Density of $g$-vector cones from triangulated surfaces
Toshiya Yurikusa

TL;DR
This paper characterizes the union of $g$-vector cones in cluster algebras from surfaces, revealing its geometric structure and implications for the connectivity of the associated exchange graph.
Contribution
It determines the closure of the union of $g$-vector cones for all clusters in surface-based cluster algebras, linking geometric and combinatorial properties.
Findings
Union of $g$-vector cones covers $ ext{R}^n$ except for a punctured surface case.
Connectedness of the exchange graph depends on the surface's puncture configuration.
Explicit description of the hyperplane for the punctured surface case.
Abstract
We study -vector cones associated with clusters of cluster algebras defined from a marked surface of rank . We determine the closure of the union of -vector cones associated with all clusters. It is equal to except for a closed surface with exactly one puncture, in which case it is equal to the half space of a certain explicit hyperplane in . Our main ingredients are laminations on , their shear coordinates and their asymptotic behavior under Dehn twists. As an application, if is not a closed surface with exactly one puncture, the exchange graph of cluster tilting objects in the corresponding cluster category is connected. If is a closed surface with exactly one puncture, it has precisely two connected components.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Density of -vector cones from triangulated surfaces
toshiya yurikusa
T. Yurikusa: Graduate School of Mathematics, Nagoya University, Chikusa-ku, Nagoya, 464-8602 Japan
Abstract.
We study -vector cones associated with clusters of cluster algebras defined from a marked surface of rank . We determine the closure of the union of -vector cones associated with all clusters. It is equal to except for a closed surface with exactly one puncture, in which case it is equal to the half space of a certain explicit hyperplane in . Our main ingredients are laminations on , their shear coordinates and their asymptotic behavior under Dehn twists. As an application, if is not a closed surface with exactly one puncture, the exchange graph of cluster tilting objects in the corresponding cluster category is connected. If is a closed surface with exactly one puncture, it has precisely two connected components.
Key words and phrases:
cluster algebra, marked surface, lamination, shear coordinate, cluster category, -tilting theory
1. Introduction
Cluster algebras, introduced by Fomin and Zelevinsky in [FZ02], are commutative algebras with generators called cluster variables. The certain tuples of cluster variables are called clusters. Their original motivation was to study total positivity of semisimple Lie groups and canonical bases of quantum groups. In recent years, it has interacted with various subjects in mathematics, for example, representation theory of quivers, Poisson geometry, integrable systems, and so on.
Let be a quiver without loops and -cycles, and let be the associated cluster algebra with principal coefficients (see Subsection 3.1). We denote by the set of clusters in . Each cluster variable in has a numerical invariant , called the -vector of [FZ07]. For each , one can define a cone
[TABLE]
in , called the -vector cone of . Note that these cones and their faces form a fan [Re14a, Theorem 8.7]. We say that is finite type if . The following result is well-known.
Theorem 1.1**.**
[Re14a, Theorem 10.6]** If a quiver is finite type, then we have
[TABLE]
In this paper, we study an analogue of Theorem 1.1 for cluster algebras defined from marked surfaces that were developed in [FG06, FG09, FST, FoT, GSV].
Let be a marked surface and a tagged triangulation of (see Subsection 2.1). We denote by the number of tagged arcs of . Fomin, Shapiro and Thurston [FST] constructed a quiver associated with . In , cluster variables correspond to tagged arcs, and clusters correspond to tagged triangulations (Theorem 3.3). Our first aim is to give the following analogue of Theorem 1.1.
Theorem 1.2**.**
If is not a closed surface with exactly one puncture, then we have
[TABLE]
where is the closure with respect to the natural topology on . If is a closed surface with exactly one puncture, then we have
[TABLE]
The second aim of this paper is to apply Theorem 1.2 to representation theory. We consider a non-degenerate potential of such that the associated Jacobian algebra is finite dimensional [DWZ]. Such a potential exists (Proposition 4.5). The potential and its Jacobian algebra have been studied by a number of researchers (see e.g. [ABCP, CL, GLS, V]). We focus on the associated cluster category in this paper.
Using the Ginzburg differential graded algebra associated with [G], Amiot [A] constructed a generalized cluster category with cluster tilting object . The -vector of each rigid object (resp., -rigid pair) in (resp., ) is a certain element in the Grothendieck group (resp., ). The -vectors of indecomposable direct summands of a cluster tilting object (resp., a -tilting pair ) form a cone in (resp., in ), called the -vector cone of (resp., ). Note that these -vector cones and their faces form a fan [DIJ]. Such a fan plays an important role in the study of scattering diagrams and their wall-chamber structures (see e.g. [B, BST, GHKK, GS, KS, Y18a]).
We denote by (resp., ) the set of isomorphism classes of basic cluster tilting objects in (resp., -tilting pairs in ). We also denote by (resp., , , ) the subset of (resp., , , ) consisting of mutation equivalence classes containing (resp., , , ). We set
[TABLE]
The following analogues of Theorem 1.2 hold.
