On the equivariant KU_G-local sphere for finite abelian groups

Since every S\mathbb{Z}_{(p)}-acyclic G-spectrum is S/p-acyclic, we have L_{S/p}L_{S\mathbb{Z}_{(p)}}\simeq L_{S/p}. If p\nmid |H|, we have

Here \Phi^H L_{S\mathbb{Z}_{(p)}}X\simeq L_{S\mathbb{Z}_{(p)}}\Phi^H X since L_{S\mathbb{Z}_{(p)}} is smashing.

The case X=S_G is treated in [bonventre2022ku_g, Lemma 6.2], and the same proof applies to general finite G-spectra. For any R/p-acyclic G-spectrum M, M \wedge R is S/p-acyclic, hence

The case where G is a finite abelian group is proved by Strickland [strickland2011multicurves, Theorem 10.3], and the same argument applies to an arbitrary finite group, using the fact that for X,Y \in \mathrm{Sp}^G,

By [hill2019equivariant, Proposition 3.2], it suffices to show that for any subgroup H\subset G, \langle \Phi^H KU_G/p\rangle = \langle \Phi^H (ECyc_+\wedge \operatorname{Inf}_{N_p}^G KU_{N_p}/p)\rangle. By Proposition 2.1 we have

On the other hand, since G is nilpotent, there is a subgroup N\subset G such that G\cong N_p\times N. For any subgroup H\subset G, we have the following diagram of groups

Therefore,

If H\cap N_p\neq \{e\}, \Phi^{H\cap N_p} KU_{H\cap N_p}/p\simeq \ast since p is invertible in \Phi^{H\cap N_p}KU_{H\cap N_p}/p. The group homomorphism H\cong (H\cap N)\times (H\cap N_p)\to H\cap N_p induces a map of ring spectra

thus \Phi^H\operatorname{Inf}_{H\cap N_p}^H KU_{H\cap N_p}/p\simeq \ast.

If H\cap N_p=\{e\}, then \Phi^H\operatorname{Inf}_{N_p}^G KU_{N_p}/p\simeq \Phi^H\operatorname{Inf}_e^H KU/p\simeq KU/p, and

Note that for the Sylow p-subgroup N_p, H\cap N_p=e if and only if p\nmid |H|. Therefore, \langle KU_G/p\rangle=\langle ECyc_+\wedge \operatorname{Inf}_{N_p}^G KU_{N_p}/p \rangle.

The proof follows the proof of the nonequivariant case in [Ravenel1984LocalizationWR, Theorem 8.4]. The equivariant Wood theorem [mathew2017nilpotence, Theorem 9.8] says that there is a cofibration

where \eta\in \pi_1(S) is the Hopf element. Then KO_G\wedge X\simeq \ast implies KU_G\wedge X\simeq \ast. Conversely, if KU_G\wedge X\simeq \ast, then multiplication by \eta induces an isomorphism on (KO_G)_*(X). Since \eta is nilpotent, it follows that (KO_G)_*(X)=0.

Apply the geometric fixed point functor \Phi^H, we have a cofibration

By the same argument in Lemma 3.3, since \eta is nilpotent, \Phi^H KU_G/p\simeq \ast if and only if \Phi^H KO_G/p\simeq \ast.

Therefore,

Since \langle KU/p\rangle=\langle KO/p \rangle, we have \langle KU_G/p\rangle=\langle ECyc_+\wedge \operatorname{Inf}_{N_p}^G KO_{N_p}/p \rangle.

Let I=ECyc_+\wedge \operatorname{Inf}_{N_p}^G KO_{N_p}. We first claim that \psi^g:I_p^\wedge \to I_p^\wedge is a map of ring spectra and let F_G be the fiber of ((\psi^g - 1)\wedge X)_p^\wedge. Then by Proposition 2.4, (I\wedge X)_p^{\wedge} is I/p-local; hence it is KU_G/p-local by Lemma 3.3. It follows that F_G is KU_G/p local.

Note that

is trivial. Smashing with X and p-completing, we see that the composite

is also trivial, which induces a map \iota: X_p^\wedge \to F_G. It suffices to show that \iota is a I/p-equivalence, or equivalently,

is an equivalence for every subgroup H\subset G. By Proposition 2.1 and Lemma 3.4, \Phi^H \operatorname{Inf}_{N_p}^G KO_{N_p}/p\simeq \ast if H is not cyclic or H\cap N_p is nontrivial, and f_H is an equivalence between trivial spectra. If H is cyclic and H\cap N_p=\{e\}, then we have

By Proposition 2.2,

Here the first equivalence follows from the fact that \Phi^H preserves fiber sequences and smash products.

It remains to prove the claim. For any prime q\neq p, let N_q denote the Sylow q-subgroup of G, and set N:=\prod_{q\neq p}N_q. Let Cyc^{N} denote the family of cyclic subgroups of N, and let ECyc^N be the associated N-space. Then G\cong N\times N_p, and

Since p\nmid |N|, the N-spectrum (ECyc^N_+)_p^\wedge is a direct summand of (S_N)_p^\wedge. In particular, it is an N-equivariant commutative ring spectrum.

Recall that for any k\in\mathbb{Z}, there is an Adams operation \psi^k on equivariant K-theory of a G-space defined in [atiyah1969group]. Hirata-Kono [hirata1982bott, Theorem3.1] shows that the Adams operation \psi^k induces a stable operation after inverting k if and only if (k,|G|)=1. Then for a chosen generator g\in \mathbb Z_p^\times, the Adams operation induces a map of ring spectra

for any p-group N_p and k\in \mathbb{Z}_p^\times/\{\pm 1\}, and

is a map of ring spectra since (\operatorname{Inf}_N^G ECyc^N_+\wedge\ \operatorname{Inf}_{N_p}^G -)_p^{\wedge} is a monoidal functor.

Since \operatorname{Inf}_{N_p}^G KO_{N_p}/p\simeq \operatorname{Inf}_{N_p}^G (KO_{N_p})_p^\wedge/p, there is an equivalence of G-spectra

So we can rewrite the fiber sequence in Proposition 3.5 by applying the functor (ECyc_+\wedge \operatorname{Inf}_{N_p}^G(-))_p^\wedge to the fiber sequence

Since (|H|,|K|)=1, for any subgroup N\subset G, the Weyl group satisfies W_G N\cong W_K L_1\times W_H L_2, where L_1=N\cap K and L_2=N\cap H. By tom-Dieck splitting,

There is a canonical inclusion

Let

be the map of H-spectra adjoint to g. We can show that f is an H-equivalence. Indeed, since S_H=\operatorname{Inf}_e^H \Sigma S^0, the H-fixed point f^H is precisely the equivalence

and the same argument shows that f^L\in \mathrm{Sp} is an equivalence for every subgroup L\subset H.

For any K\subset G, let Cyc^K be the family of all cyclic subgroups of K, then we have

Since p\nmid |N|, for any cyclic subgroup P_0\subset N_p and L_0\subset N, P_0\times L_0 is also a cyclic subgroup. By the definition of ECyc, we have

Since H=P\times L, there is an equivalence of spectra X^H\simeq (X^L)^P for any G-spectrum X.

Consider the L fixed point of ECyc_+\wedge \operatorname{Inf}_{N_p}^G E as a P-spectrum, we have

By Lemma 4.1,

thus

where the last equivalence follows from the formula

By tom-Dieck splitting formula,

If T\subset L is not cyclic, (ECyc^L_+)^T is W_L T-equivariant contractible, and

If T\subset L is cyclic, then for any subgroup K \subset W_L T,

which implies that there is an equivalence of W_L T-spaces EW_L T_+\wedge (ECyc_+^L)^T\simeq EW_L T_+, so

Therefore, we have (\Sigma_L^\infty ECyc_+^L)^L\simeq \bigvee_{(T)\in Cyc^L/L}BW_L T_+, and

Let Z be an H\times K-orbit. Then Z = (H \times K)/U for some subgroup U \subset H \times K. Let P and Q denote the images of U under the projections onto H and K, respectively. Since |H| and |K| are coprime, |U| = |P| \times |Q|, which implies U \cong P \times Q and Z \cong H/P \times K/Q. Since H acts trivially on the K-orbits and K acts trivially on the H-orbits, for P_i\subset H and Q_i\le K, i=1,2, there is a natural bijection

Let \mathcal B_{H\times K} be the H\times K-Burnside category. A transitive span

has U\cong H/R\times K/S for some R\subset H and S\subset K, and therefore determines a pair of spans

Conversely, every such pair determines a transitive H\times K-span. Since an arbitrary finite H\times K-set is a finite disjoint union of orbits, this gives a natural isomorphism

which is compatible with composition. Consequently, the assignment

defines an additive functor on the full subcategory of \mathcal B_{H\times K} spanned by the transitive H\times K-sets. Since every finite H\times K-set is a finite disjoint union of transitive H\times K-sets, it extends uniquely by additivity to a H\times K-Mackey functor.

For any subgroup H\subset G, let P=H\cap N_p and L=H\cap N, then H=P\times L. By Lemma 4.2, there is an equivalence

After p-completion, (BW_L T)_p^\wedge\simeq \ast since p\nmid |L|. Moreover, since E/p\simeq E/p\wedge ECyc^{N_p}_+, we have E_p^\wedge\simeq (E\wedge ECyc^P_+)_p^\wedge. Thus

and

We also need to determine the restriction and transfer homomorphism. For any P_1\subset P_2\subset N_p, without loss of generality, we can assume that N=\{e\}, then \operatorname{Res}_{P_1}^{P_2} and \operatorname{Tr}_{P_1}^{P_2} are inherited from those in \underline{\pi}_* E. For N_1\subset N_2\subset N, assume that N_p=\{e\}. Since p\nmid |N|, it follows by [bonventre2022ku_g, Corollary 7.12] that the restriction and transfer maps are induced by those in (\underline{A/J}_N)_p^\wedge Thus

In particular, when E=L_{KU_{N_p}/p}S_{N_p}, Corollary 3.7 implies that L_{KU_{N_p}/p}S_{N_p}\simeq (ECyc^{N_p}_+\wedge L_{KU_{N_p}/p}S_{N_p})_p^\wedge, and hence L_{KU_{N_p}/p}S_{N_p} satisfies the assumption of this proposition. Therefore,

The first statement follows from the facts that \psi^g acts on \underline{RU} as a homomorphism of N_2-Green functors [carawan2023homotopy, Lemma 3.2] and that \psi^g commutes with the complexification map c\colon \underline{RO}\to \underline{RU}.

For any K\subset N_2, every generator \tau\in RO(K;\mathbb{R}) is 1-dimensional since K is abelian, which is a sign representation, thus \psi^g(\tau)=\tau^g=\tau.

For any irreducible x\in RO(K;\mathbb{C}), let X\in RU(K) be the irreducible complex K-representation whose underlying real representation is x, then \psi^g(x) is the underlying real representation of \psi^g(X), which is irreducible by [carawan2023homotopy, Lemma 3.1].

Since \psi^g is a homomorphism of Green functors, the restriction and transfer homomorphisms are inherited from those in \underline{\pi}_* (KO_{N_2})_2^\wedge, and it suffices to compute for any orbit N_2/K.

(1) and (2): When k=0, \underline{\ker}_2\{0\}=(\underline{RO}^{\psi^g})_2^\wedge. For any V\in KU(K) such that \psi^g V=V, by [bonventre2022ku_g, Proposition 6.7], the character \chi_V takes values in \mathbb{Q}, i.e., V is in the image of R\mathbb{Q}(K)\to RU(K), which factors through RO(K). Then

where

By [serre1977linear, Proposition 35], the Schur indices for N_2 equal to 1 since N_2 is abelian. So \underline{R\mathbb{Q}_\chi}\cong \underline{R\mathbb{Q}}, and \underline{\ker}_2\{0\}\cong \underline{R\mathbb{Q}}_2^\wedge.

By Lemma 5.1, we can decompose C_K into orbits under \psi^g. Since irreducible rational K-representations are in bijection with the cyclic subgroups of K, the orbits in C_K are in bijection with \{ C_{2^t}\subset K:t\geq 2\}, and elements in M_K are in bijection with the subgroups C_{2^t}\subset K for t=0,1. Let [C_{2^t}\subset K]\subset C_K denote the orbit corresponds to the subgroup C_{2^t}\subset K. Since the degree of the 2^t-th cyclotomic polynomial

is 2^{t-1}, the orbit [C_{2^t}\subset K] has 2^{t-2} elements.

On every orbit [C_{2^t}\subset K] with t\geq 2, we can choose a basis such that \psi^g acts as the matrix

For k=8d, \psi^g-1 acts on RO(K)_2^\wedge\{u^d\} via g^{4d}M-I on every orbit, which are equivalent to diagonal matrices

using a combination of row and column operations. Thus \psi^g-1 is injective when d\neq 0.

When d=0, every orbit contributes a summand of \mathbb{Z}_2^\wedge in cokernel, so

When d\neq 0, every orbit [C_{2^t}\subset K] contributes a summand of \mathbb{Z}_2^\wedge/g^{4d}-1 in cokernel for t\leq 1, and a summand of \mathbb{Z}_2^\wedge/g^{2^td}-1 in cokernel for t\geq 2. Since g is a generator of (\mathbb{Z}_2^\wedge)^\times/\{\pm 1\}\cong 1+4\mathbb{Z}_2^\wedge, then 1-g^{2^td} is a generator of 2^{2+t+\nu_2(d)}\mathbb{Z}_2^\wedge, thus

and

(3) For k=8d+1, \pi_{8d+1} KO_{N_2}\cong RO(N_2;\mathbb{R})\{\eta u^d\}/2, thus \psi^g acts trivially on \underline{\pi}_{8d+1}KO_{N_2}.

Finally, (4)-(6) follow from the same computation as above.

Since \underline{A}_{N_2} is a representable N_2-Mackey functor, there is no nontrivial extension of \underline{A}_{N_2} by \underline{RO(-;\mathbb{R})}_{N_2}\{\eta\}/2. Consider the following commutative diagram with exact rows:

Here h is the Hurewicz map, \mathcal{R}_{N_2} is the linearization map, and

\Phi is a morphism of Mackey functors since both h and i are. It is surjective by the five lemma. If \Phi(j,c)=0, then applying \pi gives

so j\in (\underline{J}_{N_2})_2^\wedge. By definition of \theta_{N_2},

Since i is injective, c=-\theta_{N_2}(j). Thus

Since all maps involved are Mackey functor maps, the displayed kernel is a sub-Mackey functor and the quotient is a Mackey functor quotient.

Note that \operatorname{Res}^V_H X_V=0 for every proper subgroup H \subset V. Hence \theta_V(X_V) restricts to zero on all proper subgroups. The common kernel of the restriction maps

is generated by \rho_V. Thus \theta_V(X_V) is either 0 or \eta\rho_{V}. Szymik [szymik2013chromatic, Example 5.1] shows that X_V has nontrivial Hurewicz image, so \theta_V(X_V)=\eta\rho_V.

The map \theta is natural for transfer and inflations. Therefore

As a real N_2-representation, \operatorname{Tr}_K^{N_2}\operatorname{Inf}_{K/L}^{K}(\rho_{K/L}) is isomorphic to \mathbb{R}\{N_2/L\}. Its image in RO(N_2;\mathbb{R})/2 is the sum of all real one-dimensional characters of N_2 that are trivial on L, which gives the result.

If a relation j\in J(N_2) is written as an integral combination

then

The expression of j in terms of the generators X_{K,L} is not canonical, but the resulting value of \theta_{N_2}(j) is canonical because \theta_{N_2} is defined as the Hurewicz image of j.

At a general orbit N_2/H, the Mackey functor map is obtained by replacing N_2 with H:

and the same formula applies to subquotients L\subset K\subset H with K/L\cong C_2\times C_2.

The exact sequence has the form

where all elements in \underline{\mathrm{coker}}_2\{8d+2\}(N_2/K) are torsion. When d=0, \eta\in \pi_1 KO is the Hurewicz image of the Hopf element. The composite homomorphism

sends \eta to \epsilon\eta\in \pi_1^K KO_{N_2} and \operatorname{Tr}_{\ker \tau}^K(1) \eta to (\epsilon+\tau)\eta \in\pi_1^K KO_{N_2}. Then \tau\eta\in RO(K;\mathbb{R})\{\eta\}/2 lifts to an element in \pi_1^K L_{KU_{N_2}/2}S of order 2. Thus the extension problem when k=1 is trivial. For general d\in \mathbb{Z}, consider the element \mu_d\in \pi_{8d+1}L_{K(1)}S of order 2 which maps to \eta u^d \in \pi_{8d+1} KO, the same argument shows that all the extension problems are trivial.

The map

only depends on the image of a subgroup in K/2K. Indeed, for H\subset K,

and the condition H\subseteq \ker(\chi) is equivalent to H+2K\subseteq \ker(\chi). Hence the elements D_{H,H'} are in the kernel, and after quotienting by these relations we may identify the source of the map with A(K/2K)/2.

Put \overline{K}=K/2K. It remains to determine the kernel of

For a subgroup U\subset \overline{K}, the \overline{K}-set \overline{K}/U maps to

If U\subset W\subset \overline{K} and W/U\cong C_2\times C_2, with intermediate subgroups U_1,U_2,U_3, then

maps to zero. Indeed, a character trivial on U is either trivial on W, in which case it is counted four times, or has kernel one of the three intermediate subgroups, in which case it is counted twice.

Let R\subset A(\overline K)/2 be the subgroup generated by the elements \overline{B}_{W,U}. Then the quotient A(\overline K)/(2,R) is generated by the class [\overline K/\overline K] and the classes [\overline K/H] with \overline K/H\cong C_2. Their images are 1+\chi for all characters \chi of \overline K, and these elements form a basis of RO(\overline K;\mathbb R)/2. Therefore there are no further relations. Pulling this description back along K\to K/2K gives the stated generators.

First we consider the case K\cong C_4, the only generator is

Let \rho be the faithful two-dimensional rotation C_4-representation. The computation of \underline{\pi}_1 L_{KU_{C_4}/2}S_{C_4} in ? shows that

and

In general, let K\subset N_2 and H,H'\subset K such that H+2K=H'+2K, we can reduce the computation to D_{C_2,\{e\}}. Let J = H +2 K, then in A(K)/2 we have

Hence it suffices to study the generators of the form

for some subgroup L\subset K and x\in K\backslash L. Let T = L \langle x \rangle. Then

Now T/L is a cyclic 2-group. Suppose T/L \cong C_{2^m}. In A(C_{2^m})/2, let C_{2^i} denote the unique subgroup of order 2^i. Then

Each term [C_{2^m}/C_{2^i}] + [C_{2^m}/C_{2^{i+1}}] is obtained from the basic class

on the subquotient C_{2^{i+2}}/C_{2^i} \cong C_4 by inflation followed by transfer. Since \underline{\theta}_{8d+1} is compatible with inflation and transfer maps, we can compute \theta_{8d+1} of D_{H,H'} via the value of D_{C_2,\{e\}}. Therefore, for every cyclic M\cong C_{2^t} \subset K, we can compute \theta_{8d+1} by tracking the indicated summand under these inflations and transfers. Let H\leq G, and let V\in\operatorname{Irr}_{\mathbb Q}(H). Then

Since \theta_{8d+1}(D_{C_2,{e}}) is nontrivial only in the summand corresponding to \rho, the above description of transfer, together with the behavior of kernels under inflation, shows that, for any K\leq N_2, the projection of \theta_{8d+1}(D_{H,H'}) onto the summand indexed by M\cong C_{2^t}\subset K is nontrivial if and only if t\geq 2 and \ker(M) contains exactly one of H and H'. The stated formula then follows by comparing the orders of the corresponding elements.

\theta_{8d+1} is defined by the action of \underline{A}_{N_2} on \eta u^d in \underline{\pi}_{8d+1} L_{KU_{N_2}/2}S_{N_2}, so we have

When T\cong V=C_2\times C_2, B_{T,\{e\}} is the image of X_V\in J(V) under the quotient J(V)\to J(V)/2, so

For a general L and T, inflating along T\to T/L and then transferring from T to K gives

Since 2K\subseteq L, every character of K trivial on L is a real one-dimensional character. Therefore,

This proves the claimed formula.

The proof is same as that of Proposition 5.7.

The computation of \underline{\ker}_p\{k\} is given in [bonventre2022ku_g, Proposition 6.7], [carawan2023homotopy, Corollary 3.5]. The computation of \underline{\mathrm{coker}}_p\{k\} is given in [carawan2023homotopy, Proposition 3.7]. The assertion about \underline{\pi}_*L_{KU_{N_p}/p}S_{N_p} follows from the fact that \pi_* KU_{N_p} is concentrated in even degrees.

We first show that the map A(K)\otimes A(H)\to A(K\oplus H) is an isomorphism. For any X\in A(K) and Y\in A(H), there is a natural action of K\oplus H on X\times Y; thus, (X,Y)\mapsto X\times Y induces a ring homomorphism

Conversely, as observed in the proof of Lemma 4.3, any Z\in A(K\oplus H) can be expressed uniquely as

where X_i, Y_i are K-orbits and H-orbits, respectively. Therefore, there is a ring homomorphism

which is the inverse of f. Since any subgroup S \subset K \oplus H satisfies S \cong (S\cap K) \oplus (S\cap H), f commutes with \operatorname{Res}_{S}^{K \oplus H} and \operatorname{Tr}_{S}^{K \oplus H}, f induces an isomorphism of G-Green functors \underline{A}_{N_p} \otimes \underline{A}_N \to \underline{A}_G.

Consider the following commutative diagram with exact rows

there exists an isomorphism \underline{\ker}(\mathcal{R}_{N_p}\otimes \mathcal{R}_N)\cong \underline{J}_G indicated by the dashed arrow. Thus

Here the first isomorphism follows from the fact that RU(K) is a free abelian group for every subgroup K\subset G, and the second follows from the Five Lemma.

The computation follows from Proposition 4.4 and Proposition 5.13. Note that by Lemma 6.1,

Then \underline{\pi}_0(L_{KU_G/p}S_G) \cong (\underline{A/J}_G)_p^\wedge when p is odd.

For KU_G\otimes \mathbb{Q}-localization, it follows from [bonventre2022ku_g, Lemma 9.1] that L_{KU_G \otimes \mathbb{Q}} S_G \simeq H(\mathbb{Q} \otimes \underline{A/J}), which is a G-equivariant Eilenberg-MacLane spectrum. So

concentrated in degree 0. Furthermore, since \underline{\pi}_k \prod_{p} L_{KU_G/p} S_G are torsion groups for k \neq 0, -1, (\prod_{p} L_{KU_G/p} S_G)_{\mathbb{Q}} has non-trivial homotopy groups only in degrees 0 and -1. Consequently, the long exact sequence on homotopy groups induced by the arithmetic fracture square takes the following form:

and for k\neq 0,-1,-2,

Then for k\neq 0,-1,-2, \underline{\pi}_kL_{KU_G}S_G\cong\prod_p \underline{\pi}_kL_{KU_G/p}S_G.

When k=0, \underline{\pi}_0 L_{KU_G}S_G\cong \ker f. Let

we have M_2^\wedge \cong \underline{\pi}_0 L_{KU_{N_2}/2}S_{N_2}. Note that all elements of \underline{RO(-;\mathbb{R})}_{N_2}\{\eta\}/2 are 2-torsion, after p-completion for odd prime p or rationalization, M_p^\wedge \cong (\underline{A/J}_{N_2})_p^\wedge, and M\otimes \mathbb{Q}\cong \underline{A/J}_{N_2}\otimes \mathbb{Q}. Therefore,

By Proposition 6.2, f is given by

This is exactly the arithmetic pullback of M\otimes \underline{A/J}_{G/N_2}, then

When k=-1, it follows from [carawan2023homotopy, Lemma 4.14] that \mathrm{coker} \ f=0, so \underline{\pi}_{-1} L_{KU_G}S_G\cong \ker g. Here

is an inclusion induced by \mathbb{Z}\to \mathbb{Q} since \underline{\pi}_{-1}(\prod_{p} L_{KU_G/p} S_G) is torsion free, thus \underline{\pi}_{-1} L_{KU_G}S_G\cong \ker g=0.

When k=-2, \underline{\pi}_{-2} L_{KU_G}S_G\cong \mathrm{coker}\ g. Therefore,

Let R=\operatorname{CoInd}_{N_GH}^G\bigl( \operatorname{Inf}_{W_G H}^{N_G H} F(EW_G H_+, X)[\mathcal{F}_{H\not\subset}^{-1}] \bigr). For any T(H) acyclic G-spectrum M, \Phi^H(M\wedge T(H))\simeq W_G H_+\wedge \Phi^H M\simeq \ast, thus the underlying spectrum of \Phi^H M is contractible. So EW_G H_+\wedge \Phi^H M is equivariantly contractible as a W_G H-spectrum. By Proposition 2.7, there are isomorphisms

Therefore, R is T(H)-local.

The map i:X\to R induced by EW_G H_+\to S^0 is a T(H)-equivalence. Indeed, if N is not conjugate to H, then \Phi^N(i \wedge T(H)) is a map between trivial spectra. For any subgroup K conjugate to H, it suffices to compute \Phi^H(i\wedge T(H)). This is the map X\wedge W_GH_+\to F(EW_GH_+, X)\wedge W_G H_+ induced by EW_GH_+\to S^0, and it is an equivalence of underlying non-equivariant spectra.

Let R=\operatorname{CoInd}_{N_GH}^G\bigl( \operatorname{Inf}_{W_G H}^{N_G H} F(EW_G H_+, L_{E}\Phi^H X)[\mathcal{F}_{H\not\subset}^{-1}] \bigr), we need to show that R is T(H)\wedge E-local, and there exists a T(H)\wedge E-equivalence X\to R.

By Lemma 7.1, R\simeq L_{T(H)}L_{E}\Phi^H X is T(H)-local. For any T(H)\wedge E-acyclic G-spectrum M,

The second isomorphism follows from Proposition 2.7. Since M\wedge T(H)\wedge E\simeq \ast, the spectrum \Phi^H M is W_G H_+\wedge E-acyclic. Hence EW_G H+\wedge \Phi^H M is E-acyclic. It follows that [M,R]^G\cong 0, so R is E-local.

There is a map i: X\to R induced by the E-localization \Phi^H X\to L_E \Phi^H X via the isomorphism

For any N\subset G, \Phi^N (T(H)\wedge E) \simeq \ast if N is not conjugate to H. When N\sim^G H, \Phi^N(i\wedge T(H)\wedge E) is equivalent to the map

induced by the E-localization \Phi^H X\to L_{E} \Phi^H X, which is an equivalence. Therefore, \Phi^N(i\wedge T(H)\wedge E) is an equivalence for all N\subset G. Hence i is a T(H)\wedge E-equivalence.

We can decompose the (K,H)-biset G as a disjoint union

Then

This follows from the fact that

By Proposition 2.3, it suffices to show that

Indeed, \Phi^L K(T,n)\not\simeq \ast if and only if L is conjugate to T.

For L\subset G, consider the L-geometric fixed point of L_{K(H,n)}X. If L is not subconjugate to N_G H, then for every g\in N_G L\backslash G/N_G H, \operatorname{Res}_{N_GH\cap^g N_G L}^{N_G L}\widetilde{E\mathcal F_{L\not\subset}} is trivial, and

If L\subset N_G H, then \Phi^L L_{K(H,n)}X is nontrivial only if H is subconjugate to L. Since H is not subconjugate to T for all T\in I, we have

for all L\subset G, i.e., (\bigvee_{T\in I}K(T,n)) \wedge L_{K(H,n)} X\simeq \ast.

Let I be a set of representatives for the conjugacy classes of cyclic subgroups H \subset G such that p\nmid |H|. We have L_{KU_G/p}X\simeq L_{\bigvee_{H\in I}K(H,1)}X since \langle KU_G/p\rangle=\langle \bigvee_{H\in I}K(H,1)\rangle.

It follows from Lemma 7.2 that for any subgroups H_1, H_2\in I,

Here p\nmid |H_1|, by Proposition 2.2 we have

Therefore, L_{K(H_1,1)}L_{K(H_2,1)}X\simeq \ast whenever H_1 is not conjugate to H_2 in G and p\nmid |H_1|, and the pullback square in Lemma 7.5 implies that

We can order the elements of I by the order of the corresponding subgroups. For subgroups of the same order, we assign an arbitrary order, since they cannot subconjugate to one another. Based on this order, I is totally ordered. We write

where |H_i|\leq |H_{i+1}|. Let I_1=\{H_1\} and I_k=I_{k-1} \cup \{H_k\}.

We study L_{\bigvee_{H\in I}K(H,1)}S_G by induction on I_k, using the pullback diagrams in Lemma 7.5. Assume that for any k<n, L_{\bigvee_{H\in I_k}K(H,1)}X \simeq \bigvee_{H\in I_k} L_{K(H,1)}X. Then

thus L_{\bigvee_{H\in I_n}K(H,1)}X \simeq \bigvee_{H\in I_n} L_{K(H,1)}X.

Let d=\dim V=\dim V_+-\dim V_-. There is a homological homotopy-orbit spectral sequence [campbell2024hilbert, Proposition B.7]

As a G-module,

where \mathbb Z_{(p)}\{\epsilon_V\} is the G-module \mathbb Z_{(p)} with the G-action defined by

The spectral sequence therefore collapses at the E^2-page, yielding the isomorphisms

Since G=N_p\times N, the extension

gives a Lyndon-Hochschild-Serre spectral sequence

Since p\nmid |N|,the augmentation \mathbb Z_{(p)}[N]\to \mathbb Z_{(p)} admits a splitting as a homomorphism of \mathbb Z_{(p)}[N]-modules, and the trivial \mathbb Z_{(p)}[N]-module \mathbb{Z}_{(p)} is a projective \mathbb Z_{(p)}[N]-module. It follows that ,

where \mathbb Z_{(p)}\{\epsilon_V\}/N is the coinvariant of \mathbb Z_{(p)}\{\epsilon_V\} under the N-action. Thus the spectral sequence is concentrated in the zero-th row, and collapses at E^2-page. Therefore,

If \epsilon_{V,N}\equiv 1, then N acts trivially on \mathbb Z_{(p)}\{\epsilon_V\}, and

It follows from the homotopy-orbit spectral sequence that the inclusion N_p\to G induces a map

which is a \mathbb Z_{(p)}-homology equivalence. After p-localization, it induces a weak equivalence \alpha_{(p)} of p-local spectra.

If \epsilon_{V,N}\not\equiv 1, then there exists n\in N such that \epsilon_V(n)=-1. Note that a nontrivial homomorphism \epsilon_V: N \to \{\pm 1\}\cong C_2 forces p\neq 2. Then 2 is invertible in \mathbb Z_{(p)}, and \mathbb Z_{(p)}\{\epsilon_V\}/N=0. Therefore,

which implies that (EG_+\wedge_G S^V)_{(p)} is contractible.

Note that e_{(H)}^N S_{N,(p)} is the idemoptent summand of S_{N,(p)} such that

Thus \operatorname{CoInd}_{N_N H}^N (\operatorname{Inf}_{W_N H}^{N_N H} F(EW_N H_+, S_{(p)})[\mathcal{F}_{H\not\subset}^{-1}]) provides a model for e_{(H)}^N S_{N,(p)}. For any L\subset N, by Lemma 7.3, we have

It follows from Lemma 7.7 that

This leads to the stated formula for \underline M_{H,V}(N/L), since \epsilon_{V^Q,W_L Q}\equiv 1 if and only if \epsilon_{V^H,W_L H}\equiv 1.

