Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Equivariant localization and completion in cyclic homology and derived loop spaces

Equivariant localization and completion in cyclic homology and derived loop spaces We prove an equivariant localization theorem over an algebraically closed eld of characteristic zero for smooth quotient stacks by reductive groups X{G in the setting of derived loop spaces as well as Hochschild homology and its cyclic variants. We show that the derived loop spaces of the stack X{G z z and its classical z- xed point stack  pX q{G become equivalent after completion along a semisimple parameter rzs P G{{G, implying the analogous statement for Hochschild and cyclic homology of the dg category of perfect complexes PerfpX{Gq. We then prove an analogue of the Atiyah-Segal completion theorem in the setting of periodic cyclic homology, where the completion of the periodic cyclic homology of PerfpX{Gq at the identity res P G{{G is identi ed with a 2-periodic version of the derived de Rham cohomology of X{G. Together, these results identify the completed periodic cyclic homology of a stack X{G over a parameter rzs P G{{G with the 2-periodic derived de Rham cohomology of its z- xed points. Contents 1 Introduction 1 1.1 Conventions and notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 1.2 Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2 Background 7 2.1 Derived loop spaces and its variants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.2 Cyclic homology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 2.3 The equivariant cyclic bar construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 2.4 Formal and derived completions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 3 An equivariant localization theorem in derived loop spaces and Hochschild homology 22 3.1 Equivariant localization for derived loop spaces . . . . . . . . . . . . . . . . . . . . . . . . . . 22 3.2 Equivariant localization for Hochschild and cyclic homology . . . . . . . . . . . . . . . . . . . 27 4 An Atiyah-Segal completion theorem for periodic cyclic homology 29 4.1 Twisted circle actions on loop spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 4.2 Tate-equivariant functions on formal loop spaces compute analytic de Rham cohomology . . . 31 4.3 Comparing global functions on unipotent and formal loop spaces . . . . . . . . . . . . . . . . 34 1 Introduction Derived loop spaces appear naturally in questions of Hochschild homology. When X is a prestack such that the derived categories QCohpXq and IndCohpXq are compactly generated by PerfpXq and CohpXq respectively, their Hochschild homology can be computed in two ways. On one hand, it is computed by the arXiv:1708.06079v7 [math.AG] 23 Jan 2020 usual cyclic bar complex, but it can also be naturally identi ed with the global sections of a sheaf of functions or distributions on the derived loop space LX of X [BN13b]: HHpQCohpXqq  OpLXq; HHpIndCohpXqq  !pLXq: When X is a derived scheme, the loop space LX is equivalent to the shifted odd tangent bundle via a derived variant of the Hochschild-Kostant-Rosenberg theorem, re ecting the Zariski-local nature of derived loops. Introducing the S -action, results in [BN12] [Pre15] identify periodic cyclic homology with 2-periodic derived de Rham (co)homology: Tate Tate HPpPerfpXqq  OpLXq  C pX ; kqppuqq; HPpCohpXqq  !pLXq  C pX ; kqppuqq ;dR dR 2 1 where u P C pBS ; kq is the degree two universal Chern class. Furthermore, over k  C, the main result in [Bha12] identi es derived de Rham cohomology with Betti de Rham cohomology of the analyti cation, and an we have that HPpPerfpXqq  C pX ; kqppuqq. Our goal in this paper is to investigate the periodic cyclic dR homology where X is taken to be a smooth quotient stack. A diculty in understanding the Hochschild homology of stacks which does not appear for schemes is the failure of derived loop spaces to form a cosheaf for the smooth topology; informally, Hochschild homology is Zariski local but not smooth local . On the other hand, as argued in [BN12], sheaves and functions on formal loop spaces (i.e. loop spaces completed at constant loops) can be computed via a smooth covers. We begin to bridge this gap by understanding the global loop space LpX{Gq after completion over bers of the map to LpBGq  G{G as well as to its anization G{{G. We show that completion over G{G gives a certain formal loop space and completion over G{{G gives a certain unipotent loop space, formalizing a \Jordan decomposition for loop spaces" which appeared in [BN13a], where it was applied to realize a moduli stack of Langlands parameters for representations of real reductive groups. Informally, we view X{G as a family of formal loop spaces over G{G, which we in turn view as a family of unipotent loop spaces over G{{G, ultimately realizing X{G as a family of unipotent loop spaces over the ane parameter space G{{G. This paper contains two main theorems. The rst realizes well-established equivariant localization pat- terns (e.g. as in [GKM98] [Tho87] [Tho86]) in the setting of Hochschild homology via its geometric avatar, the derived loop space. The second realizes an Atiyah-Segal style (e.g. as in [AS69] [Tho88]) completion theorem identifying completed periodic cyclic homology with 2-periodic equivariant derived de Rham cohomology. Over C a theorem of Bhatt [Bha12] gives an identi cation with the Betti cohomology of the analyti cation. We mention a few related results in the literature. An analogous theory was explored by Block and Getzler in [BG94] in the setting of a compact group G acting on a compact smooth manifold M . Similar results appear in the algebraic setting when G is nite (i.e. X{G is a Deligne-Mumford stack) in Theorem 1.15 of [ACH14], and in the case of an smooth ane quotient stack in Lemma 4.11 and Proposition 4.12 of [HP16]. We generalize their statements to the case of a general smooth quasi-projective quotient stack. Our statements have also been investigated in the setting of smooth quotient stacks with nitely many orbits in [BN13a], with special attention to the case BzG{B in Theorem 3.5 of op. cit. We will now begin stating our results precisely. The following theorem is an abridged version of our geometric equivariant localization theorem for derived loop spaces, which appears in the main text as Theorem 3.1.12. Theorem A (Equivariant localization for derived loop spaces). Let G be a reductive group acting on a smooth variety X over an algebraically closed eld k of characteristic 0, and z P G a semisimple element. Note that LpX{Gq naturally lives over LpBGq; let L pX{Gq denote the completion of LpX{Gq along the ber over the semisimple orbit tG  zu{G € LpBGq and let L pX{Gq denote the completion over the saturation This depends on de nitions. Our notion of Hochschild homology is the one that is computed by a cyclic bar complex of a small dg category. Other authors have considered a variant which is obtained by extending the notion of Hochschild homology of ane schemes to stacks via at descent, which is smooth local by de nition and is equivalent to the a completed version of the former. 2 1 rzs P G{{G. For X a smooth variety with a G-action, there are functorial S -equivariant isomorphisms z z u z z u p p L p pX q{G q L pX{Gq; L p pX q{G q L pX{Gq: z 0 z 0 z z We remark brie y on the assumptions of the theorem. We assume that k is algebraically closed and that z is semisimple in reductive G since we argue via the Luna slice theorem, and we work in characteristic 0 since we model derived schemes locally as dg algebras. The assumption that X is smooth is used to compute via Koszul resolutions, and is essential since the statement is false otherwise. We note that in the case that G is a torus, these statements can be strengthened: the above maps are equivalences on a Zariski open neighborhood over z, recovering the equivariant localization in [CG10] for K -theory in the setting of periodic cyclic homology. However, for a nonabelian reductive group G, this fails even in the case when X is a point (see Remark 3.1.14 and Example 3.1.3). Our result implies the following interpretation of derived xed points, which also appears in [ACH14] as Corollary 1.12. Corollary 1.0.1. Let G be a reductive group acting on a smooth variety X and z P G semisimple. We have a natural identi cation of the derived z- xed points: z z Lp pX qq X : LpX{Gq  tzu: 0 BG z z 1 Proof. The \shift by z" map on Lp pX q{G qq is a (non-S -equivariant!) equivalence, so that z z z p p L p pX q{Gqq  Lp pX q{G q; z 0 0 z z and in particular since loop spaces commuted with ber products, X  Lp pX qq. The unipotent version of equivariant localization for loop spaces implies an equivariant localization result for Hochschild homology and its cyclic variants. The following appears in the main text as Theorems 3.2.3 and 3.2.10. Corollary 1.0.2 (Equivariant localization for Hochschild homology). Let G be reductive group acting on a smooth variety X , and z P G a semisimple element. Then we have an S -equivariant equivalence z z HHpPerfpX{Gqq HHpPerfp pX q{G qq zp 0 zp and similarly when replacing HH with its cyclic variants HC; HN; and HP . Note that since the formation of periodic cyclic homology involves a colimit, this is not automatic in that case. The fact that X is smooth gives us a cohomological boundedness of Hochschild homology, which is essential in establishing the above result. After identifying the completed derived loop spaces over a central character z, we are interested in identifying this completion with de Rham cohomology via an analogue of the Atiyah-Segal completion theorem in the setting of periodic cyclic homology. The following theorem is a consequence of Proposition 4.2.6 and Theorem 4.3.1. Theorem B (Atiyah-Segal completion for periodic cyclic homology). Let X{G be a global quotient stack where X is an algebraic space. Then, there is an equivalence HPpPerfpX{Gqq C pX{G; kqb kppuqq ep dR k where C pX{G; kq denotes the derived de Rham cohomology, and b indicates completion with respect to the dR coarsest topology induced by the derived Hodge ltration and u-adic ltration on respective tensor factors. 3 When k  C, applying a generalization of the main theorem of [Bha12] to geometric stacks, we can identify Tate functions on formal loop spaces with Betti cohomology. Corollary 1.0.3. Let X{G be a nite type global quotient stack over k  C, where X is an algebraic space. Then, there is an equivalence an an HPpPerfpX{Gqq C pX {G ;Cqb Cppuqq ep dR an an where C pX {G ;Cq denotes de Rham cohomology of the analyti cation and b indicates completion with dR respect to the coarsest topology induced by the Hodge ltration in the Cartan model for equivariant cohomology and u-adic ltration on respective tensor factors. The main technical hurdle in the proof of the theorem is that the ber of LpX{Gq over res P G{{G does not just contain formal loops but also unipotent loops. In Theorem 4.3.1 we show that this di erence vanishes after applying the Tate construction. Theorem C. For X a quasicompact algebraic space acted on by an ane algebraic group G, the pullback functor on derived global functions induces an isomorphism u Tate Tate OpL pX{Gqq OpLpX{Gqq : In particular, if U is a unipotent group, then the pullback functor induces an equivalence HPpPerfpX{Uqq HPpPerfpXqq: Following an analysis of the twisted S -rotation action on LpX{Gq above points of G{G away from the identity in Section 4.1, we relate this completion theorem to our localization theorem in Theorem 4.3.2. Theorem D. Let G be a reductive group acting on a smooth quasi-projective variety X . The periodic cyclic homology HPpPerfpX{Gqq is naturally a module over HPpPerfpBGqq  krG{{Gsppuqq. For a closed point z P G{{G, we have an identi cation of the formal completion at z with 2-periodic Betti cohomology of the z- xed points z z HPpPerfpX{Gqq C p pX q{G ; kqb kppuqq zp k dR as a module over HPpPerfpBGqq  C pBG ; kqb kppuqq, contravariantly functorial with respect to X . zp k dR Using Corollary 1.0.1, we also obtain an identi cation of the (derived) specialization of periodic cyclic homology at z P G with non-equivariant cohomology of z- xed points. Corollary 1.0.4. Let X be a smooth variety with an action of a reductive group G. For z P G semisimple, let k denote the skyscraper sheaf at rzs P G{{G. We have an equivalence L z HPpPerfpX{Gqq b k ppuqq C p pX q; kqppuqq: z 0 krG{Gsppuqq dR We indicate two natural directions in which our results may be extended. The rst is to ask what happens when X is allowed to be singular; in this case, one can study either the Hochschild homology of CohpXq or PerfpXq. In the case when X is a ber product of smooth schemes, the geometric statement follows immediately from our results, but it is unclear to us how to generalize beyond that case. If a geometric statement is out of reach, it is also of interest as to whether the global localization statements for periodic cyclic homology hold; an obstruction to applying the standard techniques of embedding a singular quotient stack into a smooth one is the lack of a devissage theorem for the periodic cyclic homology of stacks. A 4 second direction would be to categorify these results in a generalization of the Koszul duality of [Pre15] and [BN12]. In addition, it would be pleasing to have a more conceptual proof of Theorem 4.3.1. We end our introduction with a few toy examples. 1 1 Example 1.0.5. Let G  G  Spec krz; z s act on X  A  Spec krxs by scaling, i.e. assign G -weight m m |x|  1. The loop space can be calculated directly via Proposition 2.1.8 Spec krz; z ; xs{xxpz  1qy LpA {G q  : The Hochschild homology and periodic cyclic homology can also be calculated directly 1 1 G 1 1 1 HHpPerfpA {G qq  pkrz; z ; xs{xpz  1qq  krz; z s; HPpPerfpA {G qq  krz; z sppuqq m m as the S -equivariant structure on a complex concentrated in a single cohomological degree can only be realized by the zero map. Completing at any z P G gives, for t  z  z and |t|  0, 0 m 0 HPpPerfpA {G qq  krrtssppuqq m zp where t  z  z . z an On the other hand, we can compute H ppX q ; kqppuqq for each z . For z  z  1, the xed points 0 0 z z 1 1 ! ! 0 0 pX q{G  t0u{G  BS , whose 2-periodic cohomology is H pBS ; kq b kppuqq  krrsss b kppuqq 0 m 1 1 1 with |s|  2. For z  1 the xed points are A {G  C{S  BS , and the same argument applies. The identi cation krrsssppuqq  krrtssppuqq is by tu  s; in particular it is necessary to invert the degree 2 operator u. The discrepancy between the cohomological degrees of t and s is a manifestation of the Koszul duality degree-weight shearing discussed in [BN12]. Example 1.0.6 (Flag variety). Let X  G{B be the ag variety with the usual action of G. Then, X{G  BB, so LpX{Gq  B{B  G{G is the Grothendieck-Springer resolution; the ber for the map G Ñ G over any point g P G consists of the Borel subgroups containing g, i.e. the g- xed points of G{B. We identify the Hochschild homology HHpPerfpBBqq as a HHpPerfpBGqq-module by the inclusion map HHpPerfpBBqq  OpG{Gq  krHs Ð HHpPerfpBGqq  OpG{Gq  krHs where H is the universal Cartan subgroup and W is the universal Weyl group . Let s P G be a semisimple element, and rss its adjoint orbit. Completing at rss P krHs , we have W s HHpPerfpBBqq  krrhss Ð HPpPerfpBGqq  krrhss : x sp rss |Ws| In particular, the rank of HPpPerfpBBqq over HPpPerfpBGqq is |W  s|  |W s|  |W| by a theorem sp sp G of Steinberg and Pittie [Ste75]. Note that h is placed in cohomological degree zero. Applying the Tate construction, we nd that W s HPpPerfpBBqq  krrhss ppuqq: rss On the other hand, the xed points pG{Bq consist of Borel subgroups containing s; by conjugating, we s s can choose a torus such that s P T € B; let t be the Lie algebra of T . There is a G -action on pG{Bq and s s its stabilizer at every point is conjugate to B , but the action may not be transitive; thus, pG{Bq is the s s disjoint union of copies of G {B . To count the number of connected components, we count T - xed points: s s the T - xed points of G{B are also s- xed points, and furthermore each G {B contains |W | such T - xed It is known that OpGq has vanishing higher cohomology. 5 points, so we have |W|{|W s| connected components. Finally, accounting for T -equivariance, we have s an s an W s H p BpB q ; kq  krrhss Ð H pBpG q ; kq  krrtss |W|{|Ws| |W|{|W | where t is placed in cohomological degree 2. In particular, W s W s s an ! G G HPpPerfpBBqq  krrhss ppuqq  krrtss ppuqq  H ppBB q ; kq b kppuqq rss under the isomorphism uh  t. 1.1 Conventions and notation In this note, k will denote an algebraically closed eld of characteristic zero, and we work over pt  Specpkq. Unless otherwise stated, all functors and categories are derived, e.g. for an ane scheme X  SpecpAq, we denote by QCohpXq the derived category of unbounded complexes of A-modules localized with respect to quasi-isomorphisms, and b  b although we sometimes use the latter notation for emphasis (e.g. when performing calculations). We indicate a functor that is not derived by writing  or H . All gradings follow cohomological grading conventions (i.e. di erentials increase degree), unless otherwise indicated by a subscript, and HH will always denote the cochain complex of Hochschild chains rather than its cohomology groups, which we denote H pHHq. We refer to the nth cohomology group of a chain complex V by H pV q   pV q. We require a theory of 8-categories and derived algebraic geometry. Such theory has been developed by by To en and Vezzosi in [TV05] [TV08] and by Lurie in [Lur17] [Lur09b] [Lur18] [Lur09a] [Lur11a] [Lur11b] [Lur11c]. Below, we summarize some of the main de nitions. Remark 1.1.1 (8-categories). By 8-category we mean an p8; 1q-category, and we do not specify a particular model . We let S denote the 8-category of 8-groupoids or spaces and we will take for granted that the category of 8-categories is enriched in S. We let st denote the 8-category of small (stable) k-linear 8-categories whose 1-morphisms are k-linear exact functors, and Pr the category of presentable (stable) k-linear 8-categories whose 1-morphisms are functors which are k-linear exact left adjoints . Note that presentable 8-categories admit a combinatorial model structure. For such a category C P Pr , we let C P st denote its compact objects. For C P st , we let k k L L IndpCq P Pr denote its ind-completion. By [Coh16], a presentable k-linear 8-category in Pr has an k k L R associated k-linear di erential graded category in dgcat . We will denote by Fun p;q and Fun p;q k k the spaces of k-linear exact functors which are left and right adjoints respectively. For more details, see Chapter 5 of [Lur09b], Section 1.4.4 in [Lur17], and Section 6 of [Lur11a]. Remark 1.1.2 (Derived stacks). We let DRng denote the 8-category of derived rings (or derived algebras over k); during our exposition we do not insist on a particular model, but we will always compute in the op category of dg algebras over k with its projective model structure. The opposite category A  DRng is de ned to be the category of ane derived schemes. A derived scheme is as a derived locally ringed space whose 0-truncation is a scheme and whose higher homotopy groups are quasicoherent [To e14] [Lur18]. The global sections functor and derived spectrum functors induce equivalences identifying A with the category of derived schemes whose  is ane in the classical sense. We will refer to derived schemes as simply schemes, and use the term classical scheme to refer to a derived scheme X for which  pXq  X . A dg scheme [CK01] or embeddable derived scheme is de ned somewhat di erently; it is de ned to be a 0 n scheme pX;O q along with a non-positively graded sheaf of complexes O such that O  O and H pO q X X X X X A forthcoming book by Riehl and Verity [RV18] establishes the model-independence of p8; 1q-categories and its foundational properties, constructions, and theorems. In particular, by Remark 6.5 in [Lur11a], k-linear presentable categories are automatically stable. By the adjoint functor theorem, left adjoint functors commute with ltered colimits. are quasicoherent. In particular, a dg scheme Z  pX;O q admits an embedding Z Ñ X into the classical scheme X  pX;O q. Every derived scheme is locally modeled by a dg scheme. op A prestack is an 8-functor A : DRng Ñ S, and a (derived) stack is a prestack which is a sheaf for the derived  etale topology [GR17] [TV08]. We mean (derived) algebraic stack in the sense of [DG13]: an (derived) Artin 1-stack whose diagonal is quasi-separated, quasi-compact, and representable by (derived) schemes and admits an atlas by a (derived) scheme. We mean geometric stack in the sense of [BN12]: an algebraic stack whose diagonal map is ane. We say an algebraic stack is quasi-compact if it admits a quasi-compact atlas U (equivalently, if it admits an ane atlas). A map of prestacks X Ñ Y is schematic if for any scheme S and map S Ñ Y , the base change X  S to S is a scheme. A map of derived schemes f : X Ñ Y is a closed immersion if it is in the classical sense on  ; a map of algebraic stacks is a closed immersion if it is after base change to an atlas. 1.2 Acknowledgements I would like to thank David Nadler for our numerous conversations regarding this topic and for suggestions leading to and during the writing of this paper. I would also like to thank Daniel Halpern-Leistner for suggestions regarding the equivariant cyclic bar complex, David Ben-Zvi for suggestions regarding twisted circle actions and Bhargav Bhatt for explaining how his results in [Bha12] extend to the case of stacks. Finally, I would like to thank Shishir Agrawal for numerous helpful discussions, and Brian Hwang for his comments on this paper. This work was mostly carried out at UC Berkeley, in part at Cornell University, and was partially supported by NSF RTG grant DMS-1646385. 2 Background In this section we provide some basic exposition on Hochschild homology, loop spaces and derived algebraic geometry. At parts it is an informal summary of the existing literature, and at parts we provide proofs of some folklore likely known to experts. 2.1 Derived loop spaces and its variants An in-depth discussion of derived loop spaces, which we often simply refer to as loop spaces, can be found in [BN12]. We will summarize the main de nitions and prove some foundational results in the case of the loop space of an algebraic or geometric stack. These should probably be skipped on a rst reading; the essential statements for the main body of the paper are in Propositions 2.1.20 and 2.1.25, which provide an explicit description of the formal and unipotent loops of a global quotient stack. De nition 2.1.1. We consider the higher derived stack S as the locally constant sheaf on A with value 1 1 the topological circle S . Its anization is the shifted ane line BG  Spec C pS ; kq and the map S  BZ Ñ BG is induced by the map of abelian groups Z Ñ G . a a Remark 2.1.2. The stack BG is not an ane scheme since C pBS ; kq is not connective, but it still has a well-de ned functor of points; it is an example of a coane stack (in the language of [Lur11b]) or an ane stack (in the language of [To e06]). Explicitly, by Lemma 2.2.5 in [To e06] or the introduction to Section 4 of 5 6 [Lur11b], it is the right Kan extension of the classical stack sending an ane scheme S  SpecpRq to the Eilenberg-Maclane space KpR; 1q where R is considered as an abelian group under addition. The anization map S Ñ BG is given on S-points by the map of Eilenberg-Maclane spaces Kp1;Zq Ñ Kp1; Sq where we consider S as an abelian group under addition. We x an isomorphism C pS ; kq  krs where ||  1. That is, the (fully faithful) left adjoint to the restriction of a prestack (i.e. a functor DRng Ñ S) to a classical prestack (i.e. a functor Rng Ñ S). In fact, coane stacks are always left Kan extensions of classical stacks. 7 De nition 2.1.3. Let X be a prestack. We de ne the derived loop space and its variants as follows. The (derived) loop space of X is the derived mapping stack LpXq : MappS ; Xq  X  X: XX 1 1 The second presentation is a consequence of the presentation of S via the homotopy pushout S pt pt  S , and the property that derived mapping stacks take coproducts in the source to products. The evaluation map p : LX Ñ X realizes the loop space as a relative group stack over X . The derived loop space has a canonical S -action by loop rotation. The formal loop space LpXq is the completion of LpXq along constant loops X Ñ LX . It inherits a loop rotation S -action from LX . The unipotent loop space L pXq is the derived mapping stack L pXq : MappBG ; Xq 1 u and the anization map S Ñ BG de nes a map L X Ñ LX . There is a natural BG G -action a a m u 1 on L X arising from the natural G -action on BG , compatible with the S -action on LX . m a If X admits a cotangent complex, we de ne the odd tangent bundle, a linearized form of the loop space, by T r1s : Spec Sym L r1s X X X X i.e. the relative spectrum of the derived symmetric powers of the cotangent complex. There is a projection q : T r1s Ñ X and a zero section c : X Ñ T r1s induced by the structure and X X augmentation maps respectively. We write T r1s for the odd tangent bundle completed at its zero section. Both T r1s and T r1s are equipped with a natural BG G -action, where the BG -action X X a m a is encoded by the de Rham di erential and the G -action is by scaling on the bers. De nition 2.1.4. Let X be a quasicompact geometric stack. There is an exponential map p p exp : T r1s Ñ LX de ned in Section 6 of [BN12]. In particular, LX has a natural BG G -action compatible with the a m S -action. Theorem 2.1.5 (Hochschild-Kostant-Rosenberg). The exponential map is an equivalence. Proof. For the stacky case, see Section 6 of [BN12]; when X is a derived scheme, see the main theorem of [BF08]. Example 2.1.6. If X  SpecpAq, then the derived loop space op LpXq  SpecpA b Aq  SpecpC pA; Aqq Ab A is the derived spectrum of the cyclic bar complex equipped with the shue product. The rotation S -action has a combinatorial realization via the cyclic structure on the cyclic bar complex [Lod92] [Jon87]. In this op example, we think of the bar resolution B pAq Ñ A as the A b A -module obtained by tensoring A with We de ne the relative spectrum as follows: for an algebra object A P QCohpXq, we de ne the S-points for Spec A as pairs p; q where  P XpSq and  : S Ñ Spec  A which are compatible under the projection; note that  A is an algebra since S is an ane derived scheme and pullback preserves the monoidal structure on quasicoherent sheaves. The symmetric algebra functor Sym is left adjoint to the forgetful functor from the category of augmented commutative unital associative algebra objects of QCohpXq, which exists by the adjoint functor theorem. 8 the map of simplicial complexes I Ñ pt, where the unit interval I is presented by a simplicial set with two 0-simplices and one non-degenerate 1-simplex: B pAq  A b I Ñ A X  Map pI; Xq  SpecpA b Iq: DSt op The cyclic bar complex C pAq  B pAq b A is obtained by gluing the two 0-simplices of I , i.e. it is AbA the chain complex associated to the tensor product of A with the presentation of S by one 0-simplex and one non-degenerate 1-simplex: 1 1 1 C pAq  A b S LpXq  Map pS ; SpecpAqq  SpecpA b S q: DSt This makes LX into a cocyclic scheme and OpLXq into a cyclic algebra. Example 2.1.7. If X is a stack, then  pLXq is the (classical) inertia stack of X , so LX can be thought of as a derived inertia stack. In particular, let X  BG; then LpBGq  G{G is the stacky adjoint quotient (see Proposition 2.1.8 below). Note that LpBGq  BG  BG is classical since the diagonal map is at. BGBG The S -equivariant structure on OpG{Gq has a description in terms of the a cyclic algebra arising from the cyclic structure on the simplicial Cech nerve for the atlas G Ñ G{G (see Section 7.3.3 of [Lod92]). Proposition 2.1.8 (Loop space of a quotient stack). The loop space of a quotient stack LpX{Gq can be computed by the G-equivariant ber product LpX{Gq pX  Gq{G ap X{G pX  Xq{G where G acts on X  X and X  G diagonally. Proof. Note that X{G  X{G  pX  Xq{pG  Gq with action pg ; g q  px ; x q  pg x ; g x q. We write 1 2 1 2 1 1 2 2 X X X  G X  G XX   XX GG GG G G G  G G  G where the map X  G Ñ X  X sends px; gq ÞÑ px; gxq and the action of G G on X  G is pg ; g qpx; gq 1 2 pg x; g gg q. The claim follows from the \two-out-of-three" lemma for pullback squares applied to 1 2 LpX{Gq pX  Gq{G pX  Gq{pG  Gq X{G pX  Xq{G pX  Xq{pG  Gq: Remark 2.1.9. Forgetting G-equivariance, the geometric points of the loop space LpX{Gq pt are given BG by pLpX{Gq  ptqpkq  tpx; gq P Xpkq  Gpkq | g  x  xu: BG The geometric ber of the map LpX{Gq Ñ LpBGq  G{G over g P Gpkq is the xed points X pkq. The geometric ber of the evaluation map LpX{Gq Ñ X{G over x P Xpkq is the stabilizer of x in Gpkq. Example 2.1.10 (Odd tangent bundle of smooth quotient stacks). In the case of X{G where X is smooth, we have that Ñ g bO X{G X 9 where the internal di erential d is the Cartan di erential: n n 1  n  n1 n Sym pL r1sq  Sym p Ñ g bO q  p Ñ g b Ñ  Ñ Sym pg q bO q; X{G X X X X X Sym g p O  lim b ris; d : T r1s ¥k X{G Sym g i¥k The resulting de Rham complex is called the Cartan model for equivariant cohomology; see Proposition 4.12 of [HP16] for more discussion. This example can also be carried out when X is not smooth, replacing with L . 