This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

A Higher Algebraic Approach to Liftings of Modules over Derived Quotients

Ryo Ishizuka Department of Mathematics, Institute of Science Tokyo, 2-12-1 Ookayama, Meguro, Tokyo 152-8551 ishizuka.r.ac@m.titech.ac.jp
Abstract.

We show a certain existence of a lifting of modules under the self-Ext2\operatorname{Ext}^{2}-vanishing condition over the “derived quotient” by using the notion of higher algebra. This refines a work of Auslander–Ding–Solberg’s solution of the Auslander–Reiten conjecture for complete interesctions. Together with Auslander’s zero-divisor theorem, we show that the existence of such Ext\operatorname{Ext}-vanishing module over derived quotients is equivalent to being local complete intersections.

Key words and phrases:
Liftings of modules, Auslander–Reiten conjecture, derived quotients, animated rings
2020 Mathematics Subject Classification: 13D07, 55P43

1. Introduction

The Auslander–Reiten conjecture was first proposed by Auslander and Reiten in 1975 [AR75]. This conjecture has posed a significant challenge in the representation theory of algebras and (commutative) ring theory.

Conjecture 1.1 (Auslander–Reiten conjecture).

Let RR be a Noetherian local ring. If a finitely generated RR-module MM satisfies ExtRi(M,MR)=0\operatorname{Ext}_{R}^{i}(M,M\oplus R)=0 for all i1i\geq 1, then MM is a projective RR-module.

We suffice to prove the conjecture for complete Noetherian local ring RR (see, for example, [Kum23, Remark 2.3]). To date, the (partial) solutions to this conjecture have been very diverse (see, for instance, [CH10, Appendix A.2] and the introduction of [KOT22, Kim23] for recent works). These solutions mainly focus on the singularity of RR.

In particular, Auslander, Ding, and Solberg [ADS93] proved the conjecture for any local complete intersection RR. Their strategy is taking a representation RA/(f1,,fr)R\cong A/(f_{1},\dots,f_{r}) where AA is a Noetherian local ring and f1,,frf_{1},\dots,f_{r} is a regular sequence on AA and providing the following “lifting property” of modules:

Theorem 1.2 ([ADS93, Proposition 1.7]111Also, Yoshino proved this theorem for bounded below complexes whose terms are finite free in [Yos97, Lemma 3.2].).

Let AA be a Noetehrian local ring and let f1,,frf_{1},\dots,f_{r} be a regular sequence on AA. Set RA/(f1,,fr)R\coloneqq A/(f_{1},\dots,f_{r}). If a finitely generated RR-module MM satisfies ExtR2(M,M)=0\operatorname{Ext}^{2}_{R}(M,M)=0, then MM is liftable to AA, that is, there exists a finitely generated AA-module LL such that LALRML\otimes^{L}_{A}R\cong M.

If AA is a regular local ring (this is possible by Cohen’s structure theorem if a given RR is a complete Noetherian local ring), then the above theorem implies that MM has finite projective dimension, and thus the Auslander–Reiten conjecture holds for the local complete intersection RR.

Previously, as this theorem, there are a lot of desirable phenomena if we take quotients by regular sequences. However, with the advancement of DG methods and higher algebra, it has become possible to deal with quotients by any sequences of elements (not necessarily regular) by using derived quotients; let AA be a Noetherian local ring and let f1,,frf_{1},\dots,f_{r} be a (possibly non-regular) sequence of elements of AA. The derived quotient A/Lf¯A/L(f1,,fr)A/^{L}\underline{f}\coloneqq A/^{L}(f_{1},\dots,f_{r}) (Definition 3.1) is a “commutative algebra object” in D(A)D(A) and has information not only the usual quotient A/(f1,,fr)AA/(f_{1},\dots,f_{r})A but also their torsions such as (0:Afi)(0:_{A}f_{i}). In fact, as a complex of AA-modules, A/L(f1,,fr)A/^{L}(f_{1},\dots,f_{r}) is the Koszul complex Kos(f1,,fr;A)\operatorname{Kos}(f_{1},\dots,f_{r};A) and thus A/L(f1,,fr)A/^{L}(f_{1},\dots,f_{r}) is isomorphic to A/(f1,,fr)AA/(f_{1},\dots,f_{r})A if and only if f1,,frf_{1},\dots,f_{r} is a (Koszul-)regular sequence on AA. In this paper, we refine the work of Auslander–Ding–Solberg above (Theorem 1.2) by using the notion of higher algebra. This is our first main theorem:

Theorem 1.3 (Special case of Theorem 6.4).

Let AA be a Noetherian local ring and let f1,,frf_{1},\dots,f_{r} be a sequence of elements of AA. Set A/Lf¯=A/L(f1,,fr)A/^{L}\underline{f}=A/^{L}(f_{1},\dots,f_{r}) and RA/(f1,,fr)AR\coloneqq A/(f_{1},\dots,f_{r})A. If a finitely generated RR-module MM satisfies ExtA/Lf¯2(M,M)=0\operatorname{Ext}_{A/^{L}\underline{f}}^{2}(M,M)=0, then there exists a finitely generated AA-module LL such that MLALA/Lf¯M\cong L\otimes^{L}_{A}A/^{L}\underline{f}.

There are two remarks of this theorem. First, such generalization of Theorem 1.2 has been studied by using the notion of DG algebras and DG modules (see Remark 1.5). Unlike that, we use the concepts of (stable) \infty-categories, animated rings, and module spectra introduced by Lurie [HTT, HA, SAG] and Česnavičius–Scholze [ČS24].

Second, this theorem is not enough to show the Auslander–Reiten conjecture: in fact it tells us the following “full circle” result.

Theorem 1.4 (Corollary 7.2).

Let AA be a regular local ring and let f1,,frf_{1},\dots,f_{r} be a sequence of elements of AA. As above, set A/Lf¯=A/L(f1,,fr)A/^{L}\underline{f}=A/^{L}(f_{1},\dots,f_{r}) and RA/(f1,,fr)AR\coloneqq A/(f_{1},\dots,f_{r})A. Then the existence of a finitely generated RR-module MM satisfying ExtA/Lf¯2(M,M)=0\operatorname{Ext}^{2}_{A/^{L}\underline{f}}(M,M)=0 is equivalent to that f1,,frf_{1},\dots,f_{r} is a regular sequence on AA.

This theorem is deduced from our theorem (Theorem 1.3) and Auslander’s zero-divisor theorem ([Aus61, Ser75, PS73, Hoc74, Rob87], see Theorem 7.1). As a consequence, if we consider the existence of liftings of modules under such a Ext\operatorname{Ext}-vanishing condition, it is not enough to merely discuss as in [ADS93]. See also Lemma 3.11 and Example 3.12 where we prove a relationship between the Ext\operatorname{Ext}-group over derived quotients and the one over usual quotients, and give an example that ExtA/Lf¯i(M,N)\operatorname{Ext}^{i}_{A/^{L}\underline{f}}(M,N) does not vanish but ExtA/(f¯)Ai(M,N)\operatorname{Ext}^{i}_{A/(\underline{f})A}(M,N) does.

1.1. Relationship with DG methods

We end the introduction by recalling the previous work on a generalization of lifting properties of modules by using DG methods and a relationship of their results with our results.

Remark 1.5.

Let Λ\Lambda be a (commutative) Noetherian ring and let f1,,frf_{1},\dots,f_{r} be a sequence of elements of Λ\Lambda. Assume that Λ\Lambda is (f1,,fr)Λ(f_{1},\dots,f_{r})\Lambda-adically complete. Let AA be an (associative, unital, graded commutative, and positively graded) DG Λ\Lambda-algebra such that each term of AA is a finite free Λ\Lambda-module.222Any cohomologically finitely generated connective Λ\Lambda-module is quasi-isomorphic to a complex of finite free Λ\Lambda-modules. In particular, the underlying complex of any almost perfect animated Λ\Lambda-algebra is quasi-isomorphic to a complex of finite free Λ\Lambda-modules by Lemma 2.8 and Lemma 2.14. In [NS13, Corollary 3.5], they proved that if a DG Kos(f¯;A)\operatorname{Kos}(\underline{f};A)-module DD that is homologically bounded below and homologically degreewise finite333In our sense, this means that DD is a nn-truncated almost perfect module for some n0n\geq 0. See Lemma 2.8. satisfies ExtKos(f¯;A)2(D,D)=0\operatorname{Ext}^{2}_{\operatorname{Kos}(\underline{f};A)}(D,D)=0, then there exists a DG AA-module EE such that DEALKos(f¯;A)D\cong E\otimes^{L}_{A}\operatorname{Kos}(\underline{f};A). Since then, there have been approaches to the lifting problem for polynomial extensions AA[X1,,Xn]A\to A[X_{1},\dots,X_{n}] of a DG Λ\Lambda-algebra and free extensions AAX1,,XnA\to A\langle X_{1},\dots,X_{n}\rangle of a divided power DG Λ\Lambda-algebra towards the naïve lifting conjecture which leads to the Auslander–Reiten conjecture (see [NY18, OY21, NOY22] and a more comprehensive discussion in [NOY25, Appendix A]).

Next, we explain how relate our results to the similar results in Remark 1.5.

Remark 1.6.

By [DAG, §2.6], we have functors of \infty-categories

CAlgΛanϕCAlgΛdg𝜓CAlgΛE\operatorname{CAlg}^{an}_{\Lambda}\xrightarrow{\phi}\operatorname{CAlg}^{dg}_{\Lambda}\xrightarrow{\psi}\operatorname{CAlg}^{\mathbb{E}_{\infty}}_{\Lambda}

where CAlgΛan\operatorname{CAlg}^{an}_{\Lambda} is the \infty-category of animated Λ\Lambda-algebras, CAlgΛE\operatorname{CAlg}^{\mathbb{E}_{\infty}}_{\Lambda} is the \infty-category of E\mathbb{E}_{\infty}-Λ\Lambda-algebras, and CAlgΛdg\operatorname{CAlg}^{dg}_{\Lambda} be the underlying \infty-category of the ordinary category of commutative differential graded Λ\Lambda-algebras with a Quillen model structure with quasi-isomorphisms as weak equivalences. Note that any commutative differential graded Λ\Lambda-algebra is an object of CAlgΛdg\operatorname{CAlg}^{dg}_{\Lambda}. The composition θψϕ\theta\coloneqq\psi\circ\phi sends an animated ring to its underlying (connective) E\mathbb{E}_{\infty}-ring AA^{\circ} and this functor θ:CAlgΛanCAlgΛE,cn\theta\colon\operatorname{CAlg}^{an}_{\Lambda}\to\operatorname{CAlg}^{\mathbb{E}_{\infty},cn}_{\Lambda} is conservative and commutes all small limits and colimits ([DAG, Proposition 2.6.1] and [SAG, Proposition 25.1.2.2]).

If Λ\Lambda contains Q\mathbb{Q}, then ψ\psi is an equivalence of \infty-categories and ϕ\phi is fully faithful whose essential image is the connective E\mathbb{E}_{\infty}-Λ\Lambda-algebras by [HA, Proposition 7.1.4.11 and Proposition 7.1.4.20]. In particular, derived quotients in CAlgΛan\operatorname{CAlg}^{an}_{\Lambda} and Koszul complexes in CAlgΛdg\operatorname{CAlg}^{dg}_{\Lambda} define the same objects in CAlgΛE\operatorname{CAlg}^{\mathbb{E}_{\infty}}_{\Lambda} by the universal property ([KR19, Lemma 2.3.5]).

Furthermore, we can compare the \infty-categories of modules over those objects. First, for any animated ring AA, Mod(A)\operatorname{Mod}(A) and Mod(A)\operatorname{Mod}(A^{\circ}) are the same by the definition (see Definition 2.2). Second, Mod(A)\operatorname{Mod}(A^{\circ}) is equivalent to the underlying \infty-category of the model category of DG modules over the differential graded Λ\Lambda-algebra ϕ(A)\phi(A). This is because the model category of modules over the Eilenberg–Mac Lane spectrum Hϕ(A)AH\phi(A)\simeq A^{\circ} of ϕ(A)\phi(A) is (Quillen-)equivalent to the model category of DG modules over the differential graded Λ\Lambda-algebra ϕ(A)\phi(A) (see [Shi07, Theorem 1.2 and Corollary 2.15]).

Remark 1.7.

Based on Remark 1.5 and Remark 1.6 above, if Λ\Lambda contains Q\mathbb{Q}, there is no difference (in the \infty-categorical setting) between the \infty-categories CAlgΛan\operatorname{CAlg}^{an}_{\Lambda} and CAlgΛdg,cn\operatorname{CAlg}^{dg,cn}_{\Lambda}, and their module categories Mod(A)\operatorname{Mod}(A) and Mod(ϕ(A))\operatorname{Mod}(\phi(A)). Since our results (Theorem 1.3 and Theorem 6.4) only treat discrete modules over A/Lf¯A/^{L}\underline{f}, they are a special case of the previous results [NS13, Corollary 3.5] explained above if Λ\Lambda contains Q\mathbb{Q}.

However, we do not know whether our liftability results on the \infty-category of module spectra Mod(A)\operatorname{Mod}(A) are compatible with the results on the category of DG modules over the DG Λ\Lambda-algebra ϕ(A)\phi(A) in general. Otherwise or not, the methods are different: their results in [NS13] are based on DG algebras and DG modules, in particular, very calculous and concrete although our proof is conceptualized by using higher algebra. Also, since our work depends on the discreteness of a given MM (for example in Lemma 3.6 and 5.1), it is not clear whether the result Theorem 6.4 can be generalized to nn-truncated almost perfect modules as in [NS13].

Acknowledgments

The author would like to thank Kaito Kimura, Yuya Otake, Kazuma Shimomoto, and Ryo Takahashi for helpful discussions about previous works on the Auslander–Reiten conjecture and related topics, Yutaro Mikami and Masaya Sato for their valuable conversations on the notion of higher algebra, and Saeed Nasseh for his comments on the previous research by using DG methods. This work was supported by JSPS KAKENHI Grant Number 24KJ1085.

2. Animated Rings and Their Modules

In this paper, we freely use the notion of higher algebra such as stable \infty-categories and animated rings, see for example [HA, ČS24, BL22] (or a brief review in [IN24, Appendix A]). In this section, we fix some terminology and collect some lemmas about animated rings and their modules which are well-known for experts.

Notation 2.1.

Throughout this paper, we use the following notation. Let Λ\Lambda be a complete Noetherian local ring with the maximal ideal 𝔪\mathfrak{m}. Let CAlgΛan\operatorname{CAlg}^{an}_{\Lambda} be the \infty-category of animated (commutative) Λ\Lambda-algebras. Let x¯x1,,xt\underline{x}\coloneqq x_{1},\dots,x_{t} be a sequence of elements of 𝔪\mathfrak{m}.

2.1. Basic Notions

Definition 2.2.

Let AA be an animated ring (or more generally, a (connective) E\mathbb{E}_{\infty}-ring). The \infty-category of AA-modules is the \infty-category Mod(A)\operatorname{Mod}(A) of module spectra over the (underlying) E\mathbb{E}_{\infty}-ring AA^{\circ}.444In this paper, we only mention modules over animated rings instead of ones over E\mathbb{E}_{\infty}-rings. However, this makes no difference: by this definition of Mod(A)\operatorname{Mod}(A), the module category on an animated ring AA and its underlying E\mathbb{E}_{\infty}-ring AA^{\circ} are the same. So every statements on modules over animated rings are also true for modules over an connective E\mathbb{E}_{\infty}-ring. For a comparison of DG theory, please see Remark 1.6. If AA is discrete, Mod(A)\operatorname{Mod}(A) is equivalent to the (\infty-categorical enhancement of the) derived category 𝒟(A)\mathcal{D}(A) of AA-modules. For any MM and NN in Mod(A)\operatorname{Mod}(A), we denote by MapA(M,N)\operatorname{Map}_{A}(M,N) the anima of AA-module maps and by mapA(M,N)\operatorname{map}_{A}(M,N) the AA-module spectra of AA-module maps. This \infty-category Mod(A)\operatorname{Mod}(A) is stable and presentable (in particular, it has small colimits small limits by [HA, Corollary 7.1.1.5, Corollary 4.2.3.7, and Corollary 4.2.3.3]).

Definition 2.3.

Let AA be an animated ring and let MM be an AA-module. The ii-th homotopy group πi(M)\pi_{i}(M) of a module MM over an animated ring AA is that of the underlying spectra of MM. Since we use the homological notation in this paper, πi()\pi_{i}(-) corresponds to the (i)(-i)-th cohomology group Hi()H^{-i}(-): in fact, πi()\pi_{i}(-) defines a (canonical) tt-structure (Mod(A)0,Mod(A)0)(\operatorname{Mod}(A)_{\geq 0},\operatorname{Mod}(A)_{\leq 0}) on Mod(A)\operatorname{Mod}(A) (see [HA, Proposition 7.1.1.13]). An AA-module MM is connective (or animated) if πi(M)=0\pi_{i}(M)=0 for any i<0i<0. A discrete AA-module MM is an object of Mod(A)\operatorname{Mod}(A) such that πi(M)=0\pi_{i}(M)=0 for any i0i\neq 0. Note that AA-modules in this paper is always an object of Mod(A)\operatorname{Mod}(A) not of the abelian category of discrete AA-modules.

We recall the Ext\operatorname{Ext}-group in the \infty-category of AA-modules.

Definition 2.4 ([HA, Notation 7.1.1.11 and Remark 7.1.1.12]).

Let AA be an animated ring. For AA-modules MM and NN, we denote by ExtAi(M,N)\operatorname{Ext}^{i}_{A}(M,N) the discrete π0(A)\pi_{0}(A)-module π0(MapA(M,N[i]))\pi_{0}(\operatorname{Map}_{A}(M,N[i])) or πi(mapA(M,N))\pi_{-i}(\operatorname{map}_{A}(M,N)) for iZi\in\mathbb{Z}. If AA is a discrete ring and MM and NN are complexes of AA-modules, mapA(M,N)\operatorname{map}_{A}(M,N) is isomorphic to the derived functor RHomA(M,N)R\operatorname{Hom}_{A}(M,N) of the internal hom and then the AA-module ExtAi(M,N)\operatorname{Ext}^{i}_{A}(M,N) is isomorphic to the usual Yoneda Ext\operatorname{Ext}-group, that is, ExtAi(M,N)Hi(RHomA(M,N))\operatorname{Ext}^{i}_{A}(M,N)\cong H^{i}(R\operatorname{Hom}_{A}(M,N)).

Definition 2.5 ([HA, Definition 7.2.4.30] and [KST18, Definition 2.4]).

Let AA be an animated ring. We say that AA is coherent if π0(A)\pi_{0}(A) is a coherent ring and each homotopy group πi(A)\pi_{i}(A) is a finitely presented π0(A)\pi_{0}(A)-module. We say that AA is Noetherian if π0(A)\pi_{0}(A) is Noehterian and AA is coherent.

Definition 2.6 ([HA, Definition 7.2.4.1 and Proposition 7.2.4.2]).

Let AA be an animated ring. The \infty-category Mod(A)perf\operatorname{Mod}(A)^{\operatorname{perf}} is the smallest stable subcategory of Mod(A)\operatorname{Mod}(A) that contains AA and is closed under retracts. We say that an AA-module MM is perfect if it belongs to Mod(A)perf\operatorname{Mod}(A)^{\operatorname{perf}}. This is equivalent to that MM is compact in Mod(A)\operatorname{Mod}(A).

Definition 2.7 ([HA, Definition 7.2.4.10] and [SAG, Definition 2.7.0.1]).

Let AA be an animated ring and let MM be an AA-module.

  1. (a)

    MM is perfect to order nn if, for every filtered diagram {Nα}\{N_{\alpha}\} in Mod(A)0\operatorname{Mod}(A)_{\leq 0}, the map of π0(A)\pi_{0}(A)-modules

    colimαExtAi(M,Nα)ExtAi(M,colimαNα)\operatorname*{colim}_{\alpha}\operatorname{Ext}_{A}^{i}(M,N_{\alpha})\to\operatorname{Ext}_{A}^{i}(M,\operatorname*{colim}_{\alpha}N_{\alpha})

    is bijective for all i<ni<n and injective for i=ni=n.

