betaFeedback

WorksMathematics

Commutative coherent rings with GP⊈GF\mathcal{GP}\not\subseteq\mathcal{GF}

Abstract

For every field kk, we construct a commutative coherent kk-algebra admitting a Gorenstein projective module that is not Gorenstein flat.

Introduction

Throughout, rings are commutative with identity, ring homomorphisms preserve identities, and modules are unital. Tensor products (respectively, internal Hom) of complexes use coproduct (respectively, product) totalisation, with the usual Koszul signs.

Gorenstein projective modules extend Auslander–Bridger’s theory of modules of GG-dimension zero over Noetherian rings [2, 3]. An RR-module is Gorenstein projective if it is a cycle of an acyclic complex of projective modules which remains acyclic under Hom⁡R(−,W)\operatorname{Hom}_R(-, W) for every projective module WW. Enochs, Jenda and Torrecillas introduced the corresponding flat notion [4]: an RR-module is Gorenstein flat if it is a cycle of an acyclic complex of flat modules which remains acyclic under E⊗R−E \otimes_R - for every injective module EE. We denote these classes by GP(R)\mathcal{GP}(R) and GF(R)\mathcal{GF}(R), respectively.

Although projective modules are flat, the analogous inclusion GP(R)⊆GF(R)\mathcal{GP}(R) \subseteq\mathcal{GF}(R) does not follow from the definitions. The question was already raised in Holm’s 2000 master’s thesis [7 Remark 4.5(3)–(5)], where coherent rings were singled out for further investigation. A positive result of Holm states that GP(R)⊆GF(R)\mathcal{GP}(R) \subseteq\mathcal{GF}(R) when RR is coherent with finite big finitistic projective dimension [8 Proposition 3.4]. In particular, the inclusion holds for Noetherian rings of finite Krull dimension [8 Remark 3.3 and Proposition 3.4].

Later, Šaroch and Šťovíček introduced projectively coresolved Gorenstein flat modules in their study of Gorenstein flat approximation theory [5]. These are the cycles of acyclic complexes of projective modules which remain acyclic after tensoring with any injective module. They proved that the resulting class PGF(R)\mathcal{PGF}(R) satisfies PGF(R)⊆GP(R)∩GF(R)\mathcal{PGF}(R) \subseteq\mathcal{GP}(R) \cap\mathcal{GF}(R), and moreover, PGF(R)\mathcal{PGF}(R) is the left-hand class of a complete hereditary cotorsion pair [5 Theorems 4.4 and 4.9]. Iacob established equivalent criteria [6 Theorem 3] that

GP(R)=PGF(R)⟺GP(R)⊆GF(R).\mathcal{GP}(R) = \mathcal{PGF}(R) \quad\Longleftrightarrow\quad\mathcal{GP}(R) \subseteq\mathcal{GF}(R).

We show for commutative coherent algebras over an arbitrary field that, Gorenstein projective modules need not be Gorenstein flat.

Theorem 1.1 (Main theorem). For every field kk, there exists a commutative coherent kk-algebra AA admitting a Gorenstein projective module which is not Gorenstein flat.

The construction and proof are elementary; the only external input is a cohomological vanishing theorem of Dicks, Kropholler, Leary and Thomas [1 Corollary 6.9].

Proof of the main theorem

Let kk be an arbitrary field and pp a prime number distinct from char⁡(k)\operatorname{char}(k). Put ℵω=sup⁡n<ωℵn\aleph_\omega= \sup_{n<\omega} \aleph_n, where ω\omega is the first infinite ordinal. Let G=(Fp)(ℵω)G = (\mathbb{F}_p)^{(\aleph_\omega)} be an Abelian group of ℵω\aleph_{\omega} copies of coproduct of Fp\mathbb{F}_p. Let R=kGR = kG be the group algebra with augmentation map

