Summands of Syzygies over Rings of Positive Burch Index Via Canonical Resolutions
Abstract.
In recent work, Dao and Eisenbud define the notion of a Burch index, expanding the notion of Burch rings of Dao, Kobayashi, and Takahashi, and show that for any module over a ring of Burch index at least 2, its th syzygy contains direct summands of the residue field for or and all . We investigate how this behavior is explained by the bar resolution formed from appropriate differential graded (dg) resolutions, yielding a new proof that includes all , which is sharp. When the module is Golod, we use instead the bar resolution formed from resolutions to identify such summands explicitly for all and show that the number of these grows exponentially as the homological degree increases.
Key words and phrases:
Burch2020 Mathematics Subject Classification:
13D02, 13A35 (primary); 13D07, 16E45 (secondary)Introduction
The results in this paper focus on finding copies of the residue field of a local or graded-local ring appearing as direct summands of syzygies of any finitely generated module over certain rings. This continues the work of Dao and Eisenbud in [6], in which they find that this phenomenon occurs surprisingly frequently. In the first part of the paper, we give a different proof of their result which explains the homological persistance of summands of syzygies as a manifestation of the self-similarity of the bar complex. The key is identifying certain low degree non-Koszul cycles in all resolutions of the ring over an ambient regular ring. Applied to resolutions equipped with a differential graded algebra and differential graded module structure, we obtain cycles in the bar resolution of the module over the ring of interest. This shows that simple summands of syzygies are caused by the beginning of the resolution of the ring, which explains the ubiquity of this phenomenon.
These cycles spring from positivity of a numerical invariant of a local ring called the Burch index, denoted by . This was introduced by Dao and Eisenbud to refine the definition of a Burch ring, first defined by Dao, Kobayashi, and Takahashi in [7] as a ring with positive Burch index. When is a Burch ring, resolutions of over a regular ring have strong linearity properties [6, 2.4] which dominate the resolutions of all finitely generated -modules [6, 3.1, 3.7]. Rings with larger Burch index are plentiful; in particular, [6, Example 4.7, Proposition 1.5] provide large infinite classes of such rings.
Larger Burch index provides stronger results: When is a depth zero ring with Burch index at least , Dao and Eisenbud prove that has a summand for some and all and give an example to show that it need not occur for [6, 3.1, 3.6]. We give a new proof which, as a practical consequence, shows that summands are found in every degree five and beyond, giving a more uniform variant of [6, Thm 3.1] as follows. This result is a corollary of the result established in Theorem 3.7 via 1.6.
Theorem A.
Let be a local (or graded-local) ring of depth with , and let be a finitely generated (resp., finitely generated graded) -module. Then has a summand for each .
In the second part of the paper, we refine our technique by using structures on minimal resolutions of the ring and module over an ambient ring. These structures are sufficient to construct the bar resolution, which is minimal when is Golod [5]. Under this additional assumption, we build cycles in the bar resolution which precisely correspond to syzygy summands. This establishes an explicit exponential lower bound in the number of such summands in homological degree at least . In analogy with free rank, we let be the number of independent summands of a module . This result is a corollary of the result established in Theorem 4.2 via 1.6.
Theorem B.
Let be a local (or graded-local) ring with , and let be a finitely generated (resp., finitely generated graded) Golod -module. Suppose that or its completion are presented minimally as , where is regular local (or graded-local) ring and . Then for each
where denotes the minimal number of generators of .
Note that, even when is not Golod, if , it already follows from [6] that a subsequence of grows exponentially; see Remark 4.4. Here, however, noting that when , we see exponentially-many explicit cycles in all degrees providing these summands.
In their paper, Dao and Eisenbud give an example to show that Burch index 1 does not suffice for the existence of summands in syzygies. In 2.4, we explore this example to illustrate where in our methods a Burch index of at least 2 is crucial.
The paper begins with a background section, Section 1, with definitions of and the relative version for a ring map . We also remind the reader of and differential graded algebra structures and their use in constructing the bar resolution. In the preliminaries section, Section 2, we define a certain set of cycles in the resolution of over a regular ring which we call Burch cycles. We then use these cycles and the bar resolution to produce summands in syzygies when the Burch index is at least 2, first in a general setting in Section 3 and then, with more precision, in the relative Golod setting in Section 4.