2.1.1 Loop spaces of algebraic and geometric stacks We prove some technical facts which may be skipped on a rst reading. Note that a quasi-compact geometric stack is automatically QCA in the sense of [DG13]. Remark 2.1.11. The following principles are standard and will be used frequently. If X is an algebraic stack then X admits a cover by a disjoint union of ane schemes; if X is quasi-compact this disjoint union can be taken to be nite, so that X admits a cover by an ane scheme. If X is geometric (i.e. has ane diagonal), then S  T  pS  Tq  X X XX is ane for any ane schemes S; T . Lemma 2.1.12. Let X be an algebraic stack. Then LX is an algebraic stack. If X is geometric, then LX is geometric. If X is geometric and quasi-compact, then so is LX . Proof. Assume X is algebraic. That LX is algebraic follows from the fact that LX  MappS ; Xq is a nite limit, and any nite limit of algebraic stacks is algebraic. An algebraic stack X is geometric if and only if for any map from an ane U Ñ X , the stack U  U is an ane scheme. In particular, U  X is a cover X XX for LX , and we have pU  Xq  pU  Xq  U  pLX  pU  Xqq  pU  Uq  X XX LX XX X LX XX X XX which is ane since U  U is ane and the diagonal map is ane, so LX is geometric. Assume X is also quasi-compact; then it admits a cover by an ane U , and U  X is also ane since the diagonal is XX ane. Lemma 2.1.13. Let X be an algebraic stack. Then the inclusion of constant loops X Ñ LpXq is a (schematic) closed immersion. Proof. Since the diagonal map X Ñ XX is representable by schemes, so is the evaluation map LpXq Ñ X . Let U Ñ X be an atlas for X with U a scheme; its base change along the evaluation map gives a cover by a scheme U  X Ñ LX . In particular, by the two-out-of-three property of Cartesian squares, the left XX square is Cartesian U U  X U XX X LX X i.e. the base change of the inclusion of constant loops X Ñ LX along an atlas is a scheme, so it is schematic. It is a closed embedding since any map of derived schemes which admits a retract is a closed embedding, and U Ñ U  X admits a retract by universal property. XX We provide a proof for the last claim. It suces to assume all schemes are classical, since the the property of being a closed immersion depends only on classical schemes and the property of admitting a retract is 10 preserved by  . Let f : Z Ñ Y be a map of schemes admitting a retract. We can verify that f is a closed immersion ane locally on Y , so assume Y is ane. It is a closed immersion if f : O Ñ f O is surjective. Y  Z Since Y is ane, this is equivalent to OpY q Ñ OpZq being surjective, which follows since the composition on global functions OpZq Ñ OpY q Ñ OpZq is the identity. We now introduce the notion of based loops of a stack. Namely, given a point of a stack, the based loop space is the group of automorphisms of that point and the unipotent based loops consist of the unipotent automorphisms. We use these characterizations in Propositions 2.1.20 and 2.1.25 to give explicit descriptions of the formal and unipotent loop spaces of quotient stacks. De nition 2.1.14. Let X be a prestack, and x : S Ñ X be an S-point where S is an ane derived scheme. The group of based loops at x, which we denote pX; xq, is the 8-group object in prestacks over S de ned to be the Cech nerve of the map x : S Ñ X , i.e. its underlying derived stack is LX  S or equivalently the pullback pX; xq S S X: If f : X Ñ Y is a map of prestacks, with x P XpSq, then there is a natural map of 8-groups pf; xq : pX; xq Ñ pY; fpxqq. We de ne the unipotent based loops of X , denoted pX; xq, by the ber product u u L pXq  S; there is a natural map pX; xq Ñ pX; xq. Note that the unipotent based loops do not form a group. Remark 2.1.15. If X is algebraic, then based loops at x P XpSq form an 8-group object in derived schemes over S, and if X is geometric, the based loops form an 8-group object in ane derived schemes over S. Example 2.1.16. Let X be an (ane) derived scheme and x P Xpkq a geometric point. Then, pX; xq T r1s  Spec Sym x L r1s is the odd tangent space at x P Xpkq. The comultiplication on functions is X;x X given by the natural comultiplication on the symmetric algebra and antipode map by the sign morphism. Lemma 2.1.17. For any S-point x P XpSq, we have a natural identi cation of prestacks over S pX; xq  Hom pG  S; pX; xqq grp;S a where the natural map pX; xq Ñ pX; xq is identi ed with evaluation at 1 P G . 1 u 1 Proof. Let f : S Ñ S and P pX; xqpS q with base point f x. Note that 1  1 1 pX; xq  S pX; f xq  CechpS Ñ X  S q: S S u 1 We will also use to denote its image in pL XqpS q. Let  : pt Ñ BG denote the inclusion of the (additive) 0 a identity. The diagram id 1 1 S S id 1 x 1 1 BG  S X  S a S commutes, and therefore we have a map of the corresponding Cech nerves as 8-groupoids, giving us a map u 1 1 1 pX; xqpS q Ñ Hom 1pG  S ; pX; xq  S q grp;S a S The notion of an 8-group can be found in De ntion 7.2.2.1 of [Lur09b], a useful characterization in Proposition 7.2.2.4 of op. cit. . Proposition 6.1.2.11 of op. cit. shows that Cech nerves are 8-groupoid objects in any 8-category, so that pX; xq is a groupoid object in derived stacks and therefore a group object in derived stacks over S. We use the notation Hom to denote grp the space of group homomorphisms. 11 1 for every S , natural in f , inducing a map of stacks pX; xq Ñ Hom pG  S; pX; xqq: grp;S a We now produce the inverse to this map. To do so, we need to produce a map Hom pG grp;S a 1  u 1 u 1 S ; pX; f xqq Ñ pX; xqpSq natural in f : S Ñ S. First, note that pX; xqpSq  MappBG S ; Xq XpSq MappS ; Sq, so we only need to de ne a map 1  1 Hom pG  S ; pX; f xqq Ñ MappBG  S ; Xq: grp;S a a Taking geometric realizations of the 8-groups over S considered as simplicial objects in derived stack, we obtain a map BG S Ñ B pX; f xq. We compose with the map B pX; f xq Ñ X induced by universal a S S property of geometric realizations applied to the augmentation map f x : S Ñ X . We leave the veri cation that these two maps are inverses to the reader, and naturality with respect to evaluation at 1 P G (essentially since S  BZ Ñ BG is induced by the inclusion of 1 P G ). a a The following notion of a contracting action will be used in Section 4.3. De nition 2.1.18. Let X  SpecpRq be an ane scheme with a G -action. We say the G -action is m m contracting if it acts by only non-positive weights on R. In this case, the xed point locus is Y  SpecpR q, and we say the G -action contracts to Y . In particular, there are maps Y Ñ X Ñ Y . More generally, let X be a prestack with a G -action, equipped with a G -equivariant ane map p : X Ñ Y where Y m m is given the trivial action. We say the G -action contracts to Y if for any ane S and map S Ñ Y , the induced G -action on S  X contracts to S. In particular, this implies there is also a G -equivariant m Y m section Y Ñ X . Lemma 2.1.19. Let X be a quasi-compact geometric stack. The G -actions on L X and T r1s contract m X to the xed point locus of constant loops. Proof. The claim for T r1s is by de nition. For the unipotent loop space, take x P XpSq. It suces to u u show that the induced G -action on pX; xq is contracting. This follows from the description of pX; xq Hom pG ; pX; xqq, and the contracting G -action on G . grp;S a m a 2.1.2 Formal and unipotent loops over schematic maps In [BN12], it is shown that for X a scheme, LpXq  L pXq  LpXq. This is not true for stacks, but we will now show that for a schematic map f : X Ñ Y , the formal and unipotent loops of X are loops in X whose images in Y are formal and unipotent respectively. Proposition 2.1.20. Suppose that f : X Ñ Y is a map of algebraic stacks representable by schemes. Then, p p LpXq  LpY q  LpXq: LpY q Proof. It suces to show that the closed classical substack  pY  LpXqq has the same reduced points as LpY q X € LpXq. To do this, it suces to check on geometric points. Consider the diagram of classical pullbacks pX; xq pY; fpxqq Spec k Spec k f pfpxqq X fpxq Spec k Y: 12 1 Since f is schematic, x : Spec k Ñ f pfpxqq is a closed embedding of schemes (since it as a map of schemes admitting a retract; see Lemma 2.1.13). Since Y is an algebraic stack, pY; fpxqqq is a scheme, and so pX; xq Ñ pY; fpxqq is a closed embedding of schemes and a map of ane (classical) group schemes. The preimage of the identity is thus the identity, so constant loops in LY are preimages of constant loops in LX . Example 2.1.21 (Quotient stacks). In the case of quotient stacks, we have a map LpX{Gq Ñ LpBGq  G{G. The above proposition says that LpX{Gq is the completion of LpX{Gq at the closed substack of points lying over teu{G € G{G. The following is well-known, but we provide a brief argument for the reader's convenience. Proposition 2.1.22. Let H; G be groups in ane derived schemes over k. There is a natural identi cation MappBH; BGq  Hom pH; Gq{G grp where G acts on Hom pH; Gq by the adjoint action. grp Proof. Note that we assume k has characteristic zero. We de ne maps  : MappBH; BGq Ñ Hom pH; Gq{G grp and : Hom pH; Gq{G Ñ MappBH; BGq and leave the veri cation that they are strict inverses to the grp reader. These are maps of sheaves; we will restrict our attention to de ning their map on S-points  and The map  of spaces is de ned as follows. Let F P Map pBH S; BGSq be the map of sheaves whose 1 1 1 value at S Ñ S is a functor F from (right) H -torsors over S to G-torsors over S . We de ne $ , '  / ' 1 / ' F 1pH  S q Hom pH  S; G  Sq / S Sgrp & . pFqpS q Gtorsor ' / ' / ' / % - where  is de ned as follows. There is a canonical identi cation of automorphisms of the trivial torsor 1 1 1 Aut pS  Hq  HpS q, and for h P Aut pS  Hq, Fphq is an automorphism of FpS  Gq, which we S S abusively write Fphq P G as an section in GpS q. We de ne pxqphq  x  Fphq. The map is de ned as follows. The S-points of Hom pH; Gq{G are G-torsors P over S with G- grp 1 1 equivariant maps  : P Ñ Hom pH; Gq. We de ne for S Ñ S and Q a H -torsor over S grp 1 H 1 pP; qpS qpQq  Q  pP  S q S S where H acts on P on the left via . De nition 2.1.23. A map of prestacks X Ñ Y is a monomorphism, i.e. X is a substack of Y , if for any ane derived scheme S and y P Y pSq the ber product tyu  XpSq is contractible (in the category of Y pSq spaces). Proposition 2.1.24. Let X be a geometric stack. The map L X Ñ LX is a monomorphism, i.e. unipotence of a loop is a property and not a structure. Proof. Let S  SpecpRq. Consider an S-point P pLXqpSq, which determines a base point x P XpSq and a based loop g P pX; xqpSq. We wish to show that L XpSq  tp qu is contractible. Equivalently, LXpSq we wish to show that pX; xqpSq  tgu is contractible. Since X is geometric, pX; xq is derived pX;xqpSq ane. By Lemma 2.1.17 it suces to show that for a derived ane group G over S, the map of stacks Hom pG  S; Gq Ñ G is a monomorphism on S-points. Note that an 8-group object is by de nition an grp;S a 8-monoid object satisfying a condition; in particular, the map Hom pG S; Gq Ñ Hom pG S; Gq grp;S a mon;S a 13 is a monomorphism. Furthermore, the forgetful functor from CoalgpAlgpRqq to CoalgpR -modq  CoalgpRq is fully faithful, so the map Hom pG  S; Gq Ñ Hom pp O ;O q is a monomorphism. mon;S a CoalgpRq pX;xq SG It suces to show that the map induced by evaluation at 1 P G : Hom pp O ;O q  Hom pRr pX; xqs; Rrxsq Ñ Hom pRr pX; xqs; Rq SG R CoalgpRq pX;xq a CoalgpR -modq is also a monomorphism. By the calculation in Lemma 1.12 of [GG99], O b OpG q is the cofree coalgebra S k a object in CoalgpQCohpSqq, so this map is an equivalence. Proposition 2.1.25 (Unipotent loops of quotient stacks). Let G be a classical ane algebraic group over k. Then, L pBGq  U{G where U is the unipotent cone of G (i.e. the closed subvariety of unipotent elements of G). Furthermore, if G acts on a scheme X , then L pX{Gq is computed by the pullback square L pX{Gq LpX{Gq U{G G{G: Proof. It suces to prove the statement on based loops at a given x P pX{GqpSq where S be an ane derived scheme, say S  SpecpRq. The based loops in X{G for any S-point x P XpSq can be computed via a ber product pX{G; xq pBG; ppxqq  G  S S X: Let U denote the unipotent cone of G, and let U denote the closed reduced classical subscheme of pX{G;xq the inverse image of U  S € G  S. Let U the formal completion of pX{G; xq at U . We pX{G;xq pX{G;xq rst claim that pX{G; xq  U . pX{G;xq Using Proposition 2.1.17, evaluation at 1 P G provides a map  : pX{G; xq Ñ pX{G; xq, which is a monomorphism by Proposition 2.1.24. Such a map factors through U if its (classical) set-theoretic pX{G;xq image lies in the inverse image of the unipotent cone U  S. This, we can check on k-points of S, and in particular assume that S  k. The claim now follows from the classical result that for a map of linear algebraic groups over k, the image of a unipotent element must also be unipotent (i.e. its eigenvalues are all 1 P k). In particular, pX{G; xq € U € pX{G; xq. pX{G;xq For surjectivity, we de ne the inverse map : U Ñ pX{G; xq  Hom pG ; pX{G; xqq pX{G;xq grp;S a;S via the adjoint to an exponential map G  U Ñ pX{G; xq which we will now construct. We a;S S pX{G;xq take as a given that such an exponential map is constructed for classical ane algebraic groups over k, i.e. we have a map G  U Ñ G. To de ne an exponential map for pX{G; xq, we use the universal property a G of ber products and the classical fact that if an S-point g P GpSq xes x P XpSq, then so does g P GpSq for t P G pSq. More precisely, the following diagram commutes, inducing the desired exponential map p p G  U G  U a;S S a;S G pX{G;xq exp pX{G; xq G  S S X: 14 2.2 Cyclic homology In this section we give a brief overview of the basic de nitions of Hochschild homology and cyclic homology, as developed in Section 5.5 of [Lur17]. Further discussion can be found in [BN13b] [BN13b] [Hoy18] [AMR17] [AF17] [AFR18] [NS18]. De nition 2.2.1. Let Cat be a symmetric monoidal 8-category with monoidal unit 1 , and X P Cat b b b _ _ _ a 1-dualizable object with dual X , coevaluation  : 1 Ñ X b X and evaluation  : X b X Ñ 1 . We b b de ne the dimension of X by dimpXq     P End p1 q: Cat b If F : X Ñ Y is a morphism with a right dual (i.e. adjoint) G, then we can de ne dimpFq : dimpXq trpG  Fq trpF  Gq dimpY q: Remark 2.2.2. More precisely, the a choice of dualizing structure for X P Cat determines an explicit dimension dimpXq. By Proposition 4.6.1.10 in [Lur17], the space of dualizing structures on X is contractible; therefore, dimpXq is de ned uniquely up to unique isomorphism in the homotopy category HopEnd p1 qq. Cat b This fact allows us to compute Hochschild homology in two di erent ways using di erent dualizing structures on a category and know that they are equivalent without explicitly producing an equivalence. Remark 2.2.3. Lurie's proof of the Cobordism Hypothesis [Lur09c] allows for an equivalent formulation: there is a bijection between 1-dualizable objects X P Cat and framed extended Cat -valued n-dimensional b b topological eld theories Z ; for a 1-dualizable object X we de ne the dimension by dimpXq  Z pS q: 1 1 By this de nition, there is evidently an S -action on dimpXq. The relationship between S -action and its explicit realization via the cyclic structure is spelled out in Theorem 5.5.3.11 in [Lur17]. We will de ne the Hochschild homology of a category to be its dimension; we rst need to de ne a monoidal structure on 8-categories. De nition 2.2.4 (Lurie tensor product). The category Pr is equipped with a monidal structure called the Lurie tensor product in Proposition 4.8.1.15 of [Lur17], which can be thought of as an 8-analogue of the Deligne tensor product. It is equipped with a canonical functor C  D Ñ C b D pX; Y q ÞÑ X b Y: which is initial amongst functors out of C  D which preserve small colimits separately in each variable. op Remark 2.2.5. Proposition 4.8.1.17 of [Lur17] provides an explicit realization CbD  Fun pC ; Dq which is presentable by Lemma 4.8.1.16 of [Lur17]. In particular, by [Lur09b] Proposition 5.5.3.8, the Lurie tensor L L product makes Pr into a closed monoidal category with internal mapping object Fun p;q. Furthermore, by Propositions 4.8.2.10 and 4.8.2.18 in [Lur17], the Lurie tensor product induces a tensor product on k-linear presentable categories Pr . L;! De nition 2.2.6. Let Pr be the 8-category of dualizable presentable stable k-linear 8-categories, and k;_ functors which preserve compact objects. We de ne the Hochschild homology functor by L;! L HH : dim : Pr Ñ Fun pVect ; Vect q  Vect : k k k k;_ By [AFR18], the Hochschild homology as de ned above has an S -action. 15 Remark 2.2.7. The dimension of a dualizable k-linear category takes values in chain complexes under the equivalence End p1 q  Fun pVect ; Vect q  Vect : Cat b k k k b k That is, every such endofunctor F commuting with colimits is determined by its value Fpkq. In particular, a choice of dualizing structure determines an explicit Hochschild chain complex. Remark 2.2.8. Note that the Lurie tensor product is de ned on Pr , and dimension is functorial only for right dualizable maps, so Hochschild homology is only functorial for functors whose right adjoints also admit right adjoints (equivalently, whose right adjoints are also continuous). By Proposition 5.5.7.2 of [Lur09b], these are exactly the functors which preserve compact objects. Traditionally, Hochschild homology is formulated in the setting of small stable k-linear dg categories. In particular, in this setting it is possible to explicitly write out a bar complex computing the Hochschild homology. We will see that the above \large" notion of Hochschild homology is a generalization of the \small" version. The following is proven in Theorem D.7.0.7 in [Lur18] and Proposition 4.6.15 in [Lur17]. Proposition 2.2.9. If C P Pr is compactly generated, then it is dualizable. In particular, if C  IndpC q, _ 0;op then C  IndpC q, and the evaluation map is given by ind-completion via universal properties of the 0;op Yoneda pairing Homp;q : C  C Ñ Vect . Furthermore, there are natural isomorphisms L L _ Fun pC; Cq  Fun pC; Vect q b C  C b C k k which realize the coevaluation via the identity functor in Fun pC; Cq. De nition 2.2.10. Let st be the 8-category of small stable k-linear 8-categories. We de ne the Hochschild homology functor HH : dim Ind : st Ñ Fun pVect ; Vect q  Vect k k k k i.e. HHpCq is the image of k under the composition coev L  ev Vect Fun pC; Cq C b C Vect : k k Remark 2.2.11. If C is compactly generated, then HHpCq  HHpC q, so the two de nitions of Hochschild homology given above are compatible. We now seek to understand the S -action on Hochschild homology. While there are purely categorical ways to view S actions, we model the concretely on chain complexes via a mixed di erential. De nition 2.2.12. A mixed complex is a dg-module over the dg-algebra H pS ; kq  krs where ||  1. Explicitly, it is a chain complex pV; dq with a mixed di erential  of cohomological degree 1 such that d  d and   0. We are interested in the following operations on mixed complexes. 1 1 De nition 2.2.13. We de ne the S -invariants and S -coinvariants functors by S 1 V : R Hom 1 pC pES ; kq; V q  pV rruss; d uq P krruss -mod; C pS ;kq 1 1 1 1 V : V b C pES ; kq  pV ru s; d uq P krruss -mod : S C pS ;kq utors and the Tate construction by Tate S k V  V b kppuqq  lim u V 1  pV ppuqq; d uq  pV ppuqq; d uq P kppuqq -mod : krruss S 16 1 1 Remark 2.2.14. The action of C pS ; kq on C pES ; kq comes from the sweep action of chains of [GKM98]. 1 1 1 Taking a presentation of ES as a colimit of odd spheres with free S -actions, the S -invariants can be expressed as a ltered limit S 2n1 n V  lim R Hom 1 pC pS ; kq; V q  limpV rus{u ; d uq: C pS ;kq n n Remark 2.2.15. Note that the S -invariants operation (as well as the Tate construction) is not continuous as de ned above , but can be made so by considering mixed complexes in the category Indpkrs -mod q f:g: instead. De nition 2.2.16. We respectively de ne the negative cyclic homology, cyclic homology and periodic cyclic homology by S Tate HNpCq : HHpCq ; HCpCq : HHpCq 1; HPpCq : HHpCq : In this note we consider two di erent explicit models of Hochschild homology and its cyclic variants. One is the usual cyclic bar construction on a small dg category, and the other is via S -equivariant functions on the derived loop space. Example 2.2.17 (Algebraic model). Let A be a dg algebra (or more generally, a dg category) over k, and op C  A -mod the category of left dg-modules over A. For two dg categories A; B, an A b B-module de nes a continuous functor A -mod Ñ B -mod. By the dg Morita theory of [To e07], this functor op L A b B -mod Ñ Fun pA -mod; B -modq op is an equivalence. Under this equvalence, the coevaluation k -mod Ñ AbA -mod corresponds to the functor Ab , where A is considered as a bimodule over itself, and the evaulation map corresponds to b op A. k AbA In particular, the Hochschild homology is given by the usual Hochschild homology op HHpA -modq  HHpA; Aq  A b A: AbA De nition 2.2.18 (Cyclic bar complex). The cyclic nerve of a small k-linear dg category C is the cyclic vector space whose n-simplices are given by C pCq  Hom pX ; X q b Hom pX ; X q b  b Hom pX ; X q n C 0 n C n n1 C 1 0 X ;:::;X PObpCq 0 n where the face maps are given by composition, the degeneracy maps by the identity homomorphism, and the cyclic structure by rotation of the terms. Its associated chain complex, which also abusively denote by C pCq, is the cyclic bar complex which is naturally a mixed complex with the mixed di erential arising via the Connes B-operator. This mixed di erential exhibits the S -action on Hochschild homology [Hoy18]. Remark 2.2.19. One can obtain smaller models by taking objects from a set of compact generators (see Theorem 5.2 of [Kel06]) rather than all of ObpCq; for example, if C  A -perf , then the free module A is a compact generator and one recovers the classical cyclic bar complex C pA; Aq. Example 2.2.20 (Geometric model). By [BFN10], when X is a perfect stack (e.g. a quotient stack of a derived quasiprojective scheme by an ane group in characteristic zero), then QCohpXq is compactly generated by PerfpXq and we have isomorphisms QCohpXq b QCohpXq  QCohpX  Xq  Fun pQCohpXq; QCohpXqq 1 1 9 S S 1 For example, consider N  colim kr2ns. Since N is acyclic, N  0. On the other hand, pkr2nsq  u krruss so that colimpkr2nsq  kppuqq. 17 where the functors on the right are given by integral transforms. Explicitly, we identify QCohpXq QCohpXq on compact objects K P PerfpXq by K ÞÑ pX;K b q  pX;Hom pK ;qq: Letting p : X Ñ Specpkq be the map to a point and  : X Ñ X  X the diagonal, the coevaluation is given by the functor  p and the evaluation by p  . In particular, we nd that the Hochschild homology is HHpPerfpXqq  p   p k  pX;   O q  pLX;O q  OpLXq X LX 1 1 with the last isomorphism arising via base change. The S -action on OpLXq is the S -equivariant structure arising from loop rotation; for details see Remark 3.2 and Proposition 4.2 in [BN13b]. Remark 2.2.21. If X is QCA but not perfect, it is not currently known whether QCohpXq is compactly generated. On the other hand, by Theorem 4.3.1 of [DG13], QCohpXq is dualizable, so that HHpQCohpXqq  OpLXq by a similar argument. It does not appear to be known whether HHpPerfpXqq agrees with HHpQCohpXqq. 2.3 The equivariant cyclic bar construction We now de ne an explicit model for the Hochschild homology of a quotient stack. Proposition 2.3.1. Let X be a quasiprojective scheme with an action of a reductive group G, and let p : X{G Ñ BG and q : X Ñ X{G be the natural maps of stacks. Let E P PerfpX{Gq be a locally free sheaf such that q E is a compact generator of QCohpXq. De ne A  p RHom pE;Eq P AlgpQCohpBGqq: Then, the functor R Hom pE;q : QCohpX{Gq Ñ A -mod QCohpBGq is an equivalence of dg categories. Proof. Since X is quasiprojective, it admits an equivariant compact generator E of PerfpXq (not PerfpX{Gq). The functor is fully faithful since q is an atlas and q E is a generator. It is essentially surjective, since A -perf is generated by A b V for V P IrrpGq, and R Hom pE;E b V q  A b V . QCohpBGq X{G We need the following formula for G-representations. Proposition 2.3.2. Let G be a reductive ane algebraic group, and V; W two rational representations of G (i.e. V; W P QCohpBGq). Then, there is a natural equivalence G G pV b U q b pU b Wq pV b Wq : UPIrrpGq More generally, for V ; : : : ; V rational representations of G, there is a natural equivalence 0 n G  G G pV b U q b pU b V b U q b  b pU b V q pV b  b V q : 0 1 1 n n 0 n 1 2 U ;:::;U PIrrpGq 1 n 18 Proof. Consider V b W P QCohpBG  BGq as a G  G-representation. The projection formula de nes an equivalence pBG  BG;  O b pV b Wqq Ñ pBG;  pV b Wqq: BG The result follows by unwinding this equivalence using the Peter-Weyl theorem for reductive algebraic groups O  krGs  U b U , and noting that taking global sections amounts to taking G or G  G- BG UPIrrpGq invariants (note that there are no higher cohomology groups since G is reductive). The second claim results from iterating the rst, or by applying the projection formula to the diagonal BG Ñ BG . Remark 2.3.3. The map above, which we now denote , and its inverse can be written out explicitly: pv b u b u b v b u b  b u b v q  pu pu q u pu qq  v b  b v 0 1 1 n n 1 n 0 n 1 2 1 n pv b  b v q   v b  b v b  b  b  b v 0 n 0 1 1 2 n n U ;:::;U PIrrpGq 1 n G G where  P U b U is the identity map, and   pq b  b pq is the tensor products of the projections i i to the G-invariant isotypic component. De nition 2.3.4. Let A P AlgpQCohpBGqq be an associated algebra object with coaction map c : A Ñ A b krGs. The equivariant cyclic bar complex C pA; Gq, de ned in [BG94], is the mixed complex associated to the following cyclic vector space. We de ne the n-simplices by bn1 G C pA; Gq  pA b krGsq with face and degeneracy maps d pa b  b a b fq  a b  b a a b  b a b f i  0; : : : ; n  1; i 0 n 0 i i1 n d pa b  b a b fq  cpa qa b  b f; n 0 n n 0 and cyclic structure tpa b  b a b fq  cpa q b a b  b a b f; 0 n n 0 n1 s pa b  b a b fq  1 b a b  b a b f: n1 0 n 0 n Remark 2.3.5. Note that it is essential to take G-invariants for C pA; Gq to admit a cyclic sructure. In particular, if a b  b a b f is G-invariant, then 0 n cpa q b  b cpa q b f  a b  b a b f: 0 n 0 n Proposition 2.3.6. The equivariant cyclic bar complex C pG; Aq computes HHpPerfpA -mod qq: QCohpBGq Proof. Since A generates A -perf , the objects A b V generate A -perf and we have the following QCohpBGq \small" model for HHpA -perf q. The cyclic bar complex corresponding to the generating set tAbV | QCohpBGq V P IrrpGqu of A -perf has terms QCohpBGq G G G D pA; Gq  Hom pAbV ; AbV q b Hom pAbV ; AbV q bb Hom pAbV ; AbV q : n A 0 1 A 1 2 A n1 n V ;:::;V PIrrpGq 0 n G  G Using the natural isomorphism Hom pA b V; A b Wq  pA b V b Wq , we rewrite: G  G  G D pA; Gq  pV b A b V q b pV b A b V q b  b pV b A b V q : n 1 2 0 0 1 n V ;:::;V PIrrpGq 0 n 19 n Next, we can rewrite C pA; Gq using the Peter-Weyl theorem for reductive algebraic groups: bn G C pA; Gq  pV b A b V q : n 0 0 V PIrrpGq Applying Proposition 2.3.2 to each summand (i.e. for xed V ) produces an equivalence D pA; Gq Ñ 0 n C pA; Gq; the claim that it de nes a map of cyclic objects is left to the reader. Remark 2.3.7. The equivariant cyclic bar complex C pX; Gq provides us with an explicit model of the Hochschild homology HHpPerfpX{Gqq as a krGs -linear mixed complex (i.e. both the internal and mixed G G di erentials are krGs -linear). Furthermore, since pq is a left adjoint functor whose right adjoint preserves epimorphisms, the terms in C pA; Gq are also projective. 2.4 Formal and derived completions We review the notion of formal completions of derived stacks and the notion of derived completion in the derived category. This section is essentially a summary of the results in Chapter 4 of [Lur11c], Section 3.4 of [BS15], Section 15.80 of the [Sta18], and Chapter 6 of [GR14]. De nition 2.4.1. Let f : X Ñ Y be a map of derived stacks (or more generally, prestacks). The formal completion of f , written Y , is a prestack whose functor-of-points whose S-points are given by diagrams red pSq X S Y: It can also be de ned via the ber product dR Y : Y  dR X : red Lemma 2.4.2. The formal completion of a map X Ñ Y only depends on  pXq Ñ Y . In particular, if x { Z Ñ Y is a closed embedding, then Y  X  X red . Z Y  pZ Xq 0 Y Proof. This follows directly from the functor-of-points characterization of formal completions, and that red red pXq is the universal stack that canonically factors any map from a classical reduced scheme  pSq . 0 0 De nition 2.4.3. Let f : A Ñ B be a map of derived rings. Following [Lur09a], we say that f is  etale if the induced map  pAq Ñ  pBq is  etale and for every n P Z, the map  pAq b  pBq Ñ  pBq 0 0 n  pAq 0 n is an isomorphism of abelian groups. A map of derived schemes is  etale if it is for a Zariski cover, and a representable map of derived Atin stacks is  etale if it is after base change to a cover. Proposition 2.4.4. Let X; Y; Z be stacks admitting deformation theory . Suppose that f : X Ñ Y is  etale, and let Z Ñ X be any map. Then, the relative cotangent complex vanishes L  0. Furthermore, the map X{Y p p on formal completions X Ñ Y is an equivalence. Z Z Proof. The rst sentence is Proposition 2.22 in [Lur09a]. For the second, by the exact triangle for cotangent complexes we have a natural isomorphism L  L under L , and note that the formal completion of Z{Y Z{X Z a map Z Ñ X is a colimit of square-zero extensions controlled by the the map between cotangent complexes L Ñ L (see Chapter IV.5 in [GR17]). Z{X We now discuss how to compute the derived completion of a quasicoherent complex in the derived category. It is de ned abstractly a the left adjoint to an inclusion functor of complete objects which we now de ne. This notion is de ned in Chapter IV of [GR17] and is satis ed by quotient stacks. 20 De nition 2.4.5. Let A be a connective dg ring, and x an ideal I €  pAq. We de ne the full subcategory A -mod of I -nilpotent objects consisting of those modules on which I acts locally nilpotently, i.e. for each nil cycle m P H pMq there is some power of I which annihilates m. By Proposition 4.1.12 and 4.1.15 in [Lur11c], the inclusion A -mod ãÑ A -mod is continuous and preserves compact objects; therefore it has a continuous nil right adjoint , which we call the local cohomology functor. We de ne the full subcategory A -mod of I - I loc local objects as the right orthogonal to A -mod , and the full subcategory A -mod of I -complete modules nil cpl to be the right orthogonal to A -mod . The subcategory A -mod has an equivalent characterization as loc cpl those modules such that the homotopy (derived) limit x x x M M M is zero for all x P I . By Proposition 4.2.2 of [Lur11c], the inclusion of the complete objects has a left adjoint, which we call the (derived) completion functor and denote pq. The following is Proposition 4.2.5 in [Lur11c]. Proposition 2.4.6. The composition of left adjoints pq A -mod A -mod A -mod nil cpl is an equivalence. Consequently, the composition of its right adjoints A -mod A -mod A -mod cpl nil is also an equivalence. Example 2.4.7. Let A  krxs and I  pxq, then krrxss is I -complete, krx; x s{krxs is I -nilpotent, krxs is neither, and krxs{x is both. Furthermore, { 1 { pkrrxssq  pkrxsq  krx; x s{krxsr1s krx; x s{krxs  krxsr1s  krrxssr1s: I I The derived completion and local cohomology functors can each be computed in two ways. The following can be found as Propositions 15.80.10 and 15.80.17 in [Sta18] and in a global form as Proposition 3.5.1 in [BS15] and Proposition 6.7.4 in [GR14]. The statements on local cohomology are well known (and which we will not use). Proposition 2.4.8. Assume  pAq is a noetherian ring, and choose generators f ; : : : ; f of I €  pAq. We 0 1 r 0 de ne the complex ± ± 1 1 1 1 1 G   pAq  pAqr s  pAqr ; s   pAqr ; : : : ; s : 0 0 0 0 f i;j f f f f i i j 1 r The derived completion of an A-module M can be computed in two ways: L n n M  R lim M b A{pf ; : : : ; f q  R Hom pG ; Mq: pAq pAq 1 r 0 Likewise, we can compute the local cohomology of M by n n L pMq  colim R Hom pM{pf ; : : : ; f q; Mq  G b M: I  pAq 0 1 r  pAq The latter formula is the calculation of local cohomology via a Cech resolution with supports on an ane scheme. 