Ranks of $RO(G)$-graded stable homotopy groups of spheres for finite groups $G$

In ordinary stable homotopy theory, one of the most basic theorems is Serre’s Finiteness Theorem [Ser53] stating that the $n$-th stable homotopy group of the sphere, $\pi_{n}(S^{0})$, is finite for $n>0$. Since we understand that $\pi_{0}(S^{0})\cong\mathbb{Z}$, this means that rationally the structure of stable homotopy is very simple, and attention is quickly focused on torsion. Equivariantly, it is still true that rationalisation is a massive simplification, but the residual structure in the rationalisation is worth some attention.

Let $G$ be a finite group and consider the $RO(G)$-graded stable homotopy groups of the sphere [May96, Ch. IX]. The purpose of this note is to identify the crudest feature of these groups: their ranks as abelian groups. This is a straightforward deduction from well-known results, but some interesting features emerge by giving a systematic account.

Using rational equivariant stable homotopy theory, we prove the following:

We lay the groundwork for applying this theorem in Section 3 and Section 4. We then compute the ranks of the $RO(G)$-graded stable homotopy groups of spheres for various $G$ in Section 6.

In Section 5, we discuss two natural variations where the same techniques give information. Since the sphere is rationally an Eilenberg-MacLane spectrum for the Burnside Mackey functor, $S^{0}\simeq_{\mathbb{Q}}H\mathbb{A}$, we may view our methods as a calculation of the rationalization of $H^{\star}_{G}(S^{0};\mathbb{A})$ (where $\star$ denotes $RO(G)$-grading). The same methods apply to give a calculation of the rationalization of $H^{\star}_{G}(S^{0};M)$ for any Mackey functor $M$. For the second variation, we may consider the Picard-graded stable homotopy groups of spheres: invertible objects are again characterised in terms of orientations and dimension functions (see [FLM01]).

Finally, we note that our results provide a basis for understanding other large-scale phenomena in the $RO(G)$-graded stable homotopy groups of spheres. For example, Iriye [Iri83] showed that Nishida’s nilpotence theorem [Nis73] holds equivariantly: an element $\pi_{\star}^{G}(S^{0})$ is torsion if and only if it is nilpotent. A therefore explicitly describes the regions of $\pi_{\star}^{G}(S^{0})$ in which elements can be nilpotent and non-nilpotent.

For most of the paper we will work rationally, but we would like to draw conclusions about the integral situation. For completeness we include the proofs of the basic finiteness statements that permit this deduction.

The following consequence fails for infinite compact Lie groups.

A describes the $RO(G)$-graded rational homotopy groups of the sphere. By the previous lemma, this determines precisely those degrees $\alpha\in RO(G)$ for which $\pi^{G}_{\alpha}(S^{0})$ is finite.

Henceforth everything is rational. We write $G$ for a finite group, $H$ for a subgroup of $G$, and $W_{G}(H)=N_{G}(H)/H$ for the Weyl group of $H$. We use $*$ to denote $\mathbb{Z}$-graded groups and $\star$ to denote $RO(G)$-graded groups. If $V\in RO(G)$, then $|V|$ denotes its (virtual) dimension.

The second author thanks Eva Belmont, Bert Guillou, Dan Isaksen, and Mingcong Zeng for helpful discussions related to this work. The authors also thank William Balderrama for discussions concerning nilpotence, and especially for pointing out [Iri83], which answered a question in a previous version.

If we list the simple real representations $S_{1},S_{2},\ldots,S_{r}$ of $G$, we may identify $RO(G)=\mathbb{Z}^{r}$. Now, for each subgroup $H\leq G$ we have a dimension vector

$$d_{H}=(\dim(S_{1}^{H}),\ldots,\dim(S_{r}^{H})),$$

and the space of virtual representations $\alpha$ with $\alpha^{H}=0$ is

$$N_{H}=\{x\;|\;x\cdot d_{H}=0\},$$

which is isomorphic to $\mathbb{Z}^{r-1}$ as an abelian group. The only $\alpha$ for which $\pi^{G}_{\alpha}(S^{0})$ can be infinite are those lying in some $N_{H}$, and the maximum rank of $\pi^{G}_{\alpha}(S^{0})$ is the number of conjugacy classes of $H$ with $\alpha\in N_{H}$.

When $H=G$, the Weyl group $W_{G}(H)$ is trivial, and we immediately draw a useful conclusion.

If $W_{G}(H)$ is of odd order, then all the gradings in $N_{H}$ give infinite groups. In general, on each such null space $N_{H}$ we have an orientation

$$o_{H}:N_{H}\longrightarrow H^{1}(W_{G}(H);\mu_{2}).$$

As noted above, the kernel $N_{H}^{+}$ contains all even vectors of $N_{H}$ and the image of all complex representations.

The set of $\alpha$ for which $\pi^{G}_{\alpha}(S^{0})$ is infinite is $\bigcup_{H}N_{H}^{+}$. The rank $r_{\alpha}$ of $\pi^{G}_{\alpha}(S^{0})$ is the number of conjugacy classes $H$ with $\alpha\in N_{H}^{+}$.

If we choose a subgroup $H$ giving an associated fixed point vector $d_{H}$, we note that the component of the trivial representation $S_{1}$ is always 1, so that $N_{H}$ has basis $S_{2}-d_{H}(2),S_{3}-d_{H}(3),\ldots,S_{r}-d_{H}(r)$. The orientation $o_{H}$ is thus described by the homomorphisms

$$o_{H}(2),o_{H}(3),\ldots,o_{H}(r):W_{G}(H)\longrightarrow\mu_{2}$$

where $o_{H}(i)=o_{H}(S_{i}-d_{H}(i))$. Since $W_{G}(H)$ always acts trivially on the trivial representation, the orientation $o_{H}(S_{i}-d_{H}(i))=o_{H}(S_{i})$, and $o_{H}(i)$ is the determinant of $S_{i}^{H}$. Since $o_{H}$ is a homomorphism, this determines its values throughout. All the homomorphisms factor through the largest elementary abelian 2-quotient $E_{2}(H)$ of $W_{G}(H)$ (i.e., we factor out commutators and squares).

We have

$$RO(C_{2})\cong\mathbb{Z}\{1,\sigma\}$$

where $1$ is the ($1$-dimensional) trivial representation and $\sigma$ is the sign representation. Then

$$N_{e}\cong\mathbb{Z}\{1-\sigma\},\quad N_{C_{2}}\cong\mathbb{Z}\{\sigma\}.$$

Since $W_{C_{2}}(C_{2})=e$, we have

$$N_{C_{2}}^{+}=N_{C_{2}}\cong\mathbb{Z}\{\sigma\}.$$

On the other hand, $W_{C_{2}}(e)\cong C_{2}/e\cong C_{2}$ acts by $(-1)$ on $1-\sigma$, so

$$N_{e}^{+}\cong\mathbb{Z}\{2(1-\sigma)\}.$$

Each representation $V\in N_{C_{2}}^{+}\cup N_{e}^{+}$ satisfies $\operatorname{rk}\pi_{V}^{C_{2}}(S^{0})\geq 1$. Since $N_{C_{2}}^{+}\cap N_{e}^{+}=\{0\}$, we also have $\operatorname{rk}\pi_{0}^{C_{2}}(S^{0})=2$. Altogether, we find:

Let $q=\frac{p-1}{2}$. We have

$$RO(C_{p})\cong\mathbb{Z}\{1,\phi_{1},\ldots,\phi_{q}\},$$

where $1$ is the $1$-dimensional trivial representation and $\phi_{t}:C_{p}\to\operatorname{Aut}(\mathbb{R}^{2})\cong\operatorname{Aut}(\mathbb{C})$ sends the generator of $C_{p}$ to $\cdot e^{2\pi it/p}$. Then

$$N_{e}\cong\mathbb{Z}\{2-\phi_{1},\ldots,2-\phi_{q}\},\quad N_{C_{p}}\cong\mathbb{Z}\{\phi_{1},\ldots,\phi_{q}\}.$$

Since $W_{C_{p}}(e)\cong C_{p}$ and $W_{C_{p}}(C_{p})=e$ necessarily act trivially on $\mathbb{Z}$, we have

$$N_{e}^{+}=N_{e},\quad N_{C_{p}}^{+}\cong N_{C_{p}}.$$

Finally, we have

$$N_{e}^{+}\cap N_{C_{p}}^{+}\cong\mathbb{Z}\{\phi_{1}-\phi_{2},\ldots,\phi_{1}-\phi_{q}\}.$$