  2. (b)

    MM is almost perfect if it is perfect to order nn for every integer nn. This notion is equivalent to pseudo-coherence of MM in the sense of [Sta, 064Q] if AA is discrete.

Especially, if MM is connective, then MM is perfect to order 0 if and only if π0(M)\pi_{0}(M) is a finitely generated π0(A)\pi_{0}(A)-module ([SAG, Proposition 2.7.2.1(1)]). By tensor-forgetful adjunction, the property of being perfect to order nn is stable under the (derived) base change.

Lemma 2.8 ([SAG, Corollary 2.7.2.3]).

Let AA be a Noetherian animated ring and let MM be an AA-module. Then MM is perfect to order nn if and only if the following conditions hold:

  1. (a)

    MM is bounded below, that is, for any sufficiently small m0m\ll 0, πm(M)=0\pi_{m}(M)=0.

  2. (b)

    For any mnm\leq n, πm(M)\pi_{m}(M) is a finitely generated π0(A)\pi_{0}(A)-module.

Definition 2.9 ([HA, Definition 7.2.4.21]).

Let AA be an animated ring and let Iπ0(A)I\subseteq\pi_{0}(A) be an ideal. For abZ{±}a\leq b\in\mathbb{Z}\cup\{\pm\infty\}, an AA-module MM has II-complete Tor-amplitude in [a,b][a,b] if MALNM\otimes^{L}_{A}N is contained in Mod(A)[a,b]\operatorname{Mod}(A)_{[a,b]} for any II^{\infty}-torsion (or equivalently, II-torsion) discrete AA-module NN. We say that MM has Tor-amplitude n\leq n if MM is 0-complete Tor-amplitude in [,n][-\infty,n], that is, for every discrete AA-module NN, NALMN\otimes^{L}_{A}M is nn-truncated

2.2. Behavior under base change

The properties in the previous subsection are stable under base change under some conditions (Lemma 2.12 and Lemma 2.13). To show this, we need the following lemma.

Lemma 2.10.

Let AA be a discrete ring with a weakly proregular ideal IAI\subseteq A and let MM be a derived II-complete AA-module which is contained in Mod(A)[a,b]\operatorname{Mod}(A)_{[a,b]}. For any almost perfect AA-module NN, the AA-module MALNM\otimes^{L}_{A}N is derived II-complete.

Proof.

By [Yek25, Theorem 0.3 and Theorem 4.7] (or [PSY14, PSY15]), there exists a complex MM_{\bullet} of derived II-complete (discrete) AA-modules which represents MM. Since NN is almost perfect, there exists a (homologically) bounded below complex NN_{\bullet} of finite free AA-modules which represents NN ([Sta, 064Q]). Then the derived tensor product MALNM\otimes^{L}_{A}N is represented by the totalization of the double complex MANM_{\bullet}\otimes_{A}N_{\bullet}. If we set Nn=AknN_{n}=A^{\oplus k_{n}} for some integer kn0k_{n}\geq 0, the nn-th term of the totalization is

(2.1) Totn(MAN)=n=i+j(MiANj)=n=i+jMikj.\operatorname{Tot}_{n}(M_{\bullet}\otimes_{A}N_{\bullet})=\bigoplus_{n=i+j}(M_{i}\otimes_{A}N_{j})=\bigoplus_{n=i+j}M_{i}^{\oplus k_{j}}.

By MMod(A)[a,b]M\in\operatorname{Mod}(A)_{[a,b]}, MM_{\bullet} is quasi-isomorphic to MτbτaMM_{\bullet}^{\prime}\coloneqq\tau_{\leq b}\tau_{\geq a}M_{\bullet}. Since the category of derived II-complete discrete AA-modules consists of an abelian category, the complex MM_{\bullet}^{\prime} is a bounded complex of derived II-complete AA-modules. This shows that the nn-th term (2.1) of MALNM\otimes^{L}_{A}N is a finite direct sum of derived II-complete AA-modules and thus MALNM\otimes^{L}_{A}N is derived II-complete. ∎

As in the flatness, we can show the following.

Lemma 2.11 (cf. [Bha21, Lemma 5.15]).

Let AA be an animated Noetherian ring with a finitely generated ideal Iπ0(A)I\subseteq\pi_{0}(A) and let MM be a derived II-complete AA-module. Suppose that π0(A)\pi_{0}(A) is derived II-complete and MM has II-complete Tor-amplitude in [a,b][a,b]. Then MM has Tor-amplitude in [a,b][a,b].

Proof.

We have to show that MALN(MALπ0(A))π0(A)LNM\otimes^{L}_{A}N\cong(M\otimes^{L}_{A}\pi_{0}(A))\otimes^{L}_{\pi_{0}(A)}N is contained in Mod(A)[a,b]\operatorname{Mod}(A)_{[a,b]} for any discrete AA-module NN. So we can assume that AA is discrete.

Approximating NN by a discrete finitely generated π0(A)\pi_{0}(A)-module, we may assume that NN is a discrete almost perfect AA-module by Lemma 2.8. Since MM is derived II-complete and II-complete Tor-amplitude in [a,b][a,b], M=Rlimn(MALA/In)M=R\lim_{n}(M\otimes^{L}_{A}A/I^{n}) is a bounded AA-module. Using Lemma 2.10, we have MALNM\otimes^{L}_{A}N is derived II-complete. Since pro-systems {NALA/In}n1\{N\otimes^{L}_{A}A/I^{n}\}_{n\geq 1} and {N/InN}n1\{N/I^{n}N\}_{n\geq 1} are pro-isomorphic by the Artin–Rees lemma for Noetherian rings, we have

MALN(MALN)=Rlimn(MALNALA/In)Rlimn(MALN/InN).M\otimes^{L}_{A}N\cong(M\otimes^{L}_{A}N)^{\wedge}=R\lim_{n}(M\otimes^{L}_{A}N\otimes^{L}_{A}A/I^{n})\cong R\lim_{n}(M\otimes^{L}_{A}N/I^{n}N).

By the Milnor exact sequence (see, for example, [GW23, Lemma F.233]), we have the following short exact sequence of AA-modules for each kZk\in\mathbb{Z}:

0R1limnπk+1(MALN/InN)πk(MALN)limnπk(MALN/InN)0.0\to R^{1}\lim_{n}\pi_{k+1}(M\otimes^{L}_{A}N/I^{n}N)\to\pi_{k}(M\otimes^{L}_{A}N)\to\lim_{n}\pi_{k}(M\otimes^{L}_{A}N/I^{n}N)\to 0.

Since MM is II-complete Tor-amplitude in [a,b][a,b], MALN/InNM\otimes^{L}_{A}N/I^{n}N is in Mod(A)[a,b]\operatorname{Mod}(A)_{[a,b]} for all n1n\geq 1. By this assumption, MALNM\otimes^{L}_{A}N is in Mod(A)[a1,b]\operatorname{Mod}(A)_{[a-1,b]}. The exact sequence 0InN/In+1NN/In+1NN/InN00\to I^{n}N/I^{n+1}N\to N/I^{n+1}N\to N/I^{n}N\to 0 induces an exact sequence πa(MALN/In+1N)πa(MALN/InN)πa1(MALInN/In+1N)=0\pi_{a}(M\otimes^{L}_{A}N/I^{n+1}N)\to\pi_{a}(M\otimes^{L}_{A}N/I^{n}N)\to\pi_{a-1}(M\otimes^{L}_{A}I^{n}N/I^{n+1}N)=0. The Mittag–Leffler condition shows that R1limnπa(MALN/InN)R^{1}\lim_{n}\pi_{a}(M\otimes^{L}_{A}N/I^{n}N) vanishes and thus MALNM\otimes^{L}_{A}N is in Mod(A)[a,b]\operatorname{Mod}(A)_{[a,b]}. ∎

The following lemma is one of the stability of some properties under base change.

Lemma 2.12 ([SAG, Proposition 2.7.3.2]).

Let ABA\to B be a surjective555A map f:ABf\colon A\to B of animated rings is surjective if its connected component π0(f):π0(A)π0(B)\pi_{0}(f)\colon\pi_{0}(A)\to\pi_{0}(B) is a surjective map of usual rings. map of animated rings with the nilpotent kernel II of π0(A)π0(B)\pi_{0}(A)\to\pi_{0}(B). Let MM be a connective AA-module and set a connective BB-module MBMALBM_{B}\coloneqq M\otimes^{L}_{A}B. Then we have the following.

  1. (a)

    MM is perfect to order nn over AA if and only if MBM_{B} is perfect to order nn over BB.

  2. (b)

    MM is almost perfect over AA if and only if MBM_{B} is almost perfect over BB.

  3. (c)

    MM has Tor-amplitude k\leq k over AA if and only if MBM_{B} has Tor-amplitude k\leq k over BB.

  4. (d)

    MM is perfect over AA if and only if MBM_{B} is perfect over BB.

  5. (e)

    For each nZn\in\mathbb{Z}, MM is nn-connective if and only if MBM_{B} is nn-connective.

In Lemma 2.12 above, we need to assume that II is nilpotent. However, by using the topological Nakayama’s lemma (see, for example, [Mat86, Theorem 8.4]) instead of the usual Nakayama’s lemma, some proof works under our assumption without the nilpotency of II and we can show the following “topological” variant.

Lemma 2.13 (Topological variant of [SAG, Proposition 2.7.3.2]).

Let ABA\to B be a surjective map of animated rings with the kernel II of π0(A)π0(B)\pi_{0}(A)\to\pi_{0}(B). Let MM be a connective AA-module and set a connective BB-module MBMALBM_{B}\coloneqq M\otimes^{L}_{A}B. Assume that π0(A)\pi_{0}(A) is II-adically complete Noetherian and the π0(A)\pi_{0}(A)-module π0(M)\pi_{0}(M) is II-adically separated. Then we have the following.

  1. (a)

    MM is perfect to order nn over AA if and only if MBM_{B} is perfect to order nn over BB.

  2. (b)

    MM is almost perfect over AA if and only if MBM_{B} is almost perfect over BB.

  3. (c)

    If MM has Tor-amplitude k\leq k over AA, then MBM_{B} has Tor-amplitude k\leq k over BB.

  4. (d)

    If MM is perfect over AA, then MBM_{B} is perfect over BB. If AA is Noetherian, the converse also holds.

  5. (e)

    If πi(M)\pi_{i}(M) is II-adically separated for all i0i\geq 0, MM is nn-connective if and only if MBM_{B} is nn-connective for each nZn\in\mathbb{Z}.

Proof.

The only if part of (a) is the tensor-forgetful adjunction and the definition of ‘perfect to order nn’ (Definition 2.7). As in the proof of [SAG, Proposition 2.7.3.2], we show the if part by induction on nn. The case of n<0n<0 is clear. If n0n\geq 0, we have an isomorphism π0(MB)π0(M)/Iπ0(M)\pi_{0}(M_{B})\cong\pi_{0}(M)/I\pi_{0}(M) and π0(MB)\pi_{0}(M_{B}) is finite over π0(B)π0(A)/I\pi_{0}(B)\cong\pi_{0}(A)/I by [SAG, Proposition 2.7.2.1 (1)]. Since π0(M)\pi_{0}(M) is II-adically separated and π0(A)\pi_{0}(A) is II-adically complete, then π0(M)\pi_{0}(M) is also a finitely generated π0(A)\pi_{0}(A)-module by topological Nakayama’s lemma. If n=0n=0, this shows the claim; MM is perfect to order 0 over AA ([SAG, Proposition 2.7.2.1 (1)]).

For general n>0n>0, one can take a fiber sequence NAkMN\to A^{k}\to M of AA-modules for some k1k\geq 1 with a connective AA-module NN such that π0(A)kπ0(M)\pi_{0}(A)^{k}\to\pi_{0}(M) is surjective. Here we use the above argument to show that π0(M)\pi_{0}(M) is finite over π0(A)\pi_{0}(A). Taking ALB-\otimes^{L}_{A}B, we have a fiber sequence NALBBkMBN\otimes^{L}_{A}B\to B^{k}\to M_{B} of BB-modules. Since MBM_{B} is perfect to order nn, then NALBN\otimes^{L}_{A}B is perfect to order n1n-1 by [SAG, Proposition 2.7.2.1 (2)]. Since π0(A)\pi_{0}(A) is Noetherian, the finite π0(A)\pi_{0}(A)-module π0(N)\pi_{0}(N) is II-adically separated and we can apply the inductive hypothesis for NN, namely, NN is perfect to order n1n-1 over AA. By using [SAG, Proposition 2.7.2.1 (2)] again, MM is perfect to order nn.

The equivalence (b) is by (a) and the definition of almost perfectness.

We next prove (c). By the isomorphism MBBLNMALNM_{B}\otimes^{L}_{B}N\cong M\otimes^{L}_{A}N for any (discrete) BB-module NN.

The conditions (b) and (c) give the first assertion of (d) by [HA, Proposition 7.2.4.23 (4)]. We show the converse if AA is Noetherian. Again by using [HA, Proposition 7.2.4.23 (4)], MBM_{B} is almost perfect and has Tor-amplitude k\leq k over BB for some kk. By (b), MM is almost perfect over AA and it suffices to show that MM has Tor-amplitude k\leq k over AA. Any II-torsion discrete AA-module NN can be regarded as a π0(B)π0(A)/I\pi_{0}(B)\cong\pi_{0}(A)/I-module and also as a BB-module. By the assumption, MALNMBBLNM\otimes^{L}_{A}N\cong M_{B}\otimes^{L}_{B}N is kk-truncated and thus MM has II-complete Tor-amplitude k\leq k over AA. Since MM is almost perfect and π0(A)\pi_{0}(A) is II-adically complete Noetherian, πn(M)\pi_{n}(M) is derived II-complete for any nZn\in\mathbb{Z} and so is MM by [SAG, Theorem 7.3.4.1]. By Lemma 2.11, MM has Tor-amplitude k\leq k over AA and this shows the if part of (d).

Finally, we prove (e). The only if part is the stability of (nn-)connectivity under tensor products ([HA, Corollary 7.2.1.23]). The proof of the if part goes by induction on nn. The case of n0n\leq 0 is clear since MM is connective. If n>0n>0 and MBM_{B} is nn-connective, then MM is (n1)(n-1)-connective by the inductive hypothesis. This shows that 0πn1(MB)πn1(M)/Iπn1(M)0\cong\pi_{n-1}(M_{B})\cong\pi_{n-1}(M)/I\pi_{n-1}(M) and thus πn1(M)=0\pi_{n-1}(M)=0 by topological Nakayama’s lemma since πn1(M)\pi_{n-1}(M) is II-adically separated. ∎

Lemma 2.14.

Let AA be an animated Λ\Lambda-algebra. Then AA is almost perfect as an Λ\Lambda-module if and only if AA is Noetherian and π0(A)\pi_{0}(A) is finite over Λ\Lambda. In particular, if AA is almost perfect over Λ\Lambda, then πn(A)\pi_{n}(A) is complete with respect to the maximal ideal 𝔪\mathfrak{m} of Λ\Lambda for all nZn\in\mathbb{Z} and in particular AA is derived 𝔪\mathfrak{m}-complete.

Proof.

If AA is almost perfect as an Λ\Lambda-module, then πm(A)\pi_{m}(A) is finite over Λ=π0(Λ)\Lambda=\pi_{0}(\Lambda) for all mZm\in\mathbb{Z} by Lemma 2.8. In particular, π0(A)\pi_{0}(A) is Noetherian and πm(A)\pi_{m}(A) is finite over the Λ\Lambda-algebra π0(A)\pi_{0}(A). This shows AA is Noetherian. Conversely, if AA is Noetherian and π0(A)\pi_{0}(A) is finite over Λ\Lambda, then πm(A)\pi_{m}(A) is finite over Λ\Lambda for all mZm\in\mathbb{Z}. By using Lemma 2.8 again, AA is almost perfect as an Λ\Lambda-module.

The last assertion follows from the fact that πm(A)\pi_{m}(A) is finite over Λ\Lambda for all mZm\in\mathbb{Z} since any homologically 𝔪\mathfrak{m}-adically complete Λ\Lambda-module is derived 𝔪\mathfrak{m}-complete (for example [SAG, Theorem 7.3.4.1]). ∎

In the last of the proof of our main theorem, we use the following lemmas (Lemma 2.15 and Lemma 2.16) which are basic results in the discrete case.

Lemma 2.15.

Let AA be a coherent animated ring and let MM be a connective perfect AA-module. Then the following are equivalent.

  1. (1)

    MM is projective over AA in the sense of [HA, Proposition 7.2.2.6].

  2. (2)

    The Ext\operatorname{Ext}-module ExtAi(M,A)\operatorname{Ext}^{i}_{A}(M,A) vanishes for i1i\geq 1.

Proof.

The projectivity of MM over AA is equivalent to the vanishing of ExtAi(M,Q)=0\operatorname{Ext}^{i}_{A}(M,Q)=0 for any discrete (or, equivalently, connective) AA-module QQ and for any i1i\geq 1 by [HA, Proposition 7.2.2.6]. So (1) implies (2). We will show (2) \Rightarrow (1). We can assume that QQ is a discrete finitely generated AA-module, in particular, QQ is connective and almost perfect over AA by [HA, Proposition 7.1.1.13 (3) and Proposition 7.2.4.17]. As in the proof of [HA, Proposition 7.2.4.11 (5)], QQ can be represented as the colimit of a sequence of AA-modules

0=Q(1)f0Q(0)f1Q(1)f2Q(2)0=Q(-1)\xrightarrow{f_{0}}Q(0)\xrightarrow{f_{1}}Q(1)\xrightarrow{f_{2}}Q(2)\to\cdots

such that Q(j)Q(j) is connective, cofib(fj)[j]\operatorname{cofib}(f_{j})[-j] is a free AA-module AnjA^{\oplus n_{j}} of finite rank njn_{j} for j0j\geq 0.666More generally, by the same proof of [HA, Proposition 7.2.4.11], we can show that any connective AA-module QQ is the colimit of a sequence 0=Q(1)f0Q(1)f1Q(2)f20=Q(-1)\xrightarrow{f_{0}}Q(1)\xrightarrow{f_{1}}Q(2)\xrightarrow{f_{2}}\cdots of connective AA-modules such that cofib(fj)[j]\operatorname{cofib}(f_{j})[-j] is a free AA-module for any j0j\geq 0. This representation has the same spirit as the semifree resolution of DG AA-modules (see for example [NOY22, §2.12]). Applying mapA(M,())\operatorname{map}_{A}(M,(-)) for the (co)fiber sequence Q(j1)fjQ(j)cofib(fj)Q(j-1)\xrightarrow{f_{j}}Q(j)\to\operatorname{cofib}(f_{j}) for each j0j\geq 0, we have an exact sequence

ExtAi(M,Q(j1))ExtAi(M,Q(j))ExtAi(M,cofib(fj))(ExtAi+j(M,A))nj\operatorname{Ext}^{i}_{A}(M,Q(j-1))\to\operatorname{Ext}^{i}_{A}(M,Q(j))\to\operatorname{Ext}^{i}_{A}(M,\operatorname{cofib}(f_{j}))\cong(\operatorname{Ext}^{i+j}_{A}(M,A))^{\oplus n_{j}}

for each i1i\geq 1 by virtue of the equivalence cofib(fj)Anj[j]\operatorname{cofib}(f_{j})\cong A^{\oplus n_{j}}[j]. Since ExtAk(M,A)\operatorname{Ext}^{k}_{A}(M,A) vanishes for any k1k\geq 1 by the assumption, ExtAi(M,Q(j1))ExtAi(M,Q(j))\operatorname{Ext}^{i}_{A}(M,Q(j-1))\xrightarrow{\cong}\operatorname{Ext}^{i}_{A}(M,Q(j)) is an isomorphism for each i,j1i,j\geq 1. Since MM is perfect over AA, the following isomorphisms hold for each i1i\geq 1:

ExtAi(M,Q)\displaystyle\operatorname{Ext}^{i}_{A}(M,Q) ExtAi(M,colimjQ(j))colimjExtAi(M,Q(j))ExtAi(M,Q(0)).\displaystyle\cong\operatorname{Ext}^{i}_{A}(M,\operatorname*{colim}_{j}Q(j))\cong\operatorname*{colim}_{j}\operatorname{Ext}^{i}_{A}(M,Q(j))\cong\operatorname{Ext}^{i}_{A}(M,Q(0)).