One last remark: While the theorems above were stated in terms of a fixed ring relative to an ambient ring given by a minimal Cohen presentation, we prove more general results about ring homomorphisms. They are related to the statements in the introduction in 1.6.
1. Background
1.1.
Let and be local commutative Noetherian rings with a fixed surjective homomorphism and common residue field and kernel . An important application is when is the minimal Cohen presentation of a complete or positively graded ring. Modules over or are always assumed to be finitely generated.
1.2.
The results in this work apply equally well to graded-local rings with a graded-local homomorphism , by which we mean that and are -graded (or multigraded) rings whose degree pieces are a common field . In this setting, we further require that modules be graded also.
Definition 1.3.
Here we recall from [6] the definitions of the Burch ideal and the Burch index of . The Burch ideal of the map with kernel is defined to be
Note that when , , and the preimage of in is the ideal . By abuse of language, we may call elements of this ideal socle elements in descriptive text.
Define the Burch index of to be
When or , , and so and .
For a more qualitative description, note the following: The ideal records those elements of which are not minimal generators, so the iterated colon ideal describe lifts of elements of which multiply the lift of the socle outside of the set of minimal generators of . Evidently, is contained in this set: by the definition of the colon ideal, and then multiplying on the left by gives , whence . Therefore, there is a surjection
The latter is the -space of generators of which multiply some minimal generator of the lift of the socle to a minimal generator of . Therefore, for each , there are corresponding socle elements whose images in are nonzero and which satisfy .
This yields an equivalent description of the Burch index which we will use later: Setting , there is a minimal generating set of with the property that there exist whose images are independent in the -vector space (and hence also in ) such that for one has
Note that we may have for some .
Definition 1.4.
Let be a finitely generated module over a local ring . In analogy with the more traditionally studied free rank, we set the -rank of to be the maximal number of direct summands of the residue field:
where we use the notation to denote that is a direct summand of . We write when is clear from context. The use of maximum rather than supremum is justified by the fact that is finitely generated: any such map remains injective upon tensoring with . We say that an element generates a direct -summand if .
Proposition 1.5.
If is finitely-generated, then is invariant under completion.
Proof.
If , then
and this isomorphism respects the natural inclusion . The splitting then factors the identity of as . ∎
1.6.
In [6] the absolute Burch index, , of the ring is also defined by taking to be a minimal Cohen presentation, but we will not make use of this notion in this work. However, to translate our results to the special cases presented as Theorems A and B in the introduction, we describe it briefly: By the Cohen Structure Theorem, one may write the completion of as the quotient of a regular local (or graded-local) ring by an ideal in the square of the (homogeneous) maximal ideal. Let be the natural surjection. Then one has
Theorems 3.7 and 4.2 discussed below then extend to the absolute case by Proposition 1.5.
1.7.
Bar resolutions are an essential tool used in this work. Their construction involves building projective resolutions over from those over . The resolutions over must be equipped with structures. The simplest and most classical of these is that of a differential graded (dg) algebra, first introduced into commutative algebra by Tate [19]. We recall their definition below:
Definition 1.8.
A dg algebra over is complex of -modules further equipped with a chain map (the -bilinear product); writing the product as concatenation in the ordinary way, that this product is a chain map is encoded by the Leibniz rule:
where denotes the degree of ; note that by our convention, is a complex and so elements are automatically homogeneous. Denoting the differential by and the product map by , this can be restated as
We require that is a -algebra in the ordinary sense once the differential has been forgotten: in particular, must be associative and is equipped with a unit map .
A dg module over a dg algebra is a complex of -modules further equipped with a chain map which is compatible with differentials of and , as encoded by an analogous Leibniz rule.
A dg algebra resolution of over is a dg algebra with , each a projective -module, and . For an -module , a dg module resolution of over is a dg module with , each projective over , and .
1.9.
While the minimal -free resolutions of and need not support dg structures, there always exists some resolution of with a dg algebra structure [19] and some resolution of with a dg -module. When is surjective and is a finite -module, then such resolutions exist with
Definition 1.10.
Let be any resolution of over , and be any -free resolution of an -module . Any lifts of the multiplication maps and provide candidate multiplication maps and for dg algebra and module structures of and , respectively. Such lifts may fail to be associative. For some rings and modules, when and are minimal resolutions over , no such lifts are associative [3]. To rectify this, we make use of an structure, the needed details of which we recall below; see [15, 11, 12] for more of the general theory and [17, 18] for their history and origins.
