$\Delta$-filtrations and projective resolutions for the Auslander-Dlab-Ringel algebra
Teresa Conde

TL;DR
This paper investigates the structure of modules over the ADR algebra, a special class of quasihereditary algebra, providing new insights into $ ext{Δ}$-filtrations, projective covers, and counterexamples to existing claims.
Contribution
It introduces a detailed study of $ ext{Δ}$-filtrations and projective resolutions for ADR algebras, including a counterexample to a previous conjecture.
Findings
Characterization of $ ext{Δ}$-filtrations for RUSQ algebras
Determination of projective covers for specific R_A-modules
Counterexample to a claim by Auslander-Platzeck-Todorov
Abstract
The ADR algebra of an Artin algebra is a right ultra strongly quasihereditary algebra (RUSQ algebra). In this paper we study the -filtrations of modules over RUSQ algebras and determine the projective covers of a certain class of -modules. As an application, we give a counterexample to a claim by Auslander-Platzeck-Todorov, concerning projective resolutions over the ADR algebra.
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.
-filtrations and projective resolutions for the Auslander–Dlab–Ringel algebra
Teresa Conde
Institute of Algebra and Number Theory, University of Stuttgart
Pfaffenwaldring 57, 70569 Stuttgart, Germany
Abstract.
The ADR algebra of an Artin algebra is a right ultra strongly quasihereditary algebra (RUSQ algebra). In this paper we study the -filtrations of modules over RUSQ algebras and determine the projective covers of a certain class of -modules. As an application, we give a counterexample to a claim by Auslander–Platzeck–Todorov, concerning projective resolutions over the ADR algebra.
Key words and phrases:
Quasihereditary algebra, strongly quasihereditary algebra, ADR algebra
2010 Mathematics Subject Classification:
Primary 16S50, 16W70. Secondary 16G10, 16G20.
Most of this work is contained in the author’s Ph.D. thesis. This was supported by the grant SFRH/BD/84060/2012 of Fundação para a Ciência e a Tecnologia, Portugal. The author would like to express her gratitude to Stephen Donkin, Ph.D. examiner, for the simplified version of the proofs of Lemma 3.2 and Corollary 3.3 included in this article. In addition, the author would like to thank her Ph.D. supervisor, Karin Erdmann, for many useful comments.
1. Introduction
It is natural to ask whether there exist “Schur algebras” for arbitrary Artin algebras. That is, given an Artin algebra , we would like to have an -module whose endomorphism algebra is quasihereditary, so that it has finite global dimension and a highest weight theory. Such modules do exist. A suitable candidate was introduced by Auslander in [1]. He showed that the endomorphism algebra of
[TABLE]
has finite global dimension (here denotes the Loewy length of ). Subsequently, Dlab and Ringel proved that this endomorphism algebra is actually a quasihereditary algebra ([9]). For practical purposes one considers the basic version of instead. We denote this ‘Schur-like’ endomorphism algebra by and call it the Auslander–Dlab–Ringel algebra (ADR algebra) of . The original algebra is then Morita equivalent to for an idempotent in , and this is also analogous to the situation of symmetric groups and Schur algebras.
It was recently proved in [8] that the ADR algebra has a particularly neat quasihereditary structure. The ADR algebra is not only right strongly quasihereditary in the sense of Ringel ([22]); is actually a right ultra strongly quasihereditary algebra (RUSQ algebra) as defined in [8] (see also [7, Chapter 2]). The ADR algebra is not the only strongly quasihereditary algebra arising from a module theoretical context. Other examples of strongly quasihereditary algebras include: the Auslander algebras, associated to algebras of finite type; the endomorphism algebras constructed by Iyama, used in his famous proof of the finiteness of the representation dimension of Artin algebras ([16]); certain cluster-tilted algebras studied by Geiß–Leclerc–Schröer ([15], [14]), Buan–Iyama–Reiten–Scott ([4]) and Iyama–Reiten ([17]). The cluster-tilted algebras in [4], [15], [14] and [17] are actually RUSQ up to dualisation, as implicitly proved in [15] (see also [7]).
This paper complements the investigation on RUSQ algebras and on ADR algebras initiated in [8]. We start by studying the -filtrations of modules over RUSQ algebras. In Section 3, we show that the RUSQ algebras satisfy the following key property: every submodule of a direct sum of standard modules is still a direct sum of standard modules. This has several consequences. In particular, it gives rise to special (uniquely determined) filtrations of -good modules over RUSQ algebras, called -semisimple filtrations. These can be described recursively as follows. Given a -good module , let be the largest submodule of which is a direct sum of standard modules, and for define as the module satisfying the identity .
