Noncommutative supports, local cohomology and spectral sequences

  • Abhishek Banerjee

    Indian Institute of Science, Bangalore, India
  • Surjeet Kour

    Indian Institute of Technology, Delhi, India
Noncommutative supports, local cohomology and spectral sequences cover
Download PDF

This article is published open access under our Subscribe to Open model.

Abstract

The purpose of this paper is to study local cohomology in the noncommutative algebraic geometry framework of Artin and Zhang. The noncommutative spaces are obtained by base change of a Grothendieck category that is locally noetherian or strongly locally noetherian. Using what we call elementary objects and their injective hulls, we develop a theory of supports and associated primes in these categories. We apply our theory to study a general functorial setup that requires certain conditions on the injective hulls of elementary objects and gives us spectral sequences for derived functors associated with local cohomology objects, as well as generalized local cohomology and also generalized Nagata ideal transforms.

1. Introduction

If k is a field and A is a commutative k-algebra, the category ModA of A-modules determines the geometry of the affine scheme Spec(A). If A is no longer commutative, the category ModA of right A-modules plays the role of a noncommutative affine scheme. Suppose that A is a commutative graded k-algebra and Gr(A) is its category of graded modules. Then, the geometry of the projective scheme Proj(A) can be understood by means of the category QGr(A) which is the quotient of Gr(A) over torsion modules, that is, modules which are obtained as filtered colimits of bounded graded modules (see [5
M. Artin and J. J. Zhang, Noncommutative projective schemes. Adv. Math. 109 (1994), no. 2, 228287 Zbl 0833.14002 MR 1304753
]). Accordingly, the category QGr(A) plays the role of a noncommutative projective scheme in the theory of Artin and Zhang [5
M. Artin and J. J. Zhang, Noncommutative projective schemes. Adv. Math. 109 (1994), no. 2, 228287 Zbl 0833.14002 MR 1304753
], where A is a noncommutative graded k-algebra. In general, one can approach noncommutative algebraic geometry by adapting notions from commutative rings and the usual theory of schemes to arbitrary Grothendieck categories.
One of the notions that is fundamental in algebraic geometry is that of base change. Let k be a field, and let A, R be k-algebras. Then, a module over the tensor product algebra (AkR) may be described as an ordinary A-module equipped with an additional R-module action. If we replace the category of A-modules by a general k-linear abelian category 𝒮, we obtain the category 𝒮R of “R-module objects in 𝒮.” These module objects were introduced by Popescu [21
N. Popescu, Abelian categories with applications to rings and modules. London Math. Soc. Monogr. 3, Academic Press, London-New York, 1973, 467 pp. Zbl 0271.18006 MR 0340375
, p. 108] and may be treated as modules over a noncommutative base change of R by the abelian category 𝒮. An R-module object in 𝒮 consists of an object 𝒮 along with a morphism of k-algebras
ρ:R𝒮(,).
In [6
M. Artin and J. J. Zhang, Abstract Hilbert schemes. Algebr. Represent. Theory 4 (2001), no. 4, 305394 Zbl 1030.14003 MR 1863391
], Artin and Zhang systematically developed the properties of the abstract module category 𝒮R, where 𝒮 is a Grothendieck category, and often a locally noetherian or strongly locally noetherian Grothendieck category. The theory developed by Artin and Zhang in [6
M. Artin and J. J. Zhang, Abstract Hilbert schemes. Algebr. Represent. Theory 4 (2001), no. 4, 305394 Zbl 1030.14003 MR 1863391
] includes tensor products, internal hom, localization, versions of the Hilbert basis theorem, the Nakayama lemma, as well as the derived functors Tor and Ext. Several other fundamental constructions on 𝒮R, such as ind-objects, coherence and deformation theory, have been developed by Lowen and Van den Bergh [19
W. Lowen and M. Van den Bergh, Deformation theory of abelian categories. Trans. Amer. Math. Soc. 358 (2006), no. 12, 54415483 Zbl 1113.13009 MR 2238922
]. In [7
A. Banerjee, An extension of the Beauville–Laszlo descent theorem. Arch. Math. (Basel) 120 (2023), no. 6, 595604 Zbl 1527.13010 MR 4598542
], we have also studied descent properties in 𝒮R.
We note that a number of properties of the category 𝒮 extend to the module category 𝒮R, following the idea of noncommutative base change. For example, if 𝒮 is locally finitely generated, so is 𝒮R. The version of the Hilbert basis theorem proved by Artin and Zhang in [6
M. Artin and J. J. Zhang, Abstract Hilbert schemes. Algebr. Represent. Theory 4 (2001), no. 4, 305394 Zbl 1030.14003 MR 1863391
] shows that if 𝒮 is locally noetherian and R is a finitely generated k-algebra, then 𝒮R is also locally noetherian. Since Grothendieck categories abound in nature, the theory of the abstract module category 𝒮R is very general and applies to a wide variety of situations. For instance, 𝒮 could be the category of sheaves of abelian groups on a topological space, comodules over a coalgebra over a field, modules over a ringed space, quasi-coherent sheaves over a scheme or the category of comodules over a flat Hopf algebroid.
In this paper, we develop a local cohomology theory in the noncommutative algebraic geometry of Artin and Zhang [5
M. Artin and J. J. Zhang, Noncommutative projective schemes. Adv. Math. 109 (1994), no. 2, 228287 Zbl 0833.14002 MR 1304753
, 6
M. Artin and J. J. Zhang, Abstract Hilbert schemes. Algebr. Represent. Theory 4 (2001), no. 4, 305394 Zbl 1030.14003 MR 1863391
], by making use of abstract module categories. We let the k-linear Grothendieck category 𝒮=𝒮k be strongly locally noetherian, whence it follows that 𝒮R is locally noetherian for any noetherian commutative k-algebra R. We should note that our framework will give new results, even in the case we take 𝒮 to be the category of modules over a noncommutative algebra A that is strongly locally noetherian, that is, AkR is right noetherian for every commutative noetherian k-algebra R (see [4
M. Artin, L. W. Small, and J. J. Zhang, Generic flatness for strongly Noetherian algebras. J. Algebra 221 (1999), no. 2, 579610 Zbl 0958.16024 MR 1728399
]). If C is a coalgebra over a field, its category of comodules is locally noetherian (see [14
S. Dăscălescu, C. Năstăsescu, and Ş. Raianu, Hopf algebras: an introduction. Monographs and Textbooks in Pure Appl. Math. 235, Marcel Dekker, New York, 2001, 401 pp. Zbl 0962.16026 MR 1786197
, Section 2.4.8] and [23
B. Stenström, Rings of quotients: an introduction to methods of ring theory. Grundlehren Math. Wiss. 217, Springer, New York-Heidelberg, 1975, 309 pp. Zbl 0296.16001 MR 0389953
, Section V.4.3]). As with the theory in [6
M. Artin and J. J. Zhang, Abstract Hilbert schemes. Algebr. Represent. Theory 4 (2001), no. 4, 305394 Zbl 1030.14003 MR 1863391
], we can take 𝒮 to be the category Gr(A) of graded modules over a graded algebra A. The category Gr(A) is strongly locally noetherian whenever A is strongly noetherian in a graded sense (see [6
M. Artin and J. J. Zhang, Abstract Hilbert schemes. Algebr. Represent. Theory 4 (2001), no. 4, 305394 Zbl 1030.14003 MR 1863391
, Section B5]). We can also take 𝒮=QGr(A) for a noncommutative graded k-algebra A. These are the noncommutative projective schemes of Artin and Zhang [5
M. Artin and J. J. Zhang, Noncommutative projective schemes. Adv. Math. 109 (1994), no. 2, 228287 Zbl 0833.14002 MR 1304753
]. If A is strongly noetherian in a graded sense (but not necessarily commutative), the category QGr(A) is strongly locally noetherian and QGr(A)R=QGr(AkR) (see [6
M. Artin and J. J. Zhang, Abstract Hilbert schemes. Algebr. Represent. Theory 4 (2001), no. 4, 305394 Zbl 1030.14003 MR 1863391
, Section B8]).
The aims of our paper are threefold. First, we build a theory of associated primes and supports in the abstract module category 𝒮R using what we call R-elementary objects in 𝒮R. If 𝒮 is the category of modules over a commutative k-algebra A, then an R-elementary object in 𝒮R behaves somewhat like a quotient over a prime ideal. Accordingly, the injective hulls of these R-elementary objects in 𝒮R play a key role in our theory.
Our second purpose is to develop local cohomology in 𝒮R. In fact, we work with a more general framework, extending the local cohomology with respect to a pair of ideals I, JR introduced by Takahashi, Yoshino and Yoshizawa [24
R. Takahashi, Y. Yoshino, and T. Yoshizawa, Local cohomology based on a nonclosed support defined by a pair of ideals. J. Pure Appl. Algebra 213 (2009), no. 4, 582600 Zbl 1160.13013 MR 2483839
]. One of the key aspects of the theory in [24
R. Takahashi, Y. Yoshino, and T. Yoshizawa, Local cohomology based on a nonclosed support defined by a pair of ideals. J. Pure Appl. Algebra 213 (2009), no. 4, 582600 Zbl 1160.13013 MR 2483839
] is that local cohomology is based on a nonclosed support 𝕎(I,J). In order to study the two-variable local cohomology objects HI,J(__) in 𝒮R, we will need the support theory, associated primes and R-elementary objects in 𝒮R. If 𝒮R is a finitely generated object, we show that the set Ass() is finite. For an ideal IR, the question of the finiteness of the set of associated primes of local cohomology objects is very interesting in commutative algebra and is also related to Faltings’ local-global-principle for finiteness dimensions (see [16
G. Faltings, Der Endlichkeitssatz in der lokalen Kohomologie. Math. Ann. 255 (1981), no. 1, 4556 Zbl 0451.13008 MR 0611272
], and also [13
M. P. Brodmann and R. Y. Sharp, Local cohomology: an algebraic introduction with geometric applications, 2nd edn., Cambridge Stud. Adv. Math. 136, Cambridge University Press, Cambridge, 2013, 491 pp. Zbl 1263.13014 MR 3014449
, Section 9.6.2]). In fact, a number of questions related to the supports and associated primes of local cohomology have been studied extensively in the literature (see, for instance, [9
M. Brodmann, Asymptotic stability of Ass(M/InM). Proc. Amer. Math. Soc. 74 (1979), no. 1, 1618 Zbl 0395.13008 MR 0521865
-12
M. P. Brodmann and A. Lashgari Faghani, A finiteness result for associated primes of local cohomology modules. Proc. Amer. Math. Soc. 128 (2000), no. 10, 28512853 Zbl 0955.13007 MR 1664309
, 17
M. Hellus, On the set of associated primes of a local cohomology module. J. Algebra 237 (2001), no. 1, 406419 Zbl 1027.13009 MR 1813886
, 18
C. Huneke, D. Katz, and T. Marley, On the support of local cohomology. J. Algebra 322 (2009), no. 9, 31943211 Zbl 1186.13016 MR 2567416
, 22
C. Rotthaus and L. M. Şega, Some properties of graded local cohomology modules. J. Algebra 283 (2005), no. 1, 232247 Zbl 1105.13021 MR 2102081
]). We give a condition for Ass(HI,Ji()/𝒩) to be finite, where 𝒮R is finitely generated and 𝒩HI,Ji() is a finitely generated subobject. This generalizes a result of Brodmann and Faghani [12
M. P. Brodmann and A. Lashgari Faghani, A finiteness result for associated primes of local cohomology modules. Proc. Amer. Math. Soc. 128 (2000), no. 10, 28512853 Zbl 0955.13007 MR 1664309
] to the two-variable local cohomology context, as well as to the abstract module category 𝒮R. In particular, by taking 𝒩=0, we have a condition for the collection of associated primes of HI,Ji() to be finite.
Finally, using properties of the injective hull of R-elementary objects, we apply our theory to describe a general functorial setup that is similar to the one recently introduced by Àlvarez Montaner, Boix and Zarzuela in [2
J. Àlvarez Montaner, A. F. Boix, and S. Zarzuela, On some local cohomology spectral sequences. Int. Math. Res. Not. IMRN 2020 (2020), no. 19, 61976293 Zbl 1486.13024 MR 4165477
] (see also [3
J. Àlvarez Montaner, R. García López, and S. Zarzuela Armengou, Local cohomology, arrangements of subspaces and monomial ideals. Adv. Math. 174 (2003), no. 1, 3556 Zbl 1050.13009 MR 1959890
]). By considering systems of functors on 𝒮R satisfying certain conditions with respect to their values on injective hulls of R-elementary objects in 𝒮R, we obtain a general spectral sequence for their derived functors in this context. In particular, this gives us a spectral sequence for two-variable local cohomology objects in 𝒮R. In fact, it is striking that the nonclosed supports 𝕎(I,J) from [24
R. Takahashi, Y. Yoshino, and T. Yoshizawa, Local cohomology based on a nonclosed support defined by a pair of ideals. J. Pure Appl. Algebra 213 (2009), no. 4, 582600 Zbl 1160.13013 MR 2483839
] appear naturally in the theory developed in [2
J. Àlvarez Montaner, A. F. Boix, and S. Zarzuela, On some local cohomology spectral sequences. Int. Math. Res. Not. IMRN 2020 (2020), no. 19, 61976293 Zbl 1486.13024 MR 4165477
].
We conclude by showing two further applications of the spectral sequence in this formalism. The first is with generalized local cohomology functors given by derived functors of
ΓVI:SRSRlimt1Hom_R(ItV,),
where V is a finitely generated R-module and Hom¯(V,__):𝒮R𝒮R is the internal hom of an R-module V with respect to an object of 𝒮R. The second is the “generalized Nagata ideal transform” (see [2
J. Àlvarez Montaner, A. F. Boix, and S. Zarzuela, On some local cohomology spectral sequences. Int. Math. Res. Not. IMRN 2020 (2020), no. 19, 61976293 Zbl 1486.13024 MR 4165477
, Section 4] and [15
K. Divaani-Aazar and R. Sazeedeh, Cofiniteness of generalized local cohomology modules. Colloq. Math. 99 (2004), no. 2, 283290 Zbl 1072.13011 MR 2079733
]) that we extend to 𝒮R as follows:
ΔVI:SRSRlimt1Hom_R(ItV,),
where V is again a finitely generated R-module. In each case, we obtain spectral sequences for their derived functors.

