Igusa–Todorov Discrete Cluster Categories

• Igusa–Todorov discrete cluster categories are K-linear, Krull–Schmidt, 2-Calabi–Yau triangulated categories defined by arcs in a discrete infinite subset of the circle.
• They employ geometric extension theory where crossing arcs encode extensions, enabling explicit control over t-structures, torsion pairs, and triangulations.
• The framework extends to completions and negative Calabi–Yau analogues, offering novel insights in homological generation, mutation phenomena, and cluster-tilting classifications.

Igusa–Todorov discrete cluster categories are the triangulated categories $C(Z)$ attached to an admissible discrete infinite subset $Z\subseteq S^1$ with finitely many two-sided limit points. In the type-$A$ situation, $C(Z)$ is a $K$-linear, Krull–Schmidt, triangulated, $2$-Calabi–Yau category whose indecomposable objects are arcs of $Z$; crossing of arcs encodes extensions, and the resulting combinatorics is that of non-crossing arcs, triangulations, and non-crossing partitions (Gratz et al., 2021). They form an infinite but discrete counterpart of the familiar finite type-$A$ cluster categories, while retaining an explicit surface model and a remarkably rigid combinatorial control over torsion pairs, t-structures, thick subcategories, and cluster-tilting phenomena.

1. Admissible sets, arcs, and the basic category

The standard setup fixes a discrete infinite subset $Z\subseteq S^1$ with finitely many limit points

$L(Z)=\overline Z\setminus Z,\qquad |L(Z)|=n>0,$

each assumed two-sided. Writing the limit points cyclically as

$Z\subseteq S^1$0

the intervals $Z\subseteq S^1$1 are the $Z\subseteq S^1$2 sectors between consecutive limit points. An arc of $Z\subseteq S^1$3 is a $Z\subseteq S^1$4-element subset $Z\subseteq S^1$5 whose endpoints are distinct and not immediate neighbors. Igusa–Todorov’s category $Z\subseteq S^1$6 has exactly these arcs as indecomposable objects, and the suspension acts by moving endpoints one step clockwise,

$Z\subseteq S^1$7

The extension theory is geometric: $Z\subseteq S^1$8 This makes $Z\subseteq S^1$9 a discrete type-$A$0 cluster category in the geometric sense of arcs in a marked disc. The phrase “of Dynkin type $A$1” refers to the fact that the combinatorics is that of type $A$2: non-crossing arcs, triangulations, and non-crossing partitions. Unlike finite cluster categories of type $A$3, these categories have infinitely many indecomposables, non-periodic suspension, and nontrivial t-structures (Gratz et al., 2021).

2. Triangulations and cluster-tilting phenomena

In the uncompleted setting, the geometric dictionary extends from indecomposables and extensions to cluster combinatorics. One convenient notation for the underlying marked surface is the unpunctured $A$4-gon $A$5, a closed disk whose boundary marked set has $A$6 two-sided accumulation points that are not themselves marked; this is precisely the type-$A$7 discrete infinite-rank setting used as a template for later developments (Mohammadi et al., 2022). In this uncompleted case, every triangulation corresponds to a weak cluster-tilting subcategory, and the extension rule is the clean crossing rule

$A$8

Together with the exchange triangles attached to quadrilateral configurations, this yields the familiar “triangulations $A$9 weak cluster-tilting subcategories” correspondence in the discrete Igusa–Todorov categories (Çanakçı et al., 2024).

This arc model is one of the decisive structural features of the theory. It allows one to treat cluster-tilting and mutation questions by explicit combinatorics rather than by abstract approximation theory alone. A plausible implication is that the discrete nature of the marked set, together with the two-sided limit-point hypothesis, is what keeps the infinite-rank geometry tractable without collapsing it to the finite case.

3. T-structures, thick subcategories, and non-crossing partition lattices

A major structural result is the complete classification of t-structures and thick subcategories in the discrete type-$C(Z)$0 categories $C(Z)$1. A t-structure on a triangulated category $C(Z)$2 is, in this setting, a torsion pair $C(Z)$3 such that $C(Z)$4 is closed under suspension; then

$C(Z)$5

Hence classification reduces to the aisles. Gratz–Holm–Jørgensen’s earlier classification of torsion classes shows that a set of arcs defines a torsion class if and only if it satisfies the conditions PC and PTO; for aisles one further requires closure under clockwise rotation, corresponding to closure under $C(Z)$6 (Gratz et al., 2021).

