Some homological properties of ideals with cohomological dimension one

See [Me, Theorem 5.5] and [AM, Theorem 2.9].

(i)\Rightarrow(ii) Is clear.

(ii)\Rightarrow(i) Let T:=D(M), where D(-):= {\rm Hom}_{R}(-,E) denotes the Matlis dual functor and E := E_R(R/\frak{m}) is the injective hull of the residue field R/\frak{m}. Then by the adjointness we have \operatorname{Ext}^i_R(R/I,T)\simeq D(\operatorname{Tor}^R_i(R/I,M)), for all 0\leq i\leq \operatorname{cd}(I,R). In particular, the R-modules \operatorname{Ext}^i_R(R/I,T) are Artinian for all 0 \leq i \leq \operatorname{cd}(I,R). So, it follows from Lemma 2.1 and adjointness, that the R-modules \operatorname{Ext}^i_R(R/I,T)\simeq D(\operatorname{Tor}^R_i(R/I,M)) are Artinian for all i\geq 0. Now, it follows from [KLW, Lemma 1.15(a)] that, the R-modules \operatorname{Tor}^R_i(R/I,M) are finitely generated for all i\geq 0, as required.

The assertion follows from Proposition 2.2 and [Me, Theorem 2.1].

Let M,N\in \mathscr{C}(R, I)_{cof} and let f:M\longrightarrow N be an R-homomorphism. It is enough to show that the R-modules \ker f and {\rm coker} f are I-cofinite.

To this end, the exact sequence

induces an exact sequence

which using [Me, Theorem 2.1], implies that the R-module \operatorname{Tor}^R_0(R/I,{\rm im}f) is finitely generated. Now, the exact sequence

induces an exact sequence

By [Me, Theorem 2.1] the modules \operatorname{Tor}^R_1(R/I,N) and \operatorname{Tor}^R_0(R/I,N) are finitely generated R-modules, which implies that the R-modules \operatorname{Tor}^R_0(R/I,{\rm coker}f) and \operatorname{Tor}^R_1(R/I,{\rm coker}f) are finitely generated. Therefore, it follows from Corollary 2.3, that the R-module {\rm coker}f is I-cofinite. Now, the assertion follows from the following exact sequences

and

The assertion follows from Theorem 2.4.

Since N is finitely generated it follows that, N has a free resolution with finitely generated free R-modules. Now the assertion follows using Corollary 2.5 and computing the R-modules {\rm Tor}_i^R(N,M) and {\rm Ext}^i_R(N,M), using this free resolution.

Let I=(x_1,\dots,x_n) and let

be the Koszul complex of M with respect to \underline{x}=x_1,\dots,x_n.

We prove that \operatorname{H}_{i}(\underline{x}; M)=0 for all 0\leq i \leq n. By the definition we have

So, we have \operatorname{im} f_{n-1}=M. Therefore, using the hypothesis it follows that

Hence, the exact sequence

implies that \operatorname{Tor}^R_i(R/I,\operatorname{ker} f_{n-1})=0, for each i\geq 0. The exact sequence

induces the exact sequence

Now as R/I\otimes_{R}\operatorname{ker} f_{n-1} =0, it follows from (3.1.2) that

By the definition of the Koszul complex we have I\operatorname{H}_1(\underline{x};M)=0. Therefore, we have \operatorname{H}_1(\underline{x};M)\simeq R/I\otimes_{R} \operatorname{H}_1(\underline{x};M)=0. Now it follows from the exact sequence (3.1.1) that \operatorname{Tor}^R_i(R/I,\operatorname{im} f_{n-2})=0, for each i\geq 0. Moreover, the exact sequence

implies that \operatorname{Tor}^R_i(R/I,\operatorname{ker} f_{n-2})=0, for each i\geq 0. Proceeding in the same way we can see \operatorname{H}_i(\underline{x};M)=0, for all 0\leq i \leq n.