2. R-elementary objects, associated primes and injectives

Let k be a field. Throughout, we let 𝒮=𝒮k be a strongly locally noetherian k-linear category. We let R be a commutative and noetherian k-algebra. Accordingly, the category 𝒮R of R-module objects in 𝒮 is locally noetherian. By [6
M. Artin and J. J. Zhang, Abstract Hilbert schemes. Algebr. Represent. Theory 4 (2001), no. 4, 305394 Zbl 1030.14003 MR 1863391
, Section B3], there is a “tensor product” __R__:𝒮R×RMod𝒮R that is right exact in both variables. If V is a flat R-module, then the functor __RV:𝒮R𝒮R is exact. Since k is a field, the functor k__:kMod𝒮 is exact for any 𝒮. Consequently, (kR)R__:RMod𝒮R is exact for the k-algebra R (see [6
M. Artin and J. J. Zhang, Abstract Hilbert schemes. Algebr. Represent. Theory 4 (2001), no. 4, 305394 Zbl 1030.14003 MR 1863391
, Section C1]). We will typically write :=k. We also note that an R-module object in 𝒮 may be described by giving 𝒮 and a “structure map” R satisfying analogues of the usual associativity and unit conditions (see [6
M. Artin and J. J. Zhang, Abstract Hilbert schemes. Algebr. Represent. Theory 4 (2001), no. 4, 305394 Zbl 1030.14003 MR 1863391
, Section B3.14]). For notions such as finitely generated objects, noetherian objects and locally noetherian categories that we will need throughout this paper, we refer the reader to [1
J. Adámek and J. Rosický, Locally presentable and accessible categories. London Math. Soc. Lecture Note Ser. 189, Cambridge University Press, Cambridge, 1994, 316 pp. Zbl 0795.18007 MR 1294136
].
We will now introduce some notation. Let r denote the endomorphism induced on 𝒮R by rR. We set the annihilator Ann()R to be the ideal
Ann():={rRr=0:=RRrRR=}.
It is clear that if , then Ann()Ann(). For any 𝒮R and any element aR, we form the short exact sequence
0Ker(a)a=Im(a)0.
Since 𝒮R is a Grothendieck category, an object 𝒮R is said to be finitely generated (see, for instance, [23
B. Stenström, Rings of quotients: an introduction to methods of ring theory. Grundlehren Math. Wiss. 217, Springer, New York-Heidelberg, 1975, 309 pp. Zbl 0296.16001 MR 0389953
, Section V.3]) if for any directed family {i}iI of subobjects of in 𝒮R such that iIi=, there exists some i0I such that i0=. For any 𝒮R, we denote by fg() the collection of finitely generated subobjects of in 𝒮R. We also let Pr() denote the collection of finitely generated subobjects 0𝒩 in 𝒮R such that Ann(𝒩) is a prime ideal in R.

Definition 2.1.

We will say that an object 0𝒮R is R-elementary if it satisfies the following two conditions:
  1. (a)
    is finitely generated.
  2. (b)
    There is a prime ideal 𝔭R such that for any non-zero subobject , we have Ann()=𝔭.
For 𝒮R, we will say that a prime ideal 𝔭R is an associated prime of if there exists an R-elementary object such that Ann()=𝔭. We denote by Ass() the collection of associated primes of .
Our first purpose is to show that associated primes exist for every non-zero 𝒮R. For this, we first prove the following result.

Lemma 2.2.

Let 0𝒮R. Then, there exists a finitely generated subobject 0𝒩 in 𝒮R such that Ann(𝒩) is a prime ideal, that is, Pr() is non-empty.

Proof.

Because 𝒮R is locally noetherian, the object 0𝒮R must have non-zero finitely generated subobjects. Since R is noetherian, we can choose an ideal IR such that I is maximal among annihilators of finitely generated non-zero subobjects of . Suppose that I=Ann(𝒩) for some 0𝒩fg(). We claim that I is prime. Otherwise, suppose we have a, bR such that aI, bI but abI.
Since abI=Ann(𝒩), we have a𝒩b𝒩=0, which implies that b𝒩:𝒩𝒩 factors through the kernel Ker(a𝒩). Hence, π𝒩ab𝒩=0, where π𝒩a is the canonical epimorphism π𝒩a:𝒩a𝒩. Since π𝒩a is a morphism of R-module objects, we obtain ba𝒩π𝒩a=π𝒩ab𝒩=0. Since π𝒩a is an epimorphism, it follows that ba𝒩=0, that is, bAnn(a𝒩).
Since 𝒮R is locally noetherian, we note that a𝒩𝒩 is finitely generated. Since aI=Ann(𝒩), we know that a𝒩0. Clearly, I=Ann(𝒩)Ann(a𝒩). Then, I(b,I)Ann(a𝒩), which contradicts the maximality of I.  

Proposition 2.3.

Let 0𝒮R. Then, Ass()ϕ.

Proof.

Suppose that Ass()=ϕ. By Lemma 2.2, we know that Pr()ϕ. As such, we may choose 0𝒩0fg() such that 𝔭0:=Ann(𝒩0) is a prime ideal. Since 𝔭0 cannot be an associated prime of , we can choose 0𝒩0𝒩0 such that 𝔭0=Ann(𝒩0)Ann(𝒩0). Again by Lemma 2.2, we know that Pr(𝒩0)ϕ. Accordingly, we can choose 0𝒩1𝒩0 such that 𝔭1:=Ann(𝒩1) is a prime ideal. But then,
𝔭0=Ann(𝒩0)Ann(𝒩0)Ann(𝒩1)=𝔭1.
By repeating the argument, we obtain an infinite increasing chain of prime ideals in R. This contradicts the fact that R is noetherian.  

Proposition 2.4.

Suppose that 0′′0 is a short exact sequence in 𝒮R. Then, we have Ass()Ass()Ass()Ass(′′).

Proof.

From Definition 2.1, it is clear that Ass()Ass(). We now consider 𝔭Ass() and an R-elementary object such that Ann()=𝔭. We set 𝒦:=. Suppose that 𝒦=0. Then, we have
=/(+)//=′′.
Accordingly, 𝔭Ass()Ass(′′). Otherwise, suppose that 𝒦0. Then, 0𝒦= is also R-elementary with 𝔭=Ann()=Ann(𝒦). By Definition 2.1, it is clear that 𝔭Ass(𝒦)Ass().  

Proposition 2.5.

Let 𝒮R be finitely generated. Then, there exists a finite filtration
(2.1)
0=01p=
such that each successive quotient i/i1 is R-elementary. Additionally, Ass()i=1pAss(i/i1).

Proof.

Let 0. By Proposition 2.3, we can choose a subobject 01 that is R-elementary. If =1, we are done. Otherwise, we continue by applying Proposition 2.3 to /1, which gives us a filtration as in (2.1). The filtration must be finite because 𝒮R is a locally noetherian category and 𝒮R is a finitely generated object. The last assertion follows by applying Proposition 2.4 repeatedly to the short exact sequences 0i1ii/i10.  

Corollary 2.6.

Let 𝒮R be finitely generated. Then, the set Ass() is finite.

Proof.

This is clear from Proposition 2.5.  
Since 𝒮R is a Grothendieck category, a subobject 𝒩 of 𝒮R is said to be essential if 𝒩𝒩0 for every non-zero subobject 𝒩 (see, for instance, [23
B. Stenström, Rings of quotients: an introduction to methods of ring theory. Grundlehren Math. Wiss. 217, Springer, New York-Heidelberg, 1975, 309 pp. Zbl 0296.16001 MR 0389953
, Section V.2]).

Corollary 2.7.

Let 𝒮R, and let 𝒩 be an essential subobject. Then, Ass(𝒩)=Ass().

Proof.

