alphaFeedback

WorksMathematics

Local cohomology over Steenrod algebras

AI contributions
Not recorded for this paper
View details

Abstract

Let Fq\mathbb{F}_q be a finite field with q=psq = p^s elements. Let S=⨁n≥0SnS = \bigoplus_{n \geq 0} S_n be a non-negatively graded commutative, Noetherian Fq\mathbb{F}_q-algebra. Assume SS has a structure of an unstable Steenrod algebra. Let MM be a finitely generated graded unstable SS-module. Let II be a homogeneous ideal in SS. We prove that for i≥0i \geq 0 the local cohomology modules HIi(M)H^i_I(M) has a structure of a Cartan SS-module. As application we give a direct and considerably simpler proof of D. Bourguiba and S. Zarati's generalization of Landweber-Stong conjecture. We also give a necessary and sufficient condition for SS to have SrS_r-property of Serre in terms of Dickson polynomials.

Introduction

Let Fq\mathbb{F}_q be a finite field with q=psq=p^s elements. Let S=⨁n≥0SnS=\bigoplus_{n\geq0}S_n be a non-negatively graded Noetherian Fq\mathbb{F}_q-algebra with S0=FqS_0=\mathbb{F}_q. Assume SS has a structure of an unstable Steenrod algebra. Recall that this means there exists Fq\mathbb{F}_q linear maps Pi:S→S\mathcal{P}^i:S\to S with P0\mathcal{P}^0 to be the identity and satisfy the Cartan identity:

Pk(uv)=∑i+j=kPi(u)Pj(v).\mathcal{P}^k(uv)=\sum_{i+j=k}\mathcal{P}^i(u)\mathcal{P}^j(v).

We have for homogeneous uu, Pdeg⁡u(u)=uq\mathcal{P}^{\deg u}(u)=u^q and Pi(u)=0\mathcal{P}^i(u)=0 for i>deg⁡ui>\deg u. We also have that Pr(Sn)⊆Sn+r(q−1)\mathcal{P}^r(S_n)\subseteq S_{n+r(q-1)} for all n∈Zn\in\mathbb{Z}.

A large class of examples of unstable Steenrod algebras are rings of invariants SS of a finite group G⊆GLd(Fq)G\subseteq GL_d(\mathbb{F}_q). We note that GLd(Fq)GL_d(\mathbb{F}_q) is a finite group and so SS contains D∗(d)D^*(d) the ring of invariants of GLd(Fq)GL_d(\mathbb{F}_q). Dickson proved that D∗(d)D^*(d) is a polynomial ring with D∗(d)=Fq[dd,0,dd,1,…,dd,d−1]D^*(d)=\mathbb{F}_q[d_{d,0},d_{d,1},\ldots,d_{d,d-1}] with deg⁡dd,i=qd−qi\deg d_{d,i}=q^d-q^i. We follow the exposition given in [10].

One of the landmark results in modular invariant theory is the proof by D. Bourguiba and S. Zarati, see [1], of the Landweber-Stong conjecture [5]. This result states that depth⁡S≥r\operatorname{depth}S\geq r if and only if dd,d−1,…,dd,d−rd_{d,d-1},\ldots,d_{d,d-r} is an SS-regular sequence. In our paper [9] we gave an alternate and considerably simpler proof of this conjecture.

D. Bourguiba and S. Zarati prove a considerably more general result than the Landweber-Stong conjecture, see [1 1.3] A natural question was whether our techniques generalize. In this paper we show that our techniques from [9] generalize and give an alternate and considerably simpler proof of D. Bourguiba and S. Zarati’s result.

The crucial point is the notion of a Cartan SS-module which was introduced in [9].

Definition 1.1. By a Cartan SS-module we mean a graded SS-module MM equipped with Fq\mathbb{F}_q-linear maps QMi:M→MQ_M^i:M\to M for i≥0i\geq0 with QM0Q_M^0 being the identity such that for any s∈Ss\in S and m∈Mm\in M we have for all r≥1r\geq1

QMr(sm)=∑i+j=rPi(s)QMj(m).(1)Q_M^r(sm)=\sum_{i+j=r}\mathcal{P}^i(s)Q_M^j(m). \tag*{(1)}
QMr(Mn)⊆Mn+r(q−1)for all n∈Z.(2)Q_M^r(M_n)\subseteq M_{n+r(q-1)}\quad\text{for all }n\in\mathbb{Z}. \tag*{(2)}

It is easily seen that if MM is a Cartan SS-module then ann⁡SM\operatorname{ann}_S M is a P∗\mathcal{P}^*-invariant ideal of SS, see 3.10. The main result which enables all our applications is the following:

Theorem 1.2. Let S=⨁n≥0SnS=\bigoplus_{n\geq0}S_n with S0=FqS_0=\mathbb{F}_q be a Noetherian Fq\mathbb{F}_q graded algebra. Assume SS has a structure of an unstable Steenrod algebra. Let MM be a Cartan SS-module (not necessarily finitely generated). Let II be a homogeneous ideal in SS and let x=x1,…,xmx=x_1,\ldots,x_m be a set of homogeneous generators of II. Then for i≥0i\geq0 the local cohomology modules HIi(M)H_I^i(M) have the structure of a Cartan SS-module (depending on xx).

Remark 1.3. Let SS be an unstable algebra over the Steenrod algebra. Let dim⁡S=d\dim S=d. Then there exists an embedding of unstable Steenrod algebras

ψ:Fq[dd,0qt,dd,1qt,…,dd,d−1qt]⟶S\psi:\mathbb{F}_q[d_{d,0}^{q^t},d_{d,1}^{q^t},\ldots,d_{d,d-1}^{q^t}]\longrightarrow S

such that SS is a finite Fq[dd,0qt,dd,1qt,…,dd,d−1qt]\mathbb{F}_q[d_{d,0}^{q^t},d_{d,1}^{q^t},\ldots,d_{d,d-1}^{q^t}]-module via ψ\psi. For a proof of this result see [1 1.2]. We will set Fq[dd,0qt,dd,1qt,…,dd,d−1qt]\mathbb{F}_q[d_{d,0}^{q^t},d_{d,1}^{q^t},\ldots,d_{d,d-1}^{q^t}] as D(d)qtD(d)^{q^t}. We will fix such an embedding for SS. By abuse of notation we consider the embedding as an inclusion.

Applications:

I. We give a considerably simpler and a direct proof of Bourguiba and Zarati’s result.

Theorem 1.4. (with setup as in 1.3) Let MM be a finitely generated Cartan SS-module. The following assertions are equivalent

(i) depth M≥rM\geq r.

(ii) dd,d−1qt,…,dd,d−rqtd_{d,d-1}^{q^t},\ldots,d_{d,d-r}^{q^t} is an MM-regular sequence.

II. Associate primes of Cartan modules

An ideal II in SS is said to be P∗\mathcal{P}^*-invariant if P∗(I)⊆I\mathcal{P}^*(I)\subseteq I for all i≥0i\geq0. We prove

Theorem 1.5. (with setup as in 1.3) Let MM be a finitely generated Cartan SS-module. If P∈Ass⁡MP\in\operatorname{Ass}M then PP is a P∗\mathcal{P}^* invariant.

Proof of Theorem 1.5 requires that if MM is a Cartan SS-module then for any ideal JJ of SS we have HJ0(M)H_J^0(M) is a Cartan submodule of MM. This follows from Theorem 1.2. However we also give a direct proof. Theorem 1.5 is not true if MM is not finitely generated. We give a counter-example, see 8.5.

SrS_r property

Let TT be a Noetherian ring TT and let MM be a finitely generated TT-module. Set depth 0=+∞0=+\infty. Recall MM satisfies SrS_r property of Serre if

depth⁡MP≥min⁡{dim⁡MP,r}for all primes P of T.\operatorname{depth} M_P \ge\min\{\dim M_P,r\}\quad\text{for all primes }P\text{ of }T.

A natural question is what happens if dd,0qt,dd,1qt,⋯dd,r−1qtd_{d,0}^{q^t},d_{d,1}^{q^t},\cdots d_{d,r-1}^{q^t} is an MM-regular sequence. We extend our result in [9 1.6] to prove

Theorem 1.6. (with setup as in 1.3) Let MM be a finitely generated Cartan SS-module. Assume Supp⁡(M)=Spec⁡(S)\operatorname{Supp}(M)=\operatorname{Spec}(S) and that SS is equi-dimensional. The following assertions are equivalent:

(i) MM satisfies SrS_r-property of Serre.

(ii) dd,0qt,dd,1qt,⋯dd,r−1qtd_{d,0}^{q^t},d_{d,1}^{q^t},\cdots d_{d,r-1}^{q^t} is an MM-regular sequence.