The ideas in Section 3 are then applied to the ADR algebra. As a main contribution of Section 4, we prove the following (which corresponds to Lemma 4.3 and Theorem 4.4).
Theorem**.**
Let be in . Then lies in and the socle series of determines the -semisimple filtration of . Formally,
[TABLE]
for all . The factors of the -semisimple filtration of only depend on the factors of the socle series of and on the Loewy length of the projective indecomposable -modules.
Next, we describe the right minimal -approximations of rigid modules in , or equivalently, the projective covers of the -modules , with rigid. Recall that a module is said to be rigid if its radical series coincides with its socle series. We prove the following theorem in Section 5.
Theorem**.**
Let be a rigid module in , with Loewy length . Then the projective cover of in is a right minimal -approximation of .
This simple yet useful result, combined with the conclusions in Section 4, is then used to provide a counterexample to a claim by Auslander, Platzeck and Todorov in [2, §7], about the projective resolutions of modules over the ADR algebra, for which no proof was given. To be precise, we show the following.
Proposition**.**
ADR algebras need not satisfy the descending Loewy length condition on projective resolutions.
2. Preliminaries
In this section we introduce the language of preradicals and give some background on quasihereditary algebras, RUSQ algebras and the ADR algebra.
Throughout this paper the letters and shall denote arbitrary Artin algebras over some commutative artinian ring . All the modules will be finitely generated left modules. The notation will be used for the category of (finitely generated) -modules. Given a class of -modules , let be the full subcategory of consisting of all modules isomorphic to a summand of a direct sum of modules in . The additive closure of a single module is denoted by .
2.1. Preradicals
Preradicals generalise the classic notions of radical and socle of a module. The results and definitions stated in this section are elementary and most of the proofs may be found in [3], [6, Chapter 2] and [23, Chapter VI] (see also [7, Section ]).
2.1.1. Definition and first properties
Definition 2.1**.**
A preradical in is a subfunctor of the identity functor , i.e., assigns to each module a submodule , such that each morphism induces a morphism given by restriction.
A submodule of a -module is called a characteristic submodule of if , for every in . By definition, it is clear that the module is a characteristic submodule of , for every preradical and for every module . It is also evident that every preradical is an additive functor which preserves monics.
To each preradical we may associate the functor
[TABLE]
which maps to . Note that the functor preserves epics.
Example 2.2**.**
For any class of -modules, the operators defined by
[TABLE]
for in , are preradicals in . The module , called the trace of in , is the largest submodule of generated by . Symmetrically, , the reject of in , is the submodule of such that is the largest factor module of cogenerated by . If is a complete set of simple -modules, then and .
The statements below are immediate consequences of the definition of preradical.
Lemma 2.3**.**
Let be a preradical in . Let and be -modules, with , and let be a finite family of -modules. The following hold:
- (1)
; 2. (2)
; 3. (3)
.
2.1.2. Hereditary and cohereditary preradicals
We shall now look at preradicals which satisfy specific properties.
Definition 2.4**.**
A preradical is called idempotent if . Symmetrically, we say that is a radical if .
Note that is an idempotent preradical for every class of -modules . Similarly, the functor is a radical.
Definition 2.5**.**
A preradical is hereditary if , for all and in such that . Dually, a preradical is said to be cohereditary if for , and in .
Example 2.6**.**
The functors and are the typical examples of a hereditary preradical and of a cohereditary preradical, respectively.
Lemma 2.7** ([23, Chapter VI, §1]).**
Let be a preradical in . The following statements are equivalent:
- (1)
* is hereditary;* 2. (2)
* is a left exact functor;* 3. (3)
the functor preserves monics.
Moreover, any hereditary preradical is idempotent.
Remark 2.8*.*
There is a result ‘dual’ to Lemma 2.7 for cohereditary preradicals.
It is possible to construct hereditary (and cohereditary) preradicals out of special classes of modules.
Definition 2.9**.**
A class of modules in is hereditary if every submodule of a module in is generated by .
Lemma 2.10**.**
If is a hereditary class of modules then is a hereditary preradical in .
Proof.
