Spectral Excision and Descent for Almost Perfect Complexes
Abstract.
We show that almost perfect complexes of commutative ring spectra satisfy excision and -descent. These results generalize Milnor excision for perfect complexes of ordinary commutative rings and -descent for almost perfect complexes of locally noetherian derived stacks by Halpern-Leistner and Preygel, respectively.
Key words and phrases:
almost perfect complexes, v-topology, Milnor squares, commutative ring spectra, infinity-categories2020 Mathematics Subject Classification:
14A30, 18F10, 19E08, 55P43, 18N601. Introduction
1.1.
Suppose we are given a square of associative rings :
for which the map is surjective. According to Milnor (see [11, §2]), the image of under the functor which assigns to each associative ring the category of finitely generated projective -modules is a pullback diagram of categories. To work in the more general context of structured ring spectra, recall that if is a commutative ring, then a chain complex of -modules (viewed as an object of the derived category of -modules) is perfect if it is quasi-isomorphic to a bounded complex of finitely generated projective -modules (see, for example, [13, Tag 0657]). More generally, we say that is pseudo-coherent it if is quasi-isomorphic to a bounded above complex of finitely generated free -modules (see, for example, [1, p.79]). Suppose now that is an -ring in the sense of [8, 7.1.0.1] and let denote the -category of -modules (see [8, 7.1.1.2]). Then the notions of perfect and pseudo-coherent modules over commutative rings can be generalized to the setting of -rings, and we obtain the notions of perfect and almost perfect modules over , respectively (see [8, 7.2.4.1] and [8, 7.2.4.10]). We will denote by and the full subcategories of spanned by the perfect and almost perfect -modules, respectively. The constructions determine functors , where and denote the -category of connective -rings and the -category of -categories, respectively (see [8, p.1201] and [6, 3.0.0.1]).
1.2.
One of the main results in this paper is the following analogue of Milnor’s result in the setting of -rings:
Theorem 1.3.
Suppose we are given a pullback square of connective -rings
If the induced map is a surjection of commutative rings, then the diagram of -categories
determined by the extension of scalars functors is a pullback square in the -category .
Remark 1.4.
In our proof of 1.3, we will make use of the notion of a universal descent morphism of -rings in the sense of [7, D.3.1.1], which was introduced originally by Akhil Mathew in [9, 3.18]. We note that the notion of universal descent morphisms makes sense more generally for -rings. However, it is in the commutative setting that the class of universal descent morphisms has the descent property of [7, D.3.5.8], which will play an important role in our discussion of almost perfect complexes. For this reason, we focus our attention to the case of -rings (unlike Milnor’s excision, which works for associative rings).
Remark 1.5.
Remark 1.6.
Remark 1.7.
By restricting the equivalence of 1.3 to the full subcategories spanned by the dualizable objects (see the proof of 2.15), we immediately deduce that the canonical map is an equivalence of -categories. We remark that this equivalence is proven in [9, 2.23] under the additional assumption that the map is a surjection.
Remark 1.8.
According to [4, 1.17] of Markus Land and Georg Tamme, is a pullback diagram of -categories if is a pullback square of -rings for which the functor is conservative (see [4, 1.3] for more details). In the case of connective -rings, the condition appearing in [4, 1.17] is satisfied if the map is surjective (see [4, p.912] and [4, 1.3]). Consequently, 1.7, which is an immediate consequence of our more general result 1.3, can be deduced from [4, 1.17]. We note that our proof of 1.7, which is very succinct, does not involve the construction of the -ring appearing in the statement of [4, 1.17].
1.9.
Fix a finite field of elements for some prime number . Recall that an -scheme is said to be perfect if the Frobenius map is an isomorphism (see, for example, [2, 3.1]). This paper was motivated by [2, 11.2] of Bhargav Bhatt and Peter Scholze, which shows that the functor which carries each perfect quasi-compact and quasi-separated -scheme to the -category of perfect complexes on (see, for example, [7, 2.8.4.4]) is a hypercomplete sheaf with respect to the -topology of [2, 3.2] in the sense of [6, p.669]. We note that this result depends crucially on the fact that the derived tensor product can be identified with the ordinary tensor product for perfect rings; see [2, 3.16].