ε:R⟶k,∑g∈Gcgg⟼∑g∈Gcg,(finite sum).\varepsilon:R\longrightarrow k,\qquad\sum_{g\in G}c_g g\longmapsto\sum_{g\in G}c_g,\qquad\text{(finite sum).}

Put A~=k[x,y]/(x,y)2\widetilde{A}=k[x,y]/(x,y)^2, and let A=A~⊗kR=R[x,y]/(x,y)2A=\widetilde{A}\otimes_k R=R[x,y]/(x,y)^2 be the final ring. Define P=A~(ℵ0)P=\widetilde{A}^{(\aleph_0)}, and d:P⟶Pd:P\longrightarrow P via d(e2r)=xerd(e_{2r})=xe_r and d(e2r+1)=yerd(e_{2r+1})=ye_r. Then P∙=[⋯→dP→dP→d⋯ ]P_\bullet=[\cdots\xrightarrow{d}P\xrightarrow{d}P\xrightarrow{d}\cdots] is a one periodic complex. Finally fix Q∙⟶kQ_\bullet\longrightarrow k to be a projective resolution of RR-modules. Set K∙=P∙⊗kQ∙K_\bullet=P_\bullet\otimes_k Q_\bullet. We will show that Z0(K∙)∈GP(A)∖GF(A)Z_0(K_\bullet)\in\mathcal{GP}(A)\setminus\mathcal{GF}(A).

Lemma 2.1. RR is von Neumann regular, thus is coherent.

Proof. Fix arbitrary a∈Ra\in R. The support of aa is contained in a finite subgroup H⊆GH\subseteq G. Since ∣H∣|H| is invertible in kk, Maschke’s theorem implies that kHkH is semisimple. Since HH is abelian, kHkH is a finite product of fields. Hence there is b∈kHb\in kH with a=abaa=aba.

Hence RR is von Neumann regular. Consequently RR is coherent. □\square

Corollary 2.2. kk is a flat RR-module, and Hom⁡k(R,k)⊗Rk≠0\operatorname{Hom}_k(R,k)\otimes_R k\ne0.

Proof. Any module over a von Neumann regular ring is flat. Since ε\varepsilon is surjective, k≅Hom⁡k(k,k)⟶Hom⁡k(R,k)k\cong\operatorname{Hom}_k(k,k)\longrightarrow\operatorname{Hom}_k(R,k) is injective. Note that kk is flat. Hence k⊗Rkk\otimes_R k identifies a subset of Hom⁡k(R,k)⊗Rk\operatorname{Hom}_k(R,k)\otimes_R k. Consequently Hom⁡k(R,k)⊗Rk≠0\operatorname{Hom}_k(R,k)\otimes_R k\ne0. □\square

Proposition 2.3 ([1 Corollary 6.9]). For every abelian group UU and every n≥0n\ge0, one has Hn(G,ZG⊗ZU)=0H^n(G,\mathbb{Z}G\otimes_{\mathbb{Z}}U)=0, where GG acts as g⊗1g\otimes1 on ZG⊗ZU\mathbb{Z}G\otimes_{\mathbb{Z}}U.

Proof. Since ∣G∣=ℵω|G|=\aleph_\omega, the Abelian group GG has ℵ\aleph-rank [1 Notation 1.1] ω\omega. By [1 Corollary 6.9], one has Hn(G,ZG⊗ZU)=0H^n(G,\mathbb{Z}G\otimes_{\mathbb{Z}}U)=0 whenever n<ω+1n<\omega+1. □\square

Corollary 2.4. For every set JJ and every n≥0n\ge0, one has Ext⁡Rn(k,R(J))=0\operatorname{Ext}^n_R(k,R^{(J)})=0.

Proof. Clearly Ext⁡Rn(k,R(J))≅Hn(G,R(J))≅Hn(G,ZG⊗Zk(J))=0\operatorname{Ext}^n_R(k,R^{(J)})\cong H^n(G,R^{(J)})\cong H^n(G,\mathbb{Z}G\otimes_{\mathbb{Z}}k^{(J)})=0. □\square

Lemma 2.5. The one periodic complex P∙=[⋯→dP→dP→d⋯ ]P_\bullet=[\cdots\xrightarrow{d}P\xrightarrow{d}P\xrightarrow{d}\cdots] is acyclic.

