Fragments and Fusion

What this lesson covers

How the semiotic universe achieves coherence between its syntax and semantics. Fragments are finite substructures that enable local reasoning; fusion is the closure process that identifies syntactic operators agreeing on all fragments and names new operators for behaviors already present in the system. The three closure operators — semantic, syntactic, and fusion — compose into a global operator whose least fixed point defines the semiotic universe itself.

Prerequisites

The Semantic Domain and Syntactic Operators. This lesson uses concepts from both.

---

Fragments

A fragment is a finite piece of the semantic domain that is self-contained with respect to all the structure we care about. Formally, a subset $F\subseteq H$ is a modal–temporal subalgebra if it is closed under:

1. The Heyting operations: $\wedge$, $\vee$, $\Rightarrow$, $\perp$, $\top$.
2. The modality $j$.
3. The trace $G$.

A fragment is a finitely generated modal–temporal subalgebra: there exists a finite set $S\subseteq F$ such that $F$ is the smallest modal–temporal subalgebra containing $S$. The set of all fragments is denoted $\mathcal{F}(H)$.

Fragments serve the same purpose that compact objects serve in algebraic theories: they allow global properties to be checked locally. If two operators agree on every fragment, they agree everywhere that matters. If an equality holds on a fragment and is preserved by all the structure’s operations, it holds on every fragment reachable from the first.

Fragment-preserving operators

An operator $f:H^n\to H$ is fragment-preserving if $f(F^n)\subseteq F$ for every fragment $F\in \mathcal{F}(H)$. That is, applying the operator to elements of a fragment produces elements still in that fragment. Fragment-preserving operators do not break locality — they keep finite substructures closed.

This is one of the seven conditions imposed on the interpretation of syntactic operators (condition 6 in Syntactic Operators). It is not derived from the other conditions but imposed directly: the semiotic universe requires that every syntactic operation respect the fragment structure.

Fragmentwise equality

Two fragment-preserving operators $f,g:H^n\to H$ are fragmentwise equal on $F$, written $f{\equiv }_{F}g$, if they agree on all inputs from $F$:

$$f(a_1,\dots ,a_n)=g(a_1,\dots ,a_n)\text{for all }{a}_i\in F.$$

A family of fragment-preserving operators is hereditarily extensional if fragmentwise equality propagates: whenever $f{\equiv }_{F}g$ on some fragment $F$, the equality persists on every fragment $F^{\prime }$ obtained from $F$ by applying Heyting operations, $j$, $G$, or any operator in the family. Equalities, once established locally, cannot be broken by further construction.

This hereditary extensionality condition is the seventh requirement on the interpretation (condition 7 in Syntactic Operators). It ensures that fragmentwise reasoning is stable — a local observation about operator behavior remains valid as the structure grows.

Fusion

Fusion is the process that enforces coherence between syntax and semantics. It does two things:

Quotienting by fragmentwise congruence. Given a partial semiotic structure $X=(H_X,{\mathrm{Op}}_X)$ — a set of semantic objects and a set of syntactic operators — fusion identifies operators that agree on every fragment within $H_X$. If two syntactic operators $f$ and $g$ produce the same results on every finite substructure, they are treated as the same operator. The congruence relation ${\sim }_X$ formalizes this: $f{\sim }_{X}g$ when $[[f]]{\equiv }_F[[g]]$ for every fragment $F\subseteq H_X$, hereditarily.

Naming existing behaviors. If a function $h:H^n\to H$ is already present in the semantic domain — monotone, join-continuous, fragment-preserving, compatible with $j$ and $G$ — but has no syntactic name, fusion introduces one. A new operator symbol $f_h$ is added with $[[f_h]]=h$. This does not expand what the system can compute; it expands what the system can say. Every admissible semantic behavior gets a syntactic representative.

The fusion operator ${\mathcal{S}}_{\mathrm{fus}}$ leaves the semantic objects unchanged and saturates the operator set:

$${\mathcal{S}}_{\mathrm{fus}}(H_X,{\mathrm{Op}}_X)=(H_X,{\mathrm{Op}}_X^{\mathrm{fus}}).$$

This operator is monotone (enlarging inputs enlarges outputs), inflationary ($X\le {\mathcal{S}}_{\mathrm{fus}}(X)$), and idempotent (applying it twice gives the same result as applying it once). In other words, fusion is a closure operator on the lattice of partial semiotic structures.

Idempotence is the key property. Once the operator set is fusion-saturated — all fragmentwise identifications made, all admissible behaviors named — doing it again produces nothing new. The fixed points of ${\mathcal{S}}_{\mathrm{fus}}$ are the fusion-saturated structures.