The resulting parametrization is by $C(Z)$7-decorated non-crossing partitions. If $C(Z)$8 is a non-crossing partition of $C(Z)$9 and $K$0 is a decoration satisfying the sectorwise endpoint conditions determined by singleton and adjacency blocks, then

$K$1

is an aisle, and every aisle arises uniquely in this way. The order on t-structures is inclusion of aisles, and in these categories the full set of t-structures forms a genuine lattice. If $K$2 and $K$3 correspond to two t-structures, then

$K$4

$K$5

and the meet is literally the intersection of aisles: $K$6 This is a strong structural statement, because in general triangulated categories intersections of aisles need not be aisles, and t-structures need not form a lattice at all (Gratz et al., 2021).

The same paper classifies thick subcategories by non-exhaustive non-crossing partitions $K$7. For

$K$8

one has a lattice isomorphism

$K$9

The t-structure classification also identifies the coaisle via Kreweras complement, gives an explicit description of the heart, and shows that every heart is semisimple: $2$0 Non-degenerate t-structures are exactly those with every $2$1; equivalently, they are exactly those whose heart has $2$2 simples. For $2$3, bounded-above t-structures are characterized by the coarsest partition, bounded-below t-structures by the finest partition, and therefore there are no bounded t-structures in $2$4 for $2$5. Passing to Neeman’s equivalence relation, the lattice of equivalence classes of non-degenerate t-structures is isomorphic to the ordinary non-crossing partition lattice $2$6 (Gratz et al., 2021).

4. Completions, limit arcs, and completed exact structures

A natural enlargement is the Paquette–Yıldırım completion. Starting from a discrete cluster category $2$7, one inserts $2$8-indexed chains of marked points near each accumulation point, embeds into a larger discrete cluster category $2$9, and then forms a Verdier localization

$Z$0

The resulting category is Hom-finite, Krull–Schmidt, triangulated, contains the original discrete cluster category as a full subcategory, and has indecomposables corresponding to arcs in $Z$1, where

$Z$2

Thus one adds genuine limit arcs, with one or both endpoints at accumulation points (Paquette et al., 2020).

This completion is not $Z$3-Calabi–Yau in general, and $Z$4-spaces are not symmetric. Besides the usual crossing criterion, new extension cases appear when arcs share an accumulation point with an oriented rotation condition, and an arc between two accumulation points has a self-extension. Even so, cluster-tilting subcategories can be classified completely: they are precisely the geometric completions of suitable cluster-tilting subcategories in the original discrete category, with a two-sided fountain at each accumulation point and no leapfrogs. The same work constructs a cluster character with infinitely many indeterminates and proves the multiplication formula and an exchange formula for objects involving only ordinary arcs, under local Calabi–Yau conditions (Paquette et al., 2020).

The completed category also supports structures absent from the original $Z$5-CY discrete category. In the Paquette–Yıldırım completion $Z$6, torsion pairs are classified by completed precovering conditions together with a completed Ptolemy condition; aisles of t-structures are classified by half-decorated non-crossing partitions of $Z$7; aisles of co-t-structures are classified by alternating non-crossing partitions; and nontrivial recollements are classified by functorially finite thick subcategories. The paper stresses that the completion is not $Z$8-Calabi–Yau, that nontrivial co-t-structures appear, and that there are no bounded t-structures in $Z$9 for any $A$0 (Franchini, 2024).

A different response to the failure of the $A$1-CY picture in completed infinity-gons is to keep the additive category but replace its triangulated structure by an extriangulated substructure $A$2. In that setting,

$A$3

so every arc becomes rigid again, the category becomes weakly $A$4-Calabi–Yau, weak cluster-tilting subcategories are once again in bijection with triangulations, and cluster-tilting subcategories are exactly the fan triangulations (Çanakçı et al., 2024).