By the construction of the sequence {Q(j)}j1\{Q(j)\}_{j\geq-1}, Q(0)cofib(f0)[0]Q(0)\cong\operatorname{cofib}(f_{0})[-0] is a finite free AA-module. Our assumption, the vanishing of ExtAi(M,A)\operatorname{Ext}^{i}_{A}(M,A) for i1i\geq 1, shows that ExtAi(M,Q)\operatorname{Ext}^{i}_{A}(M,Q) vanishes for any i1i\geq 1 and thus MM is projective over AA. ∎

Lemma 2.16.

Let AA be an animated Noetherian ring and let LL and LL^{\prime} be almost perfect AA-modules. If LL^{\prime} is kk-truncated, then the Ext\operatorname{Ext}-module ExtAi(L,L)\operatorname{Ext}^{i}_{A}(L,L^{\prime}) is finitely generated over π0(A)\pi_{0}(A) for any iZi\in\mathbb{Z}. In particular, if LL is an almost perfect discrete AA-module, then ExtAi(L,L)\operatorname{Ext}^{i}_{A}(L,L) is finitely generated over π0(A)\pi_{0}(A) for any iZi\in\mathbb{Z}.

Proof.

Fix an integer nn. Our assumption implies that LL is perfect to order nn. By [SAG, Corollary 2.7.2.2], there exists a fiber sequence FPLF\to P\to L such that PP is a perfect AA-module and FF is nn-connective (that is, πi(F)=0\pi_{i}(F)=0 for i<ni<n). Taking mapA(,L)\operatorname{map}_{A}(-,L^{\prime}) and its long exact sequence, we have an exact sequence of π0(A)\pi_{0}(A)-modules

ExtAi1(F,L)ExtAi(L,L)ExtAi(P,L)ExtAi(F,L).\displaystyle\operatorname{Ext}^{i-1}_{A}(F,L^{\prime})\to\operatorname{Ext}^{i}_{A}(L,L^{\prime})\to\operatorname{Ext}^{i}_{A}(P,L^{\prime})\to\operatorname{Ext}^{i}_{A}(F,L^{\prime}).

Considering the tt-structure on Mod(A)\operatorname{Mod}(A) (or simply on hMod(A)\mathrm{h}\operatorname{Mod}(A)) and the nn-connectivity of FF, the first and the last terms vanish for ink1i\leq n-k-1 since LL^{\prime} is kk-truncated. We have to show that ExtAi(P,L)\operatorname{Ext}^{i}_{A}(P,L^{\prime}) is finitely generated over π0(A)\pi_{0}(A). Since PP is perfect over AA, the AA-module mapA(P,L)\operatorname{map}_{A}(P,L^{\prime}) is equivalent to PALLP^{\vee}\otimes^{L}_{A}L^{\prime} for some perfect AA-module PP^{\vee} by [HA, Proposition 7.2.4.4]. Since LL^{\prime} and PP^{\vee} are almost perfect over AA, so is the AA-module PALLP^{\vee}\otimes^{L}_{A}L^{\prime}. In particular, the homology group πi(PALL)πi(mapA(P,L))=ExtAi(P,L)\pi_{-i}(P^{\vee}\otimes^{L}_{A}L^{\prime})\cong\pi_{-i}(\operatorname{map}_{A}(P,L^{\prime}))=\operatorname{Ext}^{i}_{A}(P,L^{\prime}) is finite over π0(A)\pi_{0}(A) for any iZi\in\mathbb{Z} by Lemma 2.8 and so is ExtAi(L,L)\operatorname{Ext}^{i}_{A}(L,L^{\prime}) for ink1i\leq n-k-1 by the above exact sequence. This nn can be taken arbitrarily large and we have the desired result. The second statement follows from the first statement by taking L=LL^{\prime}=L and k=0k=0. ∎

3. Derived Quotients

In this section, we recall the notion of derived quotients and calculate some homotopy groups of (derived) tensor products (Lemma 3.5 and Lemma 3.6). Recall that (Λ,𝔪)(\Lambda,\mathfrak{m}) is a complete Noetherian local ring.

Definition 3.1.

Let AA be an animated ring and let f¯=f1,,fr\underline{f}=f_{1},\dots,f_{r} be a sequence of elements of π0(A)\pi_{0}(A). Take an AA-module MM. We denote by M/Lf¯=M/L(f1,,fr)M/^{L}\underline{f}=M/^{L}(f_{1},\dots,f_{r}) the derived quotient of MM by f¯\underline{f}, that is,

M/Lf¯MZ[X1,,Xt]LZMAL(A/Lf¯)Mod(A/Lf¯),M/^{L}\underline{f}\coloneqq M\otimes^{L}_{\mathbb{Z}[X_{1},\dots,X_{t}]}\mathbb{Z}\cong M\otimes^{L}_{A}(A/^{L}\underline{f})\in\operatorname{Mod}(A/^{L}\underline{f}),

where MZ[X1,,Xt]ZM\leftarrow\mathbb{Z}[X_{1},\dots,X_{t}]\to\mathbb{Z} is the map ×xiXi0\times x_{i}\mapsfrom X_{i}\mapsto 0. Note that A/Lf¯A/^{L}\underline{f} is an animated AA-algebra. If AA is discrete, the underlying complex of M/Lx¯M/^{L}\underline{x} is the Koszul complex Kos(M;x¯)MALKos(A;x¯)\operatorname{Kos}(M;\underline{x})\coloneqq M\otimes^{L}_{A}\operatorname{Kos}(A;\underline{x}) in D(A)D(A).

By considering the exact sequence 0Z[X]×XZ[X]Z00\to\mathbb{Z}[X]\xrightarrow{\times X}\mathbb{Z}[X]\to\mathbb{Z}\to 0, we have a fiber sequence in Mod(A)\operatorname{Mod}(A)

(3.1) M×xMM/Lx.M\xrightarrow{\times x}M\to M/^{L}x.

Moreover, if BB is an animated AA-algebra, then the derived quotient B/L(f1,,fr)B/^{L}(f_{1},\dots,f_{r}) has a natural structure of animated BB-algebras and a morphism BB/L(f1,,fr)B\to B/^{L}(f_{1},\dots,f_{r}) of animated rings since it is defined by a homotopy cocartesian square in CAlgZan\operatorname{CAlg}^{an}_{\mathbb{Z}} (see, for example, [KR19, 2.3.1]).

Notation 3.2.

Let AA be an animated Λ\Lambda-algebra and let xx be an element of 𝔪\mathfrak{m}. For n1n\geq 1, we set the derived quotient

AnA/Lxn=AZ[X]LZCAlgΛan.A_{n}\coloneqq A/^{L}x^{n}=A\otimes^{L}_{\mathbb{Z}[X]}\mathbb{Z}\in\operatorname{CAlg}^{an}_{\Lambda}.

Then we have maps of animated Λ\Lambda-algebras

AA1A2An.A\to A_{1}\leftarrow A_{2}\leftarrow\dots\leftarrow A_{n}\leftarrow\cdots.

If AA is Noetherian, so is AnA_{n} by Definition 2.5. Note that this canonical map An+1AnA_{n+1}\to A_{n} depends on nn and xx. Even if x=0x=0 and AA is discrete, this map is not an isomorphism in general (consider the map on π1\pi_{1}).

Lemma 3.3.

Let AA be an animated Λ\Lambda-algebra and let xx be an element of 𝔪\mathfrak{m}. Let MM be an AA-module. Then we have an isomorphism of AA-modules:

M/LxnMZ[X]LZ[X]/(Xn),M/^{L}x^{n}\cong M\otimes^{L}_{\mathbb{Z}[X]}\mathbb{Z}[X]/(X^{n}),

where MZ[X]Z[X]/(Xn)M\leftarrow\mathbb{Z}[X]\to\mathbb{Z}[X]/(X^{n}) is the map ×xXX¯\times x\mapsfrom X\mapsto\overline{X}. In particular, the following isomorphisms hold:

Am/Lxn\displaystyle A_{m}/^{L}x^{n} AmALAnCAlgΛan\displaystyle\cong A_{m}\otimes^{L}_{A}A_{n}\in\operatorname{CAlg}^{an}_{\Lambda}
M/Lxn\displaystyle M/^{L}x^{n} MALAnMod(An)\displaystyle\cong M\otimes^{L}_{A}A_{n}\in\operatorname{Mod}(A_{n})

for any m,n0m,n\geq 0. In particular, AiA_{i} is isomorphic to AZ[X]LZ[X]/(Xi)A\otimes^{L}_{\mathbb{Z}[X]}\mathbb{Z}[X]/(X^{i}) as an AnA_{n}-module via the canonical map AnAZ[X]LZ[X]/(Xn)AZ[X]LZ[X]/(Xi)AiA_{n}\cong A\otimes^{L}_{\mathbb{Z}[X]}\mathbb{Z}[X]/(X^{n})\to A\otimes^{L}_{\mathbb{Z}[X]}\mathbb{Z}[X]/(X^{i})\cong A_{i} for 1in11\leq i\leq n-1.

Proof.

Taking Z[X]LM-\otimes^{L}_{\mathbb{Z}[X]}M for the fiber sequence Z[X]×XnZ[X]Z[X]/(Xn)\mathbb{Z}[X]\xrightarrow{\times X^{n}}\mathbb{Z}[X]\to\mathbb{Z}[X]/(X^{n}), we have a fiber sequence of AA-modules

M×xnMMZ[X]LZ[X]/(Xn).M\xrightarrow{\times x^{n}}M\to M\otimes^{L}_{\mathbb{Z}[X]}\mathbb{Z}[X]/(X^{n}).

This shows M/LxnMZ[X]LZ[X]/(Xn)M/^{L}x^{n}\cong M\otimes^{L}_{\mathbb{Z}[X]}\mathbb{Z}[X]/(X^{n}). By this isomorphism, AnAZ[X]LZ[X]/(Xn)A_{n}\cong A\otimes^{L}_{\mathbb{Z}[X]}\mathbb{Z}[X]/(X^{n}) holds. So we have the following isomorphism of AnA_{n}-modules:

MALAn\displaystyle M\otimes^{L}_{A}A_{n} MAL(AZ[X]LZ[X]/(Xn))MZ[X]LZ[X]/(Xn)M/Lxn.\displaystyle\cong M\otimes^{L}_{A}(A\otimes^{L}_{\mathbb{Z}[X]}\mathbb{Z}[X]/(X^{n}))\cong M\otimes^{L}_{\mathbb{Z}[X]}\mathbb{Z}[X]/(X^{n})\cong M/^{L}x^{n}.

If MM is an animated Λ\Lambda-algebra AmA_{m}, the isomorphism Am/LxnAmALAnA_{m}/^{L}x^{n}\cong A_{m}\otimes^{L}_{A}A_{n} in Mod(An)\operatorname{Mod}(A_{n}) becomes an isomorphism between animated Λ\Lambda-algebras. ∎

Lemma 3.4.

Let AA be an animated Λ\Lambda-algebra and let xx be an element of 𝔪\mathfrak{m}. Then there exists a fiber sequence in Mod(An+1)\operatorname{Mod}(A_{n+1})

Ak×xn+1kAn+1An+1kA_{k}\xrightarrow{\times x^{n+1-k}}A_{n+1}\to A_{n+1-k}

for any n1n\geq 1 and 1kn1\leq k\leq n.

Proof.

By the isomorphism AnAZ[X]LZ[X]/(Xn)A_{n}\cong A\otimes^{L}_{\mathbb{Z}[X]}\mathbb{Z}[X]/(X^{n}) in Lemma 3.3, we have a fiber sequence in Mod(An+1)\operatorname{Mod}(A_{n+1})

AZ[X]LZ[X]/(Xk)×Xn+1kAZ[X]LZ[X]/(Xn+1)AZ[X]LZ[X]/(Xn+1k).A\otimes^{L}_{\mathbb{Z}[X]}\mathbb{Z}[X]/(X^{k})\xrightarrow{\times X^{n+1-k}}A\otimes^{L}_{\mathbb{Z}[X]}\mathbb{Z}[X]/(X^{n+1})\to A\otimes^{L}_{\mathbb{Z}[X]}\mathbb{Z}[X]/(X^{n+1-k}).

This gives the desired fiber sequence. ∎

Lemma 3.5.

Let AA be an animated Λ\Lambda-algebra and let xx be an element of 𝔪\mathfrak{m}. For any discrete AA-module MM, we have

TorkA(An,M)πk(AnALM){M/xnMif k=0M[xn]if k=10otherwise,\operatorname{Tor}_{k}^{A}(A_{n},M)\coloneqq\pi_{k}(A_{n}\otimes^{L}_{A}M)\cong\begin{cases}M/x^{n}M&\text{if }k=0\\ M[x^{n}]&\text{if }k=1\\ 0&\text{otherwise},\end{cases}

where M[xj]{mMxjm=0}M[x^{j}]\coloneqq\{m\in M\mid x^{j}m=0\mbox{}\} is the xjx^{j}-torsion submodule of MM for j0j\geq 0.

Proof.

We have a fiber sequence A×xnAAnA\xrightarrow{\times x^{n}}A\to A_{n} of AA-modules. Then AnALMA_{n}\otimes^{L}_{A}M is the cofiber of M×xnMM\xrightarrow{\times x^{n}}M in Mod(A)\operatorname{Mod}(A). Therefore, we have the desired calculation. ∎

The following calculation is a generalization of the same calculation for a regular element xx proved in [ADS93, Lemma 1.1] and a key to the proof of Lemma 4.6. The proof is given first, and the necessary preparations are described afterwards (Remark 3.7Corollary 3.10).

Lemma 3.6.

Let AA be an animated Λ\Lambda-algebra and let xx be an element of 𝔪\mathfrak{m}. Let MM be a discrete AnA_{n}-module. For each 1in11\leq i\leq n-1, we have the following isomorphisms of discrete π0(An)\pi_{0}(A_{n})-modules:

πk(MAnLAi){M/xiMif k=0M[xi]/xniMif k is oddM[xni]/xiMif k0 is even.\pi_{k}(M\otimes^{L}_{A_{n}}A_{i})\cong\begin{cases}M/x^{i}M&\text{if }k=0\\ M[x^{i}]/x^{n-i}M&\text{if $k$ is odd}\\ M[x^{n-i}]/x^{i}M&\text{if $k\neq 0$ is even}.\end{cases}
Proof.

The Z[X]\mathbb{Z}[X]-module Z[X]/(Xi)\mathbb{Z}[X]/(X^{i}) is quasi-isomorphic to the complex of Z[X]\mathbb{Z}[X]-modules

Z×XiZ[X]/(Xn)×XniZ[X]/(Xn)×XiZ[X]/(Xn)0.Z\coloneqq\cdots\xrightarrow{\times X^{i}}\mathbb{Z}[X]/(X^{n})\xrightarrow{\times X^{n-i}}\mathbb{Z}[X]/(X^{n})\xrightarrow{\times X^{i}}\mathbb{Z}[X]/(X^{n})\to 0.

Set the complex Z={Zk}kZZ=\{Z_{k}\}_{k\in\mathbb{Z}} and the quasi-isomorphism ZZ[X]/(Xi)Z\to\mathbb{Z}[X]/(X^{i}) in Ch(Z[X])\operatorname{Ch}(\mathbb{Z}[X]). By the Dold–Kan correspondence DK:Ch0(Ch(Z[X]))Fun(Δop,Ch(Z[X]))\operatorname{DK}\colon\operatorname{Ch}_{\geq 0}(\operatorname{Ch}(\mathbb{Z}[X]))\to\operatorname{Fun}(\Delta^{\operatorname{op}},\operatorname{Ch}(\mathbb{Z}[X])), we have a simplicial object DK(Z)\operatorname{DK}_{\bullet}(Z^{\prime}) in Ch(Z[X])\operatorname{Ch}(\mathbb{Z}[X]) for the object

(3.2) Z(×XniZ[X]/(Xn)[0]×XiZ[X]/(Xn)[0]0)Z^{\prime}\coloneqq(\cdots\xrightarrow{\times X^{n-i}}\mathbb{Z}[X]/(X^{n})[0]\xrightarrow{\times X^{i}}\mathbb{Z}[X]/(X^{n})[0]\to 0)

in Ch0(Ch(Z[X]))\operatorname{Ch}_{\geq 0}(\operatorname{Ch}(\mathbb{Z}[X])). Applying Corollary 3.10 for the simplicial Z[X]\mathbb{Z}[X]-module DK(Z)\operatorname{DK}_{\bullet}(Z^{\prime}), the geometric realization |DK(Z)|\lvert\operatorname{DK}_{\bullet}(Z^{\prime})\rvert in 𝒟(Z[X])\mathcal{D}(\mathbb{Z}[X]) is isomorphic to the totalization of ZCh0(Ch(Z[X]))Z^{\prime}\in\operatorname{Ch}_{\geq 0}(\operatorname{Ch}(\mathbb{Z}[X])), namely,

(3.3) |DK(Z)|ιZ[X](Z)Z[X]/(Xi)\lvert\operatorname{DK}_{\bullet}(Z^{\prime})\rvert\cong\iota_{\mathbb{Z}[X]}(Z)\cong\mathbb{Z}[X]/(X^{i})

in 𝒟(Z[X])=Mod(Z[X])\mathcal{D}(\mathbb{Z}[X])=\operatorname{Mod}(\mathbb{Z}[X]) where ιZ[X]\iota_{\mathbb{Z}[X]} is defined in Remark 3.7.

By Lemma 3.3, AiA_{i} is isomorphic to AZ[X]LZ[X]/(Xi)A\otimes^{L}_{\mathbb{Z}[X]}\mathbb{Z}[X]/(X^{i}) as an AnA_{n}-module. The isomorphism (3.3) is Mod(Z[X])\operatorname{Mod}(\mathbb{Z}[X]) provides the following isomorphisms of AnA_{n}-modules:

MAnLAi\displaystyle M\otimes^{L}_{A_{n}}A_{i} MAnL(AZ[X]LZ[X]/(Xi))MAnL(AZ[X]LZ)\displaystyle\cong M\otimes^{L}_{A_{n}}(A\otimes^{L}_{\mathbb{Z}[X]}\mathbb{Z}[X]/(X^{i}))\cong M\otimes^{L}_{A_{n}}(A\otimes^{L}_{\mathbb{Z}[X]}Z)
MAnL(AZ[X]L|DK(Z)|)|MAnL(AZ[X]LDK(Z))|.\displaystyle\cong M\otimes^{L}_{A_{n}}(A\otimes^{L}_{\mathbb{Z}[X]}\lvert\operatorname{DK}_{\bullet}(Z^{\prime})\rvert)\cong\lvert M\otimes^{L}_{A_{n}}(A\otimes^{L}_{\mathbb{Z}[X]}\operatorname{DK}_{\bullet}(Z^{\prime}))\rvert.