Cohomological annihilators

1.7. If MM is a finitely generated Cartan SS-module then HS+i(M)H^i_{S_+}(M) is a Cartan SS-module. It is not difficult to prove that height⁡ann⁡HS+i(M)≥d−i\operatorname{height}\operatorname{ann} H^i_{S_+}(M)\ge d-i, see 6.5. As ann⁡HS+i(M)\operatorname{ann} H^i_{S_+}(M) is P∗P^*-invariant it follows that dd,0qt∈ann⁡HS+i(M)d_{d,0}^{q^t}\in\sqrt{\operatorname{ann} H^i_{S_+}(M)} for i<d=dim⁡Si<d=\dim S. Thus some power dd,0qtd_{d,0}^{q^t} works as a cohomological annihilator. We show

Corollary 1.8. (with hypotheses as in 1.3). Let G∙=0→G0→G1→…→Gr→0G^\bullet=0\to G^0\to G^1\to\ldots\to G^r\to0 be a complex of free graded SS-modules and homogeneous maps such that Hi(G∙)H^i(G^\bullet) has finite length for all ii. Let MM be a finitely generated Cartan SS-module. Then dd,0qt∈ann⁡SHi(G∙⊗M)d_{d,0}^{q^t}\in\sqrt{\operatorname{ann}_S H^i(G^\bullet\otimes M)} for 0≤i≤d−10\le i\le d-1.

Proof. This follows from an exercise problem [3 8.1.6] and the discussion in 1.7. □

In [10 11.5.5] it is proved that if SS is a ring of invariants then Sdd,0S_{d_{d,0}} is Cohen-Macaulay. We extend this result.

Corollary 1.9. (with hypotheses as in 1.3). Let MM be a finitely generated Cartan SS-module. Then Mdd,0qtM_{d_{d,0}^{q^t}} is Cohen-Macaulay.

Remark 1.10. If dim⁡M<d\dim M<d and MM is a Cartan module then by 3.12, dd,0qt∈ann⁡M‾d_{d,0}^{q^t}\in\sqrt{\operatorname{ann}\overline{M}}. So Mdd,0qt=0M_{d_{d,0}^{q^t}}=0. Thus Corollary 1.9 has content only if dim⁡M=dim⁡S=d\dim M=\dim S=d.

Monomial ideals

Let R=Fq[X1,…,Xd]R=\mathbb{F}_q[X_1,\ldots,X_d] with the natural Steenrod algebra structure. Let II be a monomial ideal in RR. It is not difficult to show that II is a P∗P^*-invariant ideal of RR, see 2.3. As an application of this observation we prove:

Theorem 1.11. Let kk be a field of characteristic p>0p>0 containing Fq\mathbb{F}_q. Set R=Fq[X1,…,Xd]R=\mathbb{F}_q[X_1,\ldots,X_d] and T=k[X1,…,Xd]T=k[X_1,\ldots,X_d]. Let II be a monomial ideal in TT and let JJ be a homogeneous ideal in RR. Assume M=HJTi(T/I)M=H^i_{JT}(T/I) is finitely generated as a TT-module for some i≥0i\ge0. If P∈Ass⁡TMP\in\operatorname{Ass}_T M then PP is generated by linear polynomials in TT.

Finite generation of cohomology

Let SS be an unstable Steenrod algebra. If dim⁡S≥2\dim S\ge2 then there are infinitely many homogeneous prime ideals. Note there exists only finitely many P∗\mathcal{P}^{*}-invariant prime ideals. For a prime ideal PP of SS define

fP=min⁡{i∣HPi(S) is not finitely generated S-module}.f_P=\min\{i\mid H_P^i(S)\text{ is not finitely generated }S\text{-module}\}.

We prove

Theorem 1.12. (with setup as above). Assume PP is not a P∗\mathcal{P}^{*}-invariant prime ideal. Then fP≤depth⁡SPf_P\leq\operatorname{depth}S_P.

Remark 1.13. For prime ideals which are P∗\mathcal{P}^{*}-invariant, the bound in Theorem 1.12 need not hold. See 12.1 for an example.

We now describe in brief the contents of this paper. In section two we discuss a few preliminary results that we need. In Section three we discuss our notion of Cartan modules and discuss a few preliminary properties. In section four we give an alternate description of Cartan SS-modules which is convenient for us. In the next section discuss a Cartan SS-module structure on MxM_x where x∈Sx\in S is homogeneous and MM is a Cartan SS-module. In section six we prove Theorem 1.2. In the next section we give a proof of Theorem 1.4. In section eight we give proof of Theorem 1.5. In the next section we give proof of our applications III. Finally in section ten we give a proof of Corollary 1.9. In the next section we give a proof of Theorem 1.11. Finally in section twelve we give a proof of Theorem 1.12.

Preliminaries

Steenrod operators

We give the following well-known construction of Steenrod operators on polynomial rings and ring of invariants. We follow the exposition given in [10].

Let pp be a prime, q=psq=p^s and let VV be a Fq\mathbb{F}_q-vector space of dimension dd. Define P(ξ):Fq[V]→Fq[V][ξ]\mathcal{P}(\xi):\mathbb{F}_q[V]\to\mathbb{F}_q[V][\xi] (with deg⁡ξ=1−q\deg\xi=1-q) by the rules:

  1. P(ξ)(v)=v+vqξ\mathcal{P}(\xi)(v)=v+v^q\xi for all v∈V∗v\in V^*.

  2. P(ξ)(u+w)=P(ξ)(u)+P(ξ)(w)\mathcal{P}(\xi)(u+w)=\mathcal{P}(\xi)(u)+\mathcal{P}(\xi)(w) for all u,w∈Fq[V]u,w\in\mathbb{F}_q[V].

  3. P(ξ)(uw)=P(ξ)(u)P(ξ)(w)\mathcal{P}(\xi)(uw)=\mathcal{P}(\xi)(u)\mathcal{P}(\xi)(w) for all u,w∈Fq[V]u,w\in\mathbb{F}_q[V].

  4. P(ξ)(1)=1\mathcal{P}(\xi)(1)=1.

We note that P(ξ)\mathcal{P}(\xi) is a ring homomorphism of degree zero.

By separating out homogeneous components we obtain Fq\mathbb{F}_q linear maps Pi:Fq[V]→Fq[V]\mathcal{P}^i:\mathbb{F}_q[V]\to\mathbb{F}_q[V] by the requirement

P(ξ)(f)=∑i≥0Pi(f)ξi.\mathcal{P}(\xi)(f)=\sum_{i\geq0}\mathcal{P}^i(f)\xi^i.

Note P0\mathcal{P}^0 is the identity. The operations Pi\mathcal{P}^i are called the Steenrod reduced power operations over Fq\mathbb{F}_q and when p=2p=2, also denoted by SqiSq^i and referred to as Steenrod squaring operations. We will use the following Cartan formulae:

Pk(uv)=∑i+j=kPi(u)Pj(v).\mathcal{P}^k(uv)=\sum_{i+j=k}\mathcal{P}^i(u)\mathcal{P}^j(v).

We have for homogeneous uu, Pdeg⁡u(u)=uq\mathcal{P}^{\deg u}(u)=u^q and Pi(u)=0\mathcal{P}^i(u)=0 for i>deg⁡ui>\deg u. We also have that Pr(Fq[V]n)⊆Fq[V]n+r(q−1)\mathcal{P}^r(\mathbb{F}_q[V]_n)\subseteq\mathbb{F}_q[V]_{n+r(q-1)} for all n∈Zn\in\mathbb{Z}.

##

Let G⊆GL(V)G \subseteq GL(V) be a finite group. Let S=Fq[V]GS = \mathbb{F}_q[V]^G. We define an action of GG on Fq[V][ξ]\mathbb{F}_q[V][\xi] by giving the trivial action on ξ\xi. We note that σ(P(ξ)(r))=P(ξ)(σ(r))\sigma(P(\xi)(r)) = P(\xi)(\sigma(r)) for all σ∈G\sigma\in G. Thus the maps Pi\mathcal{P}^i restrict to maps from SS to SS. So SS is an unstable Steenrod algebra.

Proposition 2.3. Let R=Fq[X1,…,Xd]R = \mathbb{F}_q[X_1,\ldots,X_d] be the polynomial ring in dd-variables over Fq\mathbb{F}_q and give it the unstable Steenrod algebra structure as defined above. Let II be a monomial ideal in RR. Then

  1. II is a P∗\mathcal{P}^*-invariant ideal.

  2. R/IR/I is an unstable Steenrod algebra.

We will need the following easily proved result.

Lemma 2.4. Let SS be an unstable Steenrod algebra and let I,JI,J be P∗\mathcal{P}^*-invariant ideal. Then I+JI+J and IJIJ are P∗\mathcal{P}^*-invariant ideals. □\square

We now give

Proof of Proposition 2.3. (1) By using 2.4 repeatedly it suffices to prove that (Xi)(X_i) is P∗\mathcal{P}^*-invariant. But this is obvious.

(2) This follows from (1). □\square

Cartan Modules

In this section we describe our notion of Cartan modules. We work with the setup as in 1.3. Let Pi:S→S\mathcal{P}^i:S \to S be the ithi^{\text{th}}-Steenrod operator on SS.