Theorem 1.3**.**
Let be a tagged triangulation of a marked surface . For a non-degenerate potential of such that is finite dimensional, let and . Then we have the equalities
[TABLE]
This theorem means that -vector cones are dense in the scattering diagram of . It gives the following application.
Corollary 1.4**.**
Any basic cluster tilting object in (resp., -tilting pair in ) is contained in (resp., ). In particular, if is not a closed surface with exactly one puncture, the exchange graph of (resp., ) is connected, thus (resp., ). Otherwise, it has precisely two connected components and (resp., and ).
Notice that it was known by Plamondon [Pl13] and Ladkani [Lad13] that if is a closed surface with exactly one puncture, then the exchange graph of is not connected. Also, it was known by Qiu and Zhou [QZ] that if has non-empty boundary, then the exchange graph of is connected. Our proof is entirely different from theirs.
To prove Theorem 1.2, our main ingredient is shear coordinates on . To study coefficients in cluster algebras defined from , Fomin and Thurston [FoT] used a certain class of curves in , called laminates, and finite multi-sets of pairwise non-intersecting laminates, called laminations (see also [FG07, T]). To a laminate of , they associated an integer vector whose entries are shear coordinates of and defined for a lamination on . They showed that the map induces a bijection between the set of laminations on and . Using this bijection, some properties of cluster algebras were given (see e.g. [MSW13, Re14b]). In this paper, we want to consider not only integer vectors but real vectors.
For a multi-set of laminates of , in the same way as -vector cones, we can define a cone in , called the shear coordinate cone of with respect to . Recall that there is a natural injective map from the set of tagged arcs of to the set of laminates of (see Subsection 2.2). We denote by the set of tagged triangulations of . The following result plays an important role to prove Theorem 1.2.
Theorem 1.5**.**
Let be a tagged triangulation of a marked surface . Then we have
[TABLE]
If is a closed surface with exactly one puncture , then we have
[TABLE]
where (resp., ) is the set of tagged triangulations of tagged at in the same (resp., different) way as .
It will be interesting to understand connections between our results and known results on Teichmüller spaces such as [FG11, Ro12, Ro13].
This paper is organized as follows. In Section 2, we recall the notions of marked surfaces, laminations, and their shear coordinates. We study shear coordinates of laminates and their asymptotic behavior under Dehn twists, and prove Theorem 1.5. In Section 3, we recall cluster algebras defined from triangulated surfaces. We show that the shear coordinate of a laminate with respect to correspond with the -vector of a cluster variable in or . Consequently, Theorem 1.2 follows from Theorem 1.5. In Section 4, we recall -tilting theory, cluster tilting theory, and the relationships between them and cluster algebras. Finally, we prove Theorem 1.3 and Corollary 1.4.
Acknowledgements. The author would like to thank his supervisor Osamu Iyama for his guidance and helpful comments. He also thanks Daniel Labardini-Fragoso for valuable comments and the referees for fruitful suggestions. He is a Research Fellow of Society for the Promotion of Science (JSPS). This work was supported by JSPS KAKENHI Grant Number JP17J04270.
2. Density of shear coordinate cones from triangulated surfaces
2.1. Marked surfaces and tagged triangulations
We start with recalling the notions of [FST]. Let be a connected compact oriented Riemann surface with (possibly empty) boundary and a non-empty finite set of marked points on with at least one marked point on each boundary component. We call the pair a marked surface. Any marked point in the interior of is called a puncture. For technical reasons, we assume that is neither a monogon with at most one puncture, a digon without punctures, a triangle without punctures, nor a sphere with at most three punctures.
An arc of is a curve in with endpoints in , considered up to isotopy, such that the following conditions are satisfied:
- •
does not intersect itself except at its endpoints;
- •
is disjoint from and except at its endpoints;
- •
does not cut out an unpunctured monogon or an unpunctured digon.
An arc with two identical endpoints is called a loop. Two arcs are called compatible if they don’t intersect in the interior of . When we consider intersections of curves and , we assume that and intersect transversally in a minimum number of points. We denote by the set of their intersection points. An ideal triangulation is a maximal collection of distinct pairwise compatible arcs. A triangle with only two distinct sides is called self-folded (see Figure 2).
For an ideal triangulation , a flip at an arc replaces with another arc such that is an ideal triangulation. Notice that an arc inside a self-folded triangle can not be flipped. To make flip always possible, the notion of tagged arcs was introduced in [FST].
A tagged arc of is an arc whose each end is tagged in one of two ways, plain or notched, such that the following conditions are satisfied:
- •
does not cut out a monogon with exactly one puncture;
- •
If an endpoint of lie on , then it is tagged plain;
- •
If is a loop, then the both ends are tagged in the same way.
In the figures, we represent tagged arcs as follows:
[TABLE]
For an arc of , we define a tagged arc as follows:
- •
If does not cut out a monogon with exactly one puncture, then is the tagged arc obtained from by tagging both ends plain;
- •
If is a loop at cutting out a monogon with exactly one puncture , then there is a unique arc that connects and and does not intersect . And then is the tagged arc obtained by tagging plain at and notched at (see Figure 2).
A pair of conjugate arcs is, for a self-folded triangle , or a pair obtained from by simultaneous changing tags at each endpoint (see Figure 2).