Recall that the simplicial Z[X]\mathbb{Z}[X]-module DK(Z)\operatorname{DK}_{\bullet}(Z^{\prime})777In [HA, Construction 1.2.3.5], the value of the functor DK\operatorname{DK} is an additive category. However, since any stable \infty-category 𝒞\mathcal{C} has finite colimits, we can construct the simplicial object DK(Z)\operatorname{DK}_{\bullet}(Z^{\prime}) in 𝒞\mathcal{C} for each non-negatively chain complex with a value 𝒞\mathcal{C}. is defined by

DKn(Z)α:[n][k]ZkMod(Z[X]),\operatorname{DK}_{n}(Z^{\prime})\coloneqq\bigoplus_{\alpha\colon[n]\twoheadrightarrow[k]}Z^{\prime}_{k}\in\operatorname{Mod}(\mathbb{Z}[X]),

where Zk=Z[X]/(Xn)[0]Z^{\prime}_{k}=\mathbb{Z}[X]/(X^{n})[0] is the kk-th term of the complex ZCh0(Ch(Z[X]))Z^{\prime}\in\operatorname{Ch}_{\geq 0}(\operatorname{Ch}(\mathbb{Z}[X])) (3.2) and the sum is taken over all surjections α:[n][k]\alpha\colon[n]\twoheadrightarrow[k] in the simplex category Δ\Delta. So the nn-th term of the simplicial AnA_{n}-module MAnL(AZ[X]LDK(Z))M\otimes^{L}_{A_{n}}(A\otimes^{L}_{\mathbb{Z}[X]}\operatorname{DK}_{\bullet}(Z^{\prime})) can be written as

(3.4) MAnL(AZ[X]LDKn(Z))=MAnL(AZ[X]Lα:[n][k]Zk)\displaystyle M\otimes^{L}_{A_{n}}(A\otimes^{L}_{\mathbb{Z}[X]}\operatorname{DK}_{n}(Z^{\prime}))=M\otimes^{L}_{A_{n}}(A\otimes^{L}_{\mathbb{Z}[X]}\bigoplus_{\alpha\colon[n]\twoheadrightarrow[k]}Z_{k})
α:[n][k](MAnL(AZ[X]LZk))α:[n][k]M.\displaystyle\cong\bigoplus_{\alpha\colon[n]\twoheadrightarrow[k]}(M\otimes^{L}_{A_{n}}(A\otimes^{L}_{\mathbb{Z}[X]}Z_{k}))\cong\bigoplus_{\alpha\colon[n]\twoheadrightarrow[k]}M.

Here, we use the isomorphism AnAZ[X]LZ[X]/(Xn)=AZ[X]LZkA_{n}\cong A\otimes^{L}_{\mathbb{Z}[X]}\mathbb{Z}[X]/(X^{n})=A\otimes^{L}_{\mathbb{Z}[X]}Z_{k} of AnA_{n}-modules described in Lemma 3.3 for each k0k\geq 0. The map β:DKn(Z)DKn(Z)\beta^{*}\colon\operatorname{DK}_{n}(Z^{\prime})\to\operatorname{DK}_{n^{\prime}}(Z^{\prime}) for β:[n][n]\beta\colon[n^{\prime}]\to[n] in Δ\Delta is defined by an appropriate sum of the identity map on ZkZ_{k} and the differential, ×Xni\times X^{n-i} and ×Xi\times X^{i}. Through the above canonical isomorphism (3.4), the simplicial AnA_{n}-module MAnL(AZ[X]LDK(Z))M\otimes^{L}_{A_{n}}(A\otimes^{L}_{\mathbb{Z}[X]}\operatorname{DK}_{\bullet}(Z^{\prime})) is isomorphic to DK(×xniM×xiM0)\operatorname{DK}_{\bullet}(\cdots\xrightarrow{\times x^{n-i}}M\xrightarrow{\times x^{i}}M\to 0), that is, we have an isomorphism in Fun(Δop,Mod(An))\operatorname{Fun}(\Delta^{\operatorname{op}},\operatorname{Mod}(A_{n}));

(3.5) MAnL(AZ[X]LDK(Z))DK(×xniM×xiM0).M\otimes^{L}_{A_{n}}(A\otimes^{L}_{\mathbb{Z}[X]}\operatorname{DK}_{\bullet}(Z^{\prime}))\cong\operatorname{DK}_{\bullet}(\cdots\xrightarrow{\times x^{n-i}}M\xrightarrow{\times x^{i}}M\to 0).

The right-hand side is the simplicial AnA_{n}-module corresponds to the non-negatively chain complex of AnA_{n}-modules (×xniM×xiM0)Ch0(Mod(An))(\cdots\xrightarrow{\times x^{n-i}}M\xrightarrow{\times x^{i}}M\to 0)\in\operatorname{Ch}_{\geq 0}(\operatorname{Mod}(A_{n})). This complex is the image of the double complex of discrete π0(An)\pi_{0}(A_{n})-modules (×xniM[0]×xiM[0]0)(\cdots\xrightarrow{\times x^{n-i}}M[0]\xrightarrow{\times x^{i}}M[0]\to 0) via the (chain complex of the) canonical functor

(3.6) Ch(π0(An))ιπ0(An)Mod(π0(An))Mod(An),\operatorname{Ch}(\pi_{0}(A_{n}))\xrightarrow{\iota_{\pi_{0}(A_{n})}}\operatorname{Mod}(\pi_{0}(A_{n}))\to\operatorname{Mod}(A_{n}),

which maps M[0]M[0] to MM and preserves coproducts. The former functor ιπ0(An)\iota_{\pi_{0}(A_{n})} is defined by Remark 3.7 and the latter one is the restriction of scalars. Set the simplicial object FDK(×xniM[0]×xiM[0]0)F\coloneqq\operatorname{DK}_{\bullet}(\cdots\xrightarrow{\times x^{n-i}}M[0]\xrightarrow{\times x^{i}}M[0]\to 0) in Ch(π0(An))\operatorname{Ch}(\pi_{0}(A_{n})). The right-hand side of (3.5) is the restriction of scalar to AnA_{n} of the simplicial π0(An)\pi_{0}(A_{n})-module ιπ0(An)(F)\iota_{\pi_{0}(A_{n})}(F). We need to calculate the colimit |ιπ0(An)(F)|=colimΔopιπ0(An)(F)\lvert\iota_{\pi_{0}(A_{n})}(F)\rvert=\operatorname*{colim}_{\Delta^{\operatorname{op}}}\iota_{\pi_{0}(A_{n})}(F) in 𝒟(π0(An))\mathcal{D}(\pi_{0}(A_{n})). Applying Corollary 3.10 for FF, the geometric realization |ιπ0(An)(F)|\lvert\iota_{\pi_{0}(A_{n})}(F)\rvert in 𝒟(π0(An))\mathcal{D}(\pi_{0}(A_{n})) is isomorphic to

|ιπ0(An)DK(×xniM[0]×xiM[0]0)|ιπ0(An)(×xniM×xiM0).\lvert\iota_{\pi_{0}(A_{n})}\operatorname{DK}_{\bullet}(\cdots\xrightarrow{\times x^{n-i}}M[0]\xrightarrow{\times x^{i}}M[0]\to 0)\rvert\cong\iota_{\pi_{0}(A_{n})}(\cdots\xrightarrow{\times x^{n-i}}M\xrightarrow{\times x^{i}}M\to 0).

By using the colimit preserving property of the restriction of scalars ([HA, Corollary 4.2.3.7 (2)]), we have the following isomorphism in 𝒟(An)\mathcal{D}(A_{n}):

MAnLAi\displaystyle M\otimes^{L}_{A_{n}}A_{i} |MAnL(AZ[X]LDK(Z))||DK(×xniM×xiM0)|\displaystyle\cong\lvert M\otimes^{L}_{A_{n}}(A\otimes^{L}_{\mathbb{Z}[X]}\operatorname{DK}_{\bullet}(Z))\rvert\cong\lvert\operatorname{DK}_{\bullet}(\cdots\xrightarrow{\times x^{n-i}}M\xrightarrow{\times x^{i}}M\to 0)\rvert
ιπ0(An)(×xniM×xiM0)\displaystyle\cong\iota_{\pi_{0}(A_{n})}(\cdots\xrightarrow{\times x^{n-i}}M\xrightarrow{\times x^{i}}M\to 0)

by using the above calculation and (3.5). This shows that the ii-th homotopy group of MAnLAiM\otimes^{L}_{A_{n}}A_{i} is the desired one.

In the above proof, we use the following observation.

Remark 3.7.

Let AA be a discrete ring. Recall that the \infty-category of AA-modules Mod(A)\operatorname{Mod}(A) is equivalent to the derived \infty-category 𝒟(A)Ndg(Ch(A))\mathcal{D}(A)\coloneqq\operatorname{N_{dg}}(\operatorname{Ch}(A)^{\circ}) of (discrete) AA-modules (see [HA, Definition 1.3.5.8 and Remark 7.1.1.16]). When emphasizing the model-theoretical construction of 𝒟(A)\mathcal{D}(A), we use the notation 𝒟(A)\mathcal{D}(A) instead of Mod(A)\operatorname{Mod}(A).

By [HA, Proposition 1.3.5.15], the functor Ch(A)Ndg(Ch(A))𝐿𝒟(A)\operatorname{Ch}(A)\to\operatorname{N_{dg}}(\operatorname{Ch}(A))\xrightarrow{L}\mathcal{D}(A) exhibits 𝒟(A)\mathcal{D}(A) as the underlying \infty-category of the model category Ch(A)\operatorname{Ch}(A) with a model structure ([HA, Proposition 1.3.5.3]). Namely, the functor of \infty-categories

ιA:Ch(A)Ndg(Ch(A))𝐿𝒟(A)\iota_{A}\colon\operatorname{Ch}(A)\to\operatorname{N_{dg}}(\operatorname{Ch}(A))\xrightarrow{L}\mathcal{D}(A)

induces the equivalence of \infty-categories Ch(A)[W1]𝒟(A)\operatorname{Ch}(A)[W^{-1}]\cong\mathcal{D}(A) where WW is the class of quasi-isomorphisms in Ch(A)\operatorname{Ch}(A).

To calculate the geometric realization, we use the following construction. Our notation is based on [Bun13, Problem 4.24] and [Ara25, Definition 2.2].

Construction 3.8.

Let AA be a discrete ring and let FF be a simplicial object in Ch(A)\operatorname{Ch}(A). The totalization tot(F)\operatorname{tot}(F) of FF is the totalization Tot(M(F))\operatorname{Tot}^{\oplus}(M_{*}(F)) of the Moore complex M(F)M_{*}(F) of the simplicial chain complex FF. Explicitly, the totalization tot(F)\operatorname{tot}(F) is defined as follows:

tot(F)nTot(M(F))n=n=p+qM(F)p,q=n=q+pF([q])p\operatorname{tot}(F)_{n}\coloneqq\operatorname{Tot}^{\oplus}(M_{*}(F))_{n}=\bigoplus_{n=p+q}M_{*}(F)_{p,q}=\bigoplus_{n=q+p}F([q])_{p}

with the differential

dp,q(1)qdpF([q])+i=0q(1)iiF([q])p:F([q])pF([q])p1F([q1])pd_{p,q}\coloneqq(-1)^{q}d^{F([q])}_{p}+\sum_{i=0}^{q}(-1)^{i}\partial^{F([q])_{p}}_{i}\colon F([q])_{p}\to F([q])_{p-1}\oplus F([q-1])_{p}

where dpF([q]):F([q])pF([q])p1d^{F([q])}_{p}\colon F([q])_{p}\to F([q])_{p-1} is the differential of the chain complex F([q])Ch(A)F([q])\in\operatorname{Ch}(A) and i=0q(1)iiFp([q]):F([q])pF([q1])p\sum_{i=0}^{q}(-1)^{i}\partial^{F_{p}([q])}_{i}\colon F([q])_{p}\to F([q-1])_{p} is the alternating sum of the face maps {iFp([q])}\{\partial^{F_{p}([q])}_{i}\} of the simplicial discrete AA-module FpFun(Δop,Ch(A))F_{p}\in\operatorname{Fun}(\Delta^{\operatorname{op}},\operatorname{Ch}(A)^{\heartsuit}).

By using this, we have the following lemma which is a special case of [Ara25, Proposition 2.4 (1) and Lemma 2.8] but we recall the proof in our case.

Lemma 3.9.

Let AA be a discrete ring and let FF be a simplicial object in Ch(A)\operatorname{Ch}(A). Under the notation in Remark 3.7, we can show the following isomorphism holds:

colimΔopιAFιA(tot(F))\operatorname*{colim}_{\Delta^{\operatorname{op}}}\iota_{A}F\cong\iota_{A}(\operatorname{tot}(F))

in 𝒟(A)\mathcal{D}(A).

Proof.

Recall that the restriction of scalar 𝒟(A)𝒟(Z)\mathcal{D}(A)\to\mathcal{D}(\mathbb{Z}) is faithful (on the homotopy categories) and preserves small colimits by [HA, Corollary 4.2.3.7 and Remark 7.1.1.16]. It suffices to show that the image of the complex tot(F)\operatorname{tot}(F) in 𝒟(Z)\mathcal{D}(\mathbb{Z}) is equivalent to the colimit of the diagram ιZF:ΔopιAF𝒟(A)𝒟(Z)\iota{Z}F\colon\Delta^{\operatorname{op}}\xrightarrow{\iota_{A}F}\mathcal{D}(A)\to\mathcal{D}(\mathbb{Z}). This diagram is the same as the composition ιZF:Δop𝐹Ch(A)Ch(Z)ιZ𝒟(Z)\iota{Z}F\colon\Delta^{\operatorname{op}}\xrightarrow{F}\operatorname{Ch}(A)\to\operatorname{Ch}(\mathbb{Z})\xrightarrow{\iota_{\mathbb{Z}}}\mathcal{D}(\mathbb{Z}) and it suffices to calculate this colimit. By [Bun13, Problem 4.24] (and its proof), we have the isomorphism

colimΔopιZFιZ(tot(F))𝒟(Z)\operatorname*{colim}_{\Delta^{\operatorname{op}}}\iota{Z}F\cong\iota{Z}(\operatorname{tot}(F))\in\mathcal{D}(\mathbb{Z})

and we are done. ∎

The following typical case is used in Lemma 3.6.

Corollary 3.10.

Let AA be a discrete ring and let C=(CiC00)Ch0(A)C=(\cdots\to C_{i}\to\cdots\to C_{0}\to 0)\in\operatorname{Ch}_{\geq 0}(A) be a non-negatively complex of discrete AA-modules. Set a double complex C(Ci[0]C0[0]0)C^{\prime}\coloneqq(\cdots\to C_{i}[0]\to\cdots\to C_{0}[0]\to 0) in Ch0(Ch(A))\operatorname{Ch}_{\geq 0}(\operatorname{Ch}(A)). Then the geometric realization of the simplicial object ιADK(C)\iota_{A}\operatorname{DK}_{\bullet}(C^{\prime}) in 𝒟(A)\mathcal{D}(A) is isomorphic to ιA(C)\iota_{A}(C) in 𝒟(A)\mathcal{D}(A). Namely, we have an isomorphism

colimΔopιADK(Ci[0]C0[0]0)ιA(C1C00)\operatorname*{colim}_{\Delta^{\operatorname{op}}}\iota_{A}\operatorname{DK}_{\bullet}(\cdots\to C_{i}[0]\to\cdots\to C_{0}[0]\to 0)\cong\iota_{A}(\cdots\to C_{1}\to C_{0}\to 0)

in 𝒟(A)\mathcal{D}(A)

Proof.

By Lemma 3.9, we have the isomorphism

colimΔopιADK(C)ιA(tot(DK(C)))\operatorname*{colim}_{\Delta^{\operatorname{op}}}\iota_{A}\operatorname{DK}_{\bullet}(C^{\prime})\cong\iota_{A}(\operatorname{tot}(\operatorname{DK}_{\bullet}(C^{\prime})))

in 𝒟(A)\mathcal{D}(A). We show that the totalization tot(DK(C))\operatorname{tot}(\operatorname{DK}_{\bullet}(C^{\prime})) is quasi-isomorphic to the complex C=(C1C00)C=(\cdots\to C_{1}\to C_{0}\to 0) in Ch(A)\operatorname{Ch}(A). Since Ci=Ci[0]C^{\prime}_{i}=C_{i}[0] is concentrated in degree 0, DK(C)([q])p=(DKq(C))p=(α:[n][k]Ck[0])p=0\operatorname{DK}_{\bullet}(C^{\prime})([q])_{p}=(\operatorname{DK}_{q}(C^{\prime}))_{p}=(\oplus_{\alpha\colon[n]\twoheadrightarrow[k]}C_{k}[0])_{p}=0 holds for all p0p\neq 0. By the construction of tot()\operatorname{tot}(-) (3.8), we have

tot(DK(C))n=DK(C)([n])0=(α:[n][k]Ck[0])0=α:[n][k]Ck=DKn(C)\operatorname{tot}(\operatorname{DK}_{\bullet}(C^{\prime}))_{n}=\operatorname{DK}_{\bullet}(C^{\prime})([n])_{0}=(\oplus_{\alpha\colon[n]\twoheadrightarrow[k]}C_{k}[0])_{0}=\oplus_{\alpha\colon[n]\twoheadrightarrow[k]}C_{k}=\operatorname{DK}_{n}(C)

in Ch(A)\operatorname{Ch}(A)^{\heartsuit} and the differential is i=0q(1)iiDKn(C)0=i=0q(1)iiDKn(C)\sum_{i=0}^{q}(-1)^{i}\partial_{i}^{\operatorname{DK}_{n}(C^{\prime})_{0}}=\sum_{i=0}^{q}(-1)^{i}\partial_{i}^{\operatorname{DK}_{n}(C)}, where DKn(C)=DK(C)([n])\operatorname{DK}_{n}(C)=\operatorname{DK}_{\bullet}(C)([n]) is the nn-th term of the simplicial discrete AA-module DK(C)\operatorname{DK}_{\bullet}(C) corresponding to CCh0(A)C\in\operatorname{Ch}_{\geq 0}(A). So the totalization tot(DK(C))Ch(A)\operatorname{tot}(\operatorname{DK}_{\bullet}(C^{\prime}))\in\operatorname{Ch}(A) is the Moore complex M(DK(C))M_{*}(\operatorname{DK}_{\bullet}(C)) of the simplicial discrete AA-module DK(C)\operatorname{DK}_{\bullet}(C). The Moore complex M(DK(C))M_{*}(\operatorname{DK}_{\bullet}(C)) (in other words, unnormalized chain complex) and the normalized chain complex N(DK(C))N_{*}(\operatorname{DK}_{\bullet}(C)) are quasi-isomorphic by [HA, Proposition 1.2.3.17]. This shows that the totalization tot(DK(C))\operatorname{tot}(\operatorname{DK}_{\bullet}(C^{\prime})) is quasi-isomorphic to the complex CCh(A)C\in\operatorname{Ch}(A) because of CN(DK(C))C\cong N_{*}(\operatorname{DK}_{\bullet}(C)) in Ch(A)\operatorname{Ch}(A) ([HA, Lemma 1.2.3.11]), that is,

colimΔopιADK(C)ιA(tot(DK(C)))ιA(tot(C))ιA(C)=ιA(C1C00)\operatorname*{colim}_{\Delta^{\operatorname{op}}}\iota_{A}\operatorname{DK}_{\bullet}(C^{\prime})\cong\iota_{A}(\operatorname{tot}(\operatorname{DK}_{\bullet}(C^{\prime})))\cong\iota_{A}(\operatorname{tot}(C))\cong\iota_{A}(C)=\iota_{A}(\cdots\to C_{1}\to C_{0}\to 0)

in 𝒟(A)\mathcal{D}(A) where the first isomorphism follows from Lemma 3.9. ∎

In the last part of this section, we record the relation between ExtBi(M,N)\operatorname{Ext}^{i}_{B}(M,N) and Extπ0(B)i(M,N)\operatorname{Ext}^{i}_{\pi_{0}(B)}(M,N) for a certain animated ring BB and its modules MM and NN.

Lemma 3.11.

Let AA be an animated ring and let LL be a connective AA-module. Let BB an animated AA-algebra. Assume that the connective BB-module MLALBM\coloneqq L\otimes^{L}_{A}B is discrete. Then Extπ0(B)i(M,N)\operatorname{Ext}^{i}_{\pi_{0}(B)}(M,N) is a direct summand of ExtBi(M,N)\operatorname{Ext}^{i}_{B}(M,N) for any π0(B)\pi_{0}(B)-module NN and for each i0i\geq 0.

Proof.