21 Remark 2.4.9. A theory of formal completions is described in [GR14] with the following notation. Let X p p be a dg scheme, and Z € X a classical closed subscheme. There is a functor i : QCohpXq Ñ QCohpX q with a fully faithful (continuous) left adjoint i whose essential image is the category of quasicoherent sheaves supported on Z , and a fully faithful (non-continuous!) right adjoint i whose essential image is the category of quasicoherent sheaves complete with respect to the ideal sheaf for Z . There is an exact triangle of functors p p arising from the localization functor i i (whose essential image is cocomplete): p p i i Ñ id Ñ j j  pq| : Z ? QCohpXq  U In particular, the functor on the left is local cohomology, and the functor on the right is restriction to U . p p The (non-continuous) functor i i is the (derived) completion. The following lemma is likely well-known, but we could not nd a reference. Lemma 2.4.10. Let X be a derived scheme, i : Z € X a closed subscheme and j : U  X  Z Ñ X its complement. Let  : F Ñ G be a map of quasicoherent sheaves on X . If the derived completion  and the restriction | are isomorphisms, then  is an isomorphism. Proof. Using the above exact triangle, to show that  is an isomorphism, it suces to show that pq is an p p isomorphism, or equivalently, that pconepqq  0. To this end, note that conepq  i i conepq  0, and p p p p that i is fully faithful, so that i conepq  0, so that pconepqq  i i conepq  0. Z ? Example 2.4.11. The above is not true for non-derived completions. For example, take X  A , Z  t0u, and  : 0 Ñ M  krx; x s{krxs (M is the module of distributions supported at zero). Since M is supported at zero, M|  0, and since x M  M for all k, M  0, but  is not an isomorphism. On the other hand, U Z the derived completion of M is krrxssr1s. 3 An equivariant localization theorem in derived loop spaces and Hochschild homology 3.1 Equivariant localization for derived loop spaces The following construction de nes a notion of formal and unipotent loops near a semisimple orbit of G{G, i.e. an adjoint closed G-orbit in G consisting of semisimple elements. De nition 3.1.1 (z-formal and z-unipotent loops). Let X be a derived scheme, G a reductive group acting on X , and z P G a semisimple element. We let G denote the centralizer of z, i.e. the z- xed points under the adjoint action. We de ne Z  tgzg | g P Gu to be the closed G-orbit containing z and 1 z z U  tgzug | g P G; u P U X G u to be its saturation (here, U is the unipotent cone of G and G is the centralizer of z). We de ne the z-formal and z-unipotent loops in BG by p p x L pBGq : Z{G Ñ LpBGq  G{G; L pBGq : U {G Ñ LpBGq  G{G: z z For a quotient stack X{G, we de ne the z-formal and z-unipotent loops by the pullback squares: L pX{Gq LpX{Gq L pX{Gq LpX{Gq L pBGq LpBGq L pBGq LpBGq: This construction is functorial in representable maps over BG. Note that by Propositions 2.1.20 and u u p p 2.1.25, L pX{Gq  LpX{Gq and L pX{Gq  L pX{Gq, where e P G is the identity. 22 z Proposition 3.1.2. Let G be a reductive group, and z P G a semisimple element. The map LpBG q Ñ LpBGq is (Zariski) locally  etale at z, i.e. there is a Zariski open neighborhood of Z{G € LpBGq  G{G, and therefore a Zariski open neighborhood of U {G € LpBGq, on which the map is  etale. Proof. Let us recall the set-up of the  etale slice theorem as in [Dr e04]. Let G be a reductive group acting on an ane variety X , and x P X a closed point such that the stabilizer Z pxq is reductive. We de ne a map : X Ñ T pXq as follows. Let m be the maximal ideal for x P X ; the quotient map to the cotangent space 2 2 has a Z pxq-equivariant splitting m{m Ñ m since Z pxq is reductive, de ning a map Sym pm{m q Ñ krXs. G G Geometrically, this means choosing functions f ; : : : ; f P krXs vanishing at x whose di erentials generate 1 n the cotangent space at x, and de ning the map  : X Ñ T pXq by evaluation y ÞÑ f pyq ; df yx in a Z pxq-equivariant manner. The  etale slice is the inverse image  pNq where N € T pXq is any normal G x Z pxq 1 subspace to T pG  xq € T pXq, and the theorem tells us that the map G   pNq Ñ X is  etale. x x Specializing to our situation, where G acts on itself by the adjoint action, we produce the G -equivariant map  : G Ñ T pGq as follows. Ane locally at z, we can choose generators f ; : : : ; f such that z 1 n krGs{pf ; : : : ; f q  krG s; 1 r and the vanishing of df ; : : : ; df cuts out g € T pGq  gz (i.e. the translation of T pGq  g by central z). 1 r z e z 1 z z Thus it suces to show that g is a normal subspace to T pG  zq, since  pg q  G by construction. On z z the other hand, we have a natural isomorphism G z  G{G , inducing T pG zq  gz{g z, which produces a splitting of T pG  zq € T pGq whose kernel is g . Explicitly, z is semisimple and acts on g, so g decomposes z z z z into z-eigenspaces; g is the trivial eigenspace and g{g is isomorphic to the sum of all other eigenspaces. z G Using the fact that the map G  G Ñ G is G-equivariant, and since for any u P U we have z P G  u, it follows that every open set containing Z also contains U . Example 3.1.3. Let G be a simple reductive algebraic group and choose z P G regular semisimple. Its T reg rs centralizer is a torus T and the map G  T Ñ G is  etale with ber W  NpTq{T . Corollary 3.1.4. Let X be a prestack equipped with the action of a reductive group G over k, and let z P G be central. Then, LpX{G q Ñ LpX{Gq is  etale over a neighborhood of U {G € G{G. In particular, the z u z u p p natural maps L pX{G q Ñ L pX{Gq and L pX{G q Ñ L pX{Gq are equivalences. z z z z z z Proof. Loop spaces commute with ber products, and X{G  X{G  BG . The second claim follows BG since  etale maps induces equivalences on formal completions along isomorphic closed subschemes. It is a straightforward veri cation that the map G 1 z z 1 z U  G  tphuzh ; gq | g P G; h P G ; u P U X G u ÝÑ U  tguzg | u P U X G u z;G G z;G G is an isomorphism. We de ne two competing notions of z-invariants. De nition 3.1.5. In the set-up above, we de ne the derived z-invariants X to be the derived ber product of the diagram X X X pX  Xq where denotes the graph of the action by a closed point z P G. We de ne the classical z-invariants X to z z be  pX q{G , or equivalently, the above square considered as an underived ber product. 23 Given this notion of z-formal and z-unipotent loops, we are ready to de ne the localization map comparing z-formal and unipotent loops with the formal and unipotent loops of the classical z- xed points. De nition 3.1.6 (Localization map and formal localization map). We de ne the following maps realizing localizations from the most global to local. We de ne the global localization map via the composition z z z ` : Lp pX q{G q LpX{G q LpX{Gq z 0 z z z induced by the sequence of natural maps of quotient stacks  pX q{G Ñ X{G Ñ X{G where the rst map is the closed immeresion and the second map is the base change map along BG Ñ BG. For U{G € G{G an open subscheme, we by ` the restriction of ` to U . The map lives over the natural z;U z map LpBG q Ñ LpBGq. u u z z u The unipotent localization map ` : L p pX q{G q L pX{Gq is the base change of ` along 0 z z z u z z L pBG q LpBG q L pBGq LpBGq where the isomorphism on the left arises via Corollary 3.1.4. Note that the map lives over L pBGq. z z p p p The formal localization map ` : L p pX q{G q L pX{Gq is the base change of ` along z z 0 z z z z L pBG q LpBG q L pBGq LpBGq where the isomorphism on the left arises via Corollary 3.1.4. Note that the map lives over L pBGq. Remark 3.1.7. Applying the functor  pt, ` can be identi ed with the map induced on ber products BG z of the diagrams G z z G  p pX q  G q 0 X  G ÝÑ z z G z G z z X X  X G   pX q G  p pX q   pX qq 0 0 0 where the top map sends ph; x; gq ÞÑ ph  x; hgh q. z z z z We now investigate the map Lp pX q{G qq Ñ LpX{G q. For ease of notation, we can replace G with a reductive group G in which z is central. Our goal is to nd an open G-closed neighborhood U{G € G{G on which the above map is an equivalence. Let us rst consider the case when G  T is a torus. Lemma 3.1.8 (Finiteness of stabilizer subgroups). Let T be a torus acting on a (quasicompact) variety X . Only nitely many subgroups of T may appear as stabilizers of this action. Proof. We can work ane locally, since X has a nite T -closed Zariski cover, and may also assume that X is connected. If every point of X has stabilizer of equal dimension to T , the possible stabilizer subgroups are in bijection with a subset of the set of subgroups of the ( nite) component group T{T . If there is a point 24 x x P X whose stabilizer T € T has strictly smaller dimension, then by the Luna slice theorem (note that the stabilizer T is reductive since every subgroup is) there is a locally closed subvariety V € X such that the map a : T  V Ñ X is  etale and dominant. Any stabilizer of a point in the image of a must be a subgroup x x of T , so the problem reduces to considering (a) the action of T on V along with (b) the action of T on the compliment of V € X (which is a closed subvariety, therefore ane, of strictly smaller dimension). Note that in both cases, the dimension of either the variety or the group decreases, and that the claim is obviously true for zero-dimensional varieties and discrete ( nite) groups, so the lemma follows by induction. Corollary 3.1.9. In the above situation, let z P T and let U € T be the open neighborhood of z obtained by w z deleting the nitely many stabilizers which do not contain z. Then, for w P U , we have X € X . We bootstrap this reuslt to prove an analogous result for general reductive G. Proposition 3.1.10. Let G be a reductive group, z P G a central element and X a scheme with a G-action. Then, there is an open neighborhood U of z in G that is closed under the adjoint action and such that w P U w z implies that X € X . Proof. We de ne U as follows. First, take a maximal torus T € G containing z. By Proposition 3.1.8, there is an open neighborhood U of T with the given property. We de ne 1 1 uni U  tgtug | t P U ; g P G; u P C ptq u uni where C ptq are the unipotent elements in the centralizer of t. The set U is evidently G-closed. By Jordan 1 1 decomposition, G  U  G  pT  U q, so U is open since T  U is closed. It remains to show that if w P U , w z then X € X . Every w P U has a Jordan decomposition w  su where s is semisimple and u is unipotent; in particular, 1 1 s t z z z s  gtg for some t P U and g P G. First, note that X  g  X € g  X  X (since X is G-closed for w s z central z). By the following lemma, it follows that X € X € X . Lemma 3.1.11. Let w P G be an element of a reductive group acting on a scheme X , with Jordan decom- w s position w  su for semisimple s and unipotent u. Then, X € X . Proof. If u  e the claim is trivial, so suppose u  e. Take x P X . There is a 1-parameter subgroup of G containing s, and a choice of unipotent u  1 uniquely de nes an injective map of algebraic groups G Ñ G, assembling into an injective map of group schemes G  G Ñ G. Then, H  pG  G q X G is a closed m a m a group scheme of G which is at most two-dimensional and which contains w  su. The connected component H is either the trivial group, G  t0u, t1u  G , or the entire group G  G . If it is the trivial group m a m a then H is discrete, therefore nite, but this is impossible since the projection to G is open and there are no nite group subschemes of G . For the same reason, H cannot be G  t0u. In the remaining two cases, a m x x wH € H € G , so w P G as desired. Given this, we are now ready to prove our main theorem. Theorem 3.1.12. Let X be a smooth variety with an action of a reductive group G, and z P G semisimple. There is an open substack U{G € G{G containing z such that the S -equivariant map z z z z ` : Lp pX q{G q  z z pU X G q{G LpX{Gq  U{G z;U 0 G {G G{G is  etale. If z P G is central, then ` is an equivalence. In particular, the S -equivariant maps on z-unipotent z;U loops and z-formal loops are equivalences: u u z z u z z p p p ` : L p pX q{G q L pX{Gq; ` : L p pX q{G q L pX{Gq: 0 z z 0 z z z z 25 Proof. Assuming the rst statement holds, then the statement on z-unipotent and formal loops follows since z z z the global localization map is a composition Lp pX q{G q Ñ LpX{G q Ñ LpX{Gq: By the rst statement z z z of the theorem, since z is central in G , there is an open substack U of G {G containing both Z  G  tzu and its saturation U over which the rst map is an equivalence. By Proposition 3.1.2, the second map is etale on U , inducing an equivalence on formal completions along Z and U by the argument in Corollary 3.1.4. 2 z z We now prove the rst statement. Let U € G be an open neighborhood of z P G from Proposition w z 2 1 3.1.10, i.e. such that  pX q €  pX q for w P U . Let U € G be an open neighborhood of z on 0 0 which  is  etale, obtained via Proposition 3.1.2; note that this implies that  is open over U . De ne 2 1 z U  pU q X U € G. By construction, the map LpX{G q Ñ LpX{Gq is  etale over U . It remains to prove that the map is an equivalence when z P G is central; in this case G  G. We proceed by base changing from BG to Specpkq (i.e. forgetting equivariance). Take Y   pX q for w z shorthand, and recall that U € G is an open subscheme on which  pX q €  pX q for w P U . This means 0 0 we wish to show that the map of derived schemes i : Y  pY  Uq Ñ X  pX  Uq Y Y XX is an equivalence. Since i is a closed embedding (and therefore ane), we view the map of derived schemes ane locally as a map of di erential graded sheaves of algebras on U  G, which we abusively denote by O Ñ O  i O . Since these sheaves have coherent cohomology on XG and closed points LpX{Gq LpY {Gq LpY {Gq are dense in X  G (X  G is locally nite type over a eld), it suces to check the claim on local rings at 2 w z closed points px; wq P  pLpX{Gqq € X  G where w P U € G and x P  pX q €  pX q (via the de ning 0 0 0 property of U ). z 11 Note that Y   pX q is smooth when z P G is semisimple in reductive G by a standard argument . Consequently, the diagonal maps are local complete intersections, and Y   pX q € X is also a local complete intersection. The z-action on the cotangent space T pXq is semisimple, and determines a splitting with identi cations z  z T pXq  E ` E ; E  kerpz  q  N p pX q{Xq; E  kerpz  1q  T p pX qq: 0 1 0 0 1 0 x x x Let J denote the ideal such that O  O {J . We aim to compute Y;x X;x L L O b O ÝÑ O b O : (1) X;x Y;x O XG;px;wq O Y G;px;wq XX;px;xq Y Y;px;xq Let v ; : : : ; v be a basis of E and v ; : : : ; v a basis of E . By Nakayama's lemma we can lift this basis of 1 r 1 r1 n 0 the cotangent space to generators x ; : : : ; x P m € O such that J  px ; : : : ; x q. Furthermore, again 1 n X;x X;x 1 r by Nakayama, tx b 1  1 b x | i  1; : : : ; nu form a regular sequence for the diagonal X € X  X at i i px; xq, and likewise the images of tx b 1 1b x | i  r 1; : : : ; nu form a regular sequence for the diagonal i i Y € Y  Y at px; xq. Taking semi-free Koszul resolutions of the diagonal, as well as O as a O -module, Equa- Y G;px;wq XG;px;wq tion 1 can be rewritten 1 1 O r ; : : : ;  s O r ; : : : ;  s O r ; : : : ;  ;  ; : : : ;  s (2) 1 n r1 n r1 n XG;px;wq Y G;px;wq XG;px;wq 1 r where | |  1, and the internal di erentials are de ned by dp q  cpx q  x b 1 and dp q  x b 1. The i i i i i map on the right is a quasi-isomorphism. To show that Equation 2 is a quasi-isomorphism, it suces to produce the dotted map above making the diagram commute. Explicitly, we wish to show that the derived Choose a torus T containing z and apply the  etale slice theorem to a T -closed ane open cover. Since A and A{J are Cohen-Macaulay, any minimal generating set is automatically regular. 26 1 1 equations imposed by dp q; : : : ; dp q and by dp q; : : : ; dp q di er by an (invertible) change of variables. 1 r 1 r That is, we wish to nd an element of GL pO q which transforms r XG;px;wq tx b 1; : : : ; x b 1u into tcpx q  x b 1; : : : ; cpx q  x b 1u: 1 r 1 1 r r The element z is central, so w xes  pX q, and in particular the coaction map c preserves the ideal J . Thus, we can write cpx q  e px b 1q for some e P O . Let E  pe q denote the i ij j ij ij XG;px;wq i;j1 corresponding matrix; the matrix E I is invertible if its evaluation at w P G is invertible. The matrix Epwq is the action of w on the conormal space E  N p pX q{Xq; in particular, pE  Iqpwq is invertible if and 0 0 only if Epwq has no xed vectors. But w cannot x any vectors on a normal space of its xed-point variety, z w and since N p pX {Xq € N p pX {Xqq, the claim follows. x 0 x 0 Corollary 3.1.13. Let X; Y; Z be smooth varieties, with maps f : X Ñ Z and g : Y Ñ Z . Then z z p p L p pX q  z  pY qq  L pX  Y q z 0 0 z Z pZ q where all ber products are derived. Proof. This follows immediately since loop spaces commute with ber products. Remark 3.1.14. Note that in the case G  T is a torus, every element of T is central, so ` is an z;U equivalence. Furthermore, Proposition 3.1.8 gives an explicit description of the open set U , which is maximal, on which ` is an equivalence. 3.2 Equivariant localization for Hochschild and cyclic homology First, let us show that the completion over a closed point of the anization rzs P G{{G is the same as taking z-unipotent loops. Lemma 3.2.1. Let a : G{G Ñ G{{G be the anization, and let apzq  rzs P G{{G. The map above induces an isomorphism on completions { 1 L pBGq  a przsq{G € LpBGq: u { In particular, the map L pBGq Ñ LpBGq factors isomorphically through a przsq{G. Proof. This follows from Proposition 2.1.25 and the observation that the classical reduced ber over rzs P G z z z z z in G{G is isomorphic to U {G , where U is the unipotent cone of G . That is, if u P G is unipotent, then G G uz  zu P G has the same eigenvalues as z, so apzq  apzuq, and by Jordan composition any y P  przsq uz z z can be written uniquely in this way. Furthermore, G € G ; letting U be the unipotent elements of G , it G 1 follows from Proposition 3.1.2 that the map G  U Ñ G is a closed embedding with image  przsq. As an application of the above geometric incarnations of equivariant localization, we obtain the following equivariant localization results in Hochschild and cyclic homology. De nition 3.2.2 (Completion and localization of Hochschild homology). Let G be a reductive group. Note that HHpPerfpX{Gqq is naturally a HHpPerfpBGqq  krGs -module. Let z P G be a reductive element representing a closed point of SpecpkrGs q  G{{G. We denote by HHpPerfpX{Gqq the localization at the maximal de ning z P G{{G and HHpPerfpX{Gqq the completion at z. We de ne the analgous notions for zp HN; HC and HP , taking care to completed with respect to the ideals I rruss and I ppuqq where I € krGs z z z is the ideal de ning z P G{{G. The following is an immediate corollary of the the localization theorem for unipotent loops. 27 Theorem 3.2.3 (Equivariant localization for Hochschild homology). Let X be a smooth scheme, G a reduc- tive group, and z P G semisimple. Then the natural map on completions induced by pullback is an equivalence: z z u z z HHpPerfpX{Gqq HHpPerfp pX q{G qq  OpL p pX q{G qq: zp 0 zp 0 If z P G is central, the map on localizations is an equivalence: z z HHpPerfpX{Gqq HHpPerfp pX q{Gqq  OpLp pX q{Gqq b O : z 0 z 0 O G{{G;rzs G{{G z z u Proof. Via Lemma 3.2.1, we have a natural identi cation HHpPerfp pX q{G qq  OpL pX{Gqq. The zp z rst claim then follows directly from Theorem 3.1.12, and base change for derived completions along closed embeddings. The second claim for localizations follows by Theorem 3.1.12, since in that case the map on Hochschild homology is an equivalence for an open set containing z. Remark 3.2.4. For cyclic homology, the situation is a little more delicate. The formation of cyclic homology HC involves a ltered colimit, the formation of negative cyclic homology HN involves a co ltered limit, and the formation of periodic cyclic homology involves both. On the other hand, localization commutes with colimits and nite limits but not co ltered limits, and completion commutes with limits but not colimits. We rst need to introduce a few technical notions regarding mixed complexes, with the aim of proving that in our situation formal completions commute with the Tate construction on Hochschild homology. Analogous results and arguments can be found in [Kal15]. De nition 3.2.5. Let pV; d; q be a mixed complex. We de ne ± ¹ Tate k ` Tate k 1 V  p V u ; d uq V  p V u ; d uq  pV ru ; us; d uq where |u|  2. There are natural maps ` Tate Tate Tate V Ñ V Ñ V : Remark 3.2.6. Lemma 2.6 of [Kal15] shows that the Tate construction preserves quasi-isomorphisms, es- sentially because it is computed via the right spectral sequence. On the other hand, the other variants above do not preserve quasi-isomorphisms. In particular, they are not well-behaved in the derived category. De nition 3.2.7. We say a complex V is cohomologically bounded below (respectively, above ) if H pV q  0 for all suciently small (respectively, large) i. We say V is strictly bounded below (respectively, above ) if V if V  0 for i suciently small (respectively, large). ` Tate Tate Lemma 3.2.8. Let pV; d; q be a mixed complex. If V is strictly bounded below, then V Ñ V is an Tate Tate isomorphism. If V is strictly bounded above, then V Ñ V is an isomorphism. Proof. The proof of the rst statement appears as Corollary 2.7 in [Kal15], which we repeat for convenience. ` Tate k n2k Tate The chain complex V is, in the nth cohomological degree, the vector space u V . whereas V À ± k n2k k n2k n2k is in the nth cohomological degree the vector space p u V qp u V . Since V  0 for k¥0 k 0 large k, the product is nite and therefore a direct sum, so the map is an isomorphism. A similar argument proves the second statement. Lemma 3.2.9. Let pV ; d ;  q be a degree-wise Mittag-Leer sequential diagram of mixed complexes such that the V are uniformly cohomologically bounded above (i.e. right t-bounded). Then, the functors pq 1 Tate and pq commute with limits, i.e. Tate Tate plim V q  lim V : 28 Proof. First, since the limit is degree-wise Mittag-Leer, R lim  lim. Note that since the V are uniformly cohomologically bounded above, so is their limit. Since we are only interested in computing the usual Tate construction which respects quasi-isomorphisms, we can replace each of the V and lim V with their (strictly bounded above) truncations; the resulting complex is still Mittag-Leer. Now, note that the Tate construction commutes with limits. Theorem 3.2.10 (Equivariant localization for cyclic homology). Let X be a smooth scheme, G a reductive group, and z P G semisimple. Then the following maps on completions induced by pullback are equivalences z z u z z S HNpPerfpX{Gqq HNpPerfp pX q{G qq  OpL p pX q{G qq ; zp 0 zp 0 z z u z z HCpPerfpX{Gqq HCpPerfp pX q{G qq  OpL p pX q{G qq 1; zp 0 zp 0 z S z z u z z Tate HPpPerfpX{Gqq HPpPerfp pX q{G qq  OpL p pX q{G qq : zp 0 zp 0 z z u Proof. Via Lemma 3.2.1, we have a natural identi cation HHpPerfp pX q{G qq  OpL pX{Gqq. If we can 0 zp show that the limit diagram in the formation of the derived completion satis es the conditions of Lemma 3.2.9, then the claims would follow via Theorem 3.2.3. Take A P AlgpQCohpBGqq to be as in Proposition 2.3.1. First, note that HHpPerfpX{Gqq  C pA; Gq is cohomologically bounded above since A is cohomologically bounded. By Remark 2.3.7, its terms are projective, so the terms in the limit computing its derived completion with respect to an ideal of krGs are classical quotients. Therefore, the limit diagram satis es the conditions of Lemma 3.2.9. Remark 3.2.11. While the statement in the above theorem for negative cyclic homology was more or less automatic, the statements for cyclic homology and periodic cyclic homology are strongly dependent on the cohomological right-boundedness of Hochschild homology, which we expect to fail for CohpX{Gq where X is singular. 4 An Atiyah-Segal completion theorem for periodic cyclic homol- ogy 4.1 Twisted circle actions on loop spaces De nition 4.1.1. Let G be an ane algebraic group over k. A G-action on a prestack X is de ned to be a prestack Y over BG, along with an identi cation Y  pt  X . We often abuse notation and write BG Y  X{G and understand that the identi cation implicitly. Furthermore, if Z € G is a closed normal subgroup, a trivialization of the action of Z on X is de ned to be a G : G{Z -action on X along with an identi cation X{G  1 BG  X{G. If z P G is central, then it generates a normal closed subgroup Z ; we BG sometimes write z-trivialization to mean a Z -trivialization. Furthemore, note that any X equipped with a z-trivialization is equipped with a shift map  : X{G Ñ X{G de ned as follows. It is de ned on BG via the map of groups G Ñ G taking g ÞÑ zg  gz. It is de ned on X{G by transporting the map id 1   . across the trivialization X{G  X{G  1 BG. It is X{G id 1 z BG BG evidently an equivalence with inverse  1 . Example 4.1.2. If X is a classical scheme with a G-action, on which a normal subgroup Z € G acts trivially, then there is a canonical trivialization of the action of Z on X . De nition 4.1.3 (Twisted S -actions). Let G a linear algebraic group over k and X a scheme over k with an action of G. We de ne the following various circle actions on X{G. Let z P G be central. The loop rotation S -action, denoted , on the loop space LpX{Gq. 29 1 The z-twisting S -action, denoted  , on BG is induced via the map of groups Z  G Ñ G de ned by n n pn; gq ÞÑ z g  gz . More generally, if X is equipped with a z-trivialization, then the corresponding 1 1 1 z-twisting S action on X{G is de ned via the identi cation X{G  X{G  BG, where S acts BG 1 1 on X{G and BG trivially, but on BG via the z-twisting described above. It is evident that the maps de ning the ber product diagram are S -equivariant. 1 1 The z-twisted rotation, denoted  , on LpX{Gq is the diagonal of the S  S -action   L . Note z z that it makes sense to talk about the diagonal since  and L commute:  commutes with any group action of the form L , where is an S -action on X{G. Furthermore, for w P G central, the map  : X{G Ñ X{G is S -equivariant with respect to  on the left z zw and  on the right. In particular, Lp q : LpX{Gq Ñ LpX{Gq is equivariant with respect to  on the left w z zw and  on the right. Remark 4.1.4. Using Proposition 2.1.8, we see that the shift map Lp q : LpX{Gq Ñ LpX{Gq has an explicit realization as the multiplication by z map G{G Ñ G{G on the upper-right term in the ber product LpX{Gq  pt pX  Gq{G BG ap X{G pX  Xq{G: Example 4.1.5. On PerfpG{Gq, the rotation  acts on bers over g P G by g; the z-twisting L acts on bers over any g P G by z, and the twisted rotation  acts on bers over g P G by gz  zg. Example 4.1.6. Let G  T be a torus (in particular, every t P T is central). We can explicitly describe the S -actions on linear categories PerfpLpBTqq  PerfpT  BTq  PerfpTq where  is the character lattice of T . Let tz |  P u denote the natural basis of monomials for krTs. The rotation S -action  acts on the -summand by z . The t-twisting action  on BT acts on the -summand by the scalar z ptq. The t-twisted rotation  acts on the -summand by z ptqz . Let us take the t-twisted rotation  . We have, via the category PreMF in [Pre11], PerfpLpBTqq  PreMFpT; 1  z ptq  z q; p p PerfpLpBTqq  PreMFpT; 1  z q: ptq1 For the second identity, the zeros of 1  z ptqz meet the constant loops if and only if z ptq  1. After passing to the Tate category under the rotation action, we note that the zero locus of 1  z is smooth (it is a subgroup of T ) of codimension 1 unless   0 (in which case it has codimension zero and must be derived). Therefore, Tate PerfpLpBTqq  MFpT; 1  z ptq  z q  PerfpTq b kpuqq; Tate p p p PerfpLpBTqq  MFpT; 1  z q  PerfpTq b kppuqq: ptq1 Note that the Tate categories do not depend on the twisting at all (but the S -invariant categories do). For M a scheme and f : M Ñ G , the category PreMFpM; fq is the category PerfpM  t1uq with an extra kruss-linear structure acting by cohomological operators. 30 Proposition 4.1.7. Suppose z P G is a central element of a reductive group and acts on a quasiprojective scheme X trivially. Then, there is a krruss-linear equivalence 1 1 S ; S ; Lp q : OpLpX{Gqq OpLpX{Gqq : The same holds for formal and unipotent loop spaces. Proof. Since X is quasiprojective, it has a compact G-equivariant generator E of QCohpXq. Let A R Hom pE;Eq, so that QCohpXq  A -mod . Let c : A Ñ A b krGs be the coaction; note that QCohpBGq the z-twisted rotation is given by twisting the coaction by z, i.e. c paqp; gq : cpaqp; gzq  cpaqp; zgq. Since z acts on X trivially, this is equal to the usual coaction c, inducing an equivalence of S -invariants under the untwisted and z-twisted rotations. 4.2 Tate-equivariant functions on formal loop spaces compute analytic de Rham cohomology We now set out to prove the equivariant localization theorem for periodic cyclic homology. We rst introduce some technical notions needed to phrase the result in the 2-periodic setting. Recall the following notions for vector spaces (not chain complexes) from [Bei08]. De nition 4.2.1. A linear topological vector space is a a vector space V which admits a topology for which the vector space operations are continuous, and such that there is a system of neighborhoods at 0 consisting of subspaces. In this case, the topology is generated by this system at 0 and translations under addition. The completion V of V is the limit over the system of neighborhoods U : V : lim V {U : 0PU We say the topology is complete if the natural map V Ñ V is an isomorphism. Let V ; V be linear topological 1 2 vector spaces. We de ne a linear topological vector space, the !-tensor product V b V , via the naive tensor 1 2 product on underlying vector spaces equipped with topology by the basis consisting of open sets of the form x ! U b V V b U , where U € V and U € V are opens. We de ne V b V to be V b V completed with 1 2 1 2 1 1 2 2 1 2 1 2 respect to this topology. These notions generalize immediately to chain complexes, where we replace the notion of subspace with subcomplex. In this case, the complexes term-wise satisfy the Mittag-Leer condition and therefore lim  0, so R lim  lim. Note that as in [Kal15] V b V  lim lim V b V : 1 2 1 2 V V 1 2 Remark 4.2.2. It is unclear to us how the notion of a topological chain complex should interact with quasi-isomorphisms. We now review the constructions and results of [Bha12] and [Har75]. De nition 4.2.3. Let X be a nite-type derived stack with ane diagonal over k, and L its cotangent À ™ complex. The derived de Rham complex dR is the sum-totalization of the complex p L rns; d q, X X dR n¥0 À ™ which comes equipped with a Hodge ltration F dR  p L rns; d q. The Hodge-completed X X dR n¥k derived de Rham complex dR is the completion of dR with respect to the Hodge ltration (see Construction X X 4.1 in [Bha12]). We will denote the derived global sections of this complex to be the complex of derived de Rham cochains C pX ; kq. dR We have two competing notions of Hodge ltrations in the negative cyclic homology of PerfpX{Gq. Both play an essential role. 31 1 S Tate De nition 4.2.4. Let pV; d; q be a mixed complex. The noncommutative Hodge ltration on V and V k S is the decreasing ltration de ned by the subspaces u V . De nition 4.2.5. The odd tangent complex T r1s is ane over X , so we can consider O as an X T r1s algebra in the category QCohpXq. Furthermore, the ideal sheaf for the zero section de nes an exhaustive decreasing ltration on O , which induces an exhaustive decreasing ltration on OpT r1sq whose T r1s X completion is OpT r1sq . We call this the geometric Hodge ltration. Proposition 4.2.6. Let X be a geometric stack with a smooth cover by a variety. There is a natural quasi-isomorphism Tate OpLpXqq C pX ; kqb kppuqq dR where we consider the de Rham complex C pX ; kq as a topological chain complex with respect to the derived dR Hodge ltration, and kppuqq with respect to the noncommutative Hodge ltration. p p Proof. By Theorem 6.9 of [BN12], the exponential map T r1s Ñ LpXq is an ltration-preserving isomor- Tate S phism, so we can compute OpT r1sq instead. Note that pq commutes with totalization, so we rst compute S i j OpT r1sq  lim lim RpX; risqu ; u  d : X dR nÑ8 mÑ8 0¤i¤n 0¤j¤m Note that the ltration de ned by the limit parameter n is the geometric Hodge ltration, and the ltration de ned by the limit parameter m is the non-commutative Hodge ltration. As we take the limit with respect i j to both, it amounts to computing the direct sum complex p RpX; risqu ; u  d q with respect to dR 0¤i;j X the opens i j U   RpX; risqu ; u  d : nm dR i¥n or j¥m There is a subcomplex (in fact, a direct summand) of the direct sum complex de ned by taking the summands where j ¥ i: i j V  RpX; risqu ; u  d : dR 0¤i¤j It is a subcomplex since the de Rham di erential takes the pi; jq-summand to the pi 1; j 1q-summand. Its quotient 1 i j V  RpX; risqu ; u  d dR 0¤j i is u-torsion. Note that kppuqq is at as a (dg) krruss-module, so  b kppuqq kills u-torsion modules, and krruss in particular we have an equivalence induced by the inclusion on completions Tate p p V b kppuqq  OpT r1sq : krruss X Consider the alternative topology on V de ned by opens i j W  RpX; risqu ; u  d : km  dR ij¥k or j¥m The parameter k in this topology de nes the derived Hodge ltration in the derived de Rham complex, and the parameter m in this ltration de nes a u-adic ltration. In particular, the completion of V with respect 32 ! to the topology de ned by the W is pX; dR qb krruss. km X We claim these two topologies de ned by U and V are equivalent. Indeed, this is an easy veri cation nm km as all indices are bounded below. Explicitly, W € U and U € W . In particular, the completion nm nm km;m km with respect to this topology is the limit under the usual Hodge ltration on each summand of V de ned by letting i  j be constant, and the limit under the noncommutative Hodge ltration. Thus, via the universal property of the limit, we have a canonical equivalence p x V  pX; dR qb krruss and in particular, Tate p x ! OpLpXqq  V b kppuqq  pX; dR qb kppuqq  C pX ; kqb kppuqq: krruss X dR Remark 4.2.7. The above proposition is false if we do not consider the topologies. For example, take X 0 Tate p p BG . Then, we have OpLpBG qq  krrrtss where |t|  0, and in particular, H pOpLpBG qq q  krrtss. m m m 0 1 On the other hand, H pBG ; kq  krss  krrsss where |s|  2, so H pH pBG ; kqppuqqq  krsu s. m m Finally, we discuss results relating derived de Rham cohomology to analytic (Betti) cohomology when k  C. It is well-known by experts and is essentially a simple corollary of results in [Bha12], [Har75] and [BN12]. A general discussion can also be found in the introduction of [Kal15]. We rst de ne a few intermediate chain complexes. De nition 4.2.8. For a choice of embedding of i : X Ñ M for smooth M , the Hartshorne algebraic de Rham complex is de ned by H 1 { 1 : i O b 1 i X i O Y We de ne the term-wise Hodge-completed derived de Rham complex associated to a groupoid presentation U U of a stack as above to be the totalization of the double complex dR . Since for schemes U , 1 0 U i the de Rham cohomology computes Betti cohomology, Lemma 32 of [Beh04] implies that the cohomology of this complex is independent of choice of cover. Corollary 4.27 in [Bha12] and Theorem 1.1 in Chapter IV of [Har75] can be summarized in the following statement, identifying the derived de Rham cohomology of a possibly singular scheme X over C with its Betti cohomology. For details, see Construcion 4.25 in [Bha12]. Theorem 4.2.9 (Bhatt, Hartshorne). Let X be a nite type scheme over k  C. There is a natural H an map of sheaves of dg k-algebras dR Ñ on X which is a quasi-isomorphism. Letting j : X Ñ X an be the analyti cation map, and the analytic de Rham complex on X , the map of dg k-algebras an 1 H 1 an is a quasi-isomorphism. All together, the map j dR Ñ an is a quasi-isomorphism, X X X an and an resolves the constant sheaf C . Therefore, the hypercohomology of the derived de Rham complex an computes the Betti cohomology of X . Corollary 4.2.10. Let X be a nite type stack over k  C presented as a groupoid U U . Then, 1 0 an Betti cohomology of X is computed by the term-wise Hodge-completed derived de Rham complex associated to U : Corollary 4.2.11. Let X be a nite type derived stack over k  C. Then, there is a natural quasi- isomorphism Tate an OpLXq C pX ;Cqb kppuqq: dR Proof. By Proposition 6.4 of [BN12], the Hodge-completed graded algebra of di erential forms is a graded sheaf on the smooth site of X , i.e. the Hodge-completed derived de Rham complex for stacks is quasi- isomorphic to the term-wise Hodge-completed derived de Rham complex associated to an atlas U Ñ X of 33 the stack X . This implies that the derived de Rham complex computes Betti cohomology and in particular, using Proposition 4.2.6 there is an ltration-preserving equivalence in the derived category C pX ;Cq dR an C pX ;Cq. 4.3 Comparing global functions on unipotent and formal loop spaces We prove a completion theorem for periodic cyclic homology, assuming the following theorem, which is proven as Theorem 4.3.22. Theorem 4.3.1. Let X be a quasicompact algebraic space with an action of an ane algebraic group G. The map on functions induced by pullback is an equivalence: u Tate Tate OpL pX{Gqq OpLpX{Gqq : Theorem 4.3.2 (Atiyah-Segal completion for periodic cyclic homology). Let G be a reductive group acting on a smooth quasi-projective variety X . The periodic cyclic homology HPpPerfpX{Gqq is naturally a module over HPpPerfpBGqq  krG{{Gsppuqq. For a closed point z P G{{G, we have an identi cation of the formal completion at z with a 2-periodicization of the singular cohomology of the xed points z an z an HPpPerfpX{Gqq C ppX q {pG q ; kqb kppuqq zp dR z an ! as a module over HPpPerfpBGqq  C pBpG q ; kq b kppuqq, contravariantly functorial with respect to zp dR k X . Proof. By Theorem 3.2.3 and Theorem 3.1.12, u Tate u z z Tate; HPpPerfpX{Gqq  OpL pX{Gqq  OpL p pX q{G qq : zp 0 By Proposition 4.1.7 and Theorem 4.3.1, u z z Tate; u z z Tate z z Tate OpL p pX q{G qq  OpL p pX q{G qq  OpLp pX q{G qq : 0 0 0 Finally, by Theorem 4.2.6, we have z z Tate z an z an OpLp pX q{G qq  C p pX q {pG q ; kqppuqq: 0 0 dR It remains to prove Theorem 4.3.1. Central to our proof will be to use the fact that the map is a pro- graded isomorphism; the following lemma establishes a general situation when this is true. Let us rst clarify what we mean by a pro-graded isomorphism, and why this notion is necessary. De nition 4.3.3. A pro-graded chain complex V is an object of PropQCohpBG qq; that is, it is a l- tered limit of graded chain complexes . Letting L denote the weight n twisted one-dimensional G - n m representation, the nth homogeneous part functor is given by wtn pq : ev  PropqpBG ; bL q : PropQCohpBG qq Ñ Vect : m n m k where the functor PropqpBG ;q : PropQCohpBG qq Ñ PropVect q m m k By Proposition 1.1.3.6 of [Lur17], this category is stable. 34 is the functor induced on pro-completions from pBG ;q : QCohpBG q Ñ QCohpptq and the functor m m ev : PropVect q Ñ Vect k k takes a limit diagram and evaluates it in Vect (which has all limits); it is right adjoint to the inclusion. The underlying chain complex is given by ev  Propp q : PropQCohpBG qq Ñ Vect m k where p : pt Ñ BG is the usual atlas so that p is the forgetful functor. A map of graded chain complexes is a pro-graded isomorphism if it is an isomorphism on nth graded parts for all n. Remark 4.3.4. We require this formalism for the following reason. Let p : pt Ñ BG be the standard atlas; the pullback (forgetful functor) p does not commute with limits since the category QCohpBG q cannot di erentiate between direct sums and direct products across di erent weights. In particular, objects of QCohpBG q are Z-graded chain complexes, which are equal to the direct sum of their homogeneous pieces . Example 4.3.5. For example, completing t0u € A {G (under the usual scaling action), n n lim krxs{x  krrxss lim krxs{x  krxs: n;QCohpptq n;QCohpBG q Remark 4.3.6. One way to remedy this is to keep track of the limit diagrams by working in the category PropQCohpBG qq and apply the evaluation functor in Vect  QCohpptq rather than QCohpBG q. How- m k m ever, the category PropQCohpBG qq contains more information than we need: we do not wish to track the topologies on vector spaces as we only care about their completions. Instead, we consider a smaller category: the category of k -modules De nition 4.3.7. We de ne a functor (morally, some kind of Cartier duality) D : QCohpBG q Ñ QCohpZq as follows: it is a standard calculation that QCohpBG q  QCohpptq. For V P QCohpBG q, we m m nPZ À À denote by V the summand corresponding to n P Z. De ne Dp V q  i V where i : Specpkq Ñ Z is n n n; n n the inclusion of the point tnu. Letting r : Z Ñ Specpk q denote the anization map, we de ne a functor : PropQCohpBG qq Ñ Z Z QCohpSpec k qq  k -mod via the composition PropDq Propr q ev Z Z PropQCohpBG qq PropQCohpZqq PropQCohpSpecpk qqq QCohpSpecpk qq: De nition 4.3.8. For any n P Z, there is a natural map  : Spec k Ñ Spec k de ned by projection to the nth coordinate. Denote by k   k P QCohpSpec k q, and note that k is projective since it is a summand n n; n Z Z of the free module decomposed by k  k ` p k q. We de ne the n-stalk of a k -complex M to be n m mn M  k b M and the n-costalk to be  M  Hompk ; Mq; since k is projective the underived functor is n n n n n the derived functor. Remark 4.3.9. The notion of costalk and stalk are canonically equivalent. It is a direct veri cation that !  Z in fact, i pV q  i pV q  V . Let p P k denote the element with 0 in the nth component and 1 everywhere n n else. Let e denote the element with 1 in the nth component and 0 everywhere else. Then, p e  1, so n n n we nd that V  p V ` e V , and i pV q  kerpp q while i pV q  cokerpp q. n n n n Composing with the global sections functors recovers the underlying vector space, and composing with the costalk at n P Spec k recovers the nth homogeneous part. In particular, if we are interested in studying To see this, note that objects of QCohpBG q are chain complexes which are OpG q-coalgebras, i.e. equipped with a map m m V Ñ V b krz; z s. In particular, tensors have nite rank, so any vector can only have nitely many homogeneous parts. 35 the underlying vector space of V P PropQCohpBG qq via its homogeneous components, it suces to consider it as an object of QCohpSpecpk qq. Proposition 4.3.10. We have commutative diagrams of functors Z Z PropQCohpBG qq k -mod PropQCohpBG qq k -mod m m Z ! pSpec k ;q evPropqpBG ;bL q evPropp q m n n Vect Vect k k Proof. We factor the rst diagram as follows Propr Dq ev Z Z PropQCohpBG qq Propk -modq k -mod Propq Propp q ev Propk -modq k -mod and the second in the analogous way. So, the proposition follows from two claims: (1) that the diagrams Z ! above commute without the Pro, i.e. p  pSpec k ;q r  D and pBG ;q    r  D, and (2) that Z Z the evaluation functor Propk -modq Ñ k -mod commutes with global sections and taking costalks. The rst claim can be directly veri ed: it suces to consider abelian categories since all functors are ~ 1 exact. In particular, if V P QCohpBG q , then it is a krz; z s-comodule, i.e. there is a map 1 n V Ñ V b krz; z s  V z : The functor D takes V to the complex on Z whose value on open ane tnu P Z is V . The functor r takes DpV q to V where k acts in the natural way. Finally, we see that the global sections are exactly À À À p V  V and the costalk i p V q  Hom Zpk ; V q  V . n n n n n n k n n Since the evaluation functor is a right adjoint, we prove the second claim by showing that both the global sections functor and the costalks functor are right adjoints, and then using the general fact that right adjoints commute. The global sections functor is right adjoint to the restrictions of scalars functor (it is also left adjoint to the \corestriction of scalars" functor R Hom pk ;q). The costalks functor is right adjoint to the pushforward (and in this case is equal to the stalks functor, which is a left adjoint). Remark 4.3.11. In fact, since the global sections and (co)stalk functors are both left and right adjoints in the above situation, the above construction and proposition are valid for any iteration of taking Pro and Ind categories. De nition 4.3.12. Let V be a k -module. The support of v P V is the closed subscheme de ned by the annihilator ideal of v. Lemma 4.3.13. A pro-graded isomorphism is injective. Proof. This is the easy fact that if a map of sheaves on Specpk q is zero on stalks at closed points, then it is zero, and the observation that on Spec k , costalks and stalks coincide. Lemma 4.3.14. Suppose that f : V Ñ W is a pro-graded isomorphism of pro-graded vector spaces such that either V or W are supported at nitely many weights. Then f is an isomorphism on underlying vector spaces. More generally, let A be a sheaf of algebras on Specpk q, and f : V Ñ W a pro-graded isomorphism of sheaves of A-modules where W is generated by elements supported at nitely many weights. Then, f is an isomorphism on underlying vector spaces. 36 Proof. For the rst claim, the assumptions of the proposition imply that V and W have nite support, whose points consist entirely of closed points of k . A map being a pro-graded isomorphism means that it is an isomorphism at stalks of closed points. For the second more general claim, note that if W is an A-module, and w P W is an element of nitely many weights, say w  w  w where the w are homogeneous of weight c , then w P A  w since 1 r i i i e w  w , where e P k is the characteristic function at i P Z. In particular, W having a set of generators c i c i i supported at nitely many weights is equivalent to W having a set of homogeneous generators. Now, if f : V Ñ W is a pro-graded isomorphism, then it is injective by the previous lemma. For surjectivity, note that for a given homogeneous w P W , since f is a pro-graded isomorphism, we have a homogeneous v P V such that fpvq  w, and surjectivity follows since W is generated by homogeneous elements. The following lemma allows us to reduce statements in the derived category to statements in the abelian category. Lemma 4.3.15. A map f : V Ñ W is a pro-graded quasi-isomorphism of pro-graded complexes if and only i i i if each of the H pfq : H pV q Ñ H pWq are pro-graded isomorphisms of modules. Proof. We can take a map f : V Ñ W of complexes of k -modules. It is easy to verify that taking nth homogeneous parts (i.e. talking stalks via localization) is exact, so that if H pfq is a pro-graded isomorphism, it is an isomorphism of modules and therefore f is a quasi-isomorphism. Conversely, the global sections functor is clearly exact. De nition 4.3.16. A pro-graded dg-algebra is an object of PropAlgpQCohpBG qqq. If A is a pro-graded dg-algebra, then SpecpAq is naturally a dg-indscheme with a G -action in the sense of [GR14]. We will use the word ind-stack to mean a prestack which can be written as an inductive limit of closed embeddings of (derived) QCA stacks (in the sense of [DG13]); in practice we only need the case of a formal completion of a closed substack of a quotient stack. Recall the de nition of a contracting G -action in De nition 2.1.18. ¤0 Lemma 4.3.17. Let A be a noetherian weight Z pro-graded connective dg-algebra, which is generated in wt 0 negative weights over its weight 0 part, and let I   pA q €  pAq be the classical augmentation ideal. 0 0 The derived completion A Ñ A is a pro-graded quasi-isomorphism. Globally, if X is an ind-stack with a representable contracting G -action with xed point locus Z € X , then O Ñ O is a pro-graded quasi- m X y isomorphism of quasicoherent sheaves on X . In particular, OpXq Ñ OpX q is a pro-graded isomorphism. Proof. Choose generators f ; : : : ; f of I . By Proposition 2.4.8 we can compute the homotopy limit A via 1 r I the limit A  lim A b K pAq n n n where K is the Koszul complex for f ; : : : ; f P  pAq. Since f ; : : : ; f are of strictly negative weight, for 0 1 r n 1 r wt¥k wt¥k large n, pK q  p pAqq for any k. Furthermore, homotopy limits can be computed in the derived category of k-complexes, and in particular we can compute the homotopy limit on each graded piece. Thus, wtk wtk pA q is a computed by a limit which stabilizes at A , proving the claim. For the global claim where X is an ind-scheme, one can pass to an open ane G -closed cover (which exists since G is a torus). For m m the global claim where X is an ind-stack, one can check the equivalence on a cover of X . Recall that by Remark 6.11 of [BN12] that there are embeddings LpX{Gq ãÑ L pX{Gq ãÑ T r1s. Thus, ¤0 formal loops and unipotent loops inherit compatible G -actions and their functions are Z pro-graded. We have the following. Lemma 4.3.18. Let X be a geometric stack. The map induced by pullback OpL Xq OpLXq 37 is a pro-graded (quasi-)isomorphism. Proof. An argument is outlined in Corollary 2.7 of [BN13a]; we will repeat it for convenience. By Lemma 2.1.12, LX is geometric. The formal loops LX are the completion of the unipotent loops L X along constant loops, and the action is contracting by Lemma 2.1.19. The statement follows by Lemma 4.3.17. Example 4.3.19. The pro-graded isomorphism of Lemma 4.3.18 may fail to be an isomorphism. For example, take X  G{U where U is any unipotent subgroup of G; then we have that LppG{Uq{Gq  LpBUq  U{U Ñ LpBGq  G{G has image inside the unipotent cone of G. In particular, L pBUq  LpBUq  U{U LpBUq  p u{U: For example, if U  BG , then the map is p p OpL pBG qq  OpG  BG q  krx; s Ñ OpLpBG qq  OpG  BG q  krrxssrs a a a a a a is a pro-graded isomorphism but not an isomorphism, where |x|  0 is a generator for OpG q and ||  1 is a generator for OpBG q. Using the fact that the map is a pro-graded isomorphism, we can show in the case of a unipotent group that the map on Tate-equivariant functions is an isomorphism by a niteness argument. Essentially, we show that applying the Tate construction collapses enough of the target to produce an isomorphism. We include the following proposition as an easy precursor to the next one; it is not required in future arguments. Corollary 4.3.20. Let U be a unipotent algebraic group, and X a quasicompact algebraic space with a U -action. Then, the natural map induced by pullback u Tate Tate OpL pX{Uqq OpLpX{Uqq is an isomorphism. In particular, taking X  pt, u Tate Tate OpL pBUqq  OpU{Uq OpLpBUqq  kppuqq is an isomorphism. Proof. By Lemma 4.3.18, the map is a pro-graded isomorphism. Furthermore, applying the Tate construction, we have that OpLpX{Uqq  H pX ; kqppuqq since U is contractible, where u has cohomological degree 2 and weight 1. The statement follows from Lemma 4.3.14: for quasicompact algebraic spaces X , H pX ; kq is nite-dimensional and therefore has has a homogeneous basis. A tweaking of the above argument gives us the reductive case. Corollary 4.3.21. Let G be a reductive algebraic group, and X a quasicompact algebraic space with a G- action. Then, the natural map induced by pullback u Tate Tate OpL pX{Gqq OpLpX{Gqq is an isomorphism. Proof. Let h denote the universal Cartan of the Lie algebra g of G, and W the universal Weyl group acting on h. By Proposition 4.2.6 we have Tate ! W OpLpBGqq  C pBG; kq b ppuqq  krrhss ppuqq dR 38 W where h is in cohomological degree zero and subcomplex H pBG; kq is given by krhus . By Proposition 2.1.25, u x and the identi cation G{{G  T{{W for reductive groups G, we have that L pBGq  LpBGq t0u, and T{{W in particular by base change, OpL pBGqq  Opt0u u Tate Tate W so that OpL pBGqq  OpLpBGqq  krrhss ppuqq. In particular, u Tate Tate OpL pX{Gqq Ñ OpLpX{Gqq W Z is a map of module objects over the algebra object krrhss ppuqq in the category QCohpSpecpk qq, where h has 0 W W weight -1 and cohomological degree 0, and its cohomology groups are linear over H pkrrhss ppuqqq  krrhss . We claim that the cohomology groups of the target i Tate i ! H pOpLpX{Gqq q  H pC pX{G; kq b kppuqqq dR are nitely generated over krrhss by weight-homogeneous generators; assuming the claim, the result follows from Lemma 4.3.17 and Lemma 4.3.14. To see the claim, we trace through the identi cation of Proposition 4.2.6, keeping track of the weights. The cotangent complex L has weight -1 and u has weight 1 by convention. Thus, we nd that under the identi cation of Proposition 4.2.6, C pX{G; kq has weight 0 and u has weight 1. Therefore, it suces to show dR that the ring H pX{G; kq is nitely generated over H pBG; kq by nitely many cohomologically homogeneous generators. This can be observed via the derived Cartan model for equivariant cohomology, which computes H pX{G; kq via a double complex whose E page is H pXq b krrgss , completing the claim. 1 k We can now prove Theorem 4.3.1, which we restate for convenience. Theorem 4.3.22. Let X be a quasicompact algebraic space with an action of an ane algebraic group G. The natural map induced by pullback u Tate Tate OpL pX{Gqq OpLpX{Gqq is an isomorphism. Proof. Every ane algebraic group G embeds as a subgroup of a reductive group K . Apply the previous G G corollary to pX  Kq{K . Note that X  K is not guaranteed to be a scheme, but is always an algebraic space. References [ABV92] Je rey Adams, Dan Barbasch, and David A. Vogan Jr. The Langlands classi cation and irreducible characters for real reductive groups. Vol. 104. Progress in Mathematics. Birkh auser Boston, Inc., Boston, MA, 1992, pp. xii+318. [AC12] Dima Arinkin and Andrei C ald araru. \When is the self-intersection of a subvariety a bration?" Adv. Math. 231.2 (2012), pp. 815{842. [ACH14] Dima Arinkin, Andrei C ald araru, and M arton Hablicsek. \Formality of derived intersections and the orbifold HKR isomorphism". arXiv: 1412.5233v1 (2014). [AF17] David Ayala and John Francis. \The cobordism hypothesis". arXiv: 1705.02240v2 (2017). [AFR18] David Ayala, John Francis, and Nick Rozenblyum. \Factorization homology I: Higher categories". Adv. Math. 333 (2018), pp. 1042{1177. 39 [AMR17] David Ayala, Aaron Mazel-Gee, and Nick Rozenblyum. \Factorization homology of enriched 8- categories". arXiv: 1710.06414v1 (2017). [AS69] M. F. Atiyah and G. B. Segal. \Equivariant K -theory and completion". J. Di erential Geometry 3 (1969), pp. 1{18. [Beh04] K. Behrend. \Cohomology of stacks". Intersection theory and moduli. ICTP Lect. Notes, XIX. Abdus Salam Int. Cent. Theoret. Phys., Trieste, 2004, pp. 249{294. [Bei08] A. Beilinson. \Remarks on topological algebras". Mosc. Math. J. 8.1 (2008), pp. 1{20, 183. [BF08] Ragnar-Olaf Buchweitz and Hubert Flenner. \The global decomposition theorem for Hochschild (co-)homology of singular spaces via the Atiyah-Chern character". Adv. Math. 217.1 (2008), pp. 243{281. [BFN10] David Ben-Zvi, John Francis, and David Nadler. \Integral transforms and Drinfeld centers in derived algebraic geometry". J. Amer. Math. Soc. 23.4 (2010), pp. 909{966. [BG94] Jonathan Block and Ezra Getzler. \Equivariant cyclic homology and equivariant di erential forms". Ann. Sci.  ecole Norm. Sup. (4) 27.4 (1994), pp. 493{527. [Bha12] Bhargav Bhatt. \Completions and derived de Rham cohomology". arXiv: 1207 . 6193v1 (2012). [BN12] David Ben-Zvi and David Nadler. \Loop spaces and connections". J. Topol. 5.2 (2012), pp. 377{430. [BN13a] David Ben-Zvi and David Nadler. \Loop spaces and representations". Duke Math. J. 162.9 (2013), pp. 1587{1619. [BN13b] David Ben-Zvi and David Nadler. \Nonlinear Traces". arXiv: 1305.7175v3 (2013). [BS15] Bhargav Bhatt and Peter Scholze. \The pro- etale topology for schemes". Ast erisque 369 (2015), pp. 99{201. [CG10] Neil Chriss and Victor Ginzburg. Representation theory and complex geometry. Modern Birkh auser Classics. Reprint of the 1997 edition. Birkh auser Boston, Inc., Boston, MA, 2010, pp. x+495. [CK01] Ionut  Ciocan-Fontanine and Mikhail Kapranov. \Derived Quot schemes". Ann. Sci.  ecole Norm. Sup. (4) 34.3 (2001), pp. 403{440. [Coh16] Lee Cohn. \Di erential Graded Categories are k-linear Stable In nity Categories". arXiv:1308.2587v2 (2016). [DG13] Vladimir Drinfeld and Dennis Gaitsgory. \On some niteness questions for algebraic stacks". Geom. Funct. Anal. 23.1 (2013), pp. 149{294. [Dr e04] Jean-Marc Dr ezet. \Luna's slice theorem and applications". Algebraic group actions and quotients. Hindawi Publ. Corp., Cairo, 2004, pp. 39{89. [GG99] Ezra Getzler and Paul Goerss. \A Model Category Structure for Di erential Graded Coalgebras". https://ncatlab.org/nlab/files/GetzlerGoerss99.pdf (1999). [GKM98] Mark Goresky, Robert Kottwitz, and Robert MacPherson. \Equivariant cohomology, Koszul du- ality, and the localization theorem". Invent. Math. 131.1 (1998), pp. 25{83. [GR14] Dennis Gaitsgory and Nick Rozenblyum. \DG indschemes". Perspectives in representation theory. Vol. 610. Contemp. Math. Amer. Math. Soc., Providence, RI, 2014, pp. 139{251. 40 [GR17] Dennis Gaitsgory and Nick Rozenblyum. A study in derived algebraic geometry. Vols. I. and II. Vol. 221. Mathematical Surveys and Monographs. American Mathematical Society, Providence, RI, [Har75] Robin Hartshorne. \On the De Rham cohomology of algebraic varieties". Inst. Hautes  etudes Sci. Publ. Math. 45 (1975), pp. 5{99. [Hoy18] Marc Hoyois. \The homotopy xed points of the circle action on Hochschild homology". arXiv: 1506.07123v2 (2018). [HP16] Daniel Halpern-Leistner and Daniel Pomerleano. \Equivariant Hodge theory and non-commutative geometry". (2016). arXiv: 1507.01924. [Jon87] John D. S. Jones. \Cyclic homology and equivariant homology". Invent. Math. 87.2 (1987), pp. 403{423. [Kal15] Dmitry Kaledin. \Co-periodic cyclic homology". (2015). arXiv: 1509.08784v1. [Kel06] Bernhard Keller. \On di erential graded categories". International Congress of Mathematicians. Vol. II. Eur. Math. Soc., Zuric  h, 2006, pp. 151{190. [KL87] David Kazhdan and George Lusztig. \Proof of the Deligne-Langlands conjecture for Hecke algebras". Invent. Math. 87.1 (1987), pp. 153{215. [Lod92] Jean-Louis Loday. Cyclic homology. Vol. 301. Grundlehren der Mathematischen Wissenschaften. Appendix E by Mar a O. Ronco. Springer-Verlag, Berlin, 1992, pp. xviii+454. [Lur09a] Jacob Lurie. \Derived Algebraic Geometry IV: Deformation Theory". arXiv:0709.3091v2 (2009). [Lur09b] Jacob Lurie. Higher Topos Theory. Vol. 170. Annals of Mathematics Studies. Princeton University Press, Princeton, NJ, 2009, pp. xviii+925. isbn: 978-0-691-14049-0; 0-691-14049-9. [Lur09c] Jacob Lurie. \On the classi cation of topological eld theories". Current developments in mathe- matics, 2008. Int. Press, Somerville, MA, 2009, pp. 129{280. [Lur11a] Jacob Lurie. \Derived Algebraic Geometry VII: Spectral Schemes". (2011). url: http://www.math. harvard.edu/ lurie/. [Lur11b] Jacob Lurie. \Derived Algebraic Geometry VIII: Quasi-Coherent Sheaves and Tannaka Duality Theorems". (2011). url: http://www.math.harvard.edu/ lurie/. [Lur11c] Jacob Lurie. \Derived Algebraic Geometry XII: Proper Morphisms, Completions, and the Grothendieck Existence Theorem". (2011). url: http://www.math.harvard.edu/ lurie/. [Lur17] Jacob Lurie. Higher Algebra. 2017. url: http://www.math.harvard.edu/ lurie/. [Lur18] Jacob Lurie. Spectral Algebraic Geometry. 2018. url: http://www.math.harvard.edu/ lurie/. [NS18] Thomas Nikolaus and Peter Scholze. \On topological cyclic homology". arXiv:1707.01799v2 (2018). [Pre11] Anatoly Preygel. \Thom-Sebastiani & Duality for Matrix Factorizations". arXiv:1101.5834v1 (2011). [Pre15] Anatoly Preygel. \Ind-coherent complexes on loop spaces and connections". Stacks and categories in geometry, topology, and algebra. Vol. 643. Contemp. Math. Amer. Math. Soc., Providence, RI, 2015, pp. 289{323. 41 [RV18] Emily Riehl and Dominic Verity. Elements of 8-Category Theory. 2018. url: http://www.math. jhu.edu/ eriehl/elements.pdf. [Sta18] The Stacks Project Authors. Stacks Project. url: http://stacks.math.columbia.edu. 2018. [Ste75] Robert Steinberg. \On a theorem of Pittie". Topology 14 (1975), pp. 173{177. [Tho86] R. W. Thomason. \Lefschetz-Riemann-Roch theorem and coherent trace formula". Invent. Math. 85.3 (1986), pp. 515{543. [Tho87] R. W. Thomason. \Algebraic K -theory of group scheme actions". Algebraic topology and algebraic K -theory (Princeton, N.J., 1983). Vol. 113. Ann. of Math. Stud. Princeton Univ. Press, Princeton, NJ, 1987, pp. 539{563. [Tho88] R. W. Thomason. \Equivariant algebraic vs. topological K -homology Atiyah-Segal-style". Duke Math. J. 56.3 (1988), pp. 589{636. [To e06] Bertrand To en. \Champs anes". Selecta Math. (N.S.) 12.1 (2006), pp. 39{135. [To e07] Bertrand To en. \The homotopy theory of dg-categories and derived Morita theory". Invent. Math. 167.3 (2007), pp. 615{667. [To e14] Bertrand To en. \Derived algebraic geometry". EMS Surv. Math. Sci. 1.2 (2014), pp. 153{240. [TV05] Bertrand To en and Gabriele Vezzosi. \Homotopical algebraic geometry. I. Topos theory". Adv. Math. 193.2 (2005), pp. 257{372. [TV08] Bertrand To en and Gabriele Vezzosi. \Homotopical algebraic geometry. II. Geometric stacks and applications". Mem. Amer. Math. Soc. 193.902 (2008), pp. x+224. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Mathematics arXiv (Cornell University)

Equivariant localization and completion in cyclic homology and derived loop spaces

Chen, Harrison
Mathematics , Volume 2023 (1708) – Aug 21, 2017
Download PDF
42 pages

Loading next page...
 
/lp/arxiv-cornell-university/equivariant-localization-and-completion-in-cyclic-homology-and-derived-VTZvFCp5jT

References (56)

ISSN
0001-8708
eISSN
ARCH-3343
DOI
10.1016/j.aim.2020.107005
Publisher site
See Article on Publisher Site

Abstract

We prove an equivariant localization theorem over an algebraically closed eld of characteristic zero for smooth quotient stacks by reductive groups X{G in the setting of derived loop spaces as well as Hochschild homology and its cyclic variants. We show that the derived loop spaces of the stack X{G z z and its classical z- xed point stack  pX q{G become equivalent after completion along a semisimple parameter rzs P G{{G, implying the analogous statement for Hochschild and cyclic homology of the dg category of perfect complexes PerfpX{Gq. We then prove an analogue of the Atiyah-Segal completion theorem in the setting of periodic cyclic homology, where the completion of the periodic cyclic homology of PerfpX{Gq at the identity res P G{{G is identi ed with a 2-periodic version of the derived de Rham cohomology of X{G. Together, these results identify the completed periodic cyclic homology of a stack X{G over a parameter rzs P G{{G with the 2-periodic derived de Rham cohomology of its z- xed points. Contents 1 Introduction 1 1.1 Conventions and notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 1.2 Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2 Background 7 2.1 Derived loop spaces and its variants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.2 Cyclic homology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 2.3 The equivariant cyclic bar construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 2.4 Formal and derived completions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 3 An equivariant localization theorem in derived loop spaces and Hochschild homology 22 3.1 Equivariant localization for derived loop spaces . . . . . . . . . . . . . . . . . . . . . . . . . . 22 3.2 Equivariant localization for Hochschild and cyclic homology . . . . . . . . . . . . . . . . . . . 27 4 An Atiyah-Segal completion theorem for periodic cyclic homology 29 4.1 Twisted circle actions on loop spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 4.2 Tate-equivariant functions on formal loop spaces compute analytic de Rham cohomology . . . 31 4.3 Comparing global functions on unipotent and formal loop spaces . . . . . . . . . . . . . . . . 34 1 Introduction Derived loop spaces appear naturally in questions of Hochschild homology. When X is a prestack such that the derived categories QCohpXq and IndCohpXq are compactly generated by PerfpXq and CohpXq respectively, their Hochschild homology can be computed in two ways. On one hand, it is computed by the arXiv:1708.06079v7 [math.AG] 23 Jan 2020 usual cyclic bar complex, but it can also be naturally identi ed with the global sections of a sheaf of functions or distributions on the derived loop space LX of X [BN13b]: HHpQCohpXqq  OpLXq; HHpIndCohpXqq  !pLXq: When X is a derived scheme, the loop space LX is equivalent to the shifted odd tangent bundle via a derived variant of the Hochschild-Kostant-Rosenberg theorem, re ecting the Zariski-local nature of derived loops. Introducing the S -action, results in [BN12] [Pre15] identify periodic cyclic homology with 2-periodic derived de Rham (co)homology: Tate Tate HPpPerfpXqq  OpLXq  C pX ; kqppuqq; HPpCohpXqq  !pLXq  C pX ; kqppuqq ;dR dR 2 1 where u P C pBS ; kq is the degree two universal Chern class. Furthermore, over k  C, the main result in [Bha12] identi es derived de Rham cohomology with Betti de Rham cohomology of the analyti cation, and an we have that HPpPerfpXqq  C pX ; kqppuqq. Our goal in this paper is to investigate the periodic cyclic dR homology where X is taken to be a smooth quotient stack. A diculty in understanding the Hochschild homology of stacks which does not appear for schemes is the failure of derived loop spaces to form a cosheaf for the smooth topology; informally, Hochschild homology is Zariski local but not smooth local . On the other hand, as argued in [BN12], sheaves and functions on formal loop spaces (i.e. loop spaces completed at constant loops) can be computed via a smooth covers. We begin to bridge this gap by understanding the global loop space LpX{Gq after completion over bers of the map to LpBGq  G{G as well as to its anization G{{G. We show that completion over G{G gives a certain formal loop space and completion over G{{G gives a certain unipotent loop space, formalizing a \Jordan decomposition for loop spaces" which appeared in [BN13a], where it was applied to realize a moduli stack of Langlands parameters for representations of real reductive groups. Informally, we view X{G as a family of formal loop spaces over G{G, which we in turn view as a family of unipotent loop spaces over G{{G, ultimately realizing X{G as a family of unipotent loop spaces over the ane parameter space G{{G. This paper contains two main theorems. The rst realizes well-established equivariant localization pat- terns (e.g. as in [GKM98] [Tho87] [Tho86]) in the setting of Hochschild homology via its geometric avatar, the derived loop space. The second realizes an Atiyah-Segal style (e.g. as in [AS69] [Tho88]) completion theorem identifying completed periodic cyclic homology with 2-periodic equivariant derived de Rham cohomology. Over C a theorem of Bhatt [Bha12] gives an identi cation with the Betti cohomology of the analyti cation. We mention a few related results in the literature. An analogous theory was explored by Block and Getzler in [BG94] in the setting of a compact group G acting on a compact smooth manifold M . Similar results appear in the algebraic setting when G is nite (i.e. X{G is a Deligne-Mumford stack) in Theorem 1.15 of [ACH14], and in the case of an smooth ane quotient stack in Lemma 4.11 and Proposition 4.12 of [HP16]. We generalize their statements to the case of a general smooth quasi-projective quotient stack. Our statements have also been investigated in the setting of smooth quotient stacks with nitely many orbits in [BN13a], with special attention to the case BzG{B in Theorem 3.5 of op. cit. We will now begin stating our results precisely. The following theorem is an abridged version of our geometric equivariant localization theorem for derived loop spaces, which appears in the main text as Theorem 3.1.12. Theorem A (Equivariant localization for derived loop spaces). Let G be a reductive group acting on a smooth variety X over an algebraically closed eld k of characteristic 0, and z P G a semisimple element. Note that LpX{Gq naturally lives over LpBGq; let L pX{Gq denote the completion of LpX{Gq along the ber over the semisimple orbit tG  zu{G € LpBGq and let L pX{Gq denote the completion over the saturation This depends on de nitions. Our notion of Hochschild homology is the one that is computed by a cyclic bar complex of a small dg category. Other authors have considered a variant which is obtained by extending the notion of Hochschild homology of ane schemes to stacks via at descent, which is smooth local by de nition and is equivalent to the a completed version of the former. 2 1 rzs P G{{G. For X a smooth variety with a G-action, there are functorial S -equivariant isomorphisms z z u z z u p p L p pX q{G q L pX{Gq; L p pX q{G q L pX{Gq: z 0 z 0 z z We remark brie y on the assumptions of the theorem. We assume that k is algebraically closed and that z is semisimple in reductive G since we argue via the Luna slice theorem, and we work in characteristic 0 since we model derived schemes locally as dg algebras. The assumption that X is smooth is used to compute via Koszul resolutions, and is essential since the statement is false otherwise. We note that in the case that G is a torus, these statements can be strengthened: the above maps are equivalences on a Zariski open neighborhood over z, recovering the equivariant localization in [CG10] for K -theory in the setting of periodic cyclic homology. However, for a nonabelian reductive group G, this fails even in the case when X is a point (see Remark 3.1.14 and Example 3.1.3). Our result implies the following interpretation of derived xed points, which also appears in [ACH14] as Corollary 1.12. Corollary 1.0.1. Let G be a reductive group acting on a smooth variety X and z P G semisimple. We have a natural identi cation of the derived z- xed points: z z Lp pX qq X : LpX{Gq  tzu: 0 BG z z 1 Proof. The \shift by z" map on Lp pX q{G qq is a (non-S -equivariant!) equivalence, so that z z z p p L p pX q{Gqq  Lp pX q{G q; z 0 0 z z and in particular since loop spaces commuted with ber products, X  Lp pX qq. The unipotent version of equivariant localization for loop spaces implies an equivariant localization result for Hochschild homology and its cyclic variants. The following appears in the main text as Theorems 3.2.3 and 3.2.10. Corollary 1.0.2 (Equivariant localization for Hochschild homology). Let G be reductive group acting on a smooth variety X , and z P G a semisimple element. Then we have an S -equivariant equivalence z z HHpPerfpX{Gqq HHpPerfp pX q{G qq zp 0 zp and similarly when replacing HH with its cyclic variants HC; HN; and HP . Note that since the formation of periodic cyclic homology involves a colimit, this is not automatic in that case. The fact that X is smooth gives us a cohomological boundedness of Hochschild homology, which is essential in establishing the above result. After identifying the completed derived loop spaces over a central character z, we are interested in identifying this completion with de Rham cohomology via an analogue of the Atiyah-Segal completion theorem in the setting of periodic cyclic homology. The following theorem is a consequence of Proposition 4.2.6 and Theorem 4.3.1. Theorem B (Atiyah-Segal completion for periodic cyclic homology). Let X{G be a global quotient stack where X is an algebraic space. Then, there is an equivalence HPpPerfpX{Gqq C pX{G; kqb kppuqq ep dR k where C pX{G; kq denotes the derived de Rham cohomology, and b indicates completion with respect to the dR coarsest topology induced by the derived Hodge ltration and u-adic ltration on respective tensor factors. 3 When k  C, applying a generalization of the main theorem of [Bha12] to geometric stacks, we can identify Tate functions on formal loop spaces with Betti cohomology. Corollary 1.0.3. Let X{G be a nite type global quotient stack over k  C, where X is an algebraic space. Then, there is an equivalence an an HPpPerfpX{Gqq C pX {G ;Cqb Cppuqq ep dR an an where C pX {G ;Cq denotes de Rham cohomology of the analyti cation and b indicates completion with dR respect to the coarsest topology induced by the Hodge ltration in the Cartan model for equivariant cohomology and u-adic ltration on respective tensor factors. The main technical hurdle in the proof of the theorem is that the ber of LpX{Gq over res P G{{G does not just contain formal loops but also unipotent loops. In Theorem 4.3.1 we show that this di erence vanishes after applying the Tate construction. Theorem C. For X a quasicompact algebraic space acted on by an ane algebraic group G, the pullback functor on derived global functions induces an isomorphism u Tate Tate OpL pX{Gqq OpLpX{Gqq : In particular, if U is a unipotent group, then the pullback functor induces an equivalence HPpPerfpX{Uqq HPpPerfpXqq: Following an analysis of the twisted S -rotation action on LpX{Gq above points of G{G away from the identity in Section 4.1, we relate this completion theorem to our localization theorem in Theorem 4.3.2. Theorem D. Let G be a reductive group acting on a smooth quasi-projective variety X . The periodic cyclic homology HPpPerfpX{Gqq is naturally a module over HPpPerfpBGqq  krG{{Gsppuqq. For a closed point z P G{{G, we have an identi cation of the formal completion at z with 2-periodic Betti cohomology of the z- xed points z z HPpPerfpX{Gqq C p pX q{G ; kqb kppuqq zp k dR as a module over HPpPerfpBGqq  C pBG ; kqb kppuqq, contravariantly functorial with respect to X . zp k dR Using Corollary 1.0.1, we also obtain an identi cation of the (derived) specialization of periodic cyclic homology at z P G with non-equivariant cohomology of z- xed points. Corollary 1.0.4. Let X be a smooth variety with an action of a reductive group G. For z P G semisimple, let k denote the skyscraper sheaf at rzs P G{{G. We have an equivalence L z HPpPerfpX{Gqq b k ppuqq C p pX q; kqppuqq: z 0 krG{Gsppuqq dR We indicate two natural directions in which our results may be extended. The rst is to ask what happens when X is allowed to be singular; in this case, one can study either the Hochschild homology of CohpXq or PerfpXq. In the case when X is a ber product of smooth schemes, the geometric statement follows immediately from our results, but it is unclear to us how to generalize beyond that case. If a geometric statement is out of reach, it is also of interest as to whether the global localization statements for periodic cyclic homology hold; an obstruction to applying the standard techniques of embedding a singular quotient stack into a smooth one is the lack of a devissage theorem for the periodic cyclic homology of stacks. A 4 second direction would be to categorify these results in a generalization of the Koszul duality of [Pre15] and [BN12]. In addition, it would be pleasing to have a more conceptual proof of Theorem 4.3.1. We end our introduction with a few toy examples. 1 1 Example 1.0.5. Let G  G  Spec krz; z s act on X  A  Spec krxs by scaling, i.e. assign G -weight m m |x|  1. The loop space can be calculated directly via Proposition 2.1.8 Spec krz; z ; xs{xxpz  1qy LpA {G q  : The Hochschild homology and periodic cyclic homology can also be calculated directly 1 1 G 1 1 1 HHpPerfpA {G qq  pkrz; z ; xs{xpz  1qq  krz; z s; HPpPerfpA {G qq  krz; z sppuqq m m as the S -equivariant structure on a complex concentrated in a single cohomological degree can only be realized by the zero map. Completing at any z P G gives, for t  z  z and |t|  0, 0 m 0 HPpPerfpA {G qq  krrtssppuqq m zp where t  z  z . z an On the other hand, we can compute H ppX q ; kqppuqq for each z . For z  z  1, the xed points 0 0 z z 1 1 ! ! 0 0 pX q{G  t0u{G  BS , whose 2-periodic cohomology is H pBS ; kq b kppuqq  krrsss b kppuqq 0 m 1 1 1 with |s|  2. For z  1 the xed points are A {G  C{S  BS , and the same argument applies. The identi cation krrsssppuqq  krrtssppuqq is by tu  s; in particular it is necessary to invert the degree 2 operator u. The discrepancy between the cohomological degrees of t and s is a manifestation of the Koszul duality degree-weight shearing discussed in [BN12]. Example 1.0.6 (Flag variety). Let X  G{B be the ag variety with the usual action of G. Then, X{G  BB, so LpX{Gq  B{B  G{G is the Grothendieck-Springer resolution; the ber for the map G Ñ G over any point g P G consists of the Borel subgroups containing g, i.e. the g- xed points of G{B. We identify the Hochschild homology HHpPerfpBBqq as a HHpPerfpBGqq-module by the inclusion map HHpPerfpBBqq  OpG{Gq  krHs Ð HHpPerfpBGqq  OpG{Gq  krHs where H is the universal Cartan subgroup and W is the universal Weyl group . Let s P G be a semisimple element, and rss its adjoint orbit. Completing at rss P krHs , we have W s HHpPerfpBBqq  krrhss Ð HPpPerfpBGqq  krrhss : x sp rss |Ws| In particular, the rank of HPpPerfpBBqq over HPpPerfpBGqq is |W  s|  |W s|  |W| by a theorem sp sp G of Steinberg and Pittie [Ste75]. Note that h is placed in cohomological degree zero. Applying the Tate construction, we nd that W s HPpPerfpBBqq  krrhss ppuqq: rss On the other hand, the xed points pG{Bq consist of Borel subgroups containing s; by conjugating, we s s can choose a torus such that s P T € B; let t be the Lie algebra of T . There is a G -action on pG{Bq and s s its stabilizer at every point is conjugate to B , but the action may not be transitive; thus, pG{Bq is the s s disjoint union of copies of G {B . To count the number of connected components, we count T - xed points: s s the T - xed points of G{B are also s- xed points, and furthermore each G {B contains |W | such T - xed It is known that OpGq has vanishing higher cohomology. 5 points, so we have |W|{|W s| connected components. Finally, accounting for T -equivariance, we have s an s an W s H p BpB q ; kq  krrhss Ð H pBpG q ; kq  krrtss |W|{|Ws| |W|{|W | where t is placed in cohomological degree 2. In particular, W s W s s an ! G G HPpPerfpBBqq  krrhss ppuqq  krrtss ppuqq  H ppBB q ; kq b kppuqq rss under the isomorphism uh  t. 1.1 Conventions and notation In this note, k will denote an algebraically closed eld of characteristic zero, and we work over pt  Specpkq. Unless otherwise stated, all functors and categories are derived, e.g. for an ane scheme X  SpecpAq, we denote by QCohpXq the derived category of unbounded complexes of A-modules localized with respect to quasi-isomorphisms, and b  b although we sometimes use the latter notation for emphasis (e.g. when performing calculations). We indicate a functor that is not derived by writing  or H . All gradings follow cohomological grading conventions (i.e. di erentials increase degree), unless otherwise indicated by a subscript, and HH will always denote the cochain complex of Hochschild chains rather than its cohomology groups, which we denote H pHHq. We refer to the nth cohomology group of a chain complex V by H pV q   pV q. We require a theory of 8-categories and derived algebraic geometry. Such theory has been developed by by To en and Vezzosi in [TV05] [TV08] and by Lurie in [Lur17] [Lur09b] [Lur18] [Lur09a] [Lur11a] [Lur11b] [Lur11c]. Below, we summarize some of the main de nitions. Remark 1.1.1 (8-categories). By 8-category we mean an p8; 1q-category, and we do not specify a particular model . We let S denote the 8-category of 8-groupoids or spaces and we will take for granted that the category of 8-categories is enriched in S. We let st denote the 8-category of small (stable) k-linear 8-categories whose 1-morphisms are k-linear exact functors, and Pr the category of presentable (stable) k-linear 8-categories whose 1-morphisms are functors which are k-linear exact left adjoints . Note that presentable 8-categories admit a combinatorial model structure. For such a category C P Pr , we let C P st denote its compact objects. For C P st , we let k k L L IndpCq P Pr denote its ind-completion. By [Coh16], a presentable k-linear 8-category in Pr has an k k L R associated k-linear di erential graded category in dgcat . We will denote by Fun p;q and Fun p;q k k the spaces of k-linear exact functors which are left and right adjoints respectively. For more details, see Chapter 5 of [Lur09b], Section 1.4.4 in [Lur17], and Section 6 of [Lur11a]. Remark 1.1.2 (Derived stacks). We let DRng denote the 8-category of derived rings (or derived algebras over k); during our exposition we do not insist on a particular model, but we will always compute in the op category of dg algebras over k with its projective model structure. The opposite category A  DRng is de ned to be the category of ane derived schemes. A derived scheme is as a derived locally ringed space whose 0-truncation is a scheme and whose higher homotopy groups are quasicoherent [To e14] [Lur18]. The global sections functor and derived spectrum functors induce equivalences identifying A with the category of derived schemes whose  is ane in the classical sense. We will refer to derived schemes as simply schemes, and use the term classical scheme to refer to a derived scheme X for which  pXq  X . A dg scheme [CK01] or embeddable derived scheme is de ned somewhat di erently; it is de ned to be a 0 n scheme pX;O q along with a non-positively graded sheaf of complexes O such that O  O and H pO q X X X X X A forthcoming book by Riehl and Verity [RV18] establishes the model-independence of p8; 1q-categories and its foundational properties, constructions, and theorems. In particular, by Remark 6.5 in [Lur11a], k-linear presentable categories are automatically stable. By the adjoint functor theorem, left adjoint functors commute with ltered colimits. are quasicoherent. In particular, a dg scheme Z  pX;O q admits an embedding Z Ñ X into the classical scheme X  pX;O q. Every derived scheme is locally modeled by a dg scheme. op A prestack is an 8-functor A : DRng Ñ S, and a (derived) stack is a prestack which is a sheaf for the derived  etale topology [GR17] [TV08]. We mean (derived) algebraic stack in the sense of [DG13]: an (derived) Artin 1-stack whose diagonal is quasi-separated, quasi-compact, and representable by (derived) schemes and admits an atlas by a (derived) scheme. We mean geometric stack in the sense of [BN12]: an algebraic stack whose diagonal map is ane. We say an algebraic stack is quasi-compact if it admits a quasi-compact atlas U (equivalently, if it admits an ane atlas). A map of prestacks X Ñ Y is schematic if for any scheme S and map S Ñ Y , the base change X  S to S is a scheme. A map of derived schemes f : X Ñ Y is a closed immersion if it is in the classical sense on  ; a map of algebraic stacks is a closed immersion if it is after base change to an atlas. 1.2 Acknowledgements I would like to thank David Nadler for our numerous conversations regarding this topic and for suggestions leading to and during the writing of this paper. I would also like to thank Daniel Halpern-Leistner for suggestions regarding the equivariant cyclic bar complex, David Ben-Zvi for suggestions regarding twisted circle actions and Bhargav Bhatt for explaining how his results in [Bha12] extend to the case of stacks. Finally, I would like to thank Shishir Agrawal for numerous helpful discussions, and Brian Hwang for his comments on this paper. This work was mostly carried out at UC Berkeley, in part at Cornell University, and was partially supported by NSF RTG grant DMS-1646385. 2 Background In this section we provide some basic exposition on Hochschild homology, loop spaces and derived algebraic geometry. At parts it is an informal summary of the existing literature, and at parts we provide proofs of some folklore likely known to experts. 2.1 Derived loop spaces and its variants An in-depth discussion of derived loop spaces, which we often simply refer to as loop spaces, can be found in [BN12]. We will summarize the main de nitions and prove some foundational results in the case of the loop space of an algebraic or geometric stack. These should probably be skipped on a rst reading; the essential statements for the main body of the paper are in Propositions 2.1.20 and 2.1.25, which provide an explicit description of the formal and unipotent loops of a global quotient stack. De nition 2.1.1. We consider the higher derived stack S as the locally constant sheaf on A with value 1 1 the topological circle S . Its anization is the shifted ane line BG  Spec C pS ; kq and the map S  BZ Ñ BG is induced by the map of abelian groups Z Ñ G . a a Remark 2.1.2. The stack BG is not an ane scheme since C pBS ; kq is not connective, but it still has a well-de ned functor of points; it is an example of a coane stack (in the language of [Lur11b]) or an ane stack (in the language of [To e06]). Explicitly, by Lemma 2.2.5 in [To e06] or the introduction to Section 4 of 5 6 [Lur11b], it is the right Kan extension of the classical stack sending an ane scheme S  SpecpRq to the Eilenberg-Maclane space KpR; 1q where R is considered as an abelian group under addition. The anization map S Ñ BG is given on S-points by the map of Eilenberg-Maclane spaces Kp1;Zq Ñ Kp1; Sq where we consider S as an abelian group under addition. We x an isomorphism C pS ; kq  krs where ||  1. That is, the (fully faithful) left adjoint to the restriction of a prestack (i.e. a functor DRng Ñ S) to a classical prestack (i.e. a functor Rng Ñ S). In fact, coane stacks are always left Kan extensions of classical stacks. 7 De nition 2.1.3. Let X be a prestack. We de ne the derived loop space and its variants as follows. The (derived) loop space of X is the derived mapping stack LpXq : MappS ; Xq  X  X: XX 1 1 The second presentation is a consequence of the presentation of S via the homotopy pushout S pt pt  S , and the property that derived mapping stacks take coproducts in the source to products. The evaluation map p : LX Ñ X realizes the loop space as a relative group stack over X . The derived loop space has a canonical S -action by loop rotation. The formal loop space LpXq is the completion of LpXq along constant loops X Ñ LX . It inherits a loop rotation S -action from LX . The unipotent loop space L pXq is the derived mapping stack L pXq : MappBG ; Xq 1 u and the anization map S Ñ BG de nes a map L X Ñ LX . There is a natural BG G -action a a m u 1 on L X arising from the natural G -action on BG , compatible with the S -action on LX . m a If X admits a cotangent complex, we de ne the odd tangent bundle, a linearized form of the loop space, by T r1s : Spec Sym L r1s X X X X i.e. the relative spectrum of the derived symmetric powers of the cotangent complex. There is a projection q : T r1s Ñ X and a zero section c : X Ñ T r1s induced by the structure and X X augmentation maps respectively. We write T r1s for the odd tangent bundle completed at its zero section. Both T r1s and T r1s are equipped with a natural BG G -action, where the BG -action X X a m a is encoded by the de Rham di erential and the G -action is by scaling on the bers. De nition 2.1.4. Let X be a quasicompact geometric stack. There is an exponential map p p exp : T r1s Ñ LX de ned in Section 6 of [BN12]. In particular, LX has a natural BG G -action compatible with the a m S -action. Theorem 2.1.5 (Hochschild-Kostant-Rosenberg). The exponential map is an equivalence. Proof. For the stacky case, see Section 6 of [BN12]; when X is a derived scheme, see the main theorem of [BF08]. Example 2.1.6. If X  SpecpAq, then the derived loop space op LpXq  SpecpA b Aq  SpecpC pA; Aqq Ab A is the derived spectrum of the cyclic bar complex equipped with the shue product. The rotation S -action has a combinatorial realization via the cyclic structure on the cyclic bar complex [Lod92] [Jon87]. In this op example, we think of the bar resolution B pAq Ñ A as the A b A -module obtained by tensoring A with We de ne the relative spectrum as follows: for an algebra object A P QCohpXq, we de ne the S-points for Spec A as pairs p; q where  P XpSq and  : S Ñ Spec  A which are compatible under the projection; note that  A is an algebra since S is an ane derived scheme and pullback preserves the monoidal structure on quasicoherent sheaves. The symmetric algebra functor Sym is left adjoint to the forgetful functor from the category of augmented commutative unital associative algebra objects of QCohpXq, which exists by the adjoint functor theorem. 8 the map of simplicial complexes I Ñ pt, where the unit interval I is presented by a simplicial set with two 0-simplices and one non-degenerate 1-simplex: B pAq  A b I Ñ A X  Map pI; Xq  SpecpA b Iq: DSt op The cyclic bar complex C pAq  B pAq b A is obtained by gluing the two 0-simplices of I , i.e. it is AbA the chain complex associated to the tensor product of A with the presentation of S by one 0-simplex and one non-degenerate 1-simplex: 1 1 1 C pAq  A b S LpXq  Map pS ; SpecpAqq  SpecpA b S q: DSt This makes LX into a cocyclic scheme and OpLXq into a cyclic algebra. Example 2.1.7. If X is a stack, then  pLXq is the (classical) inertia stack of X , so LX can be thought of as a derived inertia stack. In particular, let X  BG; then LpBGq  G{G is the stacky adjoint quotient (see Proposition 2.1.8 below). Note that LpBGq  BG  BG is classical since the diagonal map is at. BGBG The S -equivariant structure on OpG{Gq has a description in terms of the a cyclic algebra arising from the cyclic structure on the simplicial Cech nerve for the atlas G Ñ G{G (see Section 7.3.3 of [Lod92]). Proposition 2.1.8 (Loop space of a quotient stack). The loop space of a quotient stack LpX{Gq can be computed by the G-equivariant ber product LpX{Gq pX  Gq{G ap X{G pX  Xq{G where G acts on X  X and X  G diagonally. Proof. Note that X{G  X{G  pX  Xq{pG  Gq with action pg ; g q  px ; x q  pg x ; g x q. We write 1 2 1 2 1 1 2 2 X X X  G X  G XX   XX GG GG G G G  G G  G where the map X  G Ñ X  X sends px; gq ÞÑ px; gxq and the action of G G on X  G is pg ; g qpx; gq 1 2 pg x; g gg q. The claim follows from the \two-out-of-three" lemma for pullback squares applied to 1 2 LpX{Gq pX  Gq{G pX  Gq{pG  Gq X{G pX  Xq{G pX  Xq{pG  Gq: Remark 2.1.9. Forgetting G-equivariance, the geometric points of the loop space LpX{Gq pt are given BG by pLpX{Gq  ptqpkq  tpx; gq P Xpkq  Gpkq | g  x  xu: BG The geometric ber of the map LpX{Gq Ñ LpBGq  G{G over g P Gpkq is the xed points X pkq. The geometric ber of the evaluation map LpX{Gq Ñ X{G over x P Xpkq is the stabilizer of x in Gpkq. Example 2.1.10 (Odd tangent bundle of smooth quotient stacks). In the case of X{G where X is smooth, we have that Ñ g bO X{G X 9 where the internal di erential d is the Cartan di erential: n n 1  n  n1 n Sym pL r1sq  Sym p Ñ g bO q  p Ñ g b Ñ  Ñ Sym pg q bO q; X{G X X X X X Sym g p O  lim b ris; d : T r1s ¥k X{G Sym g i¥k The resulting de Rham complex is called the Cartan model for equivariant cohomology; see Proposition 4.12 of [HP16] for more discussion. This example can also be carried out when X is not smooth, replacing with L . 2.1.1 Loop spaces of algebraic and geometric stacks We prove some technical facts which may be skipped on a rst reading. Note that a quasi-compact geometric stack is automatically QCA in the sense of [DG13]. Remark 2.1.11. The following principles are standard and will be used frequently. If X is an algebraic stack then X admits a cover by a disjoint union of ane schemes; if X is quasi-compact this disjoint union can be taken to be nite, so that X admits a cover by an ane scheme. If X is geometric (i.e. has ane diagonal), then S  T  pS  Tq  X X XX is ane for any ane schemes S; T . Lemma 2.1.12. Let X be an algebraic stack. Then LX is an algebraic stack. If X is geometric, then LX is geometric. If X is geometric and quasi-compact, then so is LX . Proof. Assume X is algebraic. That LX is algebraic follows from the fact that LX  MappS ; Xq is a nite limit, and any nite limit of algebraic stacks is algebraic. An algebraic stack X is geometric if and only if for any map from an ane U Ñ X , the stack U  U is an ane scheme. In particular, U  X is a cover X XX for LX , and we have pU  Xq  pU  Xq  U  pLX  pU  Xqq  pU  Uq  X XX LX XX X LX XX X XX which is ane since U  U is ane and the diagonal map is ane, so LX is geometric. Assume X is also quasi-compact; then it admits a cover by an ane U , and U  X is also ane since the diagonal is XX ane. Lemma 2.1.13. Let X be an algebraic stack. Then the inclusion of constant loops X Ñ LpXq is a (schematic) closed immersion. Proof. Since the diagonal map X Ñ XX is representable by schemes, so is the evaluation map LpXq Ñ X . Let U Ñ X be an atlas for X with U a scheme; its base change along the evaluation map gives a cover by a scheme U  X Ñ LX . In particular, by the two-out-of-three property of Cartesian squares, the left XX square is Cartesian U U  X U XX X LX X i.e. the base change of the inclusion of constant loops X Ñ LX along an atlas is a scheme, so it is schematic. It is a closed embedding since any map of derived schemes which admits a retract is a closed embedding, and U Ñ U  X admits a retract by universal property. XX We provide a proof for the last claim. It suces to assume all schemes are classical, since the the property of being a closed immersion depends only on classical schemes and the property of admitting a retract is 10 preserved by  . Let f : Z Ñ Y be a map of schemes admitting a retract. We can verify that f is a closed immersion ane locally on Y , so assume Y is ane. It is a closed immersion if f : O Ñ f O is surjective. Y  Z Since Y is ane, this is equivalent to OpY q Ñ OpZq being surjective, which follows since the composition on global functions OpZq Ñ OpY q Ñ OpZq is the identity. We now introduce the notion of based loops of a stack. Namely, given a point of a stack, the based loop space is the group of automorphisms of that point and the unipotent based loops consist of the unipotent automorphisms. We use these characterizations in Propositions 2.1.20 and 2.1.25 to give explicit descriptions of the formal and unipotent loop spaces of quotient stacks. De nition 2.1.14. Let X be a prestack, and x : S Ñ X be an S-point where S is an ane derived scheme. The group of based loops at x, which we denote pX; xq, is the 8-group object in prestacks over S de ned to be the Cech nerve of the map x : S Ñ X , i.e. its underlying derived stack is LX  S or equivalently the pullback pX; xq S S X: If f : X Ñ Y is a map of prestacks, with x P XpSq, then there is a natural map of 8-groups pf; xq : pX; xq Ñ pY; fpxqq. We de ne the unipotent based loops of X , denoted pX; xq, by the ber product u u L pXq  S; there is a natural map pX; xq Ñ pX; xq. Note that the unipotent based loops do not form a group. Remark 2.1.15. If X is algebraic, then based loops at x P XpSq form an 8-group object in derived schemes over S, and if X is geometric, the based loops form an 8-group object in ane derived schemes over S. Example 2.1.16. Let X be an (ane) derived scheme and x P Xpkq a geometric point. Then, pX; xq T r1s  Spec Sym x L r1s is the odd tangent space at x P Xpkq. The comultiplication on functions is X;x X given by the natural comultiplication on the symmetric algebra and antipode map by the sign morphism. Lemma 2.1.17. For any S-point x P XpSq, we have a natural identi cation of prestacks over S pX; xq  Hom pG  S; pX; xqq grp;S a where the natural map pX; xq Ñ pX; xq is identi ed with evaluation at 1 P G . 1 u 1 Proof. Let f : S Ñ S and P pX; xqpS q with base point f x. Note that 1  1 1 pX; xq  S pX; f xq  CechpS Ñ X  S q: S S u 1 We will also use to denote its image in pL XqpS q. Let  : pt Ñ BG denote the inclusion of the (additive) 0 a identity. The diagram id 1 1 S S id 1 x 1 1 BG  S X  S a S commutes, and therefore we have a map of the corresponding Cech nerves as 8-groupoids, giving us a map u 1 1 1 pX; xqpS q Ñ Hom 1pG  S ; pX; xq  S q grp;S a S The notion of an 8-group can be found in De ntion 7.2.2.1 of [Lur09b], a useful characterization in Proposition 7.2.2.4 of op. cit. . Proposition 6.1.2.11 of op. cit. shows that Cech nerves are 8-groupoid objects in any 8-category, so that pX; xq is a groupoid object in derived stacks and therefore a group object in derived stacks over S. We use the notation Hom to denote grp the space of group homomorphisms. 11 1 for every S , natural in f , inducing a map of stacks pX; xq Ñ Hom pG  S; pX; xqq: grp;S a We now produce the inverse to this map. To do so, we need to produce a map Hom pG grp;S a 1  u 1 u 1 S ; pX; f xqq Ñ pX; xqpSq natural in f : S Ñ S. First, note that pX; xqpSq  MappBG S ; Xq XpSq MappS ; Sq, so we only need to de ne a map 1  1 Hom pG  S ; pX; f xqq Ñ MappBG  S ; Xq: grp;S a a Taking geometric realizations of the 8-groups over S considered as simplicial objects in derived stack, we obtain a map BG S Ñ B pX; f xq. We compose with the map B pX; f xq Ñ X induced by universal a S S property of geometric realizations applied to the augmentation map f x : S Ñ X . We leave the veri cation that these two maps are inverses to the reader, and naturality with respect to evaluation at 1 P G (essentially since S  BZ Ñ BG is induced by the inclusion of 1 P G ). a a The following notion of a contracting action will be used in Section 4.3. De nition 2.1.18. Let X  SpecpRq be an ane scheme with a G -action. We say the G -action is m m contracting if it acts by only non-positive weights on R. In this case, the xed point locus is Y  SpecpR q, and we say the G -action contracts to Y . In particular, there are maps Y Ñ X Ñ Y . More generally, let X be a prestack with a G -action, equipped with a G -equivariant ane map p : X Ñ Y where Y m m is given the trivial action. We say the G -action contracts to Y if for any ane S and map S Ñ Y , the induced G -action on S  X contracts to S. In particular, this implies there is also a G -equivariant m Y m section Y Ñ X . Lemma 2.1.19. Let X be a quasi-compact geometric stack. The G -actions on L X and T r1s contract m X to the xed point locus of constant loops. Proof. The claim for T r1s is by de nition. For the unipotent loop space, take x P XpSq. It suces to u u show that the induced G -action on pX; xq is contracting. This follows from the description of pX; xq Hom pG ; pX; xqq, and the contracting G -action on G . grp;S a m a 2.1.2 Formal and unipotent loops over schematic maps In [BN12], it is shown that for X a scheme, LpXq  L pXq  LpXq. This is not true for stacks, but we will now show that for a schematic map f : X Ñ Y , the formal and unipotent loops of X are loops in X whose images in Y are formal and unipotent respectively. Proposition 2.1.20. Suppose that f : X Ñ Y is a map of algebraic stacks representable by schemes. Then, p p LpXq  LpY q  LpXq: LpY q Proof. It suces to show that the closed classical substack  pY  LpXqq has the same reduced points as LpY q X € LpXq. To do this, it suces to check on geometric points. Consider the diagram of classical pullbacks pX; xq pY; fpxqq Spec k Spec k f pfpxqq X fpxq Spec k Y: 12 1 Since f is schematic, x : Spec k Ñ f pfpxqq is a closed embedding of schemes (since it as a map of schemes admitting a retract; see Lemma 2.1.13). Since Y is an algebraic stack, pY; fpxqqq is a scheme, and so pX; xq Ñ pY; fpxqq is a closed embedding of schemes and a map of ane (classical) group schemes. The preimage of the identity is thus the identity, so constant loops in LY are preimages of constant loops in LX . Example 2.1.21 (Quotient stacks). In the case of quotient stacks, we have a map LpX{Gq Ñ LpBGq  G{G. The above proposition says that LpX{Gq is the completion of LpX{Gq at the closed substack of points lying over teu{G € G{G. The following is well-known, but we provide a brief argument for the reader's convenience. Proposition 2.1.22. Let H; G be groups in ane derived schemes over k. There is a natural identi cation MappBH; BGq  Hom pH; Gq{G grp where G acts on Hom pH; Gq by the adjoint action. grp Proof. Note that we assume k has characteristic zero. We de ne maps  : MappBH; BGq Ñ Hom pH; Gq{G grp and : Hom pH; Gq{G Ñ MappBH; BGq and leave the veri cation that they are strict inverses to the grp reader. These are maps of sheaves; we will restrict our attention to de ning their map on S-points  and The map  of spaces is de ned as follows. Let F P Map pBH S; BGSq be the map of sheaves whose 1 1 1 value at S Ñ S is a functor F from (right) H -torsors over S to G-torsors over S . We de ne $ , '  / ' 1 / ' F 1pH  S q Hom pH  S; G  Sq / S Sgrp & . pFqpS q Gtorsor ' / ' / ' / % - where  is de ned as follows. There is a canonical identi cation of automorphisms of the trivial torsor 1 1 1 Aut pS  Hq  HpS q, and for h P Aut pS  Hq, Fphq is an automorphism of FpS  Gq, which we S S abusively write Fphq P G as an section in GpS q. We de ne pxqphq  x  Fphq. The map is de ned as follows. The S-points of Hom pH; Gq{G are G-torsors P over S with G- grp 1 1 equivariant maps  : P Ñ Hom pH; Gq. We de ne for S Ñ S and Q a H -torsor over S grp 1 H 1 pP; qpS qpQq  Q  pP  S q S S where H acts on P on the left via . De nition 2.1.23. A map of prestacks X Ñ Y is a monomorphism, i.e. X is a substack of Y , if for any ane derived scheme S and y P Y pSq the ber product tyu  XpSq is contractible (in the category of Y pSq spaces). Proposition 2.1.24. Let X be a geometric stack. The map L X Ñ LX is a monomorphism, i.e. unipotence of a loop is a property and not a structure. Proof. Let S  SpecpRq. Consider an S-point P pLXqpSq, which determines a base point x P XpSq and a based loop g P pX; xqpSq. We wish to show that L XpSq  tp qu is contractible. Equivalently, LXpSq we wish to show that pX; xqpSq  tgu is contractible. Since X is geometric, pX; xq is derived pX;xqpSq ane. By Lemma 2.1.17 it suces to show that for a derived ane group G over S, the map of stacks Hom pG  S; Gq Ñ G is a monomorphism on S-points. Note that an 8-group object is by de nition an grp;S a 8-monoid object satisfying a condition; in particular, the map Hom pG S; Gq Ñ Hom pG S; Gq grp;S a mon;S a 13 is a monomorphism. Furthermore, the forgetful functor from CoalgpAlgpRqq to CoalgpR -modq  CoalgpRq is fully faithful, so the map Hom pG  S; Gq Ñ Hom pp O ;O q is a monomorphism. mon;S a CoalgpRq pX;xq SG It suces to show that the map induced by evaluation at 1 P G : Hom pp O ;O q  Hom pRr pX; xqs; Rrxsq Ñ Hom pRr pX; xqs; Rq SG R CoalgpRq pX;xq a CoalgpR -modq is also a monomorphism. By the calculation in Lemma 1.12 of [GG99], O b OpG q is the cofree coalgebra S k a object in CoalgpQCohpSqq, so this map is an equivalence. Proposition 2.1.25 (Unipotent loops of quotient stacks). Let G be a classical ane algebraic group over k. Then, L pBGq  U{G where U is the unipotent cone of G (i.e. the closed subvariety of unipotent elements of G). Furthermore, if G acts on a scheme X , then L pX{Gq is computed by the pullback square L pX{Gq LpX{Gq U{G G{G: Proof. It suces to prove the statement on based loops at a given x P pX{GqpSq where S be an ane derived scheme, say S  SpecpRq. The based loops in X{G for any S-point x P XpSq can be computed via a ber product pX{G; xq pBG; ppxqq  G  S S X: Let U denote the unipotent cone of G, and let U denote the closed reduced classical subscheme of pX{G;xq the inverse image of U  S € G  S. Let U the formal completion of pX{G; xq at U . We pX{G;xq pX{G;xq rst claim that pX{G; xq  U . pX{G;xq Using Proposition 2.1.17, evaluation at 1 P G provides a map  : pX{G; xq Ñ pX{G; xq, which is a monomorphism by Proposition 2.1.24. Such a map factors through U if its (classical) set-theoretic pX{G;xq image lies in the inverse image of the unipotent cone U  S. This, we can check on k-points of S, and in particular assume that S  k. The claim now follows from the classical result that for a map of linear algebraic groups over k, the image of a unipotent element must also be unipotent (i.e. its eigenvalues are all 1 P k). In particular, pX{G; xq € U € pX{G; xq. pX{G;xq For surjectivity, we de ne the inverse map : U Ñ pX{G; xq  Hom pG ; pX{G; xqq pX{G;xq grp;S a;S via the adjoint to an exponential map G  U Ñ pX{G; xq which we will now construct. We a;S S pX{G;xq take as a given that such an exponential map is constructed for classical ane algebraic groups over k, i.e. we have a map G  U Ñ G. To de ne an exponential map for pX{G; xq, we use the universal property a G of ber products and the classical fact that if an S-point g P GpSq xes x P XpSq, then so does g P GpSq for t P G pSq. More precisely, the following diagram commutes, inducing the desired exponential map p p G  U G  U a;S S a;S G pX{G;xq exp pX{G; xq G  S S X: 14 2.2 Cyclic homology In this section we give a brief overview of the basic de nitions of Hochschild homology and cyclic homology, as developed in Section 5.5 of [Lur17]. Further discussion can be found in [BN13b] [BN13b] [Hoy18] [AMR17] [AF17] [AFR18] [NS18]. De nition 2.2.1. Let Cat be a symmetric monoidal 8-category with monoidal unit 1 , and X P Cat b b b _ _ _ a 1-dualizable object with dual X , coevaluation  : 1 Ñ X b X and evaluation  : X b X Ñ 1 . We b b de ne the dimension of X by dimpXq     P End p1 q: Cat b If F : X Ñ Y is a morphism with a right dual (i.e. adjoint) G, then we can de ne dimpFq : dimpXq trpG  Fq trpF  Gq dimpY q: Remark 2.2.2. More precisely, the a choice of dualizing structure for X P Cat determines an explicit dimension dimpXq. By Proposition 4.6.1.10 in [Lur17], the space of dualizing structures on X is contractible; therefore, dimpXq is de ned uniquely up to unique isomorphism in the homotopy category HopEnd p1 qq. Cat b This fact allows us to compute Hochschild homology in two di erent ways using di erent dualizing structures on a category and know that they are equivalent without explicitly producing an equivalence. Remark 2.2.3. Lurie's proof of the Cobordism Hypothesis [Lur09c] allows for an equivalent formulation: there is a bijection between 1-dualizable objects X P Cat and framed extended Cat -valued n-dimensional b b topological eld theories Z ; for a 1-dualizable object X we de ne the dimension by dimpXq  Z pS q: 1 1 By this de nition, there is evidently an S -action on dimpXq. The relationship between S -action and its explicit realization via the cyclic structure is spelled out in Theorem 5.5.3.11 in [Lur17]. We will de ne the Hochschild homology of a category to be its dimension; we rst need to de ne a monoidal structure on 8-categories. De nition 2.2.4 (Lurie tensor product). The category Pr is equipped with a monidal structure called the Lurie tensor product in Proposition 4.8.1.15 of [Lur17], which can be thought of as an 8-analogue of the Deligne tensor product. It is equipped with a canonical functor C  D Ñ C b D pX; Y q ÞÑ X b Y: which is initial amongst functors out of C  D which preserve small colimits separately in each variable. op Remark 2.2.5. Proposition 4.8.1.17 of [Lur17] provides an explicit realization CbD  Fun pC ; Dq which is presentable by Lemma 4.8.1.16 of [Lur17]. In particular, by [Lur09b] Proposition 5.5.3.8, the Lurie tensor L L product makes Pr into a closed monoidal category with internal mapping object Fun p;q. Furthermore, by Propositions 4.8.2.10 and 4.8.2.18 in [Lur17], the Lurie tensor product induces a tensor product on k-linear presentable categories Pr . L;! De nition 2.2.6. Let Pr be the 8-category of dualizable presentable stable k-linear 8-categories, and k;_ functors which preserve compact objects. We de ne the Hochschild homology functor by L;! L HH : dim : Pr Ñ Fun pVect ; Vect q  Vect : k k k k;_ By [AFR18], the Hochschild homology as de ned above has an S -action. 15 Remark 2.2.7. The dimension of a dualizable k-linear category takes values in chain complexes under the equivalence End p1 q  Fun pVect ; Vect q  Vect : Cat b k k k b k That is, every such endofunctor F commuting with colimits is determined by its value Fpkq. In particular, a choice of dualizing structure determines an explicit Hochschild chain complex. Remark 2.2.8. Note that the Lurie tensor product is de ned on Pr , and dimension is functorial only for right dualizable maps, so Hochschild homology is only functorial for functors whose right adjoints also admit right adjoints (equivalently, whose right adjoints are also continuous). By Proposition 5.5.7.2 of [Lur09b], these are exactly the functors which preserve compact objects. Traditionally, Hochschild homology is formulated in the setting of small stable k-linear dg categories. In particular, in this setting it is possible to explicitly write out a bar complex computing the Hochschild homology. We will see that the above \large" notion of Hochschild homology is a generalization of the \small" version. The following is proven in Theorem D.7.0.7 in [Lur18] and Proposition 4.6.15 in [Lur17]. Proposition 2.2.9. If C P Pr is compactly generated, then it is dualizable. In particular, if C  IndpC q, _ 0;op then C  IndpC q, and the evaluation map is given by ind-completion via universal properties of the 0;op Yoneda pairing Homp;q : C  C Ñ Vect . Furthermore, there are natural isomorphisms L L _ Fun pC; Cq  Fun pC; Vect q b C  C b C k k which realize the coevaluation via the identity functor in Fun pC; Cq. De nition 2.2.10. Let st be the 8-category of small stable k-linear 8-categories. We de ne the Hochschild homology functor HH : dim Ind : st Ñ Fun pVect ; Vect q  Vect k k k k i.e. HHpCq is the image of k under the composition coev L  ev Vect Fun pC; Cq C b C Vect : k k Remark 2.2.11. If C is compactly generated, then HHpCq  HHpC q, so the two de nitions of Hochschild homology given above are compatible. We now seek to understand the S -action on Hochschild homology. While there are purely categorical ways to view S actions, we model the concretely on chain complexes via a mixed di erential. De nition 2.2.12. A mixed complex is a dg-module over the dg-algebra H pS ; kq  krs where ||  1. Explicitly, it is a chain complex pV; dq with a mixed di erential  of cohomological degree 1 such that d  d and   0. We are interested in the following operations on mixed complexes. 1 1 De nition 2.2.13. We de ne the S -invariants and S -coinvariants functors by S 1 V : R Hom 1 pC pES ; kq; V q  pV rruss; d uq P krruss -mod; C pS ;kq 1 1 1 1 V : V b C pES ; kq  pV ru s; d uq P krruss -mod : S C pS ;kq utors and the Tate construction by Tate S k V  V b kppuqq  lim u V 1  pV ppuqq; d uq  pV ppuqq; d uq P kppuqq -mod : krruss S 16 1 1 Remark 2.2.14. The action of C pS ; kq on C pES ; kq comes from the sweep action of chains of [GKM98]. 1 1 1 Taking a presentation of ES as a colimit of odd spheres with free S -actions, the S -invariants can be expressed as a ltered limit S 2n1 n V  lim R Hom 1 pC pS ; kq; V q  limpV rus{u ; d uq: C pS ;kq n n Remark 2.2.15. Note that the S -invariants operation (as well as the Tate construction) is not continuous as de ned above , but can be made so by considering mixed complexes in the category Indpkrs -mod q f:g: instead. De nition 2.2.16. We respectively de ne the negative cyclic homology, cyclic homology and periodic cyclic homology by S Tate HNpCq : HHpCq ; HCpCq : HHpCq 1; HPpCq : HHpCq : In this note we consider two di erent explicit models of Hochschild homology and its cyclic variants. One is the usual cyclic bar construction on a small dg category, and the other is via S -equivariant functions on the derived loop space. Example 2.2.17 (Algebraic model). Let A be a dg algebra (or more generally, a dg category) over k, and op C  A -mod the category of left dg-modules over A. For two dg categories A; B, an A b B-module de nes a continuous functor A -mod Ñ B -mod. By the dg Morita theory of [To e07], this functor op L A b B -mod Ñ Fun pA -mod; B -modq op is an equivalence. Under this equvalence, the coevaluation k -mod Ñ AbA -mod corresponds to the functor Ab , where A is considered as a bimodule over itself, and the evaulation map corresponds to b op A. k AbA In particular, the Hochschild homology is given by the usual Hochschild homology op HHpA -modq  HHpA; Aq  A b A: AbA De nition 2.2.18 (Cyclic bar complex). The cyclic nerve of a small k-linear dg category C is the cyclic vector space whose n-simplices are given by C pCq  Hom pX ; X q b Hom pX ; X q b  b Hom pX ; X q n C 0 n C n n1 C 1 0 X ;:::;X PObpCq 0 n where the face maps are given by composition, the degeneracy maps by the identity homomorphism, and the cyclic structure by rotation of the terms. Its associated chain complex, which also abusively denote by C pCq, is the cyclic bar complex which is naturally a mixed complex with the mixed di erential arising via the Connes B-operator. This mixed di erential exhibits the S -action on Hochschild homology [Hoy18]. Remark 2.2.19. One can obtain smaller models by taking objects from a set of compact generators (see Theorem 5.2 of [Kel06]) rather than all of ObpCq; for example, if C  A -perf , then the free module A is a compact generator and one recovers the classical cyclic bar complex C pA; Aq. Example 2.2.20 (Geometric model). By [BFN10], when X is a perfect stack (e.g. a quotient stack of a derived quasiprojective scheme by an ane group in characteristic zero), then QCohpXq is compactly generated by PerfpXq and we have isomorphisms QCohpXq b QCohpXq  QCohpX  Xq  Fun pQCohpXq; QCohpXqq 1 1 9 S S 1 For example, consider N  colim kr2ns. Since N is acyclic, N  0. On the other hand, pkr2nsq  u krruss so that colimpkr2nsq  kppuqq. 17 where the functors on the right are given by integral transforms. Explicitly, we identify QCohpXq QCohpXq on compact objects K P PerfpXq by K ÞÑ pX;K b q  pX;Hom pK ;qq: Letting p : X Ñ Specpkq be the map to a point and  : X Ñ X  X the diagonal, the coevaluation is given by the functor  p and the evaluation by p  . In particular, we nd that the Hochschild homology is HHpPerfpXqq  p   p k  pX;   O q  pLX;O q  OpLXq X LX 1 1 with the last isomorphism arising via base change. The S -action on OpLXq is the S -equivariant structure arising from loop rotation; for details see Remark 3.2 and Proposition 4.2 in [BN13b]. Remark 2.2.21. If X is QCA but not perfect, it is not currently known whether QCohpXq is compactly generated. On the other hand, by Theorem 4.3.1 of [DG13], QCohpXq is dualizable, so that HHpQCohpXqq  OpLXq by a similar argument. It does not appear to be known whether HHpPerfpXqq agrees with HHpQCohpXqq. 2.3 The equivariant cyclic bar construction We now de ne an explicit model for the Hochschild homology of a quotient stack. Proposition 2.3.1. Let X be a quasiprojective scheme with an action of a reductive group G, and let p : X{G Ñ BG and q : X Ñ X{G be the natural maps of stacks. Let E P PerfpX{Gq be a locally free sheaf such that q E is a compact generator of QCohpXq. De ne A  p RHom pE;Eq P AlgpQCohpBGqq: Then, the functor R Hom pE;q : QCohpX{Gq Ñ A -mod QCohpBGq is an equivalence of dg categories. Proof. Since X is quasiprojective, it admits an equivariant compact generator E of PerfpXq (not PerfpX{Gq). The functor is fully faithful since q is an atlas and q E is a generator. It is essentially surjective, since A -perf is generated by A b V for V P IrrpGq, and R Hom pE;E b V q  A b V . QCohpBGq X{G We need the following formula for G-representations. Proposition 2.3.2. Let G be a reductive ane algebraic group, and V; W two rational representations of G (i.e. V; W P QCohpBGq). Then, there is a natural equivalence G G pV b U q b pU b Wq pV b Wq : UPIrrpGq More generally, for V ; : : : ; V rational representations of G, there is a natural equivalence 0 n G  G G pV b U q b pU b V b U q b  b pU b V q pV b  b V q : 0 1 1 n n 0 n 1 2 U ;:::;U PIrrpGq 1 n 18 Proof. Consider V b W P QCohpBG  BGq as a G  G-representation. The projection formula de nes an equivalence pBG  BG;  O b pV b Wqq Ñ pBG;  pV b Wqq: BG The result follows by unwinding this equivalence using the Peter-Weyl theorem for reductive algebraic groups O  krGs  U b U , and noting that taking global sections amounts to taking G or G  G- BG UPIrrpGq invariants (note that there are no higher cohomology groups since G is reductive). The second claim results from iterating the rst, or by applying the projection formula to the diagonal BG Ñ BG . Remark 2.3.3. The map above, which we now denote , and its inverse can be written out explicitly: pv b u b u b v b u b  b u b v q  pu pu q u pu qq  v b  b v 0 1 1 n n 1 n 0 n 1 2 1 n pv b  b v q   v b  b v b  b  b  b v 0 n 0 1 1 2 n n U ;:::;U PIrrpGq 1 n G G where  P U b U is the identity map, and   pq b  b pq is the tensor products of the projections i i to the G-invariant isotypic component. De nition 2.3.4. Let A P AlgpQCohpBGqq be an associated algebra object with coaction map c : A Ñ A b krGs. The equivariant cyclic bar complex C pA; Gq, de ned in [BG94], is the mixed complex associated to the following cyclic vector space. We de ne the n-simplices by bn1 G C pA; Gq  pA b krGsq with face and degeneracy maps d pa b  b a b fq  a b  b a a b  b a b f i  0; : : : ; n  1; i 0 n 0 i i1 n d pa b  b a b fq  cpa qa b  b f; n 0 n n 0 and cyclic structure tpa b  b a b fq  cpa q b a b  b a b f; 0 n n 0 n1 s pa b  b a b fq  1 b a b  b a b f: n1 0 n 0 n Remark 2.3.5. Note that it is essential to take G-invariants for C pA; Gq to admit a cyclic sructure. In particular, if a b  b a b f is G-invariant, then 0 n cpa q b  b cpa q b f  a b  b a b f: 0 n 0 n Proposition 2.3.6. The equivariant cyclic bar complex C pG; Aq computes HHpPerfpA -mod qq: QCohpBGq Proof. Since A generates A -perf , the objects A b V generate A -perf and we have the following QCohpBGq \small" model for HHpA -perf q. The cyclic bar complex corresponding to the generating set tAbV | QCohpBGq V P IrrpGqu of A -perf has terms QCohpBGq G G G D pA; Gq  Hom pAbV ; AbV q b Hom pAbV ; AbV q bb Hom pAbV ; AbV q : n A 0 1 A 1 2 A n1 n V ;:::;V PIrrpGq 0 n G  G Using the natural isomorphism Hom pA b V; A b Wq  pA b V b Wq , we rewrite: G  G  G D pA; Gq  pV b A b V q b pV b A b V q b  b pV b A b V q : n 1 2 0 0 1 n V ;:::;V PIrrpGq 0 n 19 n Next, we can rewrite C pA; Gq using the Peter-Weyl theorem for reductive algebraic groups: bn G C pA; Gq  pV b A b V q : n 0 0 V PIrrpGq Applying Proposition 2.3.2 to each summand (i.e. for xed V ) produces an equivalence D pA; Gq Ñ 0 n C pA; Gq; the claim that it de nes a map of cyclic objects is left to the reader. Remark 2.3.7. The equivariant cyclic bar complex C pX; Gq provides us with an explicit model of the Hochschild homology HHpPerfpX{Gqq as a krGs -linear mixed complex (i.e. both the internal and mixed G G di erentials are krGs -linear). Furthermore, since pq is a left adjoint functor whose right adjoint preserves epimorphisms, the terms in C pA; Gq are also projective. 2.4 Formal and derived completions We review the notion of formal completions of derived stacks and the notion of derived completion in the derived category. This section is essentially a summary of the results in Chapter 4 of [Lur11c], Section 3.4 of [BS15], Section 15.80 of the [Sta18], and Chapter 6 of [GR14]. De nition 2.4.1. Let f : X Ñ Y be a map of derived stacks (or more generally, prestacks). The formal completion of f , written Y , is a prestack whose functor-of-points whose S-points are given by diagrams red pSq X S Y: It can also be de ned via the ber product dR Y : Y  dR X : red Lemma 2.4.2. The formal completion of a map X Ñ Y only depends on  pXq Ñ Y . In particular, if x { Z Ñ Y is a closed embedding, then Y  X  X red . Z Y  pZ Xq 0 Y Proof. This follows directly from the functor-of-points characterization of formal completions, and that red red pXq is the universal stack that canonically factors any map from a classical reduced scheme  pSq . 0 0 De nition 2.4.3. Let f : A Ñ B be a map of derived rings. Following [Lur09a], we say that f is  etale if the induced map  pAq Ñ  pBq is  etale and for every n P Z, the map  pAq b  pBq Ñ  pBq 0 0 n  pAq 0 n is an isomorphism of abelian groups. A map of derived schemes is  etale if it is for a Zariski cover, and a representable map of derived Atin stacks is  etale if it is after base change to a cover. Proposition 2.4.4. Let X; Y; Z be stacks admitting deformation theory . Suppose that f : X Ñ Y is  etale, and let Z Ñ X be any map. Then, the relative cotangent complex vanishes L  0. Furthermore, the map X{Y p p on formal completions X Ñ Y is an equivalence. Z Z Proof. The rst sentence is Proposition 2.22 in [Lur09a]. For the second, by the exact triangle for cotangent complexes we have a natural isomorphism L  L under L , and note that the formal completion of Z{Y Z{X Z a map Z Ñ X is a colimit of square-zero extensions controlled by the the map between cotangent complexes L Ñ L (see Chapter IV.5 in [GR17]). Z{X We now discuss how to compute the derived completion of a quasicoherent complex in the derived category. It is de ned abstractly a the left adjoint to an inclusion functor of complete objects which we now de ne. This notion is de ned in Chapter IV of [GR17] and is satis ed by quotient stacks. 20 De nition 2.4.5. Let A be a connective dg ring, and x an ideal I €  pAq. We de ne the full subcategory A -mod of I -nilpotent objects consisting of those modules on which I acts locally nilpotently, i.e. for each nil cycle m P H pMq there is some power of I which annihilates m. By Proposition 4.1.12 and 4.1.15 in [Lur11c], the inclusion A -mod ãÑ A -mod is continuous and preserves compact objects; therefore it has a continuous nil right adjoint , which we call the local cohomology functor. We de ne the full subcategory A -mod of I - I loc local objects as the right orthogonal to A -mod , and the full subcategory A -mod of I -complete modules nil cpl to be the right orthogonal to A -mod . The subcategory A -mod has an equivalent characterization as loc cpl those modules such that the homotopy (derived) limit x x x M M M is zero for all x P I . By Proposition 4.2.2 of [Lur11c], the inclusion of the complete objects has a left adjoint, which we call the (derived) completion functor and denote pq. The following is Proposition 4.2.5 in [Lur11c]. Proposition 2.4.6. The composition of left adjoints pq A -mod A -mod A -mod nil cpl is an equivalence. Consequently, the composition of its right adjoints A -mod A -mod A -mod cpl nil is also an equivalence. Example 2.4.7. Let A  krxs and I  pxq, then krrxss is I -complete, krx; x s{krxs is I -nilpotent, krxs is neither, and krxs{x is both. Furthermore, { 1 { pkrrxssq  pkrxsq  krx; x s{krxsr1s krx; x s{krxs  krxsr1s  krrxssr1s: I I The derived completion and local cohomology functors can each be computed in two ways. The following can be found as Propositions 15.80.10 and 15.80.17 in [Sta18] and in a global form as Proposition 3.5.1 in [BS15] and Proposition 6.7.4 in [GR14]. The statements on local cohomology are well known (and which we will not use). Proposition 2.4.8. Assume  pAq is a noetherian ring, and choose generators f ; : : : ; f of I €  pAq. We 0 1 r 0 de ne the complex ± ± 1 1 1 1 1 G   pAq  pAqr s  pAqr ; s   pAqr ; : : : ; s : 0 0 0 0 f i;j f f f f i i j 1 r The derived completion of an A-module M can be computed in two ways: L n n M  R lim M b A{pf ; : : : ; f q  R Hom pG ; Mq: pAq pAq 1 r 0 Likewise, we can compute the local cohomology of M by n n L pMq  colim R Hom pM{pf ; : : : ; f q; Mq  G b M: I  pAq 0 1 r  pAq The latter formula is the calculation of local cohomology via a Cech resolution with supports on an ane scheme. 21 Remark 2.4.9. A theory of formal completions is described in [GR14] with the following notation. Let X p p be a dg scheme, and Z € X a classical closed subscheme. There is a functor i : QCohpXq Ñ QCohpX q with a fully faithful (continuous) left adjoint i whose essential image is the category of quasicoherent sheaves supported on Z , and a fully faithful (non-continuous!) right adjoint i whose essential image is the category of quasicoherent sheaves complete with respect to the ideal sheaf for Z . There is an exact triangle of functors p p arising from the localization functor i i (whose essential image is cocomplete): p p i i Ñ id Ñ j j  pq| : Z ? QCohpXq  U In particular, the functor on the left is local cohomology, and the functor on the right is restriction to U . p p The (non-continuous) functor i i is the (derived) completion. The following lemma is likely well-known, but we could not nd a reference. Lemma 2.4.10. Let X be a derived scheme, i : Z € X a closed subscheme and j : U  X  Z Ñ X its complement. Let  : F Ñ G be a map of quasicoherent sheaves on X . If the derived completion  and the restriction | are isomorphisms, then  is an isomorphism. Proof. Using the above exact triangle, to show that  is an isomorphism, it suces to show that pq is an p p isomorphism, or equivalently, that pconepqq  0. To this end, note that conepq  i i conepq  0, and p p p p that i is fully faithful, so that i conepq  0, so that pconepqq  i i conepq  0. Z ? Example 2.4.11. The above is not true for non-derived completions. For example, take X  A , Z  t0u, and  : 0 Ñ M  krx; x s{krxs (M is the module of distributions supported at zero). Since M is supported at zero, M|  0, and since x M  M for all k, M  0, but  is not an isomorphism. On the other hand, U Z the derived completion of M is krrxssr1s. 3 An equivariant localization theorem in derived loop spaces and Hochschild homology 3.1 Equivariant localization for derived loop spaces The following construction de nes a notion of formal and unipotent loops near a semisimple orbit of G{G, i.e. an adjoint closed G-orbit in G consisting of semisimple elements. De nition 3.1.1 (z-formal and z-unipotent loops). Let X be a derived scheme, G a reductive group acting on X , and z P G a semisimple element. We let G denote the centralizer of z, i.e. the z- xed points under the adjoint action. We de ne Z  tgzg | g P Gu to be the closed G-orbit containing z and 1 z z U  tgzug | g P G; u P U X G u to be its saturation (here, U is the unipotent cone of G and G is the centralizer of z). We de ne the z-formal and z-unipotent loops in BG by p p x L pBGq : Z{G Ñ LpBGq  G{G; L pBGq : U {G Ñ LpBGq  G{G: z z For a quotient stack X{G, we de ne the z-formal and z-unipotent loops by the pullback squares: L pX{Gq LpX{Gq L pX{Gq LpX{Gq L pBGq LpBGq L pBGq LpBGq: This construction is functorial in representable maps over BG. Note that by Propositions 2.1.20 and u u p p 2.1.25, L pX{Gq  LpX{Gq and L pX{Gq  L pX{Gq, where e P G is the identity. 22 z Proposition 3.1.2. Let G be a reductive group, and z P G a semisimple element. The map LpBG q Ñ LpBGq is (Zariski) locally  etale at z, i.e. there is a Zariski open neighborhood of Z{G € LpBGq  G{G, and therefore a Zariski open neighborhood of U {G € LpBGq, on which the map is  etale. Proof. Let us recall the set-up of the  etale slice theorem as in [Dr e04]. Let G be a reductive group acting on an ane variety X , and x P X a closed point such that the stabilizer Z pxq is reductive. We de ne a map : X Ñ T pXq as follows. Let m be the maximal ideal for x P X ; the quotient map to the cotangent space 2 2 has a Z pxq-equivariant splitting m{m Ñ m since Z pxq is reductive, de ning a map Sym pm{m q Ñ krXs. G G Geometrically, this means choosing functions f ; : : : ; f P krXs vanishing at x whose di erentials generate 1 n the cotangent space at x, and de ning the map  : X Ñ T pXq by evaluation y ÞÑ f pyq ; df yx in a Z pxq-equivariant manner. The  etale slice is the inverse image  pNq where N € T pXq is any normal G x Z pxq 1 subspace to T pG  xq € T pXq, and the theorem tells us that the map G   pNq Ñ X is  etale. x x Specializing to our situation, where G acts on itself by the adjoint action, we produce the G -equivariant map  : G Ñ T pGq as follows. Ane locally at z, we can choose generators f ; : : : ; f such that z 1 n krGs{pf ; : : : ; f q  krG s; 1 r and the vanishing of df ; : : : ; df cuts out g € T pGq  gz (i.e. the translation of T pGq  g by central z). 1 r z e z 1 z z Thus it suces to show that g is a normal subspace to T pG  zq, since  pg q  G by construction. On z z the other hand, we have a natural isomorphism G z  G{G , inducing T pG zq  gz{g z, which produces a splitting of T pG  zq € T pGq whose kernel is g . Explicitly, z is semisimple and acts on g, so g decomposes z z z z into z-eigenspaces; g is the trivial eigenspace and g{g is isomorphic to the sum of all other eigenspaces. z G Using the fact that the map G  G Ñ G is G-equivariant, and since for any u P U we have z P G  u, it follows that every open set containing Z also contains U . Example 3.1.3. Let G be a simple reductive algebraic group and choose z P G regular semisimple. Its T reg rs centralizer is a torus T and the map G  T Ñ G is  etale with ber W  NpTq{T . Corollary 3.1.4. Let X be a prestack equipped with the action of a reductive group G over k, and let z P G be central. Then, LpX{G q Ñ LpX{Gq is  etale over a neighborhood of U {G € G{G. In particular, the z u z u p p natural maps L pX{G q Ñ L pX{Gq and L pX{G q Ñ L pX{Gq are equivalences. z z z z z z Proof. Loop spaces commute with ber products, and X{G  X{G  BG . The second claim follows BG since  etale maps induces equivalences on formal completions along isomorphic closed subschemes. It is a straightforward veri cation that the map G 1 z z 1 z U  G  tphuzh ; gq | g P G; h P G ; u P U X G u ÝÑ U  tguzg | u P U X G u z;G G z;G G is an isomorphism. We de ne two competing notions of z-invariants. De nition 3.1.5. In the set-up above, we de ne the derived z-invariants X to be the derived ber product of the diagram X X X pX  Xq where denotes the graph of the action by a closed point z P G. We de ne the classical z-invariants X to z z be  pX q{G , or equivalently, the above square considered as an underived ber product. 23 Given this notion of z-formal and z-unipotent loops, we are ready to de ne the localization map comparing z-formal and unipotent loops with the formal and unipotent loops of the classical z- xed points. De nition 3.1.6 (Localization map and formal localization map). We de ne the following maps realizing localizations from the most global to local. We de ne the global localization map via the composition z z z ` : Lp pX q{G q LpX{G q LpX{Gq z 0 z z z induced by the sequence of natural maps of quotient stacks  pX q{G Ñ X{G Ñ X{G where the rst map is the closed immeresion and the second map is the base change map along BG Ñ BG. For U{G € G{G an open subscheme, we by ` the restriction of ` to U . The map lives over the natural z;U z map LpBG q Ñ LpBGq. u u z z u The unipotent localization map ` : L p pX q{G q L pX{Gq is the base change of ` along 0 z z z u z z L pBG q LpBG q L pBGq LpBGq where the isomorphism on the left arises via Corollary 3.1.4. Note that the map lives over L pBGq. z z p p p The formal localization map ` : L p pX q{G q L pX{Gq is the base change of ` along z z 0 z z z z L pBG q LpBG q L pBGq LpBGq where the isomorphism on the left arises via Corollary 3.1.4. Note that the map lives over L pBGq. Remark 3.1.7. Applying the functor  pt, ` can be identi ed with the map induced on ber products BG z of the diagrams G z z G  p pX q  G q 0 X  G ÝÑ z z G z G z z X X  X G   pX q G  p pX q   pX qq 0 0 0 where the top map sends ph; x; gq ÞÑ ph  x; hgh q. z z z z We now investigate the map Lp pX q{G qq Ñ LpX{G q. For ease of notation, we can replace G with a reductive group G in which z is central. Our goal is to nd an open G-closed neighborhood U{G € G{G on which the above map is an equivalence. Let us rst consider the case when G  T is a torus. Lemma 3.1.8 (Finiteness of stabilizer subgroups). Let T be a torus acting on a (quasicompact) variety X . Only nitely many subgroups of T may appear as stabilizers of this action. Proof. We can work ane locally, since X has a nite T -closed Zariski cover, and may also assume that X is connected. If every point of X has stabilizer of equal dimension to T , the possible stabilizer subgroups are in bijection with a subset of the set of subgroups of the ( nite) component group T{T . If there is a point 24 x x P X whose stabilizer T € T has strictly smaller dimension, then by the Luna slice theorem (note that the stabilizer T is reductive since every subgroup is) there is a locally closed subvariety V € X such that the map a : T  V Ñ X is  etale and dominant. Any stabilizer of a point in the image of a must be a subgroup x x of T , so the problem reduces to considering (a) the action of T on V along with (b) the action of T on the compliment of V € X (which is a closed subvariety, therefore ane, of strictly smaller dimension). Note that in both cases, the dimension of either the variety or the group decreases, and that the claim is obviously true for zero-dimensional varieties and discrete ( nite) groups, so the lemma follows by induction. Corollary 3.1.9. In the above situation, let z P T and let U € T be the open neighborhood of z obtained by w z deleting the nitely many stabilizers which do not contain z. Then, for w P U , we have X € X . We bootstrap this reuslt to prove an analogous result for general reductive G. Proposition 3.1.10. Let G be a reductive group, z P G a central element and X a scheme with a G-action. Then, there is an open neighborhood U of z in G that is closed under the adjoint action and such that w P U w z implies that X € X . Proof. We de ne U as follows. First, take a maximal torus T € G containing z. By Proposition 3.1.8, there is an open neighborhood U of T with the given property. We de ne 1 1 uni U  tgtug | t P U ; g P G; u P C ptq u uni where C ptq are the unipotent elements in the centralizer of t. The set U is evidently G-closed. By Jordan 1 1 decomposition, G  U  G  pT  U q, so U is open since T  U is closed. It remains to show that if w P U , w z then X € X . Every w P U has a Jordan decomposition w  su where s is semisimple and u is unipotent; in particular, 1 1 s t z z z s  gtg for some t P U and g P G. First, note that X  g  X € g  X  X (since X is G-closed for w s z central z). By the following lemma, it follows that X € X € X . Lemma 3.1.11. Let w P G be an element of a reductive group acting on a scheme X , with Jordan decom- w s position w  su for semisimple s and unipotent u. Then, X € X . Proof. If u  e the claim is trivial, so suppose u  e. Take x P X . There is a 1-parameter subgroup of G containing s, and a choice of unipotent u  1 uniquely de nes an injective map of algebraic groups G Ñ G, assembling into an injective map of group schemes G  G Ñ G. Then, H  pG  G q X G is a closed m a m a group scheme of G which is at most two-dimensional and which contains w  su. The connected component H is either the trivial group, G  t0u, t1u  G , or the entire group G  G . If it is the trivial group m a m a then H is discrete, therefore nite, but this is impossible since the projection to G is open and there are no nite group subschemes of G . For the same reason, H cannot be G  t0u. In the remaining two cases, a m x x wH € H € G , so w P G as desired. Given this, we are now ready to prove our main theorem. Theorem 3.1.12. Let X be a smooth variety with an action of a reductive group G, and z P G semisimple. There is an open substack U{G € G{G containing z such that the S -equivariant map z z z z ` : Lp pX q{G q  z z pU X G q{G LpX{Gq  U{G z;U 0 G {G G{G is  etale. If z P G is central, then ` is an equivalence. In particular, the S -equivariant maps on z-unipotent z;U loops and z-formal loops are equivalences: u u z z u z z p p p ` : L p pX q{G q L pX{Gq; ` : L p pX q{G q L pX{Gq: 0 z z 0 z z z z 25 Proof. Assuming the rst statement holds, then the statement on z-unipotent and formal loops follows since z z z the global localization map is a composition Lp pX q{G q Ñ LpX{G q Ñ LpX{Gq: By the rst statement z z z of the theorem, since z is central in G , there is an open substack U of G {G containing both Z  G  tzu and its saturation U over which the rst map is an equivalence. By Proposition 3.1.2, the second map is etale on U , inducing an equivalence on formal completions along Z and U by the argument in Corollary 3.1.4. 2 z z We now prove the rst statement. Let U € G be an open neighborhood of z P G from Proposition w z 2 1 3.1.10, i.e. such that  pX q €  pX q for w P U . Let U € G be an open neighborhood of z on 0 0 which  is  etale, obtained via Proposition 3.1.2; note that this implies that  is open over U . De ne 2 1 z U  pU q X U € G. By construction, the map LpX{G q Ñ LpX{Gq is  etale over U . It remains to prove that the map is an equivalence when z P G is central; in this case G  G. We proceed by base changing from BG to Specpkq (i.e. forgetting equivariance). Take Y   pX q for w z shorthand, and recall that U € G is an open subscheme on which  pX q €  pX q for w P U . This means 0 0 we wish to show that the map of derived schemes i : Y  pY  Uq Ñ X  pX  Uq Y Y XX is an equivalence. Since i is a closed embedding (and therefore ane), we view the map of derived schemes ane locally as a map of di erential graded sheaves of algebras on U  G, which we abusively denote by O Ñ O  i O . Since these sheaves have coherent cohomology on XG and closed points LpX{Gq LpY {Gq LpY {Gq are dense in X  G (X  G is locally nite type over a eld), it suces to check the claim on local rings at 2 w z closed points px; wq P  pLpX{Gqq € X  G where w P U € G and x P  pX q €  pX q (via the de ning 0 0 0 property of U ). z 11 Note that Y   pX q is smooth when z P G is semisimple in reductive G by a standard argument . Consequently, the diagonal maps are local complete intersections, and Y   pX q € X is also a local complete intersection. The z-action on the cotangent space T pXq is semisimple, and determines a splitting with identi cations z  z T pXq  E ` E ; E  kerpz  q  N p pX q{Xq; E  kerpz  1q  T p pX qq: 0 1 0 0 1 0 x x x Let J denote the ideal such that O  O {J . We aim to compute Y;x X;x L L O b O ÝÑ O b O : (1) X;x Y;x O XG;px;wq O Y G;px;wq XX;px;xq Y Y;px;xq Let v ; : : : ; v be a basis of E and v ; : : : ; v a basis of E . By Nakayama's lemma we can lift this basis of 1 r 1 r1 n 0 the cotangent space to generators x ; : : : ; x P m € O such that J  px ; : : : ; x q. Furthermore, again 1 n X;x X;x 1 r by Nakayama, tx b 1  1 b x | i  1; : : : ; nu form a regular sequence for the diagonal X € X  X at i i px; xq, and likewise the images of tx b 1 1b x | i  r 1; : : : ; nu form a regular sequence for the diagonal i i Y € Y  Y at px; xq. Taking semi-free Koszul resolutions of the diagonal, as well as O as a O -module, Equa- Y G;px;wq XG;px;wq tion 1 can be rewritten 1 1 O r ; : : : ;  s O r ; : : : ;  s O r ; : : : ;  ;  ; : : : ;  s (2) 1 n r1 n r1 n XG;px;wq Y G;px;wq XG;px;wq 1 r where | |  1, and the internal di erentials are de ned by dp q  cpx q  x b 1 and dp q  x b 1. The i i i i i map on the right is a quasi-isomorphism. To show that Equation 2 is a quasi-isomorphism, it suces to produce the dotted map above making the diagram commute. Explicitly, we wish to show that the derived Choose a torus T containing z and apply the  etale slice theorem to a T -closed ane open cover. Since A and A{J are Cohen-Macaulay, any minimal generating set is automatically regular. 26 1 1 equations imposed by dp q; : : : ; dp q and by dp q; : : : ; dp q di er by an (invertible) change of variables. 1 r 1 r That is, we wish to nd an element of GL pO q which transforms r XG;px;wq tx b 1; : : : ; x b 1u into tcpx q  x b 1; : : : ; cpx q  x b 1u: 1 r 1 1 r r The element z is central, so w xes  pX q, and in particular the coaction map c preserves the ideal J . Thus, we can write cpx q  e px b 1q for some e P O . Let E  pe q denote the i ij j ij ij XG;px;wq i;j1 corresponding matrix; the matrix E I is invertible if its evaluation at w P G is invertible. The matrix Epwq is the action of w on the conormal space E  N p pX q{Xq; in particular, pE  Iqpwq is invertible if and 0 0 only if Epwq has no xed vectors. But w cannot x any vectors on a normal space of its xed-point variety, z w and since N p pX {Xq € N p pX {Xqq, the claim follows. x 0 x 0 Corollary 3.1.13. Let X; Y; Z be smooth varieties, with maps f : X Ñ Z and g : Y Ñ Z . Then z z p p L p pX q  z  pY qq  L pX  Y q z 0 0 z Z pZ q where all ber products are derived. Proof. This follows immediately since loop spaces commute with ber products. Remark 3.1.14. Note that in the case G  T is a torus, every element of T is central, so ` is an z;U equivalence. Furthermore, Proposition 3.1.8 gives an explicit description of the open set U , which is maximal, on which ` is an equivalence. 3.2 Equivariant localization for Hochschild and cyclic homology First, let us show that the completion over a closed point of the anization rzs P G{{G is the same as taking z-unipotent loops. Lemma 3.2.1. Let a : G{G Ñ G{{G be the anization, and let apzq  rzs P G{{G. The map above induces an isomorphism on completions { 1 L pBGq  a przsq{G € LpBGq: u { In particular, the map L pBGq Ñ LpBGq factors isomorphically through a przsq{G. Proof. This follows from Proposition 2.1.25 and the observation that the classical reduced ber over rzs P G z z z z z in G{G is isomorphic to U {G , where U is the unipotent cone of G . That is, if u P G is unipotent, then G G uz  zu P G has the same eigenvalues as z, so apzq  apzuq, and by Jordan composition any y P  przsq uz z z can be written uniquely in this way. Furthermore, G € G ; letting U be the unipotent elements of G , it G 1 follows from Proposition 3.1.2 that the map G  U Ñ G is a closed embedding with image  przsq. As an application of the above geometric incarnations of equivariant localization, we obtain the following equivariant localization results in Hochschild and cyclic homology. De nition 3.2.2 (Completion and localization of Hochschild homology). Let G be a reductive group. Note that HHpPerfpX{Gqq is naturally a HHpPerfpBGqq  krGs -module. Let z P G be a reductive element representing a closed point of SpecpkrGs q  G{{G. We denote by HHpPerfpX{Gqq the localization at the maximal de ning z P G{{G and HHpPerfpX{Gqq the completion at z. We de ne the analgous notions for zp HN; HC and HP , taking care to completed with respect to the ideals I rruss and I ppuqq where I € krGs z z z is the ideal de ning z P G{{G. The following is an immediate corollary of the the localization theorem for unipotent loops. 27 Theorem 3.2.3 (Equivariant localization for Hochschild homology). Let X be a smooth scheme, G a reduc- tive group, and z P G semisimple. Then the natural map on completions induced by pullback is an equivalence: z z u z z HHpPerfpX{Gqq HHpPerfp pX q{G qq  OpL p pX q{G qq: zp 0 zp 0 If z P G is central, the map on localizations is an equivalence: z z HHpPerfpX{Gqq HHpPerfp pX q{Gqq  OpLp pX q{Gqq b O : z 0 z 0 O G{{G;rzs G{{G z z u Proof. Via Lemma 3.2.1, we have a natural identi cation HHpPerfp pX q{G qq  OpL pX{Gqq. The zp z rst claim then follows directly from Theorem 3.1.12, and base change for derived completions along closed embeddings. The second claim for localizations follows by Theorem 3.1.12, since in that case the map on Hochschild homology is an equivalence for an open set containing z. Remark 3.2.4. For cyclic homology, the situation is a little more delicate. The formation of cyclic homology HC involves a ltered colimit, the formation of negative cyclic homology HN involves a co ltered limit, and the formation of periodic cyclic homology involves both. On the other hand, localization commutes with colimits and nite limits but not co ltered limits, and completion commutes with limits but not colimits. We rst need to introduce a few technical notions regarding mixed complexes, with the aim of proving that in our situation formal completions commute with the Tate construction on Hochschild homology. Analogous results and arguments can be found in [Kal15]. De nition 3.2.5. Let pV; d; q be a mixed complex. We de ne ± ¹ Tate k ` Tate k 1 V  p V u ; d uq V  p V u ; d uq  pV ru ; us; d uq where |u|  2. There are natural maps ` Tate Tate Tate V Ñ V Ñ V : Remark 3.2.6. Lemma 2.6 of [Kal15] shows that the Tate construction preserves quasi-isomorphisms, es- sentially because it is computed via the right spectral sequence. On the other hand, the other variants above do not preserve quasi-isomorphisms. In particular, they are not well-behaved in the derived category. De nition 3.2.7. We say a complex V is cohomologically bounded below (respectively, above ) if H pV q  0 for all suciently small (respectively, large) i. We say V is strictly bounded below (respectively, above ) if V if V  0 for i suciently small (respectively, large). ` Tate Tate Lemma 3.2.8. Let pV; d; q be a mixed complex. If V is strictly bounded below, then V Ñ V is an Tate Tate isomorphism. If V is strictly bounded above, then V Ñ V is an isomorphism. Proof. The proof of the rst statement appears as Corollary 2.7 in [Kal15], which we repeat for convenience. ` Tate k n2k Tate The chain complex V is, in the nth cohomological degree, the vector space u V . whereas V À ± k n2k k n2k n2k is in the nth cohomological degree the vector space p u V qp u V . Since V  0 for k¥0 k 0 large k, the product is nite and therefore a direct sum, so the map is an isomorphism. A similar argument proves the second statement. Lemma 3.2.9. Let pV ; d ;  q be a degree-wise Mittag-Leer sequential diagram of mixed complexes such that the V are uniformly cohomologically bounded above (i.e. right t-bounded). Then, the functors pq 1 Tate and pq commute with limits, i.e. Tate Tate plim V q  lim V : 28 Proof. First, since the limit is degree-wise Mittag-Leer, R lim  lim. Note that since the V are uniformly cohomologically bounded above, so is their limit. Since we are only interested in computing the usual Tate construction which respects quasi-isomorphisms, we can replace each of the V and lim V with their (strictly bounded above) truncations; the resulting complex is still Mittag-Leer. Now, note that the Tate construction commutes with limits. Theorem 3.2.10 (Equivariant localization for cyclic homology). Let X be a smooth scheme, G a reductive group, and z P G semisimple. Then the following maps on completions induced by pullback are equivalences z z u z z S HNpPerfpX{Gqq HNpPerfp pX q{G qq  OpL p pX q{G qq ; zp 0 zp 0 z z u z z HCpPerfpX{Gqq HCpPerfp pX q{G qq  OpL p pX q{G qq 1; zp 0 zp 0 z S z z u z z Tate HPpPerfpX{Gqq HPpPerfp pX q{G qq  OpL p pX q{G qq : zp 0 zp 0 z z u Proof. Via Lemma 3.2.1, we have a natural identi cation HHpPerfp pX q{G qq  OpL pX{Gqq. If we can 0 zp show that the limit diagram in the formation of the derived completion satis es the conditions of Lemma 3.2.9, then the claims would follow via Theorem 3.2.3. Take A P AlgpQCohpBGqq to be as in Proposition 2.3.1. First, note that HHpPerfpX{Gqq  C pA; Gq is cohomologically bounded above since A is cohomologically bounded. By Remark 2.3.7, its terms are projective, so the terms in the limit computing its derived completion with respect to an ideal of krGs are classical quotients. Therefore, the limit diagram satis es the conditions of Lemma 3.2.9. Remark 3.2.11. While the statement in the above theorem for negative cyclic homology was more or less automatic, the statements for cyclic homology and periodic cyclic homology are strongly dependent on the cohomological right-boundedness of Hochschild homology, which we expect to fail for CohpX{Gq where X is singular. 4 An Atiyah-Segal completion theorem for periodic cyclic homol- ogy 4.1 Twisted circle actions on loop spaces De nition 4.1.1. Let G be an ane algebraic group over k. A G-action on a prestack X is de ned to be a prestack Y over BG, along with an identi cation Y  pt  X . We often abuse notation and write BG Y  X{G and understand that the identi cation implicitly. Furthermore, if Z € G is a closed normal subgroup, a trivialization of the action of Z on X is de ned to be a G : G{Z -action on X along with an identi cation X{G  1 BG  X{G. If z P G is central, then it generates a normal closed subgroup Z ; we BG sometimes write z-trivialization to mean a Z -trivialization. Furthemore, note that any X equipped with a z-trivialization is equipped with a shift map  : X{G Ñ X{G de ned as follows. It is de ned on BG via the map of groups G Ñ G taking g ÞÑ zg  gz. It is de ned on X{G by transporting the map id 1   . across the trivialization X{G  X{G  1 BG. It is X{G id 1 z BG BG evidently an equivalence with inverse  1 . Example 4.1.2. If X is a classical scheme with a G-action, on which a normal subgroup Z € G acts trivially, then there is a canonical trivialization of the action of Z on X . De nition 4.1.3 (Twisted S -actions). Let G a linear algebraic group over k and X a scheme over k with an action of G. We de ne the following various circle actions on X{G. Let z P G be central. The loop rotation S -action, denoted , on the loop space LpX{Gq. 29 1 The z-twisting S -action, denoted  , on BG is induced via the map of groups Z  G Ñ G de ned by n n pn; gq ÞÑ z g  gz . More generally, if X is equipped with a z-trivialization, then the corresponding 1 1 1 z-twisting S action on X{G is de ned via the identi cation X{G  X{G  BG, where S acts BG 1 1 on X{G and BG trivially, but on BG via the z-twisting described above. It is evident that the maps de ning the ber product diagram are S -equivariant. 1 1 The z-twisted rotation, denoted  , on LpX{Gq is the diagonal of the S  S -action   L . Note z z that it makes sense to talk about the diagonal since  and L commute:  commutes with any group action of the form L , where is an S -action on X{G. Furthermore, for w P G central, the map  : X{G Ñ X{G is S -equivariant with respect to  on the left z zw and  on the right. In particular, Lp q : LpX{Gq Ñ LpX{Gq is equivariant with respect to  on the left w z zw and  on the right. Remark 4.1.4. Using Proposition 2.1.8, we see that the shift map Lp q : LpX{Gq Ñ LpX{Gq has an explicit realization as the multiplication by z map G{G Ñ G{G on the upper-right term in the ber product LpX{Gq  pt pX  Gq{G BG ap X{G pX  Xq{G: Example 4.1.5. On PerfpG{Gq, the rotation  acts on bers over g P G by g; the z-twisting L acts on bers over any g P G by z, and the twisted rotation  acts on bers over g P G by gz  zg. Example 4.1.6. Let G  T be a torus (in particular, every t P T is central). We can explicitly describe the S -actions on linear categories PerfpLpBTqq  PerfpT  BTq  PerfpTq where  is the character lattice of T . Let tz |  P u denote the natural basis of monomials for krTs. The rotation S -action  acts on the -summand by z . The t-twisting action  on BT acts on the -summand by the scalar z ptq. The t-twisted rotation  acts on the -summand by z ptqz . Let us take the t-twisted rotation  . We have, via the category PreMF in [Pre11], PerfpLpBTqq  PreMFpT; 1  z ptq  z q; p p PerfpLpBTqq  PreMFpT; 1  z q: ptq1 For the second identity, the zeros of 1  z ptqz meet the constant loops if and only if z ptq  1. After passing to the Tate category under the rotation action, we note that the zero locus of 1  z is smooth (it is a subgroup of T ) of codimension 1 unless   0 (in which case it has codimension zero and must be derived). Therefore, Tate PerfpLpBTqq  MFpT; 1  z ptq  z q  PerfpTq b kpuqq; Tate p p p PerfpLpBTqq  MFpT; 1  z q  PerfpTq b kppuqq: ptq1 Note that the Tate categories do not depend on the twisting at all (but the S -invariant categories do). For M a scheme and f : M Ñ G , the category PreMFpM; fq is the category PerfpM  t1uq with an extra kruss-linear structure acting by cohomological operators. 30 Proposition 4.1.7. Suppose z P G is a central element of a reductive group and acts on a quasiprojective scheme X trivially. Then, there is a krruss-linear equivalence 1 1 S ; S ; Lp q : OpLpX{Gqq OpLpX{Gqq : The same holds for formal and unipotent loop spaces. Proof. Since X is quasiprojective, it has a compact G-equivariant generator E of QCohpXq. Let A R Hom pE;Eq, so that QCohpXq  A -mod . Let c : A Ñ A b krGs be the coaction; note that QCohpBGq the z-twisted rotation is given by twisting the coaction by z, i.e. c paqp; gq : cpaqp; gzq  cpaqp; zgq. Since z acts on X trivially, this is equal to the usual coaction c, inducing an equivalence of S -invariants under the untwisted and z-twisted rotations. 4.2 Tate-equivariant functions on formal loop spaces compute analytic de Rham cohomology We now set out to prove the equivariant localization theorem for periodic cyclic homology. We rst introduce some technical notions needed to phrase the result in the 2-periodic setting. Recall the following notions for vector spaces (not chain complexes) from [Bei08]. De nition 4.2.1. A linear topological vector space is a a vector space V which admits a topology for which the vector space operations are continuous, and such that there is a system of neighborhoods at 0 consisting of subspaces. In this case, the topology is generated by this system at 0 and translations under addition. The completion V of V is the limit over the system of neighborhoods U : V : lim V {U : 0PU We say the topology is complete if the natural map V Ñ V is an isomorphism. Let V ; V be linear topological 1 2 vector spaces. We de ne a linear topological vector space, the !-tensor product V b V , via the naive tensor 1 2 product on underlying vector spaces equipped with topology by the basis consisting of open sets of the form x ! U b V V b U , where U € V and U € V are opens. We de ne V b V to be V b V completed with 1 2 1 2 1 1 2 2 1 2 1 2 respect to this topology. These notions generalize immediately to chain complexes, where we replace the notion of subspace with subcomplex. In this case, the complexes term-wise satisfy the Mittag-Leer condition and therefore lim  0, so R lim  lim. Note that as in [Kal15] V b V  lim lim V b V : 1 2 1 2 V V 1 2 Remark 4.2.2. It is unclear to us how the notion of a topological chain complex should interact with quasi-isomorphisms. We now review the constructions and results of [Bha12] and [Har75]. De nition 4.2.3. Let X be a nite-type derived stack with ane diagonal over k, and L its cotangent À ™ complex. The derived de Rham complex dR is the sum-totalization of the complex p L rns; d q, X X dR n¥0 À ™ which comes equipped with a Hodge ltration F dR  p L rns; d q. The Hodge-completed X X dR n¥k derived de Rham complex dR is the completion of dR with respect to the Hodge ltration (see Construction X X 4.1 in [Bha12]). We will denote the derived global sections of this complex to be the complex of derived de Rham cochains C pX ; kq. dR We have two competing notions of Hodge ltrations in the negative cyclic homology of PerfpX{Gq. Both play an essential role. 31 1 S Tate De nition 4.2.4. Let pV; d; q be a mixed complex. The noncommutative Hodge ltration on V and V k S is the decreasing ltration de ned by the subspaces u V . De nition 4.2.5. The odd tangent complex T r1s is ane over X , so we can consider O as an X T r1s algebra in the category QCohpXq. Furthermore, the ideal sheaf for the zero section de nes an exhaustive decreasing ltration on O , which induces an exhaustive decreasing ltration on OpT r1sq whose T r1s X completion is OpT r1sq . We call this the geometric Hodge ltration. Proposition 4.2.6. Let X be a geometric stack with a smooth cover by a variety. There is a natural quasi-isomorphism Tate OpLpXqq C pX ; kqb kppuqq dR where we consider the de Rham complex C pX ; kq as a topological chain complex with respect to the derived dR Hodge ltration, and kppuqq with respect to the noncommutative Hodge ltration. p p Proof. By Theorem 6.9 of [BN12], the exponential map T r1s Ñ LpXq is an ltration-preserving isomor- Tate S phism, so we can compute OpT r1sq instead. Note that pq commutes with totalization, so we rst compute S i j OpT r1sq  lim lim RpX; risqu ; u  d : X dR nÑ8 mÑ8 0¤i¤n 0¤j¤m Note that the ltration de ned by the limit parameter n is the geometric Hodge ltration, and the ltration de ned by the limit parameter m is the non-commutative Hodge ltration. As we take the limit with respect i j to both, it amounts to computing the direct sum complex p RpX; risqu ; u  d q with respect to dR 0¤i;j X the opens i j U   RpX; risqu ; u  d : nm dR i¥n or j¥m There is a subcomplex (in fact, a direct summand) of the direct sum complex de ned by taking the summands where j ¥ i: i j V  RpX; risqu ; u  d : dR 0¤i¤j It is a subcomplex since the de Rham di erential takes the pi; jq-summand to the pi 1; j 1q-summand. Its quotient 1 i j V  RpX; risqu ; u  d dR 0¤j i is u-torsion. Note that kppuqq is at as a (dg) krruss-module, so  b kppuqq kills u-torsion modules, and krruss in particular we have an equivalence induced by the inclusion on completions Tate p p V b kppuqq  OpT r1sq : krruss X Consider the alternative topology on V de ned by opens i j W  RpX; risqu ; u  d : km  dR ij¥k or j¥m The parameter k in this topology de nes the derived Hodge ltration in the derived de Rham complex, and the parameter m in this ltration de nes a u-adic ltration. In particular, the completion of V with respect 32 ! to the topology de ned by the W is pX; dR qb krruss. km X We claim these two topologies de ned by U and V are equivalent. Indeed, this is an easy veri cation nm km as all indices are bounded below. Explicitly, W € U and U € W . In particular, the completion nm nm km;m km with respect to this topology is the limit under the usual Hodge ltration on each summand of V de ned by letting i  j be constant, and the limit under the noncommutative Hodge ltration. Thus, via the universal property of the limit, we have a canonical equivalence p x V  pX; dR qb krruss and in particular, Tate p x ! OpLpXqq  V b kppuqq  pX; dR qb kppuqq  C pX ; kqb kppuqq: krruss X dR Remark 4.2.7. The above proposition is false if we do not consider the topologies. For example, take X 0 Tate p p BG . Then, we have OpLpBG qq  krrrtss where |t|  0, and in particular, H pOpLpBG qq q  krrtss. m m m 0 1 On the other hand, H pBG ; kq  krss  krrsss where |s|  2, so H pH pBG ; kqppuqqq  krsu s. m m Finally, we discuss results relating derived de Rham cohomology to analytic (Betti) cohomology when k  C. It is well-known by experts and is essentially a simple corollary of results in [Bha12], [Har75] and [BN12]. A general discussion can also be found in the introduction of [Kal15]. We rst de ne a few intermediate chain complexes. De nition 4.2.8. For a choice of embedding of i : X Ñ M for smooth M , the Hartshorne algebraic de Rham complex is de ned by H 1 { 1 : i O b 1 i X i O Y We de ne the term-wise Hodge-completed derived de Rham complex associated to a groupoid presentation U U of a stack as above to be the totalization of the double complex dR . Since for schemes U , 1 0 U i the de Rham cohomology computes Betti cohomology, Lemma 32 of [Beh04] implies that the cohomology of this complex is independent of choice of cover. Corollary 4.27 in [Bha12] and Theorem 1.1 in Chapter IV of [Har75] can be summarized in the following statement, identifying the derived de Rham cohomology of a possibly singular scheme X over C with its Betti cohomology. For details, see Construcion 4.25 in [Bha12]. Theorem 4.2.9 (Bhatt, Hartshorne). Let X be a nite type scheme over k  C. There is a natural H an map of sheaves of dg k-algebras dR Ñ on X which is a quasi-isomorphism. Letting j : X Ñ X an be the analyti cation map, and the analytic de Rham complex on X , the map of dg k-algebras an 1 H 1 an is a quasi-isomorphism. All together, the map j dR Ñ an is a quasi-isomorphism, X X X an and an resolves the constant sheaf C . Therefore, the hypercohomology of the derived de Rham complex an computes the Betti cohomology of X . Corollary 4.2.10. Let X be a nite type stack over k  C presented as a groupoid U U . Then, 1 0 an Betti cohomology of X is computed by the term-wise Hodge-completed derived de Rham complex associated to U : Corollary 4.2.11. Let X be a nite type derived stack over k  C. Then, there is a natural quasi- isomorphism Tate an OpLXq C pX ;Cqb kppuqq: dR Proof. By Proposition 6.4 of [BN12], the Hodge-completed graded algebra of di erential forms is a graded sheaf on the smooth site of X , i.e. the Hodge-completed derived de Rham complex for stacks is quasi- isomorphic to the term-wise Hodge-completed derived de Rham complex associated to an atlas U Ñ X of 33 the stack X . This implies that the derived de Rham complex computes Betti cohomology and in particular, using Proposition 4.2.6 there is an ltration-preserving equivalence in the derived category C pX ;Cq dR an C pX ;Cq. 4.3 Comparing global functions on unipotent and formal loop spaces We prove a completion theorem for periodic cyclic homology, assuming the following theorem, which is proven as Theorem 4.3.22. Theorem 4.3.1. Let X be a quasicompact algebraic space with an action of an ane algebraic group G. The map on functions induced by pullback is an equivalence: u Tate Tate OpL pX{Gqq OpLpX{Gqq : Theorem 4.3.2 (Atiyah-Segal completion for periodic cyclic homology). Let G be a reductive group acting on a smooth quasi-projective variety X . The periodic cyclic homology HPpPerfpX{Gqq is naturally a module over HPpPerfpBGqq  krG{{Gsppuqq. For a closed point z P G{{G, we have an identi cation of the formal completion at z with a 2-periodicization of the singular cohomology of the xed points z an z an HPpPerfpX{Gqq C ppX q {pG q ; kqb kppuqq zp dR z an ! as a module over HPpPerfpBGqq  C pBpG q ; kq b kppuqq, contravariantly functorial with respect to zp dR k X . Proof. By Theorem 3.2.3 and Theorem 3.1.12, u Tate u z z Tate; HPpPerfpX{Gqq  OpL pX{Gqq  OpL p pX q{G qq : zp 0 By Proposition 4.1.7 and Theorem 4.3.1, u z z Tate; u z z Tate z z Tate OpL p pX q{G qq  OpL p pX q{G qq  OpLp pX q{G qq : 0 0 0 Finally, by Theorem 4.2.6, we have z z Tate z an z an OpLp pX q{G qq  C p pX q {pG q ; kqppuqq: 0 0 dR It remains to prove Theorem 4.3.1. Central to our proof will be to use the fact that the map is a pro- graded isomorphism; the following lemma establishes a general situation when this is true. Let us rst clarify what we mean by a pro-graded isomorphism, and why this notion is necessary. De nition 4.3.3. A pro-graded chain complex V is an object of PropQCohpBG qq; that is, it is a l- tered limit of graded chain complexes . Letting L denote the weight n twisted one-dimensional G - n m representation, the nth homogeneous part functor is given by wtn pq : ev  PropqpBG ; bL q : PropQCohpBG qq Ñ Vect : m n m k where the functor PropqpBG ;q : PropQCohpBG qq Ñ PropVect q m m k By Proposition 1.1.3.6 of [Lur17], this category is stable. 34 is the functor induced on pro-completions from pBG ;q : QCohpBG q Ñ QCohpptq and the functor m m ev : PropVect q Ñ Vect k k takes a limit diagram and evaluates it in Vect (which has all limits); it is right adjoint to the inclusion. The underlying chain complex is given by ev  Propp q : PropQCohpBG qq Ñ Vect m k where p : pt Ñ BG is the usual atlas so that p is the forgetful functor. A map of graded chain complexes is a pro-graded isomorphism if it is an isomorphism on nth graded parts for all n. Remark 4.3.4. We require this formalism for the following reason. Let p : pt Ñ BG be the standard atlas; the pullback (forgetful functor) p does not commute with limits since the category QCohpBG q cannot di erentiate between direct sums and direct products across di erent weights. In particular, objects of QCohpBG q are Z-graded chain complexes, which are equal to the direct sum of their homogeneous pieces . Example 4.3.5. For example, completing t0u € A {G (under the usual scaling action), n n lim krxs{x  krrxss lim krxs{x  krxs: n;QCohpptq n;QCohpBG q Remark 4.3.6. One way to remedy this is to keep track of the limit diagrams by working in the category PropQCohpBG qq and apply the evaluation functor in Vect  QCohpptq rather than QCohpBG q. How- m k m ever, the category PropQCohpBG qq contains more information than we need: we do not wish to track the topologies on vector spaces as we only care about their completions. Instead, we consider a smaller category: the category of k -modules De nition 4.3.7. We de ne a functor (morally, some kind of Cartier duality) D : QCohpBG q Ñ QCohpZq as follows: it is a standard calculation that QCohpBG q  QCohpptq. For V P QCohpBG q, we m m nPZ À À denote by V the summand corresponding to n P Z. De ne Dp V q  i V where i : Specpkq Ñ Z is n n n; n n the inclusion of the point tnu. Letting r : Z Ñ Specpk q denote the anization map, we de ne a functor : PropQCohpBG qq Ñ Z Z QCohpSpec k qq  k -mod via the composition PropDq Propr q ev Z Z PropQCohpBG qq PropQCohpZqq PropQCohpSpecpk qqq QCohpSpecpk qq: De nition 4.3.8. For any n P Z, there is a natural map  : Spec k Ñ Spec k de ned by projection to the nth coordinate. Denote by k   k P QCohpSpec k q, and note that k is projective since it is a summand n n; n Z Z of the free module decomposed by k  k ` p k q. We de ne the n-stalk of a k -complex M to be n m mn M  k b M and the n-costalk to be  M  Hompk ; Mq; since k is projective the underived functor is n n n n n the derived functor. Remark 4.3.9. The notion of costalk and stalk are canonically equivalent. It is a direct veri cation that !  Z in fact, i pV q  i pV q  V . Let p P k denote the element with 0 in the nth component and 1 everywhere n n else. Let e denote the element with 1 in the nth component and 0 everywhere else. Then, p e  1, so n n n we nd that V  p V ` e V , and i pV q  kerpp q while i pV q  cokerpp q. n n n n Composing with the global sections functors recovers the underlying vector space, and composing with the costalk at n P Spec k recovers the nth homogeneous part. In particular, if we are interested in studying To see this, note that objects of QCohpBG q are chain complexes which are OpG q-coalgebras, i.e. equipped with a map m m V Ñ V b krz; z s. In particular, tensors have nite rank, so any vector can only have nitely many homogeneous parts. 35 the underlying vector space of V P PropQCohpBG qq via its homogeneous components, it suces to consider it as an object of QCohpSpecpk qq. Proposition 4.3.10. We have commutative diagrams of functors Z Z PropQCohpBG qq k -mod PropQCohpBG qq k -mod m m Z ! pSpec k ;q evPropqpBG ;bL q evPropp q m n n Vect Vect k k Proof. We factor the rst diagram as follows Propr Dq ev Z Z PropQCohpBG qq Propk -modq k -mod Propq Propp q ev Propk -modq k -mod and the second in the analogous way. So, the proposition follows from two claims: (1) that the diagrams Z ! above commute without the Pro, i.e. p  pSpec k ;q r  D and pBG ;q    r  D, and (2) that Z Z the evaluation functor Propk -modq Ñ k -mod commutes with global sections and taking costalks. The rst claim can be directly veri ed: it suces to consider abelian categories since all functors are ~ 1 exact. In particular, if V P QCohpBG q , then it is a krz; z s-comodule, i.e. there is a map 1 n V Ñ V b krz; z s  V z : The functor D takes V to the complex on Z whose value on open ane tnu P Z is V . The functor r takes DpV q to V where k acts in the natural way. Finally, we see that the global sections are exactly À À À p V  V and the costalk i p V q  Hom Zpk ; V q  V . n n n n n n k n n Since the evaluation functor is a right adjoint, we prove the second claim by showing that both the global sections functor and the costalks functor are right adjoints, and then using the general fact that right adjoints commute. The global sections functor is right adjoint to the restrictions of scalars functor (it is also left adjoint to the \corestriction of scalars" functor R Hom pk ;q). The costalks functor is right adjoint to the pushforward (and in this case is equal to the stalks functor, which is a left adjoint). Remark 4.3.11. In fact, since the global sections and (co)stalk functors are both left and right adjoints in the above situation, the above construction and proposition are valid for any iteration of taking Pro and Ind categories. De nition 4.3.12. Let V be a k -module. The support of v P V is the closed subscheme de ned by the annihilator ideal of v. Lemma 4.3.13. A pro-graded isomorphism is injective. Proof. This is the easy fact that if a map of sheaves on Specpk q is zero on stalks at closed points, then it is zero, and the observation that on Spec k , costalks and stalks coincide. Lemma 4.3.14. Suppose that f : V Ñ W is a pro-graded isomorphism of pro-graded vector spaces such that either V or W are supported at nitely many weights. Then f is an isomorphism on underlying vector spaces. More generally, let A be a sheaf of algebras on Specpk q, and f : V Ñ W a pro-graded isomorphism of sheaves of A-modules where W is generated by elements supported at nitely many weights. Then, f is an isomorphism on underlying vector spaces. 36 Proof. For the rst claim, the assumptions of the proposition imply that V and W have nite support, whose points consist entirely of closed points of k . A map being a pro-graded isomorphism means that it is an isomorphism at stalks of closed points. For the second more general claim, note that if W is an A-module, and w P W is an element of nitely many weights, say w  w  w where the w are homogeneous of weight c , then w P A  w since 1 r i i i e w  w , where e P k is the characteristic function at i P Z. In particular, W having a set of generators c i c i i supported at nitely many weights is equivalent to W having a set of homogeneous generators. Now, if f : V Ñ W is a pro-graded isomorphism, then it is injective by the previous lemma. For surjectivity, note that for a given homogeneous w P W , since f is a pro-graded isomorphism, we have a homogeneous v P V such that fpvq  w, and surjectivity follows since W is generated by homogeneous elements. The following lemma allows us to reduce statements in the derived category to statements in the abelian category. Lemma 4.3.15. A map f : V Ñ W is a pro-graded quasi-isomorphism of pro-graded complexes if and only i i i if each of the H pfq : H pV q Ñ H pWq are pro-graded isomorphisms of modules. Proof. We can take a map f : V Ñ W of complexes of k -modules. It is easy to verify that taking nth homogeneous parts (i.e. talking stalks via localization) is exact, so that if H pfq is a pro-graded isomorphism, it is an isomorphism of modules and therefore f is a quasi-isomorphism. Conversely, the global sections functor is clearly exact. De nition 4.3.16. A pro-graded dg-algebra is an object of PropAlgpQCohpBG qqq. If A is a pro-graded dg-algebra, then SpecpAq is naturally a dg-indscheme with a G -action in the sense of [GR14]. We will use the word ind-stack to mean a prestack which can be written as an inductive limit of closed embeddings of (derived) QCA stacks (in the sense of [DG13]); in practice we only need the case of a formal completion of a closed substack of a quotient stack. Recall the de nition of a contracting G -action in De nition 2.1.18. ¤0 Lemma 4.3.17. Let A be a noetherian weight Z pro-graded connective dg-algebra, which is generated in wt 0 negative weights over its weight 0 part, and let I   pA q €  pAq be the classical augmentation ideal. 0 0 The derived completion A Ñ A is a pro-graded quasi-isomorphism. Globally, if X is an ind-stack with a representable contracting G -action with xed point locus Z € X , then O Ñ O is a pro-graded quasi- m X y isomorphism of quasicoherent sheaves on X . In particular, OpXq Ñ OpX q is a pro-graded isomorphism. Proof. Choose generators f ; : : : ; f of I . By Proposition 2.4.8 we can compute the homotopy limit A via 1 r I the limit A  lim A b K pAq n n n where K is the Koszul complex for f ; : : : ; f P  pAq. Since f ; : : : ; f are of strictly negative weight, for 0 1 r n 1 r wt¥k wt¥k large n, pK q  p pAqq for any k. Furthermore, homotopy limits can be computed in the derived category of k-complexes, and in particular we can compute the homotopy limit on each graded piece. Thus, wtk wtk pA q is a computed by a limit which stabilizes at A , proving the claim. For the global claim where X is an ind-scheme, one can pass to an open ane G -closed cover (which exists since G is a torus). For m m the global claim where X is an ind-stack, one can check the equivalence on a cover of X . Recall that by Remark 6.11 of [BN12] that there are embeddings LpX{Gq ãÑ L pX{Gq ãÑ T r1s. Thus, ¤0 formal loops and unipotent loops inherit compatible G -actions and their functions are Z pro-graded. We have the following. Lemma 4.3.18. Let X be a geometric stack. The map induced by pullback OpL Xq OpLXq 37 is a pro-graded (quasi-)isomorphism. Proof. An argument is outlined in Corollary 2.7 of [BN13a]; we will repeat it for convenience. By Lemma 2.1.12, LX is geometric. The formal loops LX are the completion of the unipotent loops L X along constant loops, and the action is contracting by Lemma 2.1.19. The statement follows by Lemma 4.3.17. Example 4.3.19. The pro-graded isomorphism of Lemma 4.3.18 may fail to be an isomorphism. For example, take X  G{U where U is any unipotent subgroup of G; then we have that LppG{Uq{Gq  LpBUq  U{U Ñ LpBGq  G{G has image inside the unipotent cone of G. In particular, L pBUq  LpBUq  U{U LpBUq  p u{U: For example, if U  BG , then the map is p p OpL pBG qq  OpG  BG q  krx; s Ñ OpLpBG qq  OpG  BG q  krrxssrs a a a a a a is a pro-graded isomorphism but not an isomorphism, where |x|  0 is a generator for OpG q and ||  1 is a generator for OpBG q. Using the fact that the map is a pro-graded isomorphism, we can show in the case of a unipotent group that the map on Tate-equivariant functions is an isomorphism by a niteness argument. Essentially, we show that applying the Tate construction collapses enough of the target to produce an isomorphism. We include the following proposition as an easy precursor to the next one; it is not required in future arguments. Corollary 4.3.20. Let U be a unipotent algebraic group, and X a quasicompact algebraic space with a U -action. Then, the natural map induced by pullback u Tate Tate OpL pX{Uqq OpLpX{Uqq is an isomorphism. In particular, taking X  pt, u Tate Tate OpL pBUqq  OpU{Uq OpLpBUqq  kppuqq is an isomorphism. Proof. By Lemma 4.3.18, the map is a pro-graded isomorphism. Furthermore, applying the Tate construction, we have that OpLpX{Uqq  H pX ; kqppuqq since U is contractible, where u has cohomological degree 2 and weight 1. The statement follows from Lemma 4.3.14: for quasicompact algebraic spaces X , H pX ; kq is nite-dimensional and therefore has has a homogeneous basis. A tweaking of the above argument gives us the reductive case. Corollary 4.3.21. Let G be a reductive algebraic group, and X a quasicompact algebraic space with a G- action. Then, the natural map induced by pullback u Tate Tate OpL pX{Gqq OpLpX{Gqq is an isomorphism. Proof. Let h denote the universal Cartan of the Lie algebra g of G, and W the universal Weyl group acting on h. By Proposition 4.2.6 we have Tate ! W OpLpBGqq  C pBG; kq b ppuqq  krrhss ppuqq dR 38 W where h is in cohomological degree zero and subcomplex H pBG; kq is given by krhus . By Proposition 2.1.25, u x and the identi cation G{{G  T{{W for reductive groups G, we have that L pBGq  LpBGq t0u, and T{{W in particular by base change, OpL pBGqq  Opt0u u Tate Tate W so that OpL pBGqq  OpLpBGqq  krrhss ppuqq. In particular, u Tate Tate OpL pX{Gqq Ñ OpLpX{Gqq W Z is a map of module objects over the algebra object krrhss ppuqq in the category QCohpSpecpk qq, where h has 0 W W weight -1 and cohomological degree 0, and its cohomology groups are linear over H pkrrhss ppuqqq  krrhss . We claim that the cohomology groups of the target i Tate i ! H pOpLpX{Gqq q  H pC pX{G; kq b kppuqqq dR are nitely generated over krrhss by weight-homogeneous generators; assuming the claim, the result follows from Lemma 4.3.17 and Lemma 4.3.14. To see the claim, we trace through the identi cation of Proposition 4.2.6, keeping track of the weights. The cotangent complex L has weight -1 and u has weight 1 by convention. Thus, we nd that under the identi cation of Proposition 4.2.6, C pX{G; kq has weight 0 and u has weight 1. Therefore, it suces to show dR that the ring H pX{G; kq is nitely generated over H pBG; kq by nitely many cohomologically homogeneous generators. This can be observed via the derived Cartan model for equivariant cohomology, which computes H pX{G; kq via a double complex whose E page is H pXq b krrgss , completing the claim. 1 k We can now prove Theorem 4.3.1, which we restate for convenience. Theorem 4.3.22. Let X be a quasicompact algebraic space with an action of an ane algebraic group G. The natural map induced by pullback u Tate Tate OpL pX{Gqq OpLpX{Gqq is an isomorphism. Proof. Every ane algebraic group G embeds as a subgroup of a reductive group K . Apply the previous G G corollary to pX  Kq{K . Note that X  K is not guaranteed to be a scheme, but is always an algebraic space. References [ABV92] Je rey Adams, Dan Barbasch, and David A. Vogan Jr. The Langlands classi cation and irreducible characters for real reductive groups. Vol. 104. Progress in Mathematics. Birkh auser Boston, Inc., Boston, MA, 1992, pp. xii+318. [AC12] Dima Arinkin and Andrei C ald araru. \When is the self-intersection of a subvariety a bration?" Adv. Math. 231.2 (2012), pp. 815{842. [ACH14] Dima Arinkin, Andrei C ald araru, and M arton Hablicsek. \Formality of derived intersections and the orbifold HKR isomorphism". arXiv: 1412.5233v1 (2014). [AF17] David Ayala and John Francis. \The cobordism hypothesis". arXiv: 1705.02240v2 (2017). [AFR18] David Ayala, John Francis, and Nick Rozenblyum. \Factorization homology I: Higher categories". Adv. Math. 333 (2018), pp. 1042{1177. 39 [AMR17] David Ayala, Aaron Mazel-Gee, and Nick Rozenblyum. \Factorization homology of enriched 8- categories". arXiv: 1710.06414v1 (2017). [AS69] M. F. Atiyah and G. B. Segal. \Equivariant K -theory and completion". J. Di erential Geometry 3 (1969), pp. 1{18. [Beh04] K. Behrend. \Cohomology of stacks". Intersection theory and moduli. ICTP Lect. Notes, XIX. Abdus Salam Int. Cent. Theoret. Phys., Trieste, 2004, pp. 249{294. [Bei08] A. Beilinson. \Remarks on topological algebras". Mosc. Math. J. 8.1 (2008), pp. 1{20, 183. [BF08] Ragnar-Olaf Buchweitz and Hubert Flenner. \The global decomposition theorem for Hochschild (co-)homology of singular spaces via the Atiyah-Chern character". Adv. Math. 217.1 (2008), pp. 243{281. [BFN10] David Ben-Zvi, John Francis, and David Nadler. \Integral transforms and Drinfeld centers in derived algebraic geometry". J. Amer. Math. Soc. 23.4 (2010), pp. 909{966. [BG94] Jonathan Block and Ezra Getzler. \Equivariant cyclic homology and equivariant di erential forms". Ann. Sci.  ecole Norm. Sup. (4) 27.4 (1994), pp. 493{527. [Bha12] Bhargav Bhatt. \Completions and derived de Rham cohomology". arXiv: 1207 . 6193v1 (2012). [BN12] David Ben-Zvi and David Nadler. \Loop spaces and connections". J. Topol. 5.2 (2012), pp. 377{430. [BN13a] David Ben-Zvi and David Nadler. \Loop spaces and representations". Duke Math. J. 162.9 (2013), pp. 1587{1619. [BN13b] David Ben-Zvi and David Nadler. \Nonlinear Traces". arXiv: 1305.7175v3 (2013). [BS15] Bhargav Bhatt and Peter Scholze. \The pro- etale topology for schemes". Ast erisque 369 (2015), pp. 99{201. [CG10] Neil Chriss and Victor Ginzburg. Representation theory and complex geometry. Modern Birkh auser Classics. Reprint of the 1997 edition. Birkh auser Boston, Inc., Boston, MA, 2010, pp. x+495. [CK01] Ionut  Ciocan-Fontanine and Mikhail Kapranov. \Derived Quot schemes". Ann. Sci.  ecole Norm. Sup. (4) 34.3 (2001), pp. 403{440. [Coh16] Lee Cohn. \Di erential Graded Categories are k-linear Stable In nity Categories". arXiv:1308.2587v2 (2016). [DG13] Vladimir Drinfeld and Dennis Gaitsgory. \On some niteness questions for algebraic stacks". Geom. Funct. Anal. 23.1 (2013), pp. 149{294. [Dr e04] Jean-Marc Dr ezet. \Luna's slice theorem and applications". Algebraic group actions and quotients. Hindawi Publ. Corp., Cairo, 2004, pp. 39{89. [GG99] Ezra Getzler and Paul Goerss. \A Model Category Structure for Di erential Graded Coalgebras". https://ncatlab.org/nlab/files/GetzlerGoerss99.pdf (1999). [GKM98] Mark Goresky, Robert Kottwitz, and Robert MacPherson. \Equivariant cohomology, Koszul du- ality, and the localization theorem". Invent. Math. 131.1 (1998), pp. 25{83. [GR14] Dennis Gaitsgory and Nick Rozenblyum. \DG indschemes". Perspectives in representation theory. Vol. 610. Contemp. Math. Amer. Math. Soc., Providence, RI, 2014, pp. 139{251. 40 [GR17] Dennis Gaitsgory and Nick Rozenblyum. A study in derived algebraic geometry. Vols. I. and II. Vol. 221. Mathematical Surveys and Monographs. American Mathematical Society, Providence, RI, [Har75] Robin Hartshorne. \On the De Rham cohomology of algebraic varieties". Inst. Hautes  etudes Sci. Publ. Math. 45 (1975), pp. 5{99. [Hoy18] Marc Hoyois. \The homotopy xed points of the circle action on Hochschild homology". arXiv: 1506.07123v2 (2018). [HP16] Daniel Halpern-Leistner and Daniel Pomerleano. \Equivariant Hodge theory and non-commutative geometry". (2016). arXiv: 1507.01924. [Jon87] John D. S. Jones. \Cyclic homology and equivariant homology". Invent. Math. 87.2 (1987), pp. 403{423. [Kal15] Dmitry Kaledin. \Co-periodic cyclic homology". (2015). arXiv: 1509.08784v1. [Kel06] Bernhard Keller. \On di erential graded categories". International Congress of Mathematicians. Vol. II. Eur. Math. Soc., Zuric  h, 2006, pp. 151{190. [KL87] David Kazhdan and George Lusztig. \Proof of the Deligne-Langlands conjecture for Hecke algebras". Invent. Math. 87.1 (1987), pp. 153{215. [Lod92] Jean-Louis Loday. Cyclic homology. Vol. 301. Grundlehren der Mathematischen Wissenschaften. Appendix E by Mar a O. Ronco. Springer-Verlag, Berlin, 1992, pp. xviii+454. [Lur09a] Jacob Lurie. \Derived Algebraic Geometry IV: Deformation Theory". arXiv:0709.3091v2 (2009). [Lur09b] Jacob Lurie. Higher Topos Theory. Vol. 170. Annals of Mathematics Studies. Princeton University Press, Princeton, NJ, 2009, pp. xviii+925. isbn: 978-0-691-14049-0; 0-691-14049-9. [Lur09c] Jacob Lurie. \On the classi cation of topological eld theories". Current developments in mathe- matics, 2008. Int. Press, Somerville, MA, 2009, pp. 129{280. [Lur11a] Jacob Lurie. \Derived Algebraic Geometry VII: Spectral Schemes". (2011). url: http://www.math. harvard.edu/ lurie/. [Lur11b] Jacob Lurie. \Derived Algebraic Geometry VIII: Quasi-Coherent Sheaves and Tannaka Duality Theorems". (2011). url: http://www.math.harvard.edu/ lurie/. [Lur11c] Jacob Lurie. \Derived Algebraic Geometry XII: Proper Morphisms, Completions, and the Grothendieck Existence Theorem". (2011). url: http://www.math.harvard.edu/ lurie/. [Lur17] Jacob Lurie. Higher Algebra. 2017. url: http://www.math.harvard.edu/ lurie/. [Lur18] Jacob Lurie. Spectral Algebraic Geometry. 2018. url: http://www.math.harvard.edu/ lurie/. [NS18] Thomas Nikolaus and Peter Scholze. \On topological cyclic homology". arXiv:1707.01799v2 (2018). [Pre11] Anatoly Preygel. \Thom-Sebastiani & Duality for Matrix Factorizations". arXiv:1101.5834v1 (2011). [Pre15] Anatoly Preygel. \Ind-coherent complexes on loop spaces and connections". Stacks and categories in geometry, topology, and algebra. Vol. 643. Contemp. Math. Amer. Math. Soc., Providence, RI, 2015, pp. 289{323. 41 [RV18] Emily Riehl and Dominic Verity. Elements of 8-Category Theory. 2018. url: http://www.math. jhu.edu/ eriehl/elements.pdf. [Sta18] The Stacks Project Authors. Stacks Project. url: http://stacks.math.columbia.edu. 2018. [Ste75] Robert Steinberg. \On a theorem of Pittie". Topology 14 (1975), pp. 173{177. [Tho86] R. W. Thomason. \Lefschetz-Riemann-Roch theorem and coherent trace formula". Invent. Math. 85.3 (1986), pp. 515{543. [Tho87] R. W. Thomason. \Algebraic K -theory of group scheme actions". Algebraic topology and algebraic K -theory (Princeton, N.J., 1983). Vol. 113. Ann. of Math. Stud. Princeton Univ. Press, Princeton, NJ, 1987, pp. 539{563. [Tho88] R. W. Thomason. \Equivariant algebraic vs. topological K -homology Atiyah-Segal-style". Duke Math. J. 56.3 (1988), pp. 589{636. [To e06] Bertrand To en. \Champs anes". Selecta Math. (N.S.) 12.1 (2006), pp. 39{135. [To e07] Bertrand To en. \The homotopy theory of dg-categories and derived Morita theory". Invent. Math. 167.3 (2007), pp. 615{667. [To e14] Bertrand To en. \Derived algebraic geometry". EMS Surv. Math. Sci. 1.2 (2014), pp. 153{240. [TV05] Bertrand To en and Gabriele Vezzosi. \Homotopical algebraic geometry. I. Topos theory". Adv. Math. 193.2 (2005), pp. 257{372. [TV08] Bertrand To en and Gabriele Vezzosi. \Homotopical algebraic geometry. II. Geometric stacks and applications". Mem. Amer. Math. Soc. 193.902 (2008), pp. x+224.

Journal

Mathematics – arXiv (Cornell University)

Published: Aug 21, 2017

There are no references for this article.