Nearly Frobenius structures in some families of algebras
Dalia Artenstein, Ana Gonz\'alez, Gustavo Mata

TL;DR
This paper investigates nearly Frobenius structures across various finite-dimensional algebras, establishing conditions for their existence in radical square zero, string, and toupie algebras, and extending results to quotients of path algebras.
Contribution
It provides new criteria for the presence of nearly Frobenius structures in specific algebra families and generalizes these conditions to quotients of path algebras.
Findings
Radical square zero algebras with paths of length two are nearly Frobenius.
Non-gentle string algebras have at least one non-trivial nearly Frobenius structure.
Monomial relations in toupie algebras imply non-trivial nearly Frobenius structures.
Abstract
In this article we continue with the study started in [1] of nearly Frobenius structures in some representative families of finite dimensional algebras, as the radical square zero algebras, string algebras and the toupie algebras. We prove that the radical square zero algebras with at least one path of length two are nearly Frobenius. As for the string algebras, in the ones that are not gentle, we can afirm that there is at least one non-trivial nearly Frobenius structure. Finally, in the case of the toupie algebras, we prove that the existence of monomial relations is a suficient condition to have non-trivial nearly Frobenius structure. Using the technics developed for the previous families of algebras we prove suficient conditions for the existence of non-trivial Frobenius structures in quotients of path algebras in general.
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.
Nearly Frobenius structures in some families of algebras
Dalia Artenstein, Ana González, Gustavo Mata
Abstract
In this article we continue with the study started in [2] of nearly Frobenius structures in some representative families of finite dimensional algebras, as radical square zero algebras, string algebras and toupie algebras. We prove that such radical square zero algebras with at least one path of length two are nearly Frobenius. As for the string algebras, those who are not gentle can be endowed with at least one non-trivial nearly Frobenius structure. Finally, in the case of toupie algebras, we prove that the existence of monomial relations is a sufficient condition to have non-trivial nearly Frobenius structures. Using the technics developed for the previous families of algebras we prove sufficient conditions for the existence of non-trivial Frobenius structures in quotients of path algebras in general.
Keywords: nearly Frobenius, string algebra, radical square zero algebra.
MSC: 16W99, 16G99.
1 Introduction
It is a well known result that the Poincaré algebra associated to a compact closed manifold with trace
[TABLE]
for is a Frobenius algebra. This is not the case for a non-compact manifold , but we may ask what structure remains. The answer may be stated nowaday as follows: the cohomology algebra is a nearly Frobenius algebra.
The concept of nearly Frobenius algebra was developed in the thesis of the second author of this article, the study of these objects was motivated by the result proved in [9], which states that: the homology of the free loop space has the structure of a Frobenius algebra without counit. Later, these objets were studied in [10] and their algebraic properties were developed in [2]. In particular, the possible nearly Frobenius structures in gentle algebras were described.
In the framework of differential graded algebras, Abbaspour considers in [1] nearly Frobenius algebras that he calls open Frobenius algebras. He proves that if is a symmetric open Frobenius algebra of degree , then is an open Frobenius algebra.
The Frobenius algebra structures in quotients of path algebras have been deeply studied. In particular, there is only a small family of monomial algebras that admits Frobenius structure. The next result illustrates this assertion.
Lemma 1** ([8], Lemma 2.2).**
Let be an indecomposable monomial algebra. Then is Frobenius if and only if , or for some positive integers and , with , where the basic oriented cycle of length and is the ideal generated by the arrows.
If we are interested in finding nearly Frobenius structures the situation is quite different. A large number of families of finite dimensional algebras can be endowed with nearly Frobenius structure. In particular, in this work, we study the nearly Frobenius structures in radical square zero algebras, string algebras and toupie algebras.