The canonical map of AA-modules Bπ0(B)B\to\pi_{0}(B) induces a commutative diagram

M=LALB{M=L\otimes^{L}_{A}B}LALπ0(B){L\otimes^{L}_{A}\pi_{0}(B)}π0(LALB){\pi_{0}(L\otimes^{L}_{A}B)}π0(LALπ0(B)){\pi_{0}(L\otimes^{L}_{A}\pi_{0}(B))}\scriptstyle{\cong}

in Mod(π0(B))\operatorname{Mod}(\pi_{0}(B)). By [HA, Corollary 7.2.1.23], the lower horizontal map is an isomorphism. The inverse map gives a splitting of the upper horizontal map and thus M=LALBM=L\otimes^{L}_{A}B is a direct summand of LALπ0(B)L\otimes^{L}_{A}\pi_{0}(B) in Mod(π0(B))\operatorname{Mod}(\pi_{0}(B)). This shows that the Ext\operatorname{Ext}-group Extπ0(B)i(M,N)=Extπ0(B)i(LALB,N)\operatorname{Ext}^{i}_{\pi_{0}(B)}(M,N)=\operatorname{Ext}^{i}_{\pi_{0}(B)}(L\otimes^{L}_{A}B,N) is a direct summand of

Extπ0(B)i(LALπ0(B),N)Extπ0(B)i(MBLπ0(B),N)ExtBi(M,N)\operatorname{Ext}^{i}_{\pi_{0}(B)}(L\otimes^{L}_{A}\pi_{0}(B),N)\cong\operatorname{Ext}^{i}_{\pi_{0}(B)}(M\otimes^{L}_{B}\pi_{0}(B),N)\cong\operatorname{Ext}^{i}_{B}(M,N)

for each i0i\geq 0 and for each π0(B)\pi_{0}(B)-module NN. ∎

In Lemma 3.11, the vanishing of ExtBi(M,N)\operatorname{Ext}^{i}_{B}(M,N) implies the vanishing of its direct factor Extπ0(B)i(M,N)\operatorname{Ext}^{i}_{\pi_{0}(B)}(M,N). However, the following example shows that the converse does not hold in general. Remark that in the following example, the ring AA is not regular and the AA-module LL has infinite projective dimension. If LL has finite projective dimension (or simply AA is regular local), then the converse holds when BB is a derived quotient of AA. See Remark 7.3.

Example 3.12.

Set AZ[x]/(px)A\coloneqq\mathbb{Z}[x]/(px) for some fixed prime number pp. Let LL be an AA-module Z\mathbb{Z}. Take an animated AA-algebra BA/Lp=(Z[x]/(px))/LpB\coloneqq A/^{L}p=(\mathbb{Z}[x]/(px))/^{L}p and its module MLALBFpM\coloneqq L\otimes^{L}_{A}B\cong\mathbb{F}_{p}. Fix a discrete non-zero π0(B)\pi_{0}(B)-module NN. Then MM is discrete and Extπ0(B)i(M,N)\operatorname{Ext}^{i}_{\pi_{0}(B)}(M,N) vanishes for each i2i\geq 2 but not necessarily ExtBi(M,N)\operatorname{Ext}^{i}_{B}(M,N) for i2i\geq 2. Here is the proof.

As a complex of AA-modules, BB is isomorphic to Kos(p;Z[x]/(px))\operatorname{Kos}(p;\mathbb{Z}[x]/(px)) and thus

MZZ[x]/(px)LKos(p;Z[x]/(px))(0Z×pZ0)Fp[0]M\cong\mathbb{Z}\otimes^{L}_{\mathbb{Z}[x]/(px)}\operatorname{Kos}(p;\mathbb{Z}[x]/(px))\cong(0\to\mathbb{Z}\xrightarrow{\times p}\mathbb{Z}\to 0)\cong\mathbb{F}_{p}[0]

is discrete. The connected component π0(B)\pi_{0}(B) is isomorphic to Z[x]/(px,p)Fp[x]\mathbb{Z}[x]/(px,p)\cong\mathbb{F}_{p}[x]. The Fp[x]\mathbb{F}_{p}[x]-module M=FpM=\mathbb{F}_{p} has a free resolution 0Fp[x]×xFp[x]00\to\mathbb{F}_{p}[x]\xrightarrow{\times x}\mathbb{F}_{p}[x]\to 0, which shows that Extπ0(B)i(M,N)ExtFp[x]i(Fp,N)=0\operatorname{Ext}^{i}_{\pi_{0}(B)}(M,N)\cong\operatorname{Ext}^{i}_{\mathbb{F}_{p}[x]}(\mathbb{F}_{p},N)=0 for each i2i\geq 2. Next, we calculate the Ext\operatorname{Ext}-group

ExtBi(M,N)\displaystyle\operatorname{Ext}^{i}_{B}(M,N) Extπ0(B)i((LALB)BLπ0(B),N)Extπ0(B)i(LALπ0(B),N)\displaystyle\cong\operatorname{Ext}^{i}_{\pi_{0}(B)}((L\otimes^{L}_{A}B)\otimes^{L}_{B}\pi_{0}(B),N)\cong\operatorname{Ext}^{i}_{\pi_{0}(B)}(L\otimes^{L}_{A}\pi_{0}(B),N)
ExtFp[x]i(ZZ[x]/(px)LFp[x],N).\displaystyle\cong\operatorname{Ext}^{i}_{\mathbb{F}_{p}[x]}(\mathbb{Z}\otimes^{L}_{\mathbb{Z}[x]/(px)}\mathbb{F}_{p}[x],N).

The Z[x]/(px)\mathbb{Z}[x]/(px)-module Z\mathbb{Z} has a (periodic) free resolution ×xZ[x]/(px)×pZ[x]/(px)×xZ[x]/(px)0\cdots\xrightarrow{\times x}\mathbb{Z}[x]/(px)\xrightarrow{\times p}\mathbb{Z}[x]/(px)\xrightarrow{\times x}\mathbb{Z}[x]/(px)\to 0. This shows that ZZ[x]/(px)LFp[x]\mathbb{Z}\otimes^{L}_{\mathbb{Z}[x]/(px)}\mathbb{F}_{p}[x] can be represented by the complex ×xFp[x]×pFp[x]×xFp[x]0\cdots\xrightarrow{\times x}\mathbb{F}_{p}[x]\xrightarrow{\times p}\mathbb{F}_{p}[x]\xrightarrow{\times x}\mathbb{F}_{p}[x]\to 0, which is quasi-isomorphic to the complex of Fp[x]\mathbb{F}_{p}[x]-modules

(Fp0Fp0)k0Fp[2k](\cdots\to\mathbb{F}_{p}\to 0\to\mathbb{F}_{p}\to 0)\cong\bigoplus_{k\geq 0}\mathbb{F}_{p}[2k]

where Fp[2k]\mathbb{F}_{p}[2k] is a complex concentrated in degree 2k2k. Taking a free resolution (0Fp[x]×xFp[x]0)Ch(Fp[x])(0\to\mathbb{F}_{p}[x]\xrightarrow{\times x}\mathbb{F}_{p}[x]\to 0)\in\operatorname{Ch}(\mathbb{F}_{p}[x]) of Fp[2k]\mathbb{F}_{p}[2k] which is concentrated in degree 2k+12k+1 and 2k2k, we have

ExtFp[x]i(Fp[2k],N)\displaystyle\operatorname{Ext}^{i}_{\mathbb{F}_{p}[x]}(\mathbb{F}_{p}[2k],N) =Hi(0HomFp[x](Fp[x],N)×xHomFp[x](Fp[x],N)0)\displaystyle=H^{i}(0\to\operatorname{Hom}_{\mathbb{F}_{p}[x]}(\mathbb{F}_{p}[x],N)\xrightarrow{\times x}\operatorname{Hom}_{\mathbb{F}_{p}[x]}(\mathbb{F}_{p}[x],N)\to 0)
{N/xNifi=2kN[x]ifi=2k+10otherwise\displaystyle\cong\begin{cases}N/xN&\text{if}\ i=2k\\ N[x]&\text{if}\ i=2k+1\\ 0&\text{otherwise}\end{cases}

through the Fp[x]\mathbb{F}_{p}[x]-module structure on NN where N[x]N[x] is the submodule of xx-torsion elements of NN. The Ext\operatorname{Ext}-group ExtFp[x]i(k0Fp[2k],N)\operatorname{Ext}^{i}_{\mathbb{F}_{p}[x]}(\oplus_{k\geq 0}\mathbb{F}_{p}[2k],N) is isomorphic to

ExtFp[x]i(k0Fp[2k],N)k0ExtFp[x]i(Fp[2k],N)N/xNorN[x]\operatorname{Ext}^{i}_{\mathbb{F}_{p}[x]}(\oplus_{k\geq 0}\mathbb{F}_{p}[2k],N)\cong\bigoplus_{k\geq 0}\operatorname{Ext}^{i}_{\mathbb{F}_{p}[x]}(\mathbb{F}_{p}[2k],N)\cong N/xN\ \text{or}\ N[x]

for each i0i\geq 0. If Fp[x]\mathbb{F}_{p}[x] acts on NN by xN=0xN=0, this is a desired example (for example N=FpN=\mathbb{F}_{p}).

4. Liftings of Modules

In this section, we generalize the properties of liftings of modules over usual rings introduced in [ADS93] to ones of modules over animated rings.

Definition 4.1.

Let ABA\to B be a map of animated rings and let MM be a BB-module. A pair (L,φ)(L,\varphi) of an AA-module LL such that π0(L)\pi_{0}(L) is finite over π0(A)\pi_{0}(A) and an isomorphism φ:MLALB\varphi\colon M\xrightarrow{\cong}L\otimes^{L}_{A}B of BB-modules is called a lifting of the BB-module MM to AA. We often denote a lifting (L,φ)(L,\varphi) by LL if the isomorphism φ\varphi is clear from the context. A BB-module MM is called liftable to AA if it has a lifting to AA.

Moreover, we can show the following equivalence of liftings of discrete modules. Note that any discrete module MM over an animated ring AA is canonically a discrete module over π0(A)\pi_{0}(A).

Lemma 4.2.

Let ABA\to B be a map of animated rings and let MM be a discrete BB-module. For a given connective AA-module LL, the following are equivalent.

  1. (a)

    There exists an isomorphism π0(LALB)M\pi_{0}(L\otimes^{L}_{A}B)\cong M of discrete BB-modules such that the truncation map LALBπ0(LALB)ML\otimes^{L}_{A}B\to\pi_{0}(L\otimes^{L}_{A}B)\cong M is an isomorphism of BB-modules.

  2. (b)

    There exists an isomorphism of BB-modules LALBML\otimes^{L}_{A}B\cong M, namely, LL (and this isomorphism) is a lifting of the BB-module MM to AA.

  3. (c)

    There exists an isomorphism π(LALB)π(M)\pi_{*}(L\otimes^{L}_{A}B)\cong\pi_{*}(M) of graded π(B)\pi_{*}(B)-modules.

Proof.

(a) \Rightarrow (b) \Rightarrow (c) is clear. (c) \Rightarrow (a) follows from the assumption that πi(LALB)\pi_{i}(L\otimes^{L}_{A}B) vanishes for i0i\neq 0 and is isomorphic to MM for i=0i=0. ∎

We prove the ascending property of the finiteness and discreteness of modules.

Lemma 4.3.

Let AA be an almost perfect animated Λ\Lambda-algebra and let xx be an element of 𝔪\mathfrak{m}. Fix k1k\geq 1. Let MM be a discrete finitely generated AkA_{k}-module and let LL be an AA-module such that π0(L)\pi_{0}(L) is finite over π0(A)\pi_{0}(A). If LL is a lifting of the AkA_{k}-module MM to AA, then LL is a discrete finitely generated AA-module.

Proof.

We have a fiber sequence L×xkLLALAkML\xrightarrow{\times x^{k}}L\to L\otimes^{L}_{A}A_{k}\cong M and πi(L)×xkπi(L)\pi_{i}(L)\xrightarrow{\times x^{k}}\pi_{i}(L) is surjective for all i0i\neq 0 since MM is discrete. By Lemma 2.8 and the finiteness of MM, MLALAkM\cong L\otimes^{L}_{A}A_{k} is almost perfect over AkA_{k} and thus LL is almost perfect over AA (Lemma 2.13 (b)). Here, we use the assumption that AA is 𝔪\mathfrak{m}-adically complete Noetherian (Lemma 2.14) and π0(L)\pi_{0}(L) is finite over π0(A)\pi_{0}(A). Since πi(L)\pi_{i}(L) is finite over the Noetherian ring π0(A)\pi_{0}(A), πi(L)\pi_{i}(L) vanishes for all i1i\geq 1 by Nakayama’s lemma and thus LL is discrete. ∎

Lemma 4.4.

Let AA be an almost perfect animated Λ\Lambda-algebra and let xx be an element of 𝔪\mathfrak{m}. Fix n,k1n,k\geq 1. Take the canonical map An+kAnA_{n+k}\to A_{n} of animated Λ\Lambda-algebras. Let Ln+kL_{n+k} be an An+kA_{n+k}-module and let LnL_{n} be a discrete finitely generated AnA_{n}-module. If Ln+kL_{n+k} is a lifting of the AnA_{n}-module LnL_{n} to An+kA_{n+k}, then Ln+kL_{n+k} is a discrete finitely generated An+kA_{n+k}-module.

Proof.

The base change Ln+kAn+kLAnLnL_{n+k}\otimes^{L}_{A_{n+k}}A_{n}\cong L_{n} is discrete and almost perfect over AnA_{n} (here we use the assumption that AnA_{n} is Noetherian and Ln=π0(Ln)L_{n}=\pi_{0}(L_{n}) is finite over π0(An)\pi_{0}(A_{n})). Since An+kAnA_{n+k}\to A_{n} has the nilpotent kernel on π0\pi_{0}, the An+kA_{n+k}-module Ln+kL_{n+k} is also connective and almost perfect over An+kA_{n+k} by Lemma 2.12. Considering the fiber sequence Ak×xnAn+kAnA_{k}\xrightarrow{\times x^{n}}A_{n+k}\to A_{n} given by Lemma 3.4 and taking Ln+kAn+kLL_{n+k}\otimes^{L}_{A_{n+k}}-, we have the following fiber sequence in Mod(An+k)\operatorname{Mod}(A_{n+k}):

Ln+kAn+kLAk×xnLn+kLn+kAn+kLAn.L_{n+k}\otimes^{L}_{A_{n+k}}A_{k}\xrightarrow{\times x^{n}}L_{n+k}\to L_{n+k}\otimes^{L}_{A_{n+k}}A_{n}.

Since the last term Ln+kAn+kLAnLnL_{n+k}\otimes^{L}_{A_{n+k}}A_{n}\cong L_{n} is discrete, the map of π0(An+k)\pi_{0}(A_{n+k})-modules πi(Ln+kAn+kLAk)×xnπi(Ln+k)\pi_{i}(L_{n+k}\otimes^{L}_{A_{n+k}}A_{k})\xrightarrow{\times x^{n}}\pi_{i}(L_{n+k}) is surjective for all i0i\neq 0. Then we have xnπi(Ln+k)=πi(Ln+k)x^{n}\pi_{i}(L_{n+k})=\pi_{i}(L_{n+k}) and thus πi(Ln+k)\pi_{i}(L_{n+k}) vanishes for all i0i\neq 0 by Nakayama’s lemma. ∎

The following theorem is a key point of Auslander, Ding, and Solberg’s proof in [ADS93] which reduces the lifting problem of general quotients to the one of nilpotent quotients.

Theorem 4.5 (cf. [ADS93, Theorem 1.2]).

Let AA be an almost perfect animated Λ\Lambda-algebra and let xx be an element of 𝔪\mathfrak{m}. Take a discrete finitely generated A1A_{1}-module MM. Then the following are equivalent:

  1. (1)

    MM is liftable to AA.

  2. (2)

    There exists a sequence {Ln}n1={L1M,L2,}\{L_{n}\}_{n\geq 1}=\{L_{1}\coloneqq M,L_{2},\dots\} such that each LnL_{n} is an almost perfect animated AnA_{n}-module and Ln+1L_{n+1} is a lifting of LnL_{n} to An+1A_{n+1} for all n1n\geq 1.

  3. (3)

    There exists a sequence {Ln}n1={L1M,L2,}\{L_{n}\}_{n\geq 1}=\{L_{1}\coloneqq M,L_{2},\dots\} such that each LnL_{n} is an AnA_{n}-module and Ln+1L_{n+1} is a lifting of LnL_{n} to An+1A_{n+1} for all n1n\geq 1.

In this case, any such AnA_{n}-module LnL_{n} and any lifting LL of MM to AA are discrete and finitely generated.

Proof.

Since MM is discrete, each assumption is followed by the last assertion (Lemma 4.3 and Lemma 4.4).

(1)(2)\text{(1)}\Longrightarrow\text{(2)}: Since MM has a lifting LL to AA, we can set an AnA_{n}-module LnL/LxnLALAnL_{n}\coloneqq L/^{L}x^{n}\cong L\otimes^{L}_{A}A_{n} for each n1n\geq 1. Since MM is almost perfect over A1A_{1} and LnAnLA1ML_{n}\otimes^{L}_{A_{n}}A_{1}\cong M, LnL_{n} is almost perfect over AnA_{n} by Lemma 2.12. The isomorphisms of AnA_{n}-modules

Ln+1An+1LAn(LALAn+1)An+1LAnLALAnLn.L_{n+1}\otimes^{L}_{A_{n+1}}A_{n}\cong(L\otimes^{L}_{A}A_{n+1})\otimes^{L}_{A_{n+1}}A_{n}\cong L\otimes^{L}_{A}A_{n}\cong L_{n}.

show that the An+1A_{n+1}-module Ln+1L_{n+1} is a lifting of LnL_{n} to An+1A_{n+1}.

(2)(3)\text{(2)}\Longrightarrow\text{(3)}: This is clear.

(3)(1)\text{(3)}\Longrightarrow\text{(1)}: By using Lemma 4.4 inductively, each LnL_{n} is an almost perfect animated AnA_{n}-module. Applying Ln+1An+1LL_{n+1}\otimes^{L}_{A_{n+1}}- for the fiber sequence An×xAn+1A1A_{n}\xrightarrow{\times x}A_{n+1}\to A_{1} in Lemma 3.4, we have a fiber sequence in Mod(An+1)\operatorname{Mod}(A_{n+1})

(4.1) Ln+1An+1LAnidLn×xLn+1An+1LAn+1Ln+1An+1LA1.L_{n+1}\otimes^{L}_{A_{n+1}}A_{n}\xrightarrow{\operatorname{id}_{L_{n}}\otimes\times x}L_{n+1}\otimes^{L}_{A_{n+1}}A_{n+1}\to L_{n+1}\otimes^{L}_{A_{n+1}}A_{1}.

Furthermore, the assumption is followed by the following isomorphisms in Mod(An+1)\operatorname{Mod}(A_{n+1}):

Ln+1An+1LAn\displaystyle L_{n+1}\otimes^{L}_{A_{n+1}}A_{n} Ln,\displaystyle\cong L_{n},
Ln+1An+1LAn+1\displaystyle L_{n+1}\otimes^{L}_{A_{n+1}}A_{n+1} Ln+1,\displaystyle\cong L_{n+1},
Ln+1An+1LA1\displaystyle L_{n+1}\otimes^{L}_{A_{n+1}}A_{1} Ln+1An+1LAnAnLA1LnAnLA1L2A2LA1M.\displaystyle\cong L_{n+1}\otimes^{L}_{A_{n+1}}A_{n}\otimes^{L}_{A_{n}}A_{1}\cong L_{n}\otimes^{L}_{A_{n}}A_{1}\cong\dots\cong L_{2}\otimes^{L}_{A_{2}}A_{1}\cong M.

Then the above fiber sequence (4.1) induces a fiber sequence Ln×xLn+1ML_{n}\xrightarrow{\times x}L_{n+1}\to M in Mod(An+1)\operatorname{Mod}(A_{n+1}) and thus we have an exact sequence 0Ln×xLn+1M00\to L_{n}\xrightarrow{\times x}L_{n+1}\to M\to 0 of discrete π0(An+1)\pi_{0}(A_{n+1})-modues. Since the canonical surjective map Ln+1Ln+1An+1LAnLnL_{n+1}\twoheadrightarrow L_{n+1}\otimes^{L}_{A_{n+1}}A_{n}\cong L_{n} gives an inverse system {Ln}n1\{L_{n}\}_{n\geq 1} of discrete π0(A)\pi_{0}(A)-modules, we have an exact sequence of inverse systems of discrete π0(A)\pi_{0}(A)-modules

0{Ln}n1×x{Ln+1}n1{M}n10.0\to\{L_{n}\}_{n\geq 1}\xrightarrow{\times x}\{L_{n+1}\}_{n\geq 1}\to\{M\}_{n\geq 1}\to 0.