1.10.
Our second main result shows that the restriction to perfect schemes can be removed from [2, 11.2] if we work in the setting of spectral algebraic geometry (see 2.2 for the notion of -cover in the spectral setting). In fact, we prove this not just for perfect modules, but for more general almost perfect modules:
Theorem 1.11.
Let be a -cover of connective -rings for which the underlying map of commutative rings is of finite presentation. Then is of universal -descent: that is, for every morphism in the -category of connective -rings, the induced map is an equivalence of -categories, where denotes the Čech nerve of the map (formed in the opposite of the -category ).
Remark 1.12.
If is a morphism of noetherian simplicial commutative rings for which the induced map is a finitely presented -cover of ordinary commutative rings, it then follows from [3, 3.3.1] of Halpern-Leistner and Preygel that almost perfect complexes satisfy (not necessarily universal) descent for the morphism ; see 2.13. Our descent result 1.11 can be regarded as a generalization of their work (in the affine case): it holds without the locally noetherian assumption appearing in [3, 3.3.1] and is valid more generally for connective -rings, rather than merely for simplicial commutative rings. Moreover, the -descent for the morphism is universal (that is, satisfies descent for arbitrary base change of the map ).
Remark 1.13.
Let be a morphism of noetherian connective -rings for which the induced map is a finitely presented -cover of ordinary commutative rings. Then [3, 3.3.6] of Halpern-Leistner and Preygel shows that the functor satisfies (not necessarily universal) descent for the map (see 2.9). Using a slight variant of the proof of 1.11, we can remove the noetherian assumption on and to show that is of universal -descent; see 3.8.
Remark 1.14.
As a consequence of 1.11, we will see in 3.9 that is also of universal -descent. In particular, the functor satisfies descent with respect to the Grothendieck topology on which is characterized by the following property: a sieve is a covering if and only if it contains a finite collection of maps for which the induced map is a -cover such that the underlying map of commutative rings exhibits as a finitely presented -algebra. In contrast with [2, 11.2] for ordinary perfect schemes, the sheaf is not hypercomplete (that is, it is not true in general that if is a hypercovering with respect to the above Grothendieck topology in the sense of [7, A.5.7.1], then the composition is a limit diagram). To see this, we note that for every connective -ring , the truncation map is a -cover, so that the constant cosimplicial -ring with value is a hypercovering of with respect to the above Grothendieck topology. Consequently, if the functor is hypercomplete, then the natural map is an equivalence of -categories, which is false in general.
Remark 1.15.
Fix integers . Since the functor is a sheaf for the Grothendieck topology appearing in 1.14 by virtue of 3.9, so is the subfunctor which assigns to each connective -ring the full subcategory spanned by those perfect complexes whose Tor-amplitude is contained in ; see, for example, [8, 7.2.4.21]. As Akhil Mathew pointed out, the functor does not take values in the -category of -categories in the sense of [6, 2.3.4.1], unlike in the case of ordinary perfect schemes of [2, 11.2]. Consequently, the fact that is a sheaf does not guarantee that it is a hypercomplete sheaf (see [6, 6.5.2.9]). In fact, for a connective -ring , the canonical map is not an equivalence in general, so that is not hypercomplete.
1.16.
1.17.
Acknowledgements. The author is thankful to Bhargav Bhatt and Akhil Mathew for their helpful comments and suggestions. This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIT) (No. 2020R1A5A1016126) and the Sogang University Research Grant of 2023 (No. 202310024.01).
2. The -covers and universal descent morphisms of connective -rings
2.1.
Let be a morphism of quasi-compact and quasi-separated schemes. Recall from [2, 2.1] (see also [12, 2.9]) that is said to be a -cover if, for every valuation ring and every morphism of schemes , there exist an extension of valuation rings (that is, an injective local homomorphism) and a morphism of schemes , which fit into a commutative diagram
In the spectral setting, we define the -cover of connective -rings as follows:
Definition 2.2.
Remark 2.3.
2.4.
Lemma 2.5.
-
(i)
The collection of -covers of connective -rings contains all equivalences and is stable under composition.
-
(ii)
Suppose we are given a pushout diagram of connective -rings
If is a -cover, then so is .
