betaFeedback

WorksMathematics

On generalizations of the Cohen structure theorem

Abstract

The Cohen structure theorem is an important tool in commutative algebra and algebraic geometry. Recent work of Caviglia, Goswami, and Mesiti addresses the related question of which unital rings admit a maximal ideal for which the natural projection admits a section; for such rings, they prove a structure theorem in terms of semidirect products of rings. In this paper, we extract the underlying mechanism of their structure theorem to address when unital rings admit a nonzero proper ideal whose associated projection map has this property. We also address other settings involving ideals with conditions which are determined by ring theoretic properties of the corresponding quotient.

Introduction

Let RR be a unital commutative complete Noetherian local ring with maximal ideal m\mathfrak{m}. We say that the ring RR is equicharacteristic if RR has the same characteristic as its residue field R/mR/\mathfrak{m}. The Cohen structure theorem, which has numerous applications in commutative algebra and algebraic geometry (see e.g. [Zar47] and [2], along with the introduction of [CGM26] for more discussion on uses of the Cohen structure theorem), states that RR has a subfield FF such that the image of FF under the natural projection map R→R/mR \to R/\mathfrak{m} is R/mR/\mathfrak{m}; this property is equivalent to the natural projection map admitting a section.

Recent work of Caviglia, Goswami, and Mesiti [CGM26] examines when unital (not necessarily commutative) rings have a maximal two-sided ideal whose associated projection map admits a section. Among their main results is a structure theorem for such rings in terms of semidirect products of rings, [CGM26, Theorem 3.2] (see Section 2 for details on semidirect products of rings; these are the only semidirect products we consider) which was inspired by a communication between the second author and George Janelidze. It is natural to consider whether the underlying mechanism of this result can be adapted to other settings.

What is the importance of the condition of maximality in [CGM26, Theorem 3.2]? The main point of this paper is that, in this discourse, maximality is not at all important; We first pivot away from considering maximal ideals by examining unital rings RR which satisfy the following property.

Property 1.1. There exists a nonzero proper ideal II such that the natural projection map R→R/IR \to R/I admits a section.

Remark. All of the ideals that we consider are two-sided.

Our main result gives a structure theorem for unital rings which satisfy Property 1.1 in terms of semidirect products.

Theorem 1.2. Let RR be a unital ring.

(1) Suppose that R=R1×R2R = R_1 \times R_2 for rings R1R_1 and R2R_2 such that R1R_1 and R2R_2 are nonzero. Then RR satisfies Property 1.1 with an ideal II such that I≅R1I \cong R_1.

(2) Suppose that RR satisfies Property 1.1 with an ideal II and section s:R/I→Rs : R/I \to R. Then R≅I⋊s(R/I)R \cong I \rtimes s(R/I).

Remark. (1) Defining the semidirect product I⋊s(R/I)I \rtimes s(R/I) requires specifying two structure maps λ,ρ:s(R/I)×I→I\lambda,\rho: s(R/I) \times I \to I. As will be discussed in § 3, these maps will essentially be multiplication inside of RR.

(2) The proof of the theorem shows that one may prove similar results when one places conditions on the ideal which may, roughly speaking, be determined by the ring theoretic properties of the corresponding quotient. For example, a proper ideal II is a maximal left ideal if and only if R/IR/I is a division ring. As a porism to the proof of Theorem 1.2, if we add the condition of II being a maximal left ideal in Property 1.1, then the result holds when we further assume in part (1) that R2R_2 is a division ring; in part (2), we can further conclude that s(R/I)s(R/I) is a division ring. In other words, having Property 1.1 with the added condition of II being a maximal left ideal is tantamount to saying that RR is a semidirect product of a nonzero ring and a division ring. If RR is commutative, then similar remarks apply when we further assume that II is a prime ideal in Property 1.1 and that R2R_2 is an integral domain, or if we further assume that II is a maximal ideal in Property 1.1 and that R2R_2 is a field.

(3) Each of the isomorphisms in Theorem 1.2 and its proof are ring isomorphisms.

