• 検索結果がありません。

On Cohen-Macaulay rings

N/A
N/A
Protected

Academic year: 2022

シェア "On Cohen-Macaulay rings"

Copied!
8
0
0

読み込み中.... (全文を見る)

全文

(1)

On Cohen-Macaulay rings

Edgar E. Enochs, Overtoun M.G. Jenda

Abstract. In this paper, we use a characterization ofR-modulesN such thatf dRN = pdRN to characterize Cohen-Macaulay rings in terms of various dimensions. This is done by setting N to be the dth local cohomology functor of R with respect to the maximal ideal wheredis the Krull dimension ofR.

Keywords: injective, precovers, preenvelopes, canonical module, Cohen-Macaulay, n-Gorenstein, resolvent, resolutions

Classification: 13C14, 13D45, 13H10, 18G10

1. Introduction

Rwill denote an associative ring with a unit element,R-module will mean left R-module, and noetherian will mean left noetherian.

LetAand Bbe subcategories of R-modules. Then we recall that if A andB are objects inAandBrespectively, then theA-injective dimensionofB (denoted A−idB) or B-projective dimension of A (denoted B −pdA) is the smallest nonnegative integernsuch that ExtiR(A, B) = 0 for alli > n. Otherwise, we set A−idB=B−pdA=∞.

We defineA−idBto be the sup{A−idB :B∈ B}. Note thatA−idB=B−pdA.

Similarly,A −idB =B−pdAcan be defined. IfA −idB = 0, we will say that B isA-injective. We define theA-injective dimensionofB(denoted byA −idB) to be sup{A−idB:A∈ A, B∈ B}.

Likewise, ifA is a subcategory of right R-modules and B is a subcategory of leftR-modules, then we can defineA −f dB andA−f dBusing TorRi (A, B). B isA-flatifA −f dB= 0.

Now letM be anR-module andAbe a full subcategory ofR-modules. Then a mapψ:M →AwithAinAis said to be anA-preenvelopeofM if any diagram

M ψ - A

? . .. .. .. .. A

withA in Acan be completed.

(2)

If A contains all injective R-modules, then the preenvelopes are monomor- phisms. So that in the caseA-preenvelopes exist for allR-modules, we can define anA-resolutionofM to be an exact sequence

0→M →A→A1→. . . where

M → A, Coker(M → A) → A1, Coker(An−1 → An) → An+1 for n ≥ 1 are A-preenvelopes. We will say that M has A-resolution dimension (denoted A −rndimM)≤nif there is anA-resolution 0→M →A→A1 → · · · →An→ 0. TheA-resolution global dimensionofR (denoted by A −rngldimR) is to be the sup{A −rndimM:M ∈Mod}whereModis the category ofR-modules.

If the preenvelopes are not necessarily exact, we get a sequence, not necessarily exact, called an A-resolvent ofM, and soA-resolvent dimension (denoted A − rtdim) andA −rtgldimRcan be defined similarly.

We start in Section 2 by extending the results in Enochs-Jenda [3] to an arbi- trary ringR. In particular, we get a characterization of R-modulesN such that f dRN =pdRN in terms of the various dimensions defined above (Theorem 2.1).

By choosing an appropriate N, this theorem specializes to n-Gorenstein rings (Corollary 2.3) and Cohen-Macaulay local rings (Theorem 3.7).

If (R, m, k) is a commutative noetherian local ring of Krull dimensiondandM is a finitely generatedR-module, thenHmi (M) denotes theith local cohomology functor with respect to the maximal idealm. A finitely generatedR-moduleK is said to be acanonical moduleofRif the completion ofKRwith respect to the m-adic topology

R∼= Hom(Hmd(R), E(k))

whereE(k) denotes the injective envelope ofk(see Herzog-Kunz [7]).

A finitely generated R-module M is said to be maximal Cohen-Macaulay if depth M = d. If R is a Cohen-Macaulay ring with a canonical module, then every finitely generated R-moduleM has a maximal Cohen-Macaulay precover (see Auslander-Buchweitz [1] or Yoshino [13]), that is, a surjective mapψ:C→M withC maximal Cohen-Macaulay such that any diagram

C

. .. .. .. ..

?

C - M

withC maximal Cohen-Macaulay can be completed.

