Stable motivic invariants are eventually étale local
aa r X i v : . [ m a t h . K T ] M a r STABLE MOTIVIC INVARIANTS ARE EVENTUALLY ´ETALE LOCAL
TOM BACHMANN, ELDEN ELMANTO, AND PAUL ARNE ØSTVÆR
Abstract.
We prove a Thomason-style descent theorem for the motivic sphere spectrum, and deducean ´etale descent result applicable to all motivic spectra. To this end, we prove a new convergence resultfor the slice spectral sequence, following work by Levine and Voevodsky. This generalizes and extendsprevious ´etale descent results for special examples of motivic cohomology theories. Combined with ´etalerigidity results, we obtain a complete structural description of the ´etale motivic stable category.
Contents
1. Introduction 21.1. What is done in this paper? 21.2. Bott elements and multiplicativity of the Moore spectrum 41.3. Some applications 41.4. Overview of ´etale motivic cohomology theories 41.5. Terminology and notation 51.6. Acknowledgements 62. Preliminaries 62.1. Endomorphisms of the motivic sphere 62.2. ρ -completion 72.3. Virtual ´etale cohomological dimension 83. Bott-inverted spheres 83.1. τ -self maps 83.2. Cohomological Bott elements 93.3. Spherical Bott elements 104. Construction of spherical Bott elements 104.1. Construction via roots of unity 104.2. Lifting to higher ℓ -powers 104.3. Construction by descent 114.4. Construction over special fields 124.5. Summary of τ -self maps 135. Slice convergence 136. Spheres over fields 187. Main result 19Appendix A. Multiplicative structures on Moore objects 23A.1. Definitions and setup 23A.2. Review of Oka’s results 23A.3. Asymptotic unicity and associativity 24A.4. Uniqueness for H Z Mathematics Subject Classification.
Primary: 14F42, 55P42; Secondary: 14F20, 18N55, 18N60. Introduction
The goal of this paper is to push, as far as possible, the idea that inverting so-called “Bott elements”on certain cohomology theories for algebraic varieties results in ´etale descent. The cohomology theoriesthat we are interested in are those which are represented in Morel-Voevodsky’s stable motivic homotopycategory, and thus satisfy only Nisnevich descent. Results of this form are extremely useful when onetries to approximate the values of these cohomology theories by studying their ´etale “sheafifications”which are usually more tractable due to the presence of finer covers. For example, given appropriatefiniteness hypotheses, there is a spectral sequence whose input involves ´etale cohomology converging tothe ´etale sheafified version of these theories.The story for these motivic cohomology theories parallels the one for algebraic K -theory. Havingproved the Nisnevich descent theorem for algebraic K -theory [TT90], Thomason, in his ICM address[Tho91], explained how this “(unleashed) a pack of new fundamental results for K -theory”. One of themwas the comparison result with ´etale K -theory [Tho91, § K -theory satisfy ´etale (hyper)descent. Whencethere is a spectral sequence whose E -terms are ´etale cohomology groups, abutting to Bott-invertedalgebraic K -theory. A recent treatment of this result using completely different methods was carried outin [CM19].Levine followed Thomason’s strategy for motivic cohomology (in the form of Bloch’s higher Chowgroups), at least over fields [Lev00]. More recently, in joint work with Levine and Spitzweck, the lasttwo authors were able to prove the cases of the same result for algebraic cobordism [ELSØ17] and thusoriented theories in general. This uses Levine’s results as input, while it also recovers Thomason’s resultand generalizes Levine’s results over other bases. Another lens to view these results is through theMilnor/Bloch-Kato and Beilinson-Lichtenbaum conjectures (see [HW19] for a textbook overview), whichwere proved by Voevodsky [Voe11, Voe03]. If X is a smooth scheme over a field and ℓ is a prime invertiblein k , then there is a canonical Nisnevich/Zariski-to-´etale comparison map(1.1) H p mot ( X, Z /ℓ ( q )) → H p ´et ( X, µ ⊗ qℓ ) . Voevodsky proved that (1.1) is an isomorphism whenever p ≤ q , which allows one to compute motiviccohomology of schemes in a certain range. The Bott element of interest is a certain element on the left-hand-side of (1.1) which is of degree 0 and of positive weight (depending on the presence of roots of unityin the ground field). Therefore, inverting this element will “periodize” Voevodsky’s isomorphism. ThatBott-inversion results in ´etale descent, while an easier result, tells us that “asymptotically” invariants likemotivic cohomology and algebraic K -theory do behave like their ´etale counterparts; we refer the readerto [DFST82] for the colorful slogan: “algebraic K -theory eventually surjects onto topological K -theory.”1.1. What is done in this paper?
For any scheme S , we denote by SH ( S ) the stable motivic categoryover S . In this category we have a morphism ρ : → (1)[1] , classified by the point “ −
1” in G m . Furthermore, any motivic spectrum E ∈ SH ( S ) comes equippedwith a canonical map E → L ´et E , the ´etale localization , witnessing L ´et E as the initial motivic spectrumreceiving a map from E and satisfying ´etale hyperdescent; this map is just the unit map of the usualadjunction between SH ( S ) and its ´etale local version SH ´et ( S ). Next we state a version of the main resultof this paper. Theorem 1.2. (see Theorem 7.4) Suppose S is a Noetherian scheme of finite dimension, and let ≤ m, n ≤ ∞ . Assume the following conditions hold.(1) /ℓ ∈ S ,(2) For every s ∈ S we have vcd ℓ ( s ) < ∞ , and(3) There exists a good τ -self map (in the sense discussed in § τ : / ( ℓ n , ρ m ) → / ( ℓ n , ρ m )( r ) . Then for every E ∈ SH ( S ) ∧ ℓ,ρ the map E/ ( ℓ n , ρ m ) → E/ ( ℓ n , ρ m )[ τ − ] ∧ ℓ,ρ is an ´etale localization. TABLE MOTIVIC INVARIANTS ARE EVENTUALLY ´ETALE LOCAL 3
We note that if m, n < ∞ , then the ( ℓ, ρ )-completion is unnecessary; see Remark 3.2. A summary ofthe good τ -self maps that we managed to construct can be found in § ℓ is an odd prime. Then recall that (as discussed in e.g., § E , we have a splitting: E/ℓ n ≃ E + /ℓ n ∨ E − /ℓ n . In this situation, ´etale localization only involve the “+” part of
E/ℓ n : the map E + /ℓ n → E + /ℓ n [ τ − ]is an ´etale localization.(2) If ℓ = 2 then the situation is more complicated. In the presence of a square root of − ρ -completion is harmless and we get that E/ n → E/ n [ τ − ]is an ´etale localization. Otherwise we have to contend ourselves with ρ -completion.Indeed, dealing with the prime 2 is one of the key technical challenges of this paper that we wereable to overcome in many cases. We will now explain this point and the overall strategy of the proofof Theorem 1.2. We proceed by examining the slice spectral sequence for the Bott inverted spherespectrum over a field (Corollary 6.2), following the general strategy in [ELSØ17] for the case of algebraiccobordism. An examination of the form of the slices [RSØ19, Theorem 2.16] reveals that they arejust motivic cohomology and hence, modulo certain cases which are dealt with using the Beilinson–Lichtenbaum conjectures, satisfy ´etale descent after inverting these Bott elements due to Levine’s results[Lev00]. However, as already elaborated in [ELSØ17, Section 4], examining the slice tower of the Bottinverted sphere spectrum requires a delicate analysis — the crux point is that the process of Bott inversionis a colimit , while the slices (or, more accurately, the co-slices) try to approximate the sphere spectrumas a limit . At the prime 2, and for fields with non-finite 2-cohomological dimension (infinite R ), thesituation is worse: even the slice filtration itself is not convergent [Lev13, Remark]. From this point ofview, the situation seems hopeless.On the other hand, previous attempts to cope with this infinitary phenomenon to still obtain aThomason-style descent theorem have found success under the assumption of finite virtual cohomologicaldimension . To our knowledge, the first paper of this sort was written by the third author in [Øst03] andlater generalized in joint work with Rosenschon [RØ05]. The case of hermitian K -theory was settled in[BKSØ15].It is at this spot that we employ a different analysis from [ELSØ17], which also led to substantialimprovements even for the case of algebraic cobordism, other oriented theories and also recovers thedescent results of the papers in the previous paragraph. In loc. cit. , particularly in [ELSØ17, Section6.5] and [ELSØ17, Section 4], the last two authors (with Levine and Spitzweck) examined the resultingslice spectral sequences on Bott-inverted algebraic cobordism and on ´etale algebraic cobordism. Thisrelied on subtle convergence results on inverting elements in a spectral sequence and an analysis of theconstituent spaces and bonding maps of an ´etale-localized motivic spectra in a range. In this paper,we work directly with the slice filtration and prove a convergence result in the form of Lemma 6.1which cleanly isolates the role of the convergence of the slice filtration. At the prime 2 and with theweaker assumption that the ambient field k has finite virtual cohomological dimension, it turns out thatconvergence holds after completion with respect to the map ρ : → G m induced by the unit − ∈ k × . We prove a variant of Levine’s convergence result [Lev13], which webelieve is of independent interest (see Theorem 5.3 for a more precise version). Theorem 1.3.
Let k be a field of exponential characteristic e and ℓ > coprime to e such that vcd ℓ ( k ) < ∞ . Suppose that E/ℓ ∈ SH ( k ) is bounded below in the homotopy t -structure and suppose that there exists R ≫ for which E ρ R −−→ E ∧ G ∧ Rm is zero. Then, the filtration on π i,j ( E ) induced by Voevodsky’s slice tower f • E → E is separated andexhaustive, i.e., the filtration is convergent. We note that the finite vcd assumption also improves the range of fields and schemes for which[ELSØ17] applies. This theorem is related to Voevovodsky’s convergence conjecture, as explained in thenext remark.
TOM BACHMANN, ELDEN ELMANTO, AND PAUL ARNE ØSTVÆR
Remark 1.4.
One of the last remaining open problems in Voevodsky’s list [Voe02] is his “convergenceproblem” stated as Conjecture 13 of loc. cit. — over any perfect field, the slice filtration on a finitespectrum induces a separated filtration on its homotopy sheaves; see also [Lev13, Conjecture 1]. Thisconjecture turns out to be false in general [Lev13, Remark on page 909] and has since been modified byLevine to cover only fields of finite virtual cohomological dimension [Lev13, Conjecture 5]. Theorem 5.3can be seen as settling a derived version of this conjecture; see Remark 5.7 for details.1.2.
Bott elements and multiplicativity of the Moore spectrum.
There are two further, related,issues which we address in the paper: (1) construction of a suitable “spherical” Bott element and (2) thelack of a multiplicative structure on the motivic Moore spectrum. What is at stake in point (1) is thefact that the sphere spectrum is not oriented. In previous iterations of Thomason-style descent resultslike in [Tho85], [Jar10], [ELSØ17], one produces Bott elements out of n -th roots of unity present in theambient scheme/ring/field which is naturally an n -torsion element in an appropriate group; see [ELSØ17,Section 6.4] or [Tho85, Appendix A]. For the sphere spectrum, due to so-called “ η -logarithmic relation”in Milnor-Witt K -theory [ ab ] = [ a ] + [ b ] + η [ a ][ b ] , a root of unity is n ǫ -torsion (see Proposition 2.1), rather than n -torsion. At odd primes, the right thingto do is to look at the part of the sphere spectrum where the Milnor-Witt number n ǫ = n , i.e., theso-called “+” part.At the prime 2 the story gets more interesting, coupled with the usual complication that the mod-2Moore spectrum does not admit a multiplication. Producing a mod-2 Bott element is the first timewhere ρ -completion enters the picture: see § insufficient for our reduction steps (see the argument in Theorem 7.4),primarily because we do not know that this module structure satisfies the usual associativity axiom. Inthe appendix, we use an idea originating in the work of Davis-Lawson and Hopkins, that this action canbe made asymptotically associative if we are willing to consider the pro-system { / n } ; see § A.In any case, § ρ and η ). A search for more τ -self maps is underway and willstrengthen the scope of this paper.1.3. Some applications.
We now summarize some applications of our main results.
Theorem 1.5.
Assumptions as in Theorem 7.4, where m, n < ∞ :(1) If S is defined over a field containing a primitive ℓ -th root of unity and satisfying k × /ℓ = { } ,then there exists a Bott element τ in ∧ ℓ and SH ´et ( S ) ∧ ℓ ≃ SH ( S ) ∧ ℓ [ τ − ] . (2) In general, ( ℓ, ρ ) -complete ´etale localization is equivalent to Bousfield localization at the homologytheory / ( ℓ n , ρ m )[ τ − ] .(3) ´Etale localization is smashing on SH ( S ) ∧ ℓ,ρ .(4) If f : T → S is finite type, the base change functor f ∗ on SH ( S ) ∧ ℓ,ρ preserves ´etale local objectsso that the ´etale local sphere (in the ( ρ, ℓ ) -complete category) is stable under base change. These corollaries are discussed right after Theorem 7.4. The first statement should be considered the“model statement” — the limitation comes from the fact that the construction of higher Bott elementsrequires more multiplicative structure on the ρ -complete sphere than what we already know. In lightof this, the second and third statement are weaker but also pleasant consequences of our results. Inparticular, they tell us that on the ( ρ, ℓ )-complete categories the inclusion of ´etale-local objects preservescolimits. Lastly, base change results in motivic homotopy theory are rewarding but often hard to comeby — our results proves this by way of knowing that the τ -self maps are manifestly stable under basechange.1.4. Overview of ´etale motivic cohomology theories.
This paper completes, in many cases, ourstructural understanding of how ´etale motivic theories behave; we now sketch this. Suppose that T isa premotivic functor in the sense of [CD19] or a functor satisfying Ayoub’s axioms as in [Ayo07, 1.4.1].The most prominent examples are T = SH , DM or M od E where E is a (highly structured) motivic ring For this section, we take this in the sense of [CD19, Chapter 11]
TABLE MOTIVIC INVARIANTS ARE EVENTUALLY ´ETALE LOCAL 5 spectrum such as MGL which is defined over Z . We can also consider the ´etale local version of T , whichwe denote by T ´et and comes equipped with a premotivic adjunction L ´et : T ⇄ T ´et : i ´et . The category T ´et fits into a fracture square of the form: T ´et Q ℓ ( T ´et ) ∧ ℓ T ´et , Q ( Q ℓ ( T ´et ) ∧ ℓ ) Q . The rational part T ´et , Q coincides with T Q in many cases. In fact, a result of Cisinski-D´eglise (summarizedin, say, [ELSØ17, Theorem 12.2]) furnishes equivalences SH ´et ( S ) Q ≃ SH ( S ) + Q ≃ DM ( S ) Q , for any Noetherian, geometrically unibranch scheme S . Furthermore, as discussed in Remark 7.6, the prime-at- ℓ of T ´et part is always zero. Hence, what remain are the ℓ -complete parts of T ´et where ℓ iscoprime to the residual characteristics. In this situation, we seek two types of theorems concerning theinduced adjunction L ´et : T ∧ ℓ ⇄ ( T ´et ) ∧ ℓ : i ´et . Suslin-style rigidity If ℓ is coprime to the residual characteristics, then ( T ´et ) ∧ ℓ is described as a certain category ofsheaves over the small ´etale site.Thomason-style descent If ℓ is coprime to the residual characteristics, then the endofunctor i ´et L ´et is computed explicitlyby an inversion of a “Bott-element” τ .In conjunction, these two results give us access to the values of the cohomology theories representedin T ´et . One way to make this concrete, at least in cases where we have “Postnikov completeness” of the´etale site of a scheme X (see [CM19, Section 2] for a modern discussion when this happens) we obtain aconditionally convergent spectral sequence: H p et ( X, π ´et q,w ( E )) ⇒ [ X ( w )[ q − p ] , E [ τ − ]] = ( E [ τ − ]) p − q, − w ( X )The input of this spectral sequence is obtained as consequence of Suslin-rigidity; we note that it consistsof usual ´etale cohomology groups with coefficients in a (torsion/ ℓ -complete) sheaf of abelian groups onthe small ´etale site of X . The target is obtained as a consequence of Thomason-style descent. Thisphilosophy was already known to Thomason at the beginning in [Tho85]. This should be contrastedwith the slice spectral sequence, available in the Nisnevich/Zariski setting, where the input consists ofmotivic cohomology groups which are largely unknown.We briefly recall Suslin-style rigidity. In [Bac18c], the first author has established (again, in manycases) Suslin-style rigidity for T = SH . Earlier, analogous results were established by R¨ondigs-Østvær[RØ08a] and for T = DM by Ayoub [Ayo14] and Cisinski-D´eglise [CD16], building on the case of fieldswhere we have Suslin’s eponymous result (see [MVW06, Theorem 7.20] for an exposition).In this light, what one needs to understand is the precise gap between T and T ´et . For T = SH (andup to certain ρ -completions), this is exactly the τ -complete category, which remains mysterious.1.5. Terminology and notation.
We freely use the language of ∞ -categories, as set out in [Lur09,Lur16].1.5.1. Motivic homotopy theory. • We denote by SH ´et ( S ) the localization of SH ( S ) at the ´etale hypercovers and by SH ( S ´et ) thestabilization of the small hypercomplete ´etale ∞ -topos of S . We denote by L ´et the (various) ´etalehyperlocalization functors. We call a map E → F an ´etale localization if L ´et E ≃ L ´et F ≃ F . • We denote by E ∧ ℓ = lim n E/ℓ n (or sometimes E/ℓ ∞ ) the ℓ -completion of a spectrum; we write SH ( S ) ∧ ℓ for the category of ℓ -complete motivic spectra. • We denote by SH ( k ) ≥ , SH ( k ) ≤ the homotopy t -structure on SH ( k ) defined in [Mor04a, Section5.2]. • We put (1) = G m [ −
1] and ( n ) = (1) ∧ n . For E ∈ SH ( S ) we put E ( n ) = E ∧ ( n ). • As is standard we write T = A / A − • We denote by H Z Spitzweck’s motivic cohomology spectrum [Spi18] and write DM ( S ) for the ∞ -category of modules over H Z [RØ08b]. TOM BACHMANN, ELDEN ELMANTO, AND PAUL ARNE ØSTVÆR
Field theory.
We will adopt the following terminology concerning field theory. Let k be a field, k sep a separable closure of k and G k := Gal( k sep /k ) the absolute Galois group. Moreover, we write: • cd ℓ ( k ) for the ℓ -cohomological dimension of k in the sense of [Ser02, Section 3.1]. • vcd ( k ) := cd ( k [ √− virtual -cohomological dimension . • More generally, for an integer s we putvcd s ( k ) := max { vcd ℓ ( k ) | p | s } , where vcd ℓ ( k ) := cd ℓ ( k ( √− . Note that vcd ℓ ( k ) = cd ℓ ( k ) unless possibly if ℓ = 2 [Ser02, Proposition II.10’]. • The cohomological dimension cd( k ) of k is defined as cd( k ) = sup ℓ { cd ℓ ( k ) } .1.6. Acknowledgements.
We would like to thank Joseph Ayoub, Jeremy Hahn, Mike Hopkins, TylerLawson, Marc Levine, Denis Nardin, and Markus Spitzweck for helpful discussions. The authors thankthe Isaac Newton Institute for Mathematical Sciences for support and hospitality during the programme K -theory, algebraic cycles and motivic homotopy theory. This work was supported by EPSRC GrantNumber EP/R014604/1 and the RCN Frontier Research Group Project no. 250399 “Motivic HopfEquations.” Part of this work was carried out while Elmanto was a postdoc at the Center for Symmetryand Deformation at the University of Copenhagen, which also supported Bachmann’s visit in 2019. Weare grateful for the Center’s support. Østvær was partially supported by the Humboldt Foundation,Professor Ingerid Dal and sister Ulrikke Greve Dals prize for excellent research in the humanities, and aGuest Professorship under the auspices of The Radbound Excellence Initiative.2. Preliminaries
Endomorphisms of the motivic sphere.
We denote by h− i : → ∈ SH ( S )the map corresponding to the switch map P ∧ P → P ∧ P . Clearly we have h− i = 1. It follows thatif E ∈ SH ( S ) is such that E −→ E is an equivalence, then there is a canonical decomposition E ≃ E + ∨ E − which is characterised by the fact that h− i acts as the identity id on E + and as − id on E − [CD19, § SH ( S )[1 / + , SH ( S )[1 / − ⊂ SH ( S )[1 / E ≃ E ± . For n ∈ N we define the n -th Milnor-Witt number as n ǫ = n X i =1 h ( − i − i ∈ [ , ] SH ( S ) , where h i := 1. We will make use of the following result. Proposition 2.1.
Suppose that S is a base scheme.(1) If − is a square on S , then h− i = 1 .(2) Denote by p n : G m → G m the map corresponding to x x n (i.e., the n -th power map), and by ¯ p n : → the desuspended map. Then ¯ p n = n ǫ .Proof. (1) We claim the following more general statement: for a ∈ O × ( S ), denote by h a i : → theendomorphism in SH ( S ) induced by the endomorphism of S -schemes P → P , [ x : y ] [ ax : y ]. Thenwe have h a i = 1 . To prove the claim, we first note that the map ( x : y ) ( a x : y ) is equal to ( x : y ) ( ax : a − y ).We have a map GL ( S ) → Map( P , P ). It is thus enough to connect the matrix A = (cid:20) a − a (cid:21) tothe identity matrix in GL ( S ) via A -paths. By “Whitehead’s Lemma”, A is a product of elementarymatrices: (cid:20) a − a (cid:21) = (cid:20) /a (cid:21) (cid:20) − a (cid:21) (cid:20) −
10 1 (cid:21) (cid:20) − a − (cid:21) . Since the space of elementary matrices is A -path connected, the result follows.(2) For this proof, we will need some rudiments of the theory of framed correspondences in the senseof [EHK +
19, GP18]. Under the equivalence Σ G m ≃ P , the map p n suspends to the map q n : P → P , ( x : y ) ( x n : y n ); hence it suffices to prove the claim for this map. Further note that the projection TABLE MOTIVIC INVARIANTS ARE EVENTUALLY ´ETALE LOCAL 7 c : P → P / P \ ≃ T is an equivalence, P \ ≃ A being contractible. It follows that it suffices todetermine the map c ◦ q n : P → T which is precisely the map induced via Voevodsky’s Lemma [EHK +
19, Corollary A.1.7] from the equa-tionally framed transfer with support 0 ∈ A and framing t n (in the sense of [EHK +
19, Definition 2.1.1]).It follows from the main result of [EHK +
18] (in particular, the agreement results [EHK +
18, Theorems3.2.11, 3.3.6]) that this is stably the same as the map → induced by the corresponding framedcorrespondence via the reconstruction theorem of [EHK +
19, Theorem 3.5.11] and its generalization overany base [Hoy18, Theorem 16]. The result now follows from [EHK +
19, Proposition B.1.4], providedthat their n ǫ is the same as ours. Their h a i comes from the canonical action of O × ( S ) → Ω K ( S ) onframed correspondences, which coincides with the canonical action via fundamental classes by [EHK + (cid:3) Remark 2.2.
Given u ∈ O × ( S ) we let [ u ] denote the resulting map → G m . One may show that h− i = 1 + η [ − , where η : G m → is the geometric Hopf map.2.2. ρ -completion. In this section, we discuss ρ -completed motivic spectra in view of certain conver-gence results that we will use later; we use [MNN17, Section 2.2] as a reference for the formalism ofcomplete objects and completions; see also [RØ08a] for a previous reference in the motivic context. Weput ρ := [ −
1] : → G m ∈ SH ( S ) . By abuse of notation, we also denote the map G ∧− m → (obtained by smashing ρ with id G ∧− m ) by ρ . Wewrite /ρ for the cofiber of ρ : G ∧− m → . Recall that a morphism α : E → F ∈ SH ( S ) is a ρ -equivalence if α ∧ /ρ : E/ρ → F/ρ is an equivalence. We denote by SH ( S ) ∧ ρ the localization of SH ( S ) at the ρ -equivalences. Recall that E ∈ SH ( S ) ∧ ρ , i.e., it is ρ -complete , if andonly if for all F such that F ⊗ /ρ ≃
0, the space Map(
F, E ) is contractible.From general principles the localization SH ( S ) → SH ( S ) ∧ ρ has a fully faithful right adjoint, and thecomposite localization functor is given by [MNN17, Formula (2.22)] E E ∧ ρ := lim n E/ρ n . Remark 2.3.
Functors that preserve cofibers and smashing with G m preserve ρ -equivalences. Since G m is invertible the latter condition is in particular satisfied by symmetric monoidal functors and theiradjoints. It follows that the motivic base change functors f ∗ , f , and f ∗ preserve ρ -equivalences. Warning 2.4.
On the other hand, the functors related to the slice filtration, like f n and s n , do notinteract well with ρ -equivalences. See for example Remark 5.20.In order to streamline the exposition, in what follows we will extensively use the category SH ( S ) ∧ ρ .In many cases, however, this has very little effect: Remark 2.5. (1) Suppose RS = ∅ , i.e., − O S . Then ρ ∈ π ( ) ∗ isnilpotent, as follows, for example, from the main result of [Bac18a]. It follows that SH ( S ) ∧ ρ = SH ( S ) . (2) Similarly, over a general base, ρ is nilpotent in [1 / + and a unit in [1 / − , see [Bac18a, Lemma39]. It follows that SH ( S )[1 / ∧ ρ = SH ( S )[1 / + . Example 2.6.
Let S be essentially smooth over a Dedekind domain. Then H Z ∈ SH ( S ) is ρ -complete.Indeed it suffices to show that for X ∈ Sm S , i, j ∈ Z , we have[Σ i,j X + , H Z ∧ G ∧− dm ] = 0for d sufficiently large. This follows from the vanishing of motivic cohomology in negative weights oversuch bases [Spi18, Corollary 7.19]. For a scheme X , we denote by RX the set of pairs ( x, α ) with x ∈ X and α an ordering of k ( x ). TOM BACHMANN, ELDEN ELMANTO, AND PAUL ARNE ØSTVÆR
Virtual ´etale cohomological dimension.
We will need the following result about motivic coho-mology in large degrees.
Lemma 2.7.
For m > vcd ℓ ( k ) and a ∈ Z the map H m ( k, Z /ℓ ( a )) [ − −−−→ H m +1 ( k, Z /ℓ ( a + 1)) is an isomorphism.Proof. If a < m then both groups vanish by [SV00, Lemma 3.2(2)] ([MVW06, Lemma 5.2]), so there isnothing to prove. Thus let a ≥ m , so that H m ( k, Z /ℓ ( a )) ≃ H m ´et ( k, µ ⊗ aℓ )by the Beilinson-Lichtenbaum conjecture [Voe11, Theorem 6.17] ([MVW06, Theorem 10.2]). If ℓ is oddthen cd ℓ ( k ) = vcd ℓ ( k ) and again both groups are zero. If ℓ = 2 then µ ⊗ aℓ = Z / (cid:3) Bott-inverted spheres τ -self maps.Definition 3.1. Let ℓ be a prime with ℓ ∈ O × ( S ) and 1 ≤ m, n ≤ ∞ . If m, n < ∞ denote by / ( ℓ n , ρ m )the obvious cofiber; if n = ∞ and m < ∞ we set / ( ℓ ∞ , ρ m ) = ( /ρ m ) ∧ ℓ , and similarly if n = ∞ .Suppose given a map ˜ τ : / ( ℓ n , ρ m ) → / ( ℓ n , ρ m )( r ) . (1) We write / ( ℓ n , ρ m )[˜ τ − ] = colim (cid:20) / ( ℓ n , ρ m ) ˜ τ −→ / ( ℓ n , ρ m )( r ) ˜ τ ( r ) −−→ / ( ℓ n , ρ m )(2 r ) → . . . (cid:21) . More generally, for E ∈ SH ( S ) we set E/ ( ℓ n , ρ m )[˜ τ − ] = E ∧ / ( ℓ n , ρ m )[˜ τ − ] . (2) We call ˜ τ a τ -self map if for every map x → S , where x is the spectrum of a field, the map H Z x / ( ℓ n , ρ m ) → H Z x / ( ℓ n , ρ m )[˜ τ − ] ∧ ℓ,ρ ∈ SH ( x ) ∧ ℓ,ρ is an ´etale localization.(3) We call ˜ τ a good τ -self map if it is a τ -self map and, for every E, F ∈ SH ( S ), every morphism α : E → F/ ( ℓ n , ρ m )[˜ τ − ] ∧ ℓ,ρ factors through E → E/ ( ℓ n ′ , ρ m ′ )[˜ τ ′− ], where ˜ τ ′ is another τ -selfmap. Remark 3.2.
Unless m = ∞ or n = ∞ , E/ ( ℓ n , ρ m )[˜ τ − ] is already ( ρ, ℓ )-complete. Lemma 3.3.
Let τ : / ( ℓ n , ρ m ) → / ( ℓ n , ρ m )( r ) be a τ -self map. Then L ´et ( τ ) ∧ ℓ,ρ is an equivalence.Proof. Suppose that n = ∞ . The map L ´et ( τ ) ∧ ℓ,ρ is an equivalence if and only if L ´et ( τ /ℓ ) ∧ ℓ,ρ is (by ℓ -completeness). Since τ /ℓ is also a τ -self map, replacing τ by τ /ℓ we may assume that n < ∞ . Similarlywe may assume that m < ∞ .By continuity, any such τ -self map can be defined over a scheme of finite type over Z [1 /ℓ ], so we mayassume that S is of this form. Thus by [Bac18c, Example 5.9], the main theorem of loc. cit. applies to S and SH ´et ( S ) ∧ ℓ ≃ SH ( S ∧ ´et ) ∧ ℓ . This implies that pulling back along maps of the form Spec ( k ) → S , where k is separably closed, is a conservative family for SH ´et ( S ) ∧ ℓ . We may thus replace S by the spectrum ofa separably closed field k .Since L ´et / ( ℓ n , ρ m ) ∈ SH ≥− , in order to show that L ´et ( τ ) ∧ ℓ,ρ is an equivalence, it suffices to showthis in homology. In other words, we need to show that the map H Z / ( ℓ n , ρ m ) τ −→ H Z / ( ℓ n , ρ m )( r )is an ´etale equivalence. This is true since by part (2) of Definition 3.1, we demand that H Z / ( ℓ n , ρ m ) → H Z / ( ℓ n , ρ m )[ τ − ] ∧ ℓ,ρ is an ´etale localization. (cid:3) TABLE MOTIVIC INVARIANTS ARE EVENTUALLY ´ETALE LOCAL 9
Cohomological Bott elements.
We shall in the next subsection construct τ -self maps as multi-plication by suitable elements. In preparation, we study the analogous question for H Z . Definition 3.4.
Let S = Spec ( k ), k a field. Let 1 ≤ m, n ≤ ∞ and r >
0. By a cohomological Bottelement we mean an element τ ∈ π , − r H Z / ( ℓ n , ρ m ) such that for every field l/k and every choice ofmultiplication on H Z / ( ℓ n , ρ m ) under the standard multiplication on H Z ∧ ρ /ℓ n (in the sense of § A), themap H Z / ( ℓ n , ρ m ) | l → H Z / ( ℓ n , ρ m ) | l [ τ − ] ∧ ℓ,ρ is an ´etale localization.Here, H Z / ( ℓ n , ρ m )[ τ − ] denotes the mapping telescope of the self map given by multiplication by τ . Remark 3.5.
In light of Example 2.6, H Z /ℓ m is already ρ -complete, so we shall suppress the additionalcompletion from now on. Remark 3.6. If τ , τ are cohomological Bott elements, then so is τ τ . Indeed by cofinality, E [( τ τ ) − ](where E = H Z / ( ℓ n , ρ m )) can be computed as a colimit over N × N , with horizontal maps given bymultiplication by τ and vertical maps by τ . Computing the horizontal colimit first, it suffices to showthat multiplication by τ induces an equivalence on E [ τ − ]; this follows from the assumption that τ and τ are both cohomological Bott elements. Lemma 3.7.
Let τ ∈ π , − r H Z /ℓ and τ ∈ π , − r ′ H Z /ℓ n be elements such that τ reduces to a power of τ modulo ℓ (here n = ∞ is allowed).(1) τ is a cohomological Bott element if and only if τ is a cohomological Bott element.(2) The image τ in π , − r ′ H Z / ( ℓ n , ρ m ) is a cohomological Bott element.Proof. (1) Suppose first that n < ∞ . In D ( Z /ℓ n ) we have a cofiber sequence Z /ℓ n ℓ −→ Z /ℓ n → Z /ℓ ⊕ Z /ℓ n − [1]; hence the same holds in SH ( k ). Thus if H Z /ℓ n [ τ − ] is ´etale local then so is H Z /ℓ [ τ − ] ≃ H Z /ℓ [ τ − ]. Moreover, the converse holds if also Z /ℓ n − [ τ − ] is ´etale local; this will hold by inductionon n . Finally suppose n = ∞ . Then H Z /ℓ [ τ − ] ≃ H Z ∧ ℓ [ τ − ] ∧ ℓ /ℓ , so τ is a cohomological Bott elementif τ is. Conversely, if τ is a cohomological Bott element then H Z ∧ ℓ [ τ − ] ∧ ℓ ≃ lim n H Z /ℓ n [ τ − ] is alimit of ´etale local spectra, by what we already established, so it is ´etale local. In other words, τ is acohomological Bott element.(2) Write τ for the image. Then for any choice of compatible multiplication, the map H Z / ( ℓ n , ρ m ) → H Z / ( ℓ n , ρ m )[ τ − ] is the cofiber of multiplication by ρ m on the map H Z /ℓ n → H Z /ℓ n [ τ − ]. The latteris an ´etale localization by (1), and hence so is the former.This concludes the proof. (cid:3) Lemma 3.8.
Let l/k be a finite separable extension of degree coprime to ℓ , and τ ∈ π , − r H Z /ℓ n . Then τ is a cohomological Bott element if and only if τ | l is a cohomological Bott element.Proof. It suffices to show that H Z /ℓ n [ τ − ] is a summand of ( l/k ) ∗ H Z /ℓ n | l [ τ | − l ]. This is clear by theexistence of transfers in DM ( k ) (see e.g., Corollary C.10). (cid:3) Lemma 3.9.
Let ζ ∈ k be a primitive ℓ n -th root of unity and τ ∈ π , − H Z /ℓ n . Suppose that β ( τ ) = [ ζ ] ,where β : H , ( k, Z /ℓ n ) → H , ( k, Z ) is the integral Bockstein and [ ζ ] ∈ H , ( k, Z ) ≃ k × is the element corresponding to ζ . Then τ is acohomological Bott element.Proof. Under the equivalence Z (1) ≃ G m [SV00, Lemma 3.2(1)] ([MVW06, Theorem 4.1]), the Bocksteincorresponds to the inclusion π ( Z /ℓ n (1)) ≃ µ ℓ n ֒ → G m . It follows that L ´et ( Z /ℓ n (1)) ≃ µ ℓ n and that τ defines an equivalence Z /ℓ n ≃ L ´et ( Z /ℓ n (1)); in particular τ is an ´etale local equivalence. It remains toshow that for X ∈ Sm k the induced map H ∗∗ ( X, Z /ℓ n )[ τ − ] = colim r H ∗ , ∗ + r ( X, Z /ℓ n ) → H ∗∗ ´et ( X, Z /ℓ n )is an isomorphism. This is immediate from the solution of the Beilinson-Lichtenbaum conjecture [Voe11,Theorem 6.17]. (cid:3) Spherical Bott elements.Definition 3.10.
Let 1 ≤ m, n ≤ ∞ , S a scheme. By a spherical Bott element we mean an element τ ∈ π , − r ( / ( ℓ n , ρ m )) such that for every map x → S where x is the spectrum of a field, the inducedelement H Z ∧ τ | x ∈ π , − r ( H Z x / ( ℓ n , ρ m )) is a cohomological Bott element (see Definition 3.4). Remark 3.11.
Spherical Bott elements are stable under base change, essentially by definition.
Lemma 3.12.
Choose a multiplication on /ℓ n ∈ SH and one on ∧ ℓ /ρ m ∈ SH ( S ) . Let n ′ ≤ n andchoose a /ℓ n -module structure m on /ℓ n ′ . Let τ ∈ π , − r / ( ℓ n , ρ m ) be a spherical Bott element.Then the composite ˜ τ : / ( ℓ n ′ , ρ m ) τ ∧ id −−−→ / ( ℓ n , ρ m ) ∧ / ( ℓ n ′ , ρ m ) m −→ / ( ℓ n ′ , ρ m ) is a τ -self map. If the multiplication is chosen to be homotopy commutative, and the module structurehomotopy associative, then the τ -self map is good.Proof. We need to check that forming the mapping telescope of ˜ τ ∧ H Z | k is an ´etale localization, forevery field k . This is true basically by definition, as soon as we know that the multiplication on H Z /ℓ n induced by the one on /ℓ n is the standard one. This is indeed the case, by Corollary A.8.Goodness of the τ -self map follows from Corollary B.3. (cid:3) Construction of spherical Bott elements
Construction via roots of unity.Lemma 4.1.
Let ℓ n = 2 and suppose S contains a primitive ℓ n -th root of unity ζ . Then there exists aspherical Bott element mod ( ℓ n , ρ ∞ ) .Proof. Note that either ℓ is odd or − S . In both cases, using the notation n ǫ from § ℓ nǫ = ℓ n : ∧ ℓ,ρ → ∧ ℓ,ρ . By construction we have a cofibration sequence ∧ ℓ,ρ /ℓ n (1) β −→ G m ∧ ℓ,ρ ℓ n −→ G m ∧ ℓ,ρ , whence by Lemma 3.9 a spherical Bott element exists as soon as ℓ n [ ζ ] ∧ ρ,ℓ = 0. This is immediate fromProposition 2.1. (cid:3) Lifting to higher ℓ -powers. Given an object E in a triangulated symmetric monoidal category anda map m : E ⊗ E → E , by a derivation δ : E → E [1] we mean a map such that δ ◦ m = m ◦ (id ⊗ δ + δ ⊗ id). Lemma 4.2.
Let C be a symmetric monoidal triangulated category. Suppose E ∈ C is a non-commutative,non-associative and non-unital algebra, and let δ : E → E [1] be a derivation. Then for X ∈ C and t : X → E we have δ ( t n ) = n X i =1 t . . . tδ ( t ) t . . . t : X ⊗ n → E. Here the sum is over all possible ways of replacing one instance of t in the string t n by δ ( t ) , and bothsides are associated from the left (i.e., t n = t ( t ( . . . ( tt ) . . . )) and so on).Proof. We write the map δ ( t n +1 ) as X ⊗ n +1 ≃ X ⊗ X ⊗ n t ⊗ t ⊗ n −−−−→ E ⊗ E id ⊗ t n −−−−→ E ⊗ E m −→ E δ −→ E [1] . The definition of a derivation implies that this is the sum of the two maps tδ ( t n ) and δ ( t ) t n . The resultthus follows by induction, starting at n = 2 where we use the definition of a derivation. (cid:3) Recall the notion of a regular multiplication from § A.1.
Proposition 4.3.
Let q be a prime power, fix a regular multiplication on /q ∈ SH and assume thereexists a spherical Bott element mod ( q, ρ ∞ ) . Then for every n ≥ there exists a spherical Bott elementmod ( q n , ρ ∞ ) . TABLE MOTIVIC INVARIANTS ARE EVENTUALLY ´ETALE LOCAL 11
Proof.
Throughout the proof, all spectra are implicitly ( ρ, q )-completed.As a first step, choose regular multiplications on /q n for all n ≥ . . . /q → /q → /q consists of morphisms of homotopy unital ring spectra. This is possible by [Oka84, Lemma 5]. We shallproceed by induction on n . Thus let τ be a spherical Bott element modulo q n . We claim that τ q lifts toan element modulo q n +1 ; this will be a spherical Bott element by Lemma 3.7(1). To see this, we beginwith the cofiber sequence in [RØ08a, (3.3)] /q → /q n +1 → /q n ¯ δ −→ /q [1] . We need to show that ¯ δ ( τ q ) = 0. Let r : /q n → /q be the reduction map; then ¯ δ = rδ , where δ : /q n → /q n [1] is the coboundary. By our choice r is multiplicative, and so by Lemma 4.2 we have¯ δ ( τ q ) = r (cid:2) δ ( τ ) τ q − + τ δ ( τ ) τ q − + · · · + τ q − δ ( τ ) (cid:3) , where the sum consists of q terms.Note that q = 0 on /q , since /q has a multiplication. Suppose first that the multiplication on /q n is associative and commutative. Then each of the q terms in our sum is the same , so the sum is zero,and we are done.To finish the proof, we observe that we do not actually need the multiplication to be homotopyassociative or commutative. Since δ = u∂ (where u : → /q n is the unit map) we get ∂δ = 0. Thisimplies that any element commutes with δ ( τ ), and any two elements associate with δ ( τ ) (see SectionA.2). Thus δ ( τ ) τ m +1 (1) = δ ( τ )( τ τ m ) (2) = ( δ ( τ ) τ ) τ m (3) = ( τ δ ( τ )) τ m (2) = τ ( δ ( τ ) τ m ) , where (1) is by definition (everything being associated from the left), (2) is because everything associateswith δ ( τ ) and (3) is because everything commutes with δ ( τ ). This implies by induction that δ ( τ ) τ m isindependent of the order of multiplication, for any m , and so all the terms are the same, as before. (cid:3) Construction by descent.
Let f : S ′ → S be a morphism of schemes and σ : S ′ → S ′ be anautomorphism over S . Then f σ = f = f σ − and so σ ∗ f ∗ α ≃ f ∗ , f ∗ σ − ∗ β ≃ f ∗ . This provides us with asequence of equivalences f ∗ f ∗ f ∗ α ≃ f ∗ σ ∗ f ∗ ≃ f ∗ σ − ∗ f ∗ βf ∗ ≃ f ∗ f ∗ , where we have used that an adjoint of an equivalence is canonically equivalent to the inverse. Hence forevery object E ∈ SH ( S ) we get an automorphism σ E : f ∗ f ∗ E → f ∗ f ∗ E. This construction is natural in E .The following result is proved in Appendix C; see Corollary C.10 and the preceding paragraphs. Proposition 4.4.
Let f : S ′ → S be a finite Galois cover with group G .(1) The above incoherent construction refines to a coherent action, i.e., a functor BG → SH ( S ); ∗ 7→ f ∗ f ∗ E. (2) The unit of adjunction E → f ∗ f ∗ E refines to a G -equivariant map (for the trivial action of G on E ).(3) Suppose that E ∈ SH ( S )[1 / , / | G | ] + . Then E → f ∗ f ∗ E exhibits E as the homotopy fixed pointsof the G -action on f ∗ f ∗ E , and this limit diagram is preserved by any additive functor. Remark 4.5.
In the situation of Proposition 4.4(3), the map E → f ∗ f ∗ E is a split injection, i.e., E isa summand of f ∗ f ∗ E . See Corollary C.6 for details. Corollary 4.6.
Let q = ℓ n be an odd prime power, q = 3 . Assume /ℓ ∈ S and let S ′ be obtained from S by adjoining a primitive ℓ -th root of unity, i.e., S ′ = S × Spec ( Z ) Spec ( Z [ ζ ℓ ]) . If S ′ affords a sphericalBott element mod ( q, ρ ∞ ) then so does S . One might be concerned here about an absence of signs. We are given various maps ∧ q → /q [1], differing bypermutations of the source only . But the switch map on ∧ is the identity, whence there are no signs. Proof.
We again complete everything implicitly at ( ρ, ℓ ).If S = S ′ there is nothing to prove, so assume S ′ = S . Thus f : S ′ → S is a Galois cover with Galoisgroup G = Z / ( ℓ −
1) [Neu13, Corollary 10.4]. Consequently /q ∈ SH ( S )[1 / , / | G | ] + and so by Proposition 4.4 we get [ , /q ( m )] S = [ , /q ( m )] GS ′ . Let τ ∈ [ , /q ( m )] S ′ be a spherical Bottelement and put τ ′ = Y g ∈ G ( gτ ) ∈ [ , /q ( | G | m )] S ′ . Then for h ∈ G we have hτ ′ = Y g ( hgτ ) = τ ′ , since the multiplication in /q is commutative and associative (here we use that q = 3). In other words τ ′ is fixed by G and so defines an element τ S ∈ [ , /q ( | G | m )] S . By construction, f ∗ τ S = Q g ( g ∗ τ ). Thisis a product of spherical Bott elements and hence a spherical Bott element (see Remark 3.6). By Lemma3.8 it follows that τ S is also a spherical Bott element. (cid:3) Construction over special fields.
Quadratically closed fields.
Proposition 4.7.
Let ℓ be a prime, S = Spec ( k ) . Assume that k affords a primitive ℓ -th root of unity,and k × /ℓ ≃ { } (i.e., every element of k admits an ℓ -th root). Then there exists a spherical Bott element τ ∈ π , − ( ∧ ℓ,ρ ) . Proof.
Lemma 4.1 and its proof show that there exists τ ∈ π , − ( ∧ ρ /q ), where q = 4 if ℓ = 2 and q = ℓ else. We shall show by induction that for each n :( ∗ ) there exists a lifting τ n +1 ∈ π , − ( ∧ ρ /q n +1 ) of τ n . By the Milnor exact sequence [GJ09, Proposition VI.2.15] there is a surjection π , − ( ∧ ℓ,ρ ) → lim n π , − ( ∧ ρ /q n );hence there is a (non-canonical) lift τ ∈ π , − ( ∧ ℓ,ρ ). This will be a spherical Bott element by Lemma3.7(1).It hence remains to prove ( ∗ ). The cofiber sequence (1) ∧ ρ /q → (1) ∧ ρ /q n +1 → (1) ∧ ρ /q n → [1](1) ∧ ρ /q ≃ G m ∧ ρ /q implies that it is enough to prove the vanishing π ( G m ∧ ρ /q ) = 0.Suppose first that ℓ = 2. Then ρ is nilpotent by Remark 2.5(1), so G m ∧ ρ ≃ G m and π ( G m ∧ ρ /q ) ≃ K MW ( k ) / K MW ( k ) ≃ I ( k ) × I ( k ) /I ( k ) K M ( k ) . Since k is quadratically closed, I ( k ) = 0 [EKM08, Lemma 31.1] and hence K MW ( k ) ≃ K M ( k ). Thus K MW ( k ) / ≃ K M ( k ) / ≃ k × / ≃ { } , and we are done (here the last isomorphism again follows from the fact that k is quadratically closed).Now suppose that ℓ is odd. Then G m ∧ ρ /q ≃ G m [1 / + /ℓ and so π ( G m ∧ ρ /q ) ≃ K MW ( k )[1 / + /ℓ ≃ K M ( k ) /ℓ ≃ k × /ℓ ≃ { } . Here we have used the fact that for any field k , the group I/I ( k )[1 /
2] = 0 since it is a module over
W/I ( k ) ≃ Z / K MW is just a product.This concludes the proof. (cid:3) The real numbers. If S = Spec ( R ), then there exist τ n ∈ π , − r ( ∧ /ρ n ) lifting a power of τ ∈ π , − H Z / ∧ /ρ n is an E ∞ -ring [BS19, Lemma 7.8].Consequently this defines a spherical Bott element modulo (2 ∞ , ρ n ) by Lemma 3.9(2). TABLE MOTIVIC INVARIANTS ARE EVENTUALLY ´ETALE LOCAL 13
Summary of τ -self maps. Using Lemma 3.12 and the multiplications and module structures onMoore spectra [Oka84], as reviewed (and slightly extended in Corollary A.5) in Appendix A, we find thatthere is a good τ -self map modulo ( ℓ n , ρ m ) as soon as there is a spherical Bott element modulo ( ℓ n ′ , ρ m )for some n ′ ≥ n . In particular we have said elements in the following cases:(1) m = ∞ , ℓ odd, n < ∞ (use Corollary 4.6, Proposition 4.3, and Lemma 4.1).(2) m = ∞ , ℓ = 2, n < ∞ , √− ∈ S (using the same results).(3) m = ∞ , n = ∞ , ℓ arbitrary, S defined over a field k containing a primitive ℓ -th root of unityand satisfying k × /ℓ ≃ { } (use Proposition 4.7 and Remark 3.11).(4) m < ∞ , ℓ = 2, n ≤ ∞ , S defined over R (use § Slice convergence
In this section we provide a mild extension of Levine’s results on the convergence of the slice spectralsequence or, more precisely, convergence of the slice tower for motivic spectra satisfying certain ℓ -torsionand ρ -torsion hypotheses [Lev13]; see Theorem 5.3 for a precise statement. For this we use a very deepresult: Voevodsky’s resolution of the Milnor and Bloch-Kato conjectures [OVV07, Voe11]. We first setout our conventions on towers. Definition 5.1.
Let C be a category and E ∈ C .(1) By a tower over E we mean an object E • ∈ Fun( Z op , C /E ). We typically display towers as . . . E → E → E → E − → · · · → E or E • → E. (2) Suppose C is an abelian 1-category. Given a tower E • → E in C we define the descendingfiltration F i E = im( E i → E ) ⊂ E. We call the tower E • → E separated if0 = ∩ i F i E and convergent if in addition it is exhaustive , i.e., E = ∪ i F i E. Clearly if F : C → D is any functor and E • → E is a tower in C , then F E • → F E is a tower in D .We shall utilize this definition of convergence to detect when maps are null. Lemma 5.2.
Let E • → E be a tower in the category SH of spectra. Denote by E i /E i +1 the cofiber ofthe canonical morphism E i +1 → E i . Let k ∈ Z and assume that (a) the tower π k ( E • → E ) is convergent,and (b) π k ( E i /E i +1 ) = 0 for every i . Then π k ( E ) = 0 .Proof. It suffices to prove the result for k = 0. Let f : → E be any map. We need to show that f = 0. We shall show that f ∈ F n π E for all n ; then we are done by separatedness. By definitionof exhaustiveness, we have f ∈ F N π E for some N ; hence it suffices to show that f ∈ F n π E implies f ∈ F n +1 π E . Hence suppose f ∈ F n π E , and pick f n : → E n such that the composite → E n → E is homotopic to f ; this is possible by definition of F n π E . Since π ( E n /E n +1 ) = 0 by assumption, thecomposite → E n → E n /E n +1 is homotopic to zero, and hence f n lifts to a map f n +1 : → E n +1 . Itfollows that f ∈ F n +1 π E . (cid:3) In order to apply this result, we need a good supply of convergent towers. We shall produce them fromVoevodsky’s slice tower, using a strengthening of Levine’s convergence theorem for the slice filtration[Lev13, Theorem 7.3] that we will establish next.Recall that there is a functorial tower SH ( S ) → SH ( S ) Z ∪{∞} ; E ( f • E → E )called the slice tower [Voe02]; see also [RSØ19, Section 3] for a more extensive discussion and referencestherein. Recall the definition of virtual cohomological dimension from § E ∈ SH ( k ):(a) E ∈ SH ( k ) ≥ c for some c ∈ Z .(b) For i, j ∈ Z , K/k any finitely generated, separable field extension, and a ∈ π i ( E ) j ( K ) we have t r a = 0 for r sufficiently large.(c) There exists an integer R ≥ ρ R : E → E ∧ G ∧ Rm is homotopic to zero. Theorem 5.3.
Let k be a field of exponential characteristic e and t > coprime to e such that vcd t ( k ) < ∞ .There exists a function (5.4) Z → N , ( c, d, R, i, j, M ) N ( c, d, R, i, j, M ) such that for every x ∈ X ∈ Sm k with dim X ≤ d , ( i, j, M ) ∈ Z and E ∈ SH ( k ) satisfying (a) and (b),the following hold:(1) For n > N ( c, d, R, i, j, we have π i,j ( f n ( E ) /ρ R ) x = 0 . In particular, the tower (5.5) π i,j ( f • ( E ) /ρ R ) x → π i,j ( E/ρ R ) x is separated.(2) In addition, if E also satisfies (c), then the morphism π i,j ( f M + N ( c,d,R,i,j,M ) E ) x → π i,j ( f M E ) x is zero. In particular, the tower (5.6) π i,j ( f • E ) x → π i,j ( E ) x is separated. A few remarks are in order.
Remark 5.7.
This result is closely related to [Lev13, Conjecture 5]. Indeed the main idea in ourargument is that for fields with vcd ( k ) < ∞ , ρ is the only obstruction to nilpotence of the ideal I = I ( k ) ⊂ GW ( k ). In particular, under assumption (c), I acts nilpotently on each π i,j ( E ) x , and thus inparticular this module is I -adically complete. Levine’s conjecture thus predicts our separatedness result.Conversely, over fields of finite vcd , the I -adic and ρ -adic filtrations are commensurate on K MW ∗ ; itthus seems justified to think (over such fields) of derived ρ -completion as a form of I -adic completion.We thus view our results as establishing a derived version of Levine’s conjecture. Remark 5.8.
The function (5.4) indicates the dependence of the number N ( c, d, R, i, j, M ) on theconnectivity of E (given by c ), the bidegrees we are interested in (given by ( i, j )), the effective cover of E we are taking (given by M ) and the dimension of the scheme (or, rather, the point; this is given by d ). Note that this function does not depend on the number r that appears in condition (b). Remark 5.9.
The slice tower is always exhaustive (see, for example, [RSØ18, Lemma 3.1]); hence thetheorem implies that the towers (5.5), (5.6) are convergent.
Remark 5.10. If d is coprime to t , then (b) implies multiplication by d is an isomorphism on π i,j ( E )for all i, j , and hence so is d : E → E . In particular, under the hypotheses of Theorem 5.3, we have that E ∈ SH ( k )[1 /e ]. Remark 5.11.
It follows from Remark 2.5 that condition (c) is vacuous if k is unorderable (e.g., cd ( k ) < ∞ ), or if t is odd and E ∈ SH ( k )[1 / + . In these cases the statement of Theorem 5.3 is not quite optimal;in fact the proof shows that π i,j ( f M E ) x = 0for M ≫ i, j, c, d ). Remark 5.12. If a ∈ π p,q and F ∈ SH ( k ), then standard arguments show that a : F/a → Σ p, q F/a is the zero map (see e.g., [R¨on19, Lemma 5.2]). It follows that for E ∈ SH ( k ) ≥ c Theorem 5.3 applies to E/ ( ρ a , t b ) (and also E/t b or ( E/t b ) + , if Remark 5.11 applies).Our result implies convergence of the slice spectral sequence in novel cases. We record the following,even though we do not use it in the rest of the article. Corollary 5.13.
Let k be a field of exponential characteristic e and t > coprime to e such that vcd t ( k ) < ∞ . Suppose E ∈ SH ( k ) ≥ c for some c ∈ Z .(1) The map E ∧ t,ρ sc −→ sc( E ) ∧ t,ρ induces an isomorphism on π ∗∗ ; here sc denotes the slice completionfunctor [RSØ19, § . Not necessarily perfect, contrary to Levine’s assumption in [Lev13]. Here for a sheaf F on Sm k we denote by F x its stalk at x ∈ X . TABLE MOTIVIC INVARIANTS ARE EVENTUALLY ´ETALE LOCAL 15 (2) There is a conditionally convergent spectral sequence π p,n ( s q ( E ) ∧ t,ρ ) ⇒ π p,n ( E ∧ t,ρ ) , using the indexing conventions of [RSØ19] .Proof. (1) It is enough to show that E/ ( ρ n , t m ) → sc( E ) / ( ρ n , t m ) is a π ∗∗ -isomorphism. This is immedi-ate from Theorem 5.3(1) and Remark 5.9 (and Remark 5.12, which tells us that Theorem 5.3 applies).(2) We consider the completed slice tower · · · → f n ( E ) ∧ t,ρ → f n − ( E ) ∧ t,ρ → · · · → E ∧ t,ρ . The cones of the maps in this tower are given by the ( t, ρ )-completed slices s n ( E ) ∧ t,ρ . To prove conditionalconvergence for the corresponding spectral sequence displayed in (2), we need to show that π ∗∗ (lim n ( f n ( E ) ∧ t,ρ )) ≃ π ∗∗ (colim n ( f n ( E ) ∧ t,ρ )) ≃ π ∗∗ ( E ∧ t,ρ ) . Since limits commute we have the fiber sequencelim n ( f n ( E ) ∧ t,ρ ) ≃ (lim n f n ( E )) ∧ t,ρ → E ∧ t,ρ sc −→ sc( E ) ∧ t,ρ . Hence the claim about limits reduces to (1). For the claim about colimits, observe that for X ∈ Sm k and n < w we have map( X ( w ) , f n ( E ) ∧ t,ρ ) ≃ lim m cof(map( X ( w ) , Σ − m, − m f n ( E ) /t m ) ρ m −−→ map( X ( w ) , f n ( E ) /t m )) ≃ lim m cof(map(Σ m,m X ( w ) , f n ( E/t m )) ρ m −−→ map( X ( w ) , f n ( E/t m ))) ≃ lim m cof(map(Σ m,m X ( w ) , E/t m ) ρ m −−→ map( X ( w ) , E/t m )) ≃ map( X ( w ) , E ∧ t,ρ ) , where X ( w ) := Σ ,w Σ ∞ + X . Since the X ( w ) are compact generators of SH ( k ), we deduce that in factcolim n ( f n ( E ) ∧ t,ρ ) ≃ E ∧ t,ρ . (cid:3) Remark 5.14.
Under the assumptions of the corollary, it is in fact the case that E ∧ t,ρ → sc( E ) ∧ t,ρ is anequivalence. This requires a slightly more elaborate argument and will be treated elsewhere.In the rest of this section we prove Theorem 5.3, adapting the argument of Levine [Lev13, Theorem7.3]. Without loss of generality, we make the following standing assumptions:(1) c = 0,(2) k is perfect (using [EK18, Corollary 2.1.7] and Remark 5.10),(3) k is infinite (using standard transfer arguments; see [Lev13, Appendix A] for details),(4) x is a generic point, so X x is the spectrum of a field of transcendence degree ≤ d over k (usingunramifiedness of homotopy sheaves [Mor05, Lemma 6.4.4]).With these assumptions at play, we quickly review Levine’s approach to studying the slice filtrationvia the simplicial filtration as in [Lev13, Section 4]. For E ∈ SH ( k ) and for any M ≥ , X ∈ Sm k ,consider the mapping spectrum f M E ( X ) = map(Σ ∞ X + , f M E ) ∈ SH ;the functor X f M E ( X ) is an A -invariant Nisnevich sheaf of spectra. As elaborated in [Lev13, Section3], we have an augmented simplicial spectrum of the form E ( M ) ( X, • ) → f M E ( X ) , which is a colimit diagram. Under the Dold-Kan correspondence [Lur16, Theorem 1.2.4.1] we get theassociated filtered spectrum(5.15) sk E ( M ) ( X, • ) → sk E ( M ) ( X, • ) → · · · → sk k E ( M ) ( X, • ) → · · · → f M E ( X ) . Noting that π i ( E ) − j ( K ) = [Σ i G ∧ jm ∧ K + , E ](where K/k is a field extension) we obtain for j ≥ E p,q ( K, E, M, j ) ⇒ π p + q ( f M E ) − j ( K ) . As usual, defining F simp k π i ( f M E ) − j ( K ) := Im( π i sk k ( E ( M ) ( G ∧ jm ∧ K + , • )) → π i ( f M E ) − j ( K )) , we have the exhaustive increasing filtration(5.17) F simp0 π i ( f M E ) − j ( K ) → · · · F simp k π i ( f M E ) − j ( K ) → F simp k +1 π i ( f M E ) − j ( K ) · · · → π i ( f M E ) − j ( K ) , and the associated graded identifies with the E ∞ page [Lev13, sentence before Lemma 4.4]: gr simp p π i ( f M E ) − j ( K ) ∼ = E ∞ p,i − p ( K, E, M, j ) . The next lemma concerns the spectral sequence (5.16) and the corresponding filtration on sectionsof homotopy sheaves. To state the result we need some notation. Suppose that
F, G, H are strictly A -invariant sheaves on Sm k , H has transfers in the sense of [Mor, Chapter 4], and there is a pairing F ⊗ G → H . Then for a finitely generated separable extension K ′ of k , recall that we have the subgroup[ F G ] tr ( K ) := h tr K ′ /K F ( K ′ ) G ( K ′ ) i K ′ /K finite ⊂ H ( K ) . We refer to [Bac18b, Section 4] or [Lev11, Section 7] for details on this construction. The case that weare interested in will be F = K MWn ∼ = π ( ) n acting on G = π i ( E ) j . Lemma 5.18 (Levine) . Suppose that E ∈ SH ( k ) ≥ , i.e., (a) holds. Then for all M ≥ , j ≥ and all i ∈ Z we have:(1) The spectral sequence (5.16) is strongly convergent.(2) There is an inclusion of abelian groups E p,q ( K, E, M, j ) ⊂ M w ∈ (∆ pK ,∂ ∆ pK ) ( M ) π q + M ( E )( K ( w )) − M − j (3) F simp ∗ π i ( f M E ) − j ( K ) = 0 if ∗ < M, and F simp ∗ π i ( f M E ) − j ( K ) = π i ( f M E ) − j ( K ) if ∗ > M + i. (4) Under the canonical map π i ( f M E ) − j → π i ( E ) − j , the image of F simp M π i ( f M E ) − j ( K ) in π i ( E ) − j ( K ) is the subgroup [ K MWM π i ( E ) − M − j ] tr ( K ) . Proof.
The first two points are covered in [Lev13, Lemma 4.4], the third is immediate from (2) (see also[Lev13, Lemma 5.1]) and the last point is [Lev13, Theorem 5.3]. (cid:3)
Lemma 5.19.
Let E ∈ SH ( k ) and M ∈ Z . In the notation of Theorem 5.3:(a) If E satisfies (a) then so does f M E .(b) If E satisfies (b) then so does f M E .Proof. (a) We may assume (replacing E by E ∧ G ∧− Mm ) that M = 0. The claim is then immediate from[Bac17, Proposition 4(3)].(b) We may assume (replacing E by E ∧ Σ − i G ∧ jm ) that i = j = 0. If M < π ( f M E ) ≃ π ( E ) . Thus we may assume that M ≥
0. Consider the filtration (5.17) of π ( f M E ) ( K ). Since this filtration is finite and exhaustive, it suffices to prove the claim on the level ofassociated graded groups, which are the E ∞ -terms of the spectral sequence (5.16). The claim followssince the E ∞ -terms are subquotients of the E -terms, which by Lemma 5.18 are in turn subgroups ofsums of groups of the form π a ( E )( K ( w )) b , which satisfy (b) by assumption. (cid:3) Remark 5.20.
It is not clear that if E satisfies condition (c), then so does f M E . Instead the composite f M E ρ R −−→ f M ( E ) ∧ G ∧ Rm → f M − R ( E ) ∧ G ∧ Rm is zero. We are not going to use this observation (explicitly). Lemma 5.21.
Let E satisfy conditions (b) and (c) of Theorem 5.3. There exists a function M ( d, R ) (independent of E ) such that for n ≥ M ( d, R ) , trdeg( K/k ) ≤ d , and i, j ∈ Z we have [ K MWn π i ( E ) j ] tr ( K ) = 0 . TABLE MOTIVIC INVARIANTS ARE EVENTUALLY ´ETALE LOCAL 17
Proof.
We define the function by(5.22) M := M ( d, R ) := vcd t ( k ) + d + R + 1 . Let
K/k be of transcendence degree ≤ d and K ′ /K finite. Assumption (b) implies that every ele-ment of π i ( E ) j ( K ′ ) is a sum of elements which are ℓ m -torsion, for various ℓ | t and m ≥
0. Put F ∗ = K MW ∗ ( K ′ ) / ( ρ R , ℓ m ). Together with assumption (c), the above observation implies that it suf-fices to establish the vanishing F M = 0.If ℓ is odd then F ∗ is a quotient of K MW ∗ ( K ′ )[1 /
2] and so splits into + and − parts; moreover ρ is anisomorphism on the − part (see Remark 2.5(2)) and thus F −∗ is zero. Hence, it is enough to show that K MM ( K ′ ) /ℓ m = 0. This is true by choice of M , the fact that cd ℓ ( K ′ ) ≤ d + cd ℓ ( k ) = d + vcd ℓ ( k ) [Sta17,Tag 0F0T], and the comparison between Milnor K -theory and ´etale cohomology [Voe11].Now let ℓ = 2. Note that if M is an A -module and a ∈ A satisfies a N M = 0 for some N , then M = 0if and only if M/a = 0. Hence we may assume that m = 1. Note that h is nilpotent in F ∗ :0 = ( ηρ ) R = ( h− i − R = ( h − R = h R . It thus suffices to show that ( F M ) /h = 0. But K MW ∗ /h = I ∗ [Mor04b, Th´eor´eme 2.4.], so it is enoughto show the equality ρ R I ( K ′ ) vcd ( K ′ ) = I ( K ′ ) vcd ( K ′ )+ R . Under the isomorphism of loc. cit. , the element ρ is sent to the element 2 ∈ I ( k ) ⊂ W ( k ). Hence weneed to check that I ( K ′ ) vcd ( K ′ )+ R = 2 R I ( K ′ ) vcd ( K ′ ) . For a proof see p. 619 in [EL99]. (cid:3) The following result is the key step in the proof of Theorem 5.3. It establishes a vanishing region inthe homotopy sheaves of sufficiently effective (and connective) spectra.
Lemma 5.23.
There exists a function M ( d, R, r, s ) such that for E ∈ SH ( k ) eff ( M ( d, R, r, s )) satisfying (a) , (b) and (c) of Theorem 5.3, and i ≤ r, ≤ j ≤ s we have π i ( E ) − j ( K ) = 0 , whenever trdeg( K/k ) ≤ d .Proof. We shall define M ( d, R, r, s ) by induction on r . If i < π i ( E ) − j ( K ) = 0 for any j, K byassumption, and hence M ( d, R, r, s ) = 0 works for all r < s ∈ Z .Now suppose that M ( d, R, r ′ , s ) has been defined for all r ′ < r . We first want to define M ( d, R, r, π r ( E ) ( K ). Consider the filtration (5.17) on π r ( f n E ) ( K ) with associatedgraded gr simp p π r ( f n E ) ( K ). It follows from Lemma 5.18(2) that these groups vanish when:(1) p < n by codimension reasons, or(2) p > n + r by connectivity of E .Conversely, whenever there is a non-zero contribution we must have ( ∗ ) n ≤ p ≤ n + r , and the contri-bution comes from some π r ′ ( E ) − n ( K ( w )), wheretrdeg( K ( w ) /K ) = p − n ≤ r and r ′ = n + r − p ≤ r. Here the last two inequalities just come from ( ∗ ).We define M ( d, R, r,
0) = max { M ( d + r, R, r − , M ( d, R )) , M ( d, R ) } , where M ( d, R ) is the function from (5.22). Set n = M ( d, R ). By assumption, E is at least n -effective,so f n E ≃ E and the map α n : π r ( f n E ) ( K ) → π r ( E ) ( K )is an isomorphism. We claim that for p > n , we have gr simp p π r ( f n E ) ( X ) = 0. Indeed by the abovediscussion any contribution to this group arises from π r ′ ( E ) − n ( L ), where r ′ < r and trdeg( L/k ) ≤ d + r .This vanishes by induction and the construction of M . Now, part (3) of Lemma 5.18 implies that theimage of α n is given by F simp n π i ( f n E ) ( K ), which is identified in part (4) as the group[ K MWn π r ( E ) − n ] tr ( K ) , which vanishes by Lemma 5.21. Hence our definition of M ( d, R, r,
0) has the desired property.Finally, since π i ( E ) − j = π i ( E ∧ G ∧− jm ) , we can define M ( d, R, r, s ) = M ( d, R, r, s . This concludesthe construction of the desired function. (cid:3) Recall our standing convention that c = 0, so this means E ∈ SH ( k ) ≥ . Proof of Theorem 5.3.
Let us call a tower of sheaves F • locally null if there exists a function N ( d ) suchthat for trdeg( K/k ) ≤ d and i ≥ N ( d ) we have F i ( K ) = 0.(1) Under our standing assumptions (in particular (4)), the claim is equivalent to showing that π i (( f • E ) /ρ R ) j is locally null (with N independent of E ). Replacing E by E ∧ G ∧ jm (which still satisfies(a), (b)) replaces π i ( f M ( E ) /ρ R ) by π i ( f M − j ( E ) /ρ R ) j . It thus suffices to treat the case j = 0. Note that f M ( E ) /ρ R is ( M − R )-effective, satisfies conditions (a) and (b) by Lemma 5.19, and satisfies condition(c) by Remark 5.12 (with 2 R in place of R ). Lemma 5.23 thus shows that π i ( f M ( E ) /ρ R ) ( K ) = 0 assoon as M ≥ M ( d, R, i,
0) + R . The claim follows.(2) Consider the following commutative diagram f M ( E ) −−−−→ f M ( E ) /ρ R a −−−−→ f M − R ( E ) y y y E −−−−→ E/ρ
R b −−−−→ E. Here the vertical maps are the canonical ones, and the horizontal maps in the left hand square are theprojections. The map b is a splitting E/ρ R ≃ E ∨ Σ E ∧ G ∧− Rm → E (using condition (c)), and the map a is the unique one making the right hand square commute (using that f M ( E ) /ρ R is ( M − R )-effective).The bottom horizontal composite is id E by construction, and hence the top horizontal composite is thecanonical map f M ( E ) → f M − R ( E ). Since π i ( f • ( E ) /ρ R ) j is locally null by (1), we deduce that for M sufficiently large depending only on i, j, d , the map π i ( f M + R E ) j ( K ) → π i ( f M E ) j ( K )factors through π i ( f M + R ( E ) /ρ R ) j ( K ) = 0 . This implies the claim.Separatedeness of both towers follows immediately. This concludes the proof. (cid:3) Spheres over fields
In this section we treat a special case of our main result to which we will reduce the general case.Throughout S = Spec ( k ), where k is a field of exponential characteristic e , ℓ = e is a prime and τ : / ( ℓ ν , ρ µ ) → / ( ℓ ν , ρ µ )( r )is a τ -self map (see Definition 3.1). Here 0 ≤ µ, ν ≤ ∞ .The following is the key technical result. Lemma 6.1.
Suppose that vcd ℓ ( k ) < ∞ . Then lim n (cid:2) f n ( ) / ( ℓ ν , ρ µ )[ τ − ] / ( ℓ, ρ ) (cid:3) = 0 . Proof.
Write L = lim n L n for the limit in question. For X ∈ Sm k , w ∈ Z , spectra of the form X ( w ) :=Σ ,w Σ ∞ X + generate SH ( k ) (see, for example, [Hoy17, Proposition 6.4] when G is the trivial group), andhence it suffices to show that [ X ( w )[ i ] , L ] = 0for all w, i . By the Milnor exact sequence [GJ09, Proposition VI.2.15]0 → lim n [ X ( w )[ i + 1] , L n ] → [ X ( w )[ i ] , L ] → lim n [ X ( w )[ i ] , L n ] → , it is enough to show that for ( i, w ) fixed and n sufficiently large we have [ X ( w )[ i ] , L n ] = 0.Consider the descent spectral sequence H p Nis ( X, π q,w ( L n )) ⇒ [ X ( w )[ q − p ] , L n ] , which is strongly convergent due to the finite cohomological dimension of the Nisnevich site (see e.g., theproof of [ELSØ17, Proposition 4.3]). This implies that it suffices to prove the following claim: • For any N ∈ Z and n sufficiently large (depending on N ), all w ∈ Z and k ≤ N we have π k,w ( L n ) | X Nis = 0.By unramifiedness of homotopy sheaves [Mor05, Lemma 6.4.4], it is enough to show that π k,w ( L n ) ξ = 0for generic points ξ of ´etale extensions of X . Since homotopy sheaves commute with colimits, and w ∈ Z is arbitrary, for this it suffices to show that π k,w ( f n ( ) / ( ℓ ν , ρ µ , ℓ, ρ )) ξ = 0 . TABLE MOTIVIC INVARIANTS ARE EVENTUALLY ´ETALE LOCAL 19
Up to adding a constant to n (which only depends on µ ), it is thus enough to show the vanishing π k,w ( f n ( ) / ( ℓ, ρ )) ξ = 0 . Using Theorem 5.3(1), we find that the tower of abelian groups π k,w ( f • ( ) / ( ℓ, ρ )) ξ → π k,w ( f n ( ) / ( ℓ, ρ )) ξ is convergent (here • ≥ n , and the tower is trivially exhaustive). Hence by Lemma 5.2, it suffices toprove the following claim: • For any N ∈ Z there exists n (depending on N and trdeg( ξ/k )) such that for all m ≥ n , k ≤ N and w ∈ Z we have π k,w ( s m ( ) / ( ℓ, ρ )) ξ = 0 . By [RSØ19, Theorem 2.12], each slice s m ( )[1 /e ] is a finite sum of suspensions of motivic cohomologyspectra Σ m + s,m H Z /l, for certain s ≥ , l ≥ n , it is thus enough to show that π k,w (Σ m + s,m H Z / ( ℓ, ρ )) ξ = 0for m sufficiently large. Now π k,w (Σ m + s,m H Z /ℓ ) ξ = H m + s − k ( ξ, Z /ℓ ( m − w ))and so we need to show that multiplication by ρ induces an isomorphism on these groups, for m sufficientlylarge (depending on N and trdeg( ξ/k )), all w ∈ Z and all k ≤ N , s ≥
0. In particular s − k ≥ − N .Note also that we have vcd ℓ ( ξ ) ≤ vcd ℓ ( k ) + trdeg( ξ/k ) by [Sta17, Tag 0F0T]. The required vanishingthus follows from Lemma 2.7: we may put m ≥ n := vcd ℓ ( k ) + trdeg( ξ/k ) + N + 1. This concludes theproof. (cid:3) Corollary 6.2. If vcd ℓ ( k ) < ∞ , then / ( ℓ ν , ρ µ )[ τ − ] / ( ℓ, ρ ) is ´etale local.Proof. We have a cofiber sequence of towers . . . −−−−→ f ( ) −−−−→ f ( ) −−−−→ f ( ) = y y y . . . −−−−→ id −−−−→ id −−−−→ y y y . . . −−−−→ f ( ) −−−−→ f ( ) −−−−→ f ( ) , where f i = id /f i . Smashing with / ( ℓ ν , ρ µ )[ τ − ] / ( ℓ, ρ ) and applying the (exact) inverse limit functor,we obtain a cofiber sequencelim n (cid:2) f n ( ) / ( ℓ ν , ρ µ )[ τ − ] / ( ℓ, ρ ) (cid:3) → / ( ℓ ν , ρ µ )[ τ − ] / ( ℓ, ρ ) α −→ lim n (cid:2) f n ( ) / ( ℓ ν , ρ µ )[ τ − ] / ( ℓ, ρ ) (cid:3) . By Lemma 6.1, the map α is an equivalence. Since ´etale local spectra are closed under limits andextensions and each f n ( ) is a finite extension of the slices s m ( ) (recall that is effective), it sufficesto show that s i ( ) / ( ℓ ν , ρ µ )[ τ − ] / ( ℓ, ρ ) is ´etale local for every i . By the form of the slices of recalled inthe proof of Lemma 6.1, it suffices to show that H Z / ( ℓ ν , ρ µ )[ τ − ] / ( ℓ, ρ )is ´etale local. This holds by the definition of a τ -self map. (cid:3) Main result
In this section we establish our main ´etale localization results. Before doing so, we need some prelim-inaries.
Proposition 7.1.
Let /ℓ ∈ k , vcd ℓ ( k ) < ∞ and S a finite type k -scheme. If X ∈ Sm S is quasi-separatedand w ∈ Z , then Σ ∞ + X/ ( ℓ, ρ ) ∧ G ∧ wm ∈ SH ´et ( S ) ∧ ℓ,ρ is compact. Proof.
Since the pullback SH ´et ( S ) ∧ ℓ,ρ → SH ´et ( X ) ∧ ℓ,ρ preserves colimits, we may assume that S = X .First suppose that cd ℓ ( k ) < ∞ (e.g., ℓ odd). It is enough to show that Σ ∞ + S/ℓ is compact in SH ´et ( S ) ∧ ℓ .This is proved in [Bac18c, Corollary 5.7 and Example 5.9].Now let ℓ = 2. Write SH ( S ´et ) for the category of hypersheaves of spectra on the small ´etale site of S (see e.g., [Bac18c, § E τ ≥ n E, τ ≤ n E, π n ( E ) : SH ( S ´et ) → SH ( S ´et )coming from the standard t -structure; they all preserve filtered colimits (for τ ≤ n this follows from [Lur18,Proposition 1.3.2.7(2)] and the other cases follow from this). We thus have restricted functors E τ ≥ n ( E/ℓ ) , τ ≤ n ( E/ℓ ) , π n ( E/ℓ ) : SH ( S ´et ) ∧ ℓ → SH ( S ´et ) ∧ ℓ with the same property. Moreover E/ℓ ≃ lim n τ ≤ n ( E/ℓ ) [Bac18c, Lemma 2.16]. Via [Bac18c, Theorem6.6] we transplant these functors to SH ´et ( k ) ∧ ℓ . In order to prove that S/ ( ℓ, ρ ) is compact, it is enoughto show that the functor E [ S/ρ, E/ℓ ] preserves filtered colimits. Using the Postnikov completenessresult just recalled, there is a conditionally convergent spectral sequence[
S/ρ, π i ( E/ℓ )[ j ]] ⇒ [ S/ρ, E/ℓ [ i + j ]] . By standard arguments it is enough to show that the spectral sequence converges strongly for every E ,and that [ S/ρ, π i ( E/ℓ )[ j ]] is compatible with filtered colimits in E . Consider F = π ( E/ℓ ) ∈ SH ( S ´et ) ♥ .This is a sheaf of Z /ℓ -modules on the small ´etale site of S . There is a canonical filtration 0 → K → F → F/ℓ → K, F/ℓ sheaves of Z /ℓ -modules and an associated long exact sequence · · · → [ S/ρ, K ] → [ S/ρ, F ] → [ S/ρ, F/ℓ ] → . . . . It will thus be enough to show that the functor F [ S/ρ, F [ i ]]is compatible with filtered colimits of sheaves of Z /ℓ -vector spaces on the small ´etale site of S andvanishes for i < i > cd ( S [ √− ( S [ √− < ∞ , e.g., by [Bac18c, Example5.9]).We claim that there is a long exact sequence · · · → H i ( S, F ) ( − −−−→ H i +1 ( S, F ) → [ S/ρ, F [ i ]] → H i +1 ( S, F ) → . . . . Since ´etale cohomology commutes with filtered colimits (see e.g., [Lur18, A.2.3.2(1)]), the result willfollow from Lemma 7.2 below. To prove the claim, consider the sequence of functors and adjoints SH ( S ´et ) ∧ ℓ ⇆ SH ´et ( S ) ∧ ℓ ⇆ DM ´et ( S, Z /ℓ ) . There is an object F ∈ DM ´et ( S, Z /ℓ ) with image F in SH ( S ´et ) ∧ ℓ ; we are thus reduced to proving theanalogous result in DM ´et ( S, Z /ℓ ). To conclude we note that ρ : G ∧− m → induces multiplication by( −
1) in DM ´et ( k, Z / ≃ D ( k ´et , Z / (cid:3) Our next result follows from [MVW06, Theorem 99.13] in the case of fields.
Lemma 7.2.
Let X be a scheme with / ∈ X and set X ′ = X [ √− . Suppose F is an ´etale sheaf of Z / -vector spaces on X . Then for i > cd ( X ′ ) the map ( −
1) : H i ´et ( X, F ) → H i +1´et ( X, F ) is an isomorphism.Proof. If X = X ′ or cd ( X ′ ) = ∞ , there is nothing to prove. We may thus assume that Aut ( X ′ /X )is the group C of order 2, and that cd ( X ′ ) < ∞ . Consider the strongly convergent Hochschild–Serrespectral sequence [Mil80, Theorem III.2.20] H p ( C , H q ( X ′ , F )) ⇒ H p + q ( X, F ) . It is a module over the same spectral sequence with F = Z /
2. The class ( − ∈ H ( X, Z /
2) is detected bya class α ∈ H ( C , H ( X, Z / H ∗ ( C , Z / ≃ H ∗ ( RP ∞ , Z / ≃ Z / α ].The class α has the property that if G is any Z / C ]-module and i >
0, the multiplication map α : H i ( C , G ) → H i +1 ( C , G ) is an isomorphism. We deduce that α induces an isomorphism on allcolumns of the spectral sequence, except possibly the one containing H ( C , H ∗ ( X ′ , F )). The resultfollows. (cid:3) Indeed A := Z / C ] ≃ Z / ǫ ] /ǫ , so any finitely generated A -module is a sum of copies of A and Z /
2. For suchmodules the claim holds. A general A -module is a filtered colimit of finitely generated ones, and cohomology commuteswith filtered colimits, so the claim follows in general. TABLE MOTIVIC INVARIANTS ARE EVENTUALLY ´ETALE LOCAL 21
We deduce a weak form of Theorem 1.5(3) from the introduction, which we will use in the proof ofour main result. (The strong form will be deduced later.)
Corollary 7.3.
Let k be a field of exponential characteristic e = ℓ and vcd ℓ ( k ) < ∞ . Then L ´et : SH ( k ) ∧ ℓ,ρ → SH ( k ) ∧ ℓ,ρ is a smashing localization, or in other words for E ∈ SH ( k ) ∧ ℓ,ρ we have L ´et ( E ) ≃ E ∧ L ´et ( ) .Proof. The functor L ´et : SH ( k ) ∧ ℓ,ρ → SH ´et ( k ) ∧ ℓ,ρ will identify the target as (highly structured) modulesover the ´etale-local sphere if we can show that both the source and target are compact-rigidly generated,see e.g., [Bac18b, Lemma 22] . This is indeed the case; see [Bac18c, Example 2.3] (to obtain a familyof generators), [LYZ13, Corollary B.2] (to see that the generators are rigid) and Proposition 7.1 (to seethat the generators are compact). The result follows. (cid:3) With these preparations out of the way, we come to our main theorem. We refer to § τ -self maps we managed to construct. Theorem 7.4.
Let S be a Noetherian scheme of finite dimension, ≤ m, n ≤ ∞ . Suppose τ : / ( ℓ n , ρ m ) → / ( ℓ n , ρ m )( r ) is a τ -self map . Assume that /ℓ ∈ S and for every s ∈ S we have vcd ℓ ( s ) < ∞ . Assume further that either S is the spectrum of a field or τ is good.Then for every E ∈ SH ( S ) ∧ ℓ,ρ the map E/ ( ℓ n , ρ m ) → E/ ( ℓ n , ρ m )[ τ − ] ∧ ℓ,ρ is an ´etale localization.Proof. By Lemma 3.3, the map in question is an ´etale equivalence. We thus need to prove that E ′ := E/ ( ℓ n , ρ m )[ τ − ] ∧ ℓ,ρ is ´etale local.Let X → X ∈ Sm S be an ´etale hypercover and K = Σ ∞ + cof ( X → X ). We need to show that[ K, E ′ [ i ]( j )] = 0 for all i, j . Since E was arbitrary, replacing it by E ( i )[ j ] we may as well assume that i = j = 0. By the definition of goodness, see Definition 3.1(3), any map K → E ′ factors through K → K ′′ := K/ ( ℓ n ′ , ρ m ′ )[ τ − ] ∧ ℓ,ρ . It is thus enough to show that K ′′ = 0. It follows from [Bac18a,Corollary 14] that the collection of functors { s ∗ : SH ( S ) ∧ ℓ,ρ → SH ( s ) ∧ ℓ,ρ | s ∈ S } is conservative; alsoeach of these functors preserves colimits. It is thus enough to show that s ∗ ( K ′′ ) = 0. By construction s ∗ ( K ) is ´etale-locally equivalent to zero; hence it suffices to show that s ∗ ( K ′′ ) ≃ s ∗ ( K ) / ( ℓ n ′ , ρ m ′ )[ τ − ] ∧ ℓ,ρ is ´etale local. We have thus reduced to the case of fields.Assume now that S = Spec ( k ). Then / ( ℓ n , ρ m )[ τ − ] / ( ℓ, ρ ) is ´etale local by Corollary 6.2, and henceso is / ( ℓ n , ρ m )[ τ − ] ∧ p,ρ , being a limit of extensions of the former term. Note that E ′ ≃ E ∧ ℓ,ρ ∧ / ( ℓ n , ρ m )[ τ − ] ∈ SH ( k ) ∧ ℓ,ρ , with the smash product being formed in SH ( k ) ∧ ℓ,ρ (i.e., completed). Since ´etale localization is smashingon SH ( k ) ∧ ℓ,ρ by Corollary 7.3, we deduce that E ′ is ´etale local in SH ( k ) ∧ ℓ,ρ , and hence also in SH ( k ). (cid:3) Remark 7.5.
The only place where the assumption that S is Noetherian of finite dimension enters inthe proof of Theorem 7.4 is to ensure that the collection of functors { s ∗ : SH ( S ) → SH ( s ) | s ∈ S } is conservative. Arguing as in [Ayo14, Proposition 3.24], this remains true if S is only assumed to belocally Noetherian (with no assumption on the dimension). The above theorem holds in this slightlylarger generality. Remark 7.6.
Suppose that ℓ is not invertible on S . Let j : U = S [1 /ℓ ] ֒ → S and Z = S \ U the closedcomplement. A (not entirely trivial) argument using the Artin–Schreier sequence can be used to showthat SH ´et ( Z ) ∧ ℓ = 0. This implies that for E ∈ SH ( S ) we have L ´et E/ℓ n ≃ j ∗ L ´et j ∗ E/ℓ n . Consequently the assumption that 1 /ℓ ∈ S in Theorem 7.4 is essentially harmless.We conclude with three immediate applications. Corollary 7.7.
Assumptions as in Theorem 7.4.(1) Let ℓ be odd, n < ∞ and m = ∞ . Then E/ℓ n → ( E/ℓ n ) + [ τ − ] is an ´etale localization. (2) Let ℓ = 2 , n < ∞ , m = ∞ and assume that − a square in S . Then E/ n → E/ n [ τ − ] is an ´etale localization.(3) Suppose that S is defined over a field containing a primitive ℓ -th root of unity and satisfying k × /ℓ = { } . Then we have SH ´et ( S ) ∧ ℓ ≃ SH ( S ) ∧ ℓ [ τ − ] . Proof.
In each case the ρ -completion is unnecessary (or simplified to ( − ) + in case (1)), by Remark 2.5.In (1), (2) also the ℓ -completion is unnecessary, by Remark 3.2. (cid:3) Denote by L ∧ ´et : SH ( S ) → SH ´et ( S ) ∧ ℓ,ρ the left adjoint to the inclusion ι : SH ´et ( S ) ∧ ℓ,ρ ⊂ SH ( S ). Corollary 7.8.
Assumptions as in Theorem 7.4. Suppose that m, n < ∞ .Then the functor ι ◦ L ∧ ´et : SH ( S ) → SH ( S ) is equivalent to Bousfield localization at the homologytheory / ( ℓ n , ρ m )[ τ − ] .Proof. Let E ∈ SH ( S ) and α : E → ιL ∧ ´et E be the ( ℓ, ρ )-complete ´etale localization map. We first showthat α is a / ( ℓ n , ρ m )[ τ − ]-equivalence, or in other words that α ∧ / ( ℓ n , ρ m )[ τ − ] is an equivalence. ByTheorem 7.4, both sides are ´etale local, so it is enough to show that L ´et ( α ∧ / ( ℓ n , ρ m )[ τ − ]) ≃ L ´et ( α ) ∧ L ´et ( / ( ℓ n , ρ m )[ τ − ])is an equivalence. This is clear.It remains to show that SH ´et ( S ) ∧ ℓ,ρ consists of / ( ℓ n , ρ m )[ τ − ]-local objects. In other words, if E ∈ SH ( S ) with E ∧ / ( ℓ n , ρ m )[ τ − ] = 0, then we need to show that ιL ∧ ´et ( E ) = 0. Again by Theorem 7.4 wehave ιL ∧ ´et ( E ) / ( ℓ n , ρ m ) ≃ ιL ∧ ´et ( E/ ( ℓ n , ρ m )) ≃ E ∧ / ( ℓ n , ρ m )[ τ − ] = 0 . Since ιL ∧ ´et ( E ) is ( ℓ, ρ )-complete, we deduce that ιL ∧ ´et ( E ) = 0, as desired. (cid:3) Corollary 7.9.
Assumptions as in Theorem 7.4. Then ´etale localization is smashing on SH ( S ) ∧ ℓ,ρ .Proof. Denote by L ´et : SH ( S ) ∧ ℓ,ρ → SH ( S ) ∧ ℓ,ρ the ´etale localization functor. It suffices to show that,for E ∈ SH ( S ) ∧ ℓ,ρ , the spectrum E ′ := E ∧ L ´et ( ) ∈ SH ( S ) ∧ ℓ,ρ is ´etale local. By definition E ′ is ( ℓ, ρ )-complete, so E ′ ≃ lim n,m E ′ / ( ℓ n , ρ m ), and it suffices to show that E ′ / ( ℓ n , ρ m ) is ´etale local (for n, m sufficiently large). Theorem 7.4 implies that E ′ / ( ℓ n , ρ m ) ≃ E ∧ L ´et ( / ( ℓ n , ρ m )) ≃ E ∧ / ( ℓ n , ρ m )[ τ − n,m ] , where τ n,m denotes a τ -self map mod ( ℓ n , ρ m ) (it follows from our assumptions that this exists, at leastfor n > E ∧ / ( ℓ n , ρ m )[ τ − n,m ] is ´etale local. This concludes the proof. (cid:3) We also obtain the following new base change result, generalizing [ELSØ17, Theorem 7.5].
Theorem 7.10.
Assumptions as in Theorem 7.4. Suppose that f : T → S is a morphism of finite type.Then the functor f ∗ : SH ( S ) ∧ ℓ,ρ → SH ( T ) ∧ ℓ,ρ , preserves ´etale local spectra. In particular the canonical map f ∗ ( ´et ) S → ( ´et ) T is an equivalence in SH ( T ) ∧ ℓ,ρ .Proof. Follows from Theorem 7.4 and the fact that the Bott elements are stable under base change bydefinition. (cid:3)
Remark 7.11. If S is Noetherian, finite dimensional, and geometrically unibranch, then ´etale localiza-tion is smashing on SH ( S ) Q , since it just corresponds to “passing to the plus part” (see e.g., the proofof [Bac18c, Theorem 7.2]), or equivalently ρ -completion. In particular Corollary 7.9 holds rationally(instead of completed at ℓ ) as well. Using Remark 7.6, we see that it also holds completed at ℓ which failto be invertible on S (provided that vcd p ( s ) < ∞ for all s ∈ S ). One may deduce that ´etale localizationis smashing on SH ( S ) ∧ ρ , provided that vcd( s ) < ∞ for all s ∈ S (in addition to S being Noetherian,finite dimensional, and geometrically unibranch). Similarly f ∗ : SH ( S ) ∧ ρ → SH ( T ) ∧ ρ preserves ´etale local spectra (under these assumptions). TABLE MOTIVIC INVARIANTS ARE EVENTUALLY ´ETALE LOCAL 23
Appendix A. Multiplicative structures on Moore objects
A.1.
Definitions and setup.
Let C be a semiadditive [Lur16, Definition 6.1.6.13] symmetric monoidal ∞ -category. For n ∈ N we call /n = cof( n −→ )the mod n Moore object , which by construction comes with a map u : → /n which we call the unit .By a (homotopy) multiplication on /n we mean a map m : /n ⊗ /n → /n ∈ h C , which is compatible with the unit in the sense that m ◦ (id ⊗ u ) = id = m ◦ ( u ⊗ id) : /n → /n ∈ h C . Lemma A.1. If m is a multiplication on /n , then u : → /n is a morphism of homotopy ringobjects.Proof. Consider the diagram ⊗ id ⊗ u −−−−→ ⊗ /n u ⊗ id −−−−→ /n ⊗ /n m y y m y u −−−−→ /n /n. Here, m and the unlabelled map are canonical isomorphisms; at the same time m is the multiplicationin the canonical ring structure on . Since the top composite is u ⊗ u , we need to show that the outerrectangle commutes. The left hand square commutes by definition of a symmetric monoidal category,and the right hand square commutes by assumption. (cid:3) If C is stable, then we have the cofiber sequence n −→ u −→ /n ∂ −→ [1] . The composite δ = u ◦ ∂ : /n → /n [1]is called the coboundary . A multiplication m on /n is called regular if δ is a derivation [Oka84, Definition1], in the sense that: δ ◦ m ≃ m ◦ (id ⊗ δ + δ ⊗ id) . A.2.
Review of Oka’s results.
The case C = SH has been treated by Oka. Among other things thefollowing are shown in [Oka84, Theorem 2]. • /n has a multiplication if and only if n n is odd. • Whenever /n has a multiplication, it may be chosen to be regular. • A multiplication on /n is (homotopy) commutative if and only if n ≡ n is odd; itis (homotopy) associative if and only if n n
6≡ ± • For any multiplication on any /n , the commutator m ◦ (id − switch) : /n ∧ /n → /n factors through ∂ ∧ ∂ : /n ∧ /n → [1] ∧ [1] . Similarly, the associator m ◦ ( m ∧ id − id ∧ m ) : /n ∧ → /n factors through ∂ ∧ : /n ∧ → [1] ∧ . These statements are only interesting if the multiplication is not commutative/associative.
A.3.
Asymptotic unicity and associativity.
For a small idempotent complete ∞ -category C withfinite limits, let Pro( C ) = Ind( C op ) op denote the category of pro-objects in C [Lur18, § A.8.1, RemarkA.8.1.2]. If C has a symmetric monoidal structure, then so does Pro( C ) [Lur16, Remark 2.4.2.7, Propo-sition 4.8.1.10]. Lemma A.2.
Let C be a small stable idempotent complete symmetric monoidal ∞ -category C . Then ℓ -completion is a smashing localization on Pro( C ) . Remark A.3.
Even though Pro( C ) is not presentable, ℓ -completion (i.e., localization at the ℓ -equivalences)exists: it is clear that E lim n E/ℓ n is a ℓ -equivalence to a ℓ -complete object. Proof.
From the formula given in Remark A.3, this is the case in any symmetric monoidal stable ∞ -category where the tensor product commutes with cofiltered limits, such as Pro( C ). (cid:3) Let c : SH ω ֒ → Pro( SH ω )denote the inclusion “at constant cofiltered systems”. Proposition A.4.
Let µ i : /ℓ i ∧ /ℓ i → /ℓ i be a sequence of unital multiplication maps in SH (for i ≥ if ℓ = 2 ), such that the reductions r : /ℓ i +1 → /ℓ i are multiplicative (up to homotopy). Then thefollowing hold:(1) The pro-spectrum c ( ) ∧ ℓ is represented by the inverse system . . . r −→ /ℓ r −→ /ℓ r −→ /ℓ . (2) The multiplication on c ( ) ∧ ℓ is represented by the system of maps c ( ) ∧ ℓ ∧ c ( ) ∧ ℓ ≃ { ( /ℓ i ) ∧ } i µ i −→ { /ℓ i } i . Proof. (1) Clear by the formula for ℓ -completion in any stable category; see Remark A.3.(2) Since ℓ -completion is smashing in Pro( SH ω ) by Lemma A.2, the multiplication map is inverseto the unit map (on either side), and hence it is enough to show that the following composite map ofpro-systems is homotopic to the identity α : { /ℓ n } n ≃ { ∧ /ℓ n } n r ∧ id −−−→ { /ℓ n ∧ /ℓ n } n µ n −−→ { /ℓ n } n . Using the formula for mapping spaces in categories of pro-objects and the Milnor exact sequence, weobtain an exact sequence0 → lim m colim n [Σ /ℓ n , /ℓ m ] → [ { /ℓ n } n , { /ℓ m } m ] Pro( SH ω ) → lim m colim n [ /ℓ n , /ℓ m ] → . By assumption, α corresponds to the identity in the right hand group; it thus suffices to show that thelim -term vanishes. The (strong) dual of /ℓ n is Σ − /ℓ n , and multiplication by ℓ n is zero on /ℓ m for n sufficiently large. It follows thatcolim n [Σ /ℓ n , /ℓ m ] ≃ π ( /ℓ m ) ⊕ π ( /ℓ m ) . These groups are all finite (in fact independent of m for m ≫
0) and hence the inverse system is Mittag–Leffler, and lim = 0 as desired. (cid:3) Corollary A.5.
Let { µ i } be as in Proposition A.4. Then for n ≫ , the Moore object /ℓ can be giventhe structure of a homotopy associative /ℓ n -module.Proof. Since c ( ) ∧ ℓ is a commutative monoid, M = c ( ) ∧ ℓ /ℓ ≃ c ( ) /ℓ is an A ∞ -module under c ( ) ∧ ℓ . Notethat M is represented by the constant pro-spectrum { /ℓ } n . The multiplication map c ( ) ∧ ℓ ∧ M → M thus yields (compatible) multiplication maps /ℓ n ∧ /ℓ → /ℓ , for n sufficiently large. The associator c ( ) ∧ ℓ ∧ c ( ) ∧ ℓ ∧ M → M is zero as a map of pro-objects (the module structure being A ∞ ). Hence theassociator /ℓ n ∧ /ℓ n ∧ /ℓ → /ℓ is zero for n sufficiently large. Unitality of the multiplication ishandled similarly. (cid:3) Remark A.6.
Oka constructs similar module structures by hand (and for explicit n ) [Oka84, Section6], without addressing associativity question. TABLE MOTIVIC INVARIANTS ARE EVENTUALLY ´ETALE LOCAL 25
A.4.
Uniqueness for H Z .Lemma A.7. Let n ≥ .(1) The Moore object Z /n ∈ D ( Z ) admits a unique (up to homotopy) multiplication.(2) Let m ≥ . The Moore object Z /n ∈ D ( Z ) admits a unique (up to homotopy) structure of Z /mn -module such that the canonical map Z /mn → Z /n is a module map.Proof. For
A, B, C ∈ D ( Z ) ♥ , we have[ A ⊗ B, C ] ≃ [ π ( A ⊗ B ) , C ] . This implies that a multiplication on Z /n is the same as a multiplication in the category of abeliangroups, and similarly for the module structure of (2). But then uniqueness follows from the fact that Z → Z /mn → Z /n are surjections (and existence is even more obvious). (cid:3) Equivalently, the Eilenberg–MacLane spectrum H Z /n ∈ SH admits a unique H Z -linear multiplicationcompatible with the canonical unit map. Corollary A.8.
Let n ≥ and choose a multiplication m : /n ∧ /n → /n ∈ SH . Let e : SH → SH ( S ) denote the “constant sheaf ” functor. Then the induced multiplication on e ( /n ) ∧ H Z ≃ H Z /n is thecanonical one. Similarly for a choice of /mn -module structure on /n .Proof. Consider the commutative diagram SH e −−−−→ SH ( S ) M y y M D ( Z ) e −−−−→ DM ( S ) , where the left (respectively right) vertical functors is given by tensoring with the Eilenberg–MacLanespectrum (respectively motivic Eilenberg–MacLane spectrum). Along the left vertical arrow, by LemmaA.7, we see the multiplication on M ( /n ) induced by the choice of multiplication on /n is the same as thestandard multiplication on M ( /n ) = Z /n . It follows that the induced multiplication on M ( e ( /n )) ≃ e ( M ( /n )) is also the standard one. Let U : DM ( S ) → SH ( S ) denote the right adjoint to M ; it followsthat U M e ( /n ) ≃ U ( ) ∧ /n ≃ H Z /n carries the correct multiplication. The same argument works formodule structures. (cid:3) Appendix B. Inverting elements in homotopy rings
A variant of the following result is stated without proof on page 1 of [Art83].
Lemma B.1.
Let C be a symmetric monoidal ∞ -category in which N -indexed colimits exits and arepreserved by ⊗ in each variable separately. Let E, L ∈ C and suppose given a map τ : E → E ⊗ L .(1) There exists a canonical diagram F : N → C , informally described as E τ −→ E ⊗ L τ ⊗ id L −−−−→ E ⊗ L ⊗ → · · · . Denote its colimit by E [ τ − ] .(2) Given a map u : → E , there is a canonically induced map ¯ u : → E [ τ − ] .(3) Given a map m : E ⊗ E → E and a homotopy h as in the following diagram E ⊗ E E ⊗ L ⊗ E ⊗ L E ⊗ E ⊗ L ⊗ LE E ⊗ L E ⊗ L ⊗ L, τ ⊗ τm ≃ m ⊗ id L ⊗ id L τ h τ ⊗ id L there is a canonically induced map ¯ m : E [ τ − ] ⊗ E [ τ − ] → E [ τ − ] .(4) Suppose that C is stable and compactly generated, and L invertible. Suppose furthermore that thediagrams E ⊗ E ⊗ E E ⊗ E ⊗ E ⊗ LE E E E ⊗ L, id ⊗ u m id ⊗ τu m ≃ id ≃ τ commute in h C . (In other words, the multiplication m is homotopy left unital, and τ is given byleft multiplication by a certain homotopy element.) Then the composite E [ τ − ] ≃ E [ τ − ] ⊗ id ⊗ ¯ u −−−→ E [ τ − ] ⊗ E [ τ − ] ¯ m −→ E [ τ − ] is an equivalence. (A similar result holds for left multiplication.)Proof. (1) Let N ′ be the simplicial set with 0-cells 0 , , , · · · and for each n ∈ N a 1-cell from n to n + 1. There is a canonical map of simplicial sets N ′ → N N (here we use N to indicate the nerveof a category, for once not silently identifying categories with their nerves) which is an inner anodyneextension and hence in particular a categorical equivalence [Lur09, Lemma 2.2.5.2]. It follows thatFun( N , C ) → Fun( N ′ , C ) is an equivalence. The endomorphism τ induces an element of Fun( N ′ , C ) asdisplayed, which hence canonically lifts as claimed.(2) ¯ u exists by definition of a colimit diagram. We can make it slightly more explicit as follows. Notethat N ′ × ∆ → N N × ∆ is also a categorical equivalence [Lur09, Corollary 2.2.5.4], and consequentlymaps in Fun( N , C ) can be produced as the evident ladder diagrams, with homotopies filling the squares.We now consider the morphism of diagrams · · · E E ⊗ L E ⊗ L ⊗ L · · · , ≃ u ≃ τu ≃ τ uτ τ ⊗ id L where the homotopies are the tautological ones. The induces map on colimits is ¯ u .(3) Since ⊗ in C preserves N -indexed colimits in each variable separately (by assumption), we have E [ τ − ] ⊗ E [ τ − ] ≃ colim N × N F ⊗ F. The diagonal ∆ : N → N × N is cofinal and hence E [ τ − ] ⊗ E [ τ − ] ≃ colim N F ⊗ F ◦ ∆ ≃ colim N ′ F ⊗ F ◦ ∆ . The map 2 : N → N is also cofinal. Consequently in order to produce ¯ m , it suffices to produce homotopiesin the following diagram E ⊗ E E ⊗ L ⊗ E ⊗ L E ⊗ L ⊗ ⊗ E ⊗ L ⊗ · · · E E ⊗ L ⊗ E ⊗ L ⊗ · · · . τ ⊗ τm τ ⊗ τm m τ h τ h h Here m = m and m i for i > m and switch maps. We areprovided with h = h . If we denote the i -th square above as S i , then there is a canonical equivalence S i ≃ S ⊗ id L ⊗ i . We may thus choose h i = h ⊗ id L ⊗ i .(4) As before we have E [ τ − ] ⊗ ≃ colim N × N F ⊗ ≃ colim N ′ F ⊗ ◦ ∆. The composite in questionis the colimit of the following (composed ladder) diagram E ⊗ E ⊗ L ⊗ E ⊗ L ⊗ ⊗ · · · E ⊗ E E ⊗ L ⊗ E ⊗ L E ⊗ L ⊗ ⊗ E ⊗ L ⊗ · · · E E ⊗ L ⊗ E ⊗ L ⊗ · · · . τ ⊗ idid ⊗ u τ ⊗ id ⊗ idid ⊗ id ⊗ τu id ⊗ id ⊗ τ uτ ⊗ τm τ ⊗ τm mτ τ Here, we have suppressed the homotopies and labelled all the multiplication maps m , even if a switchmap is involved. By assumption, the vertical maps are homotopic to a composite of τ ’s and switch maps.Now, let T ∈ C be compact and denote by π : C → Set the functor [ T, − ]. Then π preserves filteredcolimits and the collection of all such functors π is conservative (by assumption). Hence it suffices to Let X n = ∆ , ` ∆ , ` · · · ` n − ∆ n − ,n , so that X n → ∆ n is inner anodyne (see e.g., [Lur09, Proof of Proposition3.2.1.13]). Let Y n = ∆ n ` X n N ′ . Since N ′ ≃ colim n X n and N N ≃ colim n ∆ n , we have colim n Y n ≃ N N . Since the class ofinner anodyne maps is weakly saturated (see [Lur09, Definition A.1.2.2]), each map N ′ → Y n is inner anodyne, and henceso is the colimiting map N ′ → N N . TABLE MOTIVIC INVARIANTS ARE EVENTUALLY ´ETALE LOCAL 27 show that when applying π to the above morphism of diagrams, we obtain an isomorphism of ind-objectsin sets. Restricting to the cofinal subcategory { , , , , . . . } of N , we obtain a diagram π ( E ⊗ L ⊗ ) π ( E ⊗ L ⊗ ) π ( E ⊗ L ⊗ ) π ( E ⊗ L ⊗ R · · · π ( E ⊗ L ⊗ ) π ( E ⊗ L ⊗ ) π ( E ⊗ L ⊗ ) π ( E ⊗ L ⊗ ) · · · , τ ˜ τ τ ˜ τ τ ˜ τ ˜ τ τ id τ id τ id id where ˜ τ n denotes an appropriate composite of τ and switch maps. We claim that the dashed identitymaps define an inverse morphism of ind-objects. Since we are working in a 1-category, we need onlyverify that all triangles commute. Since L is invertible, the switch map on L is homotopic to the identity[Dug14, Lemma 4.19], and hence ˜ τ n = τ n . This proves the desired commutativity. (cid:3) Hence, under the assumption that L is an invertible object in C and C is compactly generated (thesewill hold for all our applications), we conclude that E [ τ − ] admits “homotopy multiplication” which isnot quite, but close to, left unital. We need the next variant for modules. Lemma B.2.
In the situation of Lemma B.1, further assume that we are given an object M ∈ C equippedwith a map a : E ⊗ M → M .(1) In the situation of Lemma B.1(2), consider the map τ · : M u ⊗ id −−−→ E ⊗ M τ ⊗ id −−−→ E ⊗ L ⊗ M ≃ L ⊗ E ⊗ M id ⊗ a −−−→ L ⊗ M. Then there is a diagram F M : N → C informally described as M τ · −→ L ⊗ M id ⊗ τ · −−−→ L ⊗ L ⊗ M id ⊗ ( τ · ) ⊗ −−−−−−→ · · · . Denote its colimit by M [ τ − ] .(2) In the situation of Lemma B.1(3), suppose further that there we are given a homotopy h M : E ⊗ M E ⊗ L ⊗ L ⊗ M L ⊗ L ⊗ E ⊗ MM L ⊗ M L ⊗ L ⊗ M. τ ⊗ τ · id ⊗ a ≃ id ⊗ aτ · h M τ · Then, there is canonically induced map ¯ a : E [ τ − ] ⊗ M [ τ − ] → M [ τ − ] , such that the diagram E ⊗ M ME [ τ − ] ⊗ M [ τ − ] M [ τ − ] a ¯ a commutes.(3) Assume that we are in the situation of Lemma B.1(4) and further assume that the diagram ⊗ M E ⊗ MM M, u ⊗ id a ≃ id commutes. Then the composite M [ τ − ] ≃ ⊗ M [ τ − ] ¯ u ⊗ id −−−→ E [ τ − ] ⊗ M [ τ − ] ¯ a −→ M [ τ − ] is an equivalence.Proof. (1) By the same reasoning as in Lemma B.1(1), the described diagram exists.(2) By the same reasoning as in Lemma B.1(3), it suffices produce a transformation F ⊗ F M ◦ ∆ ⇒ F M . This is displayed in the following diagram: E ⊗ M E ⊗ L ⊗ L ⊗ M E ⊗ L ⊗ ⊗ L ⊗ ⊗ M · · · M L ⊗ ⊗ M L ⊗ ⊗ M · · · . τ ⊗ τa τ ⊗ τa a τ · h τ · h h Here, a = a and a i is a composite of the evident switch maps and a . The homotopy h is h M , while h i := h M ⊗ id L ⊗ i .(3) By the same reasoning as in Lemma B.1(4), the composite is given by taking colimits: ⊗ M ⊗ L ⊗ M ⊗ L ⊗ ⊗ M · · · E ⊗ M E ⊗ L ⊗ L ⊗ M E ⊗ L ⊗ ⊗ L ⊗ ⊗ M · · · M L ⊗ ⊗ M L ⊗ ⊗ M · · · , id ⊗ τ · u ⊗ id id ⊗ τ · τu ⊗ id ⊗ id τ u ⊗ id ⊗ id τ ⊗ τa τ ⊗ τa id ⊗ aτ τ where we have suppressed the switch maps. By assumption, the vertical composites are homotopic to acomposite of τ ’s and switch maps. It remains to note that the following diagram in sets π ( L ⊗ ⊗ M ) π ( L ⊗ ⊗ M ) π ( L ⊗ ⊗ M ) π ( L ⊗ ⊗ M ) · · · π ( L ⊗ ⊗ M ) π ( L ⊗ ⊗ M ) π ( L ⊗ ⊗ M ) π ( L ⊗ ⊗ M ) · · · τ · ˜ τ τ · ˜ τ · τ · ˜ τ · ˜ τ · τ · id τ · id τ · id id is commutative. This follows by the same reason as in Lemma B.1(4) — the switch maps are homotopicto the identity. (cid:3) Corollary B.3.
Let C be a compactly generated, stable, presentably symmetric monoidal ∞ -category, E ∈ C a homotopy unital associative ring, L ∈ P ic ( C ) and τ : → E ⊗ L a homotopy central element(i.e., left and right multiplication by τ induce homotopic maps E → E ⊗ L ). Let M be a homotopyassociative E -module (for example, E = M ).Then for any X, Y ∈ C , any map X → Y ⊗ M [ τ − ] factors through X ⊗ E [ τ − ] .Proof. We may apply the above lemmas (condition (2) is where we use centrality of τ ), and hence obtain¯ u : → E [ τ − ] , ¯ m : E [ τ − ] ⊗ M [ τ − ] → M [ τ − ]such that “right” multiplication by ¯ u is an equivalence α : M [ τ − ] ≃ M [ τ − ] ⊗ id ⊗ ¯ u −−−→ M [ τ − ] ⊗ E [ τ − ] ¯ m ′ −−→ M [ τ − ] . Here, ¯ m ′ is ¯ m composed with the switch map. Let f : X → Y ⊗ E [ τ − ] be any map and consider thecommutative diagram X ⊗ E [ τ − ] Y ⊗ M [ τ − ] ⊗ E [ τ − ] X ⊗ Y ⊗ M [ τ − ] ⊗ Y ⊗ M [ τ − ] Y ⊗ M [ τ − ] . f ⊗ id id ⊗ ¯ m ′ id ⊗ ¯ u f ⊗ id id ⊗ id ⊗ ¯ u id ⊗ α id ⊗ α − Since bottom composite is f , the result follows. (cid:3) Appendix C. Galois descent in invertible characteristic
In this section we establish some folklore results to the effect that homotopy orbits and fixed pointsare “the same” and “non-homotopical” if the group order is invertible. Then we apply this to obtainsome essentially trivial Galois descent results for SH and DM . They are well-known to experts andgeneralize [ELSØ17, Appendix C] (via a different approach). TABLE MOTIVIC INVARIANTS ARE EVENTUALLY ´ETALE LOCAL 29
C.1.
Universality of homotopy fixed points.
Let G be a finite group. Let C be an ∞ -categorywith finite coproducts and finite products. Denote by i : ∗ → BG the canonical map. The functor i ∗ : Fun( BG, C ) → C has a left adjoint i and a right adjoint i ∗ , given by i X = a g ∈ G X and i ∗ X = Y g ∈ G X. If C is furthermore pointed, i.e., ( − µ : i → i ∗ .Equivalently, there is a natural transformation i ∗ i → id; it is even natural in C [Har17, Construction3.13, Observation 3.14]. In particular, if F : C → D is a functor of pointed ∞ -categories with finitecoproducts and finite products, preserving finite coproducts, and A ∈ C , then the following diagramcommutes canonically F i A F µ A −−−−→ F i ∗ A (cid:13)(cid:13)(cid:13) y i F A µ FA −−−−→ i ∗ F A.
Now suppose that C is semiadditive (i.e., 0-semiadditive); then in particular µ A is an equivalencefor all A ∈ C . Denote by α : i → i the natural transformation corresponding by adjunction to thediagonal id → i ∗ i ≃ a G id ≃ Y G id . Inverting the middle arrow in the following spanid η −→ i ∗ i ∗ µi ∗ ←−− i i ∗ αi ∗ −−→ i i ∗ ǫ −→ id ∈ Fun(Fun(
BG, C ) , Fun(
BG, C ))yields a natural transformation T : id → id of the identity endofunctor of Fun( BG, C ). This constructionis natural in the semiadditive ∞ -category C . Remark C.1.
For A ∈ Fun(
BG, C ), the object i ∗ A has a canonical endomorphism t A : i ∗ A → i ∗ A givenby P g g . In the notation of [CSY18, Definition 2.11] we have T A ≃ R i t A . Example C.2. If C is the category of abelian groups, then Fun( BG, C ) is the category of abelian groups A with an action by G , and T A : A → A is given by a
7→ | G | P g ∈ G ga . Lemma C.3.
Let G have order n . Then T ≃ n T as endo-transformations of the identity endofunctorof Fun(
BG, C ) .Proof. This follows from [CSY18, Corollary 3.1.14]. (cid:3)
Definition C.4.
For a semiadditive ∞ -category C and a finite group G , denote byFun( BG, C ) ⊂ Fun(
BG, C )the full subcategory of those objects on which T acts invertibly. If C is additive, denote by Fun( BG, C ) r ⊂ Fun(
BG, C ) the full subcategory on which | G | − T acts invertibly. Proposition C.5.
Suppose that n = | G | is invertible in C , and assume that C is idempotent completeand stable.(1) Every object of Fun(
BG, C ) splits as A = A ⊕ A r , with A i ∈ Fun(
BG, C ) i .(2) For A ∈ Fun(
BG, C ) , B ∈ Fun(
BG, C ) r we have Map(
A, B ) = 0 = Map(
B, A ) .(3) Let p : BG → ∗ be the canonical map. Then p ∗ C ⊂ Fun(
BG, C ) .(4) The functor i ∗ : Fun( BG, C ) → C is an equivalence.(5) The functor p ∗ : C → Fun(
BG, C ) is an equivalence.Proof. It follows from Lemma C.3 that
T /n defines an idempotent of every object, and thus 1 − T /n defines a complementary idempotent [Lur16, Warning 1.2.4.8]. Since these idempotents are byconstruction preserved by all morphisms, this (together with idempotent completeness of C and henceFun( BG, C ) [Lur09, Corollaries 4.4.5.15 and 5.1.2.3]) immediately implies (1) and (2).(3) It follows from Remark C.1 and [CSY18, Proposition 3.13, Example 3.12] applied to the cartesiansquare G −−−−→ ∗ y i y ∗ i −−−−→ BG that i ∗ T A = nt A . If A = p ∗ A ′ , then t A = n and so i ∗ T A = n is an isomorphism. The result followssince i ∗ is conservative.(4) The functor i ∗ : Fun( BG, C ) → C has a left (and right) adjoint given by A ( i ∗ A ) . Since i ∗ isconservative, it suffices to show that for all A ∈ C we have i ∗ ( i ∗ ( A ) ) ≃ A . Let B ∈ C . It suffices to showthat we obtain an equivalence after applying π i Map( B, − ). Since π i Map( B, − ) : C → (Mod Z [1 /n ] ) ≤ isadditive, this reduces to C = (Mod Z [1 /n ] ) ≤ , i.e., the ordinary 1-category of Z [1 /n ]-modules. This caseis straightforward.(5) Since i ∗ p ∗ ≃ ( pi ) ∗ ≃ id, this follows from (3) and (4). (cid:3) Corollary C.6.
Assumptions as in Proposition C.5. For every A ∈ Fun(
BG, C ) we have canonicalequivalences A hG ≃ i ∗ A ≃ A hG . Proof.
The functors ( − ) hG and ( − ) hG are given by p ∗ and p , respectively. Proposition C.5 implies that p ∗ can be identified with the inclusion Fun( BG, C ) → Fun(
BG, C ), and that its common right and leftadjoint is given by A A . The result follows. (cid:3) Corollary C.7.
Let F : C → D be an additive functor of additive, idempotent complete ∞ -categorieson which | G | is invertible. Then for any A ∈ Fun(
BG, C ) we have ( F A ) hG ≃ F ( A hG ) ≃ F ( A hG ) ≃ ( F A ) hG . Proof.
Any additive functor commutes with formation of ( − ) (and i ∗ ), so this follows from CorollaryC.6. (cid:3) C.2.
Galois descent.
Fix a finite group G . In the category Fin G we have the object G , with auto-morphism group canonically isomorphic to G itself. In this way we obtain an action of the group G on G ∈ Fin G , i.e., G : BG → Fin G . Composing with the canonical functor Fin G → Span(Fin G ) we obtainan action of G on G ∈ Span(Fin G ). In the category Span(Fin G ) we also have the maps a : ∗ → G and b : G → ∗ , corresponding to the spans ∗ ←− G ≃ −→ G and G ≃ ←− G → ∗ . It is straightforward to check that these refine to G -equivariant maps, where we let G act trivially on ∗ .The composite ba : ∗ → ∗ is given by the span ∗ ← G → ∗ and denoted by ( G ).In the following result, we denote by SH ( BG ) the genuine G -equivariant stable category; see forexample [BH18, § Lemma C.8.
Consider the functor σ : Span(Fin G ) → SH ( BG )[1 / | G | , / ( G )] . Then σ ( a ) exhibits σ ( ∗ ) as σ ( G ) hG .Proof. The composite ab : G → G is homotopic to T G / | G | by construction. It follows from CorollaryC.6 that in any category where | G | is invertible, G hG is given by the summand of G corresponding tothe idempotent ab/n .After inverting | G | , we obtain a splitting G ≃ G hG ⊕ G ′ and we can write the maps a, b in matrixform as a = ( a , a ) and b = ( b b ). Since a factors through G hG we have a = 0. If we further invert( G ), then ( G ) = ba = b a + b a = b a is an equivalence, and also (cid:18) id 00 0 (cid:19) = ab = (cid:18) a b a b a b a b (cid:19) , so a b = id. It follows that a = a : → G hG is an equivalence. This was to be shown. (cid:3) Now let S be a scheme and S ′ /S a finitely presented finite ´etale scheme with a free and transitive G -action; i.e., a G -torsor, or in other words a G -Galois extension of S . We then have a functor SH ( BG ) → SH ( S ). In fact Grothendieck’s Galois theory supplies us with a functor g : Fin G → Sm S , X X × G S ′ and we have a natural transformation [BH18, Proposition 10.6] c : SH ( − ) → SH ( g ( − )) ∈ Fun(Fin G , C at) . Lemma C.9.
The functors SH ( BG ) → SH ( S ) → SH ( S )[1 / , / | G | ] + and SH ( BG ) → SH ( S ) → DM ( S, Z [1 / | G | ]) TABLE MOTIVIC INVARIANTS ARE EVENTUALLY ´ETALE LOCAL 31 invert the endomorphism ( G ) of ∈ SH ( BG ) .Proof. The problem is Zariski local on S , so we may assume that S is affine. Then since S ′ is finitelypresented, it is already defined (as a G -torsor) over some scheme T under S of finite type over Spec( Z ).We may thus assume that S is Noetherian and finite dimensional. In this situation pullback to fields isconservative [Bac18a, Corollary 14], and so we reduce to the case when S is the spectrum of a field. Inthis case the morphism A ( G ) ≃ [ , ] SH ( BG ) → [ , ] SH ( S )[1 / + ≃ Z [1 / A ( G ) ≃ [ , ] SH ( BG ) → [ , ] DM ( k ) ≃ Z . The result follows. (cid:3)
Now let E ∈ SH ( S ) and write f : S ′ → S for the canonical map. The object f ∗ f ∗ E acquires a G -action, coming from the action of G on S ′ . Corollary C.10. (1) The natural map E → f ∗ f ∗ E refines to a G -equivariant map (where the sourceis given the trivial G -action).(2) If E ∈ SH ( S )[1 / , / | G | ] + or DM ( S, Z [1 / | G | ]) then the above map presents E as ( f ∗ f ∗ E ) hG .Moreover this limit diagram is preserved by any additive functor.Proof. We claim that the G -equivariant map E → f ∗ f ∗ E is given by c ( a ) ∧ E , where a is the map fromLemma C.8. The same lemma, together with Lemma C.9 (and Corollary C.7), then implies the result.To prove the claim, we note that by the projection formula we indeed have f ∗ f ∗ E ≃ ( f ∗ f ∗ ) ∧ E ,compatibly with the G -actions. This reduces to E = . The transformation c : SH ( − ) → SH ( g ( − ))induces by passage to adjoints exchange transformations f c S ′ → c S f and c S f ∗ → f ∗ c S ′ . The former is an equivalence essentially by construction, and the latter is an equivalence because f ≃ f ∗ on both sides (compatibly). It follows that the G -action on f ∗ f ∗ SH ( S ) is the same as the one inducedby c from the G -action on f ∗ f ∗ SH ( BG ) . We have thus reduced to showing that the G -action on f ∗ f ∗ ≃ f f ∗ ≃ Σ ∞ + G ∈ SH ( BG ) is the same as the one we constructed at the beginning of thissubsection. This can be verified directly by working in Span(Fin G ). (cid:3) Remark C.11.
In the above two results, we could have more generally replaced DM by the category ofmodules over any ring spectrum in which the Hopf map η is trivial, such as algebraic cobordism MGL. References [Art83] R. D. Arthan. Localization of stable homotopy rings. In
Mathematical Proceedings of the Cambridge Philosoph-ical Society , volume 93, pages 295–302. Cambridge University Press, 1983.[Ayo07] J. Ayoub. Les six op´erations de Grothendieck et le formalisme des cycles ´evanescents dans le monde motivique.I, 2007.[Ayo14] J. Ayoub. La r´ealisation ´etale et les op´erations de Grothendieck.
Ann. Sci. ´Ec. Norm. Sup´er.(4) , 47(1):1–145,2014.[Bac17] T. Bachmann. The generalized slices of hermitian K -theory. Journal of Topology , 10(4):1124–1144, 2017.arXiv:1610.01346.[Bac18a] T. Bachmann. Motivic and real ´etale stable homotopy theory.
Compositio Mathematica , 154(5):883–917, 2018.arXiv:1608.08855.[Bac18b] T. Bachmann. On the conservativity of the functor assigning to a motivic spectrum its motive.
Duke Math. J. ,167(8):1525–1571, 06 2018. arXiv:1506.07375.[Bac18c] T. Bachmann. Rigidity in ´etale motivic stable homotopy theory. 2018. arXiv:1810.08028.[BH18] T. Bachmann and M. Hoyois. Norms in motivic homotopy theory. arXiv:1711.03061, 2018.[BKSØ15] A. J. Berrick, M. Karoubi, M. Schlichting, and P. A. Østvær. The Homotopy Fixed Point Theorem and theQuillen-Lichtenbaum conjecture in Hermitian K -theory. Adv. Math. , 278:34–55, 2015.[BS19] M. Behrens and J. Shah. C -equivariant stable homotopy from real motivic stable homotopy. arXiv preprintarXiv:1908.08378 , arXiv:1908.08378, 2019.[CD16] D.-C. Cisinski and F. D´eglise. ´Etale motives. Compos. Math. , 152(3):556–666, 2016.[CD19] D.-C. Cisinski and F. D´eglise.
Triangulated categories of mixed motives . Springer Monographs in Mathematics.Springer, Cham, [2019] c (cid:13) K -theory. arXiv:1905.06611, 2019.[CSY18] S. Carmeli, T. M. Schlank, and L. Yanovski. Ambidexterity in t ( n )-local stable homotopy theory.arXiv:1811.02057, 2018.[DFST82] W. G. Dwyer, E. M. Friedlander, V. P. Snaith, and R. W. Thomason. Algebraic K -theory eventually surjectsonto topological K -theory. Invent. Math. , 66(3):481–491, 1982. [Dug14] D. Dugger. Coherence for invertible objects and multigraded homotopy rings.
Algebraic & Geometric Topology ,14(2):1055–1106, 2014.[EHK +
18] E. Elmanto, M. Hoyois, A. A. Khan, V. Sosnilo, and M. Yakerson. Framed transfers and motivic fundamentalclasses. arXiv:1711.05248, 2018.[EHK +
19] E. Elmanto, M. Hoyois, A. A. Khan, V. Sosnilo, and M. Yakerson. Motivic infinite loop spaces. arXiv:1711.05248,2019.[EK18] E. Elmanto and A. A. Khan. Perfection in motivic homotopy theory. arXiv preprint arXiv:1812.07506 , 2018.[EKM08] R. Elman, N. Karpenko, and A. Merkurjev.
The algebraic and geometric theory of quadratic forms , volume 56of
American Mathematical Society Colloquium Publications . American Mathematical Society, Providence, RI,2008.[EL99] R. Elman and C. Lum. On the cohomological 2-dimension of fields.
Communications in Algebra , 27(2):615–620,1999.[ELSØ17] E. Elmanto, M. Levine, M. Spitzweck, and P. A. Østvær. Algebraic cobordism and ´etale cohomology.arXiv:1711.06258, 2017.[GJ09] P. G. Goerss and J. F. Jardine.
Simplicial homotopy theory . Springer Science & Business Media, 2009.[GP18] G. Garkusha and I. Panin. Framed motives of algebraic varieties (after V. Voevodsky). arXiv:1409.4372, 2018.[GSZ16] S. Gille, S. Scully, and C. Zhong. Milnor-Witt K -groups of local rings. Adv. Math. , 286:729–753, 2016.[Har17] Y. Harpaz. Ambidexterity and the universality of finite spans. arXiv:1703.09764, 2017.[Hoy17] M. Hoyois. The six operations in equivariant motivic homotopy theory.
Adv. Math. , 305:197–279, 2017.[Hoy18] M. Hoyois. The localization theorem for framed motivic spaces. arXiv:1807.04253, 2018.[HW19] C. Haesemeyer and C. A. Weibel.
The norm residue theorem in motivic cohomology , volume 200 of
Annals ofMathematics Studies . Princeton University Press, Princeton, NJ, 2019.[Jar10] J. F. Jardine.
Generalized etale cohomology theories . Modern Birkh¨auser Classics. Birkh¨auser/Springer BaselAG, Basel, 2010. Reprint of the 1997 edition [MR1437604].[Lev00] M. Levine. Inverting the motivic bott element.
K-theory , 19(1):1–28, 2000.[Lev11] M. Levine. The slice filtration and Grothendieck-Witt groups.
Pure Appl. Math. Q. , 7(4, Special Issue: Inmemory of Eckart Viehweg):1543–1584, 2011.[Lev13] M. Levine. Convergence of Voevodsky’s slice tower.
Documenta Mathematica , 18:907–941, 2013.[Lur09] J. Lurie.
Higher topos theory . Number 170. Princeton University Press, 2009.[Lur16] J. Lurie. Higher algebra, May 2016.[Lur18] J. Lurie. Spectral algebraic geometry. February 2018.[LYZ13] M. Levine, Y. Yang, and G. Zhao. Algebraic elliptic cohomology theory and flops, i. arXiv preprintarXiv:1311.2159 , 2013.[Mil80] J. S. Milne.
Etale Cohomology (PMS-33) . Princeton mathematical series. Princeton University Press, 1980.[MNN17] A. Mathew, N. Naumann, and J. Noel. Nilpotence and descent in equivariant stable homotopy theory.
Adv.Math. , 305:994–1084, 2017.[Mor] F. Morel. A -algebraic topology over a field . Lecture Notes in Mathematics, Vol. 2052. Springer Verlag.[Mor04a] F. Morel. An introduction to A -homotopy theory. pages 357–441, 2004.[Mor04b] F. Morel. Sur les puissances de l’id´eal fondamental de l’anneau de Witt. Comment. Math. Helv. , 79(4):689–703,2004.[Mor05] F. Morel. The stable A -connectivity theorems. K-Theory , 35(1-2):1–68, 2005.[Mor12] F. Morel. A -Algebraic Topology over a Field . Lecture Notes in Mathematics. Springer Berlin Heidelberg, 2012.[MVW06] C. Mazza, V. Voevodsky, and C. A. Weibel. Lecture notes on motivic cohomology , volume 2 of
Clay MathematicsMonographs . Amer. Math. Society, Providence, RI; Clay Mathematics Institute, Cambridge, MA, 2006.[Neu13] J. Neukirch.
Algebraic number theory , volume 322. Springer Science & Business Media, 2013.[Oka84] S. Oka. Multiplications on the Moore spectrum.
Memoirs of the Faculty of Science, Kyushu University. SeriesA, Mathematics , 38(2):257–276, 1984.[Øst03] P. A. Østvær. ´Etale descent for real number fields.
Topology , 42(1):197–225, 2003.[OVV07] D. Orlov, A. Vishik, and V. Voevodsky. An exact sequence for K M ∗ / Ann. of Math. (2) , 165(1):1–13, 2007.[RØ05] A. Rosenschon and P. A. Østvær. The homotopy limit problem for two-primary algebraic K -theory. Topology ,44(6):1159–1179, 2005.[RØ08a] O. R¨ondigs and P. A. Østvær. Rigidity in motivic homotopy theory.
Mathematische Annalen , 341(3):651–675,2008.[RØ08b] O. R¨ondigs and P. A. Østvær. Modules over motivic cohomology.
Advances in Mathematics , 219(2):689 – 727,2008.[R¨on19] O. R¨ondigs. Remarks on motivic Moore spectra. arXiv preprint arXiv:1910.00834 , 2019.[RSØ18] O. R¨ondigs, M. Spitzweck, and P. A. Østvær. The motivic Hopf map solves the homotopy limit problem for K -theory. Doc. Math. , 23:1405–1424, 2018.[RSØ19] O. R¨ondigs, M. Spitzweck, and P. A. Østvær. The first stable homotopy groups of motivic spheres.
Ann. ofMath. (2) , 189(1):1–74, 2019.[Ser02] J.-P. Serre.
Galois cohomology . Springer Monographs in Mathematics. Springer-Verlag, Berlin, english edition,2002. Translated from the French by Patrick Ion and revised by the author.[Spi18] M. Spitzweck. A commutative P -spectrum representing motivic cohomology over Dedekind domains. M´em.Soc. Math. Fr. (N.S.) , (157):110, 2018.[Sta17] The Stacks Project Authors. The stacks project. 2017.[SV00] A. Suslin and V. Voevodsky. Bloch-Kato conjecture and motivic cohomology with finite coefficients. In
Thearithmetic and geometry of algebraic cycles (Banff, AB, 1998) , volume 548 of
NATO Sci. Ser. C Math. Phys.Sci. , pages 117–189. Kluwer Acad. Publ., Dordrecht, 2000.
TABLE MOTIVIC INVARIANTS ARE EVENTUALLY ´ETALE LOCAL 33 [Tho85] R. W. Thomason. Algebraic K -theory and ´etale cohomology. Ann. Sci. ´Ecole Norm. Sup. (4) , 18(3):437–552,1985.[Tho91] R. W. Thomason. The local to global principle in algebraic K -theory. pages 381–394, 1991.[TT90] R. W. Thomason and T. Trobaugh. Higher algebraic K -theory of schemes and of derived categories. 88:247–435,1990.[Voe02] V. Voevodsky. Open problems in the motivic stable homotopy theory. I. 3:3–34, 2002.[Voe03] V. Voevodsky. Motivic cohomology with Z / Publ. Math. Inst. Hautes ´Etudes Sci. , (98):59–104,2003.[Voe11] V. Voevodsky. On motivic cohomology with Z /l -coefficients. Annals of mathematics , 174(1):401–438, 2011.
Department of Mathematics, Massachusetts Institute of Technology, Cambridge, MA, USA
E-mail address : [email protected] Department of Mathematics, Harvard University, Cambridge, MA, USA
E-mail address : [email protected] Department of Mathematics, University of Oslo, Norway
E-mail address ::