5. Related infinite cluster categories and neighboring frameworks

The discrete type-$A$5 categories sit alongside several closely related, but not identical, frameworks. A representation-theoretic orbit-category model of type $A$6 constructs

$A$7

shows that it is a Hom-finite Krull–Schmidt $A$8-Calabi–Yau triangulated category, and proves that its cluster-tilting subcategories correspond to compact triangulations of the infinite strip. That paper explicitly cites Igusa–Todorov strip geometry as antecedent, but does not prove equivalence with the Igusa–Todorov categories themselves (Liu et al., 2015).

For type $A$9, Igusa–Todorov’s continuous construction produces continuous Frobenius categories whose stable categories are cluster categories of type $Z\subseteq S^1$0, containing the standard finite $Z\subseteq S^1$1 cluster categories. In the continuous punctured-disk model, maximal compatible sets are laminations, and the discrete laminations are precisely the clusters. The singular objects $Z\subseteq S^1$2 and $Z\subseteq S^1$3 provide an algebraic interpretation of tagged arcs (Igusa et al., 2013). A later survey places this beside discrete infinite type-$Z\subseteq S^1$4 constructions from thread quivers, derived categories, and orbit categories, producing families $Z\subseteq S^1$5 and weak cluster categories $Z\subseteq S^1$6 modeled by punctured $Z\subseteq S^1$7-gons and their completed versions (Mohammadi et al., 2022).

Not every paper in the broader Igusa–Todorov orbit-categorical tradition is about discrete cluster categories as standalone triangulated categories. The $Z\subseteq S^1$8-cluster morphism category of a $Z\subseteq S^1$9-tilting finite algebra generalizes the hereditary Igusa–Todorov cluster morphism category, proves cubicality of the classifying space, identifies its fundamental group with the picture group, and shows that for Nakayama algebras the picture space is a locally $L(Z)=\overline Z\setminus Z,\qquad |L(Z)|=n>0,$0 cube complex and hence a $L(Z)=\overline Z\setminus Z,\qquad |L(Z)|=n>0,$1. The same paper explicitly notes that it does not define or analyze discrete cluster categories as a separate class of triangulated categories, and is relevant only indirectly to that subject (Hanson et al., 2018).

6. Generation theory and negative Calabi–Yau analogues

Recent work on the Paquette–Yıldırım completion $L(Z)=\overline Z\setminus Z,\qquad |L(Z)|=n>0,$2 introduces the notion of homologically connected objects and the hc decomposition. Every object in $L(Z)=\overline Z\setminus Z,\qquad |L(Z)|=n>0,$3 has an hc decomposition, and that decomposition determines the thick closure: if

$L(Z)=\overline Z\setminus Z,\qquad |L(Z)|=n>0,$4

is the hc decomposition, then an indecomposable $L(Z)=\overline Z\setminus Z,\qquad |L(Z)|=n>0,$5 lies in $L(Z)=\overline Z\setminus Z,\qquad |L(Z)|=n>0,$6 if and only if its complete orbit is contained in the complete orbit of some $L(Z)=\overline Z\setminus Z,\qquad |L(Z)|=n>0,$7. Classical generators are exactly the homologically connected objects with complete orbit, homological length gives an upper bound for generation time, the Orlov spectrum is bounded, and

$L(Z)=\overline Z\setminus Z,\qquad |L(Z)|=n>0,$8

(Murphy, 2023).

A further development replaces the original $L(Z)=\overline Z\setminus Z,\qquad |L(Z)|=n>0,$9-Calabi–Yau geometry by a negative Calabi–Yau one. The categories

$Z\subseteq S^1$00

are obtained as stable categories of Frobenius categories of infinite discrete symmetric Nakayama representations built using persistence-theoretic techniques. They are presented as negative Calabi–Yau analogues of the Igusa–Todorov discrete cluster categories of type $Z\subseteq S^1$01. Their indecomposables are parametrized by admissible arcs in the same $Z\subseteq S^1$02-gon $Z\subseteq S^1$03, the suspension is

$Z\subseteq S^1$04

the category is $Z\subseteq S^1$05-Calabi–Yau, and the Auslander–Reiten translation is

$Z\subseteq S^1$06

In the one-accumulation-point case,

$Z\subseteq S^1$07

This shows that the Igusa–Todorov program now includes not only the original $Z\subseteq S^1$08-Calabi–Yau discrete categories and their completions, but also explicit negative Calabi–Yau analogues with a comparable arc model (Franchini, 18 Aug 2025).