Taking the limit, Llimn1LnL\coloneqq\lim_{n\geq 1}L_{n} is an π0(A)limn1π0(An)\pi_{0}(A)\cong\lim_{n\geq 1}\pi_{0}(A_{n})-module since π0(A)\pi_{0}(A) is 𝔪\mathfrak{m}-adically complete (Lemma 2.14). So we have an exact sequence

0L×xLM00\to L\xrightarrow{\times x}L\to M\to 0

of discrete π0(A)\pi_{0}(A)-modules. This induces an isomorphism LALA1L/LxML\otimes^{L}_{A}A_{1}\cong L/^{L}x\cong M as A1A_{1}-modules because of L/xLML/xL\cong M and L[x]=0L[x]=0. To prove LL is a lifting of MM to AA, we must show that π0(L)=L\pi_{0}(L)=L is finite over π0(A)\pi_{0}(A). As explained at the begining of this proof, each LnL_{n} is almost perfect and thus a discrete finitely generated π0(An)\pi_{0}(A_{n})-module by Lemma 2.8. Applying the above argument replacing MM with LnL_{n} and xx with xnx^{n}, we have an exact sequence 0L×xnLLn00\to L\xrightarrow{\times x^{n}}L\to L_{n}\to 0 and thus LnL/xnLL_{n}\cong L/x^{n}L. As in the same proof of [ADS93, Theorem 1.2], LL is a discrete finitely generated π0(A)\pi_{0}(A)-module. ∎

Lemma 4.6 (cf. [ADS93, Lemma 1.4]).

Let AA be an almost perfect animated Λ\Lambda-algebra and let xx be an element of 𝔪\mathfrak{m}. Take a discrete finitely generated A1A_{1}-module MM and its lifting LL to AnA_{n}. For an An+1A_{n+1}-module EE, the following are equivalent:

  1. (1)

    EE is a lifting of the AnA_{n}-module LL to An+1A_{n+1}.

  2. (2)

    There exists a fiber sequence in Mod(An+1)\operatorname{Mod}(A_{n+1}) of the form MELM\to E\to L and there exists an isomorphism π0(EAn+1LA1)M\pi_{0}(E\otimes^{L}_{A_{n+1}}A_{1})\cong M of π0(A1)\pi_{0}(A_{1})-modules.

In this case, the An+1A_{n+1}-module EE is a discrete finitely generated π0(An+1)\pi_{0}(A_{n+1})-module by Lemma 4.4.

Proof.

By Lemma 4.4, LL is a discrete finitely generated π0(An)\pi_{0}(A_{n})-module.

(1)(2)\text{(1)}\Longrightarrow\text{(2)}: First note that we have a fiber sequence in Mod(An+1)\operatorname{Mod}(A_{n+1}):

A1×xnAn+1AnA_{1}\xrightarrow{\times x^{n}}A_{n+1}\to A_{n}

by Lemma 3.4. Taking base change EAn+1LE\otimes^{L}_{A_{n+1}}-, we have a fiber sequence in Mod(An+1)\operatorname{Mod}(A_{n+1})

EAn+1LA1idL×xnEAn+1LAn+1EAn+1LAn.E\otimes^{L}_{A_{n+1}}A_{1}\xrightarrow{\operatorname{id}_{L}\otimes\times x^{n}}E\otimes^{L}_{A_{n+1}}A_{n+1}\to E\otimes^{L}_{A_{n+1}}A_{n}.

Since EE is a lifting of LL to An+1A_{n+1} and LL is a lifting of MM to AnA_{n}, the following isomorphisms hold in Mod(An+1)\operatorname{Mod}(A_{n+1}):

EAn+1LA1\displaystyle E\otimes^{L}_{A_{n+1}}A_{1} (EAn+1LAn)AnLA1LAnLA1M\displaystyle\cong(E\otimes^{L}_{A_{n+1}}A_{n})\otimes^{L}_{A_{n}}A_{1}\cong L\otimes^{L}_{A_{n}}A_{1}\cong M
EAn+1LAn+1\displaystyle E\otimes^{L}_{A_{n+1}}A_{n+1} E\displaystyle\cong E
EAn+1LAn\displaystyle E\otimes^{L}_{A_{n+1}}A_{n} L.\displaystyle\cong L.

By these isomorphisms, the desired isomorphism π0(EAn+1LA1)M\pi_{0}(E\otimes^{L}_{A_{n+1}}A_{1})\cong M and the fiber sequence in Mod(An+1)\operatorname{Mod}(A_{n+1}) exist.

(2)(1)\text{(2)}\Longrightarrow\text{(1)}: Assume that there exists a fiber sequence in Mod(An+1)\operatorname{Mod}(A_{n+1}) of the form MELM\to E\to L. Since MM and LL are discrete, so is EE and thus there exists an exact sequence 0MEL00\to M\to E\to L\to 0 of finite discrete π0(An+1)\pi_{0}(A_{n+1})-modules. By Lemma 4.2, it suffices to show the isomorphism π(EAn+1LAn)π(L)\pi_{*}(E\otimes^{L}_{A_{n+1}}A_{n})\cong\pi_{*}(L) holds. The assumption that LL is a lifting of MM to AnA_{n} shows that

πk(LAnLA1){L/xLMif k=0L[x]/xn1L0if k is oddL[xn1]/xL0if k is even\pi_{k}(L\otimes^{L}_{A_{n}}A_{1})\cong\begin{cases}L/xL\cong M&\text{if }k=0\\ L[x]/x^{n-1}L\cong 0&\text{if $k$ is odd}\\ L[x^{n-1}]/xL\cong 0&\text{if $k$ is even}\end{cases}

by Lemma 3.6. Also, we have

πk(EAn+1LAn){E/xnLif k=0E[xn]/xEif k is oddE[x]/xnEif k is even.\pi_{k}(E\otimes^{L}_{A_{n+1}}A_{n})\cong\begin{cases}E/x^{n}L&\text{if }k=0\\ E[x^{n}]/xE&\text{if $k$ is odd}\\ E[x]/x^{n}E&\text{if $k$ is even}.\end{cases}

We must show that E/xnELE/x^{n}E\cong L and πk(EAn+1LAn)=0\pi_{k}(E\otimes^{L}_{A_{n+1}}A_{n})=0 for k>0k>0 but this can be shown by the same proof of [ADS93, Lemma 1.4] since π0(A)\pi_{0}(A) is a Noetherian ring and MM is a discrete Noetherian π0(A)\pi_{0}(A)-module.

5. Fiber Sequence Corresponding to a Lifting

In this section, we construct a fiber sequence θL\theta^{\prime}_{L} of A1A_{1}-modules corresponding to a lifting. This construction is based on the one for the case of complete intersections in [ADS93]. The existence of such a fiber sequence is crucial in the proof of Theorem 6.4 and one of the advantages of using higher algebra is that it can handle the derived category Mod(An)\operatorname{Mod}(A_{n}) of a non-discrete ring AnA_{n} in this manner.

Construction 5.1.

Let AA be an almost perfect animated Λ\Lambda-algebra and let xx be an element of 𝔪\mathfrak{m}. Fix n1n\geq 1. Take a discrete finitely generated A1A_{1}-module MM and its lifting LL to AnA_{n} (in this case, LL is a discrete finitely generated AnA_{n}-module by Lemma 4.4). We define two fiber sequences θL\theta_{L} (5.2) and θL\theta_{L}^{\prime} (5.10) in Mod(A1)\operatorname{Mod}(A_{1}) as follows.

Choose a map of AA-modules p:PLp\colon P\to L, where PP is a finite free AA-module and π0(p):π0(P)L\pi_{0}(p)\colon\pi_{0}(P)\to L is surjective as in the proof of [HA, Proposition 7.2.2.7] (for example, PxSAP\coloneqq\oplus_{x\in S}A where SLS\subseteq L is a system of generators of LL over π0(An)\pi_{0}(A_{n})). Taking its fiber, we have a fiber sequence of AA-modules

(5.1) ΩA(L)fib(p)P𝑝L.\Omega_{A}(L)\coloneqq\operatorname{fib}(p)\to P\xrightarrow{p}L.

Applying the base change functor ALA1/Lx-\otimes^{L}_{A}A_{1}\cong-/^{L}x (see Lemma 3.3), we can take a fiber sequence θL\theta_{L} of A1A_{1}-modules

(5.2) θL:ΩA(L)/LxP/Lxp/LxL/Lx.\theta_{L}\colon\Omega_{A}(L)/^{L}x\to P/^{L}x\xrightarrow{p/^{L}x}L/^{L}x.

In particular, this θL\theta_{L} is a distinguished triangle in the homotopy category hMod(A1)\mathrm{h}\operatorname{Mod}(A_{1}).

Next, we construct a fiber sequence θL\theta_{L}^{\prime} in Mod(A1)\operatorname{Mod}(A_{1}) from θL\theta_{L}. In the homotopy category hMod(π0(An))\mathrm{h}\operatorname{Mod}(\pi_{0}(A_{n})) of π0(An)\pi_{0}(A_{n})-modules, which is equivalent to the derived category D(π0(An))D(\pi_{0}(A_{n})), we have a sequence of complexes:

0{0}L{L}L{L}0{0}0{0}L{L}L/xL{L/xL}0{0}0{0}L/xL{L/xL}L/xL.{L/xL.}0{0}×x\scriptstyle{\times x}×x\scriptstyle{\times x}×x\scriptstyle{\times x}×xn1\scriptstyle{\times x^{n-1}}

Since LL is a lifting of MM to AnA_{n}, the homotopy group πi(LAnLA1)\pi_{i}(L\otimes^{L}_{A_{n}}A_{1}) vanishes for i0i\neq 0 and LL is discrete. By Lemma 3.6, we have L[x]=xn1LL[x]=x^{n-1}L and L[xn1]=xLL[x^{n-1}]=xL, and in particular, the map of π0(A1)\pi_{0}(A_{1})-modules

L/xL×xn1xn1L=L[x]=π1(0L×xL0)L/xL\xrightarrow{\times x^{n-1}}x^{n-1}L=L[x]=\pi_{1}(0\to L\xrightarrow{\times x}L\to 0)

is isomorphism. So the above diagram gives isomorphisms on homotopy groups;

L[x]idL[x]×xn1L/xLonπ0andL/xLidL/xLidL/xLonπ1.L[x]\xrightarrow{\operatorname{id}}L[x]\xleftarrow{\times x^{n-1}}L/xL\leavevmode\nobreak\ \text{on}\leavevmode\nobreak\ \pi_{0}\leavevmode\nobreak\ \text{and}\leavevmode\nobreak\ L/xL\xrightarrow{\operatorname{id}}L/xL\xleftarrow{\operatorname{id}}L/xL\leavevmode\nobreak\ \text{on}\leavevmode\nobreak\ \pi_{1}.

This shows that the maps L/Lx(0L×xL/xL0)L/xL[1]L/xLL/^{L}x\to(0\to L\xrightarrow{\times x}L/xL\to 0)\leftarrow L/xL[1]\oplus L/xL are isomorphisms in hMod(π0(An))\mathrm{h}\operatorname{Mod}(\pi_{0}(A_{n})). This gives rise to isomorphisms

(5.3) L/LxL/xL[1]L/xLL/^{L}x\cong L/xL[1]\oplus L/xL

in Mod(An)\operatorname{Mod}(A_{n}). Passing to Mod(An)\operatorname{Mod}(A_{n}), we have an isomorphism L/LxL/xL[1]L/xLL/^{L}x\cong L/xL[1]\oplus L/xL in Mod(An)\operatorname{Mod}(A_{n}) (and thus in Mod(A1)\operatorname{Mod}(A_{1})).

The surjective map of A1A_{1}-modules L/Lxπ0(L/Lx)ML/^{L}x\twoheadrightarrow\pi_{0}(L/^{L}x)\cong M gives the composition map q:P/Lxp/LxL/LxMq\colon P/^{L}x\xrightarrow{p/^{L}x}L/^{L}x\twoheadrightarrow M where the map p/Lx:P/LxL/Lxp/^{L}x\colon P/^{L}x\to L/^{L}x is the map in (5.2). Taking its fiber in Mod(A1)\operatorname{Mod}(A_{1})

(5.4) ΩA1(M)fib(q)P/Lx𝑞M.\Omega_{A_{1}}(M)\coloneqq\operatorname{fib}(q)\to P/^{L}x\xrightarrow{q}M.

Considering the universal property of ΩA1(M)=fib(q)\Omega_{A_{1}}(M)=\operatorname{fib}(q), the following commutative diagram whose horizontal sequences are fiber sequences in Mod(A1)\operatorname{Mod}(A_{1}) exists:

(5.5) ΩA(L)/Lx{\Omega_{A}(L)/^{L}x}P/Lx{P/^{L}x}L/Lx{L/^{L}x}ΩA1(M){\Omega_{A_{1}}(M)}P/Lx{P/^{L}x}M.{{M.}}!r\scriptstyle{\exists!r}p/Lx\scriptstyle{p/^{L}x}q\scriptstyle{q}

Taking the long exact sequences of homotopy groups, we have the following commutative diagram in the category of discrete π0(A1)\pi_{0}(A_{1})-modules for each iZi\in\mathbb{Z}:

(5.6) πi+1(P/Lx){\pi_{i+1}(P/^{L}x)}πi+1(L/Lx){\pi_{i+1}(L/^{L}x)}πi(ΩA(L)/Lx){\pi_{i}(\Omega_{A}(L)/^{L}x)}πi(P/Lx){\pi_{i}(P/^{L}x)}πi(L/Lx){\pi_{i}(L/^{L}x)}πi+1(P/Lx){\pi_{i+1}(P/^{L}x)}πi+1(M){\pi_{i+1}(M)}πi(ΩA1(M)){\pi_{i}(\Omega_{A_{1}}(M))}πi(P/Lx){\pi_{i}(P/^{L}x)}πi(M){\pi_{i}(M)}πi+1(p/Lx)\scriptstyle{\pi_{i+1}(p/^{L}x)}πi(r)\scriptstyle{\pi_{i}(r)}πi(p/Lx)\scriptstyle{\pi_{i}(p/^{L}x)}πi+1(q)\scriptstyle{\pi_{i+1}(q)}πi(q)\scriptstyle{\pi_{i}(q)}

In particular, we can show the following results:

  1. (a)

    If i2i\geq 2, then πi(L/Lx)\pi_{i}(L/^{L}x) and πi(M)\pi_{i}(M) vanish and thus πi(r):πi(ΩA(L)/Lx)πi(ΩA1(M))\pi_{i}(r)\colon\pi_{i}(\Omega_{A}(L)/^{L}x)\to\pi_{i}(\Omega_{A_{1}}(M)) is an isomorphism.

  2. (b)

    The vanishing π2(L/Lx)=0\pi_{2}(L/^{L}x)=0 implies that the map π1(r)\pi_{1}(r) is injective.

  3. (c)

    The vanishing π1(M)=0\pi_{1}(M)=0 and the isomorphy π0(L/Lx)π0(M)\pi_{0}(L/^{L}x)\cong\pi_{0}(M) imply that the sequence π1(L/Lx)π0(ΩA(L)/Lx)π0(r)π0(ΩA1(M))0\pi_{1}(L/^{L}x)\to\pi_{0}(\Omega_{A}(L)/^{L}x)\xrightarrow{\pi_{0}(r)}\pi_{0}(\Omega_{A_{1}}(M))\to 0 is exact.

  4. (d)

    If i<0i<0, the connectivity implies that πi(r)\pi_{i}(r) is an isomorphism.

By the fiber sequence L/Lx[1]ΩA(L)/LxP/LxL/^{L}x[-1]\to\Omega_{A}(L)/^{L}x\to P/^{L}x given by θL\theta_{L} (5.2), composing with rr gives maps of A1A_{1}-modules

(5.7) Mτ0(L/Lx[1])L/Lx[1]ΩA(L)/Lx𝑟ΩA1(M)M\cong\tau_{\geq 0}(L/^{L}x[-1])\to L/^{L}x[-1]\to\Omega_{A}(L)/^{L}x\xrightarrow{r}\Omega_{A_{1}}(M)

where the first isomorphism follows from (5.3). This composition MΩA1(M)M\to\Omega_{A_{1}}(M) is zero. Set the composition of the first two maps to be α:MΩA(L)/Lx\alpha\colon M\to\Omega_{A}(L)/^{L}x in Mod(A1)\operatorname{Mod}(A_{1}). Furthermore, the commutative diagram (5.5) provides ΩA1(M)P/Lxp/LxL/LxL/xL[1]L/xLL/xL[1]M[1]\Omega_{A_{1}}(M)\to P/^{L}x\xrightarrow{p/^{L}x}L/^{L}x\cong L/xL[1]\oplus L/xL\twoheadrightarrow L/xL[1]\cong M[1] in Mod(A1)\operatorname{Mod}(A_{1}). We have a sequence of maps

(5.8) M𝛼ΩA(L)/Lx𝑟ΩA1(M)M[1]M\xrightarrow{\alpha}\Omega_{A}(L)/^{L}x\xrightarrow{r}\Omega_{A_{1}}(M)\to M[1]

in Mod(A1)\operatorname{Mod}(A_{1}) such that the composition of any two successive maps is zero. In the following, we need to show that the sequence M𝛼ΩA(L)/Lx𝑟ΩA1(M)M\xrightarrow{\alpha}\Omega_{A}(L)/^{L}x\xrightarrow{r}\Omega_{A_{1}}(M) in Mod(A1)\operatorname{Mod}(A_{1}) is a fiber sequence. By Lemma 5.2 below and the discreteness of MM, it suffices to show that the homotopy long sequence induced from M𝛼ΩA(L)/Lx𝑟ΩA1(M)M[1]M\xrightarrow{\alpha}\Omega_{A}(L)/^{L}x\xrightarrow{r}\Omega_{A_{1}}(M)\to M[1] is exact.

By (a), (b), (c), and (d) above, πi(M)πi(ΩA(L)/Lx)πi(r)πi(ΩA1(M))\pi_{i}(M)\to\pi_{i}(\Omega_{A}(L)/^{L}x)\xrightarrow{\pi_{i}(r)}\pi_{i}(\Omega_{A_{1}}(M)) is exact for iZi\in\mathbb{Z}.

We next show the exactness of πi(ΩA(L)/Lx)πi(r)πi(ΩA1(M))πi1(M)\pi_{i}(\Omega_{A}(L)/^{L}x)\xrightarrow{\pi_{i}(r)}\pi_{i}(\Omega_{A_{1}}(M))\to\pi_{i-1}(M) for iZi\in\mathbb{Z}. If i2i\geq 2, the isomorphy of πi(r)\pi_{i}(r) by (a) and the vanishing of πi1(M)\pi_{i-1}(M) shows the exactness. If i=0i=0, the surjectivity of π0(r)\pi_{0}(r) by (c) shows the exactness. It remains to show the exactness for i=1i=1, that is, if an element aπ1(ΩA1(M))a\in\pi_{1}(\Omega_{A_{1}}(M)) maps to 0 via

π1(ΩA1(M))π1(P/Lx)π1(p/Lx)π1(L/Lx)π1(L/xL[1]L/xL)M\pi_{1}(\Omega_{A_{1}}(M))\to\pi_{1}(P/^{L}x)\xrightarrow{\pi_{1}(p/^{L}x)}\pi_{1}(L/^{L}x)\cong\pi_{1}(L/xL[1]\oplus L/xL)\xrightarrow{\cong}M

then it belongs to the image of π1(r)\pi_{1}(r). The first map π1(ΩA1(M))π1(P/Lx)\pi_{1}(\Omega_{A_{1}}(M))\to\pi_{1}(P/^{L}x) is an isomorphism by (5.4) and the discreteness of MM. By (5.5) and (5.6), we have the following commutative diagram:

π1(ΩA(L)/Lx){\pi_{1}(\Omega_{A}(L)/^{L}x)}π1(P/Lx){\pi_{1}(P/^{L}x)}π1(L/Lx)π1(L/xL[1]L/xL){{\pi_{1}(L/^{L}x)\cong\pi_{1}(L/xL[1]\oplus L/xL)}}π1(ΩA1(M)){\pi_{1}(\Omega_{A_{1}}(M))}π1(r)\scriptstyle{\pi_{1}(r)}π1(p/Lx)\scriptstyle{\pi_{1}(p/^{L}x)}\scriptstyle{\cong}

such that the upper horizontal sequence is exact. This shows the desired claim.

