ON THE ANDRE´-QUILLEN HOMOLOGY OF TAMBARA FUNCTORS 7 1 MICHAELA.HILL 0 2 Abstract. We liftto equivariant algebra three closely related classical alge- n braic concepts: abelian group objects in augmented commutative algebras, a J derivations,andK¨ahlerdifferentials. WedefineMackeyfunctorobjectsinthe category of Tambara functors augmented to a fixed Tambara functor R, and 2 weshowthattheusualsquare-zeroextensiongivesanequivalenceofcategories 2 betweentheseMackeyfunctorobjectsandordinarymodulesoverR. Wethen describethenaturalgeneralizationtoTambarafunctorsofaderivation,build- ] T ingontheintuitionthataTambarafunctorhasproductstwistedbyarbitrary finite G-sets, and we connect this to square-zero extensions in the expected A way. Finally,weshowthatthereisanappropriateformofK¨ahlerdifferentials . which satisfy the classical relation that derivations out of R are the same as h mapsoutoftheK¨ahlerdifferentials. t a m [ 1 1. Introduction v 9 Foundationalwork of Andr´e andQuillen defined notions of homologyand coho- 1 mologyfor commutative rings [1], [10]. This provideda naturalwayto understand 2 the deformations ofa commutative ring,connecting them to derivations,providing 6 a condition for´etale-ness,and building a naturallong-exactsequence analogousto 0 thosefromtopologyfor atriple. Unpublished workofKrizlifted this to structured . 1 ring spectra, showing that certain Postnikov invariants can be recast as Andr´e- 0 Quillen cohomology groups [6]. Basterra extended this, producing the theory of 7 1 topological Andr´e-Quillen homology of a commutative ring spectrum [2]. This : work was then extended by Basterra-Mandell, who showed that TAQ with coef- v ficients is essentially the only homology theory on commutative ring spectra and i X who explored the basics of spectrum objects in commutative ring spectra [3]. r In the G-equivariant context for a finite group G, the role of abelian groups in a non-equivariant algebra is played by Mackey functors. The category of Mackey functors is a closed symmetric monoidal category with symmetric monoidal prod- uct, the box product. In addition to the expected generalization of commutative rings to simply commutative monoids for the box product, there is a poset of gen- eralizations of the notion of commutative rings to the G-equivariant context: the incompleteTambarafunctors[4]. TheseinterpolatebetweenGreenfunctors,theor- dinarycommutativemonoids forthe boxproduct, andTambarafunctors [12]. The distinguishing feature for [incomplete] Tambara functors is the presence of certain multiplicative transfer maps, called norm maps. For a Green functor, we have no norm maps; for a Tambara functor, we have norm maps for any pair of subgroups H ⊂K of G. Theauthor wassupportedbyNSFGrantDMS-1509652. 1 2 MICHAELA.HILL This paper explores three closely related themes from classical commutative algebra in the setting of Tambara functors: square-zero extensions, derivations, and K¨ahler differentials. Strickland initiated this study, showing that in stark contrast to the classical case, Quillen’s abelian group objects in Tambara functors over a fixed Tambara functor R properly contains the category of R-modules. In particular, the Andr´e-Quillen homology groups are in general more complicated than simply the derived functors of derivations into an R-module. In this paper, we explain how to rectify this situation, showing that the correct analogue of the abelian group objects is the Mackey functor objects: Theorem. Thesquare-zeroextensiongivesanequivalenceofcategoriesbetweenthe category of R-modules and the category of Mackey functor objects in the category of S-Tambara functors augmented to R. Classically, maps into a square-zero extension are classified by derivations, and withtheappropriatenotion,suchathingistruehere. Classically,aderivationturns sums to products. We define below (Definition 4.1) a “genuine derivation” which plays the equivariant role, converting twisted products (the norms) into twisted sums (the transfers). Theorem. The set of maps from an S-Tambara functor C augmented to R to a square-zero extension R⋉M is naturally isomorphic to the set of genuine S- derivations of C into M. Finally, there is an R-module of genuine K¨ahler differentials (Definition 5.4) which receives the universal