It follows from Proposition 2.4 that Ass(𝒩)Ass(). On the other hand, let 𝔭Ass() and 0fg() be an R-elementary object such that Ann()=𝔭. Since 𝒩 is essential, we know 𝒩0. It is now clear from Definition 2.1 that 𝒩 is R-elementary and Ann(𝒩)=𝔭.  
Since 𝒮R is a Grothendieck category, every object 𝒮R has an injective hull, which we will denote by (). We will now prove the main result of this section, which describes the injectives of 𝒮R in terms of injective envelopes of R-elementary objects.

Theorem 2.8.

Let 𝒮 be a strongly locally noetherian Grothendieck category, and let R be a commutative noetherian k-algebra. Then, every injective object in 𝒮R can be expressed as a direct sum of injective hulls of R-elementary objects.

Proof.

We choose a family {i}iI of subobjects of satisfying the following two conditions:
  1. (1)
    Each i is isomorphic to the injective hull of an R-elementary object in 𝒮R.
  2. (2)
    The sum iIi is direct, that is, (i)(jij)=0 for each iI.
Since finite limits commute with filtered colimits in 𝒮R, condition (2) is equivalent to 0=(i)(jJj) for each iI and any finite JI\{i}. Accordingly, by Zorn’s lemma, we may choose a maximal such family {i}iI0 inside .
We now set :=iI0i. Since 𝒮R is locally noetherian, the direct sum of injectives must be injective. Accordingly, we can split as =′′. If ′′=0, the result is already proved. If ′′0, it follows from Proposition 2.3 that there is an R-elementary object ′′. Since ′′ is injective, the injective hull ()′′. This contradicts the maximality of the family {i}iI0.  
If TR is a multiplicatively closed set, the localization T of an object 𝒮R with respect to T is given by RR[T1] (see [6
M. Artin and J. J. Zhang, Abstract Hilbert schemes. Algebr. Represent. Theory 4 (2001), no. 4, 305394 Zbl 1030.14003 MR 1863391
, Section B6]). The localization T may also be expressed as the filtered colimit of copies of connected by morphisms t: over all tT (see [6
M. Artin and J. J. Zhang, Abstract Hilbert schemes. Algebr. Represent. Theory 4 (2001), no. 4, 305394 Zbl 1030.14003 MR 1863391
, Section B6.1]). Accordingly, the kernel of the canonical map T is given by the filtered colimit (see [6
M. Artin and J. J. Zhang, Abstract Hilbert schemes. Algebr. Represent. Theory 4 (2001), no. 4, 305394 Zbl 1030.14003 MR 1863391
, Proposition B6.2])
(2.2)
Ker(T)=tTKer(t:).
Moreover, since R[T1] is a flat R-module, it follows from [6
M. Artin and J. J. Zhang, Abstract Hilbert schemes. Algebr. Represent. Theory 4 (2001), no. 4, 305394 Zbl 1030.14003 MR 1863391
, Proposition C1.7] that localization is an exact functor on 𝒮R. If 𝔭 is a prime ideal, we denote by 𝔭 the localization with respect to R\𝔭. We now have the following definition.

Definition 2.9.

Let 𝒮R. Then, the support Supp() of consists of all prime ideals 𝔭R such that 𝔭0.

Proposition 2.10.

We now have the following result:
  1. (a)
    Let 𝒮R, and let 𝔭Supp(). Then, 𝔭Ann().
  2. (b)
    If 𝒮R is finitely generated, then Supp()=𝕍(Ann()), the collection of prime ideals containing Ann().
  3. (c)
    For a short exact sequence 0′′0 in 𝒮R, we have Supp()=Supp()Supp(′′).

Proof.

(a) If 𝔭Ann(), we choose tAnn() such that t𝔭. Since t is a unit in R𝔭, it follows that Rt=tRR𝔭=0:𝔭=RR𝔭RR𝔭=𝔭 is an isomorphism. This gives 𝔭=0, and hence 𝔭Supp(). This proves (a).
Additionally, suppose that is finitely generated and 𝔭𝕍(Ann()), but 𝔭=0. Since the colimit in (2.2) is filtered, this means that Ker(𝔭)==Ker(t) for some t𝔭. Then, tAnn() and t𝔭, which is a contradiction. This proves (b). The result of (c) follows directly from the fact that localization is exact.  
We conclude this section by establishing some properties of injective hulls in 𝒮R that we will need throughout this paper. We begin with the following result.

Lemma 2.11.

Let 𝒮R, and let TR be a multiplicatively closed subset. Then, any subobject of T𝒮R[T1] is of the form 𝒩T for some subobject 𝒩 in 𝒮R.

Proof.

Let 𝒦T𝒮R[T1] be a subobject. We set up the following two pullback diagrams
(2.3)
where the right-hand square in (2.3) follows by applying the exact functor __RR[T1] to the left-hand square. We note that T=RR[T1]=RR[T1]RR[T1]=TRR[T1] and 𝒦RR[T1]=𝒦R[T1]R[T1]RR[T1]=𝒦. From (2.3), it is now clear that 𝒩 is a subobject of and that 𝒦=𝒩RR[T1].  

Proposition 2.12.

Let 𝒮R be an injective object, and let TR be a multiplicatively closed subset. Then, the localization T is an injective object in 𝒮R[T1].

Proof.

Let {𝒫i}iI be a set of noetherian generators for 𝒮R. Since 𝒮 is strongly locally noetherian, it follows from [6
M. Artin and J. J. Zhang, Abstract Hilbert schemes. Algebr. Represent. Theory 4 (2001), no. 4, 305394 Zbl 1030.14003 MR 1863391
, Proposition B5.1, Corollary B3.17] that {𝒫iRR[T1]}iI is a set of noetherian generators for 𝒮R[T1]. We choose one of these generators 𝒫{𝒫i}iI and consider a subobject 𝒦𝒫T in 𝒮R[T1]. Applying Lemma 2.11, we have a subobject 𝒩𝒫 in 𝒮R such that the localization 𝒩T=𝒦. We note that since 𝒫𝒮R is noetherian, so is 𝒩.
Since 𝒮R is injective, it follows that we have a surjection 𝒮R(𝒫,)𝒮R(𝒩,). As mentioned above, the localization T may be expressed as a filtered colimit of copies of . Since 𝒩, 𝒫 are finitely generated, we have induced surjections 𝒮R(𝒫,T)𝒮R(𝒩,T). In other words, we have a surjection 𝒮R[T1](𝒫T,T)=𝒮R[T1](𝒫RR[T1],T)𝒮R[T1](𝒩RR[T1],T)=𝒮R[T1](𝒦,T). Applying Baer’s criterion (see [23
B. Stenström, Rings of quotients: an introduction to methods of ring theory. Grundlehren Math. Wiss. 217, Springer, New York-Heidelberg, 1975, 309 pp. Zbl 0296.16001 MR 0389953
, Proposition V.2.9]) to the Grothendieck category 𝒮R[T1], it follows that T is injective.  
We now need the following fact, which does not seem to have appeared before in the literature.

Lemma 2.13.

Let 𝒟 be a locally noetherian category. Then, a filtered colimit of essential monomorphisms in 𝒟 is an essential monomorphism.

Proof.

Let {i𝒜i}iI be a filtered system of essential monomorphisms. Then, :=limiIi𝒜:=limiI𝒜i is a monomorphism, and we claim that 𝒜 is essential. For this, we consider a subobject 0𝒜 and suppose that =0. Since every object in 𝒟 is the sum of its finitely generated subobjects, it is enough to consider the case where is finitely generated.
Accordingly, we can choose jI large enough such that the inclusion 𝒜 factors through the canonical morphism 𝒜j𝒜. Then, 𝒜j must be a monomorphism. Since j𝒜j is essential, we have j0. Since j, it follows that 0, which is a contradiction.  

Proposition 2.14.

Let TR be a multiplicatively closed subset. Let 𝒩𝒮R, and let (𝒩) be its injective hull. Then, the localization (𝒩)T is the injective hull of 𝒩T in 𝒮R[T1].

Proof.

We know that the canonical inclusion 𝒩(𝒩) is essential. Since the localization is obtained by taking a filtered colimit, it follows from Lemma 2.13 that 𝒩T(𝒩)T is essential. Since (𝒩) is injective in 𝒮R, we know from Proposition 2.12 that (𝒩)T is injective in 𝒮R[T1]. The result is now clear.  

Proposition 2.15.

Let 𝒮 be a strongly locally noetherian Grothendieck category, and let R be a commutative noetherian k-algebra. Let 𝒩 be an essential subobject in 𝒮R. Then, Supp(𝒩)=Supp().

Proof.

According to Proposition 2.14, it is clear that Supp(𝒩)=Supp((𝒩)). In general, if 𝒩 is essential, there is an inclusion (𝒩). Accordingly, Supp(𝒩)=Supp()=Supp((𝒩)).  

3. Local cohomology objects for a pair of ideals

Let I, J be ideals in R. If 𝒮R and 𝔊 is any family of subobjects of , we write 𝔊 (resp. 𝔊) for the direct sum (resp. the sum in ) of all the objects in 𝔊. We now define
(3.1)
|ΓI,J()| :={𝒩fg()InAnn(𝒩)+Jfor n1}, ΓI,J() :=|ΓI,J()|.
The definition in (3.1) is motivated by the notion of local cohomology for modules with respect to a pair of ideals introduced by Takahashi, Yoshino and Yoshizawa in [24
R. Takahashi, Y. Yoshino, and T. Yoshizawa, Local cohomology based on a nonclosed support defined by a pair of ideals. J. Pure Appl. Algebra 213 (2009), no. 4, 582600 Zbl 1160.13013 MR 2483839
]. We will say that 𝒮R is (I,J)-torsion if ΓI,J()=. We will say that 𝒮R is (I,J)-torsion free if ΓI,J()=0. When J=0, we will denote ΓI,J simply by ΓI.

Lemma 3.1.

We start with the following result:
  1. (a)
    For 𝒮R, the family |ΓI,J()| is closed under subobjects and finite sums. Further, any finitely generated 𝒩ΓI,J() lies in the family |ΓI,J()|.
  2. (b)
    ΓI,J is a left exact functor on 𝒮R.

Proof.

(a) It is clear that |ΓI,J()| is closed under subobjects. We consider 𝒩1, 𝒩2|ΓI,J()|, with In1Ann(𝒩1)+J and In2Ann(𝒩2)+J. Then, we have
In1+n2 (Ann(𝒩1)+J)(Ann(𝒩2)+J) Ann(𝒩1)Ann(𝒩2)+JAnn(𝒩1+𝒩2)+J.
Hence, |ΓI,J()| is closed under finite sums. Now if 𝒩ΓI,J() is finitely generated, it follows from the definition in (3.1) that 𝒩 lies within a finite sum of objects in |ΓI,J()|. Since |ΓI,J()| is closed under subobjects and finite sums, the result is now clear.
(b) Let ϕ: be a morphism in 𝒮R. For any 𝒩, the quotient ϕ(𝒩) satisfies Ann(𝒩)Ann(ϕ(𝒩)). Accordingly, if 𝒩|ΓI,J()|, then ϕ(𝒩)|ΓI,J()|. This determines a morphism ΓI,J(ϕ):ΓI,J()ΓI,J(). In particular, if ϕ: is a monomorphism, we have |ΓI,J()||ΓI,J()|. Hence, ΓI,J()ΓI,J().  
As defined in [24
R. Takahashi, Y. Yoshino, and T. Yoshizawa, Local cohomology based on a nonclosed support defined by a pair of ideals. J. Pure Appl. Algebra 213 (2009), no. 4, 582600 Zbl 1160.13013 MR 2483839
], we now consider for the pair of ideals (I,J) the set
𝕎(I,J)={𝔭Spec(R)In𝔭+Jfor n1}.
When J=0, we note that 𝕎(I,J)=𝕍(I), the set of prime ideals containing I.