Now, since \operatorname{H}_i(\underline{x};M)\simeq \operatorname{H}^{n-i}(\underline{x};M) for all 0\leq i \leq n, it follows that

(i)\Rightarrow(ii) We argue by induction on n. For n=0, the assertion holds by Lemma 3.1. We therefore assume, inductively, that n>0 and the result has been proved for smaller values of n. Then there is an exact sequence

Since, by the hypothesis we have (0:_MI)=0 it follows that (0:_{E_R(M)}I)=0 and hence \Gamma_I(E_R(M))=0. By [Mat, Theorem 18.5] the R-module E_R(M) is isomorph with a direct sum of a family of indecomposable injective R-modules. Let \operatorname{\frak p} be a prime ideal of R such that E_R(R/\operatorname{\frak p}) is a direct summand of E_R(M). Then we have \Gamma_I(E_R(R/\operatorname{\frak p}))=0. Therefore, form the fact that \operatorname{Ass}_R E_R(R/\operatorname{\frak p})=\{\operatorname{\frak p}\} it follows that I\not\subseteq \operatorname{\frak p}. So there exists an element a\in I such that a\not\in\operatorname{\frak p}. Then by [Mat, Theorem 18.4(iii)], multiplication by a is an automorphism on E_R(R/\operatorname{\frak p}). Therefore, multiplication by a is an automorphism on \operatorname{Tor}^R_i(R/I,E_R(R/\operatorname{\frak p})), for all i\geq 0. But, since a\in I it follows that, multiplication by a on \operatorname{Tor}^R_i(R/I,E_R(R/\operatorname{\frak p})) is the zero map, for all i\geq 0. Thus, \operatorname{Tor}^R_i(R/I,E_R(R/\operatorname{\frak p}))=0 for all i\geq 0. Since for each i\geq 0, the torsion functor \operatorname{Tor}^R_i(R/I,-) commutes with the direct sums it follows that \operatorname{Tor}^R_i(R/I,E_R(M))=0, for all i\geq 0. So, from the exact sequence (3.2.1) it follows that \operatorname{Tor}^R_i(R/I,E_R(M)/M)=0, for all i\geq 0. Hence, by applying the inductive hypothesis to the R-module E_R(M)/M we have

So, we have \operatorname{Ext}^n_{R}(R/I,M)=0. This completes the inductive step.

In view of [BS, Proposition 2.1.4] the R-module E_1:=\Gamma_I(E) is injective. Therefore, there exists an injective submodule E_2 of E such that E_1+E_2=E and E_1\cap E_2=0. Since \frac{E_1+K}{K} is a submodule of \Gamma_I(L) it follows that the R-module \frac{L}{\Gamma_I(L)} is a homomorphic image of the R-module E_2. So, in order to prove the assertion it is enough to prove that E_2\otimes_R \frac{R}{I}=0. So, we must prove that E_2=I E_2. Since \Gamma_I(R)E_2\subseteq \Gamma_I(E_2)=0, it follows that \Gamma_I(R)E_2=0. Therefore, we have

On the other hand, by [BS, Lemma 2.1.1] there is an exact sequence

for some element a\in I, which effecting the R-linear exact functor \operatorname{Hom}_R(-,E_2) induces the exact sequence

Therefore, we have \operatorname{Hom}_R(R/\Gamma_I(R),E_2)=a \operatorname{Hom}_R(R/\Gamma_I(R),E_2). Hence, we have

So, we have E_2=aE_2 and hence E_2=IE_2.

