An introduction to separated graphs and their type semigroups

Abstract: We introduce $C^*$-algebras associated with directed graphs, along with two generalizations of this concept, namely Exel-Pardo $C^*$-algebras associated with a self-similar action of a group on a directed graph, and the $C^*$-algebras associated with separated graphs. These constructions have in common that they have a dynamical behavior, being the groupoid $C^*$-algebras associated to certain topological groupoids, which are built from the combinatorial structure. An important invariant one may associate to these dynamical systems is the so-called type semigroup. We will find a formula to compute the type semigroup for a general self-similar action of a group on a row-finite graph $E$ without sources, following a paper by Kwaśniewski, Meyer and Prasad, and for any finite bipartite separated graph, following a paper by Exel and the author. In addition, we will review various results concerning the structure of the type semigroup for different dynamical systems.

• The paper introduces separated graphs and their type semigroups by extending classical graph C*-algebra theory and unifying combinatorial with dynamical approaches.
• It details the construction of separated graph C*-algebras using modified Cuntz-Krieger relations and explicit computations via Bratteli resolutions and groupoid models.
• The work examines how type semigroups capture paradoxical decompositions and inform the dichotomy between pure infiniteness and stable finiteness in operator algebras.

Introduction to Separated Graphs and Type Semigroups

This essay provides an exposition of the fundamental concepts and structural results surrounding $C^*$-algebras associated with separated graphs and their corresponding type semigroups, as articulated in "An introduction to separated graphs and their type semigroups" (2604.18304). The discussion synthesizes standard graph $C^*$-algebra theory, self-similar and groupoid generalizations, and places special emphasis on the combinatorial and dynamical nature of type semigroups—algebraic invariants that capture complex phenomena such as paradoxical decompositions. Practical implications, theoretical connections to structural dichotomies, and the current landscape of open questions in the area are addressed.

Graph $C^*$-Algebras and the Cuntz-Krieger Paradigm

Graph $C^*$-algebras are defined from directed graphs via generators (projections and partial isometries) subject to Cuntz-Krieger relations. The construction serves as an interface between combinatorial data and operator algebraic structures, generalizing both Cuntz-Krieger algebras and Leavitt path algebras. The universal property with respect to $E$-families is central, and the closure under norm yields the analytic $C^*$-algebra $C^*(E)$. Simplicity of $C^*(E)$ is governed by the cofinality of the underlying graph and the absence of cycles without entries, paralleling the classification program in the $C^*$-algebraic context.

A combinatorial monoid $M(E)$ is associated with the graph and is isomorphic to the Murray-von Neumann monoid of projections $C^*$0. The associated $C^*$1-theory is computable using the reduced adjacency matrix, providing explicit formulas for $C^*$2 and $C^*$3 in terms of graph-theoretic data.

Groupoid models further enhance the analytic framework: $C^*$4 can be realized as a groupoid $C^*$5-algebra $C^*$6, rendering the dynamics explicit via topological groupoids with clopen base sets tied to paths in the graph.

Generalizations: Self-Similar Actions and Exel-Pardo Algebras

Exel-Pardo $C^*$7-algebras extend the graph algebra framework to encompass self-similar group actions on directed graphs, incorporating additional cocycle data to encode the "self-similarity". Given $C^*$8—a group, a finite graph, and a 1-cocycle satisfying constrained compatibility—a $C^*$9-algebra $C^*$0 is defined via an extended set of relations that realize both the graph and group symmetries.

For pseudo-free self-similar actions, the associated groupoid $C^*$1 is Hausdorff. These models unify and extend self-similar group algebras (cf. Nekrashevych), Katsura algebras, and allow computation of invariants under group actions with nontrivial combinatorics. Notably, the type semigroup $C^*$2 for such groupoids is shown to be the monoid of coinvariants $C^*$3, connecting the algebraic and dynamical perspectives.