2.6.
The notion of -covers appearing in 2.2 is closely related to the notion of a universal descent morphism of -rings (which makes sense more generally for -rings): recall from [7, D.3.1.1] that a morphism of -rings is said to be a universal descent morphism if the smallest stable subcategory of which is closed under retracts and contains all -modules of the form , where is a left -module, coincides with . We note that for a finitely presented map of noetherian commutative rings, the map (regarded as a morphism of discrete -rings) is a universal descent morphism if and only if is a -cover of ordinary schemes; see [2, 11.26].
Remark 2.7.
The class of universal descent morphisms, which will play a central role in our proofs of 1.3 and 1.11, allows us to do descent theory in the spectral setting. In particular, if is a universal descent morphism of -rings, it then follows from [7, D.3.5.8] that the canonical map is an equivalence of symmetric monoidal -categories, where denotes the Čech nerve of formed in the -category .
Remark 2.8.
Let be a morphism of -rings. Let denote the Čech nerve of (so that is given by the -fold tensor product for each ). We denote by the Tot-tower of : that is, the sequence of -modules
Here denotes the limit of the diagram taken over the full subcategory spanned by those objects such that . We note that if is a universal descent morphism, then is a retract of for some integer (see the proof of [7, D.3.2.1]). We will use this fact to deduce some descending properties of universal descent morphisms; see 2.11.
2.9.
Let be a connective -ring. Recall that an -module is said to be almost connective if it is -connective for some integer (see [8, p.1201]). We will let denote the full subcategory spanned by the almost connective -modules. We note that the construction determines a functor . According to [8, 7.2.4.10], an -module is almost perfect if it is -connective for some integer and is an almost compact object of the -category of -connective -modules (that is, for every integer , is compact as an object of ). We will say that an -module is perfect to order n if, for every filtered diagram of -truncated -modules, the canonical map is bijective for and injective for ; see [7, 2.7.0.1]. We remark that an -module is almost perfect if and only if it is perfect to order for each integer (see [7, 2.7.0.2]).
Proposition 2.11.
Let be a universal descent morphism of connective -rings, let be an -module, and let be an integer. Then:
-
(i)
The -module is -connective if and only if the -module is -connective.
-
(ii)
The -module is almost connective if and only if the -module is almost connective.
-
(iii)
The -module is perfect to order if and only if the -module is perfect to order .
-
(iv)
The -module is almost perfect if and only if the -module is almost perfect.
Proof.
Assertions (ii) and (iv) follow immediately from (i) and (iii), respectively. The “only if” directions of assertions (i) and (iii) follow from [8, 7.2.1.23] and [7, 2.7.3.1], respectively. To prove “if” directions, we note that since is a universal descent morphism, it follows from [7, D.3.2.1] that is a retract of for some . To prove the “if” direction of (i), assume that is -connective. In particular, is -connective when viewed as a -module. Since is a retract of the connective -module , is -connective as desired. To complete the proof of (iii), let be a filtered diagram of -truncated -modules. We wish to show that the canonical map is bijective for and injective for . For this, it will suffice to show that is a retract of a bijection for and an injection for . Since is a retract of , we deduce that is a retract of . We are therefore reduced to proving that the horizontal map in the diagram
is bijective for and injective for , where we regard as a -module. Since the forgetful functor commutes with the formation of -truncations and preserves small colimits (see [8, 4.2.3.7]), the vertical maps are equivalences. Consequently, the desired result follows from our assumption that is perfect to order as a module over . ∎
2.12.
Definition 2.13.
Let be an -category which admits fiber products, and let be an -category. Let be a functor. We will say that a morphism in is of -descent if the composition is a limit diagram, where denotes the Čech nerve of , regarded as an augmented simplicial object of with (see [6, p.543]). That is, the canonical map is an equivalence in . In this case, we will also say that satisfies descent for . If this condition holds for every base change of along a morphism , we will say that is of universal -descent, or that satisfies universal descent for .
Theorem 2.14.
Let be a universal descent morphism of connective -rings. Then the map is of universal -descent (see 2.9) and of universal -descent.
Proof.