Definition 3.1. By a Cartan SS-module we mean a graded SS-module MM equipped with Fq\mathbb{F}_q-linear maps QMi:M→MQ^i_M:M \to M for i≥0i \ge0 with QM0Q^0_M being the identity such that for any s∈Ss \in S and m∈Mm \in M we have for all $r \ge 1

(1)

QMr(sm)=∑i+j=rPi(s)QMj(m).Q^r_M(sm)=\sum_{i+j=r}\mathcal{P}^i(s)Q^j_M(m).

(2) QMr(Mn)⊆Mn+r(q−1)Q^r_M(M_n)\subseteq M_{n+r(q-1)} for all n∈Zn \in\mathbb{Z}.

Next we define morphisms of Cartan SS-modules.

Definition 3.2. Let M,NM,N be Cartan SS-modules. By a morphism f:M→Nf:M \to N of Cartan SS-modules we mean a homogeneous SS-linear map f:M→Nf:M \to N such that the following diagram commutes for all i≥0i \ge0.

M→fNQMi↓↓QNiM→fN\begin{CD} M @>f>> N \\ @VQ^i_MVV @VVQ^i_NV \\ M @>f>> N \end{CD}

3.3. Let M,NM,N be Cartan SS-modules. We denote

Hom⁡C(S)(M,N)={f:M→N∣f is a morphism of Cartan S-modules}.\operatorname{Hom}_{\mathcal{C}(S)}(M,N)=\{f:M \to N \mid f\text{ is a morphism of Cartan }S\text{-modules}\}.

It is evident that Hom⁡C(S)(M,N)\operatorname{Hom}_{\mathcal{C}(S)}(M,N) is an abelian group. The following result follows easily from the definition of morphism of Cartan SS-modules.

Proposition 3.4. Let M,NM,N be Cartan SS-modules and let f:M→Nf:M\to N be a morphism of Cartan SS-modules. Then ker⁡f\ker f, image⁡f\operatorname{image} f, coker⁡f\operatorname{coker} f are Cartan SS-modules. Furthermore the natural inclusions ker⁡f→M\ker f\to M, image⁡f→N\operatorname{image} f\to N and the surjection N→coker⁡fN\to\operatorname{coker} f are morphisms of Cartan SS-modules.

We will need a few preliminaries.

Proposition 3.5. Let M,NM,N be Cartan SS-modules. Then M⊕NM\oplus N also has a natural Cartan SS-module structure such that

  1. the natural maps M→M⊕NM\to M\oplus N, N→M⊕NN\to M\oplus N, M⊕N→MM\oplus N\to M and M⊕N→NM\oplus N\to N are morphisms of Cartan SS-modules.

  2. if UU is a Cartan SS-module and f:M→Uf:M\to U and g:N→Ug:N\to U are morphism of Cartan SS-modules then f⊕g:M⊕N→Uf\oplus g:M\oplus N\to U is a morphism of Cartan SS-modules.

  3. if UU is a Cartan SS-module and α:U→M\alpha:U\to M and β:U→N\beta:U\to N are morphism of Cartan SS-modules then (α,β):U→M⊕N(\alpha,\beta):U\to M\oplus N is a morphism Cartan SS-modules.

Proof. Define QM⊕Nr:M⊕N→M⊕NQ^r_{M\oplus N}:M\oplus N\to M\oplus N to be

QM⊕Nr(m,n)=(Qr(m),Qr(n)).Q^r_{M\oplus N}(m,n)=(Q^r(m),Q^r(n)).

It is readily verified that this gives a Cartan SS-module structure on M⊕NM\oplus N. The assertions (1), (2) and (3) hold trivially.

We also need the following easily proved result.

Proposition 3.6. Let Mi,NjM_i,N_j be Cartan SS-modules for 1≤i≤m1\leq i\leq m and 1≤j≤n1\leq j\leq n. Set M=⨁i=1mMiM=\bigoplus_{i=1}^{m}M_i and N=⨁j=1nNjN=\bigoplus_{j=1}^{n}N_j. Give M,NM,N the induced Cartan SS-module structure. Let ψ=(ψi,j):M→N\psi=(\psi_{i,j}):M\to N where ψi,j:Mi→Nj\psi_{i,j}:M_i\to N_j. If each ψi,j\psi_{i,j} is morphism of Cartan SS-modules then ψ\psi is a morphism of Cartan SS-modules.

3.7. Let C(S)C(S) denote the category of Cartan SS-modules and Cartan morphisms. Then by 3.4, 3.5, 3.6 it follows that C(S)C(S) is an abelian category. As a consequence we obtain

Corollary 3.8. Let G∙G^\bullet be a complex of Cartan SS-modules and morphisms. Then Hi(G∙)H^i(G^\bullet) are Cartan SS-modules for all i∈Zi\in\mathbb{Z}.

3.9. An ideal II of SS is said to be P∗\mathcal{P}^*-invariant if Pi(I)⊆I\mathcal{P}^i(I)\subseteq I for all i≥0i\geq0. We note that if II is P∗\mathcal{P}^*-invariant then so is I\sqrt{I}, see [10 11.2.5].

The following Lemma is critical. We proved it in [9 3.5] for rings of invariants. Same proof works in general.

Lemma 3.10. Let MM be a Cartan SS-module. Then ann⁡M\operatorname{ann} M is a P∗\mathcal{P}^*-invariant ideal of SS.

3.11. The non-zero P∗\mathcal{P}^*-prime ideals of D(d)qtD(d)^{q^t} are (dd,0qt,⋯ ,dd,iqt)(d_{d,0}^{q^t},\cdots,d_{d,i}^{q^t}) for i=0,…,d−1i=0,\ldots,d-1; see [8 9.2.7]. We need the following result. We proved it for ring of invariants, see [9 2.8]. The same proof works in general.

Lemma 3.12. (with setup as in 1.3). Let JJ be a non-zero radical P∗\mathcal{P}^*-invariant ideal of SS. If dim⁡S/J=d−g\dim S/J=d-g then dd,0qt,⋯ ,dd,g−1qt∈Jd_{d,0}^{q^t},\cdots,d_{d,g-1}^{q^t}\in J.

An alternate description of Cartan modules

In this section we give an alternate description of Cartan modules which will be useful later. Throughout SS is an unstable Steenrod Fq\mathbb{F}_q-algebra.

The total Steenrod operation

4.1. The total Steenrod operation. This is the map defined by P(ξ):S→S[ξ]\mathcal{P}(\xi): S \to S[\xi] where P(ξ)(u)=∑i≥0Pi(u)ξi\mathcal{P}(\xi)(u) = \sum_{i\geq0} P^{i}(u)\xi^{i}. We note that P(ξ)\mathcal{P}(\xi) is a ring homomorphism. Furthermore if we set deg⁡(ξ)=1−q\deg(\xi)=1-q then P(ξ)\mathcal{P}(\xi) is a homogeneous ring homomorphism. We note that S[ξ]S[\xi] has elements of negative degrees (as deg⁡(ξ)=1−q\deg(\xi)=1-q) and so is a Z\mathbb{Z}-graded ring.

Important point

4.2. Important point: (with setup as above). As P(ξ):S→S[ξ]\mathcal{P}(\xi): S \to S[\xi] is a ring homomorphism, it induces a new SS algebra structure on S[ξ]S[\xi].

Definition 4.3. Let MM be an SS-module and let NN be an S[ξ]S[\xi]-module and let f:M→Nf:M\to N be an additive homogeneous graded map. We say ff is a Steenrod map if ff is SS-linear with NN given an SS-module structure via P(ξ)\mathcal{P}(\xi), i.e., f(um)=P(ξ)(u)f(m)f(um)=\mathcal{P}(\xi)(u)f(m) for all u∈Su\in S and for all m∈Mm\in M.

Construction of the graded subring

4.4. We need to construct a graded subring S[[ξ]]∗S[[\xi]]^* of S[[ξ]]S[[\xi]]. For a homogeneous element s∈Ss\in S set ∣s∣=deg⁡(s)|s|=\deg(s). For n∈Zn\in\mathbb{Z}, set

S[[ξ]]n∗={α=∑i≥0αiξi∣αi is homogeneous and ∣αi∣=n+i(q−1)}.S[[\xi]]^*_{n}=\left\{\alpha=\sum_{i\geq0}\alpha_i\xi^i\mid\alpha_i\text{ is homogeneous and }|\alpha_i|=n+i(q-1)\right\}.

It is readily verified that S[[ξ]]n∗S[[\xi]]^*_{n} is an additive subgroup of S[[ξ]]S[[\xi]]. Furthermore it is easy to check that

S[[ξ]]n∗S[[ξ]]m∗⊆S[[ξ]]m+n∗.S[[\xi]]^*_{n}S[[\xi]]^*_{m}\subseteq S[[\xi]]^*_{m+n}.

Set S[[ξ]]∗=⨁n∈ZS[[ξ]]n∗S[[\xi]]^*=\bigoplus_{n\in\mathbb{Z}}S[[\xi]]^*_{n}. It is a graded ring. It also contains S[ξ]S[\xi] as a graded subring.