An -algebra over a ring is a complex of -modules together with -multilinear maps of degree
called operations or multiplications, satisfying the Stasheff identities for each :
where the elements are automatically homogeneous as the underlying structure of is that of a complex. The associator of the multiplication sends to (where we write ). The third Stasheff identity is
which shows that the associator is the boundary of in . In other words, the associator is nullhomotopic with nullhomotopy .
Note that when one applies the maps in each formula above to an element, one should use the Koszul sign rule: For graded maps and , one has
where denotes the degree of .
We also require algebras to be strictly unital, in that there is a unit map such that for , . Intuitively, this just states that multiplication by the unit is associative [16, 7.2].
An -module over an -algebra is a complex of -modules together with -multilinear maps of degree
such that analogous Stasheff identities hold.
In fact, are more affordable than dg algebra structures, as manifested by the following:
Proposition 1.11 ([5], Proposition 3.6).
Let be any resolution of over and be any -resolution of an -module . Then there exists an algebra structure on and an -module structure on .
1.12.
Our essential technique to find summands of in syzygies is to use the relative bar resolution, first developed by Iyengar for dg structures [10]. In this case, the relative bar resolution arises as the totalization of the classical bar complex [8]. We use the relative bar resolution in this form to obtain Theorem 3.7. For Theorem 4.2, it is essential that the input data be minimal resolutions, and so we must make use of a generalization of the relative bar resolution to structures. This generalization was developed by Burke [5] and Positselski [16]. We recall the constructions here.
Let be a -free resolution with an algebra structure, and let be a -free resolution with an -module structure. Let be the unit map, and set , where of a complex is the complex with and differential . In this paper, since is surjective, we may identify with .
For each , define
As is traditional for the bar complex, we denote the element of by
which is said to be of bar length . The differential is the signed sum of all operations applied to every substring of appropriate bar length:
where, for visual simplicity, we suppress the exact signs, which are irrelevant to our application. The augmentation is given by
In the dg setting, and so the differential simplifies as follows. Here we do indicate the signs as we will need them. Writing concatenation in place of and in place of and , the differential is:
where refers to the degree of in the shifted complex . Note that the formula for does not have a term involving multiplication by the factor of ; heuristically, this is because the product vanishes as .
2. Preliminaries: Burch cycles
In this section, given a local (or graded-local) homomorphism , we find some distinguished cycles in any -free resolution of when the Burch index of is positive. For ease, we work in a minimal resolution but the resulting cycles therefore exist in any resolution.
2.1.
Let be the minimal resolution of over , and let be a basis of , a minimal generating set of , and . When , one or more of the generators are multiples of socle elements as described in 1.3. With for some elements and element , take any whose image in is not in the span of the image of . Then, since is in , one may write
for some elements and so the element
is a cycle in , hence a boundary, so fix with
Note that the conditions on ensure that each is a minimal generator of the set of degree one cycles : Since identified a minimal generating set, one has , so . As is independent from , the image is non-zero in , and hence is nonzero in . Therefore, the chosen preimage is part of a basis of .
The above is extended to all of by the following lemma. Note that the obvious choices of in the construction above yield and ; hence we only include those for :
Lemma 2.2.
Let be a basis of , and extend it to a basis of . The image of the set
in is linearly independent, and hence in any completion to a resolution , there are basis elements map** to each corresponding .
Furthermore, for each , we have that .
Proof.
Order the basis of lexicographically. The set is obtained from a subset of the basis by subtracting a linear combination of basis elements that are strictly smaller in the ordering, and so they are independent.
Note that in the sum, when the image of the coefficient of in is non-zero since and are linearly independent in and . ∎
Once we build cycles in the bar complex from these distinguished cycles in that generate copies of in the syzygies of a module, we show that they split out of the syzygies with the following modified version of [6, Proposition 3.4].
Lemma 2.3 ([6]).
Let be a surjective local homomorphism. Let be a map of free -modules and be nonzero in . Suppose that but , equivalently, there exists an with
Then the element generates a direct summand of isomorphic to .
Note that the conditions of the lemma are equivalent to for any chosen so that , which is what we will check when we apply the lemma.
2.4.