The paper is organized as follows. In the next section, we provide background on semidirect products of rings. In Section 3 we prove our main result by leveraging the central core of the proof of [CGM26, Theorem 3.2] in this domain. The proof of part (1) leverages the projection map onto the second factor; its kernel is an ideal which witnesses the fact that Property 1.1 holds. The proof of part (2) proceeds by considering the natural map which identifies the abelian group I⊕s(R/I)I \oplus s(R/I) with a subgroup of RR (the map Φ:I⋊s(R/I)→R\Phi: I \rtimes s(R/I) \to R below). The semidirect product structure on I⊕s(R/I)I \oplus s(R/I) upgrades Φ\Phi to a ring homomorphism. The fact that ss is a section allows one to produce an inverse Ψ:R→I⋊s(R/I)\Psi: R \to I \rtimes s(R/I) to Φ\Phi by splitting r∈Rr \in R into its “II-component” r−s([r])r-s([r]) and its “s(R/I)s(R/I)-component” s([r])s([r]). We also discuss how to prove this result in the other settings mentioned in part (2) of the remark following Theorem 1.2.

Background/Conventions

Throughout the paper, we assume all the rings which we consider are associative but not necessarily unital. If RR is a ring, we write RRR_R (respective RR_R R) when we consider RR as a right RR-module (respectively left RR-module) over itself in the natural way. For an RR-module MM (which may be left or right), we write End⁡R-Mod⁡(M)\operatorname{End}_{R\text{-}\operatorname{Mod}}(M) for the ring of RR-module endomorphisms of MM.

We recall the notion of a semidirect product of rings.

Definition. Let RR and TT be rings. Let λ:R→End⁡T-Mod⁡(TT)\lambda: R \to\operatorname{End}_{T\text{-}\operatorname{Mod}}({}_T T) be a ring homomorphism and ρ:R→End⁡T-Mod⁡(TT)\rho: R \to\operatorname{End}_{T\text{-}\operatorname{Mod}}(T_T) be a ring anti-homomorphism such that, for every s,t∈Rs,t \in R and x,y∈Tx,y \in T, we have

λ(s)∘ρ(t)=ρ(t)∘λ(s)\lambda(s) \circ\rho(t) = \rho(t) \circ\lambda(s)
ρ(s)(x)y=xλ(s)(y).\rho(s)(x)y = x\lambda(s)(y).

The semidirect product T⋊λ,ρRT \rtimes_{\lambda,\rho} R with respect to λ\lambda and ρ\rho is defined as follows. The underlying abelian group is given by T⊕RT \oplus R and we define the product of (t1,r1),(t2,r2)∈T⊕R(t_1,r_1),(t_2,r_2) \in T \oplus R via

(t1,r1)⋅(t2,r2):=(t1t2+λ(r1)(t2)+ρ(r2)(t1),r1r2).(t_1,r_1)\cdot(t_2,r_2) := (t_1t_2+\lambda(r_1)(t_2)+\rho(r_2)(t_1),r_1r_2).

Remark. It is straightforward to check that the semidirect product is a ring. We also have ring homomorphisms iT:T→T⋊λ,ρRi_T : T \to T \rtimes_{\lambda,\rho} R and iR:R→T⋊λ,ρRi_R : R \to T \rtimes_{\lambda,\rho} R given by iT(t)=(t,0)i_T(t)=(t,0) and iR(r)=(0,r)i_R(r)=(0,r).

If RR is a ring and I⊆RI \subseteq R is an ideal, we denote the image of r∈Rr \in R under the natural projection map R→R/IR \to R/I by [r][r].

Proof of Theorem 1.2/generalizations

Proof of Theorem 1.2

Proof of Theorem 1.2. We begin the proof of part (1). Consider the map

ϕ:R1⋉λ,ρR2→R2\phi: R_{1} \ltimes_{\lambda,\rho} R_{2} \to R_{2}

given by ϕ(r1,r2)=r2\phi(r_{1},r_{2})=r_{2}. We have

ker⁡(ϕ)={(r1,0):r1∈R1}≅R1;\ker(\phi)=\{(r_{1},0):r_{1}\in R_{1}\}\cong R_{1};

henceforth, we identify ker⁡(ϕ)\ker(\phi) with R1R_{1}. With this notation, we have

R1⋉λ,ρR2/R1≅R2.(1)R_{1}\ltimes_{\lambda,\rho}R_{2}/R_{1}\cong R_{2}. \tag*{(1)}

Composing this map with iR2i_{R_{2}} provides a section s:R1⋉λ,ρR2/R1→R1⋉λ,ρR2s:R_{1}\ltimes_{\lambda,\rho}R_{2}/R_{1}\to R_{1}\ltimes_{\lambda,\rho}R_{2} to the natural projection map R1⋉λ,ρR2→R1⋉λ,ρR2/R1R_{1}\ltimes_{\lambda,\rho}R_{2}\to R_{1}\ltimes_{\lambda,\rho}R_{2}/R_{1}. Indeed, we have s([(x,u)])=(0,u)s([(x,u)])=(0,u), and [(0,u)]=[(x,u)][(0,u)]=[(x,u)], so part (1) holds.