We can therefore form aCohen-Macaulayresolution

· · · →C1→C0→M →0 where

C0 →M, C1 →Ker(C0 →M), Cn+1→Ker(Cn→Cn1), n≥1

(3)

are maximal Cohen-Macaulay precovers. If there is a Cohen-Macaulay resolution 0 →Cn →Cn1 → · · · → C1 →M →0, we say that M has Cohen-Macaulay dimension (denotedCM −dim) ≤ n. We define the Cohen-Macaulay global dimension of R (denoted CM −gldimR) to be the sup{CM −dimM : M ∈ F GMod} where F GMod denotes the full subcategory of finitely generated R- modules.

The aim of Section 3 is to give characterizations of Cohen-Macaulay rings in terms of the local cohomology functor and the various dimensions that we have defined above. One of the consequences of this section is that the length of a Cohen-Macaulay resolution of a finitely generatedR-module does not exceed the Krull dimension ofR whenRhas a canonical module.

In this paper, Exti(A, B), Tori(A, B) will denote ExtiR(A, B), TorRi (A, B) respectively, and for a local ring (R, m, k), the Matlis dual Hom(M, E(k)) will be denoted byMv where E(k) is the injective envelope ofk.

2. Resolutions and resolvents

LetN be a fixedR-module. ThenAN will denote the full subcategory of all N-injectiveR-modules andBN will denote the full subcategory of allN-flat right R-modules.

In [3], we showed the existence of copure injective preenvelopes over noether- ian rings, and copure flat preenvelopes over commutative artinian rings. For an arbitrary ring R, the same proofs show the existence of AN-preenvelopes, and BN-preenvelopes in the caseN is of finite type for then Tori(−, N) preserves di- rect products by Lenzing [10]. So straight forward modifications to the proofs of the results in Section 3 and Theorem 4.1 of [3] give the following result which holds for any ringR.

Theorem 2.1. LetNbe anR-module such thatf dN=pdN. Then the following are equivalent for an integern.

(1) pdN ≤n.

(2) AN−rngldimR≤n.

(3) Everynth cosyzygy of anR-module is inAN. (4) N−f dModR≤n.

(5) N−f dF GModR≤n.

(6) Everynth syzygy of a rightR-module is inBN.

Furthermore, ifNis of finite type, then each of the above statements is equivalent to (7) BN−rtgldimR≤n.

To see that Theorem 4.1 of [3] for n-Gorenstein rings (that is, R is left and right noetherian and is of finite injective dimension at mostnover itself on either side) is a consequence of the above theorem, one observes the following:

Proposition 2.2. Let R be noetherian and {Xα} be a representative set of indecomposable injective R-modules. Set X = ⊕Xα. Then the following are

(4)

equivalent for an integern.

(1) idRR=n.

(2) pdRX =n.

(3) f dRX =n.

Proof: 1⇔2. SupposeidRR=n. ThenpdRX≤nby Jensen [9, Theorem 5.9].

IfpdRX < n, thenpdRE < nfor all injectiveR-modulesEsinceE=⊕Xβ where Xβ ∈ {Xα}, and soidRR< nby Jensen [9]. SopdRX=n, and conversely.

1⇔3. idRR =n implies thatf dRX ≤n by Enochs-Jenda [4, Theorem 4.4]

and sof dRX =nas above, and conversely.

Now we simply note thatAX-injective dimension is the copure injective dimen- sion (cid),X-flat dimension is the copure flat dimension (cf d), andBX-resolvent dimension is the copure flat resolvent dimension. Furthermore,Risn-Gorenstein if and only if pdRX ≤ n and pdXR ≤n. So if we set N = X in Theorem 2.1 above, we get the following result using Proposition 2.2 above.

Corollary 2.3 ([3, Theorem 4.1]). The following are equivalent for a left and right noetherian ringR.

(1) R isn-Gorenstein.

(2) cidM ≤nfor allR-modules(left and right)M.

(3) Everynthcosyzygy of anR-modules(left and right)is inAX. (4) cf dM ≤nfor allR-modules(left and right)M.

(5) cf dM ≤nfor all finitely generatedR-modules(left and right)M. (6) Everynthsyzygy of anR-module(left and right)is inBX.

Furthermore, ifR is commutative artinian, then each of the above statements is equivalent to

(7) Copure flat resolvent dimension of eachR-module is at mostn.