As an illustration for what goes wrong with our arguments for Theorems 3.7 and 4.2 if a map has only Burch index at most 1, here we review the example of Dao and Eisenbud of a ring of absolute Burch index 1 and a module with no syzygies containing a summand isomorphic to [6, 3.4]. We show that the cycles we built do generate modules isomorphic to , but that we cannot conclude they are direct summands.
With and , we have for that , so . Then and are the only ways to write the minimal generators as elements of .
In the minimal resolution of , as in 2.1, we have and , and the corresponding Burch cycle is . Recall we denote a preimage of this cycle by . To employ Dao-Eisenbud’s lemma to elements in the bar resolution involving this cycle, one would need that for some socle element . As the coefficients of are contained in , we have that for all .
This explains why our set of distinguished Burch cycles always involves choosing two elements , whose images are independent in and hence we require Burch index at least 2.
3. General case
Let be a local (or graded-local) homomorphism of rings, and let be a finitely generated -module. Set to be the Burch index of . In this section, we tackle the general case of any finitely generated -module when . We work first to find summands of cycles in a bar resolution constructed from nonminimal resolutions with dg structures, and show that this induces a summand of each sufficiently high syzygy of . We begin by describing the details of the resolutions we will use.
3.1.
First, as in 1.7, take to be a dg algebra resolution of over such that there is a basis of with . As described in 2.1, there is an associated distinguished set of elements which the differential maps to the cycles constructed there and which are -linearly independent by Lemma 2.2.
Next, as in 1.7, we take to be a semifree dg -module resolution of but constructed with care so that there is dg module map that is degree-wise a split injection. As described in [2], the first step is to take a -linear surjection and lift it to a dg -module map where ; this is trivially possible as . Continuing to kill cycles, one constructs a semifree dg -module resolution of with the property that there is an injection . The composition
where is any injection, is an injective dg module map. The choices of images at each step are basis elements, so splits degree-wise.
This is elucidated in the following example.
Example 3.2.
Let , , and . Then can be taken to be the Koszul complex , and , so can be defined by setting to be the identity.
We begin by finding -summands in the (possibly non-minimal) bar resolution.
Lemma 3.3.
Let be a local or graded local homomorphism, and let be a finitely generated -module. Suppose .
Let be a dg algebra resolution of over as in 1.9, and let be a semifree dg module resolution of over with a dg -module map as constructed in 3.1. Let be their associated bar resolution of over , and let . For , set
when is even and
when is odd. Then span a nonzero -vector space that is a direct summand of .
Proof.
First we verify that the elements are indeed cycles in . Applying the portion of the differential of which comes from the differentials of and yields 0 since , and the resolutions are minimal on the elements involved in the expression, that is, one has and . Note that the tensors are over , and so the element can be moved across them.
The remaining terms come from the bar portion, denoted by , of the differential. Many terms vanish from the fact that and , and the terms of involve products of adjacent factors. When is odd, this causes all terms of to vanish. When is even, the only terms remaining are
where the first term comes from the multiplication of the first two bar (interior) factors of the first term of , the second term comes from the multiplication of the last two tensor factors of the second term of , and the sign comes from the fact that and each shifted has even degree.
The elements are part of a basis of . Lifting the differential to and to a basis element , the assumptions on ensure that . Hence each of these cycles generate a -summand of the -th syzygy by Lemma 2.3, which suffices to yield the result. ∎
Remark 3.4.
As we are not presuming is -Golod, the bar resolution cannot be expected to be minimal, even when the starting resolutions are minimal. In light of this, we elect to use simpler dg structures, rather than the structure used in Theorem 4.2 1.12. Using dg structures simplified the analysis of the cycles in the proof of Lemma 3.3.
To prepare for projecting to a minimal resolution sitting inside the bar resolution in the lemma above, we prove the following.
Lemma 3.5.
Let be a local or graded. If is not a field, then cannot be a direct summand of a free -module .
Proof.
If is a summand of , then it is projective and hence free since is local or graded [4, 1.5.15(d)]. But is not free unless is a field. ∎
Lemma 3.6.
Let be a local or graded ring, Let be a free resolution of , and let be a decomposition of with a minimal resolution of and contractible. Suppose generates a direct -summand of , with and . Then generates a direct -summand of .
Proof.