Consider , with and in . The module is generated by some module which is a (finite) direct sum of modules in . Consider the pullback square
[TABLE]
As is a hereditary class, is generated by . Hence is generated by as well. Since is the largest submodule of generated by , we must have . ∎
2.2. Quasihereditary algebras, RUSQ algebras and the ADR algebra
We now introduce some notation and state basic results about RUSQ algebras and ADR algebras.
2.2.1. Quasihereditary algebras
Given an Artin algebra , we may label the isomorphism classes of simple -modules by the elements of a finite poset . Denote the simple -modules by , , and use the notation (resp. ) for the projective cover (resp. injective hull) of .
Let be the standard module with label , that is
[TABLE]
The module is the largest quotient of whose composition factors are all of the form , with . Dually, denote the costandard modules by . The module
[TABLE]
is the largest submodule of with all composition factors of the form , with .
Let be the category of all -modules which have a -filtration, that is, a filtration whose factors are standard modules. The category is defined dually. We call a -good module.
The notation will be used for the multiplicity of a simple module in the composition series of . In a similar manner, shall denote the multiplicity of in a -filtration of a module in (this is well defined).
Definition 2.11**.**
We say that is quasihereditary if the following hold for every :
- (1)
; 2. (2)
; 3. (3)
, and .
Let be quasihereditary. It was proved by Ringel in [21] (see also [13]) that there is a unique indecomposable -module (up to isomorphism) for every which has both a - and a -filtration, with one composition factor labelled by , and all the other composition factors labelled by , . The standard module can be embedded in – the corresponding monomorphism is a left minimal -approximation of and lies in .
2.2.2. Strongly and ultra strongly quasihereditary algebras
Following Ringel ([22]), a quasihereditary algebra is said to be right strongly quasihereditary if for all . This property holds if and only if the category is closed under submodules (see [10], [12, Lemma *] and [22, Appendix]).
Let be an arbitrary quasihereditary algebra, as before. Additionally, suppose that satisfies the following two conditions:
**(A1): **
for all (i.e. is right strongly quasihereditary);
**(A2): **
for all such that .
We call these algebras right ultra strongly quasihereditary algebras (RUSQ algebras, for short).
Remark 2.12*.*
It was proved in [7, §] that the definition of RUSQ algebra given in [8] is equivalent to the one above.
Let be a RUSQ algebra. It is always possible to label the elements in as
[TABLE]
for certain , so that implies that and (see [8, §5]). We shall always assume that the elements in are labelled in such a way.
The following proposition summarises some properties of the RUSQ algebras.
Proposition 2.13** ([8, §5]).**
Let be a RUSQ algebra. The following hold:
- (1)
* is closed under submodules;* 2. (2)
* for , and ;* 3. (3)
each is uniserial and has composition factors , ordered from the top to the socle; 4. (4)
; 5. (5)
for , the number of standard modules appearing in a -filtration of is given by ; 6. (6)
a module belongs to if and only if is a (finite) direct sum of modules of type .
2.2.3. The ADR algebra
Fix an Artin algebra . Given a module in , we shall denote its Loewy length by . Let have Loewy length (as a left module). We want to study the basic version of the endomorphism algebra of .
For this, let be a complete set of projective indecomposable -modules and let be the Loewy length of . Define
[TABLE]
The Auslander–Dlab–Ringel algebra of (ADR algebra of ) is defined as
[TABLE]
The projective indecomposable -modules are given by
[TABLE]
for , .
Denote the simple quotient of by and define
[TABLE]
so that labels the simple -modules. Define a partial order, , on by
[TABLE]
According to [8, §4], is a RUSQ algebra and the labelling is compatible with (1).
Theorem 2.14** ([8, §3, §4]).**
The ADR algebra is a RUSQ algebra. For every in , we have and there are short exact sequences
[TABLE]
3. -semisimple modules and -semisimple filtrations
For a quasihereditary algebra , we say that a -module is -semisimple if it is a direct sum of standard modules. Every module in has some submodule such that:
**(B): **
is -semisimple and is in .
Given a module in , we may consider the submodules of which are maximal with respect to property **(B): **. The module may have more than one such submodule (see Example in [18]). However, according to [18], the submodules of which are maximal with respect to **(B): ** are unique up to isomorphism.
Suppose now that is a RUSQ algebra. The -semisimple modules over RUSQ algebras are particularly well behaved. As we will see in Corollary 3.3, the property of being -semisimple is closed under submodules in this case. Furthermore, every module in has exactly one submodule which is maximal with respect to property **(B): **. The module is actually the unique maximal -semisimple submodule of (with respect to inclusion). Moreover, will be obtained by applying a certain hereditary preradical (as in Definition 2.5) to the module . Since still lies in , we may proceed iteratively and define the -semisimple filtration (which will be unique) and the -semisimple length of any module in .