Lemma 3.2.

Let 𝒩𝒮R be finitely generated. Then, Supp(𝒩)𝕎(I,J) if and only if InAnn(𝒩)+J for n1.

Proof.

Since 𝒩 is finitely generated, it follows from Proposition 2.10 that Supp(𝒩)=𝕍(Ann(𝒩)). First, we suppose that InAnn(𝒩)+J. Then, if 𝔭Supp(𝒩), that is, 𝔭Ann(𝒩), we see that InJ+𝔭, that is, 𝔭𝕎(I,J).
Conversely, suppose that Supp(𝒩)𝕎(I,J). If 𝔭 is a prime ideal containing Ann(𝒩), then there is n𝔭1 such that In𝔭𝔭+J. Since R is noetherian, there are only finitely many primes minimal over Ann(𝒩). It follows that we can find n1 such that Inrad(Ann(𝒩))+J. Again since R is noetherian, some power of rad(Ann(𝒩)) lies in Ann(𝒩) and the result follows.  
It follows from Lemma 3.2 that ΓI,J() is the sum of all finitely generated subobjects 𝒩fg() satisfying Supp(𝒩)𝕎(I,J). We note that this implies Supp(ΓI,J())𝕎(I,J). We will now describe the associated primes of ΓI,J().

Proposition 3.3.

For 𝒮R, we have Ass()𝕎(I,J)=Ass(ΓI,J()).

Proof.

Since ΓI,J(), we have Ass(ΓI,J())Ass(). If 𝔭Ass(ΓI,J()), we consider an R-elementary object ΓI,J() such that Ann()=𝔭. By Lemmas 3.1 and 3.2, we see that 𝔭Ass()Supp()𝕎(I,J). Hence, Ass(ΓI,J())Ass()𝕎(I,J).
Conversely, we take 𝔭Ass()𝕎(I,J) and consider an R-elementary object such that Ann()=𝔭. Since 𝔭𝕎(I,J), we can take n1 such that In𝔭+J=Ann()+J. Hence, |ΓI,J()| and therefore ΓI,J(). This gives 𝔭Ass(ΓI,J()).  
If 𝒜 is an abelian category, we recall (see, for instance, [8
A. Beligiannis and I. Reiten, Homological and homotopical aspects of torsion theories. Mem. Amer. Math. Soc. 188 (2007), no. 883, 207 pp. Zbl 1124.18005 MR 2327478
, Section 1.1]) that a torsion theory on 𝒜 consists of a pair (𝒯,) of strict and full subcategories such that 𝒜(X,Y)=0 for any X𝒯, Y and any Z𝒜 fits into a short exact sequence 0Z𝒯ZZ0 with Z𝒯𝒯 and Z. Additionally, (𝒯,) is said to be hereditary if the torsion class 𝒯 is closed under subobjects.

Proposition 3.4.

For ideals I, J, the pair (𝒯(𝒮)I,J,(𝒮)I,J) of full subcategories of 𝒮R, where
Ob(𝒯(𝒮)I,J) :={𝒮RΓI,J()=} Ob((𝒮)I,J) :={𝒮RΓI,J()=0}
forms a hereditary torsion theory in 𝒮R.

Proof.

We consider ϕ:′′ in 𝒮R, where 𝒯(𝒮)I,J, ′′(𝒮)I,J. We consider 𝒩fg(). Then, Supp(ϕ(𝒩))Supp(𝒩)𝕎(I,J) and hence ϕ()ΓI,J(′′)=0. Now for any 𝒮R, we consider the short exact sequence
0ΓI,J()/ΓI,J()0.
By Lemma 3.1, it is clear that ΓI,J()𝒯(𝒮)I,J. If /ΓI,J()(𝒮)I,J, there is some 0𝒩fg(/ΓI,J()) such that Supp(𝒩)𝕎(I,J). Then, we have a short exact sequence
0ΓI,J()𝒩0𝒩0,
where 𝒩0 is the preimage of 𝒩 in . Then, Supp(𝒩0)=Supp(ΓI,J())Supp(𝒩)𝕎(I,J) and hence 𝒩0ΓI,J(), which implies that 𝒩=0. Hence, (𝒯(𝒮)I,J,(𝒮)I,J) is a torsion theory on 𝒮R. The fact that 𝒯(𝒮)I,J is closed under subobjects is also clear from Lemma 3.2.  

Definition 3.5.

Let 𝒮R, and let I,JR be ideals. Then, for any i0, the i-th local cohomology object HI,Ji() of 𝒮R with respect to (I,J) is given by the i-th right derived functor of ΓI,J, that is, HI,Ji():=iΓI,J().
When J=0, we set HIi:=HI,0i() and refer to HI() as the local cohomology objects with respect to I.
We will now show that if 𝒮R is (I,J)-torsion, then HI,Ji()=0 for i>0. This will be achieved by a sequence of steps. We will start by showing that the torsion theory (𝒯(𝒮)I,J,(𝒮)I,J) satisfies the property that the torsion part of an injective in 𝒮R is still an injective.

Lemma 3.6.

Let 𝒮R be an injective object. Then, for any ideal IR, ΓI()=ΓI,0() is also an injective object in 𝒮R.

Proof.

We will show that for any finitely generated 𝒮R and any subobject 𝒩, a morphism ϕ:𝒩ΓI() extends to a morphism ψ:ΓI(). Since 𝒮R is locally noetherian, it has a set of finitely generated generators. Accordingly, it will follow from Baer’s criterion [23
B. Stenström, Rings of quotients: an introduction to methods of ring theory. Grundlehren Math. Wiss. 217, Springer, New York-Heidelberg, 1975, 309 pp. Zbl 0296.16001 MR 0389953
, Proposition V.2.9] that ΓI() is injective.
Since ϕ(𝒩)ΓI() is finitely generated, we can choose t1 such that Itϕ(𝒩)=0. Since is injective, the morphism 𝒩ϕΓI() extends to a morphism ψ:. In particular, ϕ(𝒩)ψ(). Since is finitely generated, so is ψ(). Applying the version of Artin–Rees lemma proved in [6
M. Artin and J. J. Zhang, Abstract Hilbert schemes. Algebr. Represent. Theory 4 (2001), no. 4, 305394 Zbl 1030.14003 MR 1863391
, Proposition D1.8], we can find c1 such that for all nc, we have
(3.2)
Inψ()ϕ(𝒩)=Inc(Icψ()ϕ(𝒩))Incϕ(𝒩).
Putting n=t+c, it follows from (3.2) that It+cψ()ϕ(𝒩)Itϕ(𝒩)=0. We now note that
(3.3)
ϕ(It+c𝒩) ϕ(𝒩), ϕ(It+c𝒩) ψ(It+c𝒩)ψ(It+c)It+cψ().
Accordingly, it follows from (3.3) that ϕ(It+c𝒩)It+cψ()ϕ(𝒩)=0. Now since
It+c𝒩 =Ker(It+c𝒩), It+c+𝒩 =Im(It+c𝒩),
the morphism 0ϕ:It+c𝒩ΓI() induces ϕ:It+c+𝒩ΓI() such that ϕIt+c=0 and ϕ𝒩=ϕ. Again since is injective, there exists ψ: extending It+c+𝒩ϕΓI(). We now see that
It+cψ()=ψ(It+c)=ϕ(It+c)=0
whence it follows that ψ()ΓI(). This proves the result.  
The next step is to make use of the directed sets 𝕎~(I,J) defined in [24
R. Takahashi, Y. Yoshino, and T. Yoshizawa, Local cohomology based on a nonclosed support defined by a pair of ideals. J. Pure Appl. Algebra 213 (2009), no. 4, 582600 Zbl 1160.13013 MR 2483839
, Section 3] as follows: the elements of 𝕎~(I,J) are ideals KR satisfying InK+J for n1. For K, K𝕎~(I,J), we will say that KK if KK. We observe that each 𝕎~(I,J) is filtered.

Lemma 3.7.

Let 𝒮R. Then, ΓI,J()=limK𝕎~(I,J)ΓK().

Proof.

We set 𝒩:=limK𝕎~(I,J)ΓK() and consider some 𝒩fg(𝒩). Since 𝕎~(I,J) is filtered, we choose K𝕎~(I,J) such that 𝒩ΓK(), that is, KmAnn(𝒩) for some m1. Since K𝕎~(I,J), we may choose n1 such that InK+J. Then, ImnKm+JAnn(𝒩)+J and hence 𝒩ΓI,J(). Conversely, we consider some 𝒩′′fg(ΓI,J()). Then, we have IlAnn(𝒩′′)+J for some l1. Then, Ann(𝒩′′)𝕎~(I,J), and we have 𝒩′′ΓAnn(𝒩′′)(). Accordingly, we have fg(ΓI,J())=fg(𝒩) and hence ΓI,J()=𝒩 because 𝒮R is locally noetherian.  

Proposition 3.8.

Let 𝒮R be an injective object. Then, for any ideals I,JR, ΓI,J() is also an injective object in 𝒮R.

Proof.

Since 𝒮R is injective, it follows from Lemma 3.6 that ΓK() is injective for each K𝕎~(I,J). We now consider some finitely generated 𝒮R, a subobject 𝒩 and a morphism ϕ:𝒩ΓI,J(). By Lemma 3.7, we know that ΓI,J()=limK𝕎~(I,J)ΓK(). Since 𝕎~(I,J) is filtered and 𝒩 is finitely generated, we can find K0𝕎~(I,J) such that ϕ factors through ΓK0(). Since ΓK0() is injective, we now have ψ:ΓK0() extending 𝒩ΓK0(). Composing with the canonical map ΓK0()limK𝕎~(I,J)ΓK()=ΓI,J(), we obtain an extension of ϕ:𝒩ΓI,J() to a morphism ψ:ΓI,J(). Since 𝒮R is a locally noetherian category, it now follows from Baer’s criterion (see [23
B. Stenström, Rings of quotients: an introduction to methods of ring theory. Grundlehren Math. Wiss. 217, Springer, New York-Heidelberg, 1975, 309 pp. Zbl 0296.16001 MR 0389953
, Proposition V.2.9]) that ΓI,J() is injective.  

Lemma 3.9.

Let 𝒮R be an (I,J)-torsion object. Then, its injective hull () is also (I,J)-torsion.

Proof.