The radical square zero algebras are finite dimensional algebras over a field such that the square of its Jacobson radical is already zero. They have been extensively studied in representation theory because their behaviour provides interesting examples and results. Some beautiful results about them are the following: a radical square zero algebra is of representation finite type if and only if its separated quiver is a finite disjoint union of Dynkin diagrams (see [4, Chapter X2]). A connected radical square zero algebra is either self-injective or CM-free (see [7]). As a consequence of the last result, we have that a connected radical square zero algebra is Gorenstein if and only if its valued quiver is either an oriented cycle with the trivial valuation or does not contain oriented cycles.
String algebras are on one hand special biserial algebras whose ideal of relations can be generated by paths and on the other hand a generalization of gentle algebras. The class of special biserial algebras was introduced by Skowroński and Waschbüsch in [14]. It has played an important role in the study of self-injective algebras. Special biserial algebras and in particular string algebras have a well-understood representation theory. In fact, if is special biserial, then it has a two-sided ideal such that the quotient is a monomial algebra, and actually a string algebra.
The other family of algebras studied in this article are the toupie algebras. They were first introduced in “Toupie algebras, some examples of Laura algebra” by Diane Castonguay, Julie Dionne, Francois Huard and Marcelo Lanzilotta (see [11] for more details). These algebras combine features of canonical algebras with monomial algebras. They are quotients of the path algebra of a finite quiver which has a source [math], a sink and branches going from [math] to . The ideal of relations can be generated by a set containing two types of relations: monomial ones, which involve arrows of one branch each, and linear combinations of branches.
This work is developed as follows. In section 2 is devoted to study radical square zero algebras, in particular we determine the Frobenius dimension for all them. In Section 3 we study the string algebras and we prove that if is a not gentle string algebra then it admits at least one non-trivial nearly Frobenius coproduct. To finish the study of the algebras mentioned above we describe, in Section 4, the toupie algebras and the conditions ensuring that a toupie algebra admits nearly Frobenius coproducts. In the last section, based on all the previous results, we give conditions on quotients of path algebras so that they admit nearly Frobenius structures.
2 Preliminaries
Throughout this article always denote a field. We denote by a finite dimensional -algebra, and by Proposition 6 of [3], can be considered as a connected algebra.
Definition 2**.**
A quiver Q=\bigl{(}Q_{0},Q_{1},s,t\bigr{)} is a quadruple consisting of two sets: (whose elements are called points, or vertices) and (whose elements are called arrows), and two maps which associate to each arrow its source and its target , respectively.
An arrow of source and target is usually denoted by . A quiver Q=\bigl{(}Q_{0},Q_{1},s,t\bigr{)} is usually denoted simply by . Thus, a quiver is nothing but an oriented graph without any restriction on the number of arrows between two points, the existence of loops or oriented cycles.
Definition 3**.**
Let Q=\bigl{(}Q_{0},Q_{1},s,t\bigr{)} be a quiver and . A path of length with source and target (or, more briefly, from to ) is a sequence
[TABLE]
where for all , s\bigl{(}\alpha_{1}\bigr{)}=a, t\bigl{(}\alpha_{k}\bigr{)}=s\bigl{(}\alpha_{k+1}\bigr{)} for each , and t\bigl{(}\alpha_{l}\bigr{)}=b. Such a path is denoted briefly by .
Definition 4**.**
*Let be a quiver. The path algebra is the -algebra whose underlying -vector space has as its basis the set of all paths \bigl{(}a|\alpha_{1},\alpha_{2},\dots,\alpha_{l}|b\bigr{)} of length in and such that the product of two basis vectors
\bigl{(}a|\alpha_{1},\alpha_{2},\dots,\alpha_{l}|b\bigr{)} and \bigl{(}c|\beta_{1},\beta_{2},\dots,\beta_{k}|d\bigr{)} of is defined by*
[TABLE]
where denotes the Kronecker delta. In other words, the product of two paths and is equal to zero if t\bigl{(}\alpha_{l}\bigr{)}\neq s\bigl{(}\beta_{1}\bigr{)} and is equal to the composed path if t\bigl{(}\alpha_{l}\bigr{)}=s\bigl{(}\beta_{1}\bigr{)}. The product of basis elements is, then, extended to arbitrary elements of by distributivity.