For any L_1\subset L_2 \subset N, the restriction of the conjugacy class (Q)\in \mathcal{C}_H(L_2) to L_1 can be decomposed into a disjoint union of conjugacy classes

The restriction map on the Q summand of \underline M_{H,V}(N/L_2)

is identified with the diagonal map \mathbb Z_{(p)}\to \bigoplus_{[a]}\mathbb Z_{(p)}. Stably, there is a stable transfer map between classifying spaces, which induces a transfer map between the corresponding Thom spaces. Thus we have stable maps

such that \operatorname{Tr}_{W_{L_1} H}^{W_{L_2} H}\circ \operatorname{Res}_{W_{L_1} H}^{W_{L_2} H}=|W_{L_2} H:W_{L_1} H|\cdot \operatorname{id} homologically. This induces the transfer map on every (Q) summand of \underline M_{H,V}(N/L_1)

Consequently, after collecting all L_1-conjugacy classes (Q_i) contained in the L_2-conjugacy class (Q), the transfer to the (Q)-summand is

For any cyclic subgroup H\subset G such that p\nmid |H|, H\subset N and W_G H\cong N_p\times W_N H. It follows from Lemma 7.2, Lemma 7.3 that for any T\subset G and V\in RO(T),

Here, in the final term, K({}^gH,1) denotes the N_G({}^gH)={}^gN_GH-equivariant Morava K-theory spectrum

Consequently, \operatorname{Res}_{T\cap {}^gN_GH}^{{}^gN_GH}K({}^gH,1) is nontrivial only if {}^gH is subconjugate to T\cap{}^gN_GH. Since {}^gH is normal in N_G{}^gH, this condition is equivalent to {}^gH\subset T. Let T\cong P\times L \subset N_p\times N be the decomposition of T into its Sylow p-subgroup and its p-complement. It follows by Proposition 2.7 that

By Lemma 7.7, if \epsilon_{V^H,W_L H}\equiv 1, then

Here the final isomorphism follows from the equivalence

which holds for every p-group P. If \epsilon_{V^H,W_L H}\not\equiv 1, then \pi_V^T L_{K(H,1)}S_G\cong 0.

For any T\subset G, the restriction and transfer maps in \underline{\pi}_V L_{K(H,1)}S_G are given by

By Lemma 4.3, it suffices to compute the restriction and transfer maps separately for P\subset N_p and L\subset N. For P_1\subset P_2\subset N_p, the restriction and transfer maps in \underline{\pi}_V L_{K(H,1)}S_G are exactly those in \underline{\pi}_{V^H} L_{KU_{N_p}/p}S_{N_p}. For L \subset N, if H is not subconjugate to L, \operatorname{Res}_L^N K(H,1) is a trivial L-spectrum, which implies that

are the natural projection and inclusion, respectively. If H is subconjugate to L and \epsilon_{V^H,W_NH}\equiv 1, then \langle \operatorname{Res}_L^N K(H,1)\rangle = \langle K(H,1) \rangle. Therefore, the restriction map \operatorname{Res}_L^N(H) is determined by the restriction of \mathcal{C}_H(N) to \mathcal{C}_L(N), while the transfer map \operatorname{Tr}_L^N(H) is given by multiplication by |W_N H:W_L H|. These maps are compatible with the restriction and transfer maps in \underline{M}_{(H),V}, which implies that \underline{\pi}_V L_{K(H,1)}S_G\cong \underline{\pi}_{V^H} L_{KU_{N_p}/p}S_{N_p}\otimes_{\mathbb Z_{(p)}}\underline{M}_{(H),V}.

Since L_{KU_G/p}S_G \simeq \bigvee_{\substack{(H)\in \mathrm{Cyc}/G\\ p\nmid |H|}} L_{K(H,1)}S_G,

which gives the stated isomorphism.

Note that a_\epsilon is a lifting of \eta\epsilon\in \pi_1^{C_4}KO_{C_4} and as the construction in the proof of Lemma 5.8, a_\epsilon is the Hurewicz image of \operatorname{Tr}_e^{C_2}(\eta)-\eta for the Hopf element \eta\in \pi_1 S. Let

the complexification map KO_{C_4}\to KU_{C_4} induces a C_4-map f:J\to J_{\mathbb{C}}. Here

The homomorphism f_*: \underline{\pi}_1 J \to \underline{\pi}_1 J_{\mathbb{C}} satisfies that

and

Therefore \operatorname{Tr}_{C_2}^{C_4}(a_\epsilon) = 2c \neq 0.