Since () is injective, Proposition 3.8 implies that ΓI,J(()) is also injective in 𝒮R. Accordingly, we have a direct sum decomposition ()=ΓI,J(())𝒩. We already know that 𝒮R is an (I,J)-torsion object, and it is clear from the definition in (3.1) that lies inside the torsion part of (). Then, 𝒩=0. But is essential in (), whence it follows that 𝒩=0, that is, ()=ΓI,J(()).  

Proposition 3.10.

Let 𝒮R be an (I,J)-torsion object. Then, HI,Ji()=0 for i>0.

Proof.

By Lemma 3.9, we know that the injective hull of an (I,J)-torsion object is also (I,J)-torsion. By Proposition 3.4, we know that the torsion class 𝒯(𝒮)I,J is closed under both quotients and subobjects. Accordingly, we can obtain a resolution consisting of injective objects that are (I,J)-torsion. Therefore, applying the functor ΓI,J leaves this sequence unchanged. By definition, the cohomology of the resolution vanishes in positive degrees, and this proves the result.  

Corollary 3.11.

Let 𝒮R. Then,
  1. (a)
    HI,Ji()HIi(/ΓI,J()) for i>0.
  2. (b)
    HI,Ji() is (I,J)-torsion for any i0.

Proof.

By Proposition 3.10, we know that HIi(ΓI,J())=0 for i>0. The result of (a) is now clear from the long exact sequence of cohomologies obtained by applying ΓI,J(__) to the short exact sequence 0ΓI,J()/ΓI,J()0. To prove (b), we note that HI,J() are, by definition, the homology objects of ΓI,J(), where is an injective resolution. Since each ΓI,J() lies in the hereditary torsion class 𝒯(𝒮)I,J, so does its subquotient HI,J().  

4. Associated primes of local cohomology objects

In this section, we study finiteness conditions on local cohomology objects and their associated primes. We give a condition for the set of associated primes of the local cohomology object HI,J(__) with respect to an ideal pair (I,J) in an abstract module category to be finite. For 𝒮R, we say that aR is a non-zero divisor on if a: is a monomorphism.

Lemma 4.1.

Let aR be a zero divisor on 𝒮R. Then, a lies in some associated prime of .

Proof.

By definition, we know that Ker(a)0. Using Proposition 2.3, we pick a prime ideal 𝔭Ass(Ker(a))Ass(). Then, 𝔭=Ann() for some R-elementary object Ker(a). Then, aAnn(Ker(a))Ann()=𝔭.  

Lemma 4.2.

Let I,JR be ideals, and let 𝒮R be a finitely generated object such that ΓI,J()=0. Then, I contains a non-zero divisor on .

Proof.

Suppose that all elements of IR are zero divisors on . By Lemma 4.1, I𝔭Ass()𝔭. Since is finitely generated, Ass() is finite by Corollary 2.6. By prime avoidance, there is some 𝔭0Ass() such that I𝔭0. Hence, 𝔭0𝕎(I,J).
Now since 𝔭0 is an associated prime of , there is an R-elementary object 0 such that 𝔭0=Ann(). Then, Supp()=𝕍(𝔭0)𝕎(I,J) and hence ΓI,J()=0, which is a contradiction.  
For any in 𝒮R and subobjects , ′′, we set
(′′:):={aRIm(a)′′}.
It is clear that any such (′′:) is an ideal in R.

Lemma 4.3.

Let 𝒩 in 𝒮R, and let 𝒦/𝒩 be a finitely generated subobject. Then, there exists a finitely generated subobject 𝒦 such that
  1. (1)
    The quotient map π:/𝒩 restricts to an epimorphism π:𝒦𝒦.
  2. (2)
    Ann(𝒦)=(𝒩:𝒦).

Proof.

We put 𝒦′′:=π1(𝒦). Since finitely generated subobjects of 𝒦′′ form a filtered system and π(𝒦′′)=𝒦, we can find 𝒦fg(𝒦′′) such that π(𝒦)=𝒦. This proves (1). To prove (2), for any aR, we consider the commutative diagram
If aAnn(𝒦), that is, a𝒦=0, we get πιa𝒦=ιa𝒦π=0, which gives Im(a𝒦)Ker(π)=𝒩, that is, a(𝒩:𝒦). Conversely, if a(𝒩:𝒦), we get ιa𝒦π=0. Since π is an epimorphism and ι is a monomorphism, we get a𝒦=0 or aAnn(𝒦).  

Proposition 4.4.

Let 𝒮R be finitely generated, and let I,JR be ideals. Suppose that i0 is such that HI,Jj() is finitely generated for all j<i. Then, for any finitely generated 𝒩HI,Ji() such that 𝒩JHI,Ji(), the collection Ass(HI,Ji()/𝒩) is finite.

Proof.

For i=0, we know that HI,J0()=ΓI,J() is finitely generated, and the result is clear. We will proceed by induction on i. By Corollary 3.11, we know that HI,Jj()HI,Jj(/ΓI,J()) for j>0. Also, we know that ΓI,J(/ΓI,J())=0. Accordingly, we may suppose that 𝒮R is finitely generated with ΓI,J()=0, and it follows from Lemma 4.2 that we can find aI which is a non-zero divisor on .
By Corollary 3.11, we know that HI,Ji() is (I,J)-torsion, that is, ΓI,J(HI,Ji())=HI,Ji(). Since 𝒩HI,Ji() is finitely generated, it follows from Lemma 3.1 that ItAnn(𝒩)+J for some t>0. Accordingly, we have at𝒩J𝒩. We now consider the short exact sequence
(4.1)
0at/at0.
Applying ΓI,J(__) to (4.1), the long exact sequence of derived functors gives us
(4.2)
HI,Jl()atHI,Jl()HI,Jl(/at)HI,Jl+1()atHI,Jl+1()
for every l0. Since 𝒮R is locally noetherian, the collection of finitely generated objects in 𝒮R is closed under extensions, subobjects and quotients (see [23
B. Stenström, Rings of quotients: an introduction to methods of ring theory. Grundlehren Math. Wiss. 217, Springer, New York-Heidelberg, 1975, 309 pp. Zbl 0296.16001 MR 0389953
, Section V.4.2]). Since HI,Jj() is finitely generated for all j<i, it follows from (4.2) that HI,Jl(/at) is finitely generated for l<i1.
Since at𝒩J𝒩, we also have the following diagram where the top row is exact and the vertical morphisms are epimorphisms
We now observe that Ker(μ¯)=μ1(J𝒩)/𝒩. This gives us the short exact sequence
(4.3)
0HI,Ji1(/at)/δ1(𝒩) =(Im(δ)+𝒩)/𝒩 Ker(μ¯) =μ1(J𝒩)/𝒩μ1(J𝒩)/(Ker(μ)+𝒩)0,
where we have used Ker(μ)=Im(δ). Since J𝒩𝒩 is finitely generated, so is its subobject μ1(J𝒩)/Ker(μ). Then, the quotient μ1(J𝒩)/(Ker(μ)+𝒩) of μ1(J𝒩)/Ker(μ) is also finitely generated.
On the other hand, we have the short exact sequence
(4.4)
0Im(ι)δ1(𝒩)=Ker(δ)δ1(𝒩)δ1(𝒩)𝛿δ(δ1(𝒩))0.
By assumption, HI,Ji1() is finitely generated, and hence so is its subquotient Im(ι)δ1(𝒩). Also since 𝒩 is finitely generated, so is δ(δ1(𝒩))𝒩. Again since 𝒮R is locally noetherian, it follows from (4.4) that δ1(𝒩) is finitely generated. Additionally, we note that
δ(JHI,Ji1(/at))Jδ(HI,Ji1(/at))JHI,Ji()𝒩JHI,Ji1(/at)δ1(𝒩).
We now put 𝒦:=HI,Ji1(/at)/δ1(𝒩) and 𝒫:=Ker(μ¯). By the induction assumption, we see that Ass(𝒦) is a finite set. Since μ1(J𝒩)/(Ker(μ)+𝒩) is finitely generated, from the short exact sequence (4.3), it now follows from Proposition 2.4 that Ass(𝒫) is finite.
Since 𝒩 is finitely generated, so is its image π(𝒩). To prove the result, it now suffices to show that
Ass(HI,Ji()/𝒩)Ass(𝒫)Ass(π(𝒩)).
Accordingly, we choose some prime ideal 𝔭Ass(HI,Ji()/𝒩)\Ass(𝒫). Then, there is an R-elementary object HI,Ji()/𝒩 such that 𝔭=Ann(). Applying Lemma 4.3, we can choose a finitely generated subobject HI,Ji() such that π:HI,Ji()HI,Ji()/𝒩 restricts to an epimorphism and (𝒩:)=Ann()=𝔭. We now consider the short exact sequence
By Proposition 2.4, we get Ass()Ass(𝒫)Ass(a¯t)Ass(𝒫)Ass(a¯t). Since 𝔭Ass() does not lie in Ass(𝒫), we obtain 𝔭Ass(a¯t)=Ass(a¯tπ())=Ass(πat).
We know that at is finitely generated. Again since HI,Ji() is (I,J)-torsion, it follows that at+sJJHI,Ji()𝒩 for some s1. But since (𝒩:)=𝔭, we get at+s𝔭, that is, a𝔭. Again since 𝔭=(𝒩:), we obtain at𝒩 and hence πatπ(𝒩). Then, 𝔭Ass(πat)Ass(π(𝒩)).  

Theorem 4.5.

Let 𝒮R be finitely generated, and let I,JR be ideals. Suppose that i0 is such that HI,Jj() is finitely generated for all j<i. Suppose that JHI,Ji() is finitely generated. Then, for any finitely generated 𝒩HI,Ji(), the collection Ass(HI,Ji()/𝒩) is finite.

Proof.

Since 𝒩 and JHI,Ji() are both finitely generated, so is 𝒩+JHI,Ji(). By Proposition 4.4, it now follows that Ass(HI,Ji()/(𝒩+JHI,Ji())) is finite. We now consider the short exact sequence
(4.5)
0(𝒩+JHI,Ji())/𝒩HI,Ji()/𝒩HI,Ji()/(𝒩+JHI,Ji())0.
Since (𝒩+JHI,Ji())/𝒩 is finitely generated and Ass(HI,Ji()/(𝒩+JHI,Ji())) is finite, the result follows by applying Proposition 2.4 to the short exact sequence (4.5).  

Corollary 4.6.

Let 𝒮R be finitely generated, and let I,JR be ideals. Suppose that i0 is such that HI,Jj() is finitely generated for all j<i. Suppose that JHI,Ji() is finitely generated. Then, the collection Ass(HI,Ji()) is finite.

Proof.

This follows directly from Theorem 4.5 by setting 𝒩=0.  
We conclude this section with the following fact, which extends a result of Brodmann, Rotthaus and Sharp [11
M. Brodmann, Ch. Rotthaus, and R. Y. Sharp, On annihilators and associated primes of local cohomology modules. J. Pure Appl. Algebra 153 (2000), no. 3, 197227 Zbl 0968.13010 MR 1783166
].

Proposition 4.7.

Let 𝒮R be such that the collection χ of maximal elements in Ass() is finite. Let IR be an ideal. Suppose that for each 𝔭χ, there is n𝔭>0 such that In𝔭𝔭=0. If n:=max{n𝔭𝔭χ}, then In=0.