Let be a quiver and be the associated path algebra. Denote by the two-sided ideal in generated by all paths of length 1, i.e. all arrows. This ideal is known as the arrow ideal.
It is easy to see that, for any we have that is a two-sided ideal generated by all paths of length . Note that we have the following chain of ideals:
[TABLE]
Definition 5**.**
A two-sided ideal in is said to be admissible if there exists such that
[TABLE]
Also, the algebra is monomial if is generated by paths.
An algebra is a connected algebra if and only if is a connected quiver (See Lema 1.7, chapter II of [5]), so, from now on, we only consider connected quivers.
Definition 6**.**
An algebra is a nearly Frobenius algebra if it admits a linear map such that
[TABLE]
commute.
In [3] was showed that the previous definition agree with the definition given in [2].
Definition 7**.**
The Frobenius space associated to an algebra is the vector space of all the possible coproducts that make it into a nearly Frobenius algebra (), see [2]. Its dimension over is called the Frobenius dimension of , that is,
[TABLE]
Gabriel’s Theorem states that if is a basic and connected finite dimensional -algebra over a algebraically closed field , there exist a quiver and an ideal of such that . This motivates us to study the existence of nearly Frobenius structures on quotients of path algebras.
3 Radical square zero algebras
Definition 8**.**
A radical square zero algebra is a finite dimensional algebra over al field such that the square of its Jacobson radical is already zero.
Now, if we consider a radical square zero algebra associated to , we can determine the general expression of a nearly Frobenius coproduct over a general vertex of and, using that the composition of two arrows is zero, determine the Frobenius dimension of the algebra .
Remark 9*.*
Let be a radical square zero algebra and . We can describe all the possible coproducts over , depending if is sink, source, or an intermediate vertex.
First, remember that if is a nearly Frobenius coproduct then \Delta\bigl{(}e_{p}\bigr{)}=\bigl{(}e_{p}\otimes 1\bigr{)}\Delta\bigl{(}e_{p}\bigr{)}=\Delta\bigl{(}e_{p}\bigr{)}\bigl{(}1\otimes e_{p}\bigr{)}, for .
- •
If is a source we can distinguish two different situations depending on the outdegree of . Let be the outdegree of .
If the situation is the following
\textstyle{p\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\beta}$$\textstyle{q_{1}\cdots} ,
then \Delta\bigl{(}e_{p}\bigr{)}=a_{0}e_{p}\otimes e_{p}+a_{1}\beta\otimes e_{p} and \Delta(\beta)=\Delta\bigl{(}e_{p}\bigr{)}(1\otimes\beta)=a_{0}e_{p}\otimes\beta+a_{1}\beta\otimes\beta=(\beta\otimes 1)\Delta\bigl{(}e_{q_{1}}\bigr{)}. We deduce that and obtain that
[TABLE]
If we have the following subquiver
[TABLE]
\Delta\bigl{(}e_{p}\bigr{)}=a_{0}e_{p}\otimes e_{p}+\sum_{i=1}^{n}b_{i}\beta_{i}\otimes e_{p}, then \Delta\bigl{(}\beta_{j}\bigr{)}=\Delta\bigl{(}e_{p}\bigr{)}\bigl{(}1\otimes\beta_{j}\bigr{)}=a_{0}e_{p}\otimes\beta_{j}+\sum_{i=1}^{n}b_{i}\beta_{i}\otimes\beta_{j}=\bigl{(}\beta_{j}\otimes 1\bigr{)}\Delta\bigl{(}e_{q_{j}}\bigr{)}. Therefore and for . Using that we conclude that for all . Then
[TABLE]
- •
If is a sink, similarly to the previous case, calling the indegree of , we can conclude that: if and is the arrow ending in , \Delta\bigl{(}e_{p}\bigr{)}=a_{1}e_{p}\otimes\alpha.
In the case that we have the following situation
[TABLE]
Reproducing the argument used in the case where is a source, we conclude that \Delta\bigl{(}e_{p}\bigr{)}=0.
- •
The last case is when is an intermediate vertex: Let be the indegree and the outdegree of . If and we have the following representation
[TABLE]
[TABLE]
[TABLE]
\displaystyle{\Delta\bigl{(}\alpha_{k}\bigr{)}=\bigl{(}\alpha_{k}\otimes 1\bigr{)}\Delta\bigl{(}e_{p}\bigr{)}=a_{0}\alpha_{k}\otimes e_{p}+\sum_{i=1}^{m}a_{i}\alpha_{k}\otimes\alpha_{i}=\Delta\bigl{(}e_{p_{k}}\bigr{)}\bigl{(}1\otimes\alpha_{k}\bigr{)}},
then , and for all , for . Similarly
\displaystyle{\Delta\bigl{(}\beta_{k}\bigr{)}=\Delta\bigl{(}e_{p}\bigr{)}\bigl{(}1\otimes\beta_{k}\bigr{)}=\sum_{j=1}^{n}b_{j}\beta_{j}\otimes\beta_{k}=\bigl{(}\beta_{k}\otimes 1\bigr{)}\Delta\bigl{(}e_{q_{k}}\bigr{)}}, then for all , for . Then
[TABLE]
If one or more arrows are loops, then the result is equal to the previous case.
If and and it is not an isolated loop, then \displaystyle{\Delta\bigl{(}e_{p}\bigr{)}=ae_{p}\otimes\alpha+b\beta\otimes e_{p}+c\beta\otimes\alpha}.
If and with loop in , then \displaystyle{\Delta\bigl{(}e_{p}\bigr{)}=a(e_{p}\otimes\alpha+\alpha\otimes e_{p})+b\alpha\otimes\alpha}.
If and with local representation of the quiver
[TABLE]
\displaystyle{\Delta\bigl{(}e_{p}\bigr{)}=a_{0}e_{p}\otimes e_{p}+a_{1}e_{p}\otimes\alpha+\sum_{j=1}^{n}b_{j}\beta_{j}\otimes e_{p}+\sum_{j=1}^{n}c_{j}\beta_{j}\otimes\alpha}, then \Delta(\alpha)=a_{0}\alpha\otimes e_{p}+a_{1}\alpha\otimes\alpha=\Delta\bigl{(}e_{p}\bigr{)}(1\otimes\alpha), therefore . Similarly,
\displaystyle{\Delta\bigl{(}\beta_{k}\bigr{)}=\Delta\bigl{(}e_{p}\bigr{)}\bigl{(}1\otimes\beta_{k}\bigr{)}=\sum_{j=1}^{n}b_{j}\beta_{j}\otimes\beta_{k}=(\beta\otimes 1)\Delta\bigl{(}e_{q_{1}}\bigr{)}}, then for al , for . Finally,
[TABLE]
If and with a loop in and arrows starting in , then
[TABLE]
If and and the quiver is locally the following
[TABLE]
As before \displaystyle{\Delta\bigl{(}e_{p}\bigr{)}=b_{1}\beta\otimes e_{p}+\sum_{i=1}^{m}c_{i}\beta\otimes\alpha_{i}}.
Finally if and with a loop in and arrows ending in , then
[TABLE]
The previous remark allows us to construct all the nearly Frobenius coproducts that the radical square zero algebra admits.
Corollary 10**.**
If with is a radical square zero algebra then .
Proof.
Since there is at least one path of length two. Let us call the intermediate vertex of that path, the indegree and the outdegree of . Then the situation is similar to the last case of the previous remark and .
∎
Finally we determine the Frobenius dimension of radical square zero algebras. For this reason we introduce the following useful notations for some special subspaces of and : , , , , where , and where . From Remark 9 we can obtain the following result:
Proposition 11**.**
If is a radical square zero algebra, then
[TABLE]
Proof.
Suppose that is a nearly Frobenius coproduct on . It is clear that .