Let

be a minimal injective resolution for M. Set K_i:={\rm ker} f_i for i\geq0. By splitting this minimal injective resolution to some short exact sequences we get the isomorphisms

Set N:=K_{t-1}/\Gamma_I(K_{t-1}). Then we have \Gamma_I(N)=0 and

So, there exists a submodule K of E:=E_{t-2} such that N\simeq E/K. By [BS, Remark 2.2.7] there is an exact sequence

For each i\geq 2 we have

Therefore, by the hypothesis we have

for all i\geq 2 and all j\geq0. Thus, by Proposition 3.2 we have \operatorname{Ext}^j_R(R/J,H^i_I(N))=0 for all i\geq 2 and all j\geq0. The exact sequence (3.4.1) yields the isomorphisms

for i\geq 2. So, we have

for all i\geq 2 and all j\geq0. Also, by [BS, Corollary 2.2.8] we have H^i_I(D_I(N))=0, for i=0,1. So \operatorname{Ext}^j_R(R/J,H^i_I(D_I(N)))=0 for all i\geq 0 and all j\geq0. Therefore, by [GBA, Lemma 2.1] we have \operatorname{Ext}^j_R(R/J,D_I(N))=0, for all integers j\geq 0. Thus by Proposition 3.2 we have \operatorname{Tor}^R_{i}(R/J,D_I(N))=0, for all integers i\geq 0. On the other hand, in view of Lemma 3.3 we have N\otimes_R R/I=0 and hence it follows from the hypothesis I \subseteq J that N\otimes_R R/J=0. Hence, using the long exact sequence induced by the exact sequence (3.4.1) it follows that \operatorname{Tor}^R_{j}(R/J,H^1_I(N))=0, for j=0,1. Therefore, \operatorname{Tor}^R_{j}(R/J,H^t_I(M))\simeq\operatorname{Tor}^R_{j}(R/J,H^1_I(N))=0, for j=0,1.

The assertion follows from Lemma 3.4.

(i)\Rightarrow(ii) Is clear.

(ii)\Rightarrow(i) Assume the opposite. Then there is an integer j>\operatorname{cd}(I,R) such that \operatorname{Tor}^R_{j}(R/I,M)\neq 0. Let \operatorname{\frak p} \in \operatorname{Supp} \operatorname{Tor}^R_{j}(R/I,M). Then localizing at \operatorname{\frak p}, without loss of generality, we may assume that (R,\operatorname{\frak m}) is a Noetherian local ring. Let T:=D(M), where D(-) denotes the Matlis dual functor. Then by the adjointness we have

for all 0\leq i\leq \operatorname{cd}(I,R). Now, by [AM, Theorem 2.9] we have H^i_I(T)=0, for all 0 \leq i \leq \operatorname{cd}(I,R). Hence, H^i_I(T)=0, for all integers i\geq 0. So, in view of [AM, Theorem 2.9] we have \operatorname{Ext}^i_R(R/I,T)=0, for all integers i\geq 0. Consequently, by the adjointness, we have

for all i\geq 0. Thus, \operatorname{Tor}^R_i(R/I,M)=0 for all i\geq 0, which is a contradiction.

(i)\Leftrightarrow(ii) See [BS, Lemma 6.3.1].

(i)\Rightarrow(iii) Follows from Proposition 3.6.

(iii)\Rightarrow(i) Assume the opposite. Then we have \operatorname{cd}(I,R)\geq 2. Let t=\operatorname{cd}(I,R). Then by Corollary 3.5 for the R-module M:=H^t_I(R) we have \operatorname{Tor}^R_{i}(R/I,M)=0 for i=0,1. So, by the hypothesis we have \operatorname{Tor}^R_{i}(R/I,M)=0, for all integers i\geq 0. Therefore, by Proposition 3.2 we have \operatorname{Ext}^i_{R}(R/I,M)=0, for all integers i\geq 0. Therefore (0:_{H^t_I(R)}I)\simeq\operatorname{Hom}_R(R/I,H^t_I(R))=0. Hence H^t_I(R)=0, which is a contradiction.