Proof.

We consider some 𝒩fg(). It is clear that (In𝒩)𝔭In𝔭=0. Since 𝒮R is locally noetherian, we know that In𝒩𝒩 is finitely generated. From (2.2), we see that for each 𝔭χ, we have
(4.6)
(In𝒩)=Ker((In𝒩)(In𝒩)𝔭=0)=tR\𝔭Ker((In𝒩)𝑡(In𝒩)).
Since the union on the right-hand side of (4.6) is filtered, we can find s𝔭R\𝔭 such that s𝔭Ann(In𝒩). Let KR be the ideal generated by the collection {s𝔭}𝔭χ. It is clear that KAnn(In𝒩). Since K is not contained in any of the prime ideals in χ, and χ is finite, it follows from prime avoidance that we can choose an element rK\𝔭χ𝔭. From Lemma 4.1, it is clear that r must be a non-zero divisor on . But multiplication by r is zero on the subobject In𝒩. Hence, In𝒩=0. It follows that In=𝒩fg()In𝒩=0.  

5. Spectral sequences for local cohomology type functors on 𝒮R

In this section, we will show that we can construct spectral sequences for “local cohomology type” functors on 𝒮R by creating an axiomatic setup similar to Àlvarez Montaner, Boix and Zarzuela [2
J. Àlvarez Montaner, A. F. Boix, and S. Zarzuela, On some local cohomology spectral sequences. Int. Math. Res. Not. IMRN 2020 (2020), no. 19, 61976293 Zbl 1486.13024 MR 4165477
]. Our main tools will be the properties of injective hulls of R-elementary objects proved in Section 2. We will then exhibit three different situations where this axiomatic setup can be used to obtain spectral sequences.
Let P be a finite poset, and let Fun(P,𝒮R) denote the category of systems of objects in 𝒮R indexed over P. We suppose from now on that 𝒮 has enough projectives. Then, it follows from Artin and Zhang [6
M. Artin and J. J. Zhang, Abstract Hilbert schemes. Algebr. Represent. Theory 4 (2001), no. 4, 305394 Zbl 1030.14003 MR 1863391
, Lemma D3.2] that 𝒮R has enough projectives. Then, we know that Fun(P,𝒮R) has enough projectives and Fun(Pop,𝒮R) has enough injectives (see, for instance, [25
C. A. Weibel, An introduction to homological algebra. Cambridge Stud. Adv. Math. 38, Cambridge University Press, Cambridge, 1994, 450 pp. Zbl 0797.18001 MR 1269324
, Section 2.3]). We also note that Fun(P,𝒮R) and Fun(Pop,𝒮R) are Grothendieck categories.
For any functor 𝔊Fun(Pop,𝒮R), we consider the complex (C(𝔊),) given by
Ck(𝔊):=p0<p1<<pk𝔊(p0)  k=j=0k+1(1)jjk:Ck(𝔊)Ck+1(𝔊),
where jk is induced by deleting the j-th term in the sequence {p0<p1<<pk<pk+1}. Because P is a finite poset, it follows from the proof of [20
A. Neeman, Triangulated categories. Ann. of Math. Stud. 148, Princeton University Press, Princeton, NJ, 2001, 449 pp. Zbl 0974.18008 MR 1812507
, Lemma A.3.2] that this complex computes the derived functor of the inverse limit over P, that is, for any 𝔊Fun(Pop,𝒮R), we have Hi(C(𝔊))=ilimpP𝔊(p).
Similarly, for any functor 𝔉Fun(P,𝒮R), we consider the complex (C(𝔉),) given by
Ck(𝔉):=p0<p1<<pk𝔉(p0)  k=j=0k+1(1)jkj:Ck+1(𝔉)Ck(𝔉),
where kj is induced by deleting the j-th term in the sequence {p0<p1<<pk<pk+1}. Then, it follows from [20
A. Neeman, Triangulated categories. Ann. of Math. Stud. 148, Princeton University Press, Princeton, NJ, 2001, 449 pp. Zbl 0974.18008 MR 1812507
, Section B.1.2] that this complex computes the derived functor of the direct limit over P, that is, for any 𝔉Fun(P,𝒮R), we have Hi(C(𝔉))=𝕃icolimpP𝔉(p).
From now onwards, we fix an ideal IR. We suppose that I may be expressed as I=Iα1Iαk, where Iα1,,Iαk are ideals of R. As in [2
J. Àlvarez Montaner, A. F. Boix, and S. Zarzuela, On some local cohomology spectral sequences. Int. Math. Res. Not. IMRN 2020 (2020), no. 19, 61976293 Zbl 1486.13024 MR 4165477
, Section 2], we let P denote the finite poset whose elements are all the possible different sums of the ideals I1,,In, ordered by reverse inclusion. By P^, we will mean the poset P^=P{1P^} obtained by adding a final element 1P^ to P (even if P already has a final element). The Alexandrov topology on P (see [2
J. Àlvarez Montaner, A. F. Boix, and S. Zarzuela, On some local cohomology spectral sequences. Int. Math. Res. Not. IMRN 2020 (2020), no. 19, 61976293 Zbl 1486.13024 MR 4165477
, Section 2.2.1]) is that whose open sets UP satisfy the property that if pU and pq, then qU. If P has the Alexandrov topology, we know (see [2
J. Àlvarez Montaner, A. F. Boix, and S. Zarzuela, On some local cohomology spectral sequences. Int. Math. Res. Not. IMRN 2020 (2020), no. 19, 61976293 Zbl 1486.13024 MR 4165477
, Lemma 2.9]) that the basic open set [p,1P^)={qPqp} is contractible for each pP.
For any pP, we will denote by Ip the sum of ideals corresponding to pP. We now fix an ideal J and an additive functor Ψ[]:𝒮RFun(P^,𝒮R) that satisfies the following conditions analogous to [2
J. Àlvarez Montaner, A. F. Boix, and S. Zarzuela, On some local cohomology spectral sequences. Int. Math. Res. Not. IMRN 2020 (2020), no. 19, 61976293 Zbl 1486.13024 MR 4165477
, Setup 4.3].
  1. (1)
    For each pP^, the functor Ψp:=Ψ[](__)(p):𝒮R𝒮R is left exact and preserves direct sums.
  2. (2)
    Let 𝒮R be an R-elementary object. Let ()𝒮R be the injective hull of , and let 𝔭:=Ann(𝒩). Then, for any maximal ideal 𝔪 in R, there are objects 𝒳(,𝔪) and 𝒴(,𝔪)𝒮R such that for pP^, we have
    Ψp(())𝔪={𝒳(,𝔪)if 𝔭𝕎(Ip,J) and 𝔭𝔪,𝒴(,𝔪)if 𝔭𝕎(Ip,J) and 𝔭𝔪,0otherwise.
  3. (3)
    For any pq, the natural transformation ΨpΨq satisfies, for any R-elementary object 𝒮R and maximal ideal 𝔪,
    Ψp(())𝔪 Ψq(())𝔪 ={id𝒳(,𝔪)if Ψp(())𝔪=𝒳(,𝔪)=Ψq(())𝔪,id𝒴(,𝔪)if Ψp(())𝔪=𝒴(,𝔪)=Ψq(())𝔪,0otherwise.
We also set Ψ:=Ψ1P^:𝒮R𝒮R. Moreover, for i0 and any 𝒮R, we let iΨ[]() denote the direct system of derived functors {iΨp()}pP.

Lemma 5.1.

Let 𝒮R be such that the localization 𝔪=0 for every maximal ideal 𝔪R. Then, =0.

Proof.

We consider any fg(). Since localization is exact, we have inclusions 𝔪𝔪=0 and hence 𝔪=0 for each maximal ideal 𝔪. Since is finitely generated, it follows from Proposition 2.10 that Supp()=𝕍(Ann()). Hence, we have 𝕍(Ann())=ϕ, that is, =0. Since 𝒮R is locally noetherian,  is the sum of all its finitely generated subobjects. Hence, =0.  

Lemma 5.2.

Let 𝒮R be an injective object. Then, the complex
C(Ψ[]())Ψ()0
is exact.

Proof.

Since localizations are exact, it follows from Lemma 5.1 that it suffices to show that C(Ψ[]())𝔪Ψ()𝔪0 is exact for each maximal ideal 𝔪. Since 𝒮R is injective, we know from Theorem 2.8 that can be expressed as a direct sum of injective hulls of R-elementary objects. Since Ψ[] and C preserve direct sums, it now suffices to check that C(Ψ[](()))𝔪Ψ(())𝔪0 is exact for each maximal ideal 𝔪, where 𝒮R is an R-elementary object and () its injective hull. We now set 𝔭:=Ann(). The rest of the proof now follows exactly as in [2
J. Àlvarez Montaner, A. F. Boix, and S. Zarzuela, On some local cohomology spectral sequences. Int. Math. Res. Not. IMRN 2020 (2020), no. 19, 61976293 Zbl 1486.13024 MR 4165477
, Lemma 4.4].  

Theorem 5.3.

Let 𝒮 be a strongly locally noetherian Grothendieck category, and let R be a commutative noetherian ring. Then, for any 𝒮R, we have a spectral sequence
E2i,j=𝕃icolimpPjΨ[]()jiΨ().

Proof.

For 𝒮R, we consider an injective resolution 0. We now consider the following bicomplex D, that is concentrated in the “second quadrant.”
(5.1)
Here, we write Dl,k:=Cl(Ψ[](k)) to facilitate cohomological notation. As in [2
J. Àlvarez Montaner, A. F. Boix, and S. Zarzuela, On some local cohomology spectral sequences. Int. Math. Res. Not. IMRN 2020 (2020), no. 19, 61976293 Zbl 1486.13024 MR 4165477
, Theorem 4.6], we can now consider the two spectral sequences associated with the first and second filtrations of this bicomplex. Since the l-th column of this complex is given by Dl,=Cl(Ψ[]()) and P is finite, the columns of this bicomplex vanish for l0 and hence both spectral sequences converge. Further, by Lemma 5.2, the rows of this bicomplex are exact up to the 0-th position, and the cohomologies at the 0-th position are given by Ψ(). It follows that the common abutment of these two spectral sequences is given by the cohomology groups of the complex
(5.2)
0Ψ(0)Ψ(1).
Since 0 is an injective resolution, it is clear that the cohomologies of (5.2) are given by Ψ(). Taking the cohomology of the columns of (5.1), one obtains E1i,j=Ci(jΨ[]()). Now applying the cohomology of the rows of (5.1), we finally obtain the spectral sequence
E2i,j=𝕃icolimpPjΨ[]()jiΨ().
This proves the result.  
In the following three subsections, we will now apply the formalism above to three separate contexts in order to obtain spectral sequences of derived functors. The first of these will be the local cohomology objects in 𝒮R with respect to a pair of ideals (I,J).

5.1. Spectral sequences for ΓI,J:𝒮R𝒮R

We continue with the ideal I=Iα1Iαk and the partially ordered set P consisting of all possible sums of the ideals {Iα1,,Iαk} as defined above. For an ideal JR and pP, we set Ψp(__):=ΓIp,J(__) and Ψ:=ΓI,J.