Separated Graph $C^*$4-Algebras: Construction and Universal Properties

A separated graph is a pair $C^*$5 where $C^*$6 is a partition of the edge set at each vertex, introducing fine-grained control over the relations imposed in the algebra. The separated graph $C^*$7-algebra $C^*$8 is defined via generators and modified Cuntz-Krieger relations that encode the partitioning.

The precise monoid $C^*$9 associated to $C^*$0, and its relationship to finitely generated projective modules and projections in $C^*$1, is clarified. However, for $C^*$2 the correspondence between the monoid at the level of the Leavitt path algebra and at the level of the $C^*$3-algebra is not fully resolved, with a canonical conjecture (conjecture of Goodearl-Ara) remaining open. In contrast, for the so-called "tame" quotient algebra $C^*$4, constructed by enforcing commutativity of final projections, a groupoid model is available, and the corresponding type semigroup admits a precise combinatorial description.

Bipartite separated graphs and their canonical Bratteli resolutions offer a mechanism to encode dynamical systems and paradoxical decompositions, supporting a reduction theory by Morita equivalence.

Type Semigroups: Definition, Computation, and Dynamical Implications

The type semigroup $C^*$5 is defined for an ample Hausdorff groupoid as the commutative semigroup generated by classes $C^*$6 of clopen subsets of the unit space, subject to additivity and invariance under bisections. There is a canonical map to the projection monoid $C^*$7, which, in favorable cases, is an isomorphism.

The semigroup encodes dynamical phenomena: its pre-order reflects (non-)paradoxical decomposability (Tarski’s theory). For topological dynamical systems, such as group actions on the Cantor set, $C^*$8 formalizes the possibility of paradoxical decompositions of open compact sets. Tarski's theorem characterizes the dichotomy between the existence of invariant measures and paradoxical decompositions in semigroup-theoretic terms. For groupoid $C^*$9-algebras, the structure of the type semigroup determines dichotomies (pure infiniteness versus stable finiteness) at the operator algebra level.

The monoid $E$0 for self-similar group actions is always a tame refinement monoid, with explicit Bratteli diagram constructions (canonical resolutions) available for separated graphs, enabling explicit computation even in combinatorially complex cases.

Cancellation and Refinement in Type Semigroups

Multiple "levels of cancellation" are considered: cancellation, separativity, and strong separativity, with structural implications in the simple setting (cf. dimension groups versus purely infinite objects). Ample groupoid type semigroups are always conical refinement monoids. For supramenable group actions, the associated type semigroup is strongly separative, precluding paradoxical decompositions. Minimal actions of amenable groups on the Cantor set are conjectured, but not known, to always yield cancellative type semigroups.

Examples are detailed showcasing wild refinement monoids (non-cancellative, stably finite, not direct limits of free monoids), constructed from symbolic dynamics (e.g., the two-sided full shift), where direct computation exhibits non-cancellation phenomena.

Realization, Embedding, and Open Problems

A central realization result, due to Wehrung, asserts that every countable conical refinement monoid occurs as the type semigroup for an action of a countable group on a totally disconnected, locally compact, zero-dimensional space. Minimality can be arranged via embedding the monoid into a simple refinement monoid.

Open problems persist: for instance, whether there exist minimal $E$1-dynamical systems (in the sense of the $E$2-paradox) such that the type semigroup contains an $E$3-paradoxical element; or whether the type semigroup of minimal actions of amenable groups on the Cantor set is always cancellative.

Conclusion

The theory of separated graph $E$4-algebras and their type semigroups articulates a deep interplay between combinatorial graph invariants, dynamics of (partial) group actions, and operator algebraic classification. The development of groupoid models and explicit semigroup computations enables analysis of fine structure (e.g., cancellation properties, dichotomy principles) guiding classification programs in symbolic dynamics and noncommutative topology. Algebraic realization theorems render the field both robust and flexible, yet subtle open questions (including realization under minimality and the extent of dichotomy phenomena) remain, inviting continued research on the algebra-dynamics correspondence and the topology of paradoxicality.