For a tagged arc , we denote by the arc obtained from by forgetting its tags. Two tagged arcs and are called compatible if the following conditions are satisfied:
- •
The arcs and are compatible;
- •
If , then at least one end of is tagged in the same way as the corresponding end of ;
- •
If and they have a common endpoint , then the ends of and at are tagged in the same way.
A partial tagged triangulation is a collection of distinct pairwise compatible tagged arcs. If a partial tagged triangulation is maximal, then it is called a tagged triangulation. Recall that we denote by the set of tagged triangulations of . We can define flips of tagged triangulations in the same way as ones of ideal triangulations. In particular, any tagged arc can be flipped.
Theorem 2.1**.**
[FST, Theorem 7.9, Proposition 7.10]** If is not a closed surface with exactly one puncture, the exchange graph of is connected, that is, any two tagged triangulations of are connected by a sequence of flips. Otherwise, it has exactly two isomorphic components: one in which all ends of tagged arcs are plain and one in which they are notched.
2.2. Laminations on marked surfaces
We recall the notions of [FoT]. A laminate of is a non-self-intersecting curve in , considered up to isotopy relative to , which is either
- •
a closed curve, or
- •
a curve whose ends are unmarked points on or spirals around punctures (either clockwise or counterclockwise),
and the following curves are not allowed (see Figure 4):
- •
a curve cutting out a disk with at most one puncture;
- •
a curve with two endpoints on such that it is isotopic to a piece of containing at most one marked point;
- •
a curve whose both ends are spirals around a common puncture in the same direction such that it does not enclose anything else.
Definition 2.2**.**
We say that two laminates of are compatible if they don’t intersect. A finite multi-set of pairwise compatible laminates of is called a lamination on (see Figure 4).
Let be a laminate of . For an ideal/tagged triangulation of , we define the shear coordinate of with respect to (see [FoT, Definition 12.2, 13.1]).
First, we assume that is an ideal triangulation. If is not inside a self-folded triangle of , then is defined by a sum of contributions from all intersections of and as follows: Such an intersection contributes (resp., ) to if a segment of cuts through the quadrilateral surrounding as in the left (resp., right) diagram of Figure 5.
Suppose that is inside a self-folded triangle of , where is a loop enclosing exactly one puncture . Then we define , where is a laminate obtained from by changing the directions of its spirals at if they exist.
Next, we assume that is a tagged triangulation. If there is an ideal triangulation satisfying , then we define , where . For an arbitrary , we can obtain a tagged triangulation from by simultaneous changing all tags at punctures (possibly ), in such a way that there is a unique ideal triangulation satisfying (see [MSW11, Remark 3.11]). Then we define b_{\operatorname{\gamma},T}(\ell)=b_{\operatorname{\gamma}^{(p_{1}\cdots p_{m})},T^{(p_{1}\cdots p_{m})}}\bigl{(}(\cdots((\ell^{(p_{1})})^{(p_{2})})\cdots)^{(p_{m})}\bigr{)}, where corresponds to .
For a multi-set of laminates of , the shear coordinate of with respect to is inductively defined by
[TABLE]
We denote by a vector . Note that the shear coordinate cone is a cone spanned by for . These vectors have the following property.
Theorem 2.3**.**
[FoT, Theorems 12.3, 13.6]** Let be a tagged triangulation of . The map sending laminations to induces a bijection between the set of laminations on and .
Example 2.4**.**
For a digon with exactly one puncture, all laminates are given as follows:
[TABLE]
We consider the following tagged triangulation :
[TABLE]
The shear coordinate is given by . Since , we have the equalities
[TABLE]
Similarly, for and , the shear coordinates and are given as follows:
[TABLE]
In particular, we have
[TABLE]
where . On the other hand, all laminations on are given by for and . Since and induces a bijection
[TABLE]
there is a bijection between the set of laminations on and .
2.3. Elementary and exceptional laminates
Non-closed laminates of are divided into two types, elementary and exceptional. For a tagged arc of , we define an elementary laminate as follows:
- •
is a laminate running along in a small neighborhood of it;
- •
If has an endpoint on a component of , then the corresponding endpoint of is located near on in the clockwise direction as in the left diagram of Figure 6;
- •
If has an endpoint at a puncture , then the corresponding end of is a spiral around clockwise (resp., counterclockwise) if is tagged plain (resp., notched) at as in the right diagram of Figure 6.
It follows from the construction that the map from the set of tagged arcs of to the set of laminates is injective. For an elementary laminate , we denote by a unique tagged arc such that . Note that, for a tagged arc , a lamination is a reflection of the elementary lamination of defined in [FoT, Definition 17.2]. Our convention is more convenient for our aim.
Elementary laminates have the following properties.
Proposition 2.5**.**
(1) Let and be tagged arcs such that . Then and are compatible if and only if and are compatible.
(2) The map induces a bijection between the set of partial tagged triangulations of without pairs of conjugate arcs and the set of laminations of consisting only of distinct elementary laminates.
Proof.