Let be the splitting sending to . Then the map sending to is well-defined, since is killed by . The post-composition with must be zero, else would generate a direct -summand of the free module , contradicting Lemma 3.5. Then since is nonzero, must be a unit . Hence the map sending to is split by ∎
Applying the above lemmas to the (possibly non-minimal) bar resolution from Lemma 3.3, we immediately obtain the main result of this section:
Theorem 3.7.
Let be a local or graded local homomorphism, and let be a finitely generated -module. Suppose .
Then for
Lemma 3.3 identifies a much larger -vector space summand of a resolution of over , but it is not clear how many of these summands become identified upon projecting to the minimal resolution. Computations suggest the lower bound in Theorem 3.7 is far from sharp, and it is possible that sharper bounds may be identified in some cases. We provide such a sharper bound in the Golod case, as described in the next section.
4. Golod case
For Golod modules , there is a bar resolution that minimally resolves . In this section, we use the minimality of the resulting resolution to establish B, which provides an exponential lower bound on the number of -summands of syzygies of , rather than the constant bound provided by A. We also establish this lower bound in one lower homological degree than that of A. In this section will be a local (or graded-local) homomorphism of Burch index at least two, and will be a -Golod module, defined in the sequel.
4.1.
The Golod property is characterized by minimality of a bar resolution of . Indeed, even when is not Golod, can still be constructed using structures on the minimal resolutions of and of [5, 10], as described in 1.12. The resulting resolution of provides an upper bound for the growth of the Betti numbers of . In particular, examining the relationship between the generating function of the ranks of with respect to those of and reveals the coefficient-wise inequality of Poincaré series
and the classical definition of the Golod property is that equality holds in the above [1], demonstrating that the Betti numbers of Golod modules attain maximal growth. In particular, one sees that for any algebra structure on the minimal -free resolution of and any -module structure on the minimal -free resolution of , the products and are all minimal [5, 6.3], that is,
An important case is when is -Golod with respect to the minimal Cohen presentation of , in which case is said to be Golod. Similarly, if is -Golod with respect to the minimal Cohen presentation of , then is said to be Golod. The Golod property then has profound implications for the homological and representation theoretic properties of ; in particular, if is Golod and not a hypersurface, than the Betti sequence of any module of infinite projective dimension must grow exponentially.
Despite the apparent specificity of Golod properties, all rings and modules are “asymptotically Golod” in the following senses: if is an ideal in a regular local ring , then is Golod for [9]. More generally, when is only assumed to be local, the map is Golod for [14]. If is a module over a Golod ring, then its syzygies are Golod for [13]. One way to obtain Golod rings of high Burch index is to take a quotient of a regular or graded ring by an ideal of the form , where is a an ideal of . For such rings, the Burch index is [6, 1.5].
The following establishes B.
Theorem 4.2.
Let be a surjective local (or graded-local) homomorphism with , and let be a finitely generated (resp., finitely generated graded) -module such that is -Golod. Then for ,
where is the minimal number of generators of .
Proof.
By Proposition 1.11, the minimal -free resolution of admits the structure of an algebra and the minimal -free resolution of admits the structure of an -module. When is -Golod, the resulting bar resolution as described in 1.12 is minimal as discussed in 4.1.
Set . As discussed in Definition 1.3, one may choose elements that minimally generate ordered so that the images of are linearly independent in . As determined in 2.1 and Lemma 2.2, these yield independent basis elements in for with the property that
Further, noting that , take any basis for for . Write where and , and choose a basis element .
Then we claim that the following elements of in degree
are independent cycles generating a -vector space summand of . Indeed they are cycles as the assumptions ensure that the differential of is minimal over and hence its image is killed by . By Lemma 2.2, the set is independent in and hence the elements for generate a -vector space in . Lastly, that these cycles are a summand of the -th syzygy follows from Lemma 2.3: The elements
are part of a basis of . Lifting the differential to and to a basis element , the assumptions on ensure that . Hence each of these cycles generate a -summand of the -th syzygy by Lemma 2.3, which suffices to yield the result. ∎
Corollary 4.3.
Let be a Golod ring and be an -module. Then the exponential growth of begins by .
Proof.
By the result mentioned in 4.1, the -module is Golod if . On the other hand, by Theorem 3.7, has a Golod direct summand (namely, ) for . In either case, Theorem 4.2 insures that the Golod module will contribute exponential growth of the by at most 4 steps later. ∎
Remark 4.4.