Construction of the graded submodule

4.5. Let MM be a graded SS-module (not necessarily finitely generated). We need to construct a graded S[[ξ]]∗S[[\xi]]^* submodule M[[ξ]]∗M[[\xi]]^* of M[[ξ]]M[[\xi]]. For a homogeneous element m∈Mm\in M set ∣m∣=deg⁡(m)|m|=\deg(m). For n∈Zn\in\mathbb{Z}, set

M[[ξ]]n∗={β=∑i≥0αiξi∣βi is homogeneous in M and ∣βi∣=n+i(q−1)}.M[[\xi]]^*_{n}=\left\{\beta=\sum_{i\geq0}\alpha_i\xi^i\mid\beta_i\text{ is homogeneous in }M\text{ and }|\beta_i|=n+i(q-1)\right\}.

It is readily verified that M[[ξ]]n∗M[[\xi]]^*_{n} is an additive subgroup of M[[ξ]]M[[\xi]]. Furthermore it is easy to check that

S[[ξ]]n∗M[[ξ]]m∗⊆M[[ξ]]m+n∗.S[[\xi]]^*_{n}M[[\xi]]^*_{m}\subseteq M[[\xi]]^*_{m+n}.

Set M[[ξ]]∗=⨁n∈ZM[[ξ]]n∗M[[\xi]]^*=\bigoplus_{n\in\mathbb{Z}}M[[\xi]]^*_{n}. It is a graded S[[ξ]]∗S[[\xi]]^*-module. It also contains M[ξ]M[\xi] as S[ξ]S[\xi]-submodule.

We now give our characterization of Cartan modules.

Theorem 4.6. (with setup as above). Give S[[ξ]]∗S[[\xi]]^* modules a S[ξ]S[\xi]-module structure via the inclusion S[ξ]↪S[[ξ]]∗S[\xi]\hookrightarrow S[[\xi]]^*. Let MM be a graded SS-module. The following assertions are equivalent:

(a) MM is a Cartan SS-module.

(b) There exists a Steenrod map QM(ξ):M→M[[ξ]]∗Q_M(\xi):M\to M[[\xi]]^* with QM(ξ)(m)0=mQ_M(\xi)(m)_0=m for all m∈Mm\in M.

Proof. (a) ⟹\Longrightarrow (b).

Define QM(ξ):M→M[[ξ]]∗Q_M(\xi): M \to M[[\xi]]^* by

QM(ξ)(m)=∑n≥0QMn(m)ξnQ_M(\xi)(m)=\sum_{n\geq0}Q_M^n(m)\xi^n

with deg⁡ξ=1−q\deg\xi=1-q. We note that QMQ_M takes homogeneous elements of degree rr to degree rr elements in M[[ξ]]∗M[[\xi]]^*. Let u∈Su\in S and m∈Mm\in M be homogeneous. We have

P(ξ)(u)QM(m)=(∑n≥0Pn(u)ξn)(∑n≥0QMn(m)ξn)=∑n≥0(∑i+j=nPi(u)QMj(m))ξn=QM(um),by Cartan’s identity.\begin{aligned} \mathcal{P}(\xi)(u)Q_M(m)=\left(\sum_{n\geq0}\mathcal{P}^n(u)\xi^n\right)\left(\sum_{n\geq0}Q_M^n(m)\xi^n\right) \\ &=\sum_{n\geq0}\left(\sum_{i+j=n}\mathcal{P}^i(u)Q_M^j(m)\right)\xi^n \\ &=Q_M(um),\quad\text{by Cartan's identity.} \end{aligned}

The result follows.

(b) ⟹\Longrightarrow (a).

Set

QM(ξ)(m)=∑r≥0QMr(m)ξrQ_M(\xi)(m)=\sum_{r\geq0}Q_M^r(m)\xi^r

Then QMr:M→MQ_M^r:M\to M are Fq\mathbb{F}_q-linear maps. Furthermore as QMQ_M is a homogeneous map and ∣ξ∣=1−q|\xi|=1-q it follows that QMr(Mn)⊆Mn+r(q−1)Q_M^r(M_n)\subseteq M_{n+r(q-1)} for all n∈Zn\in\mathbb{Z}. By assumption QM0(m)=mQ_M^0(m)=m for all m∈Mm\in M. As QMQ_M is a Steenrod map we have for u∈Su\in S, m∈Mm\in M homogeneous

QM(um)=P(ξ)(u)QM(m)=(∑n≥0Pn(u)ξn)(∑n≥0QMn(m)ξn)=∑n≥0(∑i+j=nPi(u)QMj(m))ξn.\begin{aligned} Q_M(um)=\mathcal{P}(\xi)(u)Q_M(m) \\ &=\left(\sum_{n\geq0}\mathcal{P}^n(u)\xi^n\right)\left(\sum_{n\geq0}Q_M^n(m)\xi^n\right) \\ &=\sum_{n\geq0}\left(\sum_{i+j=n}\mathcal{P}^i(u)Q_M^j(m)\right)\xi^n. \end{aligned}

The Cartan identity follows. □\square

4.7. Let xx be a non-nilpotent homogeneous element in SS. Then we can form the S[[ξ]]∗S[[\xi]]^*-graded module Sx[[ξ]]∗S_x[[\xi]]^*. It is readily verified that Sx[[ξ]]∗S_x[[\xi]]^* is a graded ring. Furthermore the natural homogeneous map S[[ξ]]∗→Sx[[ξ]]∗S[[\xi]]^*\to S_x[[\xi]]^* is a ring homomorphism. Similarly we can show that if MM is a graded SS-module then Mx[[ξ]]∗M_x[[\xi]]^* is a graded Sx[[ξ]]∗S_x[[\xi]]^*-module.

We need the following result.

Lemma 4.8. Let SS be a Steenrod Fq\mathbb{F}_q-algebra. Let x∈Sx\in S be a homogeneous non-nilpotent element. We have P(ξ)(x)=x+∑i=1∣x∣Pi(x)ξi\mathcal{P}(\xi)(x)=x+\sum_{i=1}^{|x|}\mathcal{P}^i(x)\xi^i is invertible in Sx[[ξ]]S_x[[\xi]]. Then

  1. P(ξ)(x)−1∈Sx[[ξ]]∗\mathcal{P}(\xi)(x)^{-1}\in S_x[[\xi]]^*.

  1. P(ξ)(x)0−1=1/x\mathcal{P}(\xi)(x)^{-1}_0=1/x.

Proof. (1) and (2). Let ∣x∣=c|x|=c. We show that P(ξ)(x)−1∈Sx[[ξ]]−c∗\mathcal{P}(\xi)(x)^{-1}\in S_x[[\xi]]^{*}_{-c}. Set

P(ξ)(x)−1=∑n≥0αnξn.\mathcal{P}(\xi)(x)^{-1}=\sum_{n\geq0}\alpha_n\xi^n.

We prove by induction that ∣αn∣=n(q−1)−c|\alpha_n|=n(q-1)-c for all n≥0n\geq0. We have P(ξ)(x)P(ξ)(x)−1=1\mathcal{P}(\xi)(x)\mathcal{P}(\xi)(x)^{-1}=1. So we have

∑n≥0(∑i+j=nPi(x)αj)ξn=1.\sum_{n\geq0}\left(\sum_{i+j=n}\mathcal{P}^{i}(x)\alpha_j\right)\xi^n=1.

Comparing coefficients we have xα0=1x\alpha_0=1. So α0=1/x\alpha_0=1/x has degree −c-c. This also verifies (2).

Let n≥1n\geq1. By induction suppose ∣αr∣=r(q−1)−c|\alpha_r|=r(q-1)-c for r=0,1,…,n−1r=0,1,\ldots,n-1. We have ∑i+j=nPi(x)αj=0\sum_{i+j=n}\mathcal{P}^{i}(x)\alpha_j=0. So

xαn+∑i+j=n, j≠nPi(x)αj=0.x\alpha_n+\sum_{i+j=n,\,j\ne n}\mathcal{P}^{i}(x)\alpha_j=0.

We have

∣Pi(x)αj∣=∣Pi(x)∣∣αj∣=i(q−1)+c+j(q−1)−c=n(q−1).|\mathcal{P}^{i}(x)\alpha_j|=|\mathcal{P}^{i}(x)||\alpha_j|=i(q-1)+c+j(q-1)-c=n(q-1).

So ∣αn∣=n(q−1)−c|\alpha_n|=n(q-1)-c. The result follows by induction. □\square

The Cartan SS-module structure on MxM_x

In this section we prove two critical results. Throughout SS is a Steenrod Fq\mathbb{F}_q-algebra, MM is a Cartan SS-module and x∈Sx\in S is homogeneous. We first prove

Theorem 5.1. (with setup as in 1.3). Then there exists a Cartan SS-module structure on MxM_x.

The next result is also critical.