3.1. -semisimple modules
We are interested in submodules of -good modules which are maximal with respect to property **(B): **. As a consequence of Theorem in [18], these are unique up to isomorphism.
Theorem 3.1** ([18, Theorem ]).**
Let be a quasihereditary algebra, and let be in . Any two submodules of which are maximal with respect to property **(B): ** are isomorphic.
Note that -semisimple modules are in general well behaved with respect to quotients in the following way: every -good factor module of a -semisimple module is still -semisimple. This assertion follows from the fact that is closed under taking kernels of epimorphisms ([12, Lemma ]) and from Theorem in [18].
We wish to study the -semisimple modules over a RUSQ algebra . We shall assume that the set is as described in (1). In this subsection we prove some key properties of the -semisimple modules over RUSQ algebras. Namely, we show that the property of being -semisimple is closed under taking submodules.
Lemma 3.2**.**
Let be a RUSQ algebra. Let be in and suppose that there is a short exact sequence
[TABLE]
with . If , then (5) splits.
Proof.
Note that (see Proposition 2.13, part 3). As , there is some nonzero submodule of such that . Let be the morphism which embeds in and maps to zero. By the injectivity of , extends to a map . Note that as is an injective map. Thus is a submodule of .
By part 4 of Proposition 2.13, is in . Since is closed under taking submodules, then both and lie in . Denote by the number of standard modules appearing in a -filtration of . As , and , it follows that . So either and , or and .
If , then the submodule of must coincide with (as both modules have the same Jordan–Hölder length). In this case the monic splits.
If , then as
[TABLE]
But then part of Lemma in [12] implies that (5) is a split exact sequence. ∎
We now use the previous result to give a characterisation of the -semisimple modules over a RUSQ algebra.
Corollary 3.3**.**
Let be a RUSQ algebra and let be in . Then is -semisimple if and only if the number of simple summands of coincides with the number of factors in a -filtration of . Moreover, any submodule of a -semisimple module is still -semisimple.
Proof.
Let be in . Denote by the following assertion: “the number of simple summands of coincides with the number of factors in a -filtration of ”. By parts 5 and 6 of Proposition 2.13, is true if and only if the composition factors of of type are exactly the summands of its socle. From this equivalence, it is easy to see that the truth of implies the truth of for . Let be a submodule of such that . Using Proposition 2.13, we also conclude that the truth of implies the truth of .
If is a -semisimple module then is clearly true. Suppose now that holds for . We wish to show that is -semisimple. We prove this by induction on the number of factors in a -filtration of . If the result is obvious. Suppose now that . Let be a submodule satisfying and for some . By the remark in the first paragraph, holds. Using induction, we conclude that is -semisimple. Therefore , where each is isomorphic to a standard module. Applying again the observations in the first paragraph, we deduce that assertion must hold, so, by induction, each is a -semisimple module. Using that both and are standard modules, we conclude that , where is a submodule of isomorphic to a standard module. Note that is a submodule of which has the same Jordan–Hölder length as . This implies that , which proves that is -semisimple. We have just shown that is a -semisimple module if and only if assertion holds.
Let now be a submodule of a -semisimple module . Then is true, which implies that holds. Therefore is a -semisimple module. ∎
In the next subsection we are going to show that every -good module over a RUSQ algebra has a unique maximal -semisimple submodule. First, we check that arbitrary quasihereditary algebras do not possess this property.
Example 3.4**.**
Consider the quiver
[TABLE]
and the bound quiver algebra , where is the ideal generated by the elements and . The algebra is quasihereditary with respect to the labelling poset . The modules
[TABLE]
are the corresponding standard -modules. The projective cover of the simple module with label 2 has the following structure
[TABLE]
The modules and are both maximal -semisimple submodules of . The quotient of by each of these submodules does not belong to , i.e. none of these submodules of satisfies property **(B): **.
3.2. The preradical and -semisimple filtrations
Let be an arbitrary quasihereditary algebra. As pointed out in the previous subsection, the submodules of a module in which are maximal with respect to property **(B): ** are all isomorphic, but they are not necessarily unique. We have also seen that a module in may have more than one maximal -semisimple submodule with respect to inclusion (Example 3.4). We shall prove that both these maximal submodules are unique and actually coincide when the underlying algebra is a RUSQ algebra. For this, we use the general theory of preradicals introduced in Subsection 2.1.