Remark. We note that if R is a commutative artinian ring, then Rp is quasi- Frobenius for each prime ideal P of R. Therefore R is quasi-Frobenius and so n= 0 in this case.

3. Local rings

Throughout this section, R will denote a commutative noetherian local ring with maximal idealmand residue fieldk.

We start with the following.

Lemma 3.1. The following are equivalent for a ringR and integerd≥1 (1) R is Cohen-Macaulay of dimensiond.

(2) f dHmd(R) =d.

(3) pdHmd(R) =d.

Proof: 1⇒2. See Strooker [12, Proposition 9.1.4].

(5)

2⇒1. Hmd(R) is artinian and so is an ˆR-module naturally. Sof dHmd(R) =d implies that f dRˆHmd(R) = dand thus idRˆHmd(R)v = d. Therefore, Hmd(R)v is a noetherian ˆR-module of finite injective dimension. Thus ˆR is Cohen-Macaulay (see Strooker [12, Theorem 13.1.7]) and soR is Cohen-Macaulay. Furthermore, the dimension isdfor otherwiseHmd(R) = 0.

2 ⇒ 3. f dHmd(R) = d implies that Krull dimR = d by the above. So pdHmd(R)≤d by Foxby [5, Corollary 3.4]. Sod =f dHmd(R) ≤pdHmd(R)≤d.

Thuspddm(R) =d.

3⇒2 is trivial sincef d≤pd.

Ford= 0, we have the following which is surely known and we present it here for completeness.

Lemma 3.2. The following are equivalent for a ringR.

(1) R is artinian.

(2) Hm0(R) =R.

(3) Hm0(R)6= 0andHm0(R)is flat.

(4) Hm0(R)6= 0andHm0(R)is projective.

Proof: 1⇒2,3,4.

Hm0(R)v = Hom(lim Hom(R/mt, R), E(k))

= lim

Hom(Hom(R/mt, R), E(k))

= lim

R/mt⊗E(k)

=E(k)

sinceR is complete. Thus Hm0(R) is nonzero and flat. But then Hm0(R) is free and soHm0(R) =R.

2⇒1 Hm0(R)v =E(k) is noetherian and soR is artinian.

3⇒1 follows as in Lemma 3.1 and 4⇒3 is trivial.

Corollary 3.3. Ris Gorenstein if and only if

f dRpE(k(P)) =pdRpE(k(P)) =htP for allP∈SpecR wherek(P)is the quotient ring ofR/p.

Proof: We first recall that R-Gorenstein means that idRp < ∞ for all P ∈ SpecR (see Bass [2]). If Rp has finite injective dimension, then HmRd

p(RP) = E(k(P)) whered= KrulldimRp =htP. So the result follows from the Lemmas above. Conversely, if f dRpE(k(P)) = htP, then idRˆp = htP < ∞, and so

idRp<∞.

Now let I be the full subcategory of finitely generated R-modules with finite injective dimension. We state the following, noting that ifRis Cohen-Macaulay, thenI 6= 0.

(6)

Lemma 3.4. LetR be Cohen-Macaulay. Then the following are equivalent for a finitely generatedR-moduleM.

(1) M is a maximal Cohen-MacaulayR-module.

(2) EveryR-module inI isM-injective.

(3) I has a nonzeroM-injective R-module.

Furthermore, ifRhas a canonical module K, then each of the above statements is equivalent to

(4) K isM-injective.

(5) ˆK isM-injective.

Proof: 1⇔2. We recall thatidI=depthRfor eachI∈ I, I6= 0. Furthermore, depthM+M−idI=idI (see Roberts [11]). So the result follows.

Similarly (3) implies (1), and (3) follows from (2) trivially.

1⇔4. We use the local duality Exti(M, K)⊗RRˆ ∼= HomR(Hmd−i(M), E(k)) (see Yoshino [13, Proposition 1.12] or Grothendieck [6, Theorem 6.3]). M is maximal Cohen-Macaulay if and only ifHmdi(M) = 0 for alli >0 and so if and only if Exti(M, K) = 0 fori >0.

4⇔5. We simply note that Exti(M, K)⊗RRˆ∼= Exti(M,KˆR) by Ishikawa [8,

Corollary 1.2], and so the result follows.