Let $q=\frac{p^{n}-1}{2}$. Then

$$RO(C_{p^{n}})\cong\mathbb{Z}\{1,\phi_{1},\ldots,\phi_{q}\}\cong\mathbb{Z}^{q+1}.$$

For all $0\leq m\leq n$, we have

$$N_{C_{p^{m}}}^{+}=N_{C_{p^{m}}}\cong\mathbb{Z}\{2-\phi_{i}:p^{m}\mid i\}\oplus\mathbb{Z}\{\phi_{j}:p^{m}\nmid j\}.$$

Indeed, let $\gamma$ denote a generator of $C_{p^{n}}$, so $\gamma^{p^{n-m}}$ is a generator for $C_{p^{m}}$. Since $\lambda_{i}:\gamma\mapsto\cdot e^{\frac{2\pi i}{p^{n}}}$,

$$\lambda_{i}:\gamma^{p^{n-m}}\mapsto\cdot e^{\frac{2\pi ip^{n-m}}{p^{n}}}=\cdot e^{\frac{2\pi i}{p^{m}}}.$$

Therefore $\lambda_{i}$ pulls back to a trivial $C_{p^{m}}$-representation if and only if $p^{m}\mid i$.

Describing the intersections of these subspaces gets complicated quickly. For example, if $0\leq k<m\leq n$, then

$$N_{C_{p^{k}}}^{+}\cap N_{C_{p^{m}}}^{+}\cong\mathbb{Z}\{2-\phi_{i}:p^{m}\mid i\}\oplus\mathbb{Z}\{\phi_{j}:p^{k}\nmid i\}\oplus\mathbb{Z}\{\phi_{p^{k}}-\phi_{\ell}:\ell>p^{k},\ p^{k}\mid\ell,\ p^{m}\nmid\ell\}.$$

Here, we use that $p^{m}\mid i$ implies $p^{k}\mid i$, and similarly, $p^{k}\nmid j$ implies $p^{m}\nmid j$.

Let $K=C_{2}\times C_{2}=\{e,i,j,k\}$. We have

$$RO(K)\cong\mathbb{Z}\{1,\sigma_{i},\sigma_{j},\sigma_{k}\},$$

where $1$ is the $1$-dimensional trivial representation, $\sigma_{i}$ is the $1$-dimensional representation on which $e$ and $i$ act trivially and $j$ and $k$ act by $(-1)$, and similarly for $\sigma_{j}$ and $\sigma_{k}$. Then

	$\displaystyle N_{e}$	$\displaystyle\cong\mathbb{Z}\{1-\sigma_{i},1-\sigma_{j},1-\sigma_{k}\},$
	$\displaystyle N_{\langle i\rangle}$	$\displaystyle\cong\mathbb{Z}\{1-\sigma_{i},\sigma_{j},\sigma_{k}\},$
	$\displaystyle N_{\langle j\rangle}$	$\displaystyle\cong\mathbb{Z}\{\sigma_{i},1-\sigma_{j},\sigma_{k}\},$
	$\displaystyle N_{\langle k\rangle}$	$\displaystyle\cong\mathbb{Z}\{\sigma_{i},\sigma_{j},1-\sigma_{k}\},$
	$\displaystyle N_{K}$	$\displaystyle\cong\mathbb{Z}\{\sigma_{i},\sigma_{j},\sigma_{k}\}.$

The Weyl group $W_{K}(e)\cong K$ acts nontrivially on $\sigma_{i}$, $\sigma_{j}$, and $\sigma_{k}$, so we have

$$N_{e}^{+}\cong\mathbb{Z}\{2(1-\sigma_{i}),\sigma_{i}-\sigma_{j},\sigma_{i}-\sigma_{k}\}.$$

The Weyl group $W_{K}(\langle i\rangle)\cong K/\langle i\rangle\cong\langle j\rangle\cong\langle k\rangle$ acts nontrivially on $\sigma_{i}$ but trivially on $\sigma_{j}$ and $\sigma_{k}$, so we have

$$N_{\langle i\rangle}^{+}\cong\mathbb{Z}\{2(1-\sigma_{i}),\sigma_{j},\sigma_{k}\},$$

and similarly,

$$N_{\langle j\rangle}^{+}\cong\mathbb{Z}\{\sigma_{i},2(1-\sigma_{j}),\sigma_{k}\},$$

$$N_{\langle k\rangle}^{+}\cong\mathbb{Z}\{\sigma_{i},\sigma_{j},2(1-\sigma_{k})\}.$$

Finally, since $W_{K}(K)\cong e$ must act trivially on $\mathbb{Z}$, we have

$$N_{K}^{+}\cong N_{K}\cong\mathbb{Z}\{\sigma_{i},\sigma_{j},\sigma_{k}\}.$$

To determine the ranks of $\pi^{G}_{\alpha}(S^{0})$, we now compute intersections. In the following, we let $a\in\{i,j,k\}$, $a^{\prime}\in\{i,j,k\}\setminus\{a\}$, and $a^{\prime\prime}\in\{i,j,k\}\setminus\{a,a^{\prime}\}.$ Then we have

	$\displaystyle N_{e}^{+}\cap N_{\langle a\rangle}^{+}$	$\displaystyle\cong\mathbb{Z}\{2(1-\sigma_{a}),\sigma_{a}-\sigma_{a^{\prime}}\},$
	$\displaystyle N_{e}^{+}\cap N_{K}^{+}$	$\displaystyle\cong\mathbb{Z}\{\sigma_{i}-\sigma_{j},\sigma_{i}-\sigma_{k}\},$
	$\displaystyle N_{\langle a\rangle}^{+}\cap N_{\langle a^{\prime}\rangle}^{+}$	$\displaystyle\cong\mathbb{Z}\{2(1-\sigma_{a}-\sigma_{a^{\prime}}),\sigma_{a^{\prime\prime}}\},$
	$\displaystyle N_{\langle a\rangle}^{+}\cap N_{K}^{+}$	$\displaystyle\cong\mathbb{Z}\{\sigma_{a^{\prime}},\sigma_{a^{\prime\prime}}\},$
	$\displaystyle N_{e}^{+}\cap N_{\langle a\rangle}^{+}\cap N_{\langle a^{\prime}\rangle}^{+}$	$\displaystyle\cong\{0\},$
	$\displaystyle N_{\langle a\rangle}^{+}\cap N_{\langle a^{\prime\prime}\rangle}^{+}\cap N_{K}^{+}$	$\displaystyle\cong\mathbb{Z}\{\sigma_{a^{\prime}}\},$
	$\displaystyle N_{e}^{+}\cap N_{\langle a\rangle}^{+}\cap N_{K}^{+}$	$\displaystyle\cong\mathbb{Z}\{\sigma_{a}-\sigma_{a^{\prime}}\},$
	$\displaystyle N_{\langle a\rangle}\cap N_{\langle a^{\prime}\rangle}\cap N_{\langle a^{\prime\prime}\rangle}$	$\displaystyle\cong\mathbb{Z}\{2(1-\sigma_{a}-\sigma_{a^{\prime}}-\sigma_{a^{\prime\prime}})\},$

and all $4$- and $5$-fold intersections are $\{0\}$.

We have

$$RO(D_{2p})\cong\mathbb{Z}\{1,\sigma,\phi_{1},\ldots,\phi_{q}\},$$

where $1$ is the trivial representation, $\sigma$ is the sign representation, and $\phi_{t}:D_{2p}\to\operatorname{Aut}(\mathbb{R}^{2})\cong\operatorname{Aut}(\mathbb{C})$ sends the generator of $C_{p}\subseteq D_{2p}$ to $\cdot e^{2\pi it/p}$ and the generator of $C_{2}\subseteq D_{2p}$ to reflection across the real axis. Then

	$\displaystyle N_{e}$	$\displaystyle\cong\mathbb{Z}\{1-\sigma,2-\phi_{1},\ldots,2-\phi_{q}\},$
	$\displaystyle N_{C_{2}}$	$\displaystyle\cong\mathbb{Z}\{1-\sigma,1-\phi_{1},\ldots,1-\phi_{q}\},$
	$\displaystyle N_{C_{p}}$	$\displaystyle\cong\mathbb{Z}\{1-\sigma,\phi_{1},\ldots,\phi_{q}\},$
	$\displaystyle N_{D_{2p}}$	$\displaystyle\cong\mathbb{Z}\{\sigma,\phi_{1},\ldots,\phi_{q}\}.$

Since $W_{D_{2p}}(C_{2})\cong e\cong W_{D_{2p}}(D_{2p})$, we have $N_{C_{2}}^{+}=N_{C_{2}}$ and $N_{D_{2p}}^{+}=N_{D_{2p}}$. On the other hand, $W_{D_{2p}}(e)\cong D_{2p}$ and $W_{D_{2p}}(C_{p})\cong C_{2}$, so

	$\displaystyle N_{e}^{+}$	$\displaystyle\cong\mathbb{Z}\{2(1-\sigma),2(2-\phi_{1}),\phi_{1}-\phi_{2},\ldots,\phi_{1}-\phi_{q}\},$
	$\displaystyle N_{C_{p}}^{+}$	$\displaystyle\cong\mathbb{Z}\{2(1-\sigma),2\phi_{1},\phi_{1}-\phi_{2},\ldots,\phi_{1}-\phi_{q}\}.$

We now compute intersections:

	$\displaystyle N_{e}^{+}\cap N_{C_{2}}^{+}$	$\displaystyle\cong\mathbb{Z}\{2(1-\sigma),\phi_{1}-\phi_{2},\ldots,\phi_{1}-\phi_{q}\},$
	$\displaystyle N_{e}^{+}\cap N_{C_{p}}^{+}$	$\displaystyle\cong\mathbb{Z}\{2(1-\sigma),\phi_{1}-\phi_{2},\ldots,\phi_{1}-\phi_{q}\},$
	$\displaystyle N_{e}^{+}\cap N_{D_{2p}}^{+}$	$\displaystyle\cong\mathbb{Z}\{4\sigma-2\phi_{1},\phi_{1}-\phi_{2},\ldots,\phi_{1}-\phi_{q}\},$
	$\displaystyle N_{C_{2}}^{+}\cap N_{C_{p}}^{+}$	$\displaystyle\cong\mathbb{Z}\{2(1-\sigma),\phi_{1}-\phi_{2},\ldots,\phi_{1}-\phi_{q}\},$
	$\displaystyle N_{C_{2}}^{+}\cap N_{D_{2p}}^{+}$	$\displaystyle\cong\mathbb{Z}\{\sigma-\phi_{1},\phi_{1}-\phi_{2},\ldots,\phi_{1}-\phi_{q}\},$
	$\displaystyle N_{C_{p}}^{+}\cap N_{D_{2p}}^{+}$	$\displaystyle\cong\mathbb{Z}\{\phi_{1},\dots,\phi_{q}\},$
	$\displaystyle N_{e}^{+}\cap N_{C_{2}}^{+}\cap N_{C_{p}}^{+}$	$\displaystyle\cong\mathbb{Z}\{2(1-\sigma),\phi_{1}-\phi_{2},\ldots,\phi_{1}-\phi_{q}\},$
	$\displaystyle N_{e}^{+}\cap N_{C_{2}}^{+}\cap N_{D_{2p}}^{+}$	$\displaystyle\cong\mathbb{Z}\{\phi_{1}-\phi_{2},\ldots,\phi_{1}-\phi_{q}\},$
	$\displaystyle N_{e}^{+}\cap N_{C_{p}}^{+}\cap N_{D_{2p}}^{+}$	$\displaystyle\cong\mathbb{Z}\{\phi_{1}-\phi_{2},\ldots,\phi_{1}-\phi_{q}\},$
	$\displaystyle N_{C_{2}}^{+}\cap N_{C_{p}}^{+}\cap N_{D_{2p}}^{+}$	$\displaystyle\cong\mathbb{Z}\{\phi_{1}-\phi_{2},\ldots,\phi_{1}-\phi_{q}\},$
	$\displaystyle N_{e}^{+}\cap N_{C_{2}}^{+}\cap N_{C_{p}}^{+}\cap N_{D_{2p}}^{+}$	$\displaystyle\cong\mathbb{Z}\{\phi_{1}-\phi_{2},\ldots,\phi_{1}-\phi_{q}\}.$