powers.

Theorem 2 (special models, without cardinal hypotheses). When $\kappa ={\kappa }^{<\kappa }$ fails, full saturation may be unavailable, but special models exist at every cardinal of the form $\kappa ={\beth }_{\delta }$ with $\delta$ a limit of cofinality $\omega$ (and more generally at suitable limits). A special model of cardinality $\kappa$ is an elementary union ${\bigcup }_{\lambda <\kappa }{\mathfrak{M}}_{\lambda }$ of an increasing chain with each ${\mathfrak{M}}_{{\lambda }^{+}}$ being ${\lambda }^{+}$-saturated ^{[Chang and Keisler Ch. 5]}. Special models are unique of their cardinality (back-and-forth, matching the saturation levels along the chains) and universal, and a special model is $\mathrm{cf}(\kappa )$-saturated. They are the device delivering a homogeneous-universal domain in ZFC alone — every theory has special models in arbitrarily large cardinals — and a saturated model is exactly a special model whose chain is constant.

Theorem 3 (the monster model and types as orbits). Fix $\bar{\kappa }$ larger than the size of any parameter set or model under consideration and a $\bar{\kappa }$-saturated, strongly $\bar{\kappa }$-homogeneous model $\mathfrak{C}\models T$ — the monster model ^{[Tent and Ziegler Ch. 6]}. Adopt the convention that all models are elementary submodels of $\mathfrak{C}$ of size $<\bar{\kappa }$ and all parameter sets are small ($<\bar{\kappa }$). Inside $\mathfrak{C}$ every model of $T$ of small cardinality is, up to isomorphism, an elementary submodel (by universality), every type over a small set is realized (by saturation), and for a small $A$ the group $\mathrm{Aut}(\mathfrak{C}/A)$ of automorphisms fixing $A$ pointwise acts on small tuples with $$ \mathrm{tp}(\bar a / A) = \mathrm{tp}(\bar b / A) \quad\Longleftrightarrow\quad \exists, \sigma \in \mathrm{Aut}(\mathfrak{C}/A)\ \ \sigma(\bar a) = \bar b. $$ Types over $A$ are exactly the orbits of $\mathrm{Aut}(\mathfrak{C}/A)$ on realizations — the Galois-theoretic dictionary. The set-theoretic caveat is real: a literal saturated proper-class universal domain needs a strongly inaccessible $\bar{\kappa }$ or the treatment of $\mathfrak{C}$ as a class model; the working convention is to fix $\bar{\kappa }$ beyond everything in sight and never to leave $\mathfrak{C}$. The definable and algebraic closures of $A$ are the small sets fixed setwise or with small orbit under $\mathrm{Aut}(\mathfrak{C}/A)$, recasting model-theoretic closure as fixed-point data of the automorphism group.

Theorem 4 (recursively saturated and resplendent models). When saturation in a target cardinality is unattainable, two weaker realizing properties remain. A countable model is recursively saturated when it realizes every recursive type (every type whose defining set of formulas is computably enumerable) over every finite parameter set; every consistent countable theory has a countable recursively saturated model, and these models support a fragment of the saturated-model arguments (overspill, coding) in the countable world. A model is resplendent when every consistent type in an expansion of the language by finitely many new symbols, with parameters from the model, is realized by an expansion of the model; resplendent and recursively saturated coincide for countable models. These bridge the gap to the monster: where $\mathfrak{C}$ realizes everything consistent, recursively saturated models realize everything consistent and effectively presented, which is enough for the countable structure theory and for nonstandard models of arithmetic.

Synthesis. The saturated model is the foundational reason model theory has a universal domain, and putting these together it governs all four developments through one principle — maximal realization plus enough symmetry forces uniqueness. This is exactly what Theorem 1's existence and the uniqueness theorem make precise: under $\kappa ={\kappa }^{<\kappa }$ the type-realizing elementary chain produces a saturated model, and back-and-forth shows it is the unique homogeneous-universal model of its power, generalising the countable atomic/saturated uniqueness of 42.02.03 pending from isolation-supplied to saturation-supplied witnesses. The special models of Theorem 2 are dual to the cardinal-arithmetic obstruction: where saturation fails for arithmetic reasons, the chain of increasing saturation levels recovers a universal domain in ZFC, and a saturated model is the central insight's limiting case, a special model with constant chain.