We now prove part (2). It is simple to check that the map λ:s(R/I)→End⁡R-Mod(I)(I)\lambda:s(R/I)\to\operatorname{End}_{R\text{-Mod}(I)}(I) given by λ(x)(b)=xb\lambda(x)(b)=xb is a ring homomorphism and that the map ρ:s(R/I)→End⁡R-Mod(I)(I)\rho:s(R/I)\to\operatorname{End}_{R\text{-Mod}(I)}(I) given by ρ(x)=bx\rho(x)=bx is a ring anti-homomorphism (and also that, for each x∈s(R/I)x\in s(R/I), the maps λ(x):I→I\lambda(x):I\to I and ρ(x):I→I\rho(x):I\to I are RR-module endomorphisms) satisfying the conditions (2.1) and (2.2); henceforth, we suppress the subscripts by writing I⋉s(R/I)I\ltimes s(R/I).

We show that I⋉s(R/I)≅RI\ltimes s(R/I)\cong R. Consider the map

Φ:I⋉s(R/I)→R\Phi:I\ltimes s(R/I)\to R

given by Φ(r1,r2)=r1+r2\Phi(r_{1},r_{2})=r_{1}+r_{2}. It is not hard to check that Φ\Phi is a ring homomorphism.

Consider now the map Ψ:R→I⋉s(R/I)\Psi:R\to I\ltimes s(R/I) given by Ψ(r)=(r−s([r]),s([r]))\Psi(r)=(r-s([r]),s([r])) (since [r−s([r])]=[r]−[s([r])]=[r]−[r]=[0][r-s([r])]=[r]-[s([r])]=[r]-[r]=[0], we have r−s([r])∈Ir-s([r])\in I). It is straightforward to check that Ψ\Psi is a ring homomorphism.

We now show that Φ\Phi and Ψ\Psi are inverses to each other. We have

Φ(Ψ(r))=Φ(r−s([r]),s([r]))=r.\Phi(\Psi(r))=\Phi(r-s([r]),s([r]))=r.

To show that Ψ∘Φ=id⁡\Psi\circ\Phi=\operatorname{id}, observe that

Ψ(Φ(r1,r2))=(r1+r2−s([r1+r2]),s([r1+r2])).\Psi(\Phi(r_{1},r_{2}))=(r_{1}+r_{2}-s([r_{1}+r_{2}]),s([r_{1}+r_{2}])).

so we need only show that s([r1+r2])=r2s([r_{1}+r_{2}])=r_{2}. Observe that r2=s([r2])r_{2}=s([r_{2}]), and since ss is injective by virtue of being a section, it now suffices to show that [r1+r2]=[r2][r_{1}+r_{2}]=[r_{2}], which is clear since r1∈Ir_{1}\in I. The result follows.

Other settings

We can refine the above proof in several ways. What follows is largely the same argument as in Section 3.1; we sketch the details and keep the same notation.

Recall that a proper ideal in a unital ring is a maximal left ideal if and only if the corresponding quotient is a division ring. In part (1) of Theorem 1.2, if we further assume that R2R_{2} is a division ring, then ker⁡(ϕ)≅R1\ker(\phi)\cong R_{1} is a left maximal ideal in RR for which the natural projection map has a section via the map in (3.1). If we assume in part (2) that II is a maximal left ideal, then the isomorphism argument still works as above, but we further see that s(R/I)s(R/I) is a division ring.

If we assume that RR is commutative, then the same argument applies when we add the condition in Property 1.1 that II is prime and assume that R2R_{2} is an integral domain.

AI statement

The author did not use generative AI in the production of his arguments nor in their exposition. The author used ChatGPT to proofread the paper.

Department of Mathematics, Clemson University, Clemson, SC 29634

Email address: [email protected]

References

  1. [2]Shiro Goto and Kenji Nishida, Finite modules of finite injective dimension over a Noetherian algebra, J. London Math. Soc. (2) 63 (2001), no. 2, 319–335. MR 1810132DOI

Paper details

vlnc:W_suy3g5na
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 10

downloads · 0

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

  • Mathematics

ChatGPT

View statement
Looking for related papers…
Contents