Now letCbe the full subcategory ofF G Modconsisting of all maximal Cohen- Macaulay R-modules, and I be the full subcategory of F G Mod consisting of allC-injective R-modules. It follows from Lemma 3.4 above that if R is Cohen- Macaulay, thenI is a full subcategory ofI.

If I ∈ I and 0 →I → E → E → · · · is an injective resolution of I, then 0→Hom(C, I)→Hom(C, E)→Hom(C, E)→ · · · is exact for allC inC. Fur- thermore, if· · · →C1→C0→M →0 is a Cohen-Macaulay resolution of a finitely generatedR-moduleM, then 0→Hom(M, E)→Hom(C0, E)→ · · · is exact for each injectiveE. So Hom(−,−) is right balanced by (C, Inj) onF GMod× I (see Enochs-Jenda [4]). So we obtain right derived functors Exti(M, I). We note that Exti(M, I) = Exti(M, I).

We are now in a position to prove the following.

Theorem 3.5. The following are equivalent for a ringRwith a canonical module.

(1) R is Cohen-Macaulay of dimensiond.

(2) Every finitely generated R-module has a maximal Cohen-Macaulay pre- cover andCM−gldimR=d.

(3) I 6= 0andsupI∈I{idI}=d.

Proof: 1 ⇒ 2. The first part was mentioned in Section 1. Now let K be the canonical module. ThenidKˆR=dby Lemmas 3.1 and 3.2 since ˆKR∼=Hmd(R)v. But then idKR = d since ˆR is faithfully flat. Now consider a Cohen-Macaulay resolution · · · →C1 →C0 →M →0 of a finitely generatedR-moduleM. Let

(7)

0 → Ti+1 → Ci → Ti → 0 wherei ≥n be the short exact sequence. Then we have Ext1(Ti, K)∼= Ext2(Ti1, K)∼=· · · ∼= Exti+1(M, K) since Exti(C, K) = 0 for i > 0 for all maximal Cohen-Macaulay R-modules C by Lemma 3.4. But Exti+1(M, K) = 0 for alli≥dsinceidK =d. So Ext1(Ti, K) = 0 for alli≥d.

But we also have that Extj(Td, K)∼= Extj1(Td+1, K)∼=· · · ∼= Ext1(Td+j1, K) for j ≥ 1. So Extj(Td, K) = 0 for all j ≥ 1. Therefore, Td is maximal Cohen- Macaulay, again by Lemma 3.4. ThusCM−gldimR≤d.

SupposeCM−gldimR=n < d. Then Exti(M, I) = 0 for alli > nand for all M ∈ F GMod, I ∈ I. ButK∈ I. So Exti(M, K) = Exti(M, K) = 0 for alli > n and for allM ∈ F GMod. ThusidK≤n < d, a contradiction.

2⇒3. LetC →R →0 be a maximal Cohen-Macaulay precover. ThenR is a direct summand ofC. So depthC≤ depthR. But depthC= dimR. So Ris Cohen-Macaulay. ThusI 6= 0.

CM−gldimR=dimplies that Exti(M, I) = 0 for alli > dfor allF GModM, I∈ I. So if I∈ I, thenidI ≤d. Thus supI∈I{idI} ≤d. If it were less than d, then it is easy to see thatCM−gldimR < d.

3 ⇒ 1 I 6= 0 means R is Cohen-Macaulay. So let Krull dimR = n. Then supI∈I{idI}=nsince 1⇒3. So KrulldimR=d.

Remark. It follows from part (3) of the theorem above that if R is a Cohen- Macaulay ring with a canonical module, thenI =I.

Corollary 3.6(Auslander-Buchweitz [1]). LetRbe a Cohen-Macaulay ring with a canonical module. Then inF GMod, the full subcategoriesC andI are orthog- onal. In particular,C=(C)andI = (I).

Proof: C −idI = 0 by Lemma 3.4. Furthermore, if C is a finitely generated R-module such thatC−idI= 0, thenC∈ C, by the same lemma. SoC consists precisely of allR-modulesC in F GModsuch that C−idI= 0. SoC =I. But by the preceding remark,I consists of precisely ofR-modulesI inF GModsuch thatC −idI= 0. SoI=C. ThusCandI are orthogonal. SoC=I =(C)

andI= (I).