The monster model of Theorem 3 is the realizing extreme made into a standing arena — the bridge is that strong homogeneity converts type-equality into automorphism-conjugacy, so that types become orbits and the closure operators become fixed-point data, which is exactly the symmetry and transitivity that the stability theory of 42.02.07 pending exploits to define forking and independence. Recursively saturated and resplendent models of Theorem 4 are dual to the monster in the countable world, realizing the effectively-presentable fragment of what $\mathfrak{C}$ realizes globally; this is the bridge from the abstract universal domain back to the concrete countable and arithmetic structures. The whole apparatus is the realizing mirror of the omitting machinery of 42.02.03 pending — omit maximally to get the prime model, realize maximally to get the saturated and monster models — and it is the substrate on which the categoricity theorem of 42.02.06 pending (Morley's uncountably categorical model is saturated in each uncountable power) and the stability theory of 42.02.07 pending are built.

Full proof set Master

Proposition 1 ($\kappa$-saturation reduces to $1$-types). If $\mathfrak{M}$ realizes every complete $1$-type over every $A\subseteq M$ with $|A|<\kappa$, then $\mathfrak{M}$ realizes every complete $n$-type over every such $A$.

Proof. Induct on $n$. Let $p(\bar{x})=p(x_1,\dots ,x_n)\in S_n(A)$. Its restriction $p\upharpoonright x_1$ to the first variable is a $1$-type over $A$, realized by some $a_1\in M$ by hypothesis. The set $p_1(x_2,\dots ,x_n)=\{\phi (a_1,x_2,\dots ,x_n):\phi (x_1,\dots ,x_n)\in p\}$ is a complete $(n-1)$-type over $A\cup \{a_1\}$, consistent because $a_1$ realizes $p\upharpoonright x_1$ and $p$ is consistent; and $|A\cup \{a_1\}|<\kappa$. By the induction hypothesis it is realized by ${\bar{a}}^{\prime }=(a_2,\dots ,a_n)$, and $(a_1,{\bar{a}}^{\prime })$ realizes $p$. $\square$

Proposition 2 (saturated $\Rightarrow$ homogeneous and universal). A saturated model $\mathfrak{M}$ of cardinality $\kappa$ is $\kappa$-homogeneous and universal for models of cardinality $\le \kappa$.

Proof. Homogeneity: given partial elementary $f:\bar{a}\mapsto \bar{b}$ with $|\mathrm{dom}(f)|<\kappa$ and $a\in M$, the type $\mathrm{tp}(a/\bar{a})$ pushed forward along $f$ is a type over $\bar{b}$, a set of size $<\kappa$, so $\kappa$-saturation realizes it by some $b$; then $f\cup \{a\mapsto b\}$ is partial elementary, as the pushed-forward type was defined to make $\phi (\bar{a},a)\leftrightarrow \phi (\bar{b},b)$ hold for all $\phi$. Universality: for $\mathfrak{K}\models T$ with $|K|\le \kappa$, enumerate $K$ in order type $\le \kappa$ and build a one-directional elementary chain of partial maps into $\mathfrak{M}$, at each stage realizing the pushed-forward type of the next element over the current small range using $\kappa$-saturation; the union is an elementary embedding $\mathfrak{K}\hookrightarrow \mathfrak{M}$. $\square$

Proposition 3 (uniqueness of saturated models). Two saturated models $\mathfrak{M},\mathfrak{N}\models T$ of the same cardinality $\kappa$ are isomorphic.

Proof. As in the Key-theorem proof: enumerate both universes in order type $\kappa$, build an increasing chain of partial elementary maps of size $<\kappa$ starting from $\varnothing$ (elementary by completeness of $T$), alternating forth (least missing element of $M$, type realized in $\mathfrak{N}$ by $\kappa$-saturation) and back (least missing element of $N$, type realized in $\mathfrak{M}$). Each stage's domain and range have size $<\kappa$, so saturation applies; unions are taken at limits. The resulting map has domain $M$ and range $N$ and is partial elementary, hence an isomorphism. $\square$

Proposition 4 (existence under $\kappa ={\kappa }^{<\kappa }$). If $\kappa ={\kappa }^{<\kappa }$, $T$ has a saturated model of cardinality $\kappa$.