Theorem 5.2. (with setup as in 1.3). Let x,yx,y be homogeneous. Then the natural map Mx→MxyM_x\to M_{xy} is morphism of Cartan SS-modules (the Cartan module structures on Mx,MxyM_x,M_{xy} are as constructed in Theorem 5.1).

We first give

Proof of Theorem 5.1. We may assume that Mx≠0M_x\ne0, otherwise there is nothing to prove. In particular xx is a non-nilpotent element of SS. We note that Mx[[ξ]]M_x[[\xi]] is a graded Sx[[ξ]]∗S_x[[\xi]]^*-module and so an S[ξ]S[\xi]-module via the natural map S[ξ]→Sx[[ξ]]S[\xi]\to S_x[[\xi]]. Let QM(ξ):M→M[[ξ]]∗Q_M(\xi):M\to M[[\xi]]^* be the total Steenrod operation on MM. We construct a Steenrod map QMx(ξ):Mx→Mx[[ξ]]∗Q_{M_x}(\xi):M_x\to M_x[[\xi]]^* as

QMx(ξ)(mxn)=P(ξ)(x)−nQM(ξ)(m).Q_{M_x}(\xi)\left(\frac{m}{x^n}\right)=\mathcal{P}(\xi)(x)^{-n}Q_M(\xi)(m).

We note that the right hand side always belongs to Mx[[ξ]]∗M_x[[\xi]]^*, see 4.8 and 4.7. We first show that the formula is well defined. Suppose m1/xn=m2/xlm_1/x^n=m_2/x^l. Then there exists r≥0r\geq0 such that xl+rm1=xn+rm2x^{l+r}m_1=x^{n+r}m_2. As MM is a Cartan SS-module we get P(ξ)(x)l+rQM(m1)=P(ξ)(x)n+rQM(m2)\mathcal{P}(\xi)(x)^{l+r}Q_M(m_1)=\mathcal{P}(\xi)(x)^{n+r}Q_M(m_2). So we obtain

P(ξ)(x)−nQM(ξ)(m1)=P(ξ)(x)−lQM(ξ)(m2).\mathcal{P}(\xi)(x)^{-n}Q_M(\xi)(m_1)=\mathcal{P}(\xi)(x)^{-l}Q_M(\xi)(m_2).

Thus we have a well-defined map QMx:Mx→Mx[[ξ]]∗Q_{M_x}: M_x \to M_x[[\xi]]^*. We note that

QMx(ξ)(mxn)0=P(ξ)(x)−nQM(ξ)(m)0=mxnQ_{M_x}(\xi)\left(\frac{m}{x^n}\right)_0=\mathcal{P}(\xi)(x)^{-n}Q_M(\xi)(m)_0=\frac{m}{x^n}

using the definition of Cartan modules and 4.8. Also note that QMx(ξ)Q_{M_x}(\xi) is homogeneous of degree zero. For u∈Su \in S we have

QMx(ξ)(umxn)=QMx(ξ)(umxn)=P(ξ)(x)−nQM(ξ)(um)=P(ξ)(x)−nP(ξ)(u)QM(ξ)(m)=P(ξ)(u)QMx(ξ)(mxn).\begin{aligned} Q_{M_x}(\xi)\left(u\frac{m}{x^n}\right)=Q_{M_x}(\xi)\left(\frac{um}{x^n}\right) \\ &=\mathcal{P}(\xi)(x)^{-n}Q_M(\xi)(um) \\ &=\mathcal{P}(\xi)(x)^{-n}\mathcal{P}(\xi)(u)Q_M(\xi)(m) \\ &=\mathcal{P}(\xi)(u)Q_{M_x}(\xi)\left(\frac{m}{x^n}\right). \end{aligned}

Thus QMx(ξ)Q_{M_x}(\xi) is a Steenrod map. The assertion follows from Theorem 4.6. □

Next we give

Proof of Theorem 5.2. We have nothing to show if Mx=0M_x=0 or Mxy=0M_{xy}=0. So assume that both are non-zero. So xx, xyxy (and hence yy) are non-nilpotent elements of SS. We let i:Mx→Mxyi:M_x \to M_{xy} denote the natural map i(m/xn)=ynm/(xy)ni(m/x^n)=y^nm/(xy)^n. This induces a natural map η:Mx[[ξ]]∗→Mxy[[ξ]]∗\eta:M_x[[\xi]]^* \to M_{xy}[[\xi]]^*. We have

QMxy(ξ)∘i(m/xn)=QMxy(ξ)(ynm/xnyn)=P(ξ)(xy)−nQM(ξ)(ynm)=P(ξ)(x)−nP(ξ)(y)−nP(ξ)(y)nQM(ξ)(m)=P(ξ)(x)−nQM(ξ)(m)=η∘QMx(ξ)(m/xn).\begin{aligned} Q_{M_{xy}}(\xi)\circ i(m/x^n)=Q_{M_{xy}}(\xi)(y^nm/x^ny^n) \\ &=\mathcal{P}(\xi)(xy)^{-n}Q_M(\xi)(y^nm) \\ &=\mathcal{P}(\xi)(x)^{-n}\mathcal{P}(\xi)(y)^{-n}\mathcal{P}(\xi)(y)^nQ_M(\xi)(m) \\ &=\mathcal{P}(\xi)(x)^{-n}Q_M(\xi)(m) \\ &=\eta\circ Q_{M_x}(\xi)(m/x^n). \end{aligned}

From this it easily follows that the map Mx→MxyM_x \to M_{xy} is a Cartan morphism. □

Proof of Theorem 1.2

We now give

Proof of Theorem 1.2. Let CM=0→C0→C1→⋯→Cm→0C_M=0\to C^0\to C^1\to\cdots\to C^m\to0 be the modified Čech complex on x\mathbf{x} with coefficients in MM, see [2 5.1.5]. Here

Ct=⨁1≤i1<i2<⋯<it≤mMxi1xi2⋯xit.C^t=\bigoplus_{1\leq i_1<i_2<\cdots<i_t\leq m}M_{x_{i_1}x_{i_2}\cdots x_{i_t}}.

where the differential ∂t:Ct→Ct+1\partial^t:C^t\to C^{t+1} is given by

(−1)s−1⋅nat⁡:Mxi1xi2⋯xit→(Mxi1xi2⋯xit)xjs(-1)^{s-1}\cdot\operatorname{nat}:M_{x_{i_1}x_{i_2}\cdots x_{i_t}}\to\left(M_{x_{i_1}x_{i_2}\cdots x_{i_t}}\right)_{x_{j_s}}

if {i1,…,it}={j1,…,js^,…,jt+1}\{i_1,\ldots,i_t\}=\{j_1,\ldots,\widehat{j_s},\ldots,j_{t+1}\} and 0 otherwise. It follows from 5.1 and 3.5 that CtC^t is a Cartan SS-module for all tt. By 5.2 and 3.6 it follows that ∂t\partial^t is a morphism of Cartan modules. By 3.8 it follows that HIi(M)H^i_I(M) has a structure of a Cartan SS-module. □

Remark 6.1. Suppose x\mathbf{x} and y\mathbf{y} be two different sets of homogeneous generators of II. Then we can give a Cartan SS-module structure on HIi(M)H^i_I(M) which depends on x\mathbf{x} and another Cartan SS-module structure which depends on y\mathbf{y}. We do not know whether these structures are the same. For our applications it suffices to know there is at least one Cartan SS-module structure on local cohomology modules.

Remark 6.2. We note that in our proof of Theorem 1.2 the Cartan module MM need not be finitely generated as an SS-module. As a corollary we get that for any homogeneous ideals I1,…,ItI_1,\ldots,I_t the iterated local cohomology module

HI1i1(HI2i2(⋯(HItit(M))⋯ ))H_{I_1}^{i_1}\left(H_{I_2}^{i_2}\left(\cdots\left(H_{I_t}^{i_t}(M)\right)\cdots\right)\right)

has a structure of a Cartan SS-module.

Remark 6.3. In the study of local cohomology the so called Lyubeznik functors plays an important role, see [6 p. 44]. Iterated local cohomology modules forms a large subclass of Lyubeznik functors. We do not know whether if T\mathcal{T} is a graded Lyubeznik functor on Mod⁡(S)\operatorname{Mod}(S) then T(S)\mathcal{T}(S) is a Cartan module. We however suspect that it is a Cartan SS-module.

6.4. Let EE be the ∗\ast-injective hull of Fq\mathbb{F}_q as an SS-module. Set (−)∨=∗Hom⁡S(−,E)(-)^\vee= {}^\ast\operatorname{Hom}_S(-,E). If UU is a finitely generated graded SS-module then U∨U^\vee is a ∗\ast-Artinian SS-module. If UU is a ∗\ast-Artinian SS-module then U∨U^\vee is a finitely generated graded SS-module. Furthermore the natural map U→U∨∨U \to U^{\vee\vee} is an isomorphism if either UU is finitely generated graded or when UU is a ∗\ast-Artinian SS-module.