Proof. Every element of PP is a unique finite sum v=∑r≥0(ar+brx+cry)erv=\sum_{r\ge0}(a_r+b_rx+c_ry)e_r. Since (x,y)2=0(x,y)^2=0, one has d(v)=∑r≥0(a2rx+a2r+1y)erd(v)=\sum_{r\ge0}(a_{2r}x+a_{2r+1}y)e_r. Hence d(v)=0d(v)=0 if and only if ar=0a_r=0 for every rr. Hence ker⁡d=(x,y)⋅P=im⁡d\ker d=(x,y)\cdot P=\operatorname{im}d. Consequently P∙P_\bullet is acyclic. □\square

Lemma 2.6. Each term of K∙=P∙⊗kQ∙K_\bullet=P_\bullet\otimes_k Q_\bullet is a projective AA module.

Proof. Since A~⊗kR≅A\widetilde{A}\otimes_k R\cong A, taking direct summands and coproducts complete the proof. □\square

Lemma 2.7. K∙K_\bullet and Hom⁡A(K∙,W)\operatorname{Hom}_A(K_\bullet,W) are acyclic for any projective AA-module WW.

Proof. Since K∙K_\bullet is contractible in Mod⁡k\operatorname{Mod}_k, it is acyclic in Mod⁡A\operatorname{Mod}_A.

It suffices to show the second assertion for an arbitrary free module W=A(J)W=A^{(J)}. By Corollary 2.4, Hom⁡R(Q∙,R(J))\operatorname{Hom}_R(Q_\bullet,R^{(J)}) is acyclic. Since A~\widetilde{A} is finite-dimensional over kk, there is a natural isomorphism of complexes Hom⁡R(Q∙,A(J))≅Hom⁡R(Q∙,R(J))⊗kA~\operatorname{Hom}_R(Q_\bullet,A^{(J)})\cong\operatorname{Hom}_R(Q_\bullet,R^{(J)})\otimes_k\widetilde{A}. Since Hom⁡R(Q∙,R(J))\operatorname{Hom}_R(Q_\bullet,R^{(J)}) is contractible over kk, Hom⁡R(Q∙,A(J))\operatorname{Hom}_R(Q_\bullet,A^{(J)}) is contractible over A~\widetilde{A}. Thus

Hom⁡A(K∙,A(J))≅Hom⁡A~(P∙,Hom⁡R(Q∙,A(J)))\operatorname{Hom}_A(K_\bullet,A^{(J)})\cong\operatorname{Hom}_{\widetilde{A}}\left(P_\bullet,\operatorname{Hom}_R(Q_\bullet,A^{(J)})\right)

is acyclic. □\square

Lemma 2.8. Hom⁡k(A,k)\operatorname{Hom}_k(A,k) is an injective AA module, while Hom⁡k(A,k)⊗AK∙\operatorname{Hom}_k(A,k)\otimes_A K_\bullet is not acyclic.

Proof. Since Hom⁡k(A,−)\operatorname{Hom}_k(A,-) has an exact left adjoint, it preserves injective modules. Hence Hom⁡k(A,k)\operatorname{Hom}_k(A,k) is injective. Let λ1,λx,λy\lambda_1,\lambda_x,\lambda_y be the basis of Hom⁡k(A~,k)\operatorname{Hom}_k(\widetilde{A},k) dual to 1,x,y1,x,y. We claim that λy⊗e0\lambda_y\otimes e_0 is a cycle but not a boundary in Hom⁡k(A~,k)⊗A~P∙\operatorname{Hom}_k(\widetilde{A},k)\otimes_{\widetilde{A}}P_\bullet.