Proof. Start with ${\mathfrak{M}}_0\models T$ of cardinality $\kappa$ (Löwenheim–Skolem, 42.02.01 pending). Given ${\mathfrak{M}}_{\alpha }$ of size $\kappa$, the number of subsets of size $<\kappa$ is ${\kappa }^{<\kappa }=\kappa$, and over each such subset the number of types is at most $2^{<\kappa }\le {\kappa }^{<\kappa }=\kappa$, so there are $\kappa$-many types to realize; build ${\mathfrak{M}}_{\alpha +1}⪰{\mathfrak{M}}_{\alpha }$ realizing all of them (an elementary extension exists by the diagram method of 42.02.02 pending, adding $\kappa$ new realizing elements), keeping $|{\mathfrak{M}}_{\alpha +1}|=\kappa$. At limits take unions. Let $\mathfrak{M}={\bigcup }_{\alpha <\kappa }{\mathfrak{M}}_{\alpha }$, of cardinality $\kappa$. For $A\subseteq M$ with $|A|<\kappa$ and $\mathrm{cf}(\kappa )=\kappa$ (regular, as $\kappa ={\kappa }^{<\kappa }$ forces regularity), $A\subseteq {\mathfrak{M}}_{\alpha }$ for some $\alpha <\kappa$, so any type over $A$ is realized in ${\mathfrak{M}}_{\alpha +1}\subseteq \mathfrak{M}$. Thus $\mathfrak{M}$ is $\kappa$-saturated of cardinality $\kappa$, i.e. saturated. $\square$

Proposition 5 (types are orbits in the monster). Let $\mathfrak{C}$ be $\bar{\kappa }$-saturated and strongly $\bar{\kappa }$-homogeneous, $A\subseteq \mathfrak{C}$ small, and $\bar{a},\bar{b}$ small tuples. Then $\mathrm{tp}(\bar{a}/A)=\mathrm{tp}(\bar{b}/A)$ iff some $\sigma \in \mathrm{Aut}(\mathfrak{C}/A)$ has $\sigma (\bar{a})=\bar{b}$.

Proof. ($\Leftarrow$) An automorphism fixing $A$ pointwise preserves every ${\mathcal{L}}_A$-formula, so $\sigma (\bar{a})=\bar{b}$ gives $\mathfrak{C}\models \phi (\bar{a})\leftrightarrow \phi (\bar{b})$ for all such $\phi$, i.e. equal types. ($\Rightarrow$) The map $f={\mathrm{id}}_A\cup \{\bar{a}\mapsto \bar{b}\}$ is partial elementary: on $A$ it is the identity, and for any ${\mathcal{L}}_A$-formula $\phi$, $\mathfrak{C}\models \phi (\bar{a})⟺\phi \in \mathrm{tp}(\bar{a}/A)=\mathrm{tp}(\bar{b}/A)⟺\mathfrak{C}\models \phi (\bar{b})$. Its domain $A\cup \bar{a}$ is small ($<\bar{\kappa }$), so strong $\bar{\kappa }$-homogeneity extends $f$ to an automorphism $\sigma$ of $\mathfrak{C}$; $\sigma$ fixes $A$ pointwise and $\sigma (\bar{a})=\bar{b}$. $\square$

Proposition 6 (countable recursively saturated models exist). Every consistent theory $T$ in a countable language has a countable recursively saturated model.

Proof sketch. Build a countable model by a Henkin construction in which, alongside the usual completeness and witness requirements, one interleaves recursive-type-realization requirements: enumerate all pairs $(\bar{c},p)$ with $\bar{c}$ a finite tuple of Henkin constants and $p$ a recursive type over $\bar{c}$ (the recursive types over finite tuples are countably many, being indexed by their computable enumerations). At the step for $(\bar{c},p)$, if $p$ is consistent with the theory built so far together with the diagram of $\bar{c}$, add a fresh constant $d$ and the formulas of $p$ with $d$ in place of the realized variable — consistency is preserved because $p$ is finitely satisfiable over $\bar{c}$ in the current theory and a fresh witness can be adjoined. The term model is countable and realizes every recursive type over every finite parameter set, hence is recursively saturated. (Full saturation fails for cardinality reasons; the recursive restriction is what keeps the requirement list countable.) Resplendence of the resulting countable model follows from recursive saturation by the standard equivalence for countable structures. $\square$

Connections Master

• Types and the Omitting Types Theorem 42.02.03 pending is the omitting dual this unit inverts. There the prime model is built by omitting every non-isolated type; here the saturated model is built by realizing every type over every small parameter set, and the countable saturated model exists exactly when $T$ is small — the realizing mirror of the prime model's isolated-types-dense condition. The Stone space $S_n(A)$, the realization predicate, and the back-and-forth method are imported wholesale; this unit runs the same machine to realize maximally rather than to omit maximally, and the two extremes coincide precisely in the ${\aleph }_0$-categorical case where prime equals saturated.