Claim: if with and , then .
Let be an arrow of such that and . Since , then with and . It is easy to see that , hence .
Then can be generated by vectors in for all .
Claim: if is an arrow of such that and , where and where then .
Using that is a nearly Frobenius structure we have that .
- •
- •
On the other hand the set is linearly independents, then the equations agree only if .
Finally, using Remark 9 when the indegree or outdegree are greater or equal to two, it follows the thesis statement. ∎
4 String algebras
The class of special biserial algebras was studied by Skowronski and Waschbüsch in [14] where they characterize the biserial algebras of finite representation type. The definition of these algebras can be given in terms of conditions on the associated bound quiver . A classification of the special biserial algebras which are minimal representation-infinite has been given by Ringel in [13]. There is a beautiful description of all finite-dimensional indecomposable modules over special biserial algebras: they are either string modules or band modules or non-uniserial projective-injective modules, see [6], [15].
Definition 12**.**
A bound quiver is special biserial if it satisfies the following conditions:
- (S1)
Each vertex in is the source of at most two arrows and the target of at most two arrows. 2. (S2)
For an arrow in there is at most one arrow and at most one arrow such that and .
*If the ideal I is generated by paths, the bound quiver is string.
An algebra is called special biserial (or string) if it is isomorphic to with a special biserial bound quiver (or a string bound quiver, respectively).
An algebra is called string quadratic if the ideal is generated by paths of length two.*
Remark 13*.*
Note that every gentle algebra is a string algebra. In [2] all the nearly Frobenius structures for a gentle algebra associated to are determined, where is a finite, connected and acyclic quiver. The natural question is if we can generalize this result for the family of string algebras.
The first step is to study the family of string quadratic algebras.
Remark 14*.*
If is a string quadratic algebra that is not gentle then there is at least one vertex in one of the following local situation:
[TABLE]
[TABLE]
In the first case we have that:
[TABLE]
Let us evaluate on the arrows to obtain conditions about the coefficients above.
For the arrow ,
[TABLE]
so and . For and ,
[TABLE]
and
[TABLE]
from the first equation we conclude that and from the second one and .
As a consecuence we obtain that
[TABLE]
[TABLE]
[TABLE]
Symmetrically, in the second case:
[TABLE]
[TABLE]
[TABLE]
In the third case the relations are , and and
[TABLE]
As before, we will evaluate on the arrows. First consider
[TABLE]
then and . Secondly,
[TABLE]
therefore and . Finally, for the last two arrows we obtain the following equations:
[TABLE]
[TABLE]
from the first equation we deduce that and and from the second one and . In conclusion,
[TABLE]
[TABLE]
[TABLE]
[TABLE]
In the last case we have that , , and . As before,
[TABLE]
Evaluating again on the arrows we obtain the following equations:
[TABLE]
and analogously to the previous cases, we conclude that
[TABLE]
[TABLE]
Theorem 15**.**
If is a string quadratic algebra but not a gentle algebra, then it has a non-trivial nearly Frobenius algebra structure.
Proof.
We will construct a non-zero nearly Frobenius coproduct for this type of algebras. Note that there is at least one vertex in one of the four possible cases of Remark 14. Let us describe the coproduct in each of them.
Suppose that we are locally in the first case. Analogously to the Remark 14 we can define a coproduct
[TABLE]
such that , where and are as in Remark 14 and for any other vertex .
To check that is well defined we need to verifiy that .
For the first arrow , .
In the case of and we have that and .
This proves that is a nearly Frobenius coproduct.
In the second case we can proceed in the same way as before and prove that defined as and otherwise is a nearly Frobenius coproduct.
In the third local possibility let us define as follows, , for and otherwise. It is straightforward to verifiy that is well defined and a nearly Frobenius coproduct.
Finally, in the last case we define , with and for any other vertex .
∎
Now, we study the general case, that is, is a string algebra.
Theorem 16**.**
Let be a string algebra, not quadratic. Then, admits at least one non-trivial nearly Frobenius coproduct.