genuin S-derivation from R. Theorem. ThereisanR-moduleΩ1,G andauniversalgenuineS-derivationd: R→ R/S Ω1,G . This has the property that genuine S-derivations from R to an R-module M R/S are in natural bijective correspondence with S-module maps Ω1,G →M. R/S Notational Conventions. In this paper, G will always denote a finite group. We will usually reserve the letters H and K for subgroups of G. Additionally, we will denote coefficient systems, Mackey functors, Tambara functors, and related constructions with underlined capital Roman letters to distinguish them from the non-equivariant objects. Acknowledgements. We thank Andrew Blumberg and Tyler Lawson for several helpful conversations and for their careful reading of earlier drafts. 2. Brief review of Tambara functors 2.1. Ordinary Tambara functors. Definition 2.1. LetPG denotethecategoryofpolynomialsinG-sets. Theobjects are finite G-sets, and the morphisms are isomorphism classes of diagrams f g h S ←−U −→V −→T, where two such diagrams are isomorphic if we have a commutative diagram g′ uu❦❦f❦′❦❦❦U′ //V′ ❚❚❚h❚′❚❚)) S ii❙❙❙❙❙❙∼= (cid:15)(cid:15) (cid:15)(cid:15)∼=❥❥❥❥❥❥55 T. f U //V h g EQUIVARIANT DERIVATIONS 3 Composition in this category is a bit trickier to describe, so it is convenient to name a generating collection of morphisms and then describe their commutation relations. Definition 2.2. Let f: S →T be a map of finite G-sets. Then let f = = R :=[T ←−S −→S −→S] f = f = N :=[S ←−S −→T −→T] f = = f T :=[S ←−S −→S −→T] f Then any polynomial can be written as a composite of these: f g h T ◦N ◦R =[S ←−U −→V −→T]. h g f These have the following relations. Proposition 2.3. R gives a contravariant functor from SetG into PG. N and T give covariant ones. Proposition 2.4. If we have a pullback diagram of finite G-sets S′ f′ // T′ g′ g (cid:15)(cid:15) (cid:15)(cid:15) S // T, f then we have Rg◦Nf =Nf′ ◦Rg′ and Rg◦Tf =Tf′ ◦Rg′. The interchange of N and T is trickier. Recall that if f: S → T is a map of finite G-sets, then the pullback functor f∗: SetG →SetG ↓T ↓S has a right adjoint: the dependent product . Qf Definition 2.5. AnexponentialdiagraminSetG isadiagram(isomorphictoone) of the form S oo g Aoo f′ S× A T Qh h g′ (cid:15)(cid:15) (cid:15)(cid:15) T oo A. h′ Qh Proposition 2.6. If we have an exponential diagram S oo h Aoo f′ S× A T Qg g g′ (cid:15)(cid:15) (cid:15)(cid:15) T oo A, h′ Qg then Ng◦Th =Th′ ◦Ng′ ◦Rf′. 4 MICHAELA.HILL With these morphisms, the disjoint union of finite G-sets becomes the product in the category PG. Definition 2.7. A semi-Tambara functor is a product preserving functor PG → Set. A Tambara functor is a semi-Tambara functor R for which R(T) is group- complete for all T ∈SetG. Tambarashowedthat the group-completionfunctor canbe applied to any semi- Tambara functor, giving a Tambara functor. There are several related categories of polynomials which give other flavors of Tambara functors. Recall that a subgraph of a category C is “wide” if it contains all of the objects. Definition 2.8. Inside the category PG are three important wide sub-graphs: (1) PG where the map g in a polynomial is an isomorphism, Iso (2) PG where the map g in a polynomial is an epimorphism, and Epi (3) PG where the map g in a polynomial preserves isotropy in the sense that gr for all u∈U, the stabilizer of g(u) is that of u. Proposition 2.9 ([4, Prop. 2.12]). The subgraphs PG , PG , and PG are subcat- Iso Epi gr egories of PG in which the disjoint union of finite G-sets is the product. Proposition 2.10 ([4, Prop. 4.3]). A product preserving functor PG →Set is a Iso semi-Mackey functor. Proposition 2.11 ([11, Prop. 12.11]). A product preserving functor PG →Set is gr a semi-Green functor. The category of Mackey functors is a closed symmetric monoidal category. The symmetric monoidal product is called the box product and is the Day convolution product of the tensor product of abelian groups with the Cartesian product of finite G-sets. Classically, a commutative Green functor is a commutative monoid under the box product. In particular, there is an obvious notion of the category of modules over a Green functor, and this is a symmetric monoidal category if the Green functor is commutative. Expandingout whatit meansto be a commutative monoidunder the box prod- uct, we see that a [commutative] Green functor is a Mackey functor R such that for allfinite G-sets T, R(T)is commutative ring,suchthat allrestrictionmaps are maps of commutative rings, and such that if f: T → T′ is a map of finite G-sets, then we have the Frobenius reciprocity relation a·T (b)=T (R (a)·b) f f f for all a∈R(T′) and b∈R(T). There is a similar description for Tambara functors. Proposition 2.12 ([7]). A Tambara functor is a commutative Green functor R together with norm maps NK: R(G/H)→R(G/K) H for all H ⊂ K ⊂ G. These are maps of multiplicative monoids and they satisfy certainuniversalformulaeexpressingthenormofasumandthenormofatransfer. The exact formulae for the norms of a transfer will not matter for us here; it suffices that such a formula exists. For a sum, we need slightly more information. EQUIVARIANT DERIVATIONS 5 Proposition 2.13 ([7, Thm. 2.3]). Consider the maps ∇: G/H ∐G/H → G/H and π: G/H →∗. Then we have an isomorphism of G-sets over ∗ ∇∼= F(G/H,{0,1})→∗ , Y (cid:16) (cid:17) π where {0,1}=(G/H ∐G/H)/G has a trivial action. The diagram G/H oo ∇ G/H∐G/H oo ǫ G/H ×F(G/H,{0,1}) π g (cid:15)(cid:15) (cid:15)(cid:15) ∗ oo F(G/H,{0,1}) h is an exponential diagram, where ǫ(gH,f):=(gH,f(gH))∈G/H ×{0,1}∼=G/H ∐G/H. Proposition 2.13 gives the formula for the norm of a sum of elements: NG(a+b)=T ◦N ◦R (a,b). H h g f Whendiscussingdifferentialsandthe universaldifferential,wewillneedtowork with non-unital Tambara functors. These can be defined simply from PG . Epi Definition 2.14. Anon-unital semi-Tambara functoris aproductpreserving functorPG →Set. Itisanon-unital Tambara functorifitisgroupcomplete. Epi Just as with ordinary Tambara functors, we can view a non-unital Tambara functor as a non-unital Green functor together with norm maps that satisfy the same universal formulae. 2.2. Relative Tambara functors. If S is a Tambara functor, then we can talk about Tambara functors and non-unital Tambara functors in the category of S- modules. Definition 2.15. If S is a Tambara functor, then an S-Tambara functor is a Tambara functor R together with a map S →R of Tambara functors. Let S-Tamb denote the corresponding comma category of Tambara functors equipped with a map from S. Definition2.16. Anon-unital S-Tambara functorisanS-moduleRequipped with norm maps for any surjection f: T →T′ that satisfies N (r·s)=N (r)·N (s) f f f for all s∈S(T) and r ∈R(T). Both of these have a more diagrammatic approach. Proposition2.17. LetS beaTambarafunctorandletRbea[non-unital]Tambara functor. Assume that R is a module over S, and let µ: S(cid:3)R→R be the action of S on R. Then R is a [non-unital] S-Tambara functor if and only if µ is a map of [non-unital] Tambara functors. 6 MICHAELA.HILL Remark 2.18. The category of modules over a Tambara functor S inherits a G- symmetric monoidal structure from the category of Mackey functors. The G- commutativemonoidshereareexactlytheS-Tambarafunctors,andthenon-unitial G-commutative monoids are exactly the non-unital S-Tambara functors. 3. Abelian group and Mackey functor objects We recallworkof Strickland(building onworkofQuillen) onthe homologyofa Tambara functor. Definition 3.1. Let R be an S-Tambara functor. LetS-Tamb be the comma categoryofS-Tambarafunctorswith amapto R. /R Let S-Ab denote the category of abelian group objects in S-Tamb . /R /R Let R-Mod denote the category of modules over the underlying Green functor for R in the category of Mackey functors. There is an obvious “augmentation ideal” functor I: S-Ab →R-Mod /R which assigns to an abelian group object B the kernel of B →R. In commutative rings, this functor is half of an equivalence of categories, with quasi-inverse given by the square-zero extension. Strickland shows that square-zero extensions make perfect sense here, but that these are not inverse equivalences. Proposition 3.2 ([11, Prop. 14.7]). There is a “square-zero extension functor” R⋉(−): R-Mod→S-Ab /R which sends an R-module to the square-zero extension in Green functors and which endows the module summand with trivial norms. These arenot inverseequivalences: themap R⋉(−)is notessentially surjective. Inthe square-zeroextension,the S-Tambarafunctor structureis induced bythe natural maps of Tambara functors S −→η R−I−d−⋉→0 R⋉M. The issue here is with norms in the augmentation ideal. The only condition we deduce from this being an abelian group object is that all products vanish. However, this only tells us about the restrictions of norms to various subgroups, not to the norms themselves. To better explain the failure of this equivalence and to prove the more accurate statement, we being with a simple observation. Proposition 3.3. If R and B are Tambara functors, then the set of Tambara functor maps between them has a natural extension to a coefficient system of sets: Tamb(R,B)(G/H)=TambH(i∗ R,i∗ B)⊂MackeyH(i∗ R,i∗ B). H H H H TherestrictionmapsonMackey functorsgiverisetotherestriction mapsinTamb. This provides an enrichment in coefficient systems for the category Tamb, where composition and the units are level-wise. The categories S-Tamb and S-Tamb are also enriched in coefficient systems /R and form a sub-coefficient system of Tamb. The following is an immediate application of the Yoneda Lemma. EQUIVARIANT DERIVATIONS 7 Proposition 3.4. An abelian group structure on B →R is the same as a natural lift of S-Tamb (−,B) to a coefficient system of abelian groups. /R The Yoneda Lemma also better explains the coefficient system structure here. The restriction functor i∗ from G-Tambara functors augmented over R to H- H Tambara functors augmented over i∗ R has a right adjoint: coinduction [11, Prop. H 18.3]. This has a very simple formulation: for any T ∈SetG, CoIndG(R)(T):=R(i∗ T). H H Similarly, if f: T →T′, then Tf :=Ti∗Hf Nf :=Ni∗Hf Rf :=Ri∗Hf. Since CoIndG is the right adjoint to i∗ , we have a natural map of Tambara H H functors η : R→CoIndGi∗ R. R H H This gives us the right adjoint to i∗ in the category S-Tamb: if R is an i∗ S- H H Tambara functor, then CoIndGR is an S-Tambara functor via the composite H S −η→S CoIndGi∗ S −C−o−I−nd−GH−→η CoIndGR. H H H We can also define a relative version of coinduction. Definition 3.5. If B −→f i∗ R is a Tambara functor over i∗ R, then let F (G,B) H H H be the pullback F (G,B) // CoIndGB H H FH(G,f) CoIndGHf (cid:15)(cid:15) (cid:15)(cid:15) R //CoIndGi∗ R. ηR H H Proposition 3.6. If B is an i∗ S-Tambara functor and R is an S-Tambara func- H tor, then the pullback of the structure maps gives F (G,B) the structure of an H S-Tambara functor. Proof. Consider the diagram S ηS //CoIndGi∗ SCoIndGHηB//CoIndGB ηR (cid:15)(cid:15) CoIndGHηi∗HR (cid:15)(cid:15)H Hww♣♣♣♣C♣♣o♣In♣d♣GH♣♣ǫ H R //CoIndGi∗ R. ηR H H The square commutes since η is a natural tranformation. The triangle commutes since B is an i∗ S-Tambara functor augmented to i∗ R. (cid:3) H H Proposition3.7. ThefunctorF (G,−)istheright-adjointtotherestrictionfunc- H tor i∗ in the category of Tambara functors augmented over R. H 8 MICHAELA.HILL The unit of the restriction-coinduction adjunction is induced by the natural commutative square B ηB//CoIndGi∗ B H H f CoIndGHi∗Hf (cid:15)(cid:15) (cid:15)(cid:15) R //CoIndGi∗ R. ηR H H The Yoneda Lemma now also describes the restriction maps in the coefficient system S-Tamb. Proposition 3.8. The restriction maps in S-Tamb (C,B) /R are induced by the natural maps η : B →F(G/H,B). B Tofullyunderstandthestructure,weextendthiscoefficientsystemintheobvious way to a product preserving functor S-Tamb (C,B): SetG,∐ op →Set. /R (cid:0) (cid:1) This part is also representable. Proposition3.9([4,Cor. 6.7]). IfB is a Tambarafunctor andT is afiniteG-set, then the Mackey functor B :=B(T ×−) T has a canonical Tambara functor structure. When T =G/H, we have a natural isomorphism B ∼=CoIndGi∗ B G/H H H Since the Cartesian product distributes over disjoint union, the following is im- mediate. Proposition 3.10. If B is a Tambara functor and T and T are finite G-sets, 1 2 then we have a natural isomorphism of Tambara functors B ∼=B ×B . T1∐T2 T1 T2 Combining this with the units of the restriction-coinduction adjunction then gives the following. Proposition 3.11. If B is a Tambara functor, then for any finite G-set T, there is a natural map of Tambara functors B →B . T In particular, if B is an S-Tambara functor, then B is canonically so for any T. T Using all of this we can define a version of this in the category of S-Tambara functors augmented to R. Definition 3.12. If B →R is an S-Tambara functor augmented to R and if T is a finite G-set, then let F(T,B) be the pullback F(T,B) // B T (cid:15)(cid:15) (cid:15)(cid:15) R //R . T EQUIVARIANT DERIVATIONS 9 Proposition 3.13. If B is an S-Tambara functor augmented to R and if T and 1 T are finite G-sets, then we have a natural isomorphism 2 F(T ×T ,B)∼=F T ,F(T ,B) . 1 2 2 1 (cid:0) (cid:1) Proof. Since the Cartesian product of finite G-sets is associative up to natural isomorphism, we have a natural isomorphism (B ) ∼=B . T1 T2 T1×T2 TheresultthenfollowsfromobservingthatbothTambarafunctorsarethepullback of the diagram B T1×T2 (cid:15)(cid:15) R //R . T1×T2 (cid:3) Having symmetric monoidal functors which act as symmetric monoidal powers indexedby a G-setis exactlyone ofthe waysto parse the notionofa G-symmetric monoidal category [5, Def. 3.3], so we conclude the following [5]. Theorem 3.14. With coinduction as categorical transfer maps, the category of Tambara functors augmented over R becomes a G-symmetric monoidal category. The internal tensoring with a finite G-set T is given by the functors F(T,−). This lets us reformulate Strickland’s definition. In some sense, this proposition has no real content: it is an immediate reformulation of Strickland’s result. Proposition3.15. The category S-Ab is thecategory of group-like commutative /R monoids in S-Tamb . /R Since Tamb is a G-symmetric monoidal category, we have a notion of G- /R commutative monoids [5, Def. 3.8]. Proposition 3.16. If B →R is a group-like G-commutative monoid in Tamb , /R then for all C →R, the coefficient system Tamb (C,B) /R has natural extension to a Mackey functor. Proof. Let C →R be a Tambara functor augmented to R, and let B :=S-Tamb (C,B) C /R be the coefficient system in question. By construction, the value of this at a finite G-set T is given by B (T):=S-Tamb C,F(T,B) . C /R(cid:0) (cid:1) In particular, Proposition 3.13 shows that we have a natural isomorphism of coef- ficient systems F(T,B) ∼=NT(B ), C C where NT is the endo-functor on coefficient systems of sets given by (NTM)(T′):=M(T ×T′). 10 MICHAELA.HILL Bynaturality,theG-commutativemonoidstructureofBmakesB aG-commutative C monoidin the coinduction G-symmetric monoidalstructure on coefficient systems. By [5, Thm. 5.6], this is exactly a Mackey functor structure on B . (cid:3) C Definition3.17. AMackeyfunctorobjectinTamb isagroup-likeG-commutative /R monoid in Tamb . The category of Mackey functor objects and maps is denoted /R Mackey . /R We can immediately produce a collection of such objects. Recall that a strong G-symmetricmonoidalfunctorbetweenG-symmetricmonoidalcategoriesisonefor which we have natural isomorphism F NT(−) ⇒NT F(−) . (cid:0) (cid:1) (cid:0) (cid:1) Proposition 3.18. The functor R⋉(−): R-Mod→S-Tamb /R is a strong G-symmetric monoidal functor. Proof. TheunderlyingMackeyfunctorsforCoIndG andforF(T,−)aredetermined H by the corresponding functors on Mackey functors. In this case, we have natural isomorphisms of Mackey functors augmented to R: F T,R⊕M ∼=R⊕M . (cid:0) (cid:1) T Inbothcases,theaugmentationidealhastrivialnormsandproducts,meaningthat this identification is also one of Tambara functors. (cid:3) Corollary 3.19. The functor R⋉(−): R-Mod→S-Tamb /R lifts to a functor to Mackey . /R Proof. Any Mackey functor is a group-like G-commutative monoid. A strong G- symmetric monoidal functor preserves these. (cid:3) We would like to better understand the category of Mackey functor objects augmentedtoR,andforthis,weunpacksometheexternalizedtransfermaps. Itis helpfultocomparethesewiththetransfermapsintheunderlyingMackeyfunctors. Lemma 3.20. Any Mackey functor has a unique structure as a G-commutative monoid. Proof. InMackeyfunctors,coinductionandinductionagree. Inparticular,CoIndG H is the left-adjoint to the forgetful functor as well as the right, and hence a map F(G/H,M)=CoIndGi∗ M −t−rGH→M H H is determined by its adjoint i∗ M →i∗ M. The adjoint can be computed as H H i∗ M →i∗ CoIndGi∗ M ∼=i∗ F(G/H,M)∼=F(i∗ G/H,i∗ M)−i−∗H−tr−GH→i∗ M, H H H H H H H H wherethe firstmapisthe unitofthe adjunction. Thiscorrespondstothe inclusion H/H ֒→i∗ G/H, and the composite is then just the identity map. Thus trG must H H be the adjoint to the identity mapon i∗ M, and hence is uniquely determined. (cid:3) H Corollary 3.21. IfB ∈Mackey , then all externaltransfer maps in B are maps /R of Tambara functors.