Next, we show the exactness of the sequence πi+1(ΩA1(M))πi(M)πi(α)πi(ΩA(L)/Lx)\pi_{i+1}(\Omega_{A_{1}}(M))\to\pi_{i}(M)\xrightarrow{\pi_{i}(\alpha)}\pi_{i}(\Omega_{A}(L)/^{L}x). By the discreteness of MM, it suffices to show the case of i=0i=0. The first map is

(5.9) π1(ΩA1(M))π1(P/Lx)π1(p/Lx)π1(L/Lx)M\pi_{1}(\Omega_{A_{1}}(M))\to\pi_{1}(P/^{L}x)\xrightarrow{\pi_{1}(p/^{L}x)}\pi_{1}(L/^{L}x)\cong M

and the second map is

π0(α):Mπ1(L/Lx)π0(ΩA(L)/Lx)\pi_{0}(\alpha)\colon M\cong\pi_{1}(L/^{L}x)\to\pi_{0}(\Omega_{A}(L)/^{L}x)

by (5.7). If π0(α)\pi_{0}(\alpha) maps aMa\in M to 0 in π0(ΩA(L)/Lx)\pi_{0}(\Omega_{A}(L)/^{L}x), there exists an element bπ1(P/Lx)b\in\pi_{1}(P/^{L}x) such that π1(p/Lx)(b)=aπ0(M)\pi_{1}(p/^{L}x)(b)=a\in\pi_{0}(M) by the exactness of the upper vertical sequence in (5.6). This bb maps to 0 via π1(q):π1(P/Lx)π1(M)=0\pi_{1}(q)\colon\pi_{1}(P/^{L}x)\to\pi_{1}(M)=0 and we can take cπ1(ΩA1(M))c\in\pi_{1}(\Omega_{A_{1}}(M)) which goes to bb in π1(P/Lx)\pi_{1}(P/^{L}x) by the lower exact sequence in (5.6). This cc maps to aa in π1(L/Lx)M\pi_{1}(L/^{L}x)\cong M via the first map of (5.9). This shows the desired exactness.

Therefore, the sequence (5.8) in Mod(A1)\operatorname{Mod}(A_{1}) gives a fiber sequence

(5.10) θL:M𝛼ΩA(L)/Lx𝑟ΩA1(M)\theta_{L}^{\prime}\colon M\xrightarrow{\alpha}\Omega_{A}(L)/^{L}x\xrightarrow{r}\Omega_{A_{1}}(M)

in Mod(A1)\operatorname{Mod}(A_{1}) and an exact sequence

(5.11) π0(θL):Mα0π0(ΩA(L)/Lx)r0π0(ΩA1(M))0\pi_{0}(\theta_{L}^{\prime})\colon M\xrightarrow{\alpha_{0}}\pi_{0}(\Omega_{A}(L)/^{L}x)\xrightarrow{r_{0}}\pi_{0}(\Omega_{A_{1}}(M))\to 0

of discrete π0(A1)\pi_{0}(A_{1})-modules. The two fiber sequences θL\theta_{L} and θL\theta_{L}^{\prime} give the commutative diagram whose vertical maps are distinguished triangles in hMod(A1)\mathrm{h}\operatorname{Mod}(A_{1}):

(5.12) L/Lx[1]{{L/^{L}x[-1]}}ΩA(L)/Lx{\Omega_{A}(L)/^{L}x}P/Lx{P/^{L}x}L/Lx{L/^{L}x}M{{M}}ΩA(L)/Lx{\Omega_{A}(L)/^{L}x}ΩA1(M){\Omega_{A_{1}}(M)}M[1],{{M[1],}}1\scriptstyle{-1}+1\scriptstyle{+1}1\scriptstyle{-1}r\scriptstyle{r}+1\scriptstyle{+1}

the upper sequence is θL\theta_{L} and the lower one is θL\theta^{\prime}_{L}.

In the above construction, we use the following lemma.

Lemma 5.2.

Let 𝒞\mathcal{C} be a stable \infty-category with a tt-structure (see [HA, Definition 1.2.11] for some notation). For example, 𝒞=Mod(A)\mathcal{C}=\operatorname{Mod}(A) by [HA, Remark 1.4.3.8]. Let L𝑔M𝑓NL[1]L\xrightarrow{g}M\xrightarrow{f}N\xrightarrow{h}L[1] be a sequence of maps in 𝒞\mathcal{C} such that any two successive maps are zero in 𝒞\mathcal{C}. By the universality of fib(f)\operatorname{fib}(f), we can take a map s:Lfib(f)s\colon L\to\operatorname{fib}(f) in 𝒞\mathcal{C} such that the following diagram commutes:

L{L}M{M}N{N}fib(f){\operatorname{fib}(f)}M{M}N.{N.}g\scriptstyle{g}s\scriptstyle{s}f\scriptstyle{f}f\scriptstyle{f}

Assume that MM and NN are connective and LL is discrete. Then the following are equivalent:

  1. (1)

    The map L𝑠fib(f)L\xrightarrow{s}\operatorname{fib}(f) is an isomorphism in 𝒞\mathcal{C}. In particular, L𝑔M𝑓NL\xrightarrow{g}M\xrightarrow{f}N is a fiber sequence in 𝒞\mathcal{C}.

  2. (2)

    The homotopy long sequence πi(L)πi(g)πi(M)πi(f)πi(N)\cdots\to\pi_{i}(L)\xrightarrow{\pi_{i}(g)}\pi_{i}(M)\xrightarrow{\pi_{i}(f)}\pi_{i}(N)\to\cdots in 𝒞\mathcal{C}^{\heartsuit} is an exact sequence.

Proof.

(1) \Rightarrow (2): This follows from a general fact about tt-structures.

(2) \Rightarrow (1): By the discreteness of LL, πi(f):πi(M)πi(N)\pi_{i}(f)\colon\pi_{i}(M)\to\pi_{i}(N) is an isomorphism for i0i\neq 0 and surjective for i=0i=0. Considering the long exact sequence of the fiber sequence fib(f)M𝑓N\operatorname{fib}(f)\to M\xrightarrow{f}N, we can show that πi(fib(f))=0\pi_{i}(\operatorname{fib}(f))=0 for i0i\neq 0. The fiber fib(f)\operatorname{fib}(f) is discrete. The connectivity of MM and NN implies the commutative diagram whose horizontal sequences are exact

0{0}π0(L){\pi_{0}(L)}π0(M){\pi_{0}(M)}π0(N){\pi_{0}(N)}0{0}π0(fib(f)){\pi_{0}(\operatorname{fib}(f))}πi(M){\pi_{i}(M)}π0(N).{\pi_{0}(N).}π0(g)\scriptstyle{\pi_{0}(g)}π0(s)\scriptstyle{\pi_{0}(s)}π0(f)\scriptstyle{\pi_{0}(f)}π0(f)\scriptstyle{\pi_{0}(f)}

The five lemma implies that π0(s)\pi_{0}(s) is an isomorphism and we are done. ∎

6. Ext\operatorname{Ext}-vanishings and Liftings of Modules

In this section, we prove our main theorem (Theorem 6.4). The following argument is a “derived” generalization of [ADS93]. We fix an integer n1n\geq 1.

Lemma 6.1.

Let AA be an almost perfect animated Λ\Lambda-algebra and let xx be an element of 𝔪\mathfrak{m}. Take a discrete finitely generated A1A_{1}-module MM and its lifting LL to AnA_{n}. We can take a fiber sequence θL\theta_{L}^{\prime} in Mod(A1)\operatorname{Mod}(A_{1}) and an exact sequence π0(θL)\pi_{0}(\theta_{L}^{\prime}) in Mod(A1)\operatorname{Mod}(A_{1})^{\heartsuit} by (5.10) and (5.11). Then the following are equivalent:

  1. (a)

    The map π0(α):Mπ0(ΩA(L)/Lx)\pi_{0}(\alpha)\colon M\to\pi_{0}(\Omega_{A}(L)/^{L}x) is a split injection in the category Mod(A1)\operatorname{Mod}(A_{1})^{\heartsuit} of discrete π0(A1)\pi_{0}(A_{1})-modules.

  2. (b)

    The fiber sequence θL\theta_{L}^{\prime} splits in hMod(A1)\mathrm{h}\operatorname{Mod}(A_{1}).

  3. (c)

    The map α:MΩA(L)/Lx\alpha\colon M\to\Omega_{A}(L)/^{L}x induces an isomorphism ΩA(L)/LxMΩA1(M)\Omega_{A}(L)/^{L}x\cong M\oplus\Omega_{A_{1}}(M) in hMod(A1)\mathrm{h}\operatorname{Mod}(A_{1}).

  4. (d)

    The AnA_{n}-module LL is liftable to An+1A_{n+1}.

In this case, any lifting EE of the AnA_{n}-module LL to An+1A_{n+1} is a discrete finitely generated An+1A_{n+1}-module by Lemma 4.4.

Proof.

Note that LL is a discrete finitely generated AnA_{n}-module by Lemma 4.4. (b) \Leftrightarrow (c) and (b) \Rightarrow (a) are clear. If (a) holds, we can take a map β0:π0(ΩA(L)/Lx)M\beta_{0}\colon\pi_{0}(\Omega_{A}(L)/^{L}x)\to M of discrete π0(A1)\pi_{0}(A_{1})-modules such that the composition Mπ0(α)π0(ΩA(L)/Lx)β0MM\xrightarrow{\pi_{0}(\alpha)}\pi_{0}(\Omega_{A}(L)/^{L}x)\xrightarrow{\beta_{0}}M is the identity map. Since MM is discrete, π0(α)\pi_{0}(\alpha) can be written as M𝛼ΩA(L)/Lxπ0(ΩA(L)/Lx)M\xrightarrow{\alpha}\Omega_{A}(L)/^{L}x\to\pi_{0}(\Omega_{A}(L)/^{L}x). Then the composition ΩA(L)/Lxπ0(ΩA(L)/Lx)β0M\Omega_{A}(L)/^{L}x\to\pi_{0}(\Omega_{A}(L)/^{L}x)\xrightarrow{\beta_{0}}M gives the splitting of α\alpha and thus (b) holds.

We prove (b) \Rightarrow (d). Since θL\theta_{L}^{\prime} splits in hMod(A1)\mathrm{h}\operatorname{Mod}(A_{1}), we have a map β:ΩA(L)/LxM\beta\colon\Omega_{A}(L)/^{L}x\to M in Mod(A1)\operatorname{Mod}(A_{1}) such that the composition M𝛼ΩA(L)/Lx𝛽MM\xrightarrow{\alpha}\Omega_{A}(L)/^{L}x\xrightarrow{\beta}M is (homotopically equivalent to) the identity map in Mod(A1)\operatorname{Mod}(A_{1}). Set the map γ:ΩA(L)ΩA(L)/Lx𝛽M\gamma\colon\Omega_{A}(L)\to\Omega_{A}(L)/^{L}x\xrightarrow{\beta}M in Mod(A)\operatorname{Mod}(A). Then there exists an AnA_{n}-module EE and the following commutative diagram in Mod(A)\operatorname{Mod}(A)

(6.1) ΩA(L){\Omega_{A}(L)}P{P}L{L}M{{M}}E{E}L{L}γ\scriptstyle{\gamma}g\scriptstyle{g}f\scriptstyle{f}

whose upper and lower horizontal sequences are fiber sequences in Mod(A)\operatorname{Mod}(A). In fact, EE can be taken as the cofiber of the composition L[1]ΩA(L)𝛾ML[-1]\to\Omega_{A}(L)\xrightarrow{\gamma}M in Mod(An)\operatorname{Mod}(A_{n}) and we use the fact that the restriction of scalar preserves small limits and colimits ([HA, Corollary 4.2.3.7]). Since LL is discrete, so is EE. By Lemma 4.6, we want to show that there exists an isomorphism π0(EAn+1LA1)M\pi_{0}(E\otimes^{L}_{A_{n+1}}A_{1})\cong M of π0(A1)\pi_{0}(A_{1})-modules.

Taking derived quotient /Lx-/^{L}x of the above diagram, we have a following commutative diagram in Mod(A1)\operatorname{Mod}(A_{1});

M{M}L/Lx[1]{{L/^{L}x[-1]}}ΩA(L)/Lx{\Omega_{A}(L)/^{L}x}P/Lx{P/^{L}x}L/Lx{L/^{L}x}(ΩA(L)/Lx)/Lx{(\Omega_{A}(L)/^{L}x)/^{L}x}L/Lx[1]{{L/^{L}x[-1]}}M/Lx{M/^{L}x}E/Lx{E/^{L}x}L/Lx.{L/^{L}x.}α\scriptstyle{\alpha}γ/Lx\scriptstyle{\gamma/^{L}x}β/Lx\scriptstyle{\beta/^{L}x}g/Lx\scriptstyle{g/^{L}x}f/Lx\scriptstyle{f/^{L}x}

Taking long exact sequences, we have a following commutative diagram of discrete π0(A1)\pi_{0}(A_{1})-modules;

(6.2) M{M}L/xL{L/xL}π0(ΩA(L)/Lx){\pi_{0}(\Omega_{A}(L)/^{L}x)}π0(P/Lx){\pi_{0}(P/^{L}x)}L/xL{L/xL}0{0}π0((ΩA(L)/Lx)/Lx){\pi_{0}((\Omega_{A}(L)/^{L}x)/^{L}x)}L/xL{L/xL}M{M}E/xE{E/xE}L/xL.{L/xL.}0{0}π0(α)\scriptstyle{\pi_{0}(\alpha)}\scriptstyle{\cong}π0(γ/Lx)\scriptstyle{\pi_{0}(\gamma/^{L}x)}\scriptstyle{\cong}π0(β/Lx)\scriptstyle{\pi_{0}(\beta/^{L}x)}π0(g/Lx)\scriptstyle{\pi_{0}(g/^{L}x)}π0(f/Lx)\scriptstyle{\pi_{0}(f/^{L}x)}

The isomorphism π0(ΩA(L)/Lx)π0((ΩA(L)/Lx)/Lx)\pi_{0}(\Omega_{A}(L)/^{L}x)\cong\pi_{0}((\Omega_{A}(L)/^{L}x)/^{L}x) is the identity map and thus π0(β/Lx)=π0(β)\pi_{0}(\beta/^{L}x)=\pi_{0}(\beta). So the map ML/xLπ0(ΩA(L)/Lx)π0(γ/Lx)MM\cong L/xL\to\pi_{0}(\Omega_{A}(L)/^{L}x)\xrightarrow{\pi_{0}(\gamma/^{L}x)}M is equal to π0(β)π0(α)=id\pi_{0}(\beta)\circ\pi_{0}(\alpha)=\operatorname{id}. In particular, the left lower map L/xLML/xL\to M is surjective and thus π0(f/Lx):E/xEL/xL\pi_{0}(f/^{L}x)\colon E/xE\to L/xL is an isomorphism.

We next prove (d) \Rightarrow (a). Let EE be a lifting of LL to An+1A_{n+1}. By Lemma 4.4 and Lemma 4.6, EE is a discrete finitely generated An+1A_{n+1}-module and there exists a fiber sequence M𝑔E𝑓LM\xrightarrow{g}E\xrightarrow{f}L in Mod(An+1)\operatorname{Mod}(A_{n+1}) and an isomorphism E/xEME/xE\cong M of π0(A1)\pi_{0}(A_{1})-modules. Take the fiber sequence ΩA(L)PL\Omega_{A}(L)\to P\to L in Mod(A)\operatorname{Mod}(A) defined in (5.1). Since PP is projective over AA, there exists a map γ:PE\gamma^{\prime}\colon P\to E such that PγELP\xrightarrow{\gamma^{\prime}}E\twoheadrightarrow L is PLP\to L. Since M𝑔E𝑓LM\xrightarrow{g}E\xrightarrow{f}L and ΩA(L)PL\Omega_{A}(L)\to P\to L are fiber sequences, γ\gamma^{\prime} induces a map γ:ΩA(L)M\gamma\colon\Omega_{A}(L)\to M of AA-modules making the same diagram in (6.1) commutative. As in the proof of (b) \Rightarrow (d), we obtain the same commutative diagram in (6.2). In particular, ff induces a surjective endomorphism π0(f/Lx):MM\pi_{0}(f/^{L}x)\colon M\twoheadrightarrow M on the Noetherian π0(A1)\pi_{0}(A_{1})-module MM because of E/xEML/xLE/xE\cong M\cong L/xL and thus π0(f/Lx)\pi_{0}(f/^{L}x) is furthermore injective. This shows the surjectivity of the (left lower) map ML/xLMM\cong L/xL\to M in (6.2) and thus this is also an isomorphism. Set the inverse to be ϕ\phi. By the definition of the map α:ML/xLL/Lx[1]ΩA(L)/Lx\alpha\colon M\cong L/xL\to L/^{L}x[-1]\to\Omega_{A}(L)/^{L}x in (5.7), the inverse map MϕL/xLM\xrightarrow{\phi}L/xL gives the split s:π0(ΩA(L)/Lx)π0(γ/Lx)MϕL/xLMs\colon\pi_{0}(\Omega_{A}(L)/^{L}x)\xrightarrow{\pi_{0}(\gamma/^{L}x)}M\xrightarrow{\phi}L/xL\cong M of π0(α):Mπ0(ΩA(L)/Lx)\pi_{0}(\alpha)\colon M\to\pi_{0}(\Omega_{A}(L)/^{L}x) by the left pentagon in (6.2).

Lemma 6.2.

Let AA be an almost perfect animated Λ\Lambda-algebra and let xx be an element of 𝔪\mathfrak{m}. Take a discrete finitely generated A1A_{1}-module MM and its lifting LL to AnA_{n}. Then there exists an element ob(M,L,n)\mathrm{ob}(M,L,n) in ExtA12(M,M)\operatorname{Ext}^{2}_{A_{1}}(M,M) such that its vanishing is equivalent to the existence of a discrete finitely generated An+1A_{n+1}-module EE which is a lifting of the AnA_{n}-module LL to An+1A_{n+1}. In particular, if ExtA12(M,M)\operatorname{Ext}^{2}_{A_{1}}(M,M) vanishes, then there exists such an EE.

Proof.

Applying Lemma 4.4 for MM and LL, we can show that LL is discrete. By Lemma 6.1, it suffices to show that the map α:MΩA(L)/Lx\alpha\colon M\to\Omega_{A}(L)/^{L}x in the fiber sequence θL\theta_{L}^{\prime} (5.10) splits in hMod(A1)\mathrm{h}\operatorname{Mod}(A_{1}). For the fiber sequence θL\theta_{L}^{\prime}, we can take the long exact sequence of Ext\operatorname{Ext}-groups

HomA1(ΩA(L)/Lx,M)αHomA1(M,M)ExtA11(ΩA1(M),M).\operatorname{Hom}_{A_{1}}(\Omega_{A}(L)/^{L}x,M)\xrightarrow{-\circ\alpha}\operatorname{Hom}_{A_{1}}(M,M)\to\operatorname{Ext}^{1}_{A_{1}}(\Omega_{A_{1}}(M),M).