(i)\Rightarrow(iv) Let M be an R-module such that \operatorname{Ext}^i_{R}(R/I,M)=0 for i=0,1. Then, by [AM, Theorem 2.9] we have H^i_I(M)=0 for i=0,1. On the other hand by [BS, Lemma 6.3.1] we have H^i_I(M)=0, for all integers i\geq 2. Therefore, H^i_I(M)=0, for all integers i\geq0. Hence, it follows from [AM, Theorem 2.9] that \operatorname{Ext}^i_R(R/I,M)=0 for all integers i\geq 0.

(iv)\Rightarrow(i) By [BS, Corollary 2.2.8] we have H^i_I(D_I(R))=0, for i=0,1. Hence, \operatorname{Ext}^i_R(R/I,D_I(R))=0 for i=0,1 by [AM, Theorem 2.9]. So, by the hypothesis we have \operatorname{Ext}^i_R(R/I,D_I(R))=0, for all integers i\geq 0. Hence, by Proposition 3.2 it follows that R/I\otimes_R D_I(R)=0 and so D_I(R)=ID_I(R). Now the assertion follows from [BS, Lemma 6.3.1 and Proposition 6.3.5].

(i)\Rightarrow(ii) If {\rm cd}(I,R)=2, then for the R-module M:=H^2_I(R) by Corollary 3.5 we have \operatorname{Tor}^R_{i}(R/I,M)=0 for i=0,1. Now, we claim that \operatorname{Tor}^R_{2}(R/I,M)\neq0. Assume the opposite. Then, it follows from Proposition 3.6 that \operatorname{Tor}^R_{i}(R/I,M)=0, for all integers i\geq 0. Then it follows from Proposition 3.2 that \operatorname{Ext}^i_{R}(R/I,M)=0, for all integers i\geq 0. Therefore, \operatorname{Hom}_R(R/I,H^2_I(R))=0, which implies that H^2_I(R)=0. This is a contradiction.

(ii)\Rightarrow(i) Under the given hypothesis it follows from Theorem 3.7 that {\rm cd}(I,R)\geq2. On the other hand, by [BS, Theorem 3.3.1] we have {\rm cd}(I,R)\leq2. So, we have {\rm cd}(I,R)=2.

(i)\Rightarrow(iii) If {\rm cd}(I,R)=2, then for the R-module M:=D_I(R), by [BS, Corollary 2.2.8] we have H^i_I(M)=0, for i=0,1. Hence, by [AM, Theorem 2.9] we have \operatorname{Ext}^i_R(R/I,M)=0 for i=0,1. Moreover, by [BS, Remark 2.2.7] there is an exact sequence

which induces the isomorphisms

Now, if \operatorname{Ext}^2_R(R/I,M)=0, then \operatorname{Ext}^i_R(R/I,M)=0, for i=0, 1, 2,. So, it follows from [AM, Theorem 2.9] that H^i_I(M)=0 for i=0,1,2, which is a contradiction, because H^2_I(M)\neq 0. So, for the R-module M:=D_I(R) we have \operatorname{Ext}^i_{R}(R/I,M)=0 for i=0,1 and \operatorname{Ext}^2_{R}(R/I,M)\neq0.

(iii)\Rightarrow(i) Under the given hypothesis it follows from Theorem 3.7 that {\rm cd}(I,R)\geq2. On the other hand, by [BS, Theorem 3.3.1] we have {\rm cd}(I,R)\leq2. So, we have {\rm cd}(I,R)=2.

(i) Let M be an arbitrary R-module. In order to prove the assertion we may assume \operatorname{cd}(I,M)=t\geq 2. By Corollary 3.5 we have \operatorname{Tor}^R_j(R/J,H^t_I(M))=0 for j=0,1. Now, it follows from Proposition 3.6 that \operatorname{Tor}^R_i(R/J,H^t_I(M))=0, for all integers i\geq 0. Therefore, by Proposition 3.2 we have \operatorname{Ext}^i_{R}(R/J,H^t_I(M))=0, for all integers i\geq 0. Now, if t\geq 3 then by Lemma 3.4 we have \operatorname{Tor}^R_j(R/J,H^{t-1}_I(M))=0 for j=0,1. So, it follows from Proposition 3.6 that \operatorname{Tor}^R_i(R/J,H^{t-1}_I(M))=0, for all integers i\geq 0. Therefore, by Proposition 3.2 we have \operatorname{Ext}^i_{R}(R/J,H^{t-1}_I(M))=0, for all integers i\geq 0. Proceeding in the same way we see that \operatorname{Ext}^j_R(R/J,H^i_I(M))=0, for all integers i\geq 2 and j\geq 0.