Lemma 3.5**.**
Let be a RUSQ algebra. The corresponding set of standard -modules is a hereditary class in . In particular, is a hereditary preradical in .
Proof.
Let be a submodule of a module in , so is contained in some -semisimple module . By Corollary 3.3, is still -semisimple, so it is trivially generated by . Hence the set is hereditary. Lemma 2.10 implies that is a hereditary preradical in . ∎
Remark 3.6*.*
The preradical is not usually hereditary for an arbitrary quasihereditary algebra (not even if is right strongly quasihereditary).
From now onwards we shall denote the functor by .
Definition 3.7**.**
For a RUSQ algebra , let be the hereditary preradical in .
Next, we give a description of the submodule of a module .
Proposition 3.8**.**
Let be a RUSQ algebra, and let . Then is the largest -semisimple submodule of . Furthermore, lies in . In particular, is the largest submodule of satisfying property **(B): **.
Proof.
By the definition of , there is an epic from a -semisimple module to . Note that both and are in , since this category is closed under submodules. As a consequence, must be a split epic. Hence is -semisimple. By the definition of it is clear that every -semisimple submodule of must be contained in . This shows that is the largest -semisimple submodule of .
To conclude this proof it is enough to show that lies in . We start by proving that this holds for the injective modules (recall Proposition 2.13). Note that , as is a submodule of . Since has simple socle , then has to be isomorphic to some standard module . But then we must have , and consequently is in . Let now be a finite direct sum of injective modules of type . The module still lies in because preradicals preserve finite direct sums (see part 3 of Lemma 2.3). Consider now in . By Proposition 2.13, the injective hull of is such that is a direct sum of injectives of type . By part 3 of Lemma 2.7, gives rise to a monic , and by our previous observation lies in . As is closed under submodules, the module belongs to . ∎
Example 3.9**.**
Note that for an arbitrary quasihereditary algebra the modules , , are not usually -semisimple (not even -good). Indeed, for the algebra in Example 3.4, we have , which is not -semisimple.
3.2.1. Filtrations arising from preradicals
Our next goal is to define -semisimple filtration and -semisimple length for modules in , over some RUSQ algebra . For this, some elementary results about preradicals are needed.
Let and be preradicals (over an arbitrary Artin algebra ). Write if is a subfunctor of . The functor is a preradical, and . For in define as the submodule of containing , satisfying
[TABLE]
The operator is still a predadical. By construction, .
By the characterisation of hereditary radicals given in Lemma 2.7, it follows that is hereditary if both and are hereditary. We also have that is hereditary, whenever and are both hereditary – the functor is naturally isomorphic to .
Similarly to the composition of preradicals, the operation is associative. Given a preradical , let be the identity functor in and let be the zero preradical. For , define and . We summarise the properties of these preradicals.
Lemma 3.10**.**
Let be a preradical in .
- (1)
*For every , and . * 2. (2)
For every in there is such that . 3. (3)
*For every in there is such that . *
The preradicals (and ), , give rise to special filtrations.
Lemma 3.11** ([7, §]).**
Let be a preradical. Suppose that for every nonzero -module . Given in , there is a unique integer such that , and for every satisfying . Moreover, for , we have
[TABLE]
Lemma 3.12** ([7, §]).**
Let be a hereditary preradical. Then is also a hereditary preradical, and for every . Furthermore, if for every , the following hold for and in :
- (1)
*if then ; * 2. (2)
if , then is the largest submodule of such that .
3.2.2. -semisimple filtrations and -semisimple length
Suppose once again that is a RUSQ algebra, and consider the hereditary preradical . Note that for every nonzero module in as
[TABLE]
In fact, we have . We may construct the preradicals in defined recursively in §3.2.1. Then Lemmas 3.10, 3.11 and 3.12 hold for the preradicals . In particular, is a hereditary preradical for every .
Lemma 3.13**.**
Let be a RUSQ algebra. If is in then so is , for any .
Proof.
By Proposition 3.8, the claim holds for . Suppose . Then
[TABLE]
so by induction belongs to . ∎
Given a module in , we may consider the filtration
[TABLE]
where is as defined in Lemma 3.11. The factors of this filtration are -semisimple: by Lemma 3.13 and Proposition 3.8 the modules are -semisimple. We call (6) the -semisimple filtration of .