Set the element ob(M,L,n)\mathrm{ob}^{\prime}(M,L,n) in ExtA11(ΩA1(M),M)\operatorname{Ext}^{1}_{A_{1}}(\Omega_{A_{1}}(M),M) to be the image of idMHomA1(M,M)\operatorname{id}_{M}\in\operatorname{Hom}_{A_{1}}(M,M). The existence of a splitting of α\alpha is equivalent to the vanishing of ob(M,L,n)\mathrm{ob}^{\prime}(M,L,n) by the long exact sequence. By the fiber sequence (5.4), we have a fiber sequence of A1A_{1}-modules

ExtA11(P/Lx,M)ExtA11(ΩA1(M),M)ExtA12(M,M).\operatorname{Ext}^{1}_{A_{1}}(P/^{L}x,M)\to\operatorname{Ext}^{1}_{A_{1}}(\Omega_{A_{1}}(M),M)\to\operatorname{Ext}^{2}_{A_{1}}(M,M).

The free A1A_{1}-module P/Lx=(A)/LxA1P/^{L}x=(\oplus A)/^{L}x\cong\oplus A_{1} is projective over A1A_{1}. The first term vanishes by [HA, Proposition 7.2.2.6 and Proposition 7.2.2.7]. So we have an isomorphism ExtA11(ΩA1(M),M)ExtA12(M,M)\operatorname{Ext}^{1}_{A_{1}}(\Omega_{A_{1}}(M),M)\xrightarrow{\cong}\operatorname{Ext}^{2}_{A_{1}}(M,M). Set ob(M,L,n)\mathrm{ob}(M,L,n) as the image of ob(M,L,n)ExtA11(ΩA1(M),M)\mathrm{ob}^{\prime}(M,L,n)\in\operatorname{Ext}^{1}_{A_{1}}(\Omega_{A_{1}}(M),M) in ExtA12(M,M)\operatorname{Ext}^{2}_{A_{1}}(M,M). By the above argument, the vanishing of ob(M,L,n)\mathrm{ob}(M,L,n) is equivalent to the existence of a lifting EE of LL to An+1A_{n+1}. ∎

Lemma 6.3.

Let AA be an almost perfect animated Λ\Lambda-algebra and let xx be an element of 𝔪\mathfrak{m}. Let MM be a discrete finitely generated A1A_{1}-module. If ExtA12(M,M)\operatorname{Ext}^{2}_{A_{1}}(M,M) vanishes, there exists a discrete finitely generated AA-module LL which is a lifting of the A1A_{1}-module MM to AA.

Proof.

Set L1ML_{1}\coloneqq M. By Lemma 6.2 and our assumption, there exists a discrete finitely generated A2A_{2}-module L2L_{2} such that L2L_{2} is a lifting of the A1A_{1}-module L1L_{1} to A2A_{2}. Again, by using Lemma 6.2 for L2L_{2} and MM, we can take a discrete finitely generated lifting L3L_{3} of the A2A_{2}-module L2L_{2} to A3A_{3}. By repeating this process, we can take a sequence {Ln}n0\{L_{n}\}_{n\geq 0} such that each LnL_{n} is a discrete finitely generated AnA_{n}-module and Ln+1L_{n+1} is a lifting of the AnA_{n}-module LnL_{n} to An+1A_{n+1}. By Theorem 4.5, the A1A_{1}-module MM has a lifting LL to AA. ∎

Under these preparations, we can prove the following lifting theorem.

Theorem 6.4.

Let AA be an almost perfect animated Λ\Lambda-algebra and let x1,,xtx_{1},\dots,x_{t} be a sequence of elements of 𝔪\mathfrak{m}. Set ΓjA/L(x1,,xj)\Gamma_{j}\coloneqq A/^{L}(x_{1},\dots,x_{j}) for each 1jt1\leq j\leq t. Let MM be a discrete finitely generated Γt\Gamma_{t}-module. If ExtΓt2(M,M)\operatorname{Ext}^{2}_{\Gamma_{t}}(M,M) vanishes, then there exists a discrete finitely generated AA-module LL which is a lifting of the Γt\Gamma_{t}-module MM to AA.

Proof.

Since AA is almost perfect, so is Γj\Gamma_{j} for each jj. Set LtML_{t}\coloneqq M. Our assumption is that ExtΓt2(Lt,Lt)\operatorname{Ext}^{2}_{\Gamma_{t}}(L_{t},L_{t}) vanishes. We can apply Lemma 6.3 when we set MLtM\coloneqq L_{t}, AΓt1A\coloneqq\Gamma_{t-1}, and xxtx\coloneqq x_{t}. Then there exists a discrete finitely generated Γt1\Gamma_{t-1}-module Lt1L_{t-1} and an isomorphism Lt1Γt1LΓtML_{t-1}\otimes^{L}_{\Gamma_{t-1}}\Gamma_{t}\cong M of Γt\Gamma_{t}-modules.

Next, we iterate this process. The derived quotient Γt=Γt1/Lxt\Gamma_{t}=\Gamma_{t-1}/^{L}x_{t} gives a fiber sequence Lt1×xtLt1ML_{t-1}\xrightarrow{\times x_{t}}L_{t-1}\to M of Γt1\Gamma_{t-1}-modules. Applying ExtΓt1i(Lt1,)\operatorname{Ext}^{i}_{\Gamma_{t-1}}(L_{t-1},-) for the above fiber sequence, we have an exact sequence of π0(Γt1)\pi_{0}(\Gamma_{t-1})-modules

(6.3) ExtΓt1i(Lt1,Lt1)×xt1ExtΓt1i(Lt1,Lt1)ExtΓt1i(Lt1,Lt).\operatorname{Ext}^{i}_{\Gamma_{t-1}}(L_{t-1},L_{t-1})\xrightarrow{\times x_{t-1}}\operatorname{Ext}^{i}_{\Gamma_{t-1}}(L_{t-1},L_{t-1})\to\operatorname{Ext}^{i}_{\Gamma_{t-1}}(L_{t-1},L_{t}).

The tensor-forgetful adjunction and our assumption give the isomorphism

ExtΓt12(Lt1,Lt)ExtΓt2(Lt1Γt1LΓt,Lt)ExtΓt2(Lt,Lt)=ExtΓt2(M,M)=0.\operatorname{Ext}^{2}_{\Gamma_{t-1}}(L_{t-1},L_{t})\cong\operatorname{Ext}^{2}_{\Gamma_{t}}(L_{t-1}\otimes^{L}_{\Gamma_{t-1}}\Gamma_{t},L_{t})\cong\operatorname{Ext}^{2}_{\Gamma_{t}}(L_{t},L_{t})=\operatorname{Ext}^{2}_{\Gamma_{t}}(M,M)=0.

Since π0(Γt1)π0(A)/(x1,,xt1)\pi_{0}(\Gamma_{t-1})\cong\pi_{0}(A)/(x_{1},\dots,x_{t-1}) is an 𝔪\mathfrak{m}-adically complete Noetherian local ring and Lt1L_{t-1} is a discrete finitely generated Γt1\Gamma_{t-1}-module, the base change Lt1Γt1Lπ0(Γt1)L_{t-1}\otimes^{L}_{\Gamma_{t-1}}\pi_{0}(\Gamma_{t-1}) is almost perfect over π0(Γt1)\pi_{0}(\Gamma_{t-1}) by Lemma 2.13 (b). So the Ext\operatorname{Ext}-group Extπ0(Γt1)2(Lt1Γt1Lπ0(Γt1),Lt1)ExtΓt12(Lt1,Lt1)\operatorname{Ext}^{2}_{\pi_{0}(\Gamma_{t-1})}(L_{t-1}\otimes^{L}_{\Gamma_{t-1}}\pi_{0}(\Gamma_{t-1}),L_{t-1})\cong\operatorname{Ext}^{2}_{\Gamma_{t-1}}(L_{t-1},L_{t-1}) is also a discrete finitely generated π0(Γt1)\pi_{0}(\Gamma_{t-1})-module (this also follows from Lemma 2.16). By Nakayama’s lemma, the Ext\operatorname{Ext}-group ExtΓt12(Lt1,Lt1)\operatorname{Ext}^{2}_{\Gamma_{t-1}}(L_{t-1},L_{t-1}) vanishes.

Inductively, we can take a discrete and finite Γj\Gamma_{j}-module LjL_{j} and an isomorphism LjΓjLΓj+1Lj+1L_{j}\otimes^{L}_{\Gamma_{j}}\Gamma_{j+1}\cong L_{j+1} of Γj+1\Gamma_{j+1}-modules for each 0jt10\leq j\leq t-1 (we set Γ0A\Gamma_{0}\coloneqq A). The isomorphism

L0Γ0LΓt\displaystyle L_{0}\otimes^{L}_{\Gamma_{0}}\Gamma_{t} L0Γ0LΓ1Γ1LΓtL1Γ1LΓt\displaystyle\cong L_{0}\otimes^{L}_{\Gamma_{0}}\Gamma_{1}\otimes^{L}_{\Gamma_{1}}\Gamma_{t}\cong L_{1}\otimes^{L}_{\Gamma_{1}}\Gamma_{t}
Lt1Γt1LΓtLt=M\displaystyle\cong\cdots\cong L_{t-1}\otimes^{L}_{\Gamma_{t-1}}\Gamma_{t}\cong L_{t}=M

of Γt\Gamma_{t}-modules shows that the Γ0\Gamma_{0}-module LL0L\coloneqq L_{0} is a lifting of the Γt\Gamma_{t}-module MM to AA. ∎

7. Rerevance to Auslander’s Zero-Divisor Theorem

Once we try to apply Theorem 6.4 for a regular local ring AA, the result is covered by Auslander–Ding–Solberg’s result [ADS93]. In this section, we show this result (Corollary 7.2) by combining our result (Remark 3.7) and Auslander’s zero-divisor theorem (Theorem 7.1).

Auslander [Aus61] proposed the following so-called Auslander’s zero-divisor conjecture, which already has been proved.

Theorem 7.1 ([Aus61]).

Let AA be a Noetherian local ring and let MM be a non-zero discrete finitely generated AA-module with finite projective dimension. If a sequence of elements f1,,frf_{1},\dots,f_{r} of AA is a regular sequence on MM, then f1,,frf_{1},\dots,f_{r} is also a regular sequence on AA.

If AA is a regular local ring, this is a consequence of Serre’s intersection theorem [Ser75]. The general statement follows from Peskine–Szpiro’s intersection theorem which is developed and solved by Peskine–Szpiro [PS69, PS73] in positive characteristic or essentially finite type over a field, Hochster [Hoc74] in equicharacteristic, and Roberts [Rob87] in mixed characteristic. See also a survey [Rob89] or a book [Rob98, Theorem 6.2.3]. Moreover, this theorem is one of the members of the initial homological conjecture and thus is now a consequence of, in particular, direct summand conjecture (see, for example, [Dos23, §1.6]).

By using this, we can show the “full circle” result.

Corollary 7.2.

Let AA be a regular local ring and let f1,,frf_{1},\dots,f_{r} be a sequence of elements of AA. Set A/Lf¯=A/L(f1,,fr)A/^{L}\underline{f}=A/^{L}(f_{1},\dots,f_{r}) and RA/(f1,,fr)R\coloneqq A/(f_{1},\dots,f_{r}) If there exists a discrete finitely generated non-zero RR-module MM such that ExtA/Lf¯2(M,M)=0\operatorname{Ext}^{2}_{A/^{L}\underline{f}}(M,M)=0, then f1,,frf_{1},\dots,f_{r} is a regular sequence of AA.

Proof.

If f1,,frf_{1},\dots,f_{r} is a regular sequence of AA, then RR itself can be taken as a desired RR-module MM. Conversely, by Theorem 6.4, there exists a discrete finitely generated non-zero AA-module LL such that LALA/Lf¯ML\otimes^{L}_{A}A/^{L}\underline{f}\cong M. Since AA is regular local, LL has a finite projective dimension over AA. The derived tensor product MLALA/Lf¯L/L(f1,,fr)M\cong L\otimes^{L}_{A}A/^{L}\underline{f}\cong L/^{L}(f_{1},\dots,f_{r}) is discrete and thus f1,,frf_{1},\dots,f_{r} is a (Koszul-)regular sequence on LL (see, for example, [Sta, 09CC]). The Auslander’s zero-divisor theorem (Theorem 7.1) shows that f1,,frf_{1},\dots,f_{r} is a regular sequence on AA. We are done. ∎

This result was not mentioned in [NS13], in which they proved a similar result of Theorem 6.4 by using DG methods.

Remark 7.3.

The original assumption of the Auslander–Reiten conjecture (1.1) is the vanishing of ExtRi(M,MR)\operatorname{Ext}^{i}_{R}(M,M\oplus R) for each i1i\geq 1. On the contrary, under the setting of Theorem 1.3 (or more generally Theorem 6.4), our assumption should be ExtA/Lf¯i(M,MR)=0\operatorname{Ext}^{i}_{A/^{L}\underline{f}}(M,M\oplus R)=0 for each i1i\geq 1. However, because of Corollary 7.2 (this is nontrivial if we do not have our main theorem (Theorem 6.4)), our assumption implies that f1,,frf_{1},\dots,f_{r} is a regular sequence on AA and thus those two assumptions are (trivially) equivalent.

Moreover, by Lemma 3.11 and Corollary 7.2, for a non-regular sequence f1,,frf_{1},\dots,f_{r} on the regular local ring AA, the vanishing ExtRi(M,M)\operatorname{Ext}^{i}_{R}(M,M) is strictly weaker than the vanishing ExtA/Lf¯i(M,M)\operatorname{Ext}^{i}_{A/^{L}\underline{f}}(M,M) for each i1i\geq 1 since the latter vanishing implies the regularity of f1,,frf_{1},\dots,f_{r} on AA.

References

  • [Ara25] K. Arakawa, Homotopy Limits and Homotopy Colimits of Chain Complexes, (2025). http://arxiv.org/abs/2310.00201.
  • [ADS93] M. Auslander, S. Ding, and O. Solberg, Liftings and Weak Liftings of Modules, Journal of Algebra, 156(2) (1993) 273–317.
  • [AR75] M. Auslander and I. Reiten, On a Generalized Version of the Nakayama Conjecture, Proceedings of the American Mathematical Society, 52(1) (1975) 69–74.
  • [Aus61] M. Auslander, Modules over Unramified Regular Local Rings, Illinois Journal of Mathematics, 5(4) (1961) 631–647.
  • [Bha21] B. Bhatt, Cohen-Macaulayness of Absolute Integral Closures, (2021). http://arxiv.org/abs/2008.08070.
  • [BL22] B. Bhatt and J. Lurie, Absolute Prismatic Cohomology, (2022). http://arxiv.org/abs/2201.06120.
  • [Bun13] U. Bunke, Differential Cohomology, (2013). http://arxiv.org/abs/1208.3961.
  • [CH10] L. W. Christensen and H. Holm, Algebras That Satisfy Auslander’s Condition on Vanishing of Cohomology, Mathematische Zeitschrift, 265(1) (2010) 21–40.
  • [ČS24] K. Česnavičius and P. Scholze, Purity for Flat Cohomology, Annals of Mathematics, 199(1) (2024) 51–180.
  • [Dos23] G. Dospinescu, La Conjecture Du Facteur Direct (d’après Y.André et B.Bhatt), Astérisque, 446 (2023) 141–197.
  • [GW23] U. Görtz and T. Wedhorn, Algebraic Geometry II: Cohomology of Schemes: With Examples and Exercises, Springer Studium Mathematik - Master, (2023) Springer Fachmedien.
  • [Hoc74] M. Hochster, The Equicharacteristic Case of Some Homological Conjectures on Local Rings, Bulletin of the American Mathematical Society, 80(4) (1974) 683–686.
  • [IN24] R. Ishizuka and K. Nakazato, Prismatic Kunz’s Theorem, (2024). http://arxiv.org/abs/2402.06207.
  • [Kim23] K. Kimura, Auslander–Reiten Conjecture for Normal Rings, (2023). http://arxiv.org/abs/2304.03956, to appear in Israel Journal of Mathematics.
  • [KOT22] K. Kimura, Y. Otake, and R. Takahashi, Maximal Cohen–Macaulay Tensor Products and Vanishing of Ext Modules, Bulletin of the London Mathematical Society, 54(6) (2022) 2456–2468.
  • [KR19] A. A. Khan and D. Rydh, Virtual Cartier Divisors and Blow-Ups, (2019). http://arxiv.org/abs/1802.05702.
  • [KST18] M. Kerz, F. Strunk, and G. Tamme, Algebraic K-theory and Descent for Blow-Ups, Inventiones mathematicae, 211(2) (2018) 523–577.
  • [Kum23] S. Kumashiro, The Auslander-Reiten Conjecture for Certain Non-Gorenstein Cohen-Macaulay Rings, Journal of Pure and Applied Algebra, 227(11) (2023) 107420.
  • [DAG] J. Lurie, Derived Algebraic Geometry, PhD Thesis, Massachusetts Institute of Technology, (2004).
  • [HTT] J. Lurie, Higher Topos Theory, Annals of Mathematics Studies, (no. 170) (2009) Princeton University Press.
  • [HA] J. Lurie, Higher Algebra (2017). https://www.math.ias.edu/~lurie/papers/HA.pdf.
  • [SAG] J. Lurie, Spectral Algebraic Geometry (2018). https://www.math.ias.edu/~lurie/papers/SAG-rootfile.pdf.
  • [Mat86] H. Matsumura, Commutative Ring Theory, Cambridge Studies in Advanced Mathematics, (8) (1986) Cambridge University Press.
  • [NOY22] S. Nasseh, M. Ono, and Y. Yoshino, Naïve Liftings of DG Modules, Mathematische Zeitschrift, 301(1) (2022) 1191–1210.
  • [NOY25] S. Nasseh, M. Ono, and Y. Yoshino, Diagonal tensor algebra and naïve liftings, Journal of Pure and Applied Algebra, 229 (2025) 107808..
  • [NS13] S. Nasseh and S. Sather-Wagstaff, Liftings and Quasi-Liftings of DG Modules, Journal of Algebra, 373 (2013) 162–182.
  • [NY18] S. Nasseh and Y. Yoshino, Weak Liftings of DG Modules, Journal of Algebra, 502 (2018) 233–248.
  • [OY21] M. Ono and Y. Yoshino, A Lifting Problem for DG Modules, Journal of Algebra, 566 (2021) 342–360.
  • [PS69] C. Peskine and L. Szpiro, Sur La Topologie Des Sous-Schémas Fermés d’un Shéma Localement Noethérien, Définis Comme Support d’un Faisceau Cohérent Localement de Dimension Projective Finie, Comptes Rendus Hebdomadaires des Séances de l’Académie des Sciences. Séries A et B, 269 (1969) A49–A51.
  • [PS73] C. Peskine and L. Szpiro, Dimension Projective Finie et Cohomologie Locale, Publications Mathématiques de l’IHÉS, 42 (1973) 47–119.
  • [PSY14] M. Porta, L. Shaul, and A. Yekutieli, On the Homology of Completion and Torsion, Algebras and Representation Theory, 17(1) (2014) 31–67.
  • [PSY15] M. Porta, L. Shaul, and A. Yekutieli, Cohomologically Cofinite Complexes, Communications in Algebra, 43(2) (2015) 597–615.
  • [Rob87] P. Roberts, Le Théorème d’intersection, Comptes Rendus des Séances de l’Académie des Sciences. Série I. Mathématique, 304(7) (1987) 177–180.
  • [Rob89] P. Roberts, Intersection Theorems, Commutative Algebra, (1989), 417–436.
  • [Rob98] P. C. Roberts, Multiplicities and Chern Classes in Local Algebra, Cambridge Tracts in Mathematics, (1998) Cambridge University Press.
  • [Ser75] J-P. Serre, Algèbre Locale \cdot Multiplicités, Lecture Notes in Mathematics, (1975) Springer.
  • [Shi07] B. Shipley, HZ-Algebra Spectra Are Differential Graded Algebras, American Journal of Mathematics, 129(2) (2007) 351–379.
  • [Sta] The Stacks Project Authors, Stacks Project. https://stacks.math.columbia.edu.
  • [Yek25] A. Yekutieli, Derived Complete Complexes at Weakly Proregular Ideals, Journal of Pure and Applied Algebra, 229 (2025) 107909.
  • [Yos97] Y. Yoshino, The Theory of L-Complexes and Weak Liftings of Complexes, Journal of Algebra, 188(1) (1997) 144–183.