(ii) Using [AB1, Lemma 2.2] follows from (i).

(iii) Follows from (ii).

(iv) Follows from (ii) using Proposition 3.2.

Since by the hypothesis we have J\not\subseteq \operatorname{Rad} (I) it follows that

which implies that J\not\subseteq \operatorname{\frak q} for some \operatorname{\frak q}\in \operatorname{mAss}_R R/I. Assume that k:=\operatorname{height} \operatorname{\frak q}. Then by Grothendieck’s Non-vanishing Theorem we have

which implies that \operatorname{\frak q} \in \operatorname{Supp} H^k_I(R). Hence we have \operatorname{Supp} H^k_I(R)\not\subseteq V(J) and so we have

In particular, we have \{i\in \Bbb{Z}\,\,:\,\,\operatorname{Supp} H^i_I(R)\not\subseteq V(J)\}\neq\emptyset.

Now, set \ell:= \max\{i\in \Bbb{Z}\,\,:\,\,\operatorname{Supp} H^i_I(R)\not\subseteq V(J)\} and t:=\operatorname{cd}(I\cap J,R). Then as by [BS, Corollary 3.3.3] we have \ell\leq{\rm ara}(I) and t\leq{\rm ara}(I\cap J) it follows that 0\leq\ell<\infty and 0\leq t<\infty. By the Mayer-Vietoris exact sequence for each integer i>t we have the exact sequence

which gives the exact sequence

and hence we have

So, it is clear that \ell \leq t. On the other hand, since by the hypothesis R is a domain and I\cap J\neq 0, it follows that \ell \geq 1 and t \geq 1. Hence, if t=1, then it is clear that \ell=1=t. Now, assume that t\geq2 and \ell<t. Then, we have \operatorname{Supp} H^t_I(R)\subseteq V(J). Also, by the Mayer-Vietoris exact sequence we have the exact sequence

which implies that

So, the non-zero R-module H^t_{I\cap J}(R) is J-torsion and hence we have

But, by Theorem 4.1 we have

which is a contradiction. So, we have \ell=t, whenever t\geq 2.

By the Mayer-Vietoris exact sequence we have the exact sequence

which considering that fact that \operatorname{Supp} H^k_{I+J}(R) \subseteq V(I+J)\subseteq V(J) implies that \operatorname{Supp} H^k_{I\cap J}(R)\not\subseteq V(J). (Note that by the hypothesis we have \operatorname{Supp} H^k_I(R)\not\subseteq V(J).) In particular, we have H^k_{I\cap J}(R)\neq 0. In order to prove the assertion, assume the opposite and assume that \dim \operatorname{Supp} H^k_{I\cap J}(R)=d. Then, in view of [M, Theorem 2.9] we have H^d_{\operatorname{\frak m}}(H^k_{I\cap J}(R))\neq 0. On the other hand, by Theorem 4.1 we have \operatorname{Ext}^j_R(R/\operatorname{\frak m},H^k_{I\cap J}(R) )=0, for all integers j\geq 0. Hence it follows from [AM, Theorem 2.9] that H^j_{\operatorname{\frak m}}(H^k_{I\cap J}(R))=0, for all integers j\geq 0. Therefore, we have H^d_{\operatorname{\frak m}}(H^k_{I\cap J}(R))=0, which is a contradiction.