Lemma 5.4.

Let K, JR be ideals. Then, the functor ΓK,J preserves direct sums.

Proof.

We consider a collection {β}βB of objects in 𝒮R and set :=βBβ. From the definition in (3.1), it is evident that ΓK,J()βBΓK,J(β). On the other hand, we consider 𝒩fg() such that KnAnn(𝒩)+J for n1. Since 𝒩 is finitely generated, we must have 𝒩1r for some finite subcollection {1,,r} of objects from {β}βB. Let ρl:1rl denote the canonical projections for 1lr. We now note that
KnAnn(𝒩)+JAnn(ρl(𝒩))+Jfor n1ρl(𝒩)ΓK,J(l)
for each 1lr. We now note that 𝒩ρ1(𝒩)ρr(𝒩)1r, which gives 𝒩l=1rΓK,J(l)βBΓK,J(β).  

Lemma 5.5.

Let 𝒮R be an R-elementary object, and let 𝔭:=Ann(). Let KR be an ideal. If 𝔭𝕎(K,J), then we have ΓK,J(())=0.

Proof.

Since () is an essential subobject, we see that it follows from Corollary 2.7 that Ass(())=Ass()={𝔭}. We are given 𝔭𝕎(K,J). By Proposition 3.3, we now have Ass(ΓK,J(()))=Ass(())𝕎(K,J)=ϕ. It follows that ΓK,J(())=0.  

Lemma 5.6.

Let 𝒮R be an R-elementary object, and let 𝔭:=Ann(). Let KR be an ideal.
  1. (a)
    Suppose that 𝔭𝕎(K,J). Then, we have ΓK,J(())=().
  2. (b)
    If 𝔭𝔪 for some maximal ideal 𝔪, then ()𝔪=0.

Proof.

(a) Since 𝔭=Ann() and 𝔭𝕎(K,J), we have KnAnn()+J for n1. It follows from the definition in (3.1) that is (K,J)-torsion. By Lemma 3.9, it follows that the injective hull () is also (K,J)-torsion, that is, ΓK,J(())=().
(b) Since is finitely generated, we know that Supp()=𝕍(Ann())=𝕍(𝔭). Accordingly, if 𝔭𝔪, then 𝔪=0. By Proposition 2.14, we know that ()𝔪 is the injective hull of 𝔪, and hence ()𝔪=0.  

Proposition 5.7.

For any 𝒮R, we have a spectral sequence
E2i,j=𝕃icolimpPH[],Jj()HI,Jji().

Proof.

This follows directly from Lemmas 5.45.5 and  5.6 and Theorem 5.3.  

5.2. Generalized local cohomology objects on 𝒮R

Let V be an R-module. By [6
M. Artin and J. J. Zhang, Abstract Hilbert schemes. Algebr. Represent. Theory 4 (2001), no. 4, 305394 Zbl 1030.14003 MR 1863391
, Section B4], we know that RV:𝒮R𝒮R has a right adjoint Hom¯R(V,__). The bifunctor Hom¯R(__,__):(RMod)op×𝒮R𝒮R is left exact in both variables and satisfies Hom¯R(R,)= for any 𝒮R. The right derived functors of Hom¯R(V,__) are denoted by Ext¯R(V,__). If V, W are R-modules, it also follows from the adjunction that Hom¯R(VRW,__)Hom¯R(V,Hom¯R(W,__)).
Given an R-module V and any ideal KR, we denote by VK the inverse system given by VK:={V/KtV}t1. We now set
(5.3)
ΓVI:SRSRlimt1Hom_R(ItV,),
It is immediate from (5.3) that ΓVK is left exact, and we denote by HVK(__) the right derived functors of ΓVK.

Proposition 5.8.

Let V be an R-module, and let KR be an ideal. Then, for any 𝒮R, we have
HVKi()=limt1Ext¯Ri(V/KtV,)  i0.

Proof.

For the sake of convenience, we set Fi():=limt1Ext¯Ri(V/KtV,) for any 𝒮R. From the definition in (5.3), it follows that HVK0()=F0(). The derived functors {HVK(__)} give a family of cohomological δ-functors that is universal. Similarly, for each t1, the derived functors {Ext¯R(V/KtV,__)} are universal δ-functors on 𝒮R. Since filtered colimits in 𝒮R are also exact, we see that {F} is also a family of δ-functors. For any injective 𝒮R, the derived functor Ext¯Ri(V/KtV,) vanishes for i>0, and hence so does Fi(). It follows that {F} is a universal δ-functor and since F0=HVK0, we must have Fi=HVKi for every i.  
In order to understand the functor ΓVK better, we define, for any ideal KR and 𝒮R,
γK():=aKKer(a:)=i=1nKer(ai,:),
where {a1,,an} is a set of generators for K.

Lemma 5.9.

Let KR be an ideal and 𝒮R. Then, γK()=Hom¯R(R/K,).

Proof.

If {a1,,an} is a set of generators for K, we note that R/K=(R/a1R)RR(R/anR) and hence
Hom¯R(R/K,)=Hom¯R(R/a1R,Hom¯R(R/a2R,,Hom¯R(R/anR,))).
Therefore, it suffices to check the result for a principal ideal K=(a). In that case, R𝑎RR/aR0 is a free presentation of R/aR, and it follows from [6
M. Artin and J. J. Zhang, Abstract Hilbert schemes. Algebr. Represent. Theory 4 (2001), no. 4, 305394 Zbl 1030.14003 MR 1863391
, (B4.2)] that
0Hom¯R(R/aR,)Hom¯R(R,)Hom¯R(R,a)=aHom¯R(R,)
is exact. The result is now clear.  
From Lemma 5.9 and the definition in (3.1), it now follows that for any ideal KR, we have
(5.4)
ΓK():=ΓK,0()=t1γKt()=limt1Hom¯R(R/Kt,).
For the rest of this subsection, we suppose that V is a finitely generated R-module. We note here the following fact.

Lemma 5.10.

Let V be a finitely generated R-module. Then, the functor Hom¯R(V,__):𝒮R𝒮R preserves filtered colimits.

Proof.

Let {β}βB be a filtered system of objects in 𝒮R. We will show that
(5.5)
𝒮R(𝒩RV,limβBβ) =𝒮R(𝒩,Hom¯R(V,limβBβ)) =𝒮R(𝒩,limβBHom¯R(V,β))
for any 𝒩𝒮R. Since 𝒮R is locally noetherian, any object of 𝒮R may be expressed as a filtered colimit of its finitely generated subobjects. Accordingly, it suffices to check (5.5) for 𝒩𝒮R finitely generated. If 𝒩 is finitely generated, we have
(5.6)
𝒮R(𝒩,limβBHom¯R(V,β))=limβB𝒮R(𝒩,Hom¯R(V,β))=limβB𝒮R(𝒩RV,β).
From (5.5) and (5.6), we see that it suffices to check
𝒮R(𝒩RV,limβBβ)=limβB𝒮R(𝒩RV,β)
for 𝒩𝒮R finitely generated. But we are given that V is a finitely generated R-module, that is, a quotient of Rn for some n1. Then, 𝒩RV is a quotient of 𝒩n and hence 𝒩RV is finitely generated in 𝒮R. The result is now clear.  

Proposition 5.11.

Let V be a finitely generated R-module and KR an ideal. Then, for any 𝒮R, we have ΓVK()=Hom¯R(V,ΓK()). In particular, ΓVK preserves direct sums.

Proof.

Since V is finitely generated, it follows from (5.3), (5.4) and Lemma 5.10 that
ΓVK() =limt1Hom¯R(V/KtV,)=limt1Hom¯R(V,Hom¯R(R/Kt,)) =Hom¯R(V,limt1Hom¯R(R/Kt,)) =Hom¯R(V,ΓK()).
The last statement is clear from Lemma 5.10 and from the fact that ΓK=ΓK,0 preserves direct sums.  

Lemma 5.12.

Let RR be an extension of k-algebras. For any R-module W and any 𝒩𝒮R, we have an isomorphism
Hom¯R(WRR,𝒩)Hom¯R(W,𝒩).

Proof.

We write W as the cokernel R(Y)R(X)W0 of free modules. Then, WRR is expressed as the cokernel R(Y)R(X)WRR0. By definition (see [6
M. Artin and J. J. Zhang, Abstract Hilbert schemes. Algebr. Represent. Theory 4 (2001), no. 4, 305394 Zbl 1030.14003 MR 1863391
, (B4.2)]), Hom¯R(W,𝒩) is given as the kernel of the induced map 𝒩X=Hom¯R(R(X),𝒩)Hom¯R(R(Y),𝒩)=𝒩Y. Similarly, Hom¯R(WRR,𝒩) is also given by the kernel 𝒩X=Hom¯R(R(X),𝒩)Hom¯R(R(Y),𝒩)=𝒩Y.  

Lemma 5.13.

Let RR[T1] be the localization of R with respect to a multiplicatively closed subset T. Then, for any finitely generated R-module V and any 𝒮R, we have
Hom¯R(V,)THom¯R[T1](VT,T).

Proof.

Applying Lemma 5.12, we know that Hom¯R[T1](VT,T)Hom¯R(V,T). We have noted before that the localization T is given by a filtered colimit. Since V is a finitely generated R-module, we know from Lemma 5.10 that Hom¯R(V,__) preserves filtered colimits. We now have Hom¯R(V,T)Hom¯R(V,)T.  
We now return to the ideal I=Iα1Iαk and the partially ordered set P consisting of all possible sums of the ideals {Iα1,,Iαk} as before. We fix J=0 and a finitely generated R-module V. For pP, we set Ψp(__):=ΓVIp(__) and Ψ:=ΓVI.

Lemma 5.14.

Let 𝒮R be an R-elementary object, and let 𝔭:=Ann(). Let KR be an ideal.
  1. (a)
    If 𝔭𝕎(K,0)=𝕍(K), then we have ΓVK(())=0.
  2. (b)
    Suppose that 𝔭𝕎(K,0)=𝕍(K). Then, we have ΓVK(())=Hom¯R(V,()).

Proof.

(a) Since 𝔭𝕎(K,0)=𝕍(K), we see that it follows from Lemma 5.5 that ΓK(())=ΓK,0(())=0. Further, since V is finitely generated, we have by Proposition 5.11 that ΓVK(())=Hom¯R(V,ΓK(()))=0.
(b) Since 𝔭𝕎(K,0)=𝕍(K), it follows from Lemma 5.6 that ΓK(())=ΓK,0(())=(). Since V is finitely generated, we have by Proposition 5.11 that ΓVK(())=Hom¯R(V,ΓK(()))=Hom¯R(V,()).  

Lemma 5.15.

Let 𝒮R be an R-elementary object, and let 𝔭:=Ann(). Let KR be an ideal. If 𝔭𝔪 for some maximal ideal 𝔪, then ΓVK(())𝔪=0.

Proof.