Definition 3.14**.**
The -semisimple length of a module in , denoted by , is the length of the -semisimple filtration of , i.e. it is given by the number (as in Lemma 3.11).
4. -semisimple filtrations of modules over the ADR algebra
The ADR algebra of an Artin algebra , , is our prototype of a RUSQ algebra. We now prove some results specific to the -semisimple filtrations of -good modules over the ADR algebra. Throughout this section the underlying quasihereditary algebra will be , where the poset is as defined in (2) and (3). For the proof of the next results note that the left exact functor is fully faithful since is a generator of (see [1, §8–§10]).
Lemma 4.1**.**
Let and be in , with . There is a canonical embedding
[TABLE]
and the induced morphisms
[TABLE]
are isomorphisms.
Proof.
The functor is left exact. Thus, it maps the canonical epic to the morphism , which factors as
[TABLE]
Consider the monic obtained by applying the functor to . Let be in . Then , for a map in . Since is projective, for some . So , where . This shows that is surjective, hence it is an isomorphism. The proof that is an isomorphism is analogous. ∎
Let and be in , with . We shall regard the canonical embedding in Lemma 4.1,
[TABLE]
as an inclusion of -modules. According to Lemma in [8], lies in for every in . Since the category is closed under submodules then both and are -good modules. Lemma 4.1 is hinting at a close relation between the -filtrations of the modules and . We spell out this idea below.
Corollary 4.2**.**
Let and be in , with . Write and . All the composition factors of of type appear as composition factors of its submodule . In particular, and have the same number of composition factors of type . Moreover, and lie in , , and the modules and are filtered by the same number of standard modules.
Proof.
By Lemma 4.1, all the composition factors of isomorphic to appear as composition factors of its submodule . As lies in then, by Proposition 2.13, is a direct sum of simples of type . Thus . Part 5 of Proposition 2.13 implies that the -filtrations of and have the same number of factors. ∎
As we shall see next, the socle series of an -module gives rise to the -semisimple filtration of in .
Lemma 4.3**.**
Let be in . Then
[TABLE]
Moreover, if , then
[TABLE]
Proof.
By [1, Proposition ] (see also [8, Lemma ]), is generated by projectives satisfying . This proves one of the inclusions in (7). Consider now an arbitrary morphism , with . Note that for a certain map . Clearly, . But then
[TABLE]
As was chosen arbitrarily, the other inclusion follows. This proves identity (7).
To prove the second claim in the statement of the lemma, set
[TABLE]
and assume that is isomorphic to . Recall that is isomorphic to (see Theorem 2.14). Lemma 4.1 and Corollary 4.2 imply that is contained in
[TABLE]
and that these modules have the same socle. By Corollary 3.3, is -semisimple. Finally, by the identity (7) (applied to and ), the module must be generated by projectives of type . This proves the second assertion of the lemma. ∎
Lemma 4.3 and Theorem 4.4 are very useful to compute examples. For the proof of the next result, recall the characterisation of the preradical in Subsection 3.2, namely Proposition 3.8 and Lemma 3.13.
Theorem 4.4**.**
Let be in . The socle series of induces the -semisimple filtration of . Formally,
[TABLE]
for all . In particular, .
Proof.
For satisfying we prove the claim by induction on , starting with . Note that is a direct sum of standard modules of type , so . Since the functor preserves injective hulls (see [8, Lemma ]), the modules and have the same socle. Hence the previous inclusion must be an equality.
Suppose now that , and set
[TABLE]
Since preserves injective hulls, the modules and have the same socle. But then, by Corollary 4.2, and have the same socle. Moreover, belongs to . By Lemma 4.3, must be contained in . So both and are -semisimple modules with the same socle. By Corollary 4.2, must be in . Since is closed under submodules, then is in . We must have , otherwise this factor module would have some composition factor of type . By induction, we may suppose that . Then, the identity translates to
[TABLE]
This implies that , . The same identity holds trivially for and for . ∎
5. Projective covers of modules over the ADR algebra
We would like to determine the projective covers of modules over the ADR algebra of . For a module in , the projective cover of in is the image of an epic , with domain in , through the functor . The morphism is a special kind of map: it is the right minimal -approximation of in .
The problem of finding approximations is hard in general. However, as we shall see in Theorem 5.1, it is very easy to compute right -approximations of rigid modules. Recall that a module is rigid if its radical series coincides with its socle series.