6.5. If MM is a finitely generated graded SS-module then HS+i(M)H^i_{S_+}(M) are ∗\ast-Artinian SS-module for all i≥0i \ge0. Furthermore it is well known that dim⁡HS+i(M)∨\dim H^i_{S_+}(M)^\vee is ≤i\le i, for instance see [9 10.1].

Corollary 6.6. Let MM be a finitely generated graded Cartan SS-module. Set Ji=ann⁡HS+i(M)J_i=\operatorname{ann} H^i_{S_+}(M). Then dd,0q,…,dd,d−i−1q∈Jid^q_{d,0},\ldots,d^q_{d,d-i-1}\in\sqrt{J_i} for i<di<d.

Proof. By 1.2 we get that HS+i(M)H^i_{S_+}(M) is a Cartan-module. So JiJ_i is a P∗\mathcal{P}^\ast-invariant ideal of SS. Therefore Ji\sqrt{J_i} is also a P∗\mathcal{P}^\ast-invariant ideal of SS, (see 3.9). Thus Ji=ann⁡HS+i(M)∨J_i=\operatorname{ann} H^i_{S_+}(M)^\vee and the later module has dimension ≤i\le i, see 6.5. The result follows from 3.12 □\square

Proof of Theorem 1.4

We make the convention that the depth of the zero module i +∞+\infty. In this section we give

Proof of Theorem 1.4. The assertion (ii) ⟹\Longrightarrow (i) is trivial. We prove the converse. We note that dd,0q,…,dd,d−1qd^q_{d,0},\ldots,d^q_{d,d-1} is a system of parameters of SS. So if MM is Cohen-Macaulay then we have nothing to prove.

Assume depth⁡M=d−g\operatorname{depth} M=d-g where g>0g>0. Let D=D∗(d)qD=D^\ast(d)^{q} be the (fractal) Dickson algebra. We consider MM as a DD-module. Set J=(dd,d−1q,…,dd,gq)J=(d^q_{d,d-1},\ldots,d^q_{d,g}). We note that it suffices to prove depth⁡MP≥d−g\operatorname{depth} M_P\ge d-g for all primes PP in DD which contain JJ, see [3 1.2.10]. The assertion holds if P=D+P=D_+ by our assumption. Now let P⊇JP\supseteq J and P≠D+P\ne D_+. Let D(u)D(u) be the ∗\ast-canonical module of DD. We note that dim⁡HS+i(M)∨=Ext⁡Dd−i(M,D(u))\dim H^i_{S_+}(M)^\vee=\operatorname{Ext}^{d-i}_D(M,D(u)) has dimension ≤i\le i (see 6.6). Set Kd−i=ann⁡DExt⁡Dd−i(S,D(u))K_{d-i}=\operatorname{ann}_D\operatorname{Ext}^{d-i}_D(S,D(u)). So dd,0q,…,dd,d−i−1q∈Kd−id^q_{d,0},\ldots,d^q_{d,d-i-1}\in\sqrt{K_{d-i}} by 6.6 for i<di<d.

Let P⊇JP\supseteq J be minimal. Then P=JP=J. As dd,0q∉Pd^q_{d,0}\notin P we get that Ext⁡Di(M,D(u))P=0\operatorname{Ext}^i_D(M,D(u))_P=0 for i>0i>0. So MPM_P is either Cohen-Macaulay of depth =d−g=d-g or MPM_P is zero.

Let P⊇JP\supseteq J be such that height⁡P=d−g+r\operatorname{height} P=d-g+r with r≥1r\ge1 and that P≠D+P\ne D_+. We assert that depth⁡MP≥d−g\operatorname{depth} M_P\ge d-g. If not then Ext⁡Dd−g+r−(d−g−i)(M,D(u))P≠0\operatorname{Ext}^{d-g+r-(d-g-i)}_D(M,D(u))_P\ne0 for some i>0i > 0. So we have Ext⁡r+i(M,D(u))P≠0\operatorname{Ext}^{r+i}(M,D(u))_{P} \ne0. We have U=Ext⁡r+i(M,D(u))=HS+d−r−i(M)∨U=\operatorname{Ext}^{r+i}(M,D(u))=H^{d-r-i}_{S_{+}}(M)^{\vee}. So dd,0qt,⋯ ,dd,r+i−1qt∈ann⁡DUd_{d,0}^{q^{t}},\cdots,d_{d,r+i-1}^{q^{t}}\in\sqrt{\operatorname{ann}_{D}U} by 6.6. As UP≠0U_{P}\ne0 we get that dd,0qt,⋯ ,dd,r+i−1qt∈Pd_{d,0}^{q^{t}},\cdots,d_{d,r+i-1}^{q^{t}}\in P. If t+i−1≥gt+i-1\ge g then note that D+⊆PD_{+}\subseteq P which is a contradiction. So r+i−1<gr+i-1<g. Thus PP contains d−g+r+id-g+r+i variables where i≥1i\ge1. This is a contradiction as height⁡P=d−g+r\operatorname{height}P=d-g+r. The result follows. □

Proof of Theorem 1.5

8.1. In this section we give a proof of Theorem 1.5. Throughout SS is an unstable Fq\mathbb{F}_{q}-Steenrod algebra and MM is a Cartan SS-module. The following result follows from Theorem 1.2. We give a direct proof.

Lemma 8.2. (with hypotheses as in 8.1). Let II be a homogeneous ideal in SS. Then HI0(M)H^{0}_{I}(M) is a Cartan SS-submodule of MM.

Proof. Let m∈HI0(M)m\in H^{0}_{I}(M). We show that QMi(m)∈HI0(M)Q^{i}_{M}(m)\in H^{0}_{I}(M) for all i≥0i\ge0. Say Ism=0I^{s}m=0. Set J=IsJ=I^{s}.

We prove by induction on i≥0i\ge0 that u1u2⋯ui+1QMi(m)=0u_{1}u_{2}\cdots u_{i+1}Q^{i}_{M}(m)=0 for any uj∈Ju_{j}\in J. When i=0i=0 we have QM0(m)=mQ^{0}_{M}(m)=m. So u1m=0u_{1}m=0 by our hypothesis. Now assume the result for i=0,…,r−1i=0,\ldots,r-1 where r≥1r\ge1. We have u1m=0u_{1}m=0. So QMr(u1m)=0Q^{r}_{M}(u_{1}m)=0. By Cartan identity we have

0=QMr(u1m)=u1QMr(m)+∑i+j=rj<rPi(u1)QMj(m).0=Q^{r}_{M}(u_{1}m)=u_{1}Q^{r}_{M}(m)+\sum_{\substack{i+j=r\\j<r}}P^{i}(u_{1})Q^{j}_{M}(m).

Multiply the above equation by u2⋯ur+1u_{2}\cdots u_{r+1} to obtain u1u2⋯ur+1QMr(m)=0u_{1}u_{2}\cdots u_{r+1}Q^{r}_{M}(m)=0. This will complete the inductive step.

Note we get I(i+1)sQMi(m)=0I^{(i+1)s}Q^{i}_{M}(m)=0 for all i≥0i\ge0. Hence QMi(m)∈HI0(M)Q^{i}_{M}(m)\in H^{0}_{I}(M). □

We need the following version of a result in [4 3.13].

Proposition 8.3. Let RR be a Noetherian ring and let MM be a finitely generated RR-module. Let PP be a maximal element in Ass⁡RM\operatorname{Ass}_{R}M (with respect to inclusion). Then Ass⁡RHP0(M)={P}\operatorname{Ass}_{R}H^{0}_{P}(M)=\{P\} and Ass⁡RM/HP0(M)=Ass⁡RM∖{P}\operatorname{Ass}_{R}M/H^{0}_{P}(M)=\operatorname{Ass}_{R}M\setminus\{P\}.

We prove Theorem 1.5 by induction on ∣Ass⁡SM∣|\operatorname{Ass}_{S}M| (here ∣W∣|W| denotes the cardinality of a set WW). The following result enables our induction.

Proposition 8.4. (with setup as in 8.1). Further assume MM is a finitely generated SS-module. Let PP be a maximal element in Ass⁡SM\operatorname{Ass}_{S}M (with respect to inclusion). Then PP is a P∗P^{*}-invariant prime ideal of SS.

Proof. Set N=HP0(M)N=H^{0}_{P}(M). Then NN is a Cartan submodule of MM, see 8.2. So J=ann⁡SNJ=\operatorname{ann}_{S}N is a P∗P^{*}-invariant ideal of SS, see 3.10. So J\sqrt{J} is also a P∗P^{*}-invariant ideal of SS, see 3.9.