Fusion as a reflection

Fusion is more than just a closure: it is a reflector into the subcategory of fusion-saturated structures. The set of all fusion-saturated structures ${\mathcal{U}}_{\mathrm{fus}}$ forms a reflective subcategory of the lattice of partial semiotic structures $\mathcal{U}$, and ${\mathcal{S}}_{\mathrm{fus}}$ is the left adjoint to the inclusion:

$${\mathcal{S}}_{\mathrm{fus}}(X)\le Y\text{if and only if}X\le Y$$

for every fusion-saturated $Y$. This is a standard fact about closure operators on complete lattices (Tarski, 1955), but it has a specific meaning here: any partial structure can be “freely” completed to a fusion-saturated one, and the result is the smallest fusion-saturated structure containing the original.

The three closure operators

The semiotic universe is constructed by iterating three closure operators on the lattice $\mathcal{U}=\mathcal{P}(H)\times \mathcal{P}({\mathrm{Op}}^{\mathrm{def}})$ of partial semiotic structures:

Semantic closure ${\mathcal{S}}_{\mathrm{sem}}$: expands the set of semantic objects. Starting from $H_X$, it closes under: the interpretations of all operators in ${\mathrm{Op}}_X$, all Heyting operations, the modality $j$, the trace $G$, and the least and greatest fixed points of admissible semantic endomorphisms. The operator set stays the same.

$${\mathcal{S}}_{\mathrm{sem}}(H_X,{\mathrm{Op}}_X)=(H_X^{\prime },{\mathrm{Op}}_X)$$

Syntactic closure ${\mathcal{S}}_{\mathrm{syn}}$: expands the set of operators. Starting from ${\mathrm{Op}}_X$, it closes under: lambda-definability, composition, fixed-point constructs with admissible interpretations, and — critically — finitary justification. An operator $f$ is included if there exists some fragment $F\subseteq H_X$ and some existing operator $g\in {\mathrm{Op}}_X$ such that $[[f]]{\equiv }_F[[g]]$ hereditarily. This is the compactness condition: syntactic closure is driven by finite evidence. The semantic objects stay the same.

$${\mathcal{S}}_{\mathrm{syn}}(H_X,{\mathrm{Op}}_X)=(H_X,{\mathrm{Op}}_X^{\prime })$$

Fusion ${\mathcal{S}}_{\mathrm{fus}}$: as described above — quotients operators by fragmentwise congruence and names admissible behaviors.

The global semiotic closure is their composition:

$$\mathcal{S}={\mathcal{S}}_{\mathrm{fus}}\circ {\mathcal{S}}_{\mathrm{syn}}\circ {\mathcal{S}}_{\mathrm{sem}}.$$

Each of the three is monotone and inflationary, so their composition is too. And since $\mathcal{U}$ is a complete lattice, the Knaster–Tarski fixed-point theorem guarantees that $\mathcal{S}$ has a least fixed point (Tarski, 1955).

The least fixed point

The semiotic universe is this least fixed point:

$${\mathcal{U}}_{[\mathrm{semiotic}](../../../../../linguistics/topics/semiotics/index.md)}=X^{\star }=(H_{[\mathrm{semiotic}](../../../../../linguistics/topics/semiotics/index.md)},{\mathrm{Op}}_{[\mathrm{semiotic}](../../../../../linguistics/topics/semiotics/index.md)}).$$

At the fixed point, every closure step has been exhausted:

• $H_{[\mathrm{semiotic}](../../../../../linguistics/topics/semiotics/index.md)}$ is closed under Heyting operations, $j$, $G$, all interpreted operators, and fixed points of admissible endomorphisms — semantic closure adds nothing.
• ${\mathrm{Op}}_{[\mathrm{semiotic}](../../../../../linguistics/topics/semiotics/index.md)}$ is closed under lambda-definability, composition, fixed-point constructs, and finitary justification — syntactic closure adds nothing.
• ${\mathrm{Op}}_{[\mathrm{semiotic}](../../../../../linguistics/topics/semiotics/index.md)}$ is fusion-saturated — all fragmentwise identifications have been made, all admissible behaviors have names.

The construction proceeds from an initial base structure $X_0=(H_0,{\mathrm{Op}}^{\mathrm{prim}})$ — a generating set for the Heyting–modal–comonadic structure and a set of primitive operators with valid interpretations — by iterating $\mathcal{S}$ until the fixed point is reached.

What fusion provides

The interplay of the three closures produces a structure where syntax and semantics are in full agreement:

• Every syntactic combination has a determinate semantic meaning (by interpretation).
• Every semantic behavior that can be constructed from the available operations has a syntactic name (by fusion).
• Operators that compute the same function on every finite substructure are identified (by fragmentwise congruence).
• All of this is compatible with the modal structure ($j$), the trace structure ($G$), and the fragment structure (locality).

The final lesson — The Semiotic Universe — shows that this fixed point has a universal property: it is the initial object in a 2-category of semiotic structures, making it the free modal–comonadic Heyting structure with interpreted syntax.

Tarski, A. (1955). A Lattice-Theoretical Fixpoint Theorem and Its Applications. Pacific Journal of Mathematics, 5(2), 285–309.