Theorem 5.1 (or rather consequences of this result – Corollary 5.2 and Proposition 5.3) will be very useful when dealing with examples. In Subsection 5.2, we will use Corollary 5.2 and Proposition 5.3 to give a counterexample to a claim by Auslander, Platzeck and Todorov (**[2]**) about the projective resolutions of modules over the ADR algebra.
5.1. Approximations of rigid modules
Let be a class of -modules. We recall the definition of right -approximation and of right minimal morphism. A morphism in , with in , is said to be a right -approximation of if is an epic for all in . A map in is called a right minimal morphism if every endomorphism satisfying is an automorphism.
The right -approximations of a module in are in bijection with epics in ,
[TABLE]
*where . This bijection restricts to a one-to-one correspondence between right minimal -approximations in and projective covers in . Since is a generator, the functor is particularly well behaved: it is fully faithful and it is such that the projective cover of a module in factors through its -approximation. The latter statement implies that every right -approximation is an epimorphism. *
Theorem 5.1**.**
Let be a rigid module in such that . The projective cover of in is a right minimal -approximation of .
Proof.
Let be a rigid module with Loewy length . Consider the projective cover of as an -module,
[TABLE]
We want to prove that is a right minimal -approximation. By definition, is a right minimal morphism, so it is enough to prove that every map , with , factors through . Note that this holds for , as is an epic in and is a projective -module. So suppose that . Then
[TABLE]
using that is rigid. Observe that both and are annihilated by , i.e. they lie in . Now note that the functor preserves epics. This can be seen directly, or can be deduced by looking at Example 2.6 and Remark 2.8, recalling that the composition of cohereditary preradicals is still a cohereditary preradical. Therefore we have the diagram
[TABLE]
where exists because is a projective in . Thus
[TABLE]
where denotes the inclusion of in . ∎
As an immediate consequence of Theorem 5.1, we get the following result.
Corollary 5.2**.**
Let be a rigid module in with . Suppose that is the projective cover of in . Then is the projective cover of in .
The simple modules over the ADR algebra are “linked to each other” in a neat way. When all projective indecomposable modules are rigid then the ‘glueing’ of the simple modules (and of the standard modules) is even nicer.
Proposition 5.3**.**
Let and be in . Then implies that either or . If the -module is rigid then implies that either or . In particular, the latter statement holds when all the projective indecomposable -modules are rigid.
Proof.
Observe that if and only if the simple module is a summand of . The short exact sequence (4) in the statement of Theorem 2.14 gives rise to the exact sequence
[TABLE]
where . If is a summand of the top of then either or is a summand of the top of . In the latter case, we must have by [1, Proposition ] (see also [8, Lemma ]).
If is rigid, then is also rigid. In this case, Corollary 5.2 implies that the summands of the top of are of type . ∎
The next example shows that the rigidity condition in the statement of Theorem 5.1 cannot be omitted.
Example 5.4**.**
Consider the quiver
[TABLE]
and the path algebra . Let be the -module , that is,
[TABLE]
Observe that , and that is not a rigid module.
Consider the epic and note that the simple module can be embedded in . It is not difficult to check that the epic
[TABLE]
is a right minimal -approximation of . This map is not a projective cover of , so the rigidity condition in the statement of Theorem 5.1 is necessary.
Using the approximation (8), one easily sees that the -module can be represented as
[TABLE]
5.2. An application
Motivated in part by the theory of quasihereditary algebras, Auslander, Platzeck and Todorov studied in **[2]** the homological properties of idempotent ideals. In this paper the authors defined a new class of algebras – the Artin algebras satisfying the descending Loewy length condition – and proved, in Theorem , **[2]**, that every such algebra is quasihereditary.
Definition 5.5** ([2, §7]).**
An Artin algebra satisfies the descending Loewy length condition (DLL condition, for short) if for every in , a minimal projective resolution
[TABLE]
satisfies , for all such that .
In **[2]** the authors claim that the Artin algebras of global dimension 2, the ADR algebras , and the -hereditary algebras (introduced in **[20]**) all satisfy the DLL condition. The main purpose of Theorem in **[2]** was thus to give a unified proof of results in **[11]**, **[9]** and **[5]**, already established in the literature.
It is not difficult to check that Artin algebras of global dimension and that -hereditary algebras do satisfy the DLL condition. Unfortunately, it is not true that the ADR algebra satisfies the DLL condition for every choice of . As we shall see, the Loewy length of the projectives in a projective resolution in may increase by an arbitrarily large number.