As NN is a finitely generated SS-module and PP-torsion we get that PsN=0P^{s}N=0 for some s≥1s\ge1. Thus Ps⊆JP^{s}\subseteq J. As P∈Ass⁡SNP\in\operatorname{Ass}_{S}N we get that P=(0:n)P=(0:n) for some n∈Nn\in N. We have Jn=0Jn=0. So J⊆PJ\subseteq P. We have Ps⊆J⊆PP^{s}\subseteq J\subseteq P. So P=JP=\sqrt{J} and hence a P∗P^{*}-invariant ideal of SS. □

We now give Proof of Theorem 1.5. We prove the result by induction on ∣Ass⁡SM∣|\operatorname{Ass}_S M|. We first consider the case when ∣Ass⁡SM∣=1|\operatorname{Ass}_S M|=1. Let {P}=Ass⁡SM\{P\}=\operatorname{Ass}_S M. Then trivially PP is a maximal element of Ass⁡SM\operatorname{Ass}_S M and so by 8.4 we get that PP is P∗P^*-invariant.

Now consider the case when ∣Ass⁡SM∣=r|\operatorname{Ass}_S M|=r and the result is known for all finitely generated Cartan SS-modules NN with ∣Ass⁡SN∣=r−1|\operatorname{Ass}_S N|=r-1. Let PP be a maximal element of Ass⁡SM\operatorname{Ass}_S M. Then by 8.4 we get PP is a P∗P^*-invariant prime ideal of SS. By 8.2 we get that HP0(M)H_P^0(M) is a Cartan submodule of MM. So N=M/HP0(M)N=M/H_P^0(M) is also a finitely generated Cartan SS-module. By 8.3 we get that Ass⁡SN=Ass⁡SM∖{P}\operatorname{Ass}_S N=\operatorname{Ass}_S M\setminus\{P\}. So ∣Ass⁡SN∣=r−1|\operatorname{Ass}_S N|=r-1 and by our induction hypotheses we get that QQ is a P∗P^*-invariant prime ideal of SS for every associate prime QQ of NN. Our assertion follows. □

We give an example which shows the result in Theorem 1.5 need not hold if MM is not a finitely generated Cartan SS-module.

Example 8.5. Let R=Fq[X1,…,Xd]R=\mathbb{F}_q[X_1,\ldots,X_d] with d≥2d\geq2 and the standard Steenrod algebra structure. As discussed earlier RR has infinitely many homogeneous prime ideals which are not P∗P^*-invariant. Choose one such prime PP with height g≥1g\geq1. Then M=HPg(S)M=H_P^g(S) is a not finitely generated RR-module with Ass⁡M={P}\operatorname{Ass} M=\{P\}. By 1.2 we get that MM is a Cartan SS-module.

Proof of Applications III

Proofs of result stated in our Application III

In this section we give all the proofs of result stated in our Application III. Throughout SS is a Fq\mathbb{F}_q-Steenrod algebra and MM is a finitely generated Cartan SS-module.

We will need the following result proved in [9 8.2] for invariant rings. The same proof works in general.

Lemma 9.2. Let MM be a Cartan SS-module. Assume PP is not a P∗P^*-invariant prime ideal. If Pr⊆ann⁡MP^r\subseteq\operatorname{ann} M then MP=0M_P=0.

The following result follows easily from Theorem 1.2 and 9.2.

Corollary 9.3. (with setup as in 9.1). Let PP be a prime ideal in SS which is not P∗P^*-invariant. If HPi(M)H_P^i(M) is a finitely generated module as an SS-module then HPi(M)P=0H_P^i(M)_P=0. □

The following result is crucial for our applications.

Theorem 9.4. (with setup as in 9.1). Assume further that SS is equi-dimensional and Supp⁡(M)=Spec⁡(S)\operatorname{Supp}(M)=\operatorname{Spec}(S). Let r≥0r\geq0. The following assertions are equivalent:

(i) depth⁡MP≥min⁡{height⁡P,r}\operatorname{depth} M_P\geq\min\{\operatorname{height} P,r\} for all homogeneous primes PP of SS.

(ii) depth⁡MP≥min⁡{height⁡P,r}\operatorname{depth} M_P\geq\min\{\operatorname{height} P,r\} for all homogeneous P∗P^*-invariant primes PP of SS.

Proof. The assertion (i) ⟹\Longrightarrow (ii) is trivial. We now prove (ii) ⟹\Longrightarrow (i). We have nothing to show when r=0r=0. So assume r≥1r\geq1.

Let PP be a homogeneous prime ideal of SS. Assume that height⁡P=t≥1\operatorname{height} P=t\geq1. Set

fP(M)=min⁡{i∣HPi(M) is not finitely generated S-module}.f_P(M)=\min\{i\mid H_P^i(M)\text{ is not finitely generated }S\text{-module}\}.

Set sP=min⁡{height⁡P,r}s_P=\min\{\operatorname{height} P,r\}. We make the following

Claim: If PP is a homogeneous prime ideal then fP(M)≥sPf_P(M)\geq s_P. We first prove that the Claim proves the result. Let PP be not a P∗\mathcal{P}^{*}-invariant homogeneous ideal. Then HPi(M)H^i_{\mathcal{P}}(M) are finitely generated SS-modules for i<sPi < s_{\mathcal{P}}. By 9.3 it follows that HPSPi(MP)=HPi(M)P=0H^i_{\mathcal{P}S_{P}}(M_{P})=H^i_{\mathcal{P}}(M)_{P}=0 for i<sPi < s_{P}. Thus depth MP≥sPM_{P} \ge s_{P}. The result follows.

We note that the assertion fP(M)≥sPf_{P}(M) \ge s_{P} holds trivially when r=1r=1.

Next consider the case when r≥2r \ge2 and the result holds for ≤r−1\le r-1. By induction hypothesis, fP(M)≥min⁡{height⁡P,r−1}f_{P}(M) \ge\min\{\operatorname{height} P,r-1\} for all homogeneous primes. In particular depth Mq≥min⁡{height⁡q,r−1}M_{q} \ge\min\{\operatorname{height} q,r-1\} for all homogeneous non P∗\mathcal{P}^{*}-invariant prime ideals qq of SS. We also have depth MQ≥sQM_{Q} \ge s_{Q} for all P∗\mathcal{P}^{*}-invariant prime ideals QQ by our assumption. We may assume height P≥rP \ge r, otherwise there is nothing to show. By [2 13.1.17] we have

fP(M)=min⁡{depth⁡Mq+height⁡((P+q)/q)∣q∈∗Spec⁡(S)∖V(P)}.f_{P}(M)=\min\{\operatorname{depth} M_{q}+\operatorname{height}((P+q)/q)\mid q\in{}^{*}\operatorname{Spec}(S)\setminus V(P)\}.

If height q≤r−1q\le r-1 then depth Mq=height⁡qM_{q}=\operatorname{height}q. We have

depth⁡Mq+height⁡((P+q)/q)=height⁡q+height⁡((P+q)/P)=height⁡q+height⁡(P+q)−height⁡q,=height⁡(P+q)≥r.\begin{aligned} \operatorname{depth} M_{q}+\operatorname{height}((P+q)/q)=\operatorname{height}q+\operatorname{height}((P+q)/P) \\ &=\operatorname{height}q+\operatorname{height}(P+q)-\operatorname{height}q, \\ &=\operatorname{height}(P+q)\ge r. \end{aligned}

The second equality holds as SS is equi-dimensional and catenary, see [7 Lemma 2, p. 250] for the local version; same proof works for the ∗*-local case. If height q≥rq\ge r then

depth⁡Mq+height⁡((P+q)/q)≥r−1+1=r.\operatorname{depth} M_{q}+\operatorname{height}((P+q)/q)\ge r-1+1=r.

Thus fP(M)≥rf_{P}(M)\ge r. So Claim follows. □

Surprisingly the SrS_{r} property of a Cartan-module only depends on P∗\mathcal{P}^{*}-invariant prime ideals in the following sense.

Theorem 9.5. (with hypotheses as in 9.4). The following conditions are equivalent:

(i) MM satisfies SrS_{r} property.

(ii) depth MP≥min⁡{height⁡P,r}M_{P}\ge\min\{\operatorname{height}P,r\} for all P∗\mathcal{P}^{*}-invariant homogeneous primes PP of SS.

Proof. The assertion (i) ⟹\Longrightarrow (ii) is clear.

We now prove (ii) ⟹\Longrightarrow (i). It is well-known that it suffices to prove depth MP≥min⁡{height⁡P,r}M_{P}\ge\min\{\operatorname{height}P,r\} for homogeneous primes in SS, for instance see [9 10.3] for the case when M=SM=S. By 9.4 it follows that depth MP≥min⁡{height⁡P,r}M_{P}\ge\min\{\operatorname{height}P,r\} for all homogeneous primes PP of SS. The result follows. □

Next we give

Proof of Theorem 1.6. We first prove (ii) ⟹\Longrightarrow (i). Let PP be a graded P∗\mathcal{P}^{*}-invariant prime ideal of SS. We note that if height P=gP=g then dim⁡S/P=d−g\dim S/P=d-g, as SS is equi-dimensional and universally catenary, see [7 Lemma 2, p. 250] for the local version; same proof works for the ∗*-local case. By 3.12 we get that if g≤rg\le r then PP contains dd,0qt,…,dd,g−1qtd_{d,0}^{q^{t}},\ldots,d_{d,g-1}^{q^{t}}. It follows that MPM_{P} is Cohen-Macaulay. If height P≥r+1P\ge r+1 then by 3.12 we get that PP contains dd,0,…,dd,r−1d_{d,0},\ldots,d_{d,r-1}. So depth MP≥rM_{P}\ge r. The result follows from 9.5.