Proof.
Since is not gentle there are extra monomial relations. If one of the extra relations is of lenght we are locally in one of the four cases of Remark 14 and we conclude that is nearly Frobenius.
If all the extra relations are of lenght greater that 2 choose one relation named . Locally, we are in the following situation:
[TABLE]
where from every vertex arrows might come in or out in the string scheme.
Claim if and otherwise is a coproduct in :
If ,
On the other hand,
and analogously, .
Suppose there is a path from a vertex not involved in the relation to an intermediate vertex . We can consider two different cases.
If contains the reasoning is similar to . If not, there is an arrow such that is an intermediate vertex and since is string . Then we conclude that . The case were there is a path from an intermediate vertex to another vetex not involved in is analogous and the result holds.
∎
5 Nearly Frobenius structures on toupie algebras
In this section we first prove that canonical algebras only admit the trivial nearly Frobenius algebra estructure. Then we characterize when a toupie algebra has a non-trivial nearly Frobenius structure.
Definition 17**.**
Canonical algebras were introduced in [12]. Let be a field, \mathbf{n}=\bigl{(}n_{1},\cdots,n_{t}\bigr{)} be a sequence of positive integers (weights), and \boldsymbol{\lambda}=\bigl{(}\lambda_{3},\dots,\lambda_{t}\bigr{)} be a sequence of pairwise distinct elements of . A canonical algebra of type is an algebra where is
[TABLE]
, and is the ideal in the path algebra generated by the following linear combinations of paths from [math] to :
[TABLE]
Theorem 18**.**
Let be a canonical algebra over a field , then .
Proof.
Since is a canonical algebra then with as in the previous figure, and
[TABLE]
If we have a coproduct the next condition is required
[TABLE]
This implies that the coproduct in is
[TABLE]
and the coproduct in is
[TABLE]
The coproduct in is given by
[TABLE]
Then
[TABLE]
By comparison we deduce that and
[TABLE]
Finally, if we replace by for we get that
[TABLE]
As before
[TABLE]
Then, \Delta\bigl{(}\alpha^{(2)}\bigr{)}=a\alpha^{(1)}\otimes\alpha^{(2)}=a\alpha^{(2)}\otimes\alpha^{(1)}. Therefore . This implies that ∎
Definition 19**.**
A quiver is called toupie if it has a unique source [math] and a unique sink , and, for any other vertex there is exactly one arrow having as source and exactly one arrow having as target:
[TABLE]
We say that is a toupie algebra if with a toupie quiver, and any admissible ideal.
Proposition 20**.**
If is a commutative diamond, then has . Moreover, the only nearly Frobenius structure over is the linear structure in each branch.
Proof.
The quiver associated to is of the form:
\textstyle{0\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\beta_{1}}$$\scriptstyle{\alpha_{1}}$$\textstyle{\bullet\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\beta_{2}}$$\textstyle{\bullet\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\alpha_{2}}$$\textstyle{\vdots\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\vdots\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\bullet\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\beta_{m}}$$\textstyle{\bullet\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\alpha_{n}}$$\textstyle{\omega}
and . This means that in .
If is a nearly Frobenius coproduct then
[TABLE]
and
[TABLE]
If we use that , and we deduce that
[TABLE]
where As a consequence of these equalities we have that
[TABLE]
and
[TABLE]
∎
The next corollary is a generalization of the previous proposition and is proved in an analogous way. We call the algebra involved a generalized commutative diamond.
Corollary 21**.**
If with a quiver with branches () and . Then has . Moreover, the only nearly Frobenius structure over is the linear structure in each branch.
Theorem 22**.**
Let be a toupie algebra over a field . If , the number of monomial relations, is zero and is not the linear quiver , then , except from the case of the (generalized) commutative diamond, in which .
Proof.
Let us first consider an order in the branches of the toupie algebra. Suppose that there are branches. We can modify the non-monomial relations to have the first branches to be linearly independent and for , . Let us call the length of Now, as in Theorem 18,
[TABLE]
and the coproduct in is
[TABLE]
The coproduct in is given by
[TABLE]
Then
[TABLE]
By comparison we deduce that
[TABLE]
Finally, if we replace by for we obtain that
[TABLE]
In the case that the toupie algebra is the generalized diamond see Corollary 21. If not, it has at least two linearly independent branches and ,
[TABLE]
Then, \Delta\bigl{(}\alpha^{(2)}\bigr{)}=a\alpha^{(1)}\otimes\alpha^{(2)}=a\alpha^{(2)}\otimes\alpha^{(1)}. Therefore . This implies that \Delta\bigl{(}\alpha^{(i)}\bigr{)}=0 and ∎
Next we will consider the case of toupie algebras with monomial relations.
Proposition 23**.**
Consider with a monomial ideal. Then .
Proof.
Suppose that has a relation of length of the form:
[TABLE]
Let us now define a coproduct in :
if and otherwise.
If and there exists a non-null path from to with then, by definition and on the other hand since must be included in .
If the argument is analogous.
Finally, if , the only path from to is and, so we conclude that is a coproduct.
∎
The next theorem describes toupie algebras with nearly Frobenius structures.
Theorem 24**.**
Let be a toupie algebra over a field and the number of branches with monomial relations, then
- (1)
if
- (a)
and is the linear quiver or the (generalized) commutative diamond we have that , 2. (b)
in other case , 2. (2)
if then
Proof.
-
(1)
-
(a)
This result is a consequence of Theorem 1 of [2] and the Corollary 21. 2. (b)
It is the Theorem 22. 2. (2)
If there is only one branch and has monomial relations is the case of Proposition 23. If not, using Theorem 4 of [2], the coproduct over the branches is zero except for the monomial branches, moreover, the coproduct on the first and the last arrow of the monomial branches is zero. Then, combining this result with Proposition 23 over any monomial branch we have that
[TABLE]
∎
6 Final Comment
After computing the nearly Frobenius structures in some representative classes of algebras, we observe that some local situations guarantee the existence of non-trivial nearly Frobenius structures. We summarized them in the following result.
Theorem 25**.**
Let be a finite dimensional algebra. If has a local situation in a vertex as follows
[TABLE]
[TABLE]
[TABLE]
then has a non-trivial structure of nearly Frobenius algebra.
Proof.
The coproduct in the first four cases is analogous to the ones in Theorem 15. For the last case the coproduct is the following
[TABLE]
∎
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[1] H. Abbaspour. On the Hochschild homology of open Frobenius algebras. J. Noncommut. Geom., 10 (2) pp. 709-743 (2016).
- 2[2] D. Artenstein, A. González, and M. Lanzilotta, Constructing Nearly Frobenius Algebras , Algebras and Representation Theory (2015) Volume 18, 339-367.
- 3[3] D. Artenstein, A. González, G. Mata Algebraic constructions on the Nearly Frobenius category
- 4[4] M. Auslander, I. Reiten, S. Smalø. Representation Theory of Artin Algebras , Cambridge Studies in Advanced Mathematics 36 , Cambridge University Press (1997).
- 5[5] I. Assem, D. Simson,A. Skowroński, Elements of the Representation Theory of Associative Algebra 1 : Techniques of Representation Theory, LMSST 65, Cambridge Univ. Press, Cambridge (2006).
- 6[6] M. C. R. Butler and Claus Michael Ringel, Auslander-Reiten sequences with few middle terms and applications to string algebras, Comm. Algebra 15 (1987), no. 1-2, 145–179.
- 7[7] X.-W. Chen: Algebras with radical square zero are either self-injective or CM-free , Proc. Amer. Math. Soc. 140 (1) pp. 93-98 (2012).
- 8[8] X.-W. Chen, H.-L. Huanga, Y. Ye and P. Zhang, Monomial Hopf algebras, Journal of Algebra (2004), Volume 275, Issue 1, 212-232.