In order to see this, consider the following example: define , where is a field, is the quiver
[TABLE]
and is the admissible ideal
[TABLE]
with fixed.
We may represent the projective indecomposable -modules as
[TABLE]
Note that the -module is in the socle of . Thus, using the labelling in (2), the -module contains a copy of . The module has socle , so we may consider the corresponding quotient module .
Proposition 5.6**.**
Let be the algebra introduced previously and consider the corresponding ADR algebra . Let be the -module defined above. The DLL condition fails for the -module when . Indeed, we have and , so for .
Proof.
Using that , together with Theorem 2.14, we conclude that . Since is rigid, Corollary 5.2 implies that the minimal projective presentation of is of the form
[TABLE]
We claim that and . Note that this will imply the statement in the proposition, as .
We start by showing that . To see this, note that the -module is in the socle of . Thus, the -module contains a copy of . Note that (see Proposition 2.13), so (actually this is an equality).
Using that , we deduce through a routine computation that has dimension . Consequently, has Jordan–Hölder length and . ∎
We have just shown that the module has Loewy length at most 6. One can actually prove that and, as a consequence, refine the statement of Proposition 5.6 for .
In order to emphasise the usefulness of the results in Subsection 5.1, we compute the exact Loewy length of .
Lemma 5.7**.**
The module has Loewy length equal to .
Proof.
Observe that
[TABLE]
Define . We claim that .
By Lemma 4.3 and Theorem 4.4, the module has a -semisimple filtration
[TABLE]
with factors and . In particular, has socle . Consider now the pullback diagram
[TABLE]
There is a (unique) submodule of with . In fact, by looking at the diagram above we see that , so
[TABLE]
Note that is rigid. Using Proposition 5.3 (and the structure of ), we conclude that must have top . Thus, has the following structure
[TABLE]
Recall that Lemma 2.7 applies to and . The image of the monic through the functor gives rise to the inclusion . By applying to the monics and we deduce that
[TABLE]
Now observe that is a rigid module. Corollary 5.2 implies that . Since has exactly 5 composition factors (and we have looked at them all), we conclude that . This proves the result. ∎
Remark 5.8*.*
In [19], Lin and Xi extended Dlab and Ringel’s result in [9] to endomorphism algebras of semilocal modules. The authors noticed that this class of algebras (which contains the ADR algebra) does not generally satisfy the DLL condition (see Example 3 in [19]).
Remark 5.9*.*
Although the DLL condition does not hold for the ADR algebra in general, satisfies a property similar to the DLL condition. The following was implicitly proved in [1], within the proof of Proposition .
Let be in with , and let be the right minimal -approximation of . Then is the right minimal -approximation of and .
As a consequence, the projective resolutions in come from exact sequences in whose Loewy length decreases strictly. To be precise, for every in there is an exact sequence of -modules
[TABLE]
with in satisfying for all , such that
[TABLE]
is a minimal projective resolution for (see § in [7] for details).
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[1] M. Auslander, Representation theory of Artin algebras. I, II , Comm. Algebra 1 (1974), 177–268; ibid. 1 (1974), 269–310.
- 2[2] M. Auslander, M. I. Platzeck, and G. Todorov, Homological theory of idempotent ideals , Trans. Amer. Math. Soc. 332 (1992), no. 2, 667–692.
- 3[3] L. Bican, T. Kepka, and P. Němec, Rings, modules, and preradicals , Lecture Notes in Pure and Applied Mathematics, vol. 75, Marcel Dekker Inc., New York, 1982.
- 4[4] A. B. Buan, O. Iyama, I. Reiten, and J. Scott, Cluster structures for 2-Calabi-Yau categories and unipotent groups , Compos. Math. 145 (2009), no. 4, 1035–1079.
- 5[5] W. D. Burgess and K. R. Fuller, On quasihereditary rings , Proc. Amer. Math. Soc. 106 (1989), no. 2, 321–328.
- 6[6] J. Clark, C. Lomp, N. Vanaja, and R. Wisbauer, Lifting modules , Frontiers in Mathematics, Birkhäuser Verlag, Basel, 2006, Supplements and projectivity in module theory.
- 7[7] T. Conde, On certain strongly quasihereditary algebras , Ph.D. thesis, University of Oxford, 2016.
- 8[8] by same author, The quasihereditary structure of the Auslander–Dlab–Ringel algebra , J. Algebra 460 (2016), 181–202.
