Rigidity and exotic models for $v_1$-local $G$-equivariant stable homotopy theory

We prove that the $v_1$-local $G$-equivariant stable homotopy category for $G$ a finite group has a unique $G$-equivariant model at $p=2$. This means that at the prime $2$ the homotopy theory of $G$-spectra up to fixed point equivalences on $K$-theory is uniquely determined by its triangulated homotopy category and basic Mackey structure. The result combines the rigidity result for $K$-local spectra of the second author with the equivariant rigidity result for $G$-spectra of the first author. Further, when the prime $p$ is at least $5$ and does not divide the order of $G$, we provide an algebraic exotic model as well as a $G$-equivariant exotic model for the $v_1$-local $G$-equivariant stable homotopy category, showing that for primes $p \ge 5$ equivariant rigidity fails in general.


Introduction
A basic idea in topology is looking at geometric objects up to homotopy and then ask the question to what extent the original objects can be rediscovered from this homotopy information. The language of model categories places this in a clean, formal environment: a model category is a category C with a notion of homotopy between morphisms. Thus, one can consider its homotopy category Ho(C), which, very roughly speaking, has the same objects as C but the morphisms are homotopy classes of the morphisms of C. So, just knowing Ho(C) as a category, to what extent can the model category C be recovered from this information?
To make this question more workable, one considers stable model categories. The homotopy category of a stable model category is triangulated. The triangulated structure comes from axiomatizing cofiber sequences and encodes secondary compositions such as Toda brackets. Schwede showed in [42] that all of the higher homotopy information of the stable B Constanze Roitzheim C.Roitzheim@kent.ac.uk Irakli Patchkoria irakli.patchkoria@abdn.ac.uk 1 model category of spectra Sp is contained in the triangulated structure of the stable homotopy category Ho(Sp) alone. More precisely, he showed that when given a stable model category C such that Ho(C) is triangulated equivalent to Ho(Sp), then C is Quillen equivalent to Sp. This phenomenon is referred to as rigidity. In other words the stable homotopy category Ho(Sp) is rigid.
The second author considered a K -local version of rigidity at the prime 2, showing that the higher homotopy information of K -local spectra L 1 Sp is entirely contained in the triangulated structure of their homotopy category Ho(L 1 Sp) [39]. The cases Ho(Sp) and Ho(L 1 Sp) both reduce to a core calculation involving the endomorphisms of a compact generator (in Schwede's case the stable homotopy groups of spheres π * (S 0 ), in the K -local case the stable homotopy groups of the K -local sphere π * (L 1 S 0 )), although the reduction techniques are rather different-Schwede uses the Adams filtration while the K -local proof exploits v 1 -periodicity. For further examples of rigidity see [10,25].
The K -local stable homotopy theory at odd primes exhibits a completely different feature. By results of Bousfield [7], Franke [14] and [35], it follows that for p ≥ 5 the K ( p) -local stable homotopy category is not rigid. More precisely, there is an algebraic model category C whose homotopy category is triangulated equivalent to Ho(L 1 Sp) but C is not Quillen equivalent to L 1 Sp. In other words C is an exotic model for the K ( p) -local stable homotopy category. A weaker result holds for p = 3. In this case the same happens except we do not know whether the equivalence Ho(C) ∼ Ho(L 1 Sp) is triangulated. For further examples of exotic models see [12,[34][35][36]41].
Another instance of rigidity shows up in equivariant stable homotopy theory. Let G be a finite group and Sp G denote the stable model category of orthogonal G-spectra in the sense of [32]. The first author obtained in [33] the following uniqueness result at the prime 2 which is referred to as G-equivariant rigidity: given a cofibrantly generated, proper, G-equivariant stable model category C (see Sect. 2) with a triangulated equivalence The proof of this theorem combines methods of [33,39] as well as requires some new mathematical input. It starts by constructing a Quillen adjunction Next, we compose the left derived functor − ∧ L X with the equivalence −1 obtaining an exact endofunctor We can then reduce the question to studying F on the compact generators L v 1 ∞ G/H + (H ≤ G) of Ho(L v 1 Sp G ), or more precisely, to showing that the map The proof combines a variety of interesting methods-classical calculations involving v 1 -periodicity and homotopy groups of the K -local sphere, mod-2 K -theory of BG, the model category techniques and tools from equivariant homotopy theory such as the geometric fixed points, tom Dieck splitting and double coset formula. The result provides new insight to understanding of the structure of Ho(Sp G ). Furthermore, it ventures towards studying Ho(Sp G ) in the context of chromatic homotopy theory.
In the last part of the paper we deal with the case of odd primes and show that rigidity can fail in general. More precisely, following Franke, we consider the category of twisted chain complexes of E(1) * E(1)-comodules. Our exotic algebraic model is the model category of twisted ( [1], 1)-chain complexes of Mackey functor objects in E(1) * E(1)-comodules, denoted by The main result of the cocluding part of this paper is that if p is an odd prime and p does not divide the order of the finite group G, then C ( [1]

,1) (E(1) * E(1)[M(G)]-Comod) is
Quillen equivalent to a cofibrantly generated, proper G-equivariant stable model category, thus showing that the analogous result to Theorem 1.1 does not hold if p ≥ 5. The case when p is an odd prime and the order of the finite group G is divisible by p remains open. This paper is organised as follows. We start by quickly recalling some notions and discuss notation. Then, in Sect. 3 we consider the v 1 -local model structure on G-spectra and discuss its properties. In particular, we indicate why some of the classical results in equivariant stable homotopy theory also hold in Ho(L v 1 Sp G ).
In Sect. 4 we set up the Quillen functor that is going to be part of the main theorem. In [33] it was shown that any cofibrantly generated, proper, G-equivariant stable model category can be replaced by a G-spectral one. For such a C and a cofibrant X ∈ C, there is a Quillen adjunction We discuss when this set-up factors over the localisation L v 1 Sp G . Section 5 provides the main calculation involving generators and relations of π * (L 1 S 0 ). We show that an exact functor compatible with the necessary G-equivariant structure sends ∞ G + ∧v 4 1 to an isomorphism. We also show that any exact functor with certain equivariance preoperties induces an isomorphism 9. This uses relations and Toda brackets in π * (L 1 S 0 ) as in [39]. In Sect. 6 we finish the proof that is a Quillen functor using cellular chains, Bredon homology and the equivariant Atiyah-Hirzebruch spectral sequence. This input is also new to the rigidity context and potentially has applications also in other rigidity problems.
In Sect. 7 we show that the previously constructed Quillen functor is actually a Quillen equivalence. This uses our calculation in Sect. 5 and some classical results in equivariant stable homotopy theory.
To put the rigidity result in context, in Sect. 8 we show that Ho(L v 1 Sp G ) has an algebraic model as well as an exotic G-equivariant stable model at primes p ≥ 5 with p |G|.

Background and conventions
We assume that the reader is familiar with the basic background related to stable model categories and (left) Bousfield localisation, for example see [6,21,22].
By L 1 = L K ( p) we denote Bousfield localisation with respect to p-local topological K -theory. The 1 in the notation refers to the fact that L K ( p) = L E (1) where E(1) is the Johnson-Wilson spectrum of chromatic level 1, i.e.
Throughout the paper, G is a finite group. We will use the stable model category Sp G of equivariant orthogonal G-spectra based on a complete universe constructed in [32]. This model category is enriched, tensored and cotensored over the model category of pointed G-spaces G-Top * . Moreover, for any finite dimensional orthogonal G-representation V , the functor is an equivalence of categories. Here S V is the representation sphere of V , i.e. the one-point compactification of V . In fact Sp G is an example of a G-equivariant stable model category, see e.g. [33,Definition 3.1.1]: recall that a model category C is a G-equivariant stable model category if it is enriched, tensored and cotensored over the category of pointed G-spaces G-Top * in a compatible way (i.e. the pushout-product axiom holds) and if the adjunction is a Quillen equivalence for any finite dimensional orthogonal G-representation V . (When we write adjunctions, the left adjoint is always the top arrow.) Next we also note that the model category C in Theorem 1.1 is assumed to be a cofibrantly generated G-Top * -model category in the sense of Definition 3.3.1 of [33].
We mention that all the known classical models for G-equivariant stable homtopy theory based on spaces (and not on simplicial sets) are cofibrantly generated, proper G-equivariant stable model categories. These include as already mentioned Sp G [32], the S-model structure (flat model structure, see [47,Theorem 2.3.27] and also [3]), the model category of S Gmodules [32,IV.2], the model category of G-equivariant continuous functors [5] and the model categories of G-equivariant topological symmetric spectra in the sense of [30] and [18]. We

v 1 -local G-spectra
In this section we are going to construct the v 1 -localisation of G-spectra and describe its properties. We will work at the prime 2 (omitted from notation), though most of the constructions here work at any prime (see Remark 3.8).
Throughout the paper let M denote the mod-2 Moore spectrum, which fits into the exact triangle in Ho(Sp) Here, incl is the inclusion of the bottom cell, and pinch is the map that "pinches" off the bottom cell so that only the top cell is left. As M is rationally trivial it possesses a v 1 -self map (see for example [24]) which induces an isomorphism in K -theory. Note that despite the notation, v 4 1 is not the fourth power of some existing map v 1 , it only alludes to the fact that the map has degree 8 and the element v 1 ∈ E(1) * has degree 2. By [6, Theorem 4.8] a spectrum X is K -local if and only if is an isomorphism for all n. Thus, we can think of K -localisation as "v 1 -localisation". We now translate this into the equivariant setting.

Definition 3.1
The v 1 -localisation of Sp G is defined as the left Bousfield localisation at the set of maps The notation for the resulting model category is L v 1 Sp G , and the morphisms in the homotopy category Ho( This localisation exists because Sp G is proper and cellular, and W is a set [21]. Using [21] we also see that this localisation is a cofibrantly generated, proper, G-equivariant stable model category (see [21,  We note the following immediate consequences. The notation (−) H used below will stand for both derived and point-set level categorical fixed points.
is an isomorphism for all n and all H ≤ G. By adjunction, this is equivalent to being an isomorphism for all n and all H ≤ G. By the properties of v 4 1 mentioned earlier, this is equivalent to the derived fixed points X H being K -local for all H ≤ G.
• The Quillen adjunction is also a Quillen adjunction H sends v 1 -local G-spectra to K -local spectra. (Note that in general, a Quillen adjunction is also a Quillen adjunction F : L S C − − → ←− D : G if and only if G sends fibrant objects in D to S-local objects in C, see [8]. We prove this explicitly in Lemma 4.1.)

Remark 3.2
When we work with model categories and say that an object X is local, then we will assume that it underlying fibrant. However, when working in homotopy categories we will ignore this condition. If we need to make X underlying fibrant, this can be always arranged by replacing it fibrantly in the original model category. Further, in the context of homotopy categories X H will always stand for derived fixed points, and when considering fixed points on the point-set level, then X H will denote the (non-derived) categorical fixed points.
The set { ∞ G/H + | H ≤ G} is a set of compact generators for Ho(Sp G ). We now describe the situation for Ho(L v 1 Sp G ).

Proposition 3.3
The localisation L v 1 is smashing. In particular, Proof The proof is based on results of [23]. We start by observing that an arbitrary coproduct of v 1 -local objects is v 1 -local. This follows from the fact that the domains and codomains of morphisms in W are compact in Ho(Sp G ). Reformulating, we see that L v 1 preserves arbitrary coproducts. Hence the localisation is smashing by [ We can now describe the v 1 -local G-sphere. Proof We have to show the following.
For the first point, let Z be a v 1 -local G-spectrum. This implies that Z G is a K -local spectrum. As S 0 −→ L 1 S 0 is a K -equivalence, this means that the induced map is an isomorphism. By adjunction, is an isomorphism. By Spanier-Whitehead duality (denoted by D), this is equivalent to being an isomorphism. As ε * is monoidal, this is the same as asking for to be an isomorphism. But From now on when working in Ho(L v 1 Sp G ) we will sometimes identify X with L v 1 X and X ∧ L v 1 S 0 G depending on context and convenience. Below we list some classical facts from equivariant stable homotopy theory. The references are for example [19,28,32,43]. Because of the good properties of v 1 -localisation these facts just pass to the v 1 -local G-equivariant stable homotopy theory.
Recall that G : Ho(Sp G ) → Ho(Sp) denotes the (derived) geometric fixed points functor (see e.g. [32, Section V.4]). We will use that G • ε * is isomorphic to the identity functor and G is monoidal. An immediate consequence of these properties and the previous result is the following:

Corollary 3.6 For any subgroup H of G, there is an isomorphism in Ho(Sp)
As a consequence of the latter and [20, Proposition 3.2], we see that the v 1 -localisation considered in this section is in fact a special case of the localisation considered in [20] Finally, for genuine fixed points we have:

Corollary 3.7 For any subgroup H of G, there is an isomophism in Ho(Sp)
Proof The proof uses Propositions 3.3 and 3.4. It also uses the isomorphism (Z ∧ ε * Y ) H ∼ = Z H ∧ Y in Ho(Sp) for any spectrum Y and an H -spectrum Z .
Next, we list the following facts: v 1 -local tom Dieck splitting. For a based G-CW complex X , the tom Dieck splitting says that Because of the compatibility of localisation with fixed points and because L 1 is smashing, this gives us a chain of isomorphisms in Ho(L 1 Sp)

Restriction and external transfer
Here we use Lemma 4.1. The Quillen adjunction is also a Quillen adjunction is a Quillen adunction. For this, we have to show that which is trivial by the double coset formula and because L v 1 X is a v 1 -local H -spectrum. v 1 -local Wirthmüller isomorphism The classical Wirthmüller isomorphism asserts that the canonical map is a weak equivalence. In other words, the left and right adjoints of the restriction on the homotopy category are isomorphic. This passes through the v 1 -local categories since any weak equivalence is in particular a v 1 -local equivalence.
v 1 -local double coset formula The classical double coset formula is the isomophism which also passes to the v 1 -local categories. Here c g is the conjugation map induces a Quillen adjunction It also induces a Quillen adjunction . We will often consider non-equivariant spectra as G-equivariant spectra via the inflation functor ε * : Ho(Sp) −→ Ho(Sp G ) and somtimes skip the symbol ε * . We also note that Corollary 3.6 implies that there is a v 1 -local geometric fixed points functor and a triangulated natural isomorphism Families and v 1 -localisation Let F be a family of subgroups of G. There is a stable model category of G-spectra Sp G,F based on the family F [32, Section IV.6] which comes with a Quillen adjunction Using Corollary 3.7 and Lemma 4.1 we see that this adjunction passes through v 1localisations.

Remark 3.8
The constructions and results of this section do not depend on the prime 2. An analogous construction can be performed at odd primes using 4 1 , and the same properties hold. We are using this in Sect. 8.

The Quillen functor
In order to obtain the Quillen equivalence of our main theorem, we first need a Quillen functor of which we will show in Sect. 7 that it is in fact an equivalence. The first author showed in [33, Proposition 3.1.2] that without loss of generality, our (proper, cofibrantly generated) G-equivariant stable model category C can be replaced by a Quillen equivalent G-spectral model category, i.e. by an Sp G -model category in the sense of [13]. So from now on, let us assume C to be a proper, cofibrantly generated Sp G -model category. This means that for any cofibrant X ∈ L v 1 Sp G , the G-spectral structure provides us with a Quillen pair The model category L v 1 Sp G has the same underlying category as Sp G but a different model structure, so it is not clear if is also a Quillen adjunction. For this we will need the following lemma. Proof It is clear that G sends acyclic fibrations to acyclic fibrations as N and L S N have the same cofibrations.
We still have to show that G sends fibrations to fibrations in L S N . So let g : X −→ Y be a fibration in M between fibrant objects. This implies that G(g) : G(X ) −→ G(Y ) is a fibration in N between L S N -fibrant objects. We can factor G(g) in L S N as acyclic cofibration followed by fibration As C and G(X ) are fibrant in L S N and i is an S-equivalence, i is also a weak equivalence in N . As it is also a cofibration, it is thus an acyclic cofibration in N . This implies that there is a lift H in the following diagram This means that there is a retract in In other words, G(g) is a retract of a fibration in L S N and thus a fibration itself, which is what we wanted to prove. By [13, A.2], a functor that preserves trivial fibrations as well as fibrations between fibrant objects is a right Quillen functor.

is a Quillen adjunction if and only if
Recall that we assume an equivalence of triangulated categories together with isomorphisms which are natural with respect to the restrictions, conjugations and transfers. From now on, let X be a fibrant and cofibrant replacement of (L v 1 S 0 G ) in C. We want to show the following.

Proposition 4.3 Let X be as above. Then the G-spectrum
We will perform some essential calculations in a separate section before proving this.

The calculation
This section deals with the computational part of our main argument. First, we need to recall the stable homotopy groups of the 2-local sphere and of the K -local sphere at the prime 2 in low dimensions and their generators.
The following is well-known to experts but we are not aware of a reference in literature, so we provide a proof.
We know from [1, Lemma 12.5 and proof of Theorem 12.13] that the composite is equal to 8σ . Using this and the Toda bracket we get a commutative diagram Here we used that the mod-2 Moore spectrum satisfies It remains to show that incl •ηε • pinch = 0 and incl •η 2 σ • pinch = 0. We will now show that given x ∈ π 8 S 0 (2) , then incl •ηx • pinch = 0. Indeed, since 2x = 0, using the cofiber sequence Before starting the calculations we recall the isomorphism The right hand side is generated by elements of the form α · g, where α ∈ π * (L 1 S 0 ) and g ∈ G. On the left hand side, this corresponds to the map that is α on the factor L 1 S 0 and multiplication by g on the factor ∞ G + . We also recall that there is an isomorphism and similar observations apply also here.
and F(g) = g for any g ∈ G. Then such that the following diagram commutes: Having this diagram in mind, we will once and for all identify Under these identifications F(1 ∧ incl) and F(1 ∧ pinch) correspond to 1 ∧ incl and 1 ∧ pinch, respectively. Below we will also denote for short 1 ∧ incl and 1 ∧ pinch just by incl and pinch.
The mod-2 Moore spectrum has the peculiar property that As η survives K -localisation the same identity also holds K -locally, and we also have 2I d L 1 M = 0 (see [39,Section 5.2]). We know that From the previous point we can conclude immediately that n g ν · g, m odd, n g even Because η 3 = 4ν inside π 3 S 0 and π 3 (L 1 S 0 ), we can use our calculation for F(η) together with the relation η 2 y 1 = 0 to see that Furthermore, for some integers m and n g . Hence From here we can conclude that m must be odd and n g even.
because all other summands are trivial as they are multiples of 16. Thus, We also know that for some integers k and l g . A similar argument to the previous point shows that k is odd and the l g are even.
Proof Denote ∞ G + ∧ v 4 1 by v 4 1 for short. We know that 2v 4 1 = incl •μ • pinch in Ho(Sp) (see Lemma 5.1) and hence also in Ho(L 1 Sp). So Now we turn to the case of an exact endofunctor. Then Proof This is a calculation very similar to the one in the proof of Lemma 5.2. Using exactly the same arguments as there we see that the following hold: We first show that F induces isomorphisms in degrees n = 2, . . . , 8. After this we will take care of y 0 and y 1 .
Since F(η 2 ) = η 2 , the case n = 2 immediately follows. Next, the formula for F(ν) implies that is represented by a matrix with odd determinant, i.e. an invertible matrix. This gives that is an isomorphism. Similarly the formula for F(σ ) tells us that is an isomorphism. Furthermore, since 2η = 0 and σ y 1 = 0, it follows from the formulas for F(η) and F(σ ) that F(ησ ) = ησ . Hence we conclude that F gives an isomophism in degree 8. Now we take care of dimensions n = 0 and n = 1.
F(y 0 ) = y 0 As y 0 is a torsion element, F(y 0 ) must also be torsion, so We would like to know what D is. We make use of the relation μy 0 = η 2 σ , so Substituting the results of our previous calculations, the left hand side is Because η 2 σ y 0 = μy 2 0 = 0, the last summand is zero. Hence we arrive at The last calculation is now simple as y 1 = ηy 0 . Thus, as y 1 y 0 = 0 and ηy 0 = y 1 .
We finally note that because F(g) = g, the element y 1 · g is in the image of F for any g. So η is also in the image of F, which makes F surjective and consequently bijective in degree 1.

Corollary 5.5 Let
∞ G + and F(g) = g for any g ∈ G. Then the map

is an isomorphism.
Proof Recall we have a cofiber sequence Using this, the previous proposition and the Five Lemma, one can see that the map under consideration is an isomorphism in degrees * = 0, . . . , 7. The other degrees follow from the 8-periodicity of v 4 1 , i.e. the fact that the morphism is an isomorphism in Ho(L v 1 Sp G ).

The Quillen functor (revisited)
This entire section is dedicated to the proof of Proposition 4.3, namely that for X cofibrant and X ∼ = (L v 1 S 0 G ), the G-spectrum R Hom(X , Y ) is v 1 -local for all Y . This will be an inductive argument via the subgroups H of G.
Let F denote the composite We will first explain how showing that is an isomorphism for all subgroups H of G implies the claim. Recall that we started with an equivalence , and that these isomorphisms are natural with respect to transfers, restrictions and conjugations. Let X denote a cofibrant replacement of (L v 1 S 0 G ) in C. As described earlier, we have to show that the adjunction − ∧ X : sends ∞ G/H + ∧ v 4 1 to an isomorphism. We will show this by induction on H . For convenience we choose the inverse −1 so that By our assumptions on , we see that F satisfies the conditions of Lemma 5.2. We showed the case H = {1} in Sect. 5, in other words we have already proved that F( ∞ G + ∧ v 4 1 ) is an isomorphism. This will serve as the basis for an induction on the size of the subgroup H . We will proceed to prove the claim inductively for nontrivial subgroups using isotropy separation.
Assume that H ≤ G and for all proper subgroups K of H the morphism F( ∞ G/K + ∧ v 4 1 ) is an isomorphism. We want to show that this assumption implies that is an isomorphism. Let P(H ) denote the family of proper subgroups in H and P(H |G) denote the family of subgroups in G which are proper subconjugates of H . Consider the full subcategory of Ho(Sp G ) of those objects Z such that the map is an isomorphism. Since F is exact and preserves infinite coproducts, this subcategory is localising, and hence the morphism is an equivalence. To complete the induction step we need to show that is a P(H |G)-equivalence. This follows from the next lemma:

be a cellular map of G-CW complexes that is a stable Fequivalence for some family F . Then the map
is an F -equivalence.

Proof
The key idea is that F preserves cellular chains. Let K denote the inflation ε * K of the 2-local equivariant K -theory spectrum K . Then a map A → B in Ho( Consider any U ∈ F . Since F commutes with infinite coproducts, the cellular filtrations of X and Y give us strongly convergent Atiyah-Hirzebruch type spectral sequences and Since the map f is cellular, it induces a map of the latter spectral sequences. As furthermore F( ∞ G/H + ) = L v 1 ∞ G/H + and F respects conjugations, restrictions and transfers, we can see that there is a commutative diagram for X (we omit the ∞ symbol from notation below) and similarly for Y . For the same reason, we also have a commutative diagram where the isomorphisms are the same as in the previous diagram. Roughly speaking, these facts imply that F preserves cellular chain complexes. Now this allows us to conclude that f induces an isomorphism on E 2 terms. Indeed, the first commutative diagram implies that there are isomorphisms where H U p (−, π − q K) denotes Bredon homology. On the other hand the second commutative diagram gives us a commutative diagram But the lower horizontal map is an isomoprhism since U is in F and f is a stable Fequivalence. This finishes the proof.

Now we go back to our induction to show that F(
is an isomorphism. Lemma 6.1 implies that the morphism is a P(H |G)-equivalence. Next, the commutative diagram shows that the map In order to show that it is an isomorphism, it suffices to show that the induced map on H -geometric fixed points ) is an isomorphism. For this consider the composite (N G (H ) denotes the normalizer of H in G) which we denote byF When restricted to free W G (H )-spectra,F satisfies the conditions of Lemma 5.2, implying thatF is an isomorphism. But by unraveling definitions we see that the latter map is equal to H (F( ∞ G/H + ∧v 4 1 )) which finishes the proof of the fact that the spectrum R Hom(X , Y ) is v 1 -local.
We conclude this section by the following corollary of Lemma 6.1 which will be used below:

The Quillen equivalence
We showed in Sect. 6 that the Quillen adjunction from Sect. 4 is also a Quillen adjunction. Moreover the diagram of left Quillen functors commutes on the nose. In this section we show that our adjunction is actually a Quillen equivalence. Define the exact functor F : By the commutativity of the latter diagram we see that the diagram commutes up to a natural isomorphism. Our assumptions on imply that we have for any subgroup H of G and F respects conjugations, restrictions and transfers.

with the upper horizontal arrow being an isomorphism.
Proof This follows from Corollary 6.2 and the commutative diagram

Reduction to mod-2
In order to prove the main theorem, it suffices to show that the functor F is an equivalence which amounts to showing that the map The techniques from [33] cannot be directly applied here since the source and target are not finite when * = −2 and * = 0. However after mod-2 reduction they become finite, which will be proved below. Before that we will show that the integral result follows from the combination of the rational result with the mod-2 result.

The map
is isomorphic to the map induced on homotopy groups by the morphism of mapping spectra (here f ib denotes a fibrant replacement) Since the source and target of this morphism of spectra are 2-local, it suffices to check that the morphism is a rational equivalence and induces isomorphisms on mod-2 homotopy groups. We know that the functor F preserve transfers, conjugations and restrictions between cosets. This implies that the latter map is a rational equivalence.
Checking that the latter map induces isomorphism on mod-2 homotopy groups amounts to checking that the map is an isomorphism. But this is the case if and only if the map is an isomorphism. Thus in order to prove the main theorem, it suffices to show that the last map is an isomorphism. Since the source and target are abstractly isomorphic. In fact the following lemma shows that they are finite.

Lemma 7.2 For any integer l ∈ Z, the group
Proof By duality we have an isomorphism Now Corollary 3.7 and the localised double coset formula, Wirthmüller isomorphism and tom Dieck splitting (see Sect. 3) give us an isomorphism Hence it suffices to show that for any finite group , the homotopy group is finite. The results of Adams-Baird and Ravenel (see e.g. [6,Corollary 4.4]) imply that this will follow if we show that is finite for any l ∈ Z. Recall that where K (1) is the mod-2 complex K -theory, i.e. the first Morava K -theory at prime 2. Ravenel has shown in [37] that K (1) l (B + ) is finite for any integer l. This together with η 3 = 0 in π * K O, implies that there is a finite filtration of K O l (B + ∧ M) with finite subquotients. Hence K O l (B + ∧ M) is finite which proves our claim.

Reduction to the case H = K
After the work of the previous subsection we can now follow [33, Section 5] and state certain results which immediately follow from the methods of [33]. We do not provide the explicit proofs as they are verbatim translations of the proofs contained in [33,Section 5].
Let H and K be subgroups of G. For the rest of this subsection we fix once and for all a set {g} of double coset representatives for K \G/H . Recall that for any g ∈ G, the conjugated subgroup g Hg −1 is denoted by g H . Further let stand for the map which is defined by the following commutative diagram The map is a split monomorphism. This follows exactly as [33, Proposition 5.1.1]. Now consider the commutative diagram (keeping F( Here κ F g is defined as κ g after applying F to the appropriate morphisms. Since the upper horizontal map is injection, it follows that injectivity of the right map implies injectivity of the left map. By Lemma 7.2, the source and target of the left hand map are finite of the same order. The same holds for the right hand map. Hence in order to prove the main theorem it suffices to restrict our attention to the mod-2 endomorphisms and check that the map is an isomorphism for any subgroup H of G.

Proof of the main theorem
Before completing the proof of the main theorem, we state the following proposition which is an analog of Proposition 6.3.2 in [33]. The proof follows from localising the classical isotropy separation sequence and is completely analogous to the one in [33]. We do not repeat the arguments here.

Proposition 7.3 There is a short exact sequence
Now we obtain the last ingredients of our main proof analogously to [33]. We follow the strategy of [33,Section 7.3]. Recall that what we still need to show is that is an isomorphism for any subgroup H of G. We perform an induction based on the size of H . The case H = {1} was done in Sect. 5. Consider the following compositê ThisF satisfies the conditions of Sect. 5 when restricted to free W G (H )-spectra. Hence the map (we omit the ∞ symbol) is an isomorphism. Finally, we look at the following important commutative diagram Corollary 6.2 implies that the lower left square commutes and that the lower left vertical map is an isomorphism. Other squares commute by definition. Further, according to Proposition 7.3, the lower row in this diagram is a short exact sequence and hence so is the middle one. We also note that the upper left vertical map is an isomorphism by the induction assumption and the results of Sect. 7.2. Now a diagram chase as in Section 7.3 of [33] shows that the map is surjective. Since the source and target of this map are finite of the same cardinality, this completes the proof that the map F above is an isomorphism. By the results of Sections 7.1 and 7.2, this implies that is also an isomorphism for any H , K ≤ G. Thus is an equivalence of categories, which proves our main rigidity theorem.

Exotic models at odd primes
Now let p be an odd prime and G a finite group such that p does not divide the order of G.
In this section we show that under these assumptions the triangulated category Ho(L v 1 Sp G ) has an exotic model. This exotic model is an equivariant generalisation of Franke's model [14] and, like Franke's model, relies also on the ideas of Bousfield [7]. Briefly, Franke's theorem shows the following: given a cohomological functor from a triangulated category (such as the homotopy category of a stable model category) into an abelian category, then under certain conditions this triangulated category can be realised as the derived category of twisted chain complexes. An important part of the assumptions is a relation between a "split" of the abelian category into smaller categories and the global injective dimension of this abelian category. We do not recall the conditions here and refer the reader to the original paper [14] for details. For more streamlined expositions see [40] or [35]. We also note that the proof in [14] contains a gap which is fixed for height one in [35,Section 4], see also [36] for a different approach and higher chromatic levels.

The algebraic model
We will describe the abelian category that we are dealing with and study its global injective dimension. The spectrum E(1) defines a homology theory Here E(1) * E(1)-Comod denotes the category of comodules over the flat Adams Hopf algebroid This category is an abelian category with enough injectives. Furthermore, it has global cohomological dimension equal to two [7].
In the equivariant context we need to consider Mackey functors in comodules. Let B(G) denote the Burnside category of G, see e.g. [28,Section V.9]. The category B(G) is a (pre)additive category. The objects of B(G) are the cosets of the form G/H for any subgroup H ≤ G. Morphisms are additively generated by the restrictions, conjugations and transfers.
A contravariant additive functor from B(G) to the category of abelian groups is called a Mackey functor. In this subsection we consider the category of contravariant additive functors from B(G) to the category of E(1) * E(1)-comodules, i.e. Mackey functors in comodules. We will denote this category by and refer to its objects as comodule Mackey functors. Since equivariant stable homotopy groups carry the structure of a Mackey functor [28,Section V.9], it follows that we have a homology theory The following fact is well-known and a proof can be found in [ is an equivalence of categories.
We get immediate corollaries:

Corollary 8.3 Let p be an odd prime and G a finite group such that p does not divide the order of G. Then the global cohomological dimension of the abelian category E(1) * E(1)[M(G)]-Comod is equal to two
Proof By the previous corollary it suffices to show that the cohomological dimension of E(1) * E(1)[W G (H )]-Comod is equal to two. Bousfield's result [7] tells us that the cohomological dimension of E(1) * E(1)-Comod is two. Since the order of the group W G (H ) is invertible in E(1) * E(1)-Comod, a proposition of Mitchell [31,Proposition 3.4] completes the proof.

Exotic equivalence
In this subsection we apply Franke's theorem to construct an algebraic model for L v 1 Sp G for p odd and finite group G such that p does not divide the order of G. We will check that the conditions of Franke's theorem [14] are satisfied. Franke's theorem makes use of "Eilenberg-MacLane objects" for injective objects in order to set up an Adams spectral sequence, and we are going to construct those here.

Lemma 8.4 Let J be an injective E(1) * E(1)-comodule. Then for any finite group K , the
Proof The functor is right adjoint (and also left adjoint) to the forgetful functor Additionally, the forgetful functor is exact. This implies the desired result.
Next, we show that certain injective objects in E(1) * E(1)[M(G)]-Comod are realisable. This implies that for any subgroup H ≤ G,

Proposition 8.5 Let p be an odd prime, G a finite group and suppose that p does not divide the order of G. Let I be an injective object in E(1) * E(1)[M(G)]-Comod which under the equivalence of the previous subsection corresponds to
Then by [44,Proposition III.4.28], the diagram of categories is an isomorphism for any v 1 -local spectrum Z (here Hom(−, −) * stands for graded homomorphisms). As in the previous proposition define The proof of the previous proposition implies Part (i). It remains to prove Part (ii).
We start by observing that by Proposition 3.3, it suffices to show that the map Hence in order to complete the proof, it suffices to show that the lower horizontal map is an isomorphism in this diagram. Now properties of geometric fixed points, the double coset formula and the Wirthmüller isomorphism imply that the derived fixed points I H are equivalent to We remind the reader that a twisted ( [1], 1)-chain complex is a pair (C, d), where C is a comodule Mackey functor and d : C → C [1] is a morphism of comodule Mackey functors which is a differential, i.e. d [1] • d = 0. Here [1] denotes the shift of the grading. For the model structure on the category of twisted chain complexes for general abelian categories see [9,14]. Let The equivalence is triangulated if p ≥ 5.
This theorem provides an algebraic model for Ho(L v 1 Sp G ) when p is odd and does not divide the order of G. Moreover, this equivalence does not come from a zig-zag of Quillen equivalences since the homotopy types of mapping spaces in L v 1 Sp G are not products of Eilenberg-MacLane spaces in general. Hence the model category provides an algebraic exotic model for L v 1 Sp G .

A G-equivariant stable exotic model
The exotic model constructed in the previous subsection is algebraic but it is not G-equivariant in the sense of Sect. 2. In this subsection we explain why it is Quillen equivalent to a Gequivariant stable model category. This model will then provide a counterexample which shows that if p is odd and does not divide the order of G, then the p-local version of Theorem 1.1 does not hold.
It follows from Morita theory [46] and [11,Proposition 6.3] that there is an endomorphism differential graded algebra B such that the model category C ( Let H B be the associated (cofibrant) Eilenberg-MacLane ring spectrum of B (see [45] and [29,Theorem 7.1.2.13]). Now we will describe B[M(G)]-Mod in terms of H B and thus provide the desired exotic G-equivariant stable model. By replacing H B sufficiently cofibrantly and using the fact that the inflation functor ε * : Sp −→ Sp G is a symmetric monoidal left Quillen functor, we see that there is a cofibrant G-equivariant ring spectrum ε * H B such that its underlying orthogonal ring spectrum is H B. The exotic G-equivariant stable model we are looking for is the model category ε * H B-Mod [32,Section III.7].
In what follows we will always assume that p is an odd prime, G a finite group and p does not divide the order of G. Proof Both sides are homology theories in X . Thus we only have to check that the map is an isomorphism for the sphere spectrum S 0 ( p) . This follows from the fact that p does not divide the order of G. Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.