We now finally have the following version of Theorem 2.1 for Cohen-Macaulay rings.

Theorem 3.7. The following are equivalent for a ringRand for an integerd≥1.

(1) R is Cohen-Macaulay of dimensiond.

(2) Hmd(R)−idMod=d.

(3) AHd

m(R)−rngldimR=d.

(4) Hmd(R)−f dMod=d.

(5) Hmd(R)−f dF GMod=d.

(8)

Furthermore, ifR has a canonical module, then each of the above statements is equivalent to

(6) Every finitely generated R-module has a maximal Cohen-Macaulay pre- cover andCM−gldimR=d.

Proof: The equivalence of 1 to 5 follows from Theorem 2.1 and Lemma 3.1 above.

1⇔6 is part of Theorem 3.5.

For d = 0, we have the following which easily follows from Lemma 3.2 and Theorem 3.5.

Proposition 3.8. The following are equivalent for a ringR.

(1) R is artinian.

(2) Hm0(R)6= 0and everyR-module isHm0(R)-flat.

(3) Hm0(R)6= 0and everyR-module isHm0(R)-injective.

(4) Every finitely generatedR-module is maximal Cohen-Macaulay.

References

[1] Auslander M., Buchweitz R.,The homological theory of maximal Cohen-Macaulay appro- ximations, Soc. Math. de France, Memoire38(1989), 5–37.

[2] Bass H.,On the ubiquity of Gorenstein rings, Math. Z.82(1963), 8–28.

[3] Enochs E., Jenda O.,Copure injective resolutions, flat resolvents and dimensions, Com- ment. Math. Univ. Carolinae34(1993), 202–211.

[4] ,Balanced functors applied to modules, J. Algebra92(1985), 303–310.

[5] Foxby H.-B.,Isomorphisms between complexes with applications to the homological theory of modules, Math. Scand.40(1977), 5–19.

[6] Grothendieck A.,Local Cohomology, Lecture Notes in Mathematics 41, Springer, 1967.

[7] Herzog J., Kunz E.,Der Kanonische Modul eines Cohen-Macaulay-Rings, Lecture Notes in Mathematics 238, Springer, 1971.

[8] Ishikawa T.,On injective modules and flat modules, Math. Soc. Japan17(1965), 291–296.

[9] Jensen C.,Les foncteurs d´eriv´ees delim et leurs applications en th´eorie des modules, Lecture Notes in Mathematics 254, Springer, 1972.

[10] Lenzing H.,Endlich pr¨asentierbare Moduln, Arch Math.20(1969), 262–266.

[11] Roberts P.,Homological invariants of modules over commutative rings, Semin. Math. Su- per. 15, Presses Univ. Montreal, 1980.

[12] Strooker J.,Homological questions in local algebra, London Math. Soc. Lecture Note Series 145, Cambridge Univ. Press, 1990.

[13] Yoshino Y., Cohen-Macaulay modules over Cohen-Macaulay rings, London Math. Soc.

Lecture Note Series 146, Cambridge Univ. Press, 1990.

Department of Mathematics, University of Kentucky, Lexington KY 40506-0027, USA

Department of Algebra, Combinatorics, and Analysis, Auburn University, AL 36849-5307, USA

(Received August 18, 1993)

参照

関連したドキュメント

In the same sprit, we define and study two new types of differential rings which lead to the mirrors of the corresponding results on von Neumann regular rings and principally flat

The interaction of this relation with the Cohen–Macaulay condition then leads us to study piecewise-linear Cohen–Macaulay polyhedral complexes of dimension d in Euclidean space R

Now we characterize module Jordan left derivations in full and upper triangular matrix rings over unital associative rings..

In this paper, left derivations and Jordan left derivations in full and upper triangular matrix rings over unital associative rings are characterized..

The author would like to thank the referee for his helpful advice on QF rings and general weight functions and other helpful suggestions.. The author would like to thank

We believe that our present approach is easier to understand and gives more chance to find an explicit form of the Cayley-Hamilton equation (Newton formulae) for n ≥ 3.

in [5], where the case of cohomologically trivial modules is treated, and in [15], where sums of this kind occur as well, when studying the distribution of p-class groups of

0.1. Additive Galois modules and especially the ring of integers of local fields are considered from different viewpoints. Leopoldt [L] the ring of integers is studied as a module