(1) Since (1⊗d)(λy⊗e0)=λy⊗xe0=(λyx)⊗e0=0(1\otimes d)(\lambda_y\otimes e_0)=\lambda_y\otimes xe_0=(\lambda_yx)\otimes e_0=0, λy⊗e0\lambda_y\otimes e_0 is a cycle.

(2) Every element of Hom⁡k(A~,k)⊗A~P\operatorname{Hom}_{k}(\widetilde{A},k)\otimes_{\widetilde{A}}P has a unique expression ∑r≥0μr⊗er\sum_{r\geq0}\mu_r\otimes e_r. Then

(1⊗d)(∑r≥0μr⊗er)=∑r≥0(μ2rx+μ2r+1y)⊗er=∑r≥0(μ2r(x)λ1+μ2r+1(y)λ1)⊗er.(1\otimes d)\left(\sum_{r\geq0}\mu_r\otimes e_r\right)=\sum_{r\geq0}(\mu_{2r}x+\mu_{2r+1}y)\otimes e_r=\sum_{r\geq0}(\mu_{2r}(x)\lambda_1+\mu_{2r+1}(y)\lambda_1)\otimes e_r.

Since every boundary has all its coefficients in kλ1k\lambda_1, λy⊗e0\lambda_y\otimes e_0 is not a boundary. Hence H0(Hom⁡k(A~,k)⊗A~P∙)≠0H_0(\operatorname{Hom}_{k}(\widetilde{A},k)\otimes_{\widetilde{A}}P_\bullet)\ne0. Since A~\widetilde{A} is finite-dimensional over kk, one has

Hom⁡k(A,k)≅Hom⁡k(A~,k)⊗kHom⁡k(R,k).\operatorname{Hom}_{k}(A,k)\cong\operatorname{Hom}_{k}(\widetilde{A},k)\otimes_k\operatorname{Hom}_{k}(R,k).

Consequently, Hom⁡k(A,k)⊗AK∙≅(Hom⁡k(A~,k)⊗A~P∙)⊗k(Hom⁡k(R,k)⊗RQ∙)\operatorname{Hom}_{k}(A,k)\otimes_AK_\bullet\cong\bigl(\operatorname{Hom}_{k}(\widetilde{A},k)\otimes_{\widetilde{A}}P_\bullet\bigr)\otimes_k\bigl(\operatorname{Hom}_{k}(R,k)\otimes_RQ_\bullet\bigr). Since RR is von Neumann regular, the RR-module Hom⁡k(R,k)\operatorname{Hom}_{k}(R,k) is flat, and hence

Hj(Hom⁡k(R,k)⊗RQ∙)≅{Hom⁡k(R,k)⊗Rk,j=0,0,j≠0.H_j\bigl(\operatorname{Hom}_{k}(R,k)\otimes_RQ_\bullet\bigr)\cong \begin{cases} \operatorname{Hom}_{k}(R,k)\otimes_R k, & j=0,\\ 0, & j\ne0. \end{cases}

Hence the Künneth formula (over kk) gives

H0(Hom⁡k(A,k)⊗AK∙)≅H0((Hom⁡k(A~,k)⊗A~P∙)⊗k(Hom⁡k(R,k)⊗RQ∙))H_0\bigl(\operatorname{Hom}_{k}(A,k)\otimes_AK_\bullet\bigr)\cong H_0\bigl((\operatorname{Hom}_{k}(\widetilde{A},k)\otimes_{\widetilde{A}}P_\bullet)\otimes_k(\operatorname{Hom}_{k}(R,k)\otimes_RQ_\bullet)\bigr)
≅H0(Hom⁡k(A~,k)⊗A~P∙)⊗k(Hom⁡k(R,k)⊗Rk)≠0.\cong H_0\bigl(\operatorname{Hom}_{k}(\widetilde{A},k)\otimes_{\widetilde{A}}P_\bullet\bigr)\otimes_k\bigl(\operatorname{Hom}_{k}(R,k)\otimes_Rk\bigr)\ne0.