Lastly we note that, even when is not Golod, it already follows from [6] that a subsequence of grows exponentially whenever . Indeed, note that , and so the resolution of contains a direct summand of the resolution of , Therefore, it suffices to show this fact for . We have that , and we claim that : Since , is not the hypersurface , so by examining the acyclic closure [19], we get . Then we get in turn that , and so
Continuing inductively, we see .
References
- [1] Luchezar L. Avramov “Infinite free resolutions” In Six lectures on commutative algebra, Mod. Birkhäuser Class. Birkhäuser Verlag, Basel, 2010, pp. 1–118 DOI: 10.1007/978-3-0346-0329-4˙1
- [2] Luchezar L. Avramov “Locally complete intersection homomorphisms and a conjecture of Quillen on the vanishing of cotangent homology” In Ann. of Math. (2) 150.2, 1999, pp. 455–487 DOI: 10.2307/121087
- [3] Luchezar L. Avramov “Obstructions to the existence of multiplicative structures on minimal free resolutions” In Amer. J. Math. 103.1, 1981, pp. 1–31 DOI: 10.2307/2374187
- [4] Winfried Bruns and Jürgen Herzog “Cohen-Macaulay rings” 39, Cambridge Stud. Adv. Math. Cambridge University Press, Cambridge, 1993, pp. xii+403
- [5] Jesse Burke “Higher homotopies and Golod rings”, 2015 arXiv:1508.03782 [math.AC]
- [6] Hailong Dao and David Eisenbud “Burch index, summands of syzygies and linearity in resolutions” In Bull. Iranian Math. Soc. 49.2, 2023, pp. Paper No. 10\bibrangessep10 DOI: 10.1007/s41980-023-00755-x
- [7] Hailong Dao, Toshinori Kobayashi and Ryo Takahashi “Burch ideals and Burch rings” In Algebra Number Theory 14.8, 2020, pp. 2121–2150 DOI: 10.2140/ant.2020.14.2121
- [8] Samuel Eilenberg and Saunders MacLane “Homology theories for multiplicative systems” In Trans. Amer. Math. Soc. 71, 1951, pp. 294–330 DOI: 10.2307/1990693
- [9] Jürgen Herzog, Volkmar Welker and Siamak Yassemi “Homology of powers of ideals: Artin-Rees numbers of syzygies and the Golod property” In Algebra Colloq. 23.4, 2016, pp. 689–700 DOI: 10.1142/S1005386716000584
- [10] Srikanth Iyengar “Free resolutions and change of rings” In J. Algebra 190.1, 1997, pp. 195–213 DOI: 10.1006/jabr.1996.6901
- [11] Bernhard Keller “-infinity algebras, modules and functor categories” In Trends in representation theory of algebras and related topics 406, Contemp. Math. Amer. Math. Soc., Providence, RI, 2006, pp. 67–93 DOI: 10.1090/conm/406/07654
- [12] Kenji Lefèvre-Hasegawa “Sur les A-infini catégories”, 2002
- [13] Jack Lescot “Séries de Poincaré et modules inertes” In J. Algebra 132.1, 1990, pp. 22–49 DOI: 10.1016/0021-8693(90)90250-R
- [14] Gerson Levin “Local rings and Golod homomorphisms” In J. Algebra 37.2, 1975, pp. 266–289 DOI: 10.1016/0021-8693(75)90077-0
- [15] Jean-Louis Loday and Bruno Vallette “Algebraic operads” 346, Grundlehren der mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences] Springer, Heidelberg, 2012, pp. xxiv+634 DOI: 10.1007/978-3-642-30362-3
- [16] Leonid Positselski “Two kinds of derived categories, Koszul duality, and comodule-contramodule correspondence” In Mem. Amer. Math. Soc. 212.996, 2011, pp. vi+133 DOI: 10.1090/S0065-9266-2010-00631-8
- [17] James Stasheff “On homotopy Abelian H-spaces” In Mathematical Proceedings of the Cambridge Philosophical Society 57.4 Cambridge University Press, 1961, pp. 734–745 DOI: 10.1017/S0305004100035878
- [18] Jim Stasheff “-infinity and -infinity structures” In High. Struct. 3.1, 2019, pp. 292–326
- [19] John Tate “Homology of Noetherian rings and local rings” In Illinois J. Math. 1, 1957, pp. 14–27 URL: http://projecteuclid.org/euclid.ijm/1255378502