The property of being a universal descent morphism of -rings is stable under pushouts (see [7, D.3.1.6]), so it will suffice to show that is of -descent and of -descent. Let denote the Čech nerve of , so that the canonical map is an equivalence of symmetric monoidal -categories (see 2.7). Using 2.11, we deduce the desired results by restricting to the full subcategories spanned by the almost connective and almost perfect modules, respectively. ∎
Corollary 2.15.
Let be a universal descent morphism of connective -rings. Then is of universal -descent.
Proof.
As in the proof of 2.14, it will suffice to show that is of -descent. Using 2.14, we obtain an equivalence of symmetric monoidal -categories . Then the desired result follows by restricting to the full subcategories spanned by the dualizable objects (see [8, 4.6.1.7]), because the full subcategory of spanned by the dualizable objects can be identified with by virtue of [8, 4.6.1.11]. ∎
Remark 2.16.
Remark 2.17.
The proof of 2.15 shows that every morphism which is either of universal -descent or of universal -descent is of universal -descent.
Remark 2.18.
The converse of 2.15 is not necessarily true. For example, if denotes the free -algebra over on a single generator of degree , then we will see from 3.9 that the truncation map is of universal -descent. However, the extension of scalars functor is not conservative (since the image of the localization vanishes), so that the truncation map is not a universal descent morphism (see 2.7).
3. Proof of Theorems 1.3 and 1.11
3.1.
Proof of 1.3.
We wish to show that the canonical map is an equivalence of -categories. The assertion that this map is fully faithful follows immediately from the first half of [7, 16.2.0.2], which supplies a fully faithful functor . To prove the essential surjectivity, note that we can identify the objects of with triples , where is an almost perfect -module, is an almost perfect -module, and is an equivalence of -modules. Since almost perfect modules are almost connective (that is, -connective for some integer ), we may assume that and are both -connective for some integer . Using the second part of [7, 16.2.0.2], we can choose an -connective -module such that and in and , respectively. To complete the proof, it will suffice to show that is almost perfect. Since the diagram is a pullback square, the map is a universal descent morphism in the sense of [7, D.3.1.1], so that the desired result follows from 2.11. ∎
3.2.
The proof of 1.11 will require some preliminary results.
Lemma 3.3.
Let be morphism of connective -rings. Then the canonical map is an equivalence of -categories. Moreover, it restricts to equivalences of -categories and .
Proof.
We first note that there is an equivalence of -categories (see [7, 2.5.9.3]). Then [7, 10.2.2.3] guarantees that the natural map
is an equivalence of -categories. Since the tensor product can be identified with (see [7, 10.2.1.7]), we can identify with the limit of the diagram in the -category of (not necessarily small) -categories of [6, 3.0.0.5]. The restrictions to almost perfect and perfect modules follow immediately from [7, 2.7.3.2]. ∎
3.4.
Since an almost connective module is -connective for some integer (see 2.9), we immediately obtain the following:
Corollary 3.5.
Let be morphism of connective -rings. Then the canonical map is an equivalence of -categories.
Lemma 3.6.
Let be a map of -rings which exhibits as a square-zero extension of by a -module in the sense of [8, 7.4.1.6]. Then is a universal descent morphism.
Proof.
3.7.
Proof of 1.11.
Since the collection of -covers between connective -rings for which the underlying map of commutative rings is of finite presentation is closed under pushouts (see 2.5 and [8, 7.2.1.23]), it will suffice to show that the map is of -descent (see 2.13). Let denote the Čech nerve of as in 2.7. We have a commutative diagram
Using 3.3, we see that the vertical maps are equivalences. Consequently, it will suffice to show that for each , the canonical map is of universal -descent. We proceed by induction on . We begin with the case . Note that the composition , which can be identified with , is a finitely presented -cover of ordinary commutative rings, so that it is a universal descent morphism by virtue of [10, 5.5]. Using 2.14, we deduce that it is of universal -descent. It then follows from [5, 3.1.2] that is of universal -descent. To carry out the inductive step, assume that the map is of universal -descent. Using [5, 3.1.2], it will suffice to show that the composition is of universal -descent. Since the truncation map exhibits as a square-zero extension of by (see [8, 7.4.1.28]), it is of universal -descent by virtue of 3.6 and 2.14. Combining this with the inductive hypothesis, the desired result follows from [5, 3.1.2] (which guarantees that the collection of universal -descent morphisms is closed under composition). ∎
Remark 3.8.
Corollary 3.9.
In the situation of 1.11, is of universal -descent.
3.10.
Before recording some other consequences of 1.11, let us introduce a bit of terminology:
Definition 3.11.
Let be a functor, where denotes the -category of spaces (see [6, 1.2.16.1]). Let denote the -category of quasi-coherent sheaves on ; see [7, 6.2.2.1]. We will say that an object is a perfect complex on if, for every connective -ring and every morphism , the pullback belongs to the full subcategory (see 1.1). We let denote the full subcategory of spanned by the perfect complexes on .
Remark 3.12.
In the special case where is representable by an affine spectral Deligne-Mumford stack , the full subcategory corresponds to the full subcategory under the equivalence of -categories .
Remark 3.13.
3.14.
Corollary 3.15.
Let be a quasi-compact and quasi-separated spectral algebraic space of [7, 1.6.8.1], and let denote the functor represented by in the sense of [7, 1.6.4.1]. Let be a -cover of connective -rings for which the underlying map of commutative rings exhibits as a finitely presented algebra over . Then satisfies universal descent for .
Proof.
As in the proof of 1.11, we are reduced to proving that the functor satisfies descent for the map . Let denote Čech nerve of . We have a commutative diagram
where denotes the -category of exact symmetric monoidal functors from to and is defined similarly (see [8, 2.1.3.7]). We wish to show that the upper horizontal map is an equivalence. It follows from [7, 9.6.4.2] that the vertical maps are equivalences. The desired result now follows from 3.9, which guarantees that the bottom horizontal map is an equivalence. ∎
3.16.
References
- [1] Théorie des intersections et théorème de Riemann-Roch, Lecture Notes in Mathematics, Vol. 225, Springer-Verlag, Berlin-New York, 1971, Séminaire de Géométrie Algébrique du Bois-Marie 1966–1967 (SGA 6), Dirigé par P. Berthelot, A. Grothendieck et L. Illusie. Avec la collaboration de D. Ferrand, J. P. Jouanolou, O. Jussila, S. Kleiman, M. Raynaud et J. P. Serre. MR 354655
- [2] Bhargav Bhatt and Peter Scholze, Projectivity of the Witt vector affine Grassmannian, Invent. Math. 209 (2017), no. 2, 329–423. MR 3674218
- [3] Daniel Halpern-Leistner and Anatoly Preygel, Mapping stacks and categorical notions of properness, Compos. Math. 159 (2023), no. 3, 530–589. MR 4560539
- [4] Markus Land and Georg Tamme, On the -theory of pullbacks, Ann. of Math. (2) 190 (2019), no. 3, 877–930. MR 4024564
- [5] Yifeng Liu and Weizhe Zheng, Enhanced six operations and base change theorem for higher artin stacks, 2012, Preprint.
- [6] Jacob Lurie, Higher topos theory, Annals of Mathematics Studies, vol. 170, Princeton University Press, Princeton, NJ, 2009. MR 2522659
- [7] by same author, Spectral algebraic geometry, Last update: Feb 2018, Preprint.
- [8] by same author, Higher algebra, Last update: Sep 2017, Preprint.
- [9] Akhil Mathew, The Galois group of a stable homotopy theory, Adv. Math. 291 (2016), 403–541. MR 3459022
- [10] by same author, Faithfully flat descent of almost perfect complexes in rigid geometry, J. Pure Appl. Algebra 226 (2022), no. 5, Paper No. 106938, 31. MR 4332074
- [11] John Milnor, Introduction to algebraic -theory, Annals of Mathematics Studies, No. 72, Princeton University Press, Princeton, N.J.; University of Tokyo Press, Tokyo, 1971. MR 0349811
- [12] David Rydh, Submersions and effective descent of étale morphisms, Bull. Soc. Math. France 138 (2010), no. 2, 181–230. MR 2679038
- [13] The Stacks Project Authors, Stacks Project, https://stacks.math.columbia.edu, 2018.