Here H0(Hom⁡k(A~,k)⊗A~P∙)≠0H_0(\operatorname{Hom}_{k}(\widetilde{A},k)\otimes_{\widetilde{A}}P_\bullet)\ne0 is above. Corollary 2.2 gives Hom⁡k(R,k)⊗Rk≠0\operatorname{Hom}_{k}(R,k)\otimes_Rk\ne0. □

Proof of the main theorem. Since AA is finite free over the coherent ring RR (Lemma 2.1), AA is coherent. By Lemma 2.7, Z−2(K∙)∈GP(A)Z_{-2}(K_\bullet)\in\mathcal{GP}(A). Meanwhile, the projective resolution

⋯⟶K1⟶K0⟶K−1⟶Z−2(K∙)⟶0\cdots\longrightarrow K_1\longrightarrow K_0\longrightarrow K_{-1}\longrightarrow Z_{-2}(K_\bullet)\longrightarrow0

gives Tor⁡1A(Hom⁡k(A,k),Z−2(K∙))≅H0(Hom⁡k(A,k)⊗AK∙)≠0\operatorname{Tor}^{A}_{1}(\operatorname{Hom}_{k}(A,k),Z_{-2}(K_\bullet))\cong H_0(\operatorname{Hom}_{k}(A,k)\otimes_AK_\bullet)\ne0 by Lemma 2.8. Since Hom⁡k(A,k)\operatorname{Hom}_{k}(A,k) is injective, Z−2(K∙)Z_{-2}(K_\bullet) is not Gorenstein flat. □

Acknowledgements and reflections on AI-assisted research

The author is grateful to his advisor, Pu Zhang, for urging him to think further about this problem; this encouragement was instrumental in finding the elementary approach presented here. The author also thanks Junhong Chen, Jian Liu, Xue-Song Lu Tianyang Sun (and JIUCHONG team at the University of Science and Technology of China) for their help with his earlier attempts to solve this problem.

The author thanks ChatGPT for finding complicated counterexamples in earlier arXiv versions. Had the author not returned to think carefully about the problem, the early counterexample might have remained one that few were willing to read: the problem might have been regarded as “solved”, yet remained shrouded in fog.

References

References

  1. [1]W. Dicks, P. H. Kropholler, I. J. Leary and S. Thomas, Classifying spaces for proper actions of locally-finite groups, J. Group Theory 5 (2002), no. 4, 453–480.DOI
  2. [2]M. Auslander and M. Bridger, Stable Module Theory, Memoirs of the American Mathematical Society, no. 94, American Mathematical Society, Providence, RI, 1969.DOI
  3. [3]E. E. Enochs and O. M. G. Jenda, Gorenstein injective and projective modules, Math. Z. 220 (1995), 611–633.DOI
  4. [4]E. E. Enochs, O. M. G. Jenda, and B. Torrecillas, Gorenstein flat modules, Nanjing Daxue Xuebao Shuxue Bannian Kan 10 (1993), 1–9.
  5. [5]J. Šaroch and J. Šťovíček, Singular compactness and definability for Σ-cotorsion and Gorenstein modules, Selecta Math. (N.S.) 26 (2020), no. 2, Paper No. 23.DOI
  6. [6]A. Iacob, Projectively coresolved Gorenstein flat and Ding projective modules, Comm. Algebra 48 (2020), no. 7, 2883–2893.arxiv.org/abs/2001.09234
  7. [7]N. Ding, Y. Li, and L. Mao, Strongly Gorenstein flat modules, J. Aust. Math. Soc. 86 (2009), no. 3, 323–338.DOI
  8. [8]H. Holm, Gorenstein homological dimensions, J. Pure Appl. Algebra 189 (2004), 167–193.

Paper details

vlnc:W_mgn82acw
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC0-1.0
Usage

views · 0

downloads · 0

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

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents