ScalingStacks

10. Trace maps

In this section we apply the work of the preceding sections to give a universal characterization of the topological Dennis trace map K→T​H​HK\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}THH [16] and the cyclotomic trace map K→T​CK\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}TC [17]. More generally, we identify all natural transformations of additive functors from KK-theory to T​H​HTHH: they are the multiples of the topological Dennis trace map. This identification provides a very satisfying conceptual construction of the cyclotomic trace map, and of course makes it clear that all known definitions are consistent.

10.1. T​H​HTHH as a localization invariant

We begin by observing that T​H​HTHH provides a localizing invariant of small stable ∞\infty-categories. Although it is possible to do this directly in the setting of ∞\infty-categories (see for instance the more general discussion of topological chiral homology in [53, §5.3] or the constructions outlined in [8, 5.1.1]), we use existing constructions in the setting of spectral categories in order to ease technical difficulties that arise in the subsequent construction of T​RTR and T​CTC. Our basic sources for this material are [10] and [11].

Recall that for a small spectral category 𝒞{\mathcal{C}} we can define T​H​H​(𝒞)THH({\mathcal{C}}) in terms of the Hochschild-Mitchell cyclic nerve for spectral categories [10, §3]. The cyclic nerve is defined as the simplicial object

Nqcyc​𝒞=⋁𝒞⁡(cq−1,cq)∧⋯∧𝒞⁡(c0,c1)∧𝒞⁡(cq,c0),N^{\cyc}_{q}{\mathcal{C}}=\bigvee{\mathcal{C}}(c_{q-1},c_{q})\wedge\dotsb\wedge{\mathcal{C}}(c_{0},c_{1})\wedge{\mathcal{C}}(c_{q},c_{0}),

where the sum is over the (q+1)(q+1)-tuples (c0,…,cq)(c_{0},\dotsc,c_{q}) of objects of 𝒞{\mathcal{C}}. This becomes a simplicial object using the usual cyclic bar construction face and degeneracy maps: The unit maps of 𝒞{\mathcal{C}} induce the degeneracy maps, and the composition maps in 𝒞{\mathcal{C}} (along with the twist map at the end) induce the face maps. We denote the geometric realization as Ncyc​𝒞N^{\cyc}{\mathcal{C}}.

The spectrum Ncyc​𝒞N^{\cyc}{\mathcal{C}} has the correct homotopy type only when 𝒞{\mathcal{C}} has cofibrant mapping spectra [10, 3.1]. Since the cofibrant objects in the Morita model structure on small spectral categories reviewed in theorem 2.2 have cofibrant mapping spectra [74, 4.18], we can define the functor

T​H​H:=Ncyc∘Q:Cat𝒮⟶𝒮THH:=N^{\cyc}\circ Q\colon\Cat_{\mathcal{S}}\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}\mathcal{S}

where QQ denotes the cofibrant replacement functor in the Morita model structure on Cat𝒮\Cat_{\mathcal{S}}. Since this construction preserves Morita equivalences [10, 5.12] (and in fact DK-equivalences [10, 5.9]) the functor descends to the level of ∞\infty-categories.

0NQV

Lemma 10.1. The functor

T​H​H:Cat𝒮⟶𝒮THH\colon\Cat_{\mathcal{S}}\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}\mathcal{S}

induces a functor of ∞\infty-categories

T​H​H:Cat∞ex≃N⁡((Cat𝒮)c)​[W−1]⟶N⁡((𝒮)c)​[W−1]≃𝒮∞.THH\colon\Cat_{\infty}^{\ex}\simeq\mathrm{N}((\Cat_{\mathcal{S}})^{\mathrm{c}})[W^{-1}]\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}\mathrm{N}((\mathcal{S})^{\mathrm{c}})[W^{-1}]\simeq{\mathcal{S}}_{\infty}.

This definition of T​H​HTHH as a functor of ∞\infty-categories lets us deduce the following proposition from known properties of T​H​HTHH in the setting of spectral categories.

0NQW

Proposition 10.2. T​H​HTHH is a localizing invariant of small stable ∞\infty-categories.

0NQX

Proof. The cyclic bar construction commutes with filtered homotopy colimits of spectral categories. Furthermore, T​H​H​(−)THH(-) takes exact sequences of spectral categories to exact sequences of spectra [10, 7.1]. Therefore, the induced functor on ∞\infty-categories is a localizing invariant. ∎

The force of the co-representability result for algebraic KK-theory (theorem 7.13) is that it implies, via the spectral Yoneda lemma, the following identification of the spectrum of natural transformations of additive functors K→EK\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}E.

0NQY

Theorem 10.3. Given an additive invariant E:Cat∞ex⟶𝒮∞E\colon\Cat_{\infty}^{\ex}\longrightarrow{\mathcal{S}}_{\infty} with values in the stable ∞\infty-category of spectra, we have a natural equivalence

Nat⁡(K,E)≃E⁡(𝒮∞ω),\Nat(K,E)\simeq E({\mathcal{S}}_{\infty}^{\omega}),

where Nat⁡(K,E)\Nat(K,E) denotes the spectrum of natural transformations from KK to EE as additive invariants from small stable ∞\infty-categories to spectra.

0NQZ

Proof. By theorem 6.10, we can describe the additive invariants KK and EE as elements of FunL​(ℳadd,𝒮∞)\mathrm{Fun}^{\mathrm{L}}({\mathcal{M}}_{\mathrm{add}},{\mathcal{S}}_{\infty}). The equivalence

Nat⁡(Map⁡(𝒰add​(𝒮∞ω),−),E)≃E⁡(𝒮∞ω)\Nat(\mathrm{Map}({\mathcal{U}}_{\mathrm{add}}({\mathcal{S}}_{\infty}^{\omega}),-),E)\simeq E({\mathcal{S}}_{\infty}^{\omega})

follows from 7.13 and the spectral Yoneda lemma. ∎

In particular, applying theorem 10.3 to T​H​HTHH yields the following corollary.

0NR0

Corollary 10.4. We have an equivalence of spectra

Nat⁡(K⁡(−),T​H​H​(−))≅T​H​H​(𝒮ω)≃T​H​H​(𝕊)≃𝕊.\Nat(K(-),THH(-))\cong THH({\mathcal{S}}^{\omega})\simeq THH(\mathbb{S})\simeq\mathbb{S}.

Passing to π0\pi_{0} on both sides we obtain an isomorphism between homotopy classes of natural transformations and π0​(𝕊)≅ℤ\pi_{0}(\mathbb{S})\cong{\mathbb{Z}}.

10.2. The topological Dennis trace map

Next, we want to characterize the topological Dennis trace K→T​H​HK\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}THH in terms of the classification of corollary 10.4. To do this, we briefly recall the construction of the topological Dennis trace map for spectral categories and verify that it descends to provide a natural transformation of additive invariants from KK-theory to T​H​HTHH on ∞\infty-categories. We rely on the work of [10, §5].

Recall that any small spectral category 𝒜{\mathcal{A}} is equivalent in the Morita model structure to the small spectral category 𝒜^perf\widehat{{\mathcal{A}}}_{\perf} of perfect modules. The category 𝒜^perf\widehat{{\mathcal{A}}}_{\perf} admits the structure of a Waldhausen category by restriction from the spectral model structure on 𝒜^\widehat{{\mathcal{A}}}. As explained in [10, §15], without loss of generality we can work with categories enriched in EKMM SS-modules. In this case, because all objects are fibrant the weak equivalences of the Waldhausen structure on 𝒜^perf\widehat{{\mathcal{A}}}_{\perf} are compatible with the spectral enrichment in the sense of [11, §1].

Now, following [11, §5], we construct a trace using the perspective of [57] by “mixing” a cyclic bar construction and Waldhausen’s S∙S_{\bullet} construction; this is definition [11, 5.12]. Upon passage to underlying ∞\infty-categories, we end up with a natural transformation of localizing invariants.

0NR1

Lemma 10.5. The topological Dennis trace above induces a natural transformation of localizing invariants

K⟶T​H​H.K\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}THH.
0NR2

Proof. It is clear from the construction of the trace described above that it descends to a natural transformation of functors of ∞\infty-categories K→T​H​HK\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}THH. One checks on each side that the trace commutes with filtered homotopy colimits of spectral categories, and so the result follows. ∎

As a corollary, we know that there exists an element x∈ℤx\in{\mathbb{Z}} such that xx corresponds to the homotopy class of the topological Dennis trace under the identification of corollary 10.4. The following theorem identifies this element as the unit.

0NR3

Theorem 10.6. The topological Dennis trace is (up to homotopy) the natural transformation given by the identity element 1∈π0​(T​H​H​(𝕊))≅π0​(𝕊)≅ℤ1\in\pi_{0}(THH(\mathbb{S}))\cong\pi_{0}(\mathbb{S})\cong{\mathbb{Z}}.

0NR4

Proof. Given a point ϕ\phi in Nat⁡(K⁡(−),T​H​H​(−))\Nat(K(-),THH(-)) (a specific natural transformation, that is), we can describe the corresponding element in π0​(𝕊)\pi_{0}(\mathbb{S}) as the homotopy class represented by the composite

𝕊\textstyle{\mathbb{S}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Map⁡(𝒰add​(𝒮∞ω),𝒰add​(𝒮∞ω))≃K⁡(𝕊)\textstyle{\mathrm{Map}({\mathcal{U}}_{\mathrm{add}}({\mathcal{S}}_{\infty}^{\omega}),{\mathcal{U}}_{\mathrm{add}}({\mathcal{S}}_{\infty}^{\omega}))\simeq K(\mathbb{S})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}ϕ\scriptstyle{\phi}T​H​H​(𝕊)≃𝕊,\textstyle{THH(\mathbb{S})\simeq\mathbb{S},}

where the first map picks out the identity map in Map⁡(𝒰add​(𝒮∞ω),𝒰add​(𝒮∞ω))\mathrm{Map}({\mathcal{U}}_{\mathrm{add}}({\mathcal{S}}_{\infty}^{\omega}),{\mathcal{U}}_{\mathrm{add}}({\mathcal{S}}_{\infty}^{\omega})). There is also a classical map i:𝕊→K⁡(𝕊)i\colon\mathbb{S}\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}K(\mathbb{S}) constructed (for instance) as the canonical inclusion of the finite sets into finite spaces. Waldhausen’s calculations [86, §5] imply that the homotopy class of ii is represented by 1∈π0​(K​(𝕊))1\in\pi_{0}(K(\mathbb{S})). On the other hand, since the identity map is the unit for the multiplication on π0​(Map⁡(𝒰add​(𝒮∞ω),𝒰add​(𝒮∞ω)))\pi_{0}(\mathrm{Map}({\mathcal{U}}_{\mathrm{add}}({\mathcal{S}}_{\infty}^{\omega}),{\mathcal{U}}_{\mathrm{add}}({\mathcal{S}}_{\infty}^{\omega}))) induced by the composition, it must also be represented by 1∈π0​(K​(𝕊))1\in\pi_{0}(K(\mathbb{S})). Finally, specializing to the case when ϕ\phi is the topological Dennis trace, Waldhausen [86, 5.2] proves that the composite

𝕊\textstyle{\mathbb{S}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}i\scriptstyle{i}K⁡(𝕊)\textstyle{K(\mathbb{S})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}T​H​H​(𝕊)≃𝕊\textstyle{THH(\mathbb{S})\simeq\mathbb{S}}

is homotopic to the identity. ∎

10.3. T​CTC and the cyclotomic trace map

Much of the interest in the topological Dennis trace comes from the fact that the T​H​HTHH spectrum comes with an additional equivariant structure (generalizing the classical connection between the cyclic bar construction of a space and the free loop space) which allows a refinement into a theory called T​CTC, the topological cyclic homology. The topological Dennis trace lifts to a map

K⟶T​CK\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}TC

called the cyclotomic trace map [17].

We will once again apply co-representability and point-set models to characterize the cyclotomic trace. However, when attempting to apply our co-representability results to T​CTC, we run into certain obstacles. Although T​CTC is Morita invariant and satisfies localization [10], it does not preserve filtered colimits and is therefore not a localizing (or additive) invariant. Nonetheless, we can adapt our results to characterize the cyclotomic trace in this setting.

We begin by recalling the definition of T​CTC in the context of spectral categories. Our review is brief; once again, we refer the interested reader to [11, §5] and [10, §4] for authoritative treatment. Fix a prime pp. For a spectral category 𝒞{\mathcal{C}}, we can realize T​H​H​(𝒞)THH({\mathcal{C}}) as a cyclotomic S1S^{1}-spectrum. It is convenient to use the Bokstedt model of T​H​HTHH, which is revised in detail in [10, §4]. Since the Bokstedt model is naturally weakly equivalent to Ncyc​Q​𝒞N^{\cyc}Q{\mathcal{C}} for any spectral category 𝒞{\mathcal{C}} [10, 3.1], we can leverage the work above. A cyclotomic structure is additional structure on an equivariant spectrum arising from the self-equivalence S1/H≅S1S^{1}/H\cong S^{1} (for finite H⊂S1H\subset S^{1}) that models the structure of the free loop space.

Roughly speaking, what we have is a set of compatible maps

ϕH​T⟶T\phi^{H}T\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}T

for finite H⊂S1H\subset S^{1}. The equivariant structure allows us to consider the associated non-equivariant spectra

T​Rn​(𝒞)=T​H​H​(𝒞)Cpn−1,TR^{n}({\mathcal{C}})=THH({\mathcal{C}})^{C_{p^{n-1}}},

the fixed points with respect to the induced Cpn−1C_{p^{n-1}} action. The inclusion of fixed points and the cyclotomic structure give rise to maps FF and RR respectively

F,R:T​Rn⟶T​Rn−1.F,R\colon TR^{n}\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}TR^{n-1}.

We define T​Cn​(𝒞)TC^{n}({\mathcal{C}}) to be the homotopy equalizer

holimF,R⁡T​Rn​(𝒞)⟶T​Rn−1​(𝒞).\holim_{F,R}TR^{n}({\mathcal{C}})\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}TR^{n-1}({\mathcal{C}}).

We then have that

T​C​(𝒞)=holimn⁡T​Cn​(𝒞),TC({\mathcal{C}})=\holim_{n}TC^{n}({\mathcal{C}}),

where we form the homotopy limit over the maps induced by the restriction RR; this definition is equivalent to the one originally given in [17].

The work of [10, §5] produces a description of T​H​H​(𝒞)THH({\mathcal{C}}) for a spectral category 𝒞{\mathcal{C}} as a cyclotomic spectrum, and hence constructions of T​RnTR^{n}, T​CnTC^{n}, and T​CTC. Moreover, a cyclotomic trace map K→T​CK\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}TC is constructed which arises from compatible maps K→T​CnK\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}TC^{n}. As above, we import these constructions into the setting of ∞\infty-categories. First, we have the following lemma:

0NR5

Lemma 10.7. The functors

T​Rn,T​Cn,T​C:Cat𝒮⟶𝒮TR^{n},TC^{n},TC\colon\Cat_{\mathcal{S}}\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}\mathcal{S}

induces functors of ∞\infty-categories

T​Rn,T​Cn,T​C:Cat∞ex≃N⁡((Cat𝒮)c)​[W−1]⟶N⁡((𝒮)c)​[W−1]≃𝒮∞.TR^{n},TC^{n},TC\colon\Cat_{\infty}^{\ex}\simeq\mathrm{N}((\Cat_{\mathcal{S}})^{\mathrm{c}})[W^{-1}]\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}\mathrm{N}((\mathcal{S})^{\mathrm{c}})[W^{-1}]\simeq{\mathcal{S}}_{\infty}.
0NR6

Proof. By [10, 3.9], maps which induce equivalences on T​H​HTHH induce equivalences on T​RnTR^{n}, T​CnTC^{n}, and T​CTC. As a consequence, the result follows from lemma 10.1. ∎

Next, we observe that each of the objects T​CnTC^{n} provides a localizing invariant.

0NR7

Proposition 10.8. The functor T​CnTC^{n} is a localizing invariant of stable ∞\infty-categories with values in the stable ∞\infty-category of spectra.

0NR8

Proof. The localization theorem of [10, 7.1] implies {T​Cn​(𝒞)}\{TC^{n}({\mathcal{C}})\} takes strict-exact sequences of spectral categories to strict exact sequences of small stable ∞\infty-categories. Thus, we need to show that T​Rn​(𝒞)TR^{n}({\mathcal{C}}) preserve filtered colimits. We know this for T​H​HTHH, and the result now follows inductively from consideration of the fundamental cofibration sequence (e.g., [39, 2.1.4])

T​H​H​(𝒞)Cpn−1⟶T​Rn​(𝒞)⟶T​Rn−1​(𝒞)THH({\mathcal{C}})_{C_{p^{n-1}}}\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}TR^{n}({\mathcal{C}})\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}TR^{n-1}({\mathcal{C}})

(where the left-hand term denotes the homotopy orbit space) and the fact that homotopy orbits commute with filtered colimits. ∎

The topological Dennis trace lifts through the constructions of T​CnTC^{n}, essentially by construction. Roughly speaking (see [11, 5.12] for a detailed construction), the trace is induced by an “inclusion of objects” map S∙​𝒞→T​H​H​(S∙​𝒞)S_{\bullet}{\mathcal{C}}\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}THH(S_{\bullet}{\mathcal{C}}), given by taking an object to its identity map in the 0-skeleton. Since the trace lands in the fixed set with respect to the spacewise S1S^{1}-action on the cyclotomic T​H​HTHH spectrum, it is compatible with the maps RR and FF (see e.g., [40, 1.2] for a more detailed discussion of this). Moreover, we can check that the trace descends to a natural transformation of localizing invariant using lemma 10.5.

0NR9

Lemma 10.9. The topological Dennis trace above induces a natural transformation of localizing invariants

K⟶T​Cn.K\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}TC^{n}.

Furthermore, the cyclotomic trace K→T​CK\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}TC provides a natural transformation in this setting which is assembled from natural transformations of localizing invariants.

0NRA

Lemma 10.10. The natural transformations of localizing invariants

K⟶T​CnK\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}TC^{n}

induce a natural transformation of spectrum-valued functors

K⟶T​Cn.K\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}TC^{n}.

Although T​C=holim⁡T​CnTC=\holim TC^{n} is not itself a localizing invariant (it does not preserve filtered colimits in general), any natural transformation of functors K→T​CK\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}TC is equivalent to the data of compatible maps to each T​CnTC^{n}. Therefore, if we consider the spectrum of natural transformations of functors to spectra from K→T​CK\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}TC which restrict to localizing invariants on each component, the spectrum is given by the limit (in the ∞\infty-category of spectra)

limnNat⁡(K⁡(−),T​Cn​(−)).\lim_{n}\Nat(K(-),TC^{n}(-)).

Finally, this yields the following characterization of the cyclotomic trace.

0NRB

Theorem 10.11. After pp-completion, the set of homotopy classes of compatible localizing invariants {K→TCn}\{K\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}TC^{n}\} is isomorphic to ℤp\mathbb{Z}_{p}. The cyclotomic trace is represented by 1∈ℤp1\in\mathbb{Z}_{p}.

0NRC

Proof. Using theorem 10.3 as in the proof of corollary 10.4, we see that the spectrum of natural transformations K→T​CK\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}TC which restrict to localizing invariants on each component can be computed as the limits (in the ∞\infty-category of spectra)

limnNat⁡(K⁡(−),T​Cn​(−))≃limnT​Cn​(𝕊)=T​C​(𝕊).\lim_{n}\Nat(K(-),TC^{n}(-))\simeq\lim_{n}TC^{n}(\mathbb{S})=TC(\mathbb{S}).

Completing at the prime pp, recall that T​C​(𝕊)≃𝕊∨Σ​C​P−1∞TC(\mathbb{S})\simeq\mathbb{S}\vee\Sigma CP^{\infty}_{-1} [68, §1]. Since π0​(Σ​C​P−1∞)=0\pi_{0}(\Sigma CP^{\infty}_{-1})=0, we deduce that the set of homotopy classes of compatible invariants is ℤp\mathbb{Z}_{p}. Furthermore, using the argument for theorem 10.6 and passing to the limit, we can identify the class of the cyclotomic trace by understanding the homotopy class of the composite

𝕊⟶K⁡(𝕊)⟶T​C​(𝕊)⟶T​H​H​(𝕊)≃𝕊\mathbb{S}\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}K(\mathbb{S})\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}TC(\mathbb{S})\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}THH(\mathbb{S})\simeq\mathbb{S}

(after pp-completion). An elaboration of Waldhausen’s results [86, §5] (see [17, §5] or [68, §1]) implies that this homotopy class is the identity (i.e., the unit splits the trace T​C​(𝕊)→T​H​H​(𝕊)≃𝕊TC(\mathbb{S})\mathchoice{\longrightarrow}{\rightarrow}{\rightarrow}{\rightarrow}THH(\mathbb{S})\simeq\mathbb{S}, which gives the identification of T​C​(𝕊)TC(\mathbb{S}) above), and so using the work of theorem 10.6 we again deduce that the cyclotomic trace is represented by the unit. ∎

Original mathematics by the credited authors. Source collection and HTML conversion remain in progress.

Andrew J. Blumberg, David Gepner, Goncalo Tabuada

Original source: arXiv:1001.2282v4