Next we prove (i) ⟹\Longrightarrow (ii). The extension D∗(d)qt⊆SD^{*}(d)^{q^{t}}\subseteq S is a finite extension. As (dd,0qt,…,dd,r−1qt)(d_{d,0}^{q^{t}},\ldots,d_{d,r-1}^{q^{t}}) is a prime ideal of height rr in D∗(d)D^{*}(d), it follows that (dd,0qt,…,dd,r−1qt)S(d_{d,0}^{q^t},\ldots,d_{d,r-1}^{q^t})S has height rr as an ideal in SS. As MM satisfies SrS_r it follows from [3 1.2.10] that grade⁡((dd,0,…,dd,r−1)S,M)=r\operatorname{grade}((d_{d,0},\ldots,d_{d,r-1})S,M)=r. The result follows. □\square

Proof of Corollary 1.9

In this section we give

Proof of Corollary 1.9. By 1.10 we may assume dim⁡M=dim⁡S=d\dim M=\dim S=d. We work with the setup as in 1.3. Set D=D(d)qtD=D(d)^{q^t}. Let D(u)D(u) be the canonical module of DD. By local duality we get that Hmi(M)∨=Ext⁡Dd−i(M,D(u))H_{\mathfrak{m}}^{i}(M)^{\vee}=\operatorname{Ext}_{D}^{d-i}(M,D(u)). So dd,0qt∈ann⁡Ext⁡Dd−i(M,D(u))d_{d,0}^{q^t}\in\sqrt{\operatorname{ann}\operatorname{Ext}_{D}^{d-i}(M,D(u))} for i<di<d. We note that we have a finite extension Dd,0qt↪Sd,0qtD_{d,0}^{q^t}\hookrightarrow S_{d,0}^{q^t}. So it suffices to show that Md,0qtM_{d,0}^{q^t} is Cohen-Macaulay as a Dd,0qtD_{d,0}^{q^t}-module. Let PP be any homogeneous prime in DD not containing dd,0qtd_{d,0}^{q^t}. Then it follows that Ext⁡Di(M,D(u))P=0\operatorname{Ext}_{D}^{i}(M,D(u))_P=0 for i>0i>0. It follows that MPM_P is Cohen-Macaulay. The result follows. □\square

Proof of Theorem 1.11

In this section we prove Theorem 1.11. To prove this result we need Theorem 23.3 from [7]. Unfortunately there is a typographical error in the statement of Theorem 23.3 in [7]. So we state it here.

Theorem 11.1. Let φ:A→B\varphi:A\to B be a homomorphism of Noetherian rings, and let EE be an AA-module and GG a BB-module. Suppose that GG is flat over AA; then we have the following:

(i) if p∈Spec⁡A\mathfrak{p}\in\operatorname{Spec} A and G/pG≠0G/\mathfrak{p}G\ne0 then

aφ(Ass⁡B(G/pG))=Ass⁡A(G/pG)={p}.a_{\varphi}\bigl(\operatorname{Ass}_{B}(G/\mathfrak{p}G)\bigr)=\operatorname{Ass}_{A}(G/\mathfrak{p}G)=\{\mathfrak{p}\}.

(ii) Ass⁡B(E⊗AG)=⋃p∈Ass⁡A(E)Ass⁡B(G/pG)\operatorname{Ass}_{B}(E\otimes_A G)=\bigcup_{\mathfrak{p}\in\operatorname{Ass}_{A}(E)}\operatorname{Ass}_{B}(G/\mathfrak{p}G).

Remark 11.2. In [7] Ass⁡A(E⊗G)\operatorname{Ass}_{A}(E\otimes G) is typed instead of Ass⁡B(E⊗G)\operatorname{Ass}_{B}(E\otimes G). Also note that φ(P)=P∩A\varphi(\mathfrak{P})=\mathfrak{P}\cap A for P∈Spec⁡B\mathfrak{P}\in\operatorname{Spec} B.

Proof. The inclusion R→TR\to T is faithfully flat. Also note that as II is a monomial ideal it is extended from RR. Say I=LTI=LT for a monomial ideal LL of RR. Set N=Hmj(R/L)N=H_{\mathfrak{m}}^{j}(R/L). Then M=N⊗RTM=N\otimes_R T. As R→TR\to T is faithfully flat we get that NN is a finitely generated RR-module. By 2.3 we get that LL is a P∗P^*-invariant ideal of RR. So R/LR/L is a Cartan RR-module. By 1.2 we get that NN is a Cartan RR-module. It is also finitely generated as an RR-module. Thus if P∈Ass⁡RNP\in\operatorname{Ass}_{R}N then PP is a P∗P^*-invariant ideal in RR. So by [10 11.4.1] we get that PP is generated by some linear homogeneous polynomials of RR. In 11.1 we take A=RA=R, B=TB=T and G=TG=T. We note that if P∈Spec⁡(R)P\in\operatorname{Spec}(R) is homogeneous and generated by linear polynomials then PTPT is a prime in TT. The result follows from 11.1. □\square

Proof of Theorem 1.12

In this section we give Proof of Theorem 1.12. Set t=depth⁡SPt = \operatorname{depth} S_P. Suppose if possible fP>tf_P > t. Then note that HPt(S)H_P^t(S) is finitely generated SS-module. Then by 9.3 we get that HPt(S)P=0H_P^t(S)_P = 0. But HPSPt(SP)=HPt(S)P=0H_{P S_P}^t(S_P) = H_P^t(S)_P = 0. This is a contradiction as t=depth⁡SPt = \operatorname{depth} S_P. □

We give an example which shows that the bound in Theorem 1.12 need not hold for P∗P^*-invariant ideals.

Example 12.1. Let II be a monomial ideal in R=Fq[X1,…,Xn]R = \mathbb{F}_q[X_1,\ldots,X_n] such that S=R/IS = R/I is generalized Cohen-Macaulay but not Cohen-Macaulay. We note that SS is an unstable Steenrod Fq\mathbb{F}_q-algebra, see 2.3. We note that fS+=d=dim⁡Sf_{S_+} = d = \dim S while depth⁡SS+<dim⁡SS+=d\operatorname{depth} S_{S_+} < \dim S_{S_+} = d. For a specific example see [11 3.5].

Department of Mathematics, Indian Institute of Technology Bombay, Powai, Mumbai 400 076, India

Email address: [email protected]

References

  1. [1]D. Bourguiba and S. Zarati, Depth and the Steenrod algebra, With an appendix by J. Lannes. Invent. Math. 128 (1997), no. 3, 589–602.DOI
  2. [2]M. Brodmann and R. Y. Sharp, Local Cohomology: An algebraic introduction with geometric applications, Cambridge University Press, 60, 1998.DOI
  3. [3]W. Bruns and J. Herzog, Cohen-Macaulay Rings, revised edition, Cambridge Studies in Advanced Mathematics, 39. Cambridge University Press, 1998.
  4. [4]D. Eisenbud, Commutative algebra, With a view toward algebraic geometry, Grad. Texts in Math., 150 Springer-Verlag, New York, 1995.
  5. [5]P. S. Landweber and R. E. Stong, The depth of rings of invariants over finite fields, Number theory (New York, 1984–1985), 259–274. Lecture Notes in Math., 1240 Springer-Verlag, Berlin, 1987.DOI
  6. [6]G. Lyubeznik, Finiteness properties of local cohomology modules (an application of D - modules to commutative algebra), Invent. Math. 113 (1993), no. 1, 41–55.DOI
  7. [7]H. Matsumura, Commutative ring theory, Cambridge Studies in Advanced Mathematics, vol. 8, Cambridge University Press, Cambridge, 1986.DOI
  8. [8]M. D. Neusel and L. Smith, Invariant theory of finite groups, Math. Surveys Monogr., 94 American Mathematical Society, Providence, RI, 2002.
  9. [9]T. J. Puthenpurakal, Annihilators of local cohomology modules over modular invariant rings and Dickson polynomials. Preprint, arXiv:2603.16491.
  10. [10]L. Smith, Polynomial invariants of finite groups, Res. Notes Math., 6 A K Peters, Ltd., Wellesley, MA, 1995.DOI
  11. [11]N. Terai and K. Yoshida, Buchsbaum Stanley-Reisner rings with minimal multiplicity, Proc. Amer. Math. Soc. 134 (2006), no. 1, 55–65.DOI

Paper details

vlnc:W_hjeg7bms
v1 · latestHistory (1)

published · Fresh — added to the Hub within the last 30 days

added ·

Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

CC BY 4.0

Original source

  • Mathematics

Not recorded for this paper · View statement →

Looking for related papers…
Contents