• Cardinal exponentiation and the continuum function 42.03.03 supplies the arithmetic that controls existence. The hypothesis $\kappa ={\kappa }^{<\kappa }$ under which a saturated model of size $\kappa$ exists, the GCH instances that make it hold at every successor, and the $\beth$-fixed-point cofinality conditions for special models are statements about cardinal exponentiation; when the arithmetic fails, the special-model construction of this unit is the ZFC-safe substitute, and the set-theoretic caveat on the monster (a strongly inaccessible $\bar{\kappa }$, or a class model) is a large-cardinal question fixed there.

• Categoricity and Morley's theorem, co-produced as 42.02.06 pending, presupposes the saturation theory of this unit. Morley's theorem realizes the unique uncountably categorical model as a saturated model in each uncountable power, and the upward and downward transfer of categoricity is run through the existence and uniqueness of saturated models established here; the two-cardinal theorems and the Morley-rank analysis both operate inside a saturated or monster model where every relevant type is realized.

• Stability and forking, co-produced as 42.02.07 pending, is developed entirely inside the monster model of this unit. The symmetry and transitivity of $\mathrm{Aut}(\mathfrak{C}/A)$ acting on types — types as orbits, definable and algebraic closure as fixed-point data — is the substrate on which forking independence, the definability of types over models, and the independence theorem are defined; stability theory takes the monster as its standing universe and the saturation/homogeneity machinery as its working hypotheses throughout.

Historical & philosophical context Master

The systematic theory of saturated models entered model theory in 1962 with Michael Morley and Robert Vaught's "Homogeneous universal models," which abstracted two earlier constructions to arbitrary first-order theories: Bjarni Jónsson's homogeneous-universal relational systems (1960) and Roland Fraïssé's amalgamation construction of the countable ultrahomogeneous limit of a class of finite structures ^{[Morley and Vaught 1962]}. Morley and Vaught defined $\kappa$-saturation, proved existence of saturated models under the cardinal-arithmetic hypothesis $\kappa ={\kappa }^{<\kappa }$ (so under the generalized continuum hypothesis for regular $\kappa$), and established uniqueness up to isomorphism by the back-and-forth method, identifying the saturated model with the unique homogeneous-universal model of its cardinality. The construction subsumed Felix Hausdorff's ${\eta }_{\alpha }$-orderings and Pavel Urysohn's universal metric space as instances of a single model-theoretic phenomenon.

The special-model device, realizing a homogeneous-universal domain in ZFC without cardinal-arithmetic assumptions through an elementary union of models of increasing saturation, was developed by Morley, Vaught, and H. Jerome Keisler and codified in Chang and Keisler's Model Theory ^{[Chang and Keisler Ch. 5]}. The monster model — a single $\bar{\kappa }$-saturated, strongly $\bar{\kappa }$-homogeneous universal domain inside which all models are elementary submodels and types are orbits of the automorphism group — became the standing convention of stability theory through the work of Saharon Shelah, and its Galois-theoretic reading of types as $\mathrm{Aut}(\mathfrak{C}/A)$-orbits, with definable and algebraic closure as fixed-point data, is the framework of the modern treatments by Tent and Ziegler and by Pillay ^{[Tent and Ziegler Ch. 6]}. The set-theoretic caveat that a literal saturated proper-class universal domain requires either a strongly inaccessible cardinal or the treatment of the monster as a class model traces to this period, and the recursively saturated and resplendent models that recover a countable fragment of the theory were introduced by Jon Barwise, John Schlipf, and Andrzej Ehrenfeucht in the 1970s.

Bibliography Master

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

@book{changkeisler1990,
  author    = {Chang, Chen Chung and Keisler, H. Jerome},
  title     = {Model Theory},
  edition   = {3},
  series    = {Studies in Logic and the Foundations of Mathematics},
  volume    = {73},
  publisher = {North-Holland},
  year      = {1990}
}

@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 and Cambridge University Press},
  year      = {2012}
}

@book{hodges1993modeltheory,
  author    = {Hodges, Wilfrid},
  title     = {Model Theory},
  series    = {Encyclopedia of Mathematics and its Applications},
  volume    = {42},
  publisher = {Cambridge University Press},
  year      = {1993}
}

@article{morleyvaught1962,
  author  = {Morley, Michael and Vaught, Robert L.},
  title   = {Homogeneous universal models},
  journal = {Mathematica Scandinavica},
  volume  = {11},
  year    = {1962},
  pages   = {37--57}
}

@article{jonsson1960,
  author  = {J{\'o}nsson, Bjarni},
  title   = {Homogeneous universal relational systems},
  journal = {Mathematica Scandinavica},
  volume  = {8},
  year    = {1960},
  pages   = {137--142}
}

@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}
}