betaFeedback

WorksMathematics

A divisorial domain with non-Prüfer normalization

Abstract

We answer a 1968 question of Heinzer by constructing a two-dimensional local divisorial domain whose normalization is not Prüfer.

Heinzer's question and its status in the literature

Let DD be a domain with fraction field KK. For DD-submodules UU, VV of KK, write (U:V)={z∈K∣zV⊆U}(U : V) = \{z \in K \mid zV \subseteq U\}. The divisorial closure of a nonzero fractional ideal II is Iv=(D:(D:I))I_v = (D : (D : I)). The domain DD is divisorial if I=IvI = I_v for every such II. Its normalization D‾\overline{D} is its integral closure in KK. In 1968 Heinzer [2] asked whether D‾\overline{D} must be a Prüfer domain whenever DD is divisorial. A Prüfer domain is one in which each nonzero finitely generated ideal is invertible.

Two classical cases give an affirmative answer. An integrally closed divisorial domain is Prüfer [2 Theorem 5.1]. A Noetherian divisorial domain that is not a field has dimension one [2 Corollary 4.3], so its normalization is Dedekind by the Krull–Akizuki theorem. Gabelli, Houston and Picozza [1 Questions 3.7 and 3.8] restated Heinzer’s question and proved it equivalent to asking whether every ww-divisorial domain is a UMtUMt-domain. A domain is ww-divisorial if each of its ww-ideals is divisorial. In particular, every divisorial domain is ww-divisorial.

In 2026 Tran, Vo and Nguyen [6] answered a question of Vasconcelos by constructing a one-dimensional coherent local domain with non-Prüfer normalization. Kim [3 Theorem A] proved that the same domain is Gorenstein Prüfer. In particular, each of its nonzero finitely generated ideals is divisorial. Heinzer’s question requires the identity I=IvI = I_v also for infinitely generated ideals.

Theorem 1.1. For each prime pp, put k=Fp(t1,t2,…)k = \mathbb{F}_p(t_1, t_2, \ldots), where the tit_i are algebraically independent. There exists a two-dimensional local divisorial domain RR of characteristic pp with normalization S=k[[x,y]]S = k[[x, y]]. In particular, RR is not Prüfer, and RR fails the UMtUMt condition.

The construction uses a derivation to obtain a subring of a prescribed normal domain, an approach also used by Olberding [4]. Here the derivation takes values in the injective hull of the residue field of SS. This choice controls annihilators in every quotient R/cRR/cR by a nonzero nonunit.

A derivation with prescribed values

Write n=(x,y)S\mathfrak{n} = (x, y)S and let E=ES(k)E = E_S(k) be the injective hull of k=S/nk = S/\mathfrak{n}. For an ideal HH and a module MM over the same ring, put M[H]={u∈M∣Hu=0}M[H] = \{u \in M \mid Hu = 0\}, and write M[c]=M[(c)]M[c] = M[(c)].

Lemma 2.1. There is an Fp\mathbb{F}_p-derivation δ:S→E\delta: S \to E which is surjective as a map of additive groups.

Proof. The equality E=⋃j≥1E[nj]E = \bigcup_{j \ge1} E[\mathfrak{n}^j] holds over the Noetherian local ring SS [5 Lemma 47.7.3]. Each E[nj]=Hom⁡S(S/nj,E)E[\mathfrak{n}^j] = \operatorname{Hom}_S(S/\mathfrak{n}^j, E) has finite length. Indeed Hom⁡S(−,E)\operatorname{Hom}_S(-, E) is exact and sends the simple module kk to kk, the socle of EE. As kk is countable, so is EE. Enumerate its elements as e1,e2,…e_1, e_2, \ldots.

In the ordinary module of Kähler differentials ΩS/Fp\Omega_{S/\mathbb{F}_p}, the dtidt_i are SS-linearly independent. For each jj, the derivation ∂/∂tj\partial/\partial t_j of kk extends coefficientwise to SS: the product rule holds since each coefficient of a product is a finite sum. The corresponding SS-linear functional on ΩS/Fp\Omega_{S/F_p} takes dtidt_i to the Kronecker delta δij\delta_{ij}. On the free submodule ⨁iS dti\bigoplus_i S\,dt_i, send dtidt_i to eie_i. Injectivity of EE extends this map to all of ΩS/Fp\Omega_{S/F_p}. Its composite with the universal derivation has δ(ti)=ei\delta(t_i)=e_i, so its image contains every element of EE. □\square

Fix such a δ\delta and put R=ker⁡δR=\ker\delta. This is a local domain with maximal ideal m=R∩n\mathfrak m=R\cap\mathfrak n, since an element r∈Rr\in R invertible in SS satisfies δ(r−1)=−r−2δ(r)=0\delta(r^{-1})=-r^{-2}\delta(r)=0. In characteristic pp we have Sp⊆RS^p\subseteq R, so SS is integral over RR. For each s∈Ss\in S, a power of n\mathfrak n kills δ(s)\delta(s). Choose NN with xpNδ(s)=0x^{p^N}\delta(s)=0. Then xpNx^{p^N} and xpNsx^{p^N}s both belong to RR, so Frac⁡R=Frac⁡S\operatorname{Frac}R=\operatorname{Frac}S. Since SS is integrally closed, R‾=S\overline{R}=S. Integrality gives dim⁡R=dim⁡S=2\dim R=\dim S=2. In a local Prüfer domain every nonzero finitely generated ideal is principal, whereas dim⁡k(n/n2)=2\dim_k(\mathfrak n/\mathfrak n^2)=2. Thus SS is not Prüfer.

Divisoriality of all ideals

Lemma 3.1. Let JJ be a square-zero ideal of a commutative ring BB. Suppose A=B/JA=B/J is complete Noetherian local with residue field ℓ\ell and J≃EA(ℓ)J\simeq E_A(\ell) as an AA-module. Then B≃EB(ℓ)B\simeq E_B(\ell) as BB-modules. Every ideal LL of BB satisfies L=Ann⁡B(Ann⁡BL)L=\operatorname{Ann}_B(\operatorname{Ann}_B L).

Proof. Since JJ is nilpotent, BB is local with residue field ℓ\ell. Set W=EB(ℓ)W=E_B(\ell). Its submodule W[J]=Hom⁡B(A,W)W[J]=\operatorname{Hom}_B(A,W) is injective over AA by adjunction and contains ℓ\ell essentially. Hence it is EA(ℓ)E_A(\ell). Fix an isomorphism θ:J→W[J]\theta:J\to W[J] and extend it by injectivity to a BB-linear map f:B→Wf:B\to W.

Matlis duality identifies scalar multiplication A→End⁡A(J)A\to\operatorname{End}_A(J) with the completion map [5 Lemma 47.7.5], hence with an isomorphism. In particular JJ is faithful over AA. If f(b)=0f(b)=0, then θ(bj)=jf(b)=0\theta(bj)=jf(b)=0 for every j∈Jj\in J. Faithfulness gives b∈Jb\in J, where ff is injective. To see surjectivity, take w∈Ww\in W. The map j↦jwj\mapsto jw has image in W[J]W[J] because J2=0J^2=0. Composing with θ−1\theta^{-1} gives an endomorphism of JJ, hence multiplication by some bˉ∈A\bar b\in A. Lift it to b∈Bb\in B. Then jw=θ(bj)=jf(b)jw=\theta(bj)=jf(b) for every jj, so w−f(b)∈W[J]=f(J)w-f(b)\in W[J]=f(J).

Finally, if b∉Lb\notin L, the nonzero cyclic submodule generated by b+Lb+L in B/LB/L has a quotient ℓ\ell in which its generator survives. Embed ℓ\ell in WW and extend the resulting homomorphism to B/LB/L. Under W≃BW\simeq B, the image of 1+L1+L annihilates LL and has nonzero product with bb. This proves the double-annihilator identity. □\square

Proposition 3.2. For 0≠c∈m0\ne c\in\mathfrak m, the map B=R/cR→A=S/cSB=R/cR\to A=S/cS is onto. Its kernel JJ is square-zero and is isomorphic to EA(k)E_A(k) as an AA-module.

Proof. Multiplication by cc on EE is onto. For any e∈Ee\in E, the SS-linear map cS→EcS\to E sending cc to ee extends to SS, as cc is regular and EE is injective. Lemma 2.1 therefore gives, for any s∈Ss\in S, some t∈St\in S with cδ(t)=δ(s)c\delta(t)=\delta(s). Then s−ct∈Rs-ct\in R, proving surjectivity.

The kernel is J=(cS∩R)/cRJ=(cS\cap R)/cR. If cs,ct∈Rcs,ct\in R, then cδ(s)=cδ(t)=0c\delta(s)=c\delta(t)=0, whence δ(cst)=0\delta(cst)=0 and (cs)(ct)∈cR(cs)(ct)\in cR. Thus J2=0J^2=0. The assignment

J⟶E[c],cs+cR⟼δ(s)J\longrightarrow E[c],\qquad cs+cR\longmapsto\delta(s)

is well-defined and injective because ker⁡δ=R\ker\delta=R and cc is regular. It is onto because δ(S)=E\delta(S)=E, and cs∈Rcs\in R whenever cδ(s)=0c\delta(s)=0. It is AA-linear since R→AR\to A is onto and δ(rs)=rδ(s)\delta(rs)=r\delta(s) for r∈Rr\in R. Finally E[c]=Hom⁡S(A,E)E[c]=\operatorname{Hom}_S(A,E) is injective over AA and contains kk essentially, so it is EA(k)E_A(k). □\square

Let II be any nonzero proper ideal of RR, and choose 0≠c∈I0\ne c\in I. The ring S/cSS/cS is complete Noetherian local, so Lemma 3.1 applies to B=R/cRB=R/cR by Proposition 3.2. For cR⊆H⊆RcR\subseteq H\subseteq R the fractional colon (cR:H)(cR:H) lies in RR, because c∈Hc\in H, and contains cRcR. Consequently

Ann⁡B(H/cR)=(cR:H)/cR.\operatorname{Ann}_B(H/cR)=(cR:H)/cR.

Applying the double-annihilator identity to I/cRI/cR yields

I=(cR:(cR:I))=(cR:c(R:I))=(R:(R:I))=Iv.I=(cR:(cR:I))=(cR:c(R:I))=(R:(R:I))=I_v.

The identity is immediate for I=RI=R. Clearing a denominator extends it to every nonzero fractional ideal because (aI)v=aIv(aI)_v=aI_v for a∈Frac⁡(R)×a\in\operatorname{Frac}(R)^\times. This proves the divisoriality assertion of Theorem 1.1.

An upper to zero

A nonzero prime of D[Z]D[Z] contracting to zero in DD is called an upper to zero. A domain DD is a UMT-domain if each upper to zero is a maximal tt-ideal of D[Z]D[Z]. Here HtH_t is the union of FvF_v over the nonzero finitely generated subideals F⊆HF\subseteq H. In particular every nonzero divisorial ideal is a tt-ideal, since H⊆Ht⊆HvH\subseteq H_t\subseteq H_v.

Corollary 4.1. In R[Z]R[Z], the kernel QQ of evaluation at y/x∈Frac⁡Ry/x\in\operatorname{Frac} R is an upper to zero strictly contained in the proper tt-ideal m[Z]\mathfrak{m}[Z].

Proof. Evaluation is injective on RR and its target is a field. Its kernel is nonzero since xpZp−yp∈Qx^pZ^p-y^p\in Q. If ∑i=0driZi∈Q\sum_{i=0}^{d}r_iZ^i\in Q, then ∑i=0drixd−iyi=0\sum_{i=0}^{d}r_ix^{d-i}y^i=0 in SS. The monomials xd−iyix^{d-i}y^i give a kk-basis of nd/nd+1\mathfrak{n}^d/\mathfrak{n}^{d+1}, so the residue of every rir_i in kk is zero. Thus Q⊆m[Z]Q\subseteq\mathfrak{m}[Z], and xpx^p witnesses strict inclusion.

Since m=mv\mathfrak{m}=\mathfrak{m}_v, it is an intersection of principal fractional ideals of RR. Extending each to R[Z]R[Z] preserves this intersection, as membership is determined coefficientwise. Hence m[Z]\mathfrak{m}[Z] is divisorial in R[Z]R[Z], and therefore a proper tt-ideal. This proves the remaining assertion of Theorem 1.1. □

Acknowledgment

The counterexample and its proof were generated through discussions with ChatGPT 6 Astra, which also assisted in drafting the manuscript.

Department of Mathematics, National University of Singapore, Singapore 119076

Email address: [email protected]

URL: https://vh-tran.github.io/

References

  1. [1]Stefania Gabelli, Evan Houston, and Giampaolo Picozza, w-divisoriality in polynomial rings, Comm. Algebra 37 (2009), no. 3, 1117–1127, doi:10.1080/00927870802278982.DOI
  2. [2]William Heinzer, Integral domains in which each non-zero ideal is divisorial, Mathematika 15 (1968), no. 2, 164–170, doi:10.1112/S0025579300002527.DOI
  3. [3]Hwankoo Kim, Gorenstein Prüfer domains with non-Prüfer integral closure, Preprint, arXiv:2610.15444, 2026.arxiv.org/abs/2610.15444
  4. [4]Bruce Oberding, A counterpart to Nagata idealization, J. Algebra 365 (2012), 199–221, doi:10.1016/j.jalgebra.2012.05.002.
  5. [5]The Stacks Project Authors, Injective hull of the residue field, The Stacks Project, Section 47.7 (Tag 0821).
  6. [6]Viet-Hoang Tran, Thieu N. Vo, and Tan M. Nguyen, A one-dimensional coherent domain with non-Prüfer integral closure, Preprint, arXiv:2609.21013, 2026.arxiv.org/abs/2609.21013

Paper details

vlnc:W_4amcc3z4
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 7

downloads · 0

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

  • Mathematics

ChatGPT 6 Astra

View statement
Looking for related papers…
Contents