If 𝔭𝕎(K,0)=𝕍(K), we already have ΓVK(())=0 by Lemma 5.14 (a). Otherwise, suppose 𝔭𝕎(K,0)=𝕍(K). Then, Lemma 5.14 (b) gives us ΓVK(())=Hom¯R(V,()). Because V is finitely generated, applying Lemma 5.13 gives us
ΓVK(())𝔪=Hom¯R𝔪(V𝔪,()𝔪)
for any maximal ideal 𝔪. Since 𝔭𝔪, we have by Lemma 5.6 that ()𝔪=0, which shows that ΓVK(())𝔪=0.  

Proposition 5.16.

Let V be a finitely generated R-module. Then, for any 𝒮R, we have a spectral sequence
E2i,j=𝕃icolimpPHVIpj()HVIji().

Proof.

This follows directly from Lemmas 5.14 and 5.15 and Theorem 5.3.  

5.3. Generalized Nagata ideal transforms on 𝒮R

We let V be a finitely generated R-module and KR be an ideal. We define the generalized Nagata ideal transform ΔVK on 𝒮R as follows:
ΔVK:𝒮R𝒮Rlimt1Hom¯R(KtV,).
It is immediate that the functor ΔVK is left exact. Since V is finitely generated, it is also clear from Lemma 5.10 that ΔVK preserves direct sums. We now need the following result.

Lemma 5.17.

Let 𝒮R be an injective object. Then, the functor Hom¯R(__,):RMod𝒮R is exact.

Proof.

We already know that Hom¯R(__,) is left exact. Let VV be an inclusion of R-modules. We will show that Hom¯R(V,)Hom¯R(V,) is an epimorphism. Accordingly, we set 𝒮R to be the cokernel Hom¯R(V,)Hom¯R(V,)0.
In this section, we have assumed that 𝒮 has enough projectives. We choose an epimorphism 𝒫 in 𝒮 with 𝒫𝒮 projective. This induces an epimorphism 𝒫RR in 𝒮R. Also, we know that composing the structure map R of 𝒮R with the morphism =kR induced by the unit gives the identity in 𝒮. Accordingly, R is an epimorphism in 𝒮R, since the underlying morphism in 𝒮 is an epimorphism. We now have an epimorphism 𝒫RR in 𝒮R.
Since 𝒫𝒮 is projective, the adjoint isomorphism 𝒮R(𝒫R,__)𝒮k(𝒫,__) shows that 𝒫R𝒮R is projective. Hence, the following sequence is exact:
(5.7)
𝒮R(𝒫R,Hom¯R(V,))𝒮R(𝒫R,Hom¯R(V,))𝒮R(𝒫R,)0.
Since k is a field, the inclusion VV induces a monomorphism (𝒫R)RV=𝒫V𝒫V=(𝒫R)RV. Since 𝒮R is injective, it now follows that 𝒮R((𝒫R)RV,)𝒮R((𝒫R)RV,) is an epimorphism. Using (5.7), we now have 𝒮R(𝒫R,)=0, which shows that =0.  
Using Lemma 5.17, we see that if 𝒮R is an injective object, then
0Hom¯R(V/KtV,)Hom¯R(V,)Hom¯R(KtV,)0
is exact for any t1. Taking filtered colimits, we have a short exact sequence
(5.8)
0ΓVK() =limt1Hom¯R(V/KtV,)Hom¯R(V,)limt1Hom¯R(KtV,) =ΔVK()0.
For our setup with I=Iα1Iαk and J=0, we now set Ψp:=ΔVIp(__) for each pP and Ψ:=ΔVI.

Lemma 5.18.

Let 𝒮R be an R-elementary object, and let 𝔭:=Ann(). Let KR be an ideal.
  1. (a)
    If 𝔭𝕎(K,0)=𝕍(K), then we have ΔVK(())=0.
  2. (b)
    If 𝔭𝕎(K,0)=𝕍(K), we have ΔVK(())=Hom¯R(V,()).

Proof.

(a) If 𝔭𝕎(K,0)=𝕍(K), it follows from Lemma 5.14 (b) that
ΓVK(())=Hom¯R(V,())
Accordingly, the short exact sequence in (5.8) gives us ΔVK(())=0.
(b) If 𝔭𝕎(K,0)=𝕍(K), it follows from Lemma 5.14 (a) that ΓVK(())= 0. Accordingly, the short exact sequence in (5.8) gives us ΔVK(())=Hom¯R(V,()). This proves the result.  

Lemma 5.19.

Let 𝒮R be an R-elementary object, and let 𝔭:=Ann(). Let KR be an ideal. If 𝔭𝔪 for some maximal ideal 𝔪, then ΔVK(())𝔪=0.

Proof.

If 𝔭𝕎(K,0), we already know from Lemma 5.18 (a) that ΔVK(())= 0. Otherwise, suppose that 𝔭𝕎(K,0), so that it follows from Lemma 5.18 (b) that ΔVK(())=Hom¯R(V,()). Because V is finitely generated, applying Lemma 5.13 gives us ΔVK(())𝔪=Hom¯R𝔪(V𝔪,()𝔪) for any maximal ideal 𝔪. Since 𝔭𝔪, we have by Lemma 5.6 (b) that ()𝔪=0, which shows that ΔVK(())𝔪=0.  

Proposition 5.20.

Let V be a finitely generated R-module. For any 𝒮R, we have a spectral sequence
E2i,j=𝕃icolimpPjΔVIp()jiΔVI().

Proof.

This follows directly from Lemmas 5.18 and 5.19 and Theorem 5.3.  

References

  1. [1] J. Adámek and J. Rosický, Locally presentable and accessible categories. London Math. Soc. Lecture Note Ser. 189, Cambridge University Press, Cambridge, 1994, 316 pp. Zbl 0795.18007 MR 1294136
  2. [2] J. Àlvarez Montaner, A. F. Boix, and S. Zarzuela, On some local cohomology spectral sequences. Int. Math. Res. Not. IMRN 2020 (2020), no. 19, 61976293 Zbl 1486.13024 MR 4165477
  3. [3] J. Àlvarez Montaner, R. García López, and S. Zarzuela Armengou, Local cohomology, arrangements of subspaces and monomial ideals. Adv. Math. 174 (2003), no. 1, 3556 Zbl 1050.13009 MR 1959890
  4. [4] M. Artin, L. W. Small, and J. J. Zhang, Generic flatness for strongly Noetherian algebras. J. Algebra 221 (1999), no. 2, 579610 Zbl 0958.16024 MR 1728399
  5. [5] M. Artin and J. J. Zhang, Noncommutative projective schemes. Adv. Math. 109 (1994), no. 2, 228287 Zbl 0833.14002 MR 1304753
  6. [6] M. Artin and J. J. Zhang, Abstract Hilbert schemes. Algebr. Represent. Theory 4 (2001), no. 4, 305394 Zbl 1030.14003 MR 1863391
  7. [7] A. Banerjee, An extension of the Beauville–Laszlo descent theorem. Arch. Math. (Basel) 120 (2023), no. 6, 595604 Zbl 1527.13010 MR 4598542
  8. [8] A. Beligiannis and I. Reiten, Homological and homotopical aspects of torsion theories. Mem. Amer. Math. Soc. 188 (2007), no. 883, 207 pp. Zbl 1124.18005 MR 2327478
  9. [9] M. Brodmann, Asymptotic stability of Ass(M/InM). Proc. Amer. Math. Soc. 74 (1979), no. 1, 1618 Zbl 0395.13008 MR 0521865
  10. [10] M. Brodmann and M. Hellus, Cohomological patterns of coherent sheaves over projective schemes. J. Pure Appl. Algebra 172 (2002), no. 2–3, 165182 Zbl 1011.13009 MR 1906872
  11. [11] M. Brodmann, Ch. Rotthaus, and R. Y. Sharp, On annihilators and associated primes of local cohomology modules. J. Pure Appl. Algebra 153 (2000), no. 3, 197227 Zbl 0968.13010 MR 1783166
  12. [12] M. P. Brodmann and A. Lashgari Faghani, A finiteness result for associated primes of local cohomology modules. Proc. Amer. Math. Soc. 128 (2000), no. 10, 28512853 Zbl 0955.13007 MR 1664309
  13. [13] M. P. Brodmann and R. Y. Sharp, Local cohomology: an algebraic introduction with geometric applications, 2nd edn., Cambridge Stud. Adv. Math. 136, Cambridge University Press, Cambridge, 2013, 491 pp. Zbl 1263.13014 MR 3014449
  14. [14] S. Dăscălescu, C. Năstăsescu, and Ş. Raianu, Hopf algebras: an introduction. Monographs and Textbooks in Pure Appl. Math. 235, Marcel Dekker, New York, 2001, 401 pp. Zbl 0962.16026 MR 1786197
  15. [15] K. Divaani-Aazar and R. Sazeedeh, Cofiniteness of generalized local cohomology modules. Colloq. Math. 99 (2004), no. 2, 283290 Zbl 1072.13011 MR 2079733
  16. [16] G. Faltings, Der Endlichkeitssatz in der lokalen Kohomologie. Math. Ann. 255 (1981), no. 1, 4556 Zbl 0451.13008 MR 0611272
  17. [17] M. Hellus, On the set of associated primes of a local cohomology module. J. Algebra 237 (2001), no. 1, 406419 Zbl 1027.13009 MR 1813886
  18. [18] C. Huneke, D. Katz, and T. Marley, On the support of local cohomology. J. Algebra 322 (2009), no. 9, 31943211 Zbl 1186.13016 MR 2567416
  19. [19] W. Lowen and M. Van den Bergh, Deformation theory of abelian categories. Trans. Amer. Math. Soc. 358 (2006), no. 12, 54415483 Zbl 1113.13009 MR 2238922
  20. [20] A. Neeman, Triangulated categories. Ann. of Math. Stud. 148, Princeton University Press, Princeton, NJ, 2001, 449 pp. Zbl 0974.18008 MR 1812507
  21. [21] N. Popescu, Abelian categories with applications to rings and modules. London Math. Soc. Monogr. 3, Academic Press, London-New York, 1973, 467 pp. Zbl 0271.18006 MR 0340375
  22. [22] C. Rotthaus and L. M. Şega, Some properties of graded local cohomology modules. J. Algebra 283 (2005), no. 1, 232247 Zbl 1105.13021 MR 2102081
  23. [23] B. Stenström, Rings of quotients: an introduction to methods of ring theory. Grundlehren Math. Wiss. 217, Springer, New York-Heidelberg, 1975, 309 pp. Zbl 0296.16001 MR 0389953
  24. [24] R. Takahashi, Y. Yoshino, and T. Yoshizawa, Local cohomology based on a nonclosed support defined by a pair of ideals. J. Pure Appl. Algebra 213 (2009), no. 4, 582600 Zbl 1160.13013 MR 2483839
  25. [25] C. A. Weibel, An introduction to homological algebra. Cambridge Stud. Adv. Math. 38, Cambridge University Press, Cambridge, 1994, 450 pp. Zbl 0797.18001 MR 1269324

Cite this article

Abhishek Banerjee, Surjeet Kour, Noncommutative supports, local cohomology and spectral sequences. J. Noncommut. Geom. 20 (2026), no. 4, pp. 1197–1224

DOI 10.4171/JNCG/693