(1) Since transforms and just around the marked points of , it is enough to consider neighborhoods of their endpoints. In particular, if and have no common endpoints, the assertion holds. Suppose that and have at least one common endpoint. Since , is not a pair of conjugate arcs. Thus and are compatible if and only if the ends of and at each common endpoint are tagged in the same way. By the definition of , it is equivalent that and are compatible.
(2) If two distinct tagged arcs and satisfying are compatible, then is a pair of conjugate arcs, in which case and are not compatible. Therefore, the assertion follows from (1). ∎
Laminates which are neither closed nor elementary are called exceptional. They are characterized as follows.
Proposition 2.6**.**
A laminate is exceptional if and only if it is one of the following curves (Figure 7):
- •
a curve enclosing exactly one puncture whose both endpoints lie on a common boundary segment;
- •
a curve enclosing exactly one puncture whose both ends are spirals around a common puncture in the same direction.
Proof.
Applying the same transformation as to a non-closed laminate and forgetting its tags, we obtain a unique ideal arc. In general, an ideal arc is not obtained from a tagged arc by forgetting its tags if and only if is a loop cutting out a monogon with exactly one puncture. Therefore, is not elementary if and only if it is one of the desired cases. ∎
Note that exceptional laminates coincide with excluded curves for quasi-laminations in [Re14b]. To interpret shear coordinates of exceptional laminates as ones of elementary laminates, we introduce the following notations. For an exceptional laminate of , elementary laminates and are given by
[TABLE]
In particular, () is a pair of conjugate arcs. For a lamination on , we denote by the multi-set of elementary laminates obtained from by replacing exceptional laminates with and .
Example 2.7**.**
In Example 2.4, and are exceptional, and , , and . Thus we have the equalities
[TABLE]
In general, the same property as Example 2.7 holds for arbitrary exceptional laminates.
Lemma 2.8**.**
Let be a tagged triangulation of . For an exceptional laminate of , we have
[TABLE]
Proof.
By Proposition 2.6, there is a unique puncture enclosed by . We only need to prove
[TABLE]
for any . If is not incident to , then (2.1) is clear. We assume that is incident to . Let be tagged arcs of incident to winding clockwisely around such that the following conditions are satisfied (see Figure 8):
- •
crosses them at points in this order;
- •
The segment of from to , that of from to , and that of from to form a contractible triangle.
Note that if these arcs contains a pair of conjugate arcs, then we can choice the order of and .
Moreover, is different from . Indeed, if , considering triangles with a side , there is a tagged arc of incident to such that crosses it before or after , a contradiction.
The contributions to at except at coincide with the contributions to and at them. We denote by (resp, , ) the contribution to (resp., , ) at for . To prove (2.1), we only need to show
[TABLE]
First, we assume that neither nor form a pair of conjugate arcs. Then it is easy to give the following values:
[TABLE]
Therefore, (2.2) holds.
Second, we assume that is a pair of conjugate arcs tagged in the different ways at , in which case . Then by exchanging and if necessary (see Figure 9), we have
[TABLE]
Therefore, (2.2) holds.
Finally, we assume that is a pair of conjugate arcs tagged in the same way at . We define a set as follows: If is a pair of conjugate arcs, then ; Otherwise, . Then we have for and , and
[TABLE]
Therefore, (2.2) holds. Moreover, it follows from the symmetry for the case that is a pair of conjugate arcs tagged in the same way at . Consequently, (2.1) holds for any . ∎
For a lamination on , we have decompositions
[TABLE]
where (resp., , ) consists of all elementary (resp., exceptional, closed) laminates in . For multi-sets and of laminates of , we define non-multi-sets
[TABLE]
The following properties are used to prove Theorem 1.5 in Subsection 2.5.
Proposition 2.9**.**
Let be a lamination on with . Then the following properties hold:
- (1)
.
- (2)
* is a partial tagged triangulation of .*
Moreover, we take a set of tagged arcs of such that is a tagged triangulation. Then we have the equality
- (3)
.
Proof.
(1) The assertion immediately follows from Lemma 2.8.
(2) By Proposition 2.5(2), is a partial tagged triangulation of . Since is a lamination, any laminate in is compatible with all laminates in . Then, by Proposition 2.5(1), any tagged arc of is compatible with all tagged arcs of . Moreover, is a partial tagged triangulation since is a pair of conjugate arcs for . Therefore,
[TABLE]
is a partial tagged triangulation of .
(3) Since coincides with the multiplicity of , we have the equalities
[TABLE]
.
2.4. Shear coordinates and Dehn twists
We consider the Dehn twist along a closed laminate and its effect on shear coordinates. In this subsection, we fix an ideal or tagged triangulation , a closed laminate of and its direction. We denote by the Dehn twist of along defined from the direction of as follows:
[TABLE]
The aim of this subsection is to prove the following.
Theorem 2.10**.**
Let be a closed laminate and a laminate of intersecting with , and let . Then there is such that for any , we have
[TABLE]
First, we assume that is an annulus without punctures and is its ideal triangulation consisting of arcs crossing in order of occurrence along (we can have even if ), that is,
[TABLE]
where two vertical lines are identified. Any elementary laminate of intersects with at most once since they intersect in a minimal number of points. We assume that intersects with . We define the direction of as crossing from left to right:
[TABLE]
Let such that the starting point of is on the triangle of with sides and , where . In particular, intersects at least one of and . Thus intersects with the () diagonals either or of in order. In the former (resp., latter) case, we say that intersects with in ascending (resp., descending) order:
[TABLE]
Proposition 2.11**.**
Let be an annulus without punctures and an elementary laminate of intersecting with .
- (1)
If intersects with in ascending order, then so is and for , we have
[TABLE]
If and intersect with in descending order, then for , we have
[TABLE]
- (2)
There is such that intersects with in ascending order.
Proof.
(1) We only prove the first assertion since the proof of the second assertion is similar. Suppose that intersects with in ascending order. If , then the assertion holds since only transforms around an intersection point of and as follows:
[TABLE]
If , then and are given as follows:
[TABLE]
Then the assertion is directly given by enumerating their shear coordinates.
(2) Suppose that intersects with in descending order. If , then intersects with in ascending order. If , then intersects with in descending order and . By the induction, the assertion holds. ∎
Next, we consider an arbitrary marked surface and its ideal triangulation . For in Theorem 2.10, we construct an annulus and its triangulation as follows: Let be the arcs of crossing in order of occurrence along ( and can be the same even if ). Hence crosses triangles in this order. For , let be a copy of the triangle , hence has the sides and . Then an annulus and its triangulation are obtained by gluing along the edges , that is,
[TABLE]
in , where two vertical lines are identified. In particular, if is inside a self-folded triangle of , then the corresponding triangles are given by
[TABLE]
For a laminate of intersecting with , let and a laminate of corresponding to the connected segment of in containing as follows:
[TABLE]
Proposition 2.12**.**
Let be an ideal triangulation of , , and a laminate of intersecting with . We assume that for all , intersects with in ascending order. Then we have
[TABLE]
Proof.
The proof is divided into the following three cases (1)–(3).
(1) If is not a side of triangles of , then the assertion is clear.
(2) We assume that is a side of some triangle of and . If is not inside a self-folded triangle of , then the construction of preserves the quadrilateral surrounding . Therefore, we have
[TABLE]
Suppose that is inside a self-folded triangle of enclosing a puncture . Recall that , where is a laminate obtained from by changing the directions of its spirals at if they exist (see Subsection 2.2). Thus the assertion follows from the previous case if for all , intersects with in ascending order. This is checked as follows: If no ends of are spirals at , then , hence it is clear. Otherwise, , and and are identified in the natural way. Then and are only different in that their ends around are given by
[TABLE]
Therefore, also intersects with in ascending order.
(3) We assume that is a side of some triangle of and . Then we prove . In this case, is not inside a self-folded triangle of . Indeed, if is inside a self-folded triangle of , then is either or , hence it is a contradiction. Therefore, there is the quadrilateral surrounding of . Since for all , intersects with in ascending order, the Dehn twist affects as follows:
[TABLE]
Therefore, it gives . ∎
We are ready to prove Theorem 2.10.
Proof of Theorem 2.10.
First of all, we prove Theorem 2.10 for an ideal triangulation . For , by Proposition 2.11(2) there exists such that intersects with in ascending order. Thus for
[TABLE]
intersects with in ascending order for each . Therefore, by Theorem 2.12, we have
[TABLE]
For an arbitrary tagged triangulation , we recall that there is a unique ideal triangulation satisfying , where is obtained from by simultaneous changing all tags at some punctures if necessary (see Subsection 2.2 for details). Then the shear coordinate of a laminate is equal to , where is a laminate obtained from by changing the directions of its spirals at these punctures if they exist. Since the change of directions of its spirals and the Dehn twist are compatible, the proof of Theorem 2.10 comes down to the case of ideal triangulations. ∎
2.5. Proof of Theorem 1.5
In this subsection, we fix a tagged triangulation of . By Theorem 2.3, to prove the first assertion of Theorem 1.5, we only need to show that for each lamination on ,
[TABLE]
To prove (2.6), we need some preparation. We have decompositions (2.4) of . By Proposition 2.9(2), is a partial tagged triangulation of . Then we take a set of tagged arcs of such that is a tagged triangulation.
Lemma 2.13**.**
Let be a lamination on .
- (1)
Any closed laminate in does not intersect with tagged arcs of , but it intersects with at least one tagged arc of .
- (2)
.
Proof.
Since is a lamination, any closed laminate in does not intersect with all laminates in . Thus does not intersect with all tagged arcs of since and transform laminates just around the marked points of . Moreover, since is a tagged triangulation and is not contractible, intersects with at least one tagged arc of , hence it is of .
By , we have and it is a tagged triangulation. The desired equality is given by Proposition 2.9(3). ∎
Let be all distinct closed laminates in and the multiplicity of in for . By Lemma 2.13(1), is not zero. In particular, is equal to . We fix the direction of for and consider the Dehn twists . Since are not intersect, are commutative. We set
[TABLE]
Proposition 2.14**.**
Let be a lamination on . Then we have
[TABLE]
Proof.
By Theorem 2.10, for and , we have
[TABLE]
This equality gives
[TABLE]
thus
[TABLE]
Since by Proposition 2.9(1), we have
[TABLE]
where the last equality is given by Lemma 2.13(2). ∎
Proof of the first assertion of Theorem 1.5.
Since is a tagged triangulation of for any , Proposition 2.14 finishes the proof of (2.6). Hence the assertion holds. ∎
To prove the second assertion of Theorem 1.5, we give the following results in a more general setting.
Proposition 2.15**.**
Let be an ideal triangulation of without self-folded triangles.
- (1)
For a laminate , we have
[TABLE]
- (2)
For a tagged arc of whose both endpoints are punctures, we have
[TABLE]
Proof.
(1) Fix a direction of . For a closed laminate , let be in (2.5). For a non-closed laminate , we define a polygon and its ideal triangulation consisting of triangles and in the same way, where we need a slight modification at the spirals. If the starting (resp., ending) end of is a spiral around a puncture , then the triangles of incident to are for (resp., for ) as follows:
[TABLE]
Let be an arbitrary laminate of and we consider the ideal triangulation consisting of triangles ( if is closed). We call a left (resp., right) triangle if a side of is a boundary segment of on the left (resp., right) side of . Then we have
[TABLE]
Therefore, we have
[TABLE]
On the other hand, since has no self-folded triangles, we have
[TABLE]
Thus the assertion follows from (2.7).
(2) We consider as above and define the direction of from by the obvious way. If the starting point of is tagged plain (resp., notched), then is a right (resp., left) triangle. If the ending point of is tagged plain (resp., notched), then is a left (resp., right) triangle. Thus the assertion also follows from (2.7). ∎
Proof of the second assertion of Theorem 1.5.
Let be a closed surface with exactly one puncture and its tagged triangulation. In this case, all ends of tagged arcs of are tagged plain or they are tagged notched. Thus we can assume that is an ideal triangulation without self-folded triangles. Let and be tagged arcs of tagged plain and notched, respectively. We only need to show that
[TABLE]
that is,
[TABLE]
It immediately follows from Proposition 2.15(2). ∎
2.6. Example of Proposition 2.14
For an annulus with exactly two marked points, all laminates are given as follows:
[TABLE]
where is closed and is elementary for . Their shear coordinates with respect to
[TABLE]
are given by
[TABLE]
For a tagged triangulation of , the shear coordinate cone is given by for some . Then the set of integer vectors which are not contained in these shear coordinate cones is . Taking and , Proposition 2.14 means that
[TABLE]
It is described in the above picture.
3. Cluster algebras
3.1. Cluster algebras and triangulated surfaces
We briefly recall cluster algebras with principal coefficients [FZ07]. For that, we need to prepare some notations. Let and be the field of rational functions in variables over .
Definition 3.1**.**
A seed with coefficients is a pair consisting of the following data:
- (a)
is a free generating set of over .
- (b)
is a quiver without loops and -cycles whose vertices are .
Then we refer to as the cluster, to each as a cluster variable and as a coefficient.
For a seed with coefficients, the mutation in direction is defined as follows:
- (a)
is defined by
[TABLE]
where .
- (b)
is the quiver obtained from by the following steps:
- (i)
For any path , add an arrow .
- (ii)
Reverse all arrows incident to .
- (iii)
Remove a maximal set of disjoint -cycles.
We remark that is an involution, that is, we have . Moreover, it is elementary that is also a seed with coefficients.
For a quiver without loops and -cycles whose vertices are . The framed quiver associated with is the quiver obtained from by adding vertices and arrows . We fix a seed with coefficients, called the initial seed. We also call each the initial cluster variable.
Definition 3.2**.**
The cluster algebra with principal coefficients for the initial seed is a -subalgebra of generated by the cluster variables and the coefficients obtained by all sequences of mutations from .
One of the remarkable properties of cluster algebras with principal coefficients is the strongly Laurent phenomenon [FZ07, Proposition 3.6], that is, . We consider the -grading in given by
[TABLE]
where are the standard basis vectors in . Every cluster variable of is homogeneous with respect to the -grading, and its degree is called -vector of [FZ07, Proposition 6.1]. We denote by the set of clusters in and by the set of cluster variables in .
Let be a tagged triangulation of . Fomin, Shapiro and Thurston [FST] constructed a quiver without loops and -cycles as follows: Any tagged triangulation is obtained by gluing together a number of puzzle pieces in Table 1 and by simultaneous changing all tags at some punctures (see [FST, Remark 4.2] for details). The vertices of are arcs of and its arrows are obtained as in Table 1 for puzzle pieces of , where we remove arrows incident to .
Thus we have the cluster algebra associated with .
We denote by the set of tagged triangulations of obtained from by sequences of flips, and by the set of tagged arcs of each tagged triangulation contained in . Cluster algebras defined from triangulated surfaces have the following properties.
Theorem 3.3**.**
Let be a tagged triangulation of .
- (1)
[FST, Theorem 7.11]**[FoT, Theorem 6.1]** There is a bijection
[TABLE]
Moreover, it induces a bijection
[TABLE]
which sends to the initial cluster in and commutes with flips and mutations.
- (2)
[Lab10, Theorem 10.0.5]**[Lab09b, Theorem 7.1]**[Re14b, Proposition 5.2]** For each , we have
[TABLE]
Note that, in another way, Theorem 3.3(2) can be directly given by the cluster expansion formula in [Y18b]. Moreover, it was proved in [FeT, Theorem 8.6] for orbifolds in the same way as [Re14b, Proposition 5.2].
3.2. Proof of Theorem 1.2
We recall the following notion to prove Theorem 1.2.
Definition 3.4**.**
[BQ] Let be an arbitrary marked surface. The tagged rotation of a tagged arc of is the tagged arc defined as follows:
- •
If has an endpoint on a component of , then is obtained from by moving to the next marked point on in the counterclockwise direction;
- •
If has an endpoint at a puncture , then is obtained from by changing its tags at .
By Theorem 2.1, we have
[TABLE]
Let be the same surface as oriented in the opposite direction and . For a tagged arc or laminate of , we denote by the corresponding one of . In particular, the tagged triangulation of is naturally induced by and we have . By Theorem 3.3(1), the composition of maps , and gives a bijection
[TABLE]
Moreover, it induces a bijection
[TABLE]
which sends to the initial cluster in and commutes with flips and mutations.
Theorem 3.5**.**
Let be a tagged triangulation of . Then for each , we have
[TABLE]
Proof.
For a tagged arc of , the equalities
[TABLE]
hold. Since , Theorem 3.3(2) gives
[TABLE]
for , hence . ∎
Proof of Theorem 1.2.
By Theorems 3.3 and 3.5, we have
[TABLE]
If is a closed surface with exactly one puncture, then and coincide with and in Theorem 1.5, respectively. Therefore, the assertion follows from Theorem 1.5 and (3.1). ∎
3.3. Example for a cluster algebra
For the tagged triangulation in Subsection 2.6, the quiver is the Kronecker quiver . The set is described by
[TABLE]
where
[TABLE]
The corresponding -vectors are as follows:
[TABLE]
The -vector cones of clusters are reflections of the corresponding shear coordinate cones in Subsection 2.6 as follows:
[TABLE]
4. Representation theory
4.1. -tilting theory and cluster tilting theory
In this subsection, we recall -tilting and cluster tilting theory to prepare for the proofs of Theorem 1.3 and Corollary 1.4.
First, we recall -tilting theory [AIR]. Let be a finite dimensional algebra over a field. We denote by (resp., ) the category of finitely generated (resp., finitely generated projective) left -modules. We denote by the Auslander-Reiten translation of and by is the number of non-isomorphic indecomposable direct summands of . Let and . We say that a pair is
- •
-rigid if ;
- •
-tilting if is -rigid and
- •
basic if and are basic;
- •
a direct summand of if is a direct summand of and is a direct summand of ;
- •
indecomposable if is basic and .
Recall that we denote by the set of isomorphism classes of basic -tilting pairs in . For and an indecomposable direct summand of , there is a unique indecomposable -rigid pair such that [AIR, Theorem 0.4]. Therefore, one can define mutations in .
Let be a decomposition of , where is an indecomposable projective -module. Then form a basis for , thus there is a natural bijection between and . Let . There is a minimal projective presentation of
[TABLE]
We set
[TABLE]
called the -vector of . We denote by the set of isomorphism classes of indecomposable -rigid pairs in . The -vector of is .
For our aim, we also need to consider the opposite algebra of . For , the notation denotes the transpose of . We define . Then gives .
Theorem 4.1**.**
[AIR, Theorem 2.14]**[F, Subsection 3.4]** There is a bijection
[TABLE]
given by such that
[TABLE]
where is a maximal projective direct summand of . The map induces a bijection
[TABLE]
which sends to and commutes with mutations.
Next, we recall cluster tilting theory in -Calabi-Yau triangulated categories. Let be a Hom-finite Krull-Schmidt -Calabi-Yau triangulated category. We call rigid if . We denote by the category of all direct summands of finite direct sums of copies of . We call cluster tilting if . We denote by the set of isomorphism classes of indecomposable rigid objects in . Recall that we denote by the set of isomorphism classes of basic cluster tilting objects in . We assume that has cluster tilting objects, that is, . In this case, any maximal rigid object in is cluster tilting [ZZ, Theorem 2.6]. Iyama and Yoshino [IY] gave mutations in (see also [BMRRT]).
Let be a decomposition of , where is indecomposable. Then form a basis for , thus there is a natural bijection between and . For and , there is a triangle
[TABLE]
where . We define
[TABLE]
called the -vector of with respect to .
There is a close relationship between cluster tilting theory and -tilting theory as follows.
Theorem 4.2**.**
[AIR, Theorem 4.1]** Let and . Then there is a bijection
[TABLE]
such that
[TABLE]
for . Moreover, it induces a bijection
[TABLE]
which sends to and commutes with mutations.
For , we denote by (resp., ) the set of indecomposable direct summands of an object in (resp., ). Clearly, the map in Theorem 4.2 gives bijections
[TABLE]
4.2. Representation theory and cluster algebras
We consider a relationship between representation theory and cluster algebras to prove Theorem 1.3 and Corollary 1.4. For a quiver with potential , we have the associated Jacobian algebra , Ginzburg differential graded algebra , and generalized cluster category (see e.g. [A, DWZ, G, K08, K11] for details). The following is the main result in the additive categorification of cluster algebras.
Theorem 4.3**.**
Let be a quiver without loops and -cycles and a non-degenerate potential of such that is finite dimensional.
- (1)
[A, Theorem 2.1]** The category is a Hom-finite Krull-Schmidt -Calabi-Yau triangulated category with a cluster tilting object .
- (2)
[FK, Theorem 6.3]**[CKLP, Corollary 3.5]** There is a bijection
[TABLE]
such that
[TABLE]
for . Moreover, it induces a bijection
[TABLE]
which sends to the initial cluster in and commutes with mutations.
Note that the map in Theorem 4.3(2) is called the cluster character associated with (see e.g. [BY, CC, Pa, Pl11a, Pl11b]).
We also study and . We have
[TABLE]
where is the opposite quiver of and is a non-degenerate potential of corresponding to .
Corollary 4.4**.**
Let be a quiver without loops and -cycles and a non-degenerate potential of such that is finite dimensional. Then there is a bijection
[TABLE]
such that
[TABLE]
for . Moreover, it induces a bijection
[TABLE]
which sends to the initial cluster in and commutes with mutations.
Proof.
Let be the following composition:
[TABLE]
By Theorems 4.1, 4.2 and 4.3, it induces a bijection between and which sends to the initial cluster in and commutes with mutations. Moreover, we have the equalities
[TABLE]
for . ∎
For a tagged triangulation of , we consider a non-degenerate potential of such that is finite dimensional. It is known that such a potential exists.
Proposition 4.5**.**
Let be a tagged triangulation of . Then there is a non-degenerate potential of such that is finite dimensional.
Proof.
For a sphere with exactly four punctures, such a potential was given in [GG] (see also [GLS]). Suppose that is not a sphere with exactly four punctures. Labardini-Fragoso [Lab09a, Lab16] defined a potential of for any tagged triangulation of , and showed that it is non-degenerate except for a sphere with exactly five punctures, which in this case was proved in [GLS]. Finite dimensionally of was proved in [Lab09a] for with non-empty boundary and in [Lad12] for with empty boundary, where it was proved independently in [TV] for spheres. ∎
4.3. Proofs of Theorem 1.3 and Corollary 1.4
We keep the notations in the previous subsection. Let and , where and are indecomposable. We define -vector cones
[TABLE]
Proof of Theorem 1.3.
Let be a tagged triangulation of and a non-degenerate potential of such that is finite dimensional. By Theorem 4.2, we have
[TABLE]
for . Therefore, we only need to prove the assertion for . By Theorem 4.3(2) and Corollary 4.4, the equalities
[TABLE]
hold. Thus the assertion follows from Theorem 1.2. ∎
Proof of Corollary 1.4.
A -vector cone has dimension for any [DK, Theorem 2.4]. For , and have no intersections except for their boundaries [DIJ, Corollary 6.7]. Thus there are no cluster tilting objects in by Theorem 1.3. The assertion follows from Theorem 2.1. ∎
4.4. Example for representation theory
For the tagged triangulation in Subsection 2.6, the quiver is the Kronecker quiver . The set is as follows:
[TABLE]
The corresponding -vectors
[TABLE]
coincide with the -vectors of the corresponding cluster variables as in Subsection 3.3.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[AIR] T. Adachi, O. Iyama and I. Reiten, τ 𝜏 \tau - tilting theory , Compos. Math. Vol. 150, 3 (2014) 415–452.
- 2[A] C. Amiot, Cluster categories for algebras of global dimension 2 2 2 and quivers with potential , Ann. Inst. Fourier (Grenoble) 59, no. 6 (2009) 2525–2590.
- 3[ABCP] I. Assem, T. Brüstle, G. Charbonneau-Jodoinand and P. Plamondon, Gentle algebras arising from surface triangulations , Algebra Number Theory 4, no. 2 (2010) 201–229.
- 4[B] T. Bridgeland, Scattering diagrams, Hall algebras and stability conditions , Algebr. Geom. 4 (5) (2017) 523–561.
- 5[BQ] T. Brüstle and Y. Qiu, Tagged mapping class groups: Auslander-Reiten translation , Math. Z. 279 (2015) 1103–1120.
- 6[BST] T. Brüstle, D. Smith and H. Treffinger, Wall and chamber structure for finite-dimensional algebras , Adv. Math. Vol. 354 (2019) 106746.
- 7[BY] T. Brüstle and D. Yang, Ordered exchange graphs , Advances in Representations of Algebras (ICRA Bielefeld 2012) 135–193.
- 8[BMRRT] A. B. Buan, R. Marsh, M. Reineke, I. Reiten and G. Todorov, Tilting theory and cluster combinatorics , Adv. Math